This action might not be possible to undo. Are you sure you want to continue?

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

**Michael S. Triantafyllou Franz S. Hover
**

Department of Ocean Engineering Massachusetts Institute of Technology Cambridge, Massachusetts USA

Maneuvering and Control of Marine Vehicles Latest Revision: November 5, 2003 c ◦Michael S. Triantafyllou and Franz S. Hover

Contents

1 KINEMATICS OF MOVING FRAMES 1.1 Rotation of Reference Frames . . . . . . 1.2 Diﬀerential Rotations . . . . . . . . . . . 1.3 Rate of Change of Euler Angles . . . . . 1.4 Dead Reckoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1

2

4

5

5

5

6

7

8

8

8

9

10

11

11

12

12

12

13

14

15

15

16

17

17

17

18

20

21

21

22

22

22

23

23

2 VESSEL INERTIAL DYNAMICS 2.1 Momentum of a Particle . . . . . . . . . . . . . 2.2 Linear Momentum in a Moving Frame . . . . . 2.3 Example: Mass on a String . . . . . . . . . . . 2.3.1 Moving Frame Aﬃxed to Mass . . . . . 2.3.2 Rotating Frame Attached to Pivot Point 2.3.3 Stationary Frame . . . . . . . . . . . . . 2.4 Angular Momentum . . . . . . . . . . . . . . . 2.5 Example: Spinning Book . . . . . . . . . . . . . 2.5.1 x-axis . . . . . . . . . . . . . . . . . . . 2.5.2 y-axis . . . . . . . . . . . . . . . . . . . 2.5.3 z-axis . . . . . . . . . . . . . . . . . . . 2.6 Parallel Axis Theorem . . . . . . . . . . . . . . 2.7 Basis for Simulation . . . . . . . . . . . . . . . 3 NONLINEAR COEFFICIENTS IN DETAIL 3.1 Helpful Facts . . . . . . . . . . . . . . . . . . 3.2 Nonlinear Equations in the Horizontal Plane . 3.2.1 Fluid Force X . . . . . . . . . . . . . . 3.2.2 Fluid Force Y . . . . . . . . . . . . . . 3.2.3 Fluid Moment N . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

4 VESSEL DYNAMICS: LINEAR CASE 4.1 Surface Vessel Linear Model . . . . . . . . . . . 4.2 Stability of the Sway/Yaw System . . . . . . . . 4.3 Basic Rudder Action in the Sway/Yaw Model . 4.3.1 Adding Yaw Damping through Feedback 4.3.2 Heading Control in the Sway/Yaw Model 4.4 Response of the Vessel to Step Rudder Input . . 4.4.1 Phase 1: Accelerations Dominate . . . . 4.4.2 Phase 3: Steady State . . . . . . . . . . 4.5 Summary of the Linear Maneuvering Model . . 4.6 Stability in the Vertical Plane . . . . . . . . . . i

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

5 SIMILITUDE 5.1 Use of Nondimensional Groups . . . . . 5.2 Common Groups in Marine Engineering 5.3 Similitude in Maneuvering . . . . . . . . 5.4 Roll Equation Similitude . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

23

23

25

27

29

30

30

30

30

33

33

33

34

34

34

34

35

35

35

35

36

37

37

38

39

39

39

40

41

41

42

43

44

44

44

44

45

46

47

48

6 CAPTIVE MEASUREMENTS 6.1 Towtank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Rotating Arm Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Planar-Motion Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 STANDARD MANEUVERING TESTS 7.1 Dieudonn´ Spiral . . . . . . . . . . . . . e 7.2 Zig-Zag Maneuver . . . . . . . . . . . . . 7.3 Circle Maneuver . . . . . . . . . . . . . . 7.3.1 Drift Angle . . . . . . . . . . . . 7.3.2 Speed Loss . . . . . . . . . . . . 7.3.3 Heel Angle . . . . . . . . . . . . . 7.3.4 Heeling in Submarines with Sails . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8 STREAMLINED BODIES 8.1 Nominal Drag Force . . . . . . . . . . . . . . . . . 8.2 Munk Moment . . . . . . . . . . . . . . . . . . . . 8.3 Separation Moment . . . . . . . . . . . . . . . . . . 8.4 Net Eﬀects: Aerodynamic Center . . . . . . . . . . 8.5 Role of Fins in Moving the Aerodynamic Center . . 8.6 Aggregate Eﬀects of Body and Fins . . . . . . . . . 8.7 Coeﬃcients Zw , Mw , Zq , and Mq for a Slender Body 9 SLENDER-BODY THEORY 9.1 Introduction . . . . . . . . . . . . . . . . . . 9.2 Kinematics Following the Fluid . . . . . . . 9.3 Derivative Following the Fluid . . . . . . . . 9.4 Diﬀerential Force on the Body . . . . . . . . 9.5 Total Force on a Vessel . . . . . . . . . . . . 9.6 Total Moment on a Vessel . . . . . . . . . . 9.7 Relation to Wing Lift . . . . . . . . . . . . . 9.8 Convention: Hydrodynamic Mass Matrix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10 PRACTICAL LIFT CALCULATIONS 10.1 Characteristics of Lift-Producing Mechanisms 10.2 Jorgensen’s Formulas . . . . . . . . . . . . . . 10.3 Hoerner’s Data: Notation . . . . . . . . . . . 10.4 Slender-Body Theory vs. Experiment . . . . . 10.5 Slender-Body Approximation for Fin Lift . . . ii

.2 Linearized Dynamics . . . . . . . . . 14. . . . . . .2. . . . . . . . . .2 Partial Fractions: Unique Poles . 13. .4 Vehicle Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Basic Relations . . . . . . . . 13. . .3. . . . .1 Force Balance . . . . .3 Ampere’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . 14 TOWING OF VEHICLES 14. . . . . . . . . . . . . . . . . . . . . 15. . . . . . . . 15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Cable Strumming . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . .2 DC Motors . .1 Permanent Field Magnets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12. . . . . . . 15. . . . . . . . . . . . .1 Derivation . . . . . .2. . . . . . . . . . .1. . 13. . . . . .2 Solution for Steady Conditions 12. . . . . . . . 12. . . .2 Critical Angle . . . . . . . . . .3 Series Windings . . . . . . . . .1 Introduction . . . . . . . .1. . . . . . . . . . .2 Shunt or Independent Field Windings 13. . . . . . . . . . . . . . . .3. . . . . . . . . .6 Stability in Linear Systems . . . 49 49 49 50 50 50 51 52 54 54 56 56 56 57 57 57 58 58 58 58 59 60 60 61 62 64 65 65 67 68 68 70 72 73 73 73 74 74 75 75 75 13 ELECTRIC MOTORS 13. . . . . . . . . . . . . .4 Partial Fractions: Complex-Conjugate Poles . .1. . . . . . .2 Faraday’s Law . . . . . . . . . . . . iii . . . . . . . . . . . . . . . . . . . .1. . .3 Three-Phase Synchronous Motor . . . . . . . . . . . . 14. 12. . .2. . . . . .1 Origin of Lift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Example: Partial Fractions with Complex Poles . . . . . 11. . . . . . . . . . . . .1 Concepts . . 13. . . . . . . . . . . . . . . . . . . . . . . . . . . 12.2 Steady Propulsion of Vessels . . 12. . . . . . . . . . . . . . . . . . . . . 15. . . . . . . . . .3 Unsteady Propulsion Models . . . . . 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . .2 Damped Axial Motion 14. . . . . . . . . . . . . 12 PROPELLERS AND PROPULSION 12. . . . . . . . . . . . . . . . . . . . . . .1 One-State Model: Yoerger et al. . . . . . . . . .3 Example: Partial Fractions with Unique Real Poles 15. . . . . . 13. . . . . . . . . . . . . . . . . . . . 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Three-Phase Induction Motor . . . . . . . . .2. . . . . 14. . . . . . . 14. . . . . . . . 14. . . . . . . . . . . . . . . . . . . .3 Engine/Motor Models . . . . . . . .1 Partial Fractions . . .2 Two-State Model: Healey et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Basic Characteristics . . 12. . . . . . . . . . . . . . . .2. .1 Statics . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Ring Fins . .4 Force . .11 FINS AND LIFTING SURFACES 11. . . . . . . . . . . . . . . . . . 14. 13.2 Three-Dimensional Eﬀects: Finite Length . . . . . . . . . . . . . . . . . . . . .2. . 15 TRANSFER FUNCTIONS & STABILITY 15. . . . . . . 13. .

. . Inputs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Deﬁnition . . . . . . . . . .3 Nonlinear Control . . . . . . . . . . . . . . . . . . . . . . . 16. . . . . . . . 18.3 Converting a Transfer Function into a State-Space Model 16. . .6 Block Diagrams of Systems . . . . . 15. . . . . . and Outputs . . .4. . .2. . . . . . 16. . . . . . . . . . . . . 18.2. 16. . . . .2 The Need for Modeling . . . . . . . 16. . . . . . . . . .1 Introduction .3 Modal Decomposition of Response 17. . . . . . . . . . . . . 18. .4. . . . . . . . . . . . . . . . . . 16. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Robust Performance . .1 Mapping Theorem . . . . . . . . . .7 Stability �� Poles in LHP . . .3 Primary Transfer Functions . . . .2. . . . 16. . .4. . . 16. . 17. . .5 Heuristic Tuning . . . . . . . . . . . . . . . . . . . .3 PID Controllers .1 Standard State-Space Form . . . . . . . . . 16 CONTROL FUNDAMENTALS 16. . . . . . .4 Example: PID Control . . . . .2 Block Diagrams: General Case . . . . . . . . . . . . . . . 18 CONTROL SYSTEMS – LOOPSHAPING 18. . .1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 16. . . . . . . . . . . . .15. . . . . .2. . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . .3 Proportional-Integral-Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . 76 76 76 76 76 77 77 77 77 78 78 79 79 80 80 80 81 81 81 81 82 83 83 83 83 84 84 84 85 86 86 86 87 87 88 89 89 90 91 91 . . . . . . . . . . . . . . . . . . . . . . . . . .3 Forced Response and Controllability . . . . . . . . . . . .2. . . . . . . . .2 Nyquist Criterion . .2 Matrix Exponential . . 16. . . . .1. . . . 16. . 18. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17. . . . . . .2 Modal Canonical Form . . . . . 17 MODAL ANALYSIS 17. . . . . . . . 18. . . . . . .2. . . . . . . . . . . . .2 Roots of Stability – Nyquist Criterion . . . . .8 General Stability . . . . . . . . .4 Design for Robustness . . . . . . . . . . . . . . 16. . . . . . . . . . . . .2 Proportional-Derivative Only . . . . . . . . . . . . . . . . . . . . . .4 Plant Output and Observability . . . . . . . iv . . .2. . . . . . .7 The Recipe for Loopshaping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . .6. . . . . 18. . . . . . . . . . . . . . . . . . . . . . . . 17. . . . .1 Proportional Only . . . . . . . . . . . . . . . . . .1 Fundamental Feedback Loop . . . .1. . . . . . . . . . . . . . .2 Converting a State-Space Model into a Transfer Function 16. . . . . . . . . . . . . . . 17. . . . . 16.6. . . . . . . . . . . . . . . . . 18. . . . . . . . . . . . . . . . . .1 Introduction . . . . . . . . 16. . . . . . . . . . .1. . . . . . . 18. . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Design for Nominal Performance . . . .3 Robustness on the Nyquist Plot . . . . . . . . . . . . . . . . . . . . . .6. . . . 16. . . . . .1 Plants. . . . . . . .6 Implications of Bode’s Integral . . 16. . . . . . . .2 Representing Linear Systems . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . . . . . . . . . . .1 Vectors .4 Vector Cross Product . . . . . . . . .3 Proof that �H trace(−ΦΓC T H T ) = −ΦΓC T 20. . . . . . . .1 Introduction . . . . . . . . . . . � 20. . .4 Gradient Method Solution for the General Case 19. . . . . . . 22. . . . . 19. . . . . . . .2. . . . .2. . . . . .2 A Special Property of the LQR Solution 21. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Determinant . . . . . . . . . . . . .6 Optimal Full-State Feedback . . . . v . . . . . . . . . . . 22 APPENDIX 1: MATH FACTS 22. . . . . . . . . . .5 Three Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . � 20. 22. . . . . . .4 Common Matrices . 19. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 Inverse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Full-State Feedback . . . . . . . . . .7. . . . . . . . . . . . . . . . . . . . . . . . .2 Multiplying a Vector by a Matrix 22. . . .1 Proof that E(eT W e) = trace(ΓW ) . . . . . . . . . . . . 21 LOOP TRANSFER RECOVERY 21. .3 Step 1: An Equation for Γ . . . . 22. . . . . . . . . . . . . .5 Transpose . . 22. .2. . . 20. . . . . . . .6 Combination of LQR and KF . . . . . . . . . . . . . . . . . . . . . . . . . . 22. . . . . . . . . . . . . . . . . . . . . . . . . . 22. . .1 Deﬁnition . . . . . . .19 LINEAR QUADRATIC REGULATOR 19. .2 Problem Statement . . . . . . . 20.3 The Maximum Principle . . . . . . . . . . . . . . . . . . . . . . . . .2.2. . . . . . . . . . 22. . 19. . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . ˙ 20. . . . . . .7 Properties and Use of the LQR . . . . . . .7. . .4 Step 2: H as a Function of Γ . . . . . . . . . . . .5 Properties of the Solution . . . . .5 LQR Solution . . . . . . . . . . . .8 Proof of the Gain and Phase Margins . .2 Proof that �H trace(−ΦHCΓ) = −ΦT ΓC T . . . . . . . . 22. . . . . . . . . . .1 Introduction . . . 92 92 93 93 94 95 96 96 97 98 98 98 99 100 101 102 103 103 104 104 105 105 105 106 107 108 109 110 110 110 111 111 112 112 112 112 113 113 114 114 115 115 20 KALMAN FILTER 20. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Multiplying a Matrix by a Matrix 22. .1.2. . . . . . . . . . . . . 21. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20. . . . . . . . . . . . . . . . . . . . . . . . . . .8 Trace . . . . . . . . . . . . .3 Vector Dot or Inner Product . . . 19. . . . . . . . . . . . . . .1. . . . . . . . . . . . . . 21. . . . . . . .2 Matrices . . . 20. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Vector Magnitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . .7. . . . . . . . . . . . . 19. . . . . . . . . . . . . . . . . . . . . .4 Proof of the Separation Principle . . . . . . . 20. . . . . . . . . . . . .4 Usage of the Loop Transfer Recovery . . . . . . . . . 22. . . . . . . . . . . . . .1. . . . . . . . . . . . . . . . 22. . . . . . . . .1. . . . . . . 21. . . . 19. . . . . .7 Proofs of the Intermediate Results . . . . . . . . . . . . . . . . . . . . . . . .1 Deﬁnition . . . . . . . . . . . . . . . . . . . . . . . . . . . 22. . . .7. . . . . . . . . . .3 The Loop Transfer Recovery Result . . . . . . . . . . . .

. . . . . .5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . .1. . . . . . . . . . . . . . .2. . . . . 22. 25. . . . 24 APPENDIX 3: LQR VIA DYNAMIC PROGRAMMING 130 24. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25. . . . . . . 133 134 134 134 134 135 135 135 136 136 137 137 138 140 141 142 vi . . .2 Derivative of a Scalar with Respect to a Vector . . . . . . . . . . . .1 Free versus Column Vector . . . . .3. . . . . . . .4 A and B Perturbations as an LMI . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . 25. . . . . . . . . . . . . . .1 Lyapunov’s Second Method . . . . . 25.3.5. . . . 23. . . . . 22. . .2 Convergence . . . . . . . . . . . . . . . .3 Franklin Inequality . . . .3 Parametric Uncertainty in A and B Matrices . . . . . . . . . . . . . . 25. . . . . .1.3 Dot and Cross Product . . . . .11 Singular Value . . . . . . .2 Matrix Inequality Deﬁnition . . . . . . . . . . . . . . . .4 Solution of Diﬀerential Equations by Laplace 22. . . . . . . .6 Schur Complement of a Nine-Block Matrix . . . . . . . . . . . . . . . . . 22. . . 22. . . . . . . . . . . . . .1. .2 Uncertainty in B . . 23. . . . . . . . .3 Laplace Transform . . . . . . . . . .4 Derivation of Kirchhoﬀ’s Relations . . . . 23. . . . . . . . . . . . . . . . . . . . . . . . . . . 22. .4 Schur Complement . . . . . .2.3 Uncertainty in A . . . . . . 23. . .1. . . .5 Nomenclature . . 22. . .1 General Case . . . . . . . . . . . . . . . . 25. . . . . . .4 Background for the Mapping Theorem . . . . . . . . . . . .2 Kirchhoﬀ’s Relations . . . . . . 25. . . . . . . . . . . . . . . . . . . . . . . . . 131 24. . . . . . . .3 Fluid Inertia Terms . . . .3. . . . . . . . . . . . . . . . . . .3. . .7 Quadratic Optimization with a Linear Constraint 25. . . . . . . . . . . . . . . . . 23. . . 25. . . . . . . . . . . .2 Dynamic Programming and Full-State Feedback . . . . . . 25. . . . . . . .10 Modal Decomposition . . . . . . . . 115 117 118 118 118 119 119 121 121 124 124 126 126 127 130 130 130 130 23 APPENDIX 2: ADDED MASS VIA LAGRANGIAN DYNAMICS 23. . . . . . . . . . . . . . . . . . . . . . .5. . . . . . . . . . . . . 25. . . .1. . . . . . .3. . . 23. . . . . . . . .1. . . . . . . . 23. . . . . . .1 Kinetic Energy of the Fluid . . . . .5 Proof of Schur Complement Sign . . . . . . . . . . . . .1 Example in the Case of Discrete States . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . 22. . . . . . . . . .3 Convolution Theorem . . . . . . . .4 Input Nonlinearities . . . . . . . . . . . . .1 Deﬁnition . . . . . . . .22. . . . . . 25. . . . . . . . . . . . 25. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Comments on Linear Matrix Inequalities (LMI’s) . . . . . . . . . Transform . . . . . . . . . . . . . . . . . 25. . . . . . . . . . . . . . . . . . . 132 25 Further Robustness of the LQR 25. . .1. . . . . . . . . . . . .2. . . . . . . . .9 Eigenvalues and Eigenvectors . . . . . . . . . . . . . . . . . . . . . . . .1 Tools . . . . . . . . .

Before the ﬁrst rotation. including the Euler angles and the Euler parameters (quaternions). pitch. z=z’ z’’ z’’’ y’’’ y y’=y’’ x’ x x’’=x’’’ Figure 1: Successive application of three Euler angles transforms the original coordinate frame into an arbitrary orientation. For example. some of them might use the same axis twice. it is easy to see that any possible orientation can be achieved. but with more abstraction. xb x xb 0 1 0 � ⎭ (1) . roll]: starting from an orientation identical to some inertial frame. then about the new pitch axis. We have cos δ sin δ 0 ⎛ ⎞ 1 γ = ⎝ − sin δ cos δ 0 ⎠ γ b0 = R(δ)γ 0 . One of the problems with Euler angles is that for certain speciﬁc values the transformation exhibits discontinuities.1 1 1. which we now derive through successive rotations of the three Euler angles. pitch. Needless to say. For the x x moment.1 KINEMATICS OF MOVING FRAMES Rotation of Reference Frames We denote through a subscript the speciﬁc reference system of a vector. Successively rotating the system about three of its own principal axes. then about the newer still roll axis. and in a body-reference frame γ b . A ﬁrst question is: what is the coordinate of a point ﬁxed in inertial space. referenced to a rotated body frame? The transformation takes the form of a 3×3 matrix. and has a solid link with the intuitive notions of roll. consider the sequence of [yaw. the body-referenced coordinate matches that of the inertial frame: γ b0 = γ . Now rotate the x x movable frame yaw axis (z) through an angle δ. rotate the movable system about its yaw axis. The former method involves successive rotations about the principle axes. Quaternions present a more elegant and robust method. Let a vector ex pressed in the inertial frame be denoted as γ . Tape three pencils together to form a right-handed three-dimensional coordinate system. We will develop the equations of motion using Euler angles. we assume that the origins of these frames are coincident. there are many valid Euler angle rotation sets possible to reach a given orientation. but that the body frame has a diﬀerent angular orientation. The angular orientation has several well-known descriptions. and yaw.

rotate the body system an angle ω about its newest x-axis: 1 0 0 ⎛ 3 γ b = ⎝ 0 x cos ω sin ω ⎞ γ b2 = R(ω)γ 2 . γ b2 = ⎛ x xb 0 1 ⎝ ⎠x sin χ 0 cos χ � ⎭ � ⎭ (2) Finally. including quaternions. we should note that the rotation matrix R is universal to all representations of orientation. ω)γ . x (4) All of the transformation matrices. x x x where γ 0 is the location of the moving origin. x (5) 1. In the case that the movable (body) reference frame has a diﬀerent origin than the inertial frame. as written. ω). using the small angle assumption to ignore higher-order terms gives 1 ζδ −ζχ ⎛ 1 ζω ⎞ R � ⎝ −ζδ ⎠ 1 ζχ −ζω � ⎭ (6) . are speciﬁc to Euler angles. expressed in inertial coordinates. i.e. are orthonormal: their inverse is equivalent to their transpose. Now apply the second rotation. xb ⎠x 0 − sin ω cos ω (3) This represents the location of the original point. χ. The roles of the trigonometric functions. including R(δ. γ b3 . in the fully-transformed body-reference frame. and to the order in which we performed the rotations. pitch about the new y-axis by the angle χ: cos χ 0 − sin χ 0 ⎞ γ b1 = R(χ)γ 1 .. Additionally.2 1 KINEMATICS OF MOVING FRAMES Rotation about the z-axis does not change the z-coordinate of the point. We will use the notation γ b instead of γ b3 from here on.2 Diﬀerential Rotations Now consider small rotations from one frame to another. we have γ = γ 0 + RT γ b . the other axes are modiﬁed according to basic trigonometry. χ. The three independent x x x rotations can be cascaded through matrix multiplication (order matters!): x γ b = R(ω)R(χ)R(δ)γ x ⎭ � cχcδ cχsδ −sχ ⎛ cωcδ + sωsχsδ sωcχ ⎞ γ = ⎝ −cωsδ + sωsχcδ ⎠x sωsδ + cωsχcδ −sωcδ + cωsχsδ cωcχ = R(δ.

1. This same cross-product relationship can be derived in the second frame as well: γ r γ b (t) = Rγ = γ − ζ E × γ x r r r γ b (t + ζt) = γ. x r r T (9) Dividing by the diﬀerential time step gives γ ζE ζγ x ×γ = r ζt ζt = � × γ. γ x γ b = γ − ζE × γ x x γ x γ = γ b + ζE × γ b. Since R−1 = RT . the order of γ the small rotations does not matter. R comprises the identity plus a part equal to the γ γ (negative) cross-product operator [−ζ E×]. instead of in inertial space. with respect to the ﬁrst frame as follows: γ (t) = γ x r γ r γ (t + ζt) = R γ = γ + ζ E × γ.2 Diﬀerential Rotations 0 ζδ −ζχ ⎛ 0 ζω ⎞ + I3×3 . x such that γ ζγ b x ζE ×γ = r ζt ζt = � × γ. the time rate of change of a constant radius vector is the cross-product of the rotation rate vector � and the radius vector itself. = ⎝ −ζδ ⎠ 0 ζχ −ζω � ⎭ 3 where I3×3 donotes the identity matrix. ζδ]. . so that r we can write the location of the point before and after the rotation. γ r (11) (12) On a rotating body whose origin point is ﬁxed. Small rotations are completely decoupled. x x (7) (8) We now ﬁx the point of interest on the body. z]. ζχ. the vector of Euler angles ordered with the axes [x. The γ resultant derivative is in the moving body frame. γ r (10) γ where the rotation rate vector � � dE/dt because the Euler angles for this inﬁnitesimal rotation are small and decoupled. calling its location in the body frame γ (radius). y. where ζ E = [ζω. The diﬀerential rotations occur over a time step ζt. we have also R−1 = I3×3 + ζ E ×.

and determine the eﬀects on the rotation vector. plus a component due to rotation of this γ frame. The velocity equation can be generalized to any body-referenced vector f : γ �f df γ = + � × f. γ dt �t (16) 1. An important part of any simulation is the evolution of the Euler angles.4 1 KINEMATICS OF MOVING FRAMES In the case that the radius vector changes with respect to the body frame. the Euler yaw angle (applied ﬁrst) is deﬁnitely not about the ﬁnal body yaw axis. v (15) �t where γo is the body-referenced velocity of the origin. with the sequence [yaw. the pitch and roll rotations moved the axis. The total velocity of the particle is v equal to the velocity of the reference frame origin. we need an additional term: dγ b x �γ r = � ×γ+ . The ﬁrst Euler angle undergoes two additional rotations.roll]. and the ﬁnal Euler angle no additional rotations: ⎧ 0 ⎧ ⎧ dω ⎧ 0 ⎧ ⎥ ⎢ ⎥ ⎢ dt ⎥ 0 + � = R(ω)R(χ) γ + R(ω) dζ 0 ⎧ ⎧ dt ⎧ ⎧ ⎧ dδ ⎧ ⎨ ⎣ ⎣ 0 ⎨ ⎣ 0 ⎨ dt ⎡ ⎧ ⎢ ⎤ ⎡ ⎤ ⎡ ⎤ (17) Taking the inverse gives 1 0 − sin χ ⎧ ⎢ ⎛ cos ω sin ω cos χ ⎞ = ⎝ 0 ⎠ ⎧ 0 − sin ω cos ω cos χ ⎣ � � ⎭⎡ dω dt dζ dt dδ dt ⎤ ⎧ ⎥ ⎧ ⎨ . Since the physics determine rotation rate � . = �(E)γ ⎭ (18) . we seek a γ γ mapping � ∀ dE/dt. dt �t dt This result is often written in terms of body-referenced velocity γ : v γ = � ×γ+ v γ r (13) (14) �γ r + γo .pitch. allowing the origin to move as well gives �γ dγ o r x dγ b x = �×γ+ γ r + . γ The idea is to consider small changes in each Euler angle. 1 sin ω tan χ cos ω tan χ γ dE ⎛ cos ω − sin ω ⎞ � = ⎝ 0 ⎠γ dt 0 sin ω/ cos χ cos ω/ cos χ γ �. the second angle one rotation.3 Rate of Change of Euler Angles Only for the case of inﬁnitesimal Euler angles is it true that the time rate of change of the Euler angles equals the body-referenced rotation rate. γ r dt �t Finally. For example.

2. First. Nonetheless. some of the most remarkable feats of navigation in history have depended on dead reckoning. dynamics expressed in the bodyreferenced frame need extra attention. In applications where this is a real possibility. if the estimated longitudinal speed over ground is U . linear momentum for a particle obeys the equality d γ (19) F = (mγ ) v dt A rigid body consists of a large number of these small particles. The summations we use below can be generalized to integrals quite easily. currents and vehicle sideslip will cause this to be in error.1.1 Momentum of a Particle Since the body moves with respect to an inertial frame. which can be indexed. and z-axis up. and other marine vehicles has the body-referenced xaxis forward.4 Dead Reckoning The measurement of heading and longitudinal speed gives rise to one of the oldest methods of navigation: dead reckoning. ignoring the lateral velocity leads to the evolution of Cartesian coordinates: x = U cos δ ˙ y = U sin δ. v dt (20) γ γ where Fi is the external force acting on the particle and Ri is the net force exerted by all the other surrounding particles (internal forces). because of the division by cos χ. 3β/2}. For most ocean vessels. and hence this otherwise useful equation for propagating the angular orientation of a body fails when the vehicle rotates about the intermediate y-axis by ninety degrees. A common frame for ships. and the estimated heading is δ. as long as it is not on the yaw axis! 1. 2 VESSEL INERTIAL DYNAMICS We consider the rigid body dynamics with a coordinate system aﬃxed on the body. the singularity is acceptable.4 Dead Reckoning 5 Singularities exist in � at χ = {β/2. This will be the sense of our body-referenced coordinate system here. Quite simply. We have d γ γ Fi + R i = (miγi ) . Since the collection of particles is not driven apart by the internal forces. we must have equal and opposite internal forces such that . for example in orbiting satellites and robotic arms. y-axis to port (left). quaternions provide a seamless mapping. submarines. ˙ Needless to say.

θ x. Further deﬁning the center of gravity vector γG v v γ r r . z. and yaw about the z-axis is positive turning left. and we now consider how the r momentum equation depends on this motion. pitch about the y-axis is positive bow-down. u. v i=1 dt (22) Note that the particle velocities are not independent. φ y. roll about the x axis is positive counterclockwise. w. (21) Then summing up all the particle momentum equations gives N ⎫ i=1 γ Fi = N ⎫ d (miγi ) . because the particles are rigidly at tached.6 2 VESSEL INERTIAL DYNAMICS N ⎫ i=1 γ Ri = 0. ψ Figure 2: Convention for the body-referenced coordinate system on a vessel: x is forward. v. in which the particle i resides at bodyreferenced radius vector γ. 2. and γi = γo + � × γi . the body translates and rotates. with origin 0. and z is heave upwards. y is sway to the left. Now consider a body reference frame. r γ = m + �× �t dt i=1 where m = such that �N i=1 mi . Looking forward from the vessel bridge.2 Linear Momentum in a Moving Frame The expression for total velocity may be inserted into the summed linear momentum equation to give N ⎫ i=1 N ⎫ i=1 γ Fi = d (mi (γo + � × γi )) v γ r dt ⎬ � (23) N ⎫ �γo v d miγi .

with radius r. being swung around around in a circle at speed U . (24) �γo v d γ Fi = m + m (γ × γG ). Z} (external force. γ � r � r � � γ r � and the resulting three linear momentum equations are ⎬ dq dr �u X = m + qw − rv + zG − yG + (qyG + rzG )p − (q 2 + r 2 )xG �t dt dt ⎬ � �v dr dp 2 2 Y = m + ru − pw + xG − zG + (rzG + pxG )q − (r + p )yG �t dt dt ⎬ � �w dp dq 2 2 Z = m + pv − qu + yG − xG + (pxG + qyG )r − (p + q )zG . The vector equation at the start is dγ � �γo v γ × γG + � × (γ × γG ) . body coordinates). �t dt dt (27) Note that about half of the terms here are due to the mass center being in a diﬀerent location than the reference frame origin. the complete vector equation in body coor dinates is �γo v dγ � γ N =m × γG + � × (γ × γG ) .e. � r (25) �t dt i=1 Using the expansion for total derivative again.3 Example: Mass on a String Consider a mass on a string. The centrifugal force can be computed in at least three diﬀerent ways. Y.3 Example: Mass on a String 7 mγG = r we have N ⎫ N ⎫ i=1 r miγi . zg } (body-referenced location of center of mass) = {p. 2. The last term in the previous equation simpliﬁes using the vector triple product identity � × (γ × γG ) = (γ · γG )γ − (γ · �)γG . in body coordinates) = {X.. w} (body-referenced velocity) = {xG . q. r γ � r F = + � × γo + γ v �t dt i=1 Now we list some conventions that will be used from here on: ⎫ � � (26) γo v γG r � γ γ F = {u. v. i.2. γG = γ r ≥ 0. + � × γo + γ v r γ � r F =m �t dt � � . r} (rotation vector. yG .

Uζω. 0}T .8 2. gives γo v � γ γG r �γo v �t �γ � �t such that γ F = mγ × γo = m{0. 0}T . 0}T = {0.3. 0}T � {0. 0. v ˙ Since ω = U/r. Aﬃxing the moving reference frame to the pivot point of the string.2 Rotating Frame Attached to Pivot Point = {U. with the local x oriented forward and y inward towards the circle center. can also be used for the calculation. r.3. 0.3. U sin ζω. A frame ﬁxed in inertial space. v . 0. U/r}T = {0. 0}T . U 2 /r. 0. 0}T = {0. � � r 2. with the same orientation as above but allowing it to rotate with the string. U/r}T = {0. it follows easily that in the ﬁxed frame dγ /dt = {0. 2.3. 0}T = {0. as before.3 Stationary Frame = {0. 0. we have γo v � γ γG r �γo v �t �γ � �t giving the same result: γ F = mγ × (γ × γG ) = m{0. and momentarily coincident with the frame on the mass (2.1). 0. U 2 /r. U 2 /r. as the string travels through a small arc ζω. 0}T = {0. 0. 0}T . 0}T = {0. 0.1 Moving Frame Aﬃxed to Mass 2 VESSEL INERTIAL DYNAMICS Aﬃxing a reference frame on the mass. � v The force of the string pulls in on the mass to create the circular motion. 0. 0}T = {0. vector subtraction gives ζγ = {0. 0}T . In this case. 0}T .

the summed particle equation is γ γ (Mi + γi × Fi ) = r γi × r d (miγi ). Similar to the case for linear momentum. r � � r The summation in the ﬁrst term of the right-hand side is recognized simply as mγG . Ixx Ixy Ixz ⎛ ⎞ I = ⎝ Iyx Iyy Iyz ⎠ Izx Izy Izz Ixx = Iyy = Izz = N ⎫ 2 2 mi (yi + z ) i i=1 N ⎫ i=1 N ⎫ i=1 ⎡ �N 2 ⎧ ⎢ �i=1 mi ((yi N 2 i=1 mi ((xi ⎧ ⎣ �N 2 ⎭ N ⎫ i=1 mi � �γ � �γ � − · γi γi (γi · γi ) r r r r �t �t � � � ⎤ (30) 2 ˙ ˙ ˙)x ⎥ + zi )p − (yi q + zi r i ) ⎧ 2 + zi )q − (xi p + zi r i ) . We have N ⎫ N ⎫ �γo v �γ � γ γ miγi × r (Mi + γi × Fi ) = r r + � × γo + miγi × γ v r × γi + �t �t i=1 i=1 i=1 N ⎫ i=1 N ⎫ ⎬ � � � miγi × (γ × (γ × γi )).2. summed internal moments cancel. and the r ﬁrst term becomes �γo v + � × γo . v dt (28) γ where Mi is an external moment on the particle i.4 Angular Momentum 9 2. ˙ ˙ ˙)y ⎧ ⎨ 2 ˙ ˙ ˙ i=1 mi ((xi + yi )r − (xi p + yi q)zi ) � (inertia matrix) 2 mi (x2 + zi ) i 2 mi (x2 + yi ) i N ⎫ i=1 Ixy = I = − yx m i xi y i (cross-inertia) . γ v mγG × r �t The second term expands as (using the triple product) N ⎫ ⎬ � (29) �γ � × γi miγi × r r �t i=1 � � = = Employing the deﬁnitions of moments of inertia.4 Angular Momentum N ⎫ i=1 N ⎫ i=1 For angular momentum.

e. ˙ ˙ (32) (Izz − Iyy )rq (I − Izz )rp ⎧ xx ⎣ (Iyy − Ixx )qp ⎤ ⎧ ⎥ ⎧ ⎨ . with no external forces or moments. 2.. but oﬀers no similar condensation: N ⎫ i=1 i=1 N ⎫ i=1 N ⎫ m i xi z i mi y i z i . The � third term can be worked out along the same lines. M.10 2 VESSEL INERTIAL DYNAMICS Ixz = Izx = − Iyz = Izy = − the second term of the angular momentum right-hand side collapses neatly into I�γ /�t. i.5 Example: Spinning Book Consider a homogeneous rectangular block with Ixx < Iyy < Izz and all oﬀ-diagonal moments of inertia are zero. are . The linearized angular momentum equations. N } be the total moment acting on the body. the left side of Equation 28. miγi × ((γ · γi )γ − (γ · �)γi ) = r � r � � γ r = N ⎫ i=1 miγi × �(γ · γi ) r γ � r ⎤ ⎧ ⎥ (31) Iyz (q 2 − r 2 ) + Ixz pq − Ixy pr I (r 2 − p2 ) + Ixy rq − Iyz pq = ⎧ xz 2 ⎣ Ixy (p − q 2 ) + Iyz pr − Ixz qr ⎡ ⎧ ⎢ ⎡ �N ⎧ ⎢ � i=1 mi (yi r − zi q)(xi p + yi q + zi r) N i=1 mi (zi p − xi r)(xi p + yi q + zi r) ⎧ ⎣ �N i=1 ⎡ ⎧ ⎢ ⎧ mi (xi q − yi p)(xi p + yi q + zi r) ⎨ ⎤ ⎧ ⎥ ⎧ ⎨ + γ Letting M = {K. the complete moment equations are K = Ixx p + Ixy q + Ixz r + ˙ ˙ ˙ (Izz − Iyy )rq + Iyz (q 2 − r 2 ) + Ixz pq − Ixy pr + m [yG (w + pv − qu) − zG (v + ru − pw)] ˙ ˙ M = Iyx p + Iyy q + Iyz r + ˙ ˙ ˙ (Ixx − Izz )pr + Ixz (r 2 − p2 ) + Ixy qr − Iyz qp + m [zG (u + qw − rv) − xG (w + pv − qu)] ˙ ˙ N = Izx p + Izy q + Izz r + ˙ ˙ ˙ (Iyy − Ixx )pq + Ixy (p2 − q 2 ) + Iyz pr − Ixz qr + m [xG (v + ru − pw) − yG (u + qw − rv)] .

2. where the ζ preﬁx indicates a small value compared to Δ. The imaginary exponent indicates that the solution is of the form ζq(t) = > ζq(0)cos ∂t. �t2 Izz Ixx Here the second coeﬃcient has negative sign. The interesting result is that rotations about the x and z axes are stable. Diﬀerentiate the second equation to obtain � 2 ζq �ζr + (Ixx − Izz )Δ =0 2 �t �t Iyy Substitution of this result into the third equation yields � 2 ζq Iyy Izz 2 + (Ixx − Izz )(Ixx − Iyy )Δ2 ζq = 0. �t A simpler expression is ζq +∂ζq = 0. p = Δ + ζp. and indicates no change in ζp. 2. and r = ζr. This is easily demonstrated experimentally with a book or a tennis racket. q = ζq.2. that is. . which has response ζq(t) = ζq(0)e −∂t . The exponent is real now.2 y-axis � Now suppose q = Δ+ζq: diﬀerentiate the ﬁrst equation and substitute into the third equation to obtain � 2 ζp + (Iyy − Ixx )(Iyy − Izz )Δ2 ζp = 0.5. while rotation about the y axis is not. � and the solution grows without bound. and hence ∂ ≈ 0. both coeﬃcients of the diﬀerential equation are positive. when ζq ¨ ˙(0) = 0. it too oscillates. The ﬁrst equation above is uncoupled from the others. and therefore ∂ < 0.5 Example: Spinning Book 11 dp + (Izz − Iyy )rq = 0 dt dq Iyy + (Ixx − Izz )pr = 0 dt dr Izz + (Iyy − Ixx )qp = 0. For spin about the x-axis. since the small term ζqζr can be ignored. following ζp(t) = ζp(0)e −∂t .1 x-axis In the case of the x-axis. with constant angular rate Δ. Since the perturbation ζr is coupled. it oscillates but does not grow.5. dt Ixx We consider in turn the stability of rotations about each of the main axes.

Iyz .7 Basis for Simulation γ γ Except for external forces and moments F and M . for example.3 z-axis 2 VESSEL INERTIAL DYNAMICS Finally. 2. 2 �t Iyy Ixx The coeﬃcients are positive. 2. Ixz small enough that they can be ignored. we now have the necessary terms for writing a full nonlinear simulation of a rigid body. and vice versa.12 2. location of the body origin. Note that translation of MMOI using the parallel axis theorem must be either to or from a frame resting exactly at the center of gravity. Sometimes a translation of coordinates to the mass center will make the cross-inertial terms Ixy . so that the equations of motion are signiﬁcantly reduced. and ζx.5. x .6 Parallel Axis Theorem Often. There are twelve states. the mass center of an body is at a diﬀerent location than a more convenient measure ment point. The parallel axis theorem allows one to translate the mass moments of inertia referenced to the mass center into another frame with parallel orientation. v • �. The formulas are: Ixx Iyy Izz Iyz Ixz Ixy = = = = = = ¯ Ixx + m(ζy 2 + ζz 2 ) ¯ Iyy + m(ζx2 + ζz 2 ) ¯ Izz + m(ζx2 + ζy 2 ) ¯ Iyz − mζyζz ¯ Ixz − mζxζz ¯ Ixy − mζxζy. comprising the following components: • γo . is the translation of the x-axis to the new frame. the geometric center of a vessel for example. (33) ¯ where I represents an MMOI in the axes of the mass center. body rotation rate vector. so bounded oscillations occur. r 0 as in the spinning book example. in body coordinates. γ • γ . the vector of body-referenced velocities. in inertial space. let r = Δ+ζr: diﬀerentiate the ﬁrst equation and substitute into the second equation to obtain � 2 ζp + (Ixx − Izz )(Iyy − Izz )Δ2 ζp = 0. in this case γG = γ also.

E = �(E)γ (35) (34) 3 NONLINEAR COEFFICIENTS IN DETAIL The method of hydrodynamic coeﬃcients is a somewhat blind series expansion of the ﬂuid force in an attempt to provide a framework on which to base experiments and calculations to evaluate these terms. The basic assumption in using the method of hydrodynamic coeﬃcients is that the forces have no memory eﬀects. based on Newton’s second law. We retain only ﬁrst order acceleration terms. we expect the inertia terms from the ﬂuid to be linearly dependent on acceleration. or when large vortices are shed. this is a useful property to us in order to eliminate a certain number of coeﬃcients which are either zero or very small. . 2. We do not include terms coupling velocities and accelerations. a computation of these forces. past motions have no impact on the ﬂuid forces at the present moment. x while the Euler angles follow: ˙ γ γ �. because then the vorticity in the ﬂuid aﬀects the ﬂuid forces for a considerable time after they have been shed – i. Euler angle vector. 3.13 γ • E. a propeller introduces an asymmetry port/starboard since it rotates in a certain direction (unless it is a twin-screw ship with counter-rotating propellers). i. Unless there is a reason not to. since a number of symmetries and basic laws can be applied to reduce the number of unknown coeﬃcients. Still. i.e.e. Based on Newton’s second law. We employ the following basic facts and assumptions to derive the ﬂuid forces acting on a ship. The basic diﬁculty. We consider port/starboard symmetry. The Cartesian position propagates according to ˙ γ v γ = RT (E)γo . we are not totally ignorant about these forces. at least today and in the near future. because under certain conditions such eﬀects can not be ignored. submarine or vehicle: 1. we expect inertia forces to depend on acceleration alone. We will show later some methods which allow us to incorporate the eﬀect of shed vorticity.e. the intractability of the governing equations of motion of viscous ﬂuid prohibits. In ships. until they move suﬃciently far away from the body. with some coupling which generally requires a 6 × 6 matrix inverse. This is not correct when the ﬂow separates. Again. but in such cases we limit this asymmetry to only propeller-related terms. This is the purpose of this section. The derivatives of body-referenced velocity and rotation rate come from Equations 27 and 32.

1 Helpful Facts To exploit symmetries. so their inclusion is motivated by physical arguments. we ﬁnd sometimes necessary to use coeﬃcients such as Y|v|v . serve well the purpose of deriving equations which are suﬃciently accurate in a wide parametric range... 3. The proof is to consider the Taylor series expansion of f (x) about x = 0: 1 d2 f 1 d3 f df 2 (0)x + (0)x3 + .e. We retain up to third order terms. for n even: dn f (0) = 0 dxn (40) (41) . using the anti-symmetry condition: f (x) = −f (−x) to ﬁnd that. in general.14 3 NONLINEAR COEFFICIENTS IN DETAIL 4. we consider the following simple facts: • A function f (x) which is symmetric in x has zero odd derivatives with respect to x at x = 0. but most existing models employ this assumption Finally. This is a practical consideration and has been found to. providing a drag-related term.. whose strict deﬁnition would be: Y |v |v = �2Y (v = 0) �v� |v | (36) A Taylor series expansion would not include such terms. f (x) = f (0) + (0)x + 2 3 dx 2! dx 3! dx Symmetry gives that for all x: f (x) = f (−x) (38) (37) The only way that (38) is satisﬁed given the expression (37) is that all odd derivatives of f (x) must be zero. i. The proof is exactly analogous to the result above. for n odd: dn f (0) = 0 dxn (39) • A function f (x) which is anti-symmetric in x has zero even derivatives with respect to x at x = 0. It does not constitute an absolute rule.

r. −ζ) (44) (45) (46) These relations imply that all odd derivatives of X with respect to v at v = 0 are zero.e. when r = 0 and ζ = 0. employing the assumptions above. ζ = 0) X(u. r = 0. v = 0.3. a symmetric function of r when v = 0 and ζ = 0. ζ = 0) = X(u. −v. valid up to third order: X = Xe + Xu u + Xu u + Xuu u2 + Xuuu u3 + Xvv v 2 + Xrr r 2 + Xνν ζ 2 + ˙ ˙ Xrv rv + Xrν rζ + Xvν vζ + Xvvu v 2 u + +Xrru r 2 u + Xννu ζ 2 u + Xrνu rζu + Xrvu rvu + +Xvνu vζu + Xrvν rvζ (43) We have used three basic properties.1 Fluid Force X (42) By denoting the rudder angle as ζ. r = 0. v = 0. r = 0.2 Nonlinear Equations in the Horizontal Plane 15 3. v = 0.2. which include the ﬂuid forces: m(u − vr − xG r 2 ) = X ˙ m(v + ur + xG r) = Y ˙ ˙ Izz r + mxG (v + ur) = N ˙ ˙ where we have assumed that w = 0. ζ = 0) X(u. yG = 0. we will derive the governing nonlinear equations of motion in the horizontal plane (surge. we derive the following expression for the ﬂuid force X.2 Nonlinear Equations in the Horizontal Plane To demonstrate the methodology. q = 0. zG = 0. 2. ζ = 0) = X(u. a symmetric function of ζ when r = 0 and v = 0. a symmetric function of v when r = 0 and ζ = 0. v = 0. independent of the forward velocity. We start by re-stating the inertia terms driven by external forces. ζ) = X(u. ζ = 0) = 0 �v 3 (47) .. must be: 1. that the ﬂuid force X. sway and yaw). v. We will not include drag related terms of the form of equation 36 for the time being. r = 0. i. 3. −r. 3. This is a result of port/starboard symmetry and is expressed as: X(u. r = 0. p = 0. and similarly for r and ζ. v = 0. For example: �3X (u.

: Y (u. ζ = 0) Y (u. −ζ) (50) Hence. v = 0. −r. ζ = 0) = 0 �v 3 �u or equivalently Xvvvu = 0. ζ = 0) = −Y (u. and ζ) must be zero. or: Yvv = 0. the even derivatives of Y with respect to v (and then r. ζ) = −Y (u. Xruu = 0.2 Fluid Force Y (48) (49) In the case of the ﬂuid foce Y . v = 0. Xννν = 0. Finally. which break the symmetry. except for propeller eﬀects. For example. since relations such as (44) hold for all forward velocities u. Xvvv = 0. r = 0. −v. we will include terms such as Yu to allow for the propeller asymmetry (twin propeller ships with counter-rotating propellers have zero such terms). Xrrr = 0. we derive the following expansion for Y : Y = Ye + Yu u + Yuu u2 + Yv v + Yr r + Yv v + Yr r + Yν ζ + Yνu ζu + Yvu vu + ˙˙ ˙˙ 2 2 Yru ru + Yvuu vu + Yruu ru + Yνuu ζu2 + Yvvv v 3 + Yrrr r 3 + Yννν ζ 3 + Yrrν r 2 ζ + Yrrv r 2 v + Yvvr v 2 r + Yvvν v 2 ζ + Yvrν vrζ + Yννr ζ 2 r + Yννv ζ 2 v (52) . r = 0. r = 0. Xr = 0.e. v = 0. i. r = 0. Xvuu = 0. Xνuu = 0 3.2. due to port/starboard symmetry. v = 0. Xvu = 0. ζ = 0. Xνu = 0. the symmetries provide the following zero coeﬃcients: Xv = 0. In summary. Yvvu = 0 Yrr = 0. ζ = 0. Xru = 0. v = 0.16 3 NONLINEAR COEFFICIENTS IN DETAIL which implies that Xvvv = 0. in analogy with the previous section. r. r = 0. the symmetry implies that this force must be an antisymmetric function of v when r = 0. Yννu = 0 (51) Finally. For this reason. the force Y should not be aﬀected by u when v = 0. Also. from (47) we derive: �4X (u. it means that derivatives with respect to u of expression (44) are aso true. Yrru = 0 Yνν = 0. r = 0. v. and likewise for r and ζ. ζ = 0) Y (u. ζ = 0) = −Y (u. Xν = 0.

and eliminating higher-order terms.17 3. Since the vessel is symmetric about the x − z plane. we ﬁrst ﬁnd that the following coeﬃcients are zero: Nvv = 0.1 VESSEL DYNAMICS: LINEAR CASE Surface Vessel Linear Model We ﬁrst discuss some of the hydrodynamic parameters which govern a ship maneuvering in the horizontal plane. i. We then have at the outset �u X = m − rv − xG r 2 �t � � �v �r Y = m + ru + xG �t �t � � �r �v N = Izz + mxG + ru . so positive r has the vessel turning left. �t �t (56) � � (55) . where U >> u. Nrru = 0 Nνν = 0. Nvvu = 0 Nrr = 0. zG is inconsequential. The body x-axis is forward and the y-axis is to port. (53) (54) 4 4.3 Fluid Moment N The derivation for N follows the same exact steps as for the side force Y . �t �t Letting u = U + u.2. the same symmetries apply.e. We will consider motions only in the horizontal plane. this set is X = m �u �t � � �v �r Y = m + rU + xG �t �t � � �r �v N = Izz + mxG + rU . As a result. yG = 0. Nννu = 0 Hence. we derive an analogous expansion for the moment: N = Ne + Nu u + Nuu u2 + Nv v + Nr r + Nv v + Nr r + Nν ζ + Nνu ζu + ˙˙ ˙˙ 2 Nvu vu + Nru ru + Nvuu vu + Nruu ru2 + Nνuu ζu2 + Nvvv v 3 + Nrrr r 3 + Nννν ζ 3 + Nrrν r 2 ζ + Nrrv r 2 v + Nvvr v 2 r + Nvvν v 2 ζ + Nvrν vrζ + Nννr ζ 2 r + Nννv ζ 2 v. which means χ = ω = p = q = w = 0.

Xy . We have so far. with no current. Thus Xv = 0. it must have zero slope with v at the origin. Y ∗ .1 Second. Xv . which is always invertible. Nδ } are all zero. this topic will be discussed later.2 Stability of the Sway/Yaw System ˙ Consider the homogeneous system γ = Aγ: s s s1 = A11 s1 + A12 s2 ˙ s2 = A21 s1 + A22 s2 . {Xx . We assume that no hydrodynamic forces depend on the position of the vessel. N ∗ }. Nu } = 0. we do not consider nonlin ear acceleration terms. Nu . It should be noted that some nonlinear terms related to those we have eliminated above are not zero. Yx . First. Xr . but in general Xvv = 0 only if the vessel is bow-stern symmetric. or higher time derivatives. and therefore are of immediate interest. r} and external force/moment vector F = {Y ∗ . 1 . Yuu = 0 because of hull symmetry. in a homoge neous sea. (m − Xu )u = Xu u + X ∗ ˙ ˙ (m − Yv )v + (mxG − Yr )r = Yv v + (Yr − mU )r + Y ∗ ˙ ˙ ˙ ˙ (mxG − Nv )v + (Izz − Nr )r = Nv v − (Nr − mxG U )r + N ∗ . or wind eﬀects. wave. consider Xv : since this longitudinal force would have the same sign regardless of the sign of v (because of side-to-side hull symmetry). With the state vector γ γ = {v. considering only the linear hydrodynamic terms. 4. For instance. as noted in the last chapter. since ˙ ˙ ˙ ˙ ﬂuid particle acceleration relates linearly with pressure or force. or s γ (60) (61) The matrix M is a mass or inertia matrix. a state-space representation of s the sway/yaw system is ⎬ m − Yv mxG − Yr ˙ ˙ mxG − Nv Izz − Nr ˙ ˙ � dγ s dt ˙ Mγ s ˙ γ s ˙ γ s = = = = Yv Yr − mU Nv Nr − mxG U Pγ + F s γ γ M −1 Pγ + M −1 F s γ Aγ + B F . Yu . and B is a gain matrix for the control and disturbance inputs. Note that the surge equation is decoupled from the sway and yaw. ˙ ˙ ˙ ˙ (57) (58) (59) The right side here carries also the imposed forces from a thruster(s) and rudder(s) {X ∗ . Yy . N ∗ }.18 4 VESSEL DYNAMICS: LINEAR CASE A number of coeﬃcients can be discounted. Finally. Nx . The same argument shows that {Xr . ˙ Note that the linearized heave/pitch dynamics of a submarine do depend on the pitch angle. Xδ . Ny . Yδ . The last form of the equation is a standard one wherein A represents the internal dynamics of the system. but that sway and yaw themselves are coupled. s ⎬ � γ + F . Yu .

Yr . if the vessel is reasonably balanced with regard to forward and aft areas with respect to the origin. we would simply use s. Yr } take very small values in comparison with ˙ ˙ the others. the terms {Nv . both large and negative. For the second condition. and ˙ ˙ A11 = Yv m−Yv ˙ Nr Izz −Nr ˙ < 0 A22 = < 0. we require . valid for many vessels with the origin is at the geometric center. these are the terms Yv and Nr . we have only to look at the numerators. in the lan guage of the Laplace transform. First. Next. Linear drag and ˙ ˙ ˙ rotational drag are signiﬁcant also. The denominator for A’s components reduces to (m − Yv )(Izz − Nr ). To wit. the added mass term −Yv is of the order of the vessel’s material mass ˙ m. A necessary and suﬃcient condition for stability of this ODE system is that each coeﬃcient must be greater than zero: −A11 − A22 > 0 A11 A22 − A12 A21 > 0 The components of A for the sway/yaw problem are (Izz − Nr )Yv + (Yr − mxG )Nv ˙ ˙ (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ −(Izz − Nr )(mU − Yr ) − (Yr − mxG )(mxG U − Nr ) ˙ ˙ (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ (Nv − mxG )Yv + (m − Yv )Nv ˙ ˙ (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ −(Nv − mxG )(mU − Yr ) − (m − Yv )(mxG U − Nr ) ˙ ˙ . Both Yv and Nr take large negative values. (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ (64) A11 = A12 = A21 = A22 = (65) The denominators are identical.2 Stability of the Sway/Yaw System We can rewrite the second equation as s2 = � 19 d(·) − A22 dt �−1 A21 s1 (62) and substitute into the ﬁrst equation to give s1 + (−A11 − A22 )s1 + (A11 A22 − A12 A21 )s1 = 0. xG = 0. For stability. If the origin is at the center of mass. and similarly Nr � −Izz . Hence the ﬁrst condition for stability is met: −A11 − A22 > 0. Nv . since the denominators of the Aij are identical. let xG � 0. ¨ ˙ (63) Note that these operations are allowed because the derivative operator is linear. and can be simpliﬁed.4.

The terms of C compete.3 Basic Rudder Action in the Sway/Yaw Model Rudders are devices which develop large lift forces due to an angle of attack with respect 1 to the oncoming ﬂuid. positive xG increases the (positive) inﬂuence of C’s ﬁrst term.20 4 VESSEL DYNAMICS: LINEAR CASE (Izz − Nr )Yv (m − Yv )Nr ˙ ˙ − [Nv Yv + (m − Yv )Nv ] [−(Izz − Nr )(mU − Yr ) + Yr Nr ] > 0. We have . When only the largest terms are considered for a vessel. 4. Cl (∂) = � �∂ �∂=0 � (69) Since the rudder develops force (and a small moment) far away from the body origin. generally making stability critical on the (usually negative) Nv . (67) C is called the vessels stability parameter. The basic form is as follows: L = 2 πAU 2 Cl (∂). ˙ ˙ ˙ ˙ (66) The ﬁrst term is the product of two large negative and two large positive numbers. Adding more surface area aft drives Nv more positive. The lift coeﬃcient Cl is normally linear with ∂ near ∂ = 0. The second part of the second term contains mU . which has a large positive value. We will assume that ∂ is small enough that the linear relationship applies: �Cl � � ∂. Stability can also be improved by moving the center of gravity forward. y δ r x Figure 3: Convention for positive rudder angle in the vessel reference system. but the rudder stalls when the angle of attack reaches a critical value. where ∂ is the angle of attack. (68) Since Nr and Yv are both negative. say a distance l aft. the moment equation is quite simple. Nonzero xG shows up as follows: C = Yv (Nr − mxG U ) + Nv (mU − Yr ) > 0. and thereafter develops much less lift. and yaw/sway stability depends closely on the magnitude and sign of Nv . a simpler form is common: C = Yv Nr + Nv (mU − Yr ) > 0. increasing stability as expected.

4. 4. On the other hand. which are used to track δdesired .2 Heading Control in the Sway/Yaw Model Considering just the yaw equation of motion. with a rudder. but in real conditions. 4.. In this case the rudder still opposes yaw and sway perturbations and acts to stabilize the vessel. Thus. � 2 �∂ �∂=0 1 �Cl � � Y∂ = πA U2 � � 2 �∂ ∂=0 � � (70) (71) The stability coeﬃcient C resulting from the addition of a control law ζ = kr r. ˙ (74) Since all coeﬃcients are positive (recall Nν < 0). v = 0. A properly-balanced vessel just achieves stability with zero rudder action.1 Adding Yaw Damping through Feedback 1 �Cl � � N∂ = − πA lU 2 . is C = Yv (Nr − mxG U + kr Nν ) + Nv (mU − Yr − kr Yν ). (72) Yν is small positive. even when ζ = 0. where kr > 0 is a feedback gain.3 Basic Rudder Action in the Sway/Yaw Model 21 Note the diﬀerence between the rudder angle expressed in the body frame. Angle of attack is inﬂuenced by ζ. Hence C becomes more positive. a vessel which is overly stable (C >> 0 with no rudder action) is unmaneuverable. the system equation becomes a homogeneous.e. This use of an error signal . ˙ (73) Employing the control law ζ = kδ δ. as well as v/U and lr. ζ. for example Nv and Nr . we have ¨ ˙ (Izz − Nr )δ + (mxG U − Nr )δ = Nν ζ. Generally speaking. and the total angle of attack ∂. i.3. The control law ζ = kδ (δ − δdesired ) + kr r is the basis for heading autopilots. settling under second-order dynamics to χ(∗) = 0. A positive rudder deﬂection (deﬁned to have the same sense as the yaw angle) causes a negative yaw perturbation. ζ = ∂ and Nν = N∂ . so that a reasonable amount of control will provide good maneuvering capabilities. the hydrodynamic characteristics of the vessel depend strongly on the rudder.. secondorder ODE: ¨ ˙ (Izz − Nr )δ + (mxG U − Nr )δ − Nν kδ δ = 0. in tank testing with v = 0. but Nν is large and negative. since Yv is negative. the equation gives a stable χ response. other hydrodynamic derivatives are augmented to capture the necessary eﬀects. and a very small positive sway perturbation. etc.3. Control system limitations and the stalling of rudders make obvious the fact that even a very large control gain kr cannot completely solve stability problems of a poorly-designed vessel with an inadequate rudder.

(77) Izz − Nr ˙ The ﬁrst phase is followed by a period (Phase 2). In this case. since the velocities are zero. (78) = N v Yν − Y v N ν r C Note that the denominator is the stability parameter.22 4 VESSEL DYNAMICS: LINEAR CASE to drive an actuator is in fact the essence of feedback control. (76) m − Yv ˙ and the vessel moves to the left. 4. a controller to calculate ζ. (80) r −Yv Nν ζ The radius goes up directly with C. R= � � � � . indicating that too stable a ship has poor turning performance. We see also that increasing the rudder area increases Nν .4. the vessel is in a steady turning condition. mxG − Nv Izz − Nr r ˙ Nν ˙ ˙ Since Yr and mxG are comparatively small in the ﬁrst row. (79) C With C > 0. Increasing the deﬂection ζ to reduce R works only to the point of stalling. This turning rate equation can also be used to estimate turning radius R: r=− U UC = . since Nν < 0: Nν ζ . decreasing R as desired. The equation looks like this: m − Yv mxG − Yr v ˙ Yν ˙ ˙ = ζ. the positive v-direction.4 4. and the accelerations are zero. and an actuator to implement the corrective action.4.2 Phase 3: Steady State When the transients have decayed. The steady turning rate is thus approximated by Yv N ν ζ. we have ˙ v ˙(0) = ⎬ �� � � � (75) Yν ζ . If the vessel is unstable (C < 0). we require sensors to obtain r and δ. the steady-state yaw rate is negative. The initial yaw is in the negative r-direction.1 Response of the Vessel to Step Rudder Input Phase 1: Accelerations Dominate When the rudder ﬁrst moves. acceleration terms dominate. r ˙(0) = 4. in which many terms are competing and contributing to the transient response. The system equations reduce to ζ v (mxG U − Nr )Yν + (Yr − mU )Nν . it turns in the opposite direction than expected.

On the basis of physical laws. 3.1 SIMILITUDE Use of Nondimensional Groups For a consistent description of physical processes. lb denotes the vertical separation of the center of gravity and the center of buoyancy. independent quantities. 5 5. for which the dimensional weight/buoyancy forces and moments are invariant with speed. and “ hands-oﬀ” stability.4. Among other eﬀects. with xG = 0 and no actuation: �w − Uq = Zw w + Zw w + Zq˙ q + Zq q + (B − W ) ˙ m ˙ ˙ �t dq = Mw w + Mw w + Mq˙ q + Mq q − Blb sin χ. and the actual vessel loses forward speed during maneuvering. and they are independent of the longitudinal dynamics. This fact is not true in the vertical plane. ˙ Iyy ˙ ˙ dt � � (81) (82) The last term in each equation is a hydrostatic eﬀect induced by opposing net buoyancy B and weight W . Because buoyancy eﬀects do not change with speed. creating the so-called righting moment which nearly all underwater vehicles possess. the dynamic properties and hence stability of the vehicle may change with speed. The linearized sway/yaw dynamics of a surface vessel are strongly coupled. The Buckingham β-theorem provides a basis for all nondimensionalization. Very tight ma neuvering requires the nonlinear inertial components and hydrodynamic terms. some quantities are dependent on other.6 Stability in the Vertical Plane Stability in the horizontal plane changes very little as a function of speed. because drag and lift eﬀects generally scale with U 2 . The design parameter C should be slightly greater than zero for easy turning. 4.5 Summary of the Linear Maneuvering Model 23 4. The case C < 0 should only be considered under active feedback control. The analysis is valid only up to small angles of attack and turning rates. we require that all terms in an equation must have the same units. Let a quantity Qn be given as a function of a set of n − 1 other quantities: . For example.5 Summary of the Linear Maneuvering Model We conclude our discussion of the yaw/sway model by noting that 1. and apply the technique to maneuvering. We form nondimensional groups out of the dimensional ones in this section. the nonlinear equations couple surge to the other motions. consider the case of heave and pitch. 2.

through a drain nozzle of diameter d. and one is a unique function of the other two. Suppose we have a block of mass m resting on a frictionless horizontal surface. and therefore k = 3. mass]. but suppose also that there are only k independent ones. Consider the ﬂow rate Q of water from an open bucket of height h. To . Hence.24 5 SIMILITUDE There are n variables here. The dimensional function is X(T ) = fQ (m. not far at all from the known result that X(T ) = F T 2 /2m! Example. and the (MKS) units of these quantities are: [Q] [h] [d] [π] [µ] [g] = = = = = = m3 /s m m kg/m3 kg/ms m/s2 There are only three units that appear: [length. π. (83) Example. so n = 4. The theorem asserts that there are n − k dimensionless groups βi that can be formed. g is the acceleration due to gravity. We have Q = fQ (h. βn−k−1 ). Qn−1 ). At time zero. · · · . where the water density is π. Simple term-cancellation gives β1 = X(T )m/F T 2 . d. No other parameters aﬀect the ﬂow rate. Q2 . time. T ). We have n = 6. g). The (MKS) units are [X(·)] [m] [F ] [T ] = = = = m kg kgm/s2 s. and thus k = 3. a steady force of magnitude F is applied. µ. · · · . β1 assumes only a constant (but unknown) value. the distance that the block has moved as of time T . and the functional equivalence is reduced to βn−k = fψ (β1 . k is equivalent to the number of physical unit types encountered. (84) Qn = fQ (Q1 . and its absolute viscosity µ. We want to know X(T ). F. There is just one nondimensional group in this relationship. β2 . only three non-dimensional groups exist.

5. . g is the acceleration due to gravity. as we now show. and U the propeller inlet velocity. µ d d2 gh The uncertainty about where to use h and d. . Cavitation number: where U is the speed of the vessel. they could easily change places in the ﬁrst two groups. making sure that each of the original (dimensional) quantities is represented. ζ= P√ − P v . with no constraints about the actual physics taking place. and L is the waterline length of the vessel. Froude number: U Fr = ≈ . and then β5 = β1 /β4 = πd gh/µ to establish an oriﬁce Reynolds number independent of Q. Since h and d have the same units. 5. such as in waves on the ocean surface. it relates the vessel speed U to water wave speeds of wavelength L. The Froude number appears in problems involving pressure boundary conditions. ≈ Three plausible ﬁrst groups are: β1 = πQ/dµ. The power of the β-theorem is primarily in reducing the number of parameters which must be considered independently to characterize a process. Intuition and additional manipulations come in handy. the theorem reduced the number of independent parameters from ﬁve to two.2 Common Groups in Marine Engineering 25 arrive at a set of valid groups.. β1 is recognized as a Reynolds number pertaining to the oriﬁce ﬂow. β2 ) −∀ � ≈ � Q πd gh h ≈ = fψ . Note that all six quantities appear at least once. e. but products and fractions of groups are ≈ themselves valid groups. we must create three nondimensional quantities. gL (85) 2. β2 = dπ gh/µ. 1 πU 2 2 (86) where P√ represents the ambient total pressure.2 Common Groups in Marine Engineering One frequently encounters the following groups in ﬂuid mechanics and marine engineering: 1.g. In the ﬂow example. Pv the vapor pressure of the ﬂuid. β2 is more awkward. A low cavitation number means that the Bernoulli pressure loss across the lifting surface will cause the ﬂuid to vaporize. Roughly speaking. and the questionable importance of h/d as a group are remnants of the theorem. and possible deterioration of the material. However. and we may construct β4 = β≈ 2 = Q/d2 gh to nondimensionalize 1 /β Q with a pressure velocity. We ﬁnally have the useful result β4 = fψ (β5 . where η is the wavelength. degradation of performance. causing bubbles. and the other two terms have a nice physical meaning. and β3 = h/d. β5 is a Reynolds number. Intuition is that h/d is immaterial. the phase speed of � a surface wave is V = ηg/2β.

26 3. (inertia) Fi = π (gravity) Fg = (pressure) Fp = (shear) Fs = (surface tension) Fθ = Thus the groups listed above can be written as Fi Fg Fp ζ= Fi Fi Re = Fs Fi W = Fθ U2 gl P � πU 2 πU l � µ πU 2 l � θ � Fr = When testing models. and l normalizes length. l represents one of many length scales. and π is density. 4. Reynolds number: Re = Ul . Reynolds number is a ratio of ﬂuid inertial pressures to viscous pressures: When Re is high. Various forces on the box scale as �v �v dxdydz + πv dxdydz � πU 2 l2 �t �x πgdxdydz � πgl3 P dxdy � P l2 �v µ dxdy � µU l �z θdx � θl. Weber number: W = πU 2 l . µ is absolute viscosity. bluﬀ body wakes. and ﬂow across a plate. it is imperative to maintain as many of the nondimensional groups as possible of the full-scale system. Given that [θ] = N/m (MKS). The Weber number indicates the importance of surface tension. Since Re appears in many applications. viscous eﬀects are negligible. Re can be used to characterize pipe ﬂow. µ/π 5 SIMILITUDE (87) where U is velocity. dz]. This holds for the geometry of the body and the kinematics . θ (88) where θ is the surface tension of a ﬂuid. πU 2 normalizes pressure. among others. To appreciate the origins of these terms from a ﬂuid particle’s point of view. consider a box having side lengths [dx. dy.

For surface vessels.5. and fur thermore are readily accessible to the user. First. ≈ the viscous eﬀects are negligible. As a result. however. denoted with a prime symbol: v∗ = ˙ v∗ r∗ ˙ L v ˙ U2 1 = v U L2 = r ˙ U2 (89) . few ﬂuids with this property are workable in a large testing tank. Then Reynolds number scaling im plies directly λmodel = λship /η3/2 . Unfortunately. it is a routine procedure to employ turbulence stimulators to achieve ﬂow that would normally occur with ship-scale Re. since l is large. F r. 5. ˙ ˙ ˙ ˙ These equations can be nondimensionalized in a standard way. ≈ Froude number similitude requires Umodel = Uship / η. θ. both of which are important for surface vessels. time. W ). and hence surface tension plays no role.e. g. W is very large. we write the functional relationship for drag as a starting point: Cr = D 1 πAU 2 2 = fQ (π. l) = fψ (Re.3 Similitude in Maneuvering The linear equations of motion for the horizontal yaw/sway problem are: (m − Yv )v − Yv v + (mU − Yr )r + (mxG − Yr )r = Y ˙ ˙ ˙ ˙ (Izz − Nr )r + (mxG U − Nr )r + (mxG − Nv )v − Nv v = N. by using the quantities [U. Next. we look ≈ at Re = U l/λ and F r = U/ gl. 000. π]: these three values provide the necessary units of length. µ. Under inviscid i.. we set gmodel = gship ..e. Consider the example of nozzle ﬂow from a bucket. With this achieved. Cf is not sensitive to Re. one then tries to match F r closely. the surface roughness. the model and the true vessel operate in the same gravity ﬁeld. and all of the relevant groups governing ﬂuid dynamics. and submarines near the surface. Above a critical value Re � 500. i. so that usually η >> 1. additionally. First. the ﬂow rate is Q = βd 2gh/4. This rate cannot be achieved for lower-Re conditions because of ﬂuid drag in the oriﬁce. 2 conditions. In a vessel. we create nondimensional states. and mass. Suppose that lship = ηlmodel . accurate scaling of Re for large vessels to model scale is quite diﬃcult.3 Similitude in Maneuvering 27 of the ﬂow. Suppose that we conduct a model test in which Re is abnormally large. U. L.

(m∗ − Yv∗ )v ∗ − Yv∗ v ∗ + (m∗ U ∗ − Yr∗ )r ∗ + (m∗ x∗G − Yr∗ )r ∗ = Y ∗ ˙ ˙ ˙ ˙ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ (Izz − Nr )r + (m xG U − Nr )r + (m xG − Nv )v − Nv v ∗ = N ∗ . and every moment with 2 πU 2 L3 . U 5 SIMILITUDE We follow a similar procedure for the constant terms as follows. for consistency with our previous expressions: m∗ = ∗ Izz = m 1 πL3 2 (90) x∗G = U∗ = Yv∗ = ˙ Yv∗ = Yr∗ = ˙ Yr∗ = Y∗ = ∗ Nv = ˙ ∗ Nv = ∗ Nr = ˙ ∗ Nr = Izz 1 πL5 2 xG L U =1 U Yv ˙ 1 πL3 2 Yv 1 πU L2 2 Yr ˙ 1 πL4 2 Yr 1 πU L3 2 Y 1 πU 2 L2 2 Nv ˙ 1 πL4 2 Nv 1 πU L3 2 Nr ˙ 1 πL5 2 Nr 1 πU L4 2 N . 1 πU 2 L3 2 N∗ = 1 1 Note that every force has been normalized with 2 πU 2 L2 . ˙ ˙ ˙ ˙ (91) . Thus we arrive at a completely equivalent set of nondimensional system equations. time has been also nondimensionalized with L/U . including a factor of 1/2 with π.28 r∗ = L r.

The nondimensional terms are ∗ Ixx = ∗ Kp = ˙ ∗ Kp = ∗ Kω = Ixx 1 πL5 2 Kp ˙ 1 πL5 2 Kp 1 πU L4 2 Kω 1 πU 2 L3 2 2 (93) p∗ = ˙ p∗ leading to the equivalent system L p ˙ U2 L = p. In the case of a submarine. since it appears explicitly in the nondimensional righting moment ∗ term. and GM is the metacentric height. (94) U2 Note that the roll angle δ was not nondimensionalized. ˙ ˙ (92) For a surface vessel. and has the form Kω = −πg∞(GM ). 5. since B and h are ﬁxed. U 2gL − − − ∞∗ (GM ∗ )ω = K ∗ .4 Roll Equation Similitude 29 Since ﬂuid forces and moments generally scale with U 2 . the roll moment Kω is based on metacentric stability. . and also has a strong inﬂuence on Kp .5. this Kω needs to be maintained in model tests. and therefore require special attention. the nondimensionalized description holds for a range of velocities. and h is the righting arm. where B is the buoyant force. where ∞ is the displaced ﬂuid volume of the vessel. the righting moment has the form Kω = −Bh. Let us carry out a similar normalization of the simpliﬁed roll equation (Ixx − Kp )p − Kp p − Kω ω = K. The nondimensional coeﬃcient becomes ∗ (Ixx ∗ Kp )p∗ ˙ ˙ ∗ Kp p∗ � � ∗ Kω = − 1 Bh . πU 2 L3 2 ∗ ∗ Kω again depends strongly on U . The Froude number has a very strong inﬂuence on roll stability.4 Roll Equation Similitude Certain nondimensional coeﬃcients may arise which depend explicitly on U .

The coeﬃcients Yv and Nv can also be obtained ˙ by running with a ﬁxed angle of attack. they should both be implemented in model testing. the vessel is ﬁxed on an arm of length R. In model tests. need not be matched as long as the scale model λ attains turbulent ﬂow (supercritical Re). rotating at constant rate r: the vessel forward speed is U = rR. which can each impose independent sway motions.2 Rotating Arm Device On a rotating arm device. The model moves in pure sway if ya (t) = yb (t). the Froude number F r = �U . the measurement is made over one rotation only. The connection points are a distance l forward and aft from the vessel origin. and the basic lifting surface models are not diﬃcult. The slope of the curve at zero angle determines Yv and Nv respectively. The available computational approaches should be considered as well. which scales the inﬂuence of surface waves. 6. Note that the lateral force also contains the component (m − Yv )r 2 R. gL must be maintained between model and full-scale surface vessels. modern sensors and computer control systems make possible the estimation of certain coeﬃcients based on open-water tests of a model or a full-scale design. 6. tow the vehicle at diﬀerent angles of attack. in a pure yaw motion about the mid-length point if ya (t) = −yb (t). measuring sway force and yaw moment. these are very good for predicting added mass in particular. or in a combination sway and yaw motion. which scales the eﬀect of viscosity.1 Towtank In a towtank. One can use turbulence stimulators near the bow if necessary. The idea is to measure the crossbody force and yaw moment as a function of r. (95) . Usually a sinusoidal motion is imposed: ya (t) = a cos �t yb (t) = b cos(�t + ω). For example. 6.3 Planar-Motion Mechanism With a planar motion mechanism. Rudder derivatives can be computed also by towing with various rudder angles.30 6 CAPTIVE MEASUREMENTS 6 CAPTIVE MEASUREMENTS Before making the decision to measure hydrodynamic derivatives. giving the coeﬃcients Yr and Nr . Finally. test results for many hull-forms have already been published. higherorder terms can be generated if the points deviate from a straight line. Since the control surface(s) and propeller(s) aﬀect the coeﬃcients. one forward and one aft. so that the vessel does not re-enter its own wake. Reynolds number Re = UL . therefore producing variable yaw. the vessel is towed at constant forward speed U . Finally. but is held by two posts. a preliminary search of the literature may turn up useful estimates.

˙ ˙ ˙ ˙ We have v = (ya + ya )/2 and r = (yb − ya )/2l. then (m − Yv )v − Yv v + (mU − Yr )r + (mxG − Yr )r = Ya + Yb ˙ ˙ ˙ ˙ (Izz − Nr )r + (mxG U − Nr )r + (mxG − Nv )v − Nv v = (Yb − Ya )l. we obtain four independent equations: a� 2 (m − Yv ) − (1 + cos ω) − ˙ 2 � � a� Yv − sin ω + 2 � � a� sin ω + (mU − Yr ) − 2l � � a� 2 (mxG − Yr ) − (cos ω − 1) = Fa cos χa + Fb cos χb ˙ 2l a� 2 (m − Yv ) − (− sin ω) − ˙ 2 � � a� Yv − (1 + cos ω) + 2 � � a� (mU − Yr ) − (cos ω − 1) + 2l � � a� 2 (mxG − Yr ) − (− sin ω) = −Fa sin χa − Fb sin χb ˙ 2l a� 2 (Izz − Nr ) − (cos ω − 1) + ˙ 2l � � � � � � (99) (100) (101) . 2l v = − 31 (96) (97) (98) Equating the sine terms and then the cosine terms. these become ˙ ˙ ˙ ˙ a� (sin �t(1 + cos ω) + cos �t sin ω) 2 a� 2 v = − ˙ (cos �t(1 + cos ω) − sin�t sin ω) 2 a� r = − (sin �t(cos ω − 1) + cos �t sin ω) 2l a� 2 r = − ˙ (cos �t(cos ω − 1) − sin �t sin ω) . When a = b.3 Planar-Motion Mechanism and the transverse forces on the posts are measured and approximated as Ya (t) = Fa cos(�t + χa ) Yb (t) = Fb cos(�t + χb ). If linearity holds.6.

ω. χb ]. Remarkably. we have ˙ ˙ a� 2 (mxG − Yr ) − (−2) ˙ 2l � � a� (mU − Yr ) − (−2) 2l � � a� 2 (Izz − Nr ) − (−2) ˙ 2l � � a� (mxG U − Nr ) − (−2) 2l � � = Fa cos χa + Fb cos χb = −Fa sin χa − Fb sin χb = l(Fb cos χb − Fa cos χa ) = l(−Fb sin χb + Fa sin χa ). Fb . Izz . a. Yv . χa . Nr . Nv . exactly what is required to ﬁnd the eight coeﬃcients [Yv .32 � � 6 CAPTIVE MEASUREMENTS a� (mxG U − Nr ) − sin ω + 2l � � a� 2 (mxG − Nv ) − (1 + cos ω) − ˙ 2 � � a� sin ω = l(Fb cos χb − Fa cos χa ) Nv − 2 a� 2 (Izz − Nr ) − (− sin ω) + ˙ 2l � � a� (mxG U − Nr ) − (cos ω − 1) + 2l � � a� 2 (mxG − Nv ) − (− sin ω) − ˙ 2 � � a� Nv − (1 + cos ω) = l(−Fb sin χb + Fa sin χa ) 2 � � (102) In this set of four equations. For ω = 180o . a� 2 (m − Yv ) − (2) ˙ 2 � � a� −Yv − (2) 2� � a� 2 (mxG − Nv ) − (2) ˙ 2 � � a� −Nv − (2) 2 � � = Fa cos χa + Fb cos χb = −Fa sin χa − Fb sin χb = l(Fb cos χb − Fa cos χa ) = l(−Fb sin χb + Fa sin χa ). we obtain [Fa . Yr . xG ]. Nv . (103) to be solved respectively for [Yv . It turns out that the two cases of ω = 0 (pure sway motion) and ω = 180o (pure yaw motion) yield a total of eight independent equations. (104) . we can write the eight solutions ˙ ˙ ˙ ˙ directly: For ω = 0. Nv ]. Yv . From the experiment. Nv . �]. and from the rigid-body model we have [m. we know from the imposed motion the values [U. Nr ]. Yr .

Proceed back up to 15→ . to -15→ . and keep it there until steady yaw rate is maintained for one minute. Repeat. Repeat in decrements of -5→ . the eight linear coeﬃcients for a surface vessel maneu ˙ ˙ vering. The PMM can be driven with more complex trajectories which will target speciﬁc nonlinear terms. 2. suppose that the ﬁrst 15→ rudder deﬂection causes the vessel to turn right. Turn rudder quickly by 15→ .33 to be solved for [Yr . The test reveals if the vessel has a memory eﬀect. The Dieudonn´ maneuver has the vessel path following a growing spiral. 4. It is easy to see that the rudder in this case has to be used excessively driving the vessel back and forth. can be deduced from two tests with a planar motion mechanism. 7 STANDARD MANEUVERING TESTS This section describes some of the typical maneuvering tests which are performed on full-scale vessels. No changes in speed setting are made after this point. We say that the vessel is unstable. manifested as a hysteresis in yaw rate r. With zero rudder. 5. 3. is still to the right.2 Zig-Zag Maneuver 1. and will require a negative rudder action to pull out of the turn. to assess stability and performance. But if the corrective action causes the vessel to turn left at all. Reduce rudder angle by 5→ . Achieve steady speed and direction for one minute. For example. Deﬂect the rudder to 20→ . but that the yaw rate at zero rudder. 3. and hold until the vessel turns to -20≤ with respect to the starting heading. Yr . The vessel has gotten “stuck” here. achieve steady speed for one minute. 7. for a given speed. and clearly a poor design. 2. on the way negative. and hold until the vessel turns 20→ . Thus. 7. 4. and keep it there until steady yaw rate is maintained for one minute. the same memory eﬀect may occur. and that some curve ﬁtting will be needed in any event. . Deﬂect the rudder to -20→ . Nr . and then a con e tracting spiral in the opposite direction. Nr ].1 Dieudonn´ Spiral e 1. We note that the nonlinear terms will play a signiﬁcant role if the motions are too large.

the yaw overshoot (amount the vessel exceeds ±20→ when the rudder has turned the other way). In a surface vessel. and the total period for the 20→ oscillations. . (106) with Yr small. but the center of gravity is near the waterline or above. v < 0 when r > 0 for most vessels. the ﬂuid forces act below the waterline. yaw.3 Circle Maneuver From a steady speed. Of course. the speed loss. and the angle of heel ω. After steady state is reached again. ˙ ˙ ˙ ˙ (105) The coeﬃcients for r are quite small.3. When the rudder is ﬁrst deﬂected. 7. The vessel responds by turning in a circle. When steady turning conditions are reached (Phase 3). If the turning circle has radius R (measured from the vessel origin). 7. similar tests can be made with diﬀerent rudder angles and diﬀerent threshold vessel headings. At this location.34 7 STANDARD MANEUVERING TESTS This maneuver establishes several important characteristics of the yaw response. zero yaw rate condition. the drift angle α. hydrodynamic forces equalize the centrifugal force mU r and the rudder force Yν ζ.2 Speed Loss Speed loss occurs primarily because of drag induced by the drift angle. and roll caused by the center of gravity. the vessel ultimately rolls to port (negative) for a positive rudder action. there is no apparent lateral velocity. the rudder is moved to a new setting.1 Drift Angle The drift angle is the equivalent to angle of attack for lifting surfaces.3. and describes how the vessel “skids” during a turn. inertial terms dominate (Phase 1) and the sway equation is (m − Yv )v − (Yr − mxG )r = Yν ζ. A vessel which drifts very little may have very little speed loss. The vessel-reference velocity components are thus u = U cos α and v = −U sin α. parameters of interest are the turning diameter. These are: the response time (time to reach a given heading). Because the centrifugal force acts above the waterline. 7.3. and thus the vessel ﬁrst rolls to starboard (positive) ˙ for a positive rudder action. A line along the vessel centerline reaches closest to the true center of the turning circle at a point termed the turning center. and it appears to an observer there that the vessel turns about this point. The sway equation is −Yv v + (mU − Yr )r = Yν ζ. 7. then the speed tangential to the circle is U = rR. and |Yv v | > |Yν ζ |. which may or may not exist on the physical vessel.3 Heel Angle Heel during turning occurs as a result of the intrinsic coupling of sway.

and hence area Ao is usually that of the wetted surface. the sign of FA is negative. submarines have the mass center below the rudder action point. we see that if the rudder is straightened out at this point. In the steady state. but not necessarily a zero moment. d’Alembert’s paradox predicts zero net force. also. the keel out of the turn. which has the form 1 (109) FA = − πCA Ao U 2 . and Yν ζ (intermediate). The inertial equation (m − Yv )v − (Yr − mxG )r = Yν ζ ˙ ˙ ˙ ˙ (107) is dominated by mv (acting low). the asymmetric location of the stagnation points. toward the center of the turn. which have the rigid mass center above the center of ﬂuid forcing. From the sway equation.2 Munk Moment Any shape other than a sphere generates a moment when inclined in an inviscid ﬂow. ˙ −Yv v + (mU − Yr )r = Yν ζ.1 STREAMLINED BODIES Nominal Drag Force A symmetric streamlined body at zero angle of attack experiences only a drag force. (108) The drift angle α keeps the Yv -force. and again centrifugal force mU r causes the bottom of the submarine to move out of the turn. .4 Heeling in Submarines with Sails Submarines typically roll into a turn during all phases. the roll angle of a submarine with a sail is always into the turn. Hence. 2 The drag coeﬃcient CA has both pressure and skin friction components. because it opposes the vehicle’s x-axis. but then out of the turn in the steady state. Unlike surface vessels. Because ˙ ˙˙ |Yv | > m.3. −Yv v (acting high). acting high. 7. 8 8. both initially and in the steady state. the roll will momentarily be even worse! In summary. Note that the A-subscript will be used to denote zero angle of attack conditions. the vessel overshoots the ﬁnal port roll angle as the vessel slows. 8. This Munk moment arises for a simple reason. The heel angle declines as the speed of the submarine drops.35 The transition between the inertially-dominated and steady-turning regimes includes an overshoot: in the above formulas. the vessel rolls under the sail. the vessel rolls into the turn initially. and additionally feel the eﬀects of a large sail above both.

Let ∂ represent the angle of attack. if it is moving horizontally.3 Separation Moment In a viscous ﬂuid. The formation of the vortex depends on the angle of attack.e. taken to be positive with the nose up – this equates to a negative pitch angle δ in vehicle coordinates. We will limit the present discussion to the vertical plane. The zero-∂ drag force FA is modiﬁed by the vortex shedding: 1 Fa = − πCa Ao U 2 . a decomposition using apparent velocity. For a small angle of attack. or from slender body approximation (to follow).36 8 STREAMLINED BODIES where pressure is highest on the front of the body (decelerating ﬂow) and lowest on the back (accelerating ﬂow). 2 �∂ Fn = (111) With a positive angle of attack. in the sense that it acts to turn the vehicle perpendicular to the ﬂow. a helical vortex may form and convect downstream. The added mass terms Azz and Axx can be estimated from analytical expressions (available only for regular shapes such as ellipsoids). ﬂow over a streamlined body is similar to that of potential ﬂow. i. and Azz along the vehicle’s z-axis z (up). from numerical calculation. Since vortices correlate with low pressure. this force is in the positive z-direction. In this latter area. with the exceptions of the boundary layer. The Munk moment is: 1 Mm = − (Azz − Axx )U 2 sin 2∂ 2 � −(Azz − Axx )U 2 ∂. the transverse force Fn can be written in the same form as for control surfaces: 1 πCn Ao U 2 2 1 �Cn � π ∂Ao U 2 . but it also induces drag. and it may cover a larger area (increasing the stabilizing moment and drag) for a larger angle of attack. the eﬀect of such a vortex is stabilizing. but similar arguments can be used to describe the horizontal plane.. Consider a symmetric body with added mass components Axx along the vehicle (slender) xaxis (forward). 8. The Munk moment is always destabilizing. (112) The last relation is based on writing CA (U cos δ)2 as (CA cos2 δ)U 2 . (110) Azz > Axx for a slender body. and the negative sign indicates a negative pitch with respect to the vehicle’s pitch axis. where 2 Ca = CA cos2 δ. . and a small region near the trailing end.

If the origin lies behind AC. for towed vehicles. Each constant ρ is taken as positive.4 Net Eﬀects: Aerodynamic Center 37 8. and in this case. Hence. the net moment is stabilizing. First we simplify: Fa = −ρa Fn = ρ n ∂ Mm = −ρm ∂. 8. Fixed surfaces induce lift and drag on the body: 1 L = πAf U 2 Cl (∂) � ρL ∂ 2 1 D = − πAf U 2 Cd � −ρD (constant) 2 (115) . The net moment about O is zero if we select ln = ρm . (113) where ln denotes the (positive) distance between O and the application point of Fn . The Munk moment is a pure couple which does not depend on a reference point. the towpoint must be located forward of AC. with a nose-up angle of attack. As a ﬁnal note. AC moves inﬁnitely far forward as viscosity eﬀects diminish. the aerodynamic center is signiﬁcantly ahead of the nose. the moment is destabilizing. For self-propelled vehicles. ρn (114) and the location of O is then called the aerodynamic center or AC. and their mag nitudes determine the stability of a hullform. if the vehicle’s origin lies in front of AC. In many cases with very streamlined bodies. the mass center must be forward of AC for stability. The point AC has an intuitive explanation: it is the location on the hull where Fn would act to create the total moment. and the signs reﬂect orientation in the vehicle reference frame.5 Role of Fins in Moving the Aerodynamic Center Control surfaces or ﬁxed ﬁns are often attached to the stern of a slender vehicle to enhance directional stability. and write the total pitch moment about O as: M = M m + F n ln = (−ρm + ρn ln )∂. a rigid sting would have to extend at least to AC in order for stable towing.8.4 Net Eﬀects: Aerodynamic Center The Munk moment and the moment induced by separation are competing. since the Munk moment persists even in inviscid ﬂow. Similarly. We pick a temporary origin O for Fn however.

this latter term is large. so that lf might be very small.e. (118) (117) The Munk moment ρm opposes the aggregate eﬀects of vorticity lift ρn and the ﬁns’ lift and drag ρL + ρD . there is no net gain in stability. referenced to an arbitrary body point O. If we say that the ﬁns act a distance lf behind the temporary origin O. Equation 118 contains two length measurements. (116) All of the forces are pushing the vehicle up. it can be recalled that ln = ρm . Setting Fa and Fn to account for the fuselage and ﬁns. since both eﬀects usually act near the stern.. With very large ﬁns. ρn Hence. The summed forces on the body are thus: X = � Z = � Fa − |D | cos ∂ + |L| sin ∂ −ρa − ρD + ρL ∂2 Fn + |L| cos ∂ + |D | sin ∂ ρn ∂ + ρL ∂ + ρD ∂. it is possible to group ˆ ˆ them into a condensed form. We solve for lf : lf = ρ m + ρ n lf n . and thus AC can be referenced to any other ﬁxed point easily. The moment about O vanishes if ρm = ρn ln + lf (ρL + ρD ). the ﬁns act directly in the denominator to shorten lf . let lf n denote the (positive) distance that the ﬁns are located behind Fn . this is the case of AC moving aft toward the ﬁns. 8. A vehicle with excessively large ﬁns will be diﬃcult to turn and maneuver. lf n < 0. Note that if the ﬁns are located forward of the vortex shedding force Fn . this is likely a small number. Without ﬁns. with ρ > 0 and ∂ > 0. and are signed again to match the vehicle frame. lf is reduced. ρn + ρ L + ρ D (119) This is the distance that AC is located forward of the ﬁns. we have . To solve it explicitly.38 8 STREAMLINED BODIES These forces act somewhere on the ﬁn. and that the moment carried by the ﬁns themselves is very small (compared to the moment induced by Llf ) the total moment is as follows: M = (−ρm + ρn ln )∂ + (ρL + ρD )lf ∂. but since AC is referenced to the ﬁns.6 Aggregate Eﬀects of Body and Fins Since all of the terms discussed so far have the same dependence on ∂. i.

we consider the contribution of the ﬁns only – slender body theory. with an instantaneous pitch angle of χ. 2 (120) 8. provides good results for the hull. The angle of attack seen by the ﬁn is a combination of a part due to χ and a part linear with q: lf q U and so lateral force and moment derivatives (for the ﬁn alone) emerge as ∂f � −χ + 1 Zq = − πCl∗ Af Ulf 2 1 ∗ 2 Mq = − πCl Af Ulf . The ﬁn center of pressure is located a distance lf aft of the body origin.7 Coeﬃcients Zw .7 Coeﬃcients Zw .1 SLENDER-BODY THEORY Introduction Consider a slender body with d << L. and Mq for a Slender Body 39 1 ˆ ˆ ˆ X = Fa � − πCa Ao U 2 2 1 ˆ∗ ˆ 2 ˆ Z = Fn � π C n A o U ∂ 2 1 ˆ∗ ˆ ˆ M = −Fn xAC � − Cn Ao U 2 xAC ∂. Mw . Mw . 2 (123) w u � (121) (124) 9 9. but we require that the lateral variations are small and . and we assume that the vehicle is moving horizontally. as well as their derivatives.8. 2 n The rotation of the vessel involves complex ﬂow. The body could be asymmetric in cross-section. which depends on both w and q. To start. that is mostly straight. Zq . U We can then write several linear hydrodynamic coeﬃcients easily: 1 ˆ∗ ˆ (122) Zw = − πCn Ao U 2 1 ˆ∗ ˆ Mw = πC Ao UxAC . Zq . or even ﬂexible. introduced shortly. and Mq for a Slender Body The angle of attack ∂ is related to the cross-body velocity w as follows: ∂ = − tan−1 � − � w for U >> w.

. the body is ﬂexible. ≥ and we can write wn = �zb �zb . we will consider the two-dimensional heave/surge problem only. not the vehicle. t). this usage has an indirect connection with the derivative described in our previous discussion of rigid-body dynamics. (126) �t The ﬁrst component is the time derivative in the body frame. can be found in the books by Newman and Blevins. 4 The added mass is equal to the mass of the water displaced by the cylinder. t). The eﬀects are integrated along the length to approximate lift force and moment. The time-dependence indicates that the conﬁguration is free to change with time. The local geometry is described by the location of the centerline. and the second due to the deﬂection of the particle by the inclined body. 9. if the diameter of a body of revolution is d(s). i. −U �t �x wn � The ﬂuid derivative operator in action is as follows: dzb wn = = dt � � � −U zb . We use the notion of a ﬂuid derivative for this purpose: the operator d/dt indicates a derivative taken in the frame of the passing particle. The velocity of a ﬂuid particle normal to the body at x is wn (t. is to think of the body as a longitudinal stack of thin sections. includ ing ﬂat plates and ellipses. where the nominal added mass value for a cylinder is β (125) ζma = π d2 ζx. Many formulas for added mass of two-dimensional sections. then �zb /�t will contain w. sin ∂ � tan ∂ = �zb /�x. If the body reference frame is rotated to the ﬂow. The equation above turns out to be a good approximation for a number of two-dimensional shapes. if d is taken as the width dimension presented to the ﬂow. Note that the curvilinear coordinate s is nearly equal to the body-reference (linear) coordinate x. For the purposes of explaining the theory. as well as for simple three-dimensional bodies. As one example. For small angles.40 9 SLENDER-BODY THEORY smooth along the length.e.2 Kinematics Following the Fluid The added mass forces and moments derive from accelerations that ﬂuid particles experience when they encounter the body. The idea of the slender-body theory. that is. Slender-body theory is accurate for small ratios d/L. then we can compute ζma (x). under these assumptions. if w = 0. x): �zb cos ∂ − U sin ∂. �t �x � . and local angle ∂(x. each having an easily- computed added mass. Hence. except near the ends of the body. it has vertical location (in body coordinates) of zb (x.

t)] = lim [µ(x + ζx. Let µ(x.t. t + ζt): �µ �µ ζt + ζx + h.o. then the diﬀerential inertial force on the body here is the derivative (following the ﬂuid) of the momentum: ζF = − d [ma (x. t) + and noting that ζx = −U ζt. 9.9. t)] νt�0 ζt dt � ⎬ �µ �µ . dt (127) Note that we could here let the added mass vary with time also – this is the case of a changing cross-section! The lateral velocity of the point zb (x) in the body-reference frame is �zb = w − xq. ζx �t �x We now restrict ourselves to a rigid body. t)] ζx. t + ζt) = µ(x. t)wn (x. t)(w(t) − xq(t) − U∂(x. t).3 Derivative Following the Fluid A more formal derivation for the ﬂuid derivative operator is quite simple. �t �x µ(x + ζx. we have ζF � � −U =− [ma (x. �t such that wn = w − xq − U∂. t) represent some property of a ﬂuid particle. d 1 [µ(x. −U = �t �x The second equality can be veriﬁed using a Taylor series expansion of µ(x + ζx. t))] .4 Diﬀerential Force on the Body If the local transverse velocity is wn (x. The ﬂuid is convected downstream with respect to the body.. so that neither ma nor ∂ may change with time. � ζF = ma (x)(−w + xq + U ˙ ˙) [ma (x)(w − xq − U∂)] . ζx �x (130) � � (128) (129) . t + ζt) − µ(x.3 Derivative Following the Fluid 41 9. Taking the derivative.

∂ could take any value(s) along the body. . computed with strip theory is � xN xT Z = ζF dx (131) where xT represents the coordinate of the tail. however. It is interesting to note that both Zw and Z∂(xT ) depend on a nonzero base area. Ex panding. the strip theory thus provides (132) Zw ˙ Zq˙ Zw Zq Z∂(xT ) = = = = = −m33 −m35 −Uma (xT ) UxT ma (xT ) U 2 ma (xT ). since their diﬀerence depends only on how the body coordinate system is oriented to the ﬂow. Another noteworthy fact is that the lift force depends only on ∂ at the tail. Additionally. ˙ ˙ x=xT xN Z = � ma (x) [−w + xq dx + U ˙ ˙] � xN We made use here of the added mass deﬁnitions � xN xT m33 = ma (x)dx xN m35 = − � xT xma (x)dx. In general. The two terms are clearly related. and xN is the coordinate of the nose. ˙ ˙ In terms of the linear hydrodynamic derivatives. with no eﬀect on Z. the added mass ma at the nose is zero. so this should come as no surprise. so that a simpler form occurs: Z = −m33 w − m35 q − Uma (xT )(w − xT q − U∂(xT )). for vessels with pointed noses and ﬂat tails.42 9 SLENDER-BODY THEORY 9. we have � [ma (x)(w − xq − U∂)] dx xT xT �x = −m33 w − m35 q + Uma (x)(w − xq − U∂)|x=xN . potential ﬂow estimates do not create lift (or drag) forces for a smooth body.5 Total Force on a Vessel The net lift force on the body.

the linear hydrodynamic moment derivatives are Mw ˙ Mq˙ Mw Mq M∂ = = = = = −m35 −m55 UxT ma (xT ) + Um33 −Ux2 ma (xT ) + Um35 T 2 −U xT ma (xT ) − U 2 m33 . The integral above contains the product ma (x)∂(x). As with the lift force. For simplicity. The derivative Mw is closely-related to the Munk moment. depending only on the orientation of the body frame to the ﬂow. The Munk moment (an exact result) may therefore be used to make a correction to the second term in the slender-body approximation above of Mw . we now assume that ∂ is in fact constant on the length. Finally.6 Total Moment on a Vessel 43 9. . which must be calculated if ∂ changes along the length.6 Total Moment on a Vessel A similar procedure can be applied to the moment predictions from slender body theory (again for small ∂): M = − = = � � xN xT xζF dx � � xN xT xN xT � � � −U x [ma (x)(w − xq − U∂)] dx �t �x � xN � ˙ ˙)dx − U x [ma (x)(w − xq + U∂)] dx. whose linearization would provide Mw = (m33 − m11 )U . leading to M = −m35 w − m55 q + UxT ma (xT )(w − xT q − U∂) + ˙ ˙ U m33 w + Um35 q − U 2 m33 ∂. xma (x)(w − xq �x xT Then we make the further deﬁnition m55 = � xN xT x2 ma (x)dx. (note that m35 = m53 ) and use integration by parts to obtain M = −m35 w − m55 q − Uxma (x)(w − xq − U∂)|x=xN + ˙ ˙ x=xT U � xN xT ma (x)(w − xq − U∂)dx. Mw and M∂ are closely related.9.

Inserting this approximation into the lift formula. which itself is stabilizing. and can be accurately modeled with slender body theory. . The lift coeﬃcient slope is approximated by (Hoerner) 1 (133) Cl∗ ⇐ β(AR). and so on. and using the slender-body estimate for Zw . Where does the slender-body predict the force will act? Recalling that Mw = Um33 + UxT ma (xT ). By listing the body-referenced velocities in the order γ = [u. The blunt trailing end induces lift as a product of added mass eﬀects. v. and are convected continuously into the wake.8 Convention: Hydrodynamic Mass Matrix A Hydrodynamic derivatives that depend on accelerations are often written as components of a mass matrix A. The lift force from the latter is 1 of the form L = − 2 πU ACl∗ w . where M is the mass matrix of the material vessel and F is a γ we write (M + A)γ s generalized force. This location will tend to stabilize the wing. the blunt trailing end. we obtain β L = − πs2 Uw. The helical body vortices are stable and symmetric in this condition. ˙ ˙ 10 10. (134) 4 Now we look at a slender body approximation of the same force: The added mass at the tail is ma (xT ) = πs2 β/4. 9. usually toward the stern of the vehicle. s ˙ = F . w.7 Relation to Wing Lift There is an important connection between the slender body theory terms involving added mass at the tail (ma (xT )).44 10 PRACTICAL LIFT CALCULATIONS 9. The ﬁns can often be properly modeled using experimental data parametrized with aspect ratio and several other geometric quantities.1 PRACTICAL LIFT CALCULATIONS Characteristics of Lift-Producing Mechanisms At a small angle of attack. the product of chord (long) and span (short). where A = cs. The low pressure associated with the vortices provides the suction force. A blunt trailing edge also induces some drag. q. Therefore A33 = −Zw . A5. a slender body experiences transverse force due to: helical body vortices. and low aspect-ratio wing theory.3 = −Mw . in the sense that it acts to orient the wing parallel to the incoming ﬂow. r]. 4 Slender-body theory is thus able to recover exactly the lift of a low-aspect ratio wing. and since m33 = 0 for a front-back symmetric wing. and ﬁns. we calculate for lift: Z = −ma (xT )Uw β = − πs2 Uw. p. the estimated lift force acts at the trailing edge. 2 where (AR) is the aspect ratio.

vorticity sheds as if from a bluﬀ-body. and greatly increases drag by widening the wake. axial force. In the limit of a 90→ angle of attack. This area could be the frontal projected area. The helical vortices can become bigger while remaining stable. Ab = 0 for a body that tapers to a point at the stern. this shedding is nonsymmetric. and reference area Ar . Jorgensen gives the coeﬃcients as follows: ∂ Ap Ab sin 2∂ cos + Cd sin2 ∂ Ar 2 Ar n = CAo cos2 ∂ � � ∂ ∞ − Ab (L − xm ) A p xm − x c sin2 ∂. measured back from the nose. and the rest peels away from the body. Some of it convects downstream. The normal force. The body travels at speed U . and there is little axial convection. The parameters used here are • Ab : stern base area. . • Cdn : crossﬂow drag coeﬃcient. Let the body have length L. and angle of attack ∂. The formulas provide a good systematic procedure for design. although the complete equations do account for more complex shapes. and we now present one of them due to Jorgensen. 1 πU 2 Ar L 2 CN = CA = CM = (135) The moment Mxm is taken about a point xm . or the wetted area. If “normal” Reynolds number Ren = U sin ∂D/λ. = − sin 2∂ cos − Cdn Ar L 2 Ar L CN = CA CM (136) (137) (138) We have listed only the formulas for the special case of circular cross-section. 10. and moment coeﬃcients are deﬁned as follows: FN 1 πU 2 Ar 2 FA 1 πU 2 Ar 2 M xm . We call the area of the stern the base area. and are best suited to vehicles with a blunt trailing edge. equivalent to that of an inﬁnite circular cylinder. and appears explicitly in the formula for Cm . which is also not a constraint in the complete equations. the planform area. Further.10.2 Jorgensen’s Formulas 45 As the angle of attack becomes larger. but eventually will split randomly. this location is arbitrary. the approximations in the ﬁn and slender-body anal ysis will break down.2 Jorgensen’s Formulas There are some heuristic and theoretical formulas for predicting transverse force and moment on a body at various angles of attack. we have assumed that L >> D.

xAC = xm + 10. Two moment coeﬃcients are: . The idea here is that the fore and aft components of crossﬂow drag cancel out.. i. Hoerner. There are several terms that match exactly the slender-body theory approximations for small ∂. Similarly. dynamic pressure is given as q = 2 πU 2 . • Ap : planform area. if the moment is referenced to the center of the planform area. Ren < 3 × 105 10 PRACTICAL LIFT CALCULATIONS – Cd ⇐ 0. Thus. and are described here.6. 1 First. • ∞: body volume. • xc : distance from the nose backwards to the center of the planform area. the moment seeks to move the AC forward on the body. while the second uses the square frontal area. while CN is always positive.e. CY = CYd D/L. wings. It con tains a large amount of aerodynamic data from many diﬀerent types of vehicles. modiﬁed by cos2 ∂. whether or not Ab = 0.46 – Cd ⇐ 1. 7 × 105 < Ren . A few notations are used throughout the books. based on wetted surface area. but the eﬀect is moderated by the lift force. such that two typical body lift coeﬃcients are: CY = C Yd Y DLq Y = 2 . the axial force is simply the zero-∂ result. (139) CN As written. The aerodynamic center (again referenced toward the stern. and other common engineering shapes. and the entire ﬁrst term in the moment (CM ). 3 × 105 < Ren < 7 × 105 – Cd ⇐ 0.3 Hoerner’s Data: Notation An excellent reference for experimental data is the two-volume set by S. D q The ﬁrst version uses the rectangular planform area as a reference. • CAo : axial drag at zero angle of attack. through a sin2 ∂-dependence. the moment coeﬃcient is negative if the moment destabilizes the body. we note that the second term in CM disappears if xm = xc .006 for slender streamlined bodies.2. In both cases. Finally. CAo � 0. These are the ﬁrst term in the normal force (CN ). we see that if Ab = 0. In the formula for normal force. only drag forces act to create lift. scaling of U 2 into the body principle directions is all that is required.002−0.3. from the nose) can be found after the coeﬃcients are computed: CM L. Hence. both frictional and form. It depends on Re = U L/λ.

4 Slender-Body Theory vs.027/deg for a long cylinder and −0. Because this coeﬃcient scales roughly with wetted area. The point at which the viscous forces act can then be estimated as the following distance aft of the nose: . the net moment is measured. 10. Experiment 47 CN = C N1 N LD 2 q N1 = . which achieves an equivalent lift of β(AR)/2 = 0. this acts at a point on the latter half of the vehicle. Note that the negative sign here is consistent with Hoerner’s convention that destabilizing moments have negative sign. This section gives the formula for this location in Hoerner’s notation.009 for L/D = 4. acting to destabilize the vehicle. and gives two further examples of how well the slender-body theory matches experiments.003/deg. The following relation holds for these deﬁnitions 1 C N1 = C N + C Y d . ˜ For L/D > 6. 2 The lift and moment coeﬃcients are strongly dependent on angle of attack. and the reference length is body length L. Note that the reference area for moment is the square frontal projection. comprising both the destabilizing part due to the potential ﬂow. stabilizing the angle. Experiment In an experiment.015/deg. it also reduces to −0.0027 for (AR) = 10 � D/L.018/deg for a long ellipsoid. the slender-body (pure added mass) estimates give Cnb � −0. The value compares well with −0. It follows from above that Cn·b = Cnb + Cydb /2. and N1 is taken about the nose. and the stabilizing part due to vortex shedding or a blunt tail.4 Slender-Body Theory vs. �b and so on.10. where b is the angle of attack. it changes little with L/D. Hoerner uses the notation Cnb = Cn·b Cyb �CN �b �CN1 = �b �Cy = . usually in degrees. proportional to LD. It can be compared with a low-aspect ratio wing. The experimental lift force is typically given by CY b � 0. Comparison of the measurements and the theory allows us to place the action point of the suction force. LD 2 q where N is the moment taken about the body mid-point.

both vehicles are equally unstable. 10. L/D ˜ Cnb Cyb Cydb Cn·b Cnb x/L symmetric 6.0044/deg. due solely to the blunt end condition.7 -0. In the table following are values from Hoerner (p.0037 0. Consider a body. The slender-body lift force associated with this end is simply the product of speed U and local added mass ma (xT ) (in our previous notation). Expressed in terms of the Hoerner reference area D 2 . where Af is the total area of the ﬁn pair.025 0.012 0. The lift force dependence on the blunt tail is not diﬃcult to see.48 10 PRACTICAL LIFT CALCULATIONS ˜ x Cn·b − Cnb = (140) L Cydb The calculation uses experimental values of Cydb and Cn·b . With respect to the area βr 2 .60 In comparing the two body shapes. but it acts too close to the midpoint to add any stability there. 13-2. At the body midpoint. we see that the moment at the nose is much more stable (positive) for the body with a blunt trailing edge. (141) 2 such that the ﬁrst term in parentheses is an eﬀective area.0031 0.4 for the data in the table. This is the case if the vehicle body tapers to a point where the ﬁns have their trailing edge. The blunt-tailed geometry has a much larger lift force. acting at the tail end of the vehicle. The slender-body approximation of lift as a result of blunt-end conditions is Z = βs2 πU 2 ∂. and the second is a lift coeﬃcient. where we made the assumption here that 2r/D = 0. It comes out to be 1 Z = πU 2 (βr 2 )(2∂).0031 (stable) -0. is consistent with measurements.0093 (unstable) 0.63 blunt 6. the moment slope referenced to the nose.012 0.5 Slender-Body Approximation for Fin Lift Let us now consider two ﬁns of span s each. however. substitution will give a lift curve slope of Cl∗ = β (AR). 2 . Figure 2) for a symmetric and a blunt-ended body. using slender-body theory. This lift diﬀerence. with trailing edge radius r.7 -0. the equivalent lift coeﬃcient is Cydb = 0. (142) Letting the aspect ratio be (AR) = (2s)2 /Af . the lift curve slope is therefore 2/rad.021 0.0094 (unstable) 0.0012 (stable) -0.

Another description for the wing-tip vortices is that the pressure diﬀerence across the surface simply causes ﬂow around the end.2 Three-Dimensional Eﬀects: Finite Length Since all practical lifting surfaces have ﬁnite length. which requires that ﬂuid particle streamlines do not wrap around the trailing edge of the surface. Lift is always deﬁned to act in a direction perpendicular to the ﬂow. 11 FINS AND LIFTING SURFACES Vessels traveling at signiﬁcant speed typically use rudders. and is quite accurate for (AR) � 1. Since bound circulation cannot end abruptly at the wing end. lift arises from circulation �: �= � γ s V · dγ. overestimating the lift by about 30% when (AR) = 2. and other streamlined control surfaces to maneuver. The theory of lifting surfaces centers on the Kutta condition. A critical parameter which governs the extent of three-dimensional eﬀects is the aspect ratio: span2 span . and a lifting surface requires circulation in order to meet the Kutta condition. (143) and then L = −πU �. This continuation causes induced velocities at the tips. This fact is true for a non-stalled surface at any angle of attack. More formally. leading to so-called wing-tip vortices. and some induced drag. with little drag penalty. (144) = chord area The second representation is useful for non-rectangular control surfaces. If the wing is attached AR = . sharp trailing edge. and worsening further as (AR) grows. it is clear that the ﬂuid must travel faster over one side of the surface than the other. Since the separation point on the front of the section rotates with the angle of attack. 11. The reduced Bernoulli pressure this induces can then be thought of as the lift-producing mechanism. Their utility arises mainly from the high lift forces they can develop. and drag in the same direction as the ﬂow. It is inadequate for higher-aspect ratio wings however. Circulation is the integral of velocity around the cross-section. and a smooth surface. it continues on in the ﬂuid.49 This is known as Jones’ formula.1 Origin of Lift A lifting surface is nominally an extrusion of a streamlined cross-section: the cross-section has a rounded leading edge. elevators. but instead rejoin with streamlines from the other side of the wing at the trailing edge. Prandtl’s inviscid theory provides some insight. 11. The eﬀective span is taken to be the length between the free ends of a symmetric wing. the ﬂow near the ends may be threedimensional.

and has one empirical description Cl∗ = 1 1 2ψ∂ ¯ + 1 ψ(AR) + 1 2ψ(AR)2 . since they are omnidirectional. (146) where ∂ is in radians. but also stalls earlier. Lift is written as 1 L = πU 2 ACl . 1 0. In many marine vessels and vehicles. xA � c/4. The eﬀective aspect ratio for a ring of diameter d is given as AR = The eﬀective area of the ring is taken as Ae = β dc. The aspect ratio is a strong determinant of wing performance: for a given angle of attack. and AR is the eﬀective aspect ratio.1 PROPELLERS AND PROPULSION Introduction We discuss in this section the nature of steady and unsteady propulsion. For high AR. this fact creates both a net force and a net moment. the eﬀective span is twice the physical value. the ¯ theoretical and maximum value for Cl∗ is 2β.63 + ψ(AR) (149) 12 12. βc (147) and we thus have L = 1 πU 2 Ae Cl∗ ∂. For low AR. (145) 2 where A is the single-side area of the surface. For angles of attack ∂ below stall. The lift generated on a surface is the result of a distributed pressure ﬁeld. where one formula for C ∗ is l 2 Cl∗ = 1 . say) or an electric motor drives the . A single equivalent force acts at a so-called center of action xA . 11. 2 (148) 4d . a larger aspect ratio achieves a higher lift value. and structurally more robust than cantilevered plane surfaces.3 Ring Fins Ring ﬁns are useful when space allows. xA � c/2. which depends mainly on the aspect ratio. an engine (diesel or gas turbine. the lift coeﬃcient Cl is nearly linear with ∂: Cl = Cl∗ ∂. by reﬂection.90. where Cl∗ is called the lift coeﬃcient slope. When AR ∀ ∗. ∂ � 0. measured back from the front of the wing. and in this case the eﬀective aspect ratio is therefore twice the physical value.50 12 PROPELLERS AND PROPULSION to a wall.

because of similitude (using diameter D in place of L when the propeller is involved). reducers.e. such as would be caused by a towing bar or hawser.. the ﬂow around the hull interacts with the ﬂow through the propeller. when the hull is towed with no propeller we use a t subscript. not behind a hull. we list several low-order models of thrusters. there are no additional forces or moments on the vessel. steady propulsion conditions are generally optimized for fuel eﬃciency. when the propeller is run in open water. When subscripts are not used. commercial surface vessels spend the vast majority of their time operating in open-water and at constant speed. In the second section. We use an sp subscript to indicate speciﬁcally self-propulsion conditions.2 Steady Propulsion of Vessels The notation we will use is as given in Table 1. An approximation of the transient behavior of a system can be made using the quasi-static assumption. we do not distinguish between the magnitude of forces in model and full-scale vessels. Self-propelled conditions refer to the propeller being installed and its propelling the vessel. In this case. and there are two diﬀerent ﬂow conditions to consider. i. Large.2 Steady Propulsion of Vessels 51 propeller through a linkage of shafts. Conversely. and the eﬀects of each part are important in the response of the net system. generalization to either condition is implied. Finally. we use an o subscript. Furthermore. which have recently been used to model and simulate truly unsteady conditions. and bearings. 12.12. Rsp Rt T ne nm np η Qe Qp �g Pe Pp D U Up Qm f fm N N N Hz Hz Hz Nm Nm W W m m/s m/s Nm kg/s kg/s hull resistance under self-propulsion towed hull resistance (no propeller attached) thrust of the propeller rotational speed of the engine maximum value of ne rotational speed of the propeller gear ratio engine torque propeller torque gearbox eﬃciency engine power propeller shaft power propeller diameter vessel speed water speed seen at the propeller maximum engine torque fuel rate (or energy rate in electric motor) maximum value of f Table 1: Nomenclature .

indicates that the incoming velocity seen by the propeller is only 90% of the vessel’s speed. In practical terms. ρ2 ] are usually positive.1. the prop creates some extra thrust. (150) The gear ratio η is usually large. ρ1 . indicating that the propeller turns much more slowly than the driving engine or motor. and therefore ﬁt the approximate form: KT (J ) = α1 − α2 J KQ (J ) = ρ1 − ρ2 J. Thrust and torque co eﬃcients are typically nearly linear over a range of J .1 Basic Characteristics 12 PROPELLERS AND PROPULSION In the steady state. (151) and power follows as Pp = �g Pe . the four coeﬃcients [α1 . . Behind a vessel moving at speed U . Up may not be the same as the speed of the vessel U .2. πnp D KT = The open-water propeller eﬃciency can be written then as �o = To U J (U )KT = . as shown in the ﬁgure. note that in the wake of the vessel. the wake fraction comes about this way: Suppose the open water thrust of a propeller is known at a given U and np . α2 . A propeller operating in open water can be characterized by two nondimensional parameters which are both functions of J : To (thrust coeﬃcient) πn2 D 4 p Qp KQ = 2 o 5 (torque coeﬃcient). w is referred to as the wake fraction. (155) As written. 2βnp Qpo 2βKQ (154) (152) (153) This eﬃciency divides the useful thrust power by the shaft power. for example. The following relations deﬁne the gearbox: ne = ηnp Qp = �g ηQe . We call J = Up /np D the advance ratio of the prop when it is exposed to a water speed Up . The propeller is operating in a wake. force balance in self-propulsion requires that Rsp = Tsp .52 12. A typical wake fraction of 0. We next introduce three factors useful for scaling and parameterizing our mathematical models: • Up = U (1 − w). and with the propeller spinning at the same np . for any ﬂow condition.

with no propeller.2 KT 0. = � o �R (1 − w) (156) .2 Steady Propulsion of Vessels 53 1.02 0.0 0. The thrust deduction is particularly useful. and can be estimated from published values. Note that in comparison with the wake fraction.6 J 0. and how they interact.4 0. t is a small positive number. and rotation rate np . translates self-propelled torque to open water torque.10 0. • Qpo = �R Qpsp . A common measure of eﬃciency. a propeller will increase the resistance of the vessel by creating low-pressure on its intake side (near the hull).2 as a typical value.0 0. for the same incident velocity Up . w is then chosen so that the open water thrust coeﬃcient matches what is observed. which makes Rsp > Rt .00 0. t is called the thrust deduction even though it is used to model resistance of the hull. Often.12. �QP = Rt U 2βnp Qpsp To (1 − t)Up �R = 2βnp (1 − w)Qpo (1 − t) .06 K Q 0. rotative eﬃciency equalizes torque instead of thrust. and the self-propelled torque.8 1. is based on the towed resis tance.6 0. as well as the turbulence induced by the hull.8 K η 0 T η0 0.0 K Q Figure 4: Typical thrust and torque coeﬃcients. the quasi-propulsive eﬃciency. �R is meant to account for spatial variations in the wake of the vessel which are not captured by the wake fraction. The wake fraction can also be estimated by making direct velocity measurements behind the hull. w scales U at the prop and thus J. with 0.0 0. it is obviously speciﬁc to both the hull and the propeller(s).08 0. In this case.2 0. • Rt = Rsp (1 − t). if only the towed resistance of a hull is known. which may be greater than one. The rotative eﬃciency �R .04 0. thrust T .4 0.

3 Engine/Motor Models The torque-speed maps of many engines and motors ﬁt the form Qe = Qm F (f /fm .2. Suppose that the towed resistance is of the form 1 (157) Rt = πCr Aw U 2 . For example. The vessel speed can be computed by recalling that J (U ) = U/np D and Up = U (1 − w). 2ζ �� ν � (158) The last equation predicts the steady-state advance ratio of the vessel. Equating the self-propelled thrust and resistance then gives Tsp = Rsp To = Rt /(1 − t) 1 πCr Aw U 2 KT (J (Up ))πn2 D 4 (1 − t) = p 2 2 Up 1 πCr Aw (α1 − α2 J (Up ))πn2 D 4 (1 − t) = p (1 − w)2 2 Cr Aw J (Up )2 α1 − α2 J (Up ) = 2 (1 − t)(1 − w)2 2D J (Up ) = � −α2 + � 2 α2 + 4α1 ζ . More speciﬁcally. The ratio (1−t)/(1−w) is often called the hull eﬃciency. but which are in competition. if F () has the form . and on the hull. but it is clear that we need now to ﬁnd np . and thus that �o is the open-water propeller eﬃciency at this speed. since it relies on the towed resistance and in general Rt > Rsp . 12. A high rotative eﬃciency and open water propeller eﬃciency (at Up ) obviously contribute to an eﬃcient overall system. gas turbines roughly ﬁt the curves shown in the ﬁgure (Rubis). This requires a torque equation.2 Solution for Steady Conditions The linear form of KT and KQ (Equation 156) allows a closed-form solution for the steadyoperating conditions. 2 where Cr is the resistance coeﬃcient (which will generally depend on Re and F r). ne /nm ). and we see that a small thrust deduction t and a large wake fraction w are beneﬁcial eﬀects. depending only on the propeller open characteristics. (159) where F () is the characteristic function. It follows that To (Up ) = Tsp .2.54 12 PROPELLERS AND PROPULSION To and Qpo are values for the inﬂow speed Up . 12. which was used to complete the above equation. which necessitates a model of the drive engine or motor. and Aw is the wetted area. The quasi-propulsive eﬃciency can be greater than one.

The dynamic response of the coupled propulsion and ship systems. The manipulations begin by equating the engine and propeller torque: Qpo (J (Up )) 2 5 πnp D KQ (J (Up )) 2 5 πnp D (ρ1 − ρ2 J (Up )) n2 p (nm /η)2 = �R Qps p (J (Up )) = �R �g ηQe = �R �g ηQm F (f /fm . under the assumption of quasi-static propeller conditions. f2 . ne /nm ) = − a fm nm fm np = −∂1 + ∂2 .2 Steady Propulsion of Vessels 55 Q e /Q m f3 f2 f1 n e /n m Figure 5: Typical gas turbine engine torque-speed characteristic for increasing fuel rates f1 . f f ne +b + c +d F (f /ff m. f3 . ne /nm ) �R �g ηQm np = −∂1 + ∂2 5 (ρ − ρ J (U ))(n /η)2 πD 1 (nm /η) 2 p m � � 2 σ2 ∂1 + 4σ∂2 . . this is left as a problem for the reader. ˙ (162) Making the necessary substitutions creates a nonlinear model with f as the input. is given by (m + ma )u = Tsp − Rsp ˙ 2βIp np = �g ηQe − Qps p . 2 � � −σ∂1 + np = (nm /η) �� σ � (161) Note that the fuel rate enters through both ∂1 and ∂2 . nm /η � � � � (160) then a closed-form solution for ne (and thus np ) can be found.12.

56

12 PROPELLERS AND PROPULSION

12.3

Unsteady Propulsion Models

When accurate positioning of the vehicle is critical, the quasi-static assumption used above does not suﬃce. Instead, the transient behavior of the propulsion system needs to be consid ered. The problem of unsteady propulsion is still in development, although there have been some very successful models in recent years. It should be pointed out that the models de scribed below all pertain to open-water conditions and electric motors, since the positioning problem has been central to bluﬀ vehicles with multiple electric thrusters. We use the subscript m to denote a quantity in the motor, and p for the propeller. 12.3.1 One-State Model: Yoerger et al.

The torque equation at the propeller and the thrust relation are Ip �p = ηQm − K� �p |�p | ˙ T = Ct �p |�p |. (163) (164)

where Ip is the total (material plus ﬂuid) inertia reﬂected to the prop;the propeller spins at �p radians per second. The diﬀerential equation in �p pits the torque delivered by the motor against a quadratic-drag type loss which depends on rotation speed. The thrust is then given as a static map directly from the rotation speed. This model requires the identiﬁcation of three parameters: Ip , K� , and Ct . It is a ﬁrst-order, nonlinear, low-pass ﬁlter from Qm to T , whose bandwidth depends directly on Qm . 12.3.2 Two-State Model: Healey et al.

The two-state model includes the velocity of a mass of water moving in the vicinity of the blades. It can accommodate a tunnel around the propeller, which is very common in thrusters for positioning. The torque equation, similarly to the above, is referenced to the motor and given as Im �m = −K� �m + Kv V − Qp /η. ˙ (165)

Here, K� represents losses in the motor due to spinning (friction and resistive), and Kv is the gain on the input voltage (so that the current ampliﬁer is included in Kv ). The second dynamic equation is for the ﬂuid velocity at the propeller: ˙ πALρUp = −πA�α(Up − U )|Up − U | + T. (166)

Here A is the disc area of the tunnel, or the propeller disk diameter if no tunnel exists. L denotes the length of the tunnel, and ρ is the eﬀective added mass ratio. Together, πALρ is the added mass that is accelerated by the blades; this mass is always nonzero, even if there is no tunnel. The parameter �α is called the diﬀerential momentum ﬂux coeﬃcient across the propeller; it may be on the order of 0.2 for propellers with tunnels, and up to 2.0 for open propellers.

57 The thrust and torque of the propeller are approximated using wing theory, which invokes lift and drag coeﬃcients, as well as an eﬀective angle of attack and the propeller pitch. However, these formulae are static maps, and therefore introduce no new dynamics. As with the one-state model of Yoerger et al., this version requires the identiﬁcation of the various coeﬃcients from experiments. This model has the advantage that it creates a thrust overshoot for a step input, which is in fact observed in experiments.

13

ELECTRIC MOTORS

Modern underwater vehicles and surface vessels are making increased use of electrical ac tuators, for all range of tasks including weaponry, control surfaces, and main propulsion. This section gives a very brief introduction to the most prevalent electrical actuators: The DC motor, the AC induction motor, and the AC synchronous motor. For the latter two technologies, we consider the case of three-phase power, which is generally preferred over single-phase because of much higher power density; three-phase motors also have simpler starting conditions. AC motors are generally simpler in construction and more robust than DC motors, but at the cost of increased control complexity. This section provides working parametric models of these three motor types. As with gas turbines and diesel engines, the dynamic response of the actuator is quite fast compared to that of the system being controlled, say a submarine or surface vessel. Thus, we concentrate on portraying the quasi-static properties of the actuator – in particular, the torque/speed characteristics as a function of the control settings and electrical power applied. The discussion below on these various motors is generally invertible (at least for DC and AC synchronous devices) to cover both motors (electrical power in, mechanical power out) and generators (mechanical power in, electrical power out). We will only cover motors in this section, however; a thorough treatment of generators can be found in the references listed. The book by Bradley (19??) has been drawn on heavily in the following.

13.1

13.1.1

Basic Relations

Concepts

First we need the notion of a magnetic ﬂux, analagous to an electrical current, denoted �; a common unit is the Weber or Volt-second. The ﬂux density

B = �/A

(167)

is simply the ﬂux per unit area, given in Teslas such that 1T = 1W/m2 . Correponding to electrical ﬁeld (Volts/m) is the magnetic ﬁeld intensity H, in Amperes/meter: � B = ; µo µr Aµo µr

H=

(168)

58

13 ELECTRIC MOTORS

µo ⇐ 4β × 10− 7Henries/meter is the permeability of free space, and µr is a (dimensionless) relative permeability. The product µo µr represents the real permeability of the material, and is thus the analog of electrical conductivity. A small area A or low relative permeability drives up the ﬁeld intensity for a given ﬂux �. 13.1.2 Faraday’s Law

The voltage generated in a conductor experiencing a time rate of change in magnetic ﬂux is given as e=− d� dt (169)

This voltage is commonly called the back-electromotive force or back-e.m.f., since it typically opposes the driving current; it is in fact a limiting factor in DC motors. 13.1.3 Ampere’s Law

**Current passing through a conductor in a closed loop generates a perpendicular magnetic ﬁeld intensity given by I=
**

�

Hdl.

(170)

An important point is that N circular wraps of the same conductor carrying current I induce the ﬁeld H = βDN I, where D is the diameter of the circle. 13.1.4 Force

Forces are generated from the orthogonal components of magnetic ﬂux density B and current I: F = I × B. (171)

The units of this force is N/m, and so represents a distributed force on the conductor.

13.2

DC Motors

The DC motor in its simplest form can be described by three relations: ea = K�� V = e a + R a ia T = K�ia , where

13.2 DC Motors • K is a constant of the motor • � is the airgap magnetic ﬂux per pole (Webers) • � is the rotational speed of the motor (rad/s) • ea is the back-e.m.f. • V is the applied voltage • Ra is the armature resistance on the rotor • ia is the current delivered to the armature on the rotor • T is developed torque

59

The magnetic ﬁeld in a typical motor is stationary (on the stator) and is created by perma nent magnets or by coils, i.e., Faraday’s law. Current is applied to the rotor armature through γ γ γ slip rings, and thus the force on each conductor in the armature is given by F = ia × B. Back-e.m.f. is created because the conductors in the rotor rotate through the stationary ﬁeld, causing a relative rate of change of ﬂux. The armature voltage loop contains the backe.m.f. plus the resistive losses in the windings. As expected, torque scales with the product of magnetic ﬂux and current. There are three main varieties of DC motors, all of which make use of the relations above. Speed control of the DC motor is primarily through the voltage V , either directly or through pulse-width modulation, but the stator ﬂux could also be controlled in the shunt/independent conﬁgurations. 13.2.1 Permanent Field Magnets

Here, the magnetic ﬁeld is created by permanent magnets arranged in the stator, imposing a steady �. The product K � is generally written as kt , the torque constant of a DC motor, and has units of N m/A. When SI units are used, kt also describes back-e.m.f.. The three basic relations are thus rewritten ea = k t � V = e a + R a ia T = k t ia , which leads via substitution to Ra T 1 , or V − � = kt kt kt T = [V − kt �] . Ra

� �

The ﬁeld windings are commonly placed in series with the rotor circuit.3 Series Windings � � When the ﬁeld windings are arranged in series with the rotor circuit. in parallel with it (shunt connection). in the sense that current and hence the ﬁeld strength can be modulated easily. no torque is created. 13.2. where Ks is a constant of the ﬁeld winding. but that it also scales negatively with the motor speed. This additional connection requires V = ea + (Ra + Rs )ia . the speed starts to degrade as shown in the ﬁgure. It follows through the substitutions that T = KKs i2 ∀ a Ia = � T KKs V Ra + R s − � = ≈ .2 Shunt or Independent Field Windings The ﬁeld created by the stator can be strengthened by replacing the permanent magnets with electromagnets. Hence. however. more commonly. or. We have Ra T 1 V − . At very high torques and currents. which makes these motor types ideal for self-regulation applications. Thus. compared with the permanent magnet case above. at the speed � = V /Kt . or. they may be powered from a completely separate circuit. the maximum torque is created at zero speed. The latter two cases are eﬀectively equivalent. the ﬂux is � = K s ia . Additionally. directly through current ia .60 13 ELECTRIC MOTORS This result indicates that the torque developed scales linearly with the applied voltage. which gives torque directly. � = K� K� with the important property that the second term in brackets is small due to the increased ﬁeld strength. the motor speed is eﬀectively independent of torque. through a variable resistance in the shunt case. KKs KKs T . the ﬁeld winding brings a new resistance Rs into the voltage loop.2. Control of these motors is through the voltage V . 13. the total available ﬂux will be reduced because of ﬁeld armature reactance.

so that the ﬁrst term dominates. 13. Equating the electrical (all three phases) and mechanical power gives . and so the rotor follows the rotating stator ﬁeld at the same speed. Then. let δ denote the angle between ia and V . there is no net torque. including increasing torque with increasing speed. which can oﬀset the speed droop due to ﬁeld armature reaction ef fects in the shunt motor. j = −1. leading to a nonlinear torque/speed characteristic. The stator ﬁeld windings are driven with three (balanced) phases of an AC supply. it should be observed that under light loading. These achieve other torque/speed curves. and referred to as com pound DC motors. b) shunt or independent ﬁeld winding. Some variations on the series and shunt connections are common. The torque exerted on the rotor tries to align the two ﬁelds. the voltage loop equation for a single phase on the stator gives V = ea + jia X.13. Note that if the rotor speed lags that of the stator ﬁeld.3 Three-Phase Synchronous Motor 61 The eﬀects of resistance are usually quite small. and X is the reactance (armature and stator leakage) of the machine. ea . where V . hence the name synchronous motor. At the same time. ≈ ia are now phasors (magnitude of V and ea measured with respect to and ground). As with the DC motors. A simple model of the synchronous motor is straightforward. c) series ﬁeld winding. such that a moving ﬁeld is created which rotates around the stator.3 Three-Phase Synchronous Motor The rotor is either ﬁtted with permanent magnets or supplied with DC current to create a static ﬁeld on the rotor. the series motor may well self-destruct since there is no intrinsic upper limit to speed! T V T T V V a) ω b) ω c) ω Figure 6: Torque-speed characteristics of three types of DC motors: a) permanent ﬁeld magnets. and the series ﬁeld winding ﬁnds wide application in railway locomotives. The starting torque from this kind of motor is exceptionally high.

the synchronous motor can slip one or more poles. . Letting the rotor speed be �r . or T = 2� X p Vab ea. causing a large transient disturbance. and thus the rotor lags the stator ﬁeld when the motor is powering. Speed control of the three-phase synchronous machine is generally through the frequency of the three-phase power supply.ab sin ζ T = 2� X where • p is the number of poles on the rotor. and the rotational speed of the stator ﬁeld ≈ • Vab is the line-to-line voltage.f.ab is the line-to-line back-e. If the load torque exceeds the available torque. a potential ﬁeld will be induced on any conductor on the rotor. Next. equal to 3ea • delta is the angle by which the rotor ﬁeld lags the stator ﬁeld (rad) • X is the synchronous reactance The torque scales with sin ζ. 13.62 13 ELECTRIC MOTORS 3V ia cos δ = T �. In the case ≥ of a squirrel-cage rotor design. • � is both the rotational speed of the rotor. two poles means one north pole and one south pole. we see immediately that if � = �r . the rotor is made of conductor bars which are shorted out through rings at the ends..m.4 Three-Phase Induction Motor Like the synchronous machine. with the assumption that adequate voltage and current are available. in a generator. �. the stator ﬁeld lags the rotor. Torque is then generated through the familiar F = I ×B relation. and hence the potential ﬁeld will cause a real current ﬂow. etc. The fact of unequal ﬁeld and rotor speeds in the induction motor is related to several unique eﬀects. It follows from the voltage loop equation that ia cos δX = ea sin ζ ∀ p 3V ea sin ζ . equal to 3V ≈ • ea. let ζ denote the angle between the phasors ea and V . the induction machine has windings on the stator to create a rotating magnetic ﬁeld at frequency �. leading to torque-speed characteristic which diﬀers signiﬁcantly from both the DC the AC synchronous machines.

The . say 0. they form an ideal inductor: N r Ir = N s Is . for reasons described below. for a constant voltage at the rotor. = s The factor of s in the rotor inductance occurs because the ﬁeld seen by the rotor is actually rotating at s�. � 63 (172) a slip ratio of zero means � = �r (and hence zero torque because the magnetic ﬁeld seen by the rotor is constant) while a slip ratio of one implies the rotor is stopped. Finally. but it rotates more slowly: s �r = �o sin �r t = �o sin s�t. In per-unit terms. On the rotor. and this then indicates that the rotor impedance. and equivalent resistance R and inductance X.13.1-0. Most induction motors are designed to operate at a small positive slip ratio. the number of ampere-turns on stator and rotor is equivalent. seen from the stator. Next. since the magnetic ﬂux lines pass through the rotor. Hence the rotor voltage is er = N d�r /dt = N �o s� cos s�t. the associated voltage is es = N d�/dt = N �o � cos �t. also scales inversely with the slip ratio: Zrs = 1 [Rr + jsXr ] s 1 Rr + jXr . the same ﬂux applies. The mechanical power is Pm = (1 − s)�T . Then the RMS voltage of the stator and rotor sides of the inductive coupling are related by Es 1 = . we construct the (one phase) Thevenin equivalent circuit of the stator: it has a voltage source Vt . that is. for which we set N = Nr = Ns and hence Ir = Is .4 Three-Phase Induction Motor First we deﬁne the slip ratio s= � − �r . Er s (174) The voltage seen at the stator scales inversely with the slip ratio. giving the following current I=� Vt (Rr /s + R)2 + (Xr + X)2 . Next. reﬂected to the stator. we need to express the torque/speed characteristic of the machine. This is to be paired with the rotor resistance and inductance. the current in the rotor and stator are equivalent.2. (173) We consider per-unit quantities from here on. If the stator ﬂux at a particular location is �s = �o sin �t. while the power delivered across the airgap is Pgap = I 2 Rr /s.

and do not have to be self-propelled. wing theory.5 2 Figure 7: Torque-speed characteristics of a typical three-phase induction motor. For example. etc.5 ω /ω m 1 1. s� [(Rr /s + R)2 + (Xr + X)2 ] Maximum torque is developed at a slight slippage. even a heavy steel cable with 2cm diameter. and in this section. The cables are generally circular in cross section. Some towed vehicles might have active lifting surfaces or thrusters for attitude control. linearization. which can generally be handled with the same tools as other vehicles. with decreased values at lower speeds.. Modern cables can easily exceed 5000m in length. towed vehicles are often streamlined. s=1 s=0 Torque −1 −0. Combining the mechanical power with the torque equation gives T = 3Pgap Pm = (1 − s)� � 3Vt2 Rr = .64 14 TOWING OF VEHICLES actual power dissipated in the copper is related to the real rotor resistance: Ploss = I 2 Rr . On the other hand. we concentrate on cable mechanics more than vehicle characteristics. and usually need good directional stability. and hence the mechanical power is Pm = 3(Pgap − Ploss ) = 3Pgap (1 − s). and may carry power conductors and . towed vehicles may be quite heavy in water. It follows that the eﬃciency of the motor is simply � = 1 − s.e. Some basic guidelines for vehicle design are given at the end of this section. 14 TOWING OF VEHICLES Vehicles which are towed have some similarities to the vehicles that have been discussed so far. i. if they are to be supported by a cable. slender-body theory. The cable itself is an important factor in the behavior of the complete towed system.5 0 0..

1 14. The heavy systems will generally transmit surface motions and tensions to the towed vehicle much more easily than light-tether systems. but rather look at heavy systems. upwards along the cable is the positive direction.14. such as a remotely-operated vehicle. We will not discuss light systems speciﬁcally here. Tension and hydrostatic pressure will elongate a cable. multiple communication channels (ﬁber optic).5 2 Figure 8: Eﬀects of constant Volts/Hz speed control.1 Statics 65 s=1 s=0 Torque −1 −0. The towed array is a relatively high-velocity system that nominally streams out horizontally behind the vessel.5 ωm/ωbase 1 1.1. and must have large propulsors to control the tether if there are currents. We employ the curvilinear axial coordinate s. Light-tether systems are characterized by neutrally-buoyant (or nearly so) cables. and one main division may be made simply of the density of the cable. or a vehicle capable of maneuvering itself. Most of the analysis can be adapted to either case. voltage is constant above electrical frequency �base . Heavy systems. but the eﬀect is usually a small percentage of the total length. operates at low speed. on the other hand. which we take to be zero at the bottom end of the cable. 14. we assume for the moment that that it is inextensible. as in a towed array. The free-body diagram shown . Cable systems come in a variety of conﬁgurations. with either a minimal vehicle at the end. An ROV. the rationale is that gravity will tend to keep the cable vertical and make the deployment robust against currents and towing speed. The extreme L/D ratio for these cables obviates any bending stiﬀness eﬀects.5 0 0. in contrast. employ a heavy cable and possibly a heavy weight. however.1 Statics Force Balance For the purposes of deriving the static conﬁguration of a cable in a ﬂow.

leading to 1 Rt = − πCt dU 2 cos2 δ 2 1 Rn = − πCn dU 2 sin2 δ. • T (s): local tension. 2 (177) (178) Note that we simpliﬁed the drag laws from a usual form v |v | to v 2 . In both cases. Force balance in the tangential and normal coordinates gives two coupled equations for T and δ: dT = Wn sin δ − Rt ds dδ T = Wn cos δ + Rn .66 14 TOWING OF VEHICLES y x s+ ds ds Rt φ T+dT U ds Rn ds Wn T has the following components: • Wn : net in-water weight of the cable per unit length. • Rt (s): external tangential force. per unit length. U horizontal toward the left. and the normal drag scales with a crossﬂow drag coeﬃcient Cn . • Rn (s): external normal force. 0 √ δ √ β/2. since as drawn. • δ(s): local inclination angle. per unit length. is to be projected onto the relevant axes. the tangential drag is controlled by a frictional drag coeﬃcient Ct . ds (175) (176) The external forces are primarily ﬂuid drag. . the ﬂuid velocity vector.

14.1 Statics

67

φc

W

The equations for T and δ can be integrated along the cable coordinate s to ﬁnd the cable’s static conﬁguration. Two boundary conditions are needed, and the common case is that a force balance on the vehicle, dominated by drag, weight, and the cable tension, provides both T (0) and δ(0). For example, a very heavy but low-drag vehicle will impose δ(0) � β/2, with T (0) equal to the in-water weight of the vehicle. With regard to Cartesian coordinates x, y, the cable conﬁguration follows dx = cos δ ds dy = sin δ. ds

(179) (180)

The simultaneous integration of all four equations (T, δ, x, y) deﬁnes the cable conﬁguration, and current dependency may be included, say U is a function of y. 14.1.2 Critical Angle

For very deep systems, the total weight of cable will generally exceed that the vehicle. This gives rise to a conﬁguration in which the cable is straight for a majority of its length, but turns as necessary at the vehicle end, to meet the bottom boundary condition. In the straight part of the cable, normal weight and drag components are equalized. The uniform angle is called the critical angle δc , and can be approximated easily. Let the relative importance of weight be given as Wn , πCn dU 2

ζ=

68

14 TOWING OF VEHICLES

so that the condition dδ/ds = 0 requires from the force balance 1 2 sin δc = 0. 2

ζ cos δc −

We are considering the case of 0 < δc < β/2. Substituting sin2 δc = 1 − cos2 δc , we solve a quadratic equation and keep only the positive solution: cos δc = ≈ ζ 2 + 1 − 1. (181)

In the case of a very heavy cable, ζ is large, and the linear approximation of the square root ≈ 1 + σ ⇐ 1 + σ/2 gives cos δc � δc 1 −∀ 2ζ β 1 − . � 2 2ζ

(182)

For a very light cable, ζ is small; the same approximation gives cos δc � 1 − ζ −∀ δc � ≈ 2ζ.

The table below gives some results of the exact solution, and the approximations.

ζ exact 0.1 0.44 0.2 0.61 0.5 0.91 1.0 1.14 2.0 1.33 5.0 1.47

ζ >> 1 ζ << 1 0.45 0.63 0.57 1.00 1.07 1.41 1.32 1.47 -

14.2

14.2.1

Linearized Dynamics

Derivation

The most direct procedure for deriving useful linear dynamic equations for a planar cable problem is to consider the total tension and angle as made up of static parts summed with dynamic parts: ¯ ˜ T (s, t) = T (s) + T (s, t) ¯ ˜ δ(s, t) = δ(s) + δ(s, t).

14.2 Linearized Dynamics

69

We also write the axial deﬂection with respect to the static conﬁguration as p(s, t), and the ˜ lateral deﬂection q(s, t). It follows that δ = �q/�s. Now augment the two static conﬁguration equations with inertial components: ¯ �T ˜ �2p �T 1 �p ¯ ˜ m 2 = + − Wn sin(δ + δ) − πCt d U cos δ + �t �s �s 2 �t � � ˜ ¯ �2q ¯ ˜ ¯ ˜ �δ + �δ − Wn cos(δ + δ) + (m + ma ) 2 = (T + T ) �t �s �s

1 �q πCn d U sin δ − 2 �t

� �2

� �2

.

Here the material mass of the cable per unit length is m, and its transverse added mass is ma . Note that avoiding the drag law form v |v | again, we have implicitly assumed that U cos δ > |�p/�t| and U sin δ > |�q/�t|. If it is not the case, say U = 0, then equivalent linearization can be used for the quadratic drag. ¯ Now we perform the trigonometry substitutions in the weight terms, let δ � δ for the calculation of drag, and substitute the constitutive (Hooke’s) law ˜ �2p �T = EA 2 . �s �s The static solution cancels out of both governing equations, and keeping only linear terms we obtain �2p �2p ¯ �q ¯ �p m 2 = EA 2 − Wn cos δ − πCt dU cos δ �t �s �s �t ¯ �2q �2q �p �δ �q ¯ ¯ �q ¯ (m + ma ) 2 = T 2 + EA + Wn sin δ − πCn dU sin δ . �t �s �s �s �s �t ¯˜ The axial dynamics (p) couples with the lateral equation through the weight term −Wn cos δδ. ˜ ¯ The lateral dynamics (q) couples with the axial through the term T �δ/�s. An additional ¯ weight term Wn sin δ�q/�s also appears. The uncoupled dynamics are both in the form of damped wave equations �2p �p �2p = EA 2 + bt �t2 �t �s �2q �q �2q ¯ �q ¯ (m + ma ) 2 + bn = T 2 + Wn sin δ , �t �t �s �s m ¯ ¯ where we made the substitution bt = πCt dU cos δ and bn = πCn dU sin δ. To a linear ap proximation, the out-of-plane vibrations of a cable are also governed by the second equation above.

and can develop longitudinal resonant conditions (next section).70 14 TOWING OF VEHICLES Because of light damping in the tangential direction. while low-frequency motions occur sluggishly. and only after a signiﬁcant delay time. and primes to indicate spatial derivatives. t) = P cos �t. We let p(s. At the surface. is a very slow. the lateral cable motions are heavily damped. damped nonlinear ﬁlter. Highfrequency vessel motions in the horizontal plane are completely missed by the vehicle.2 Damped Axial Motion nψ L lateral � ¯ T (s) m+ma � EA m SLOW � ¯ T (L/2) m+ma sine/cosine Ct � O(0.01) YES � EA m O � nψ L � Bessel function Cn � O(1) NO Mode Shape. such that disturbances only travel a few tens or hundreds of meters before they dissipate. axial wave speed FAST natural frequency mode shape damping disturbances travel down cable? 14. t) = p(s) cos �t. since it is lightly damped and forced by the heaving of vessels in seas. so that ˜ EA m� 2 − i�bt p + ˜ p = 0. at the lower end. The axial direction is of particular interest. ¨ (185) �s These top and bottom behaviors comprise the boundary conditions for the wave equation. ˜ EA ∗∗ � � (186) . The nature of the lateral response. Consider a long cable governed by the damped wave equation mp + bt p = EAp∗∗ ¨ ˙ (183) We use over-dots to indicate time derivatives. in and out of the towing plane. we impose the motion p(L. In contrast.2. (184) while the towed vehicle. is an undamped mass responding to the local tension variations: �p(0. heavy cables easily transmit motions and tensions along their length. t). t) = Mp(0.

14. where the static tension is highest. In general. and investi gating the singularity of p. T = −EAP k ζ cos kL − sin kL There are two dangerous situations: (189) � ¯ ˜ • The maximum tension is T + |T | and must be less than the working load of the cable. simplifying the result to p = P cos ks/ cos kL. 6 � � Employing the deﬁnition for ζ. Natural Frequency. for which ζ cos kL� sin kL. The top and bottom boundary conditions give. (188) ζ cos kL − sin kL In the case that M ∀ 0. ˜ ¯ • If |T | > T .2 Linearized Dynamics This admits the solution p(s) = c1 cos ks + c2 sin ks. k= P = c1 cos kL + c2 sin kL 0 = c1 + ζc2 . It is possible to compute the dynamic tension via T = EAp∗ . the system is dominated by the vehicle mass. we arrive at mL (kL)2 1− M 2 � � 2 = (kL) � (kL)2 1− . This is normally problematic at the top of the cable. We ˜ obtain ζ sin ks + cos ks ˜ . We start with better approximations for sin() and cos(): (kL)2 ζ 1− 2 � � (kL)2 = kL 1 − . which is only a correct answer if M >> mL. but we ﬁnd ˜ = that a ﬁrst-order approximation yields � = EA/LM . the cable will unload completely and then reload with extremely high impulsive forces. where ζ = EAk/� 2 M . 6 � . respectively. The natural frequency can be found by letting bt = 0. Some higher order terms need to be kept. where ˜ 71 m� 2 − i�bt .. ˜ p=P ˜ ˜ Dynamic Tension. These can be combined to give the solution ζ cos ks − sin ks . i. the scalar ζ ∀ ∗. (187) EA Note that k is complex when bt ≥= 0. and recalling that � 2 = k 2 EA/m.e. ¯ where T is low. it occurs primarily at the vehicle. kL << 1. This is known as snap loading.

it may be undesirable to expose the vehicle to these motions. Strumming of amplitude d/2 or greater can occur for 0. This drag penalty decreases the critical angle of towing. Strumming of cables is caused by the proximity of a preferred vortex formation frequency � S to the natural frequency of the structure �n . In very deep water. if πc = m/A is the density of the cable. then ⎦ � � � 14 TOWING OF VEHICLES �= EA/L . where S is the Strouhal number. including obvious fatigue when the amplitudes and frequencies are high.2 for a nonoscillating cable to as high as 3. Some solutions in use today are: stable vessels (e.6rad/s 2000m 3. about 0. and the mass that is oscillating is M + mL/2. SWATH).7rad/s 5000m 1.0. so that larger lengths of cable are needed to reach a given depth. heave compensation through an active crane.20 for a large range of Re.0rad/s 1000m 7.g.5rad/s 14.72 If we match up to second order in kL. M + mL/2 This has the familiar form of the square root of a stiﬀness divided by a mass: the stiﬀness of the cable is EA/L.5. and the towed system lags further and further behind the surface vessel. a clump weight below which a light cable is employed. The book by Blevins is a good general reference. and an S-shaped length of cable at the bottom formed with ﬂotation balls. The most noteworthy issue with towing is that the vibrations may cause the normal drag coeﬃcient Cn to increase dramatically – from about 1. The preferred frequency of vortex formation is given by the empirical relation �S = 2βSU/d. the eﬀects of mL/2 dominate. .. we have the approximation 1 �� L � 2E . The static tension will increase accordingly as well. L = 500m �n = 15. The natural frequencies near wave excitation at the surface vessel must be taken into account in any design or deployment.3 Cable Strumming Cable strumming causes a host of problems. and πc = 7000kg/m3 .6 < �S /�n < 2. Even if a cable can withstand the eﬀects of resonance. πc A few examples are given below for a steel cable with E = 200×109 P a.16-0. This latter frequency can be obtained as a zero of the lightly-damped Bessel function solution of the lateral dynamics equation above.

so that the vehicle enters the water ﬁns ﬁrst. 3. 5. 15 TRANSFER FUNCTIONS & STABILITY The reader is referred to Laplace Transforms in the section MATH FACTS for preliminary material on the Laplace transform. This will ensure that high-frequency disturbances do not induce excessive pitching. away from the surface. in the context of control systems. as the fundamental link between pole locations and stability. 4. (s + p1 )(s + p2 ) · · · (s + pn ) (190) Although for the moment we are discussing the signal Y (s). The towpoint must be located above the vehicle center of in-water weight. Meeting all of these criteria simultaneously is no small feat. Partial fractions are presented here. For this reason.4 Vehicle Design 73 14. The combined center of mass (material and added mass) should be longitudinally between the towpoint and the aerodynamic center. and self-stabilizes with U > 0.4 Vehicle Design The physical layout of a towed vehicle is amenable to the analysis tools of self-propelled vehicles. . and that the vehicle can be quite heavy in water. The towpoint should be forward of the aerodynamic center. 2.1 Partial Fractions Solving linear time-invariant systems by the Laplace Transform method will generally create a signal containing the (factored) form Y (s) = K(s + z1 )(s + z2 ) · · · (s + zm ) . with the main exceptions that the towpoint presents a large mean force as well as some disturbances. 15. and nearer the towpoint. so that the vehicle pitches downward at small U . full-scale experiments are commonly used in the design process.14. for basic roll and pitch stability. and hence the net lift force is downward. The center of buoyancy should be behind the in-water center of weight. The towpoint should be longitudinally forward of the center of in-air weight. later we will see that dynamic systems are described in the same format: in that case we call the impulse response G(s) a transfer function. Here are basic guidelines to be considered: 1. A system transfer function is identical to its impulse response. and the performance of the device is very sensitive to small perturbations in the geometry. since L(ζ(t)) = 1. for towing stability reasons.

evaluate them at s = −pi . We have s+6 a1 a2 = + . and solve for ai .3 Example: Partial Fractions with Unique Real Poles G(s) = s(s + 6) e−2s . if this is not the case. it is clear that the magnitude of Y (s) will go to zero at the zeros.74 15 TRANSFER FUNCTIONS & STABILITY The constants −zi are called the zeros of the transfer function or signal. 15. and −pi are the poles. How to ﬁnd ai ? A simple rule applies: multiply the right-hand sides of the two equations above by (s + pi ). we can initially work with G(s)/se−2s . (191) s + p1 s + p2 s + pn in the special case that all of the poles are unique and real. Partial fraction expansions alter the form of Y (s) so that the simple transform pairs can be used to ﬁnd the time-domain output signals. (s + 4)(s − 1) Since we have a pure delay and m = n. and once all these are found it is a simple matter to go back to the transform table and look up the time-domain responses. it is a fact that Y (s) is equivalent to a2 an a1 + +··· . Note that in this example.2 Partial Fractions: Unique Poles Under the condition m < n. giving (s + 4)(s − 1) s+4 s−1 a1 = a2 Thus 2 7 L−1 (G(s)/se−2s ) = − e−4t + et −∀ 5 5 8 7 g(t) = ζ(t − 2) + e−4(t−2) + et−2 . and to inﬁnity at the poles. then we have to divide the numerator by the denominator as necessary to ﬁnd a simple form. Y (s) = 15. Viewed in the complex plane. We must have m < n. The coeﬃcient ai is termed the residual associated with the i’th pole. we were able to apply the derivative operator s after expanding the partial = � (s+6)(s+4) (s+4)(s−1) s=−4 (s+6)(s−1) (s+4)(s−1) s=1 � � = − = 7 5 2 5 � . the only one left. 5 5 The impulse response is needed to account for the step change at t = 2.

i. 15. also � s+1 (s + j)(s − j) = a3 = 1.4 Partial Fractions: Complex-Conjugate Poles A complex-conjugate pair of poles should be kept together. but this condition rarely occurs in applications. s=0 Working out the inverse transforms from the table of pairs.6 Stability in Linear Systems In linear systems.5 Example: Partial Fractions with Complex Poles G(s) = b1 s + b 2 a3 s+1 = + : s(s + j)(s − j) (s + j)(s − j) s � s+1 s � s=−j = [b1 s + b2 ]s=−j −∀ ⎬ 1 + j = −b1 j + b2 −∀ b1 = −1 b2 = 1. multiply through by (s + p1 )(s + p2 ). If one exponent is 0. then one element of the response grows without bound. exponential stability occurs when all the real exponents of e are strictly negative. the response never decays or grows in amplitude. and 2 then evaluate at s = −p1 . . For cases where a second derivative must be taken. The more traditional method. If at least one real exponent is positive. with the following procedure: employ the form Y (s) = b1 s + b 2 a3 + + ···. (s + p1 )(s + p2 ) s + p3 (192) where p1 = p� (complex conjugate). special care should be used when accounting for the signal slope discontinuity at t = 0.15.e.. this is called marginal stability. 15. and the system is unstable. m → n + 1. The signals decay within an exponential envelope. 15. The case of repeated real roots may be handled elegantly. As before. exempliﬁed by Ogata. we have simply (noting that ψ = 0) g(t) = − cos t + sin t + 1(t).4 Partial Fractions: Complex-Conjugate Poles 75 fractions. may prove easier to work through.

15. GPS gyrocompass joint angle inﬂation. the relationship between stability and pole locations is especially clear. u 0 y y The output asymptotically approaches zero. within a decaying exponential envelope. 16 16.S.8 General Stability There are two deﬁnitions. If γ(0) = γ and |fi (t)| < ρ. then |yi (t)| < ∂e−ρt . and unstable parts to a positive real part. The unit step function 1(t) has a pole at zero. economy Nuclear reactor Inputs elevator. Marginally stable parts correlate with a zero real part. ﬁnite. Exponential stability is easy to check for linear systems.7 Stability �� Poles in LHP In the context of partial fraction expansions. rudder angle valve position fed interest rate. and distill alcohol. All of the other pairs exhibit the same property: A system is stable if and only if all of the poles occur in the left half of the complex plane. for ﬁnite ∂ and ρ > 0. ﬂy jets. hdg heading tip position prosperity power level Sensors altimeter.1 16. In linear time-invariant systems. which apply to systems with input γ (t) and output γ(t). etc. It is the fundamental concept of controller design that a set of input variables acts through a given “plant” to create an output. M1 temp. pressure . we employ controllers to steer ships. ∂ > 0 and ﬁnite. and so on. 2. then |yi (t)| < ∂.1 CONTROL FUNDAMENTALS Introduction Plants. position electric motors and hydraulic actuators. Many target systems are physical. and Outputs Controller design is about creating dynamic systems that behave in useful ways. the two deﬁnitions are identical. but for nonlinear systems. ρ > 0 and y 0. the exponential e−at has a pole at −a. Feedback control then uses sensed plant outputs to apply corrective inputs: Plant Jet aircraft Marine vessel Hydraulic robot U. rudder. Controllers are also applied in macro economics and many other important. neutron ﬂux Outputs altitude.. Bounded-Input Bounded-Output (BIBO). Inputs. cooling.76 16 CONTROL FUNDAMENTALS 15. non-physical systems. etc. u y 1. Exponential. If γ (t) = γ and γ (0) = γo . BIBO stability is usually easier to achieve.1.

Beyond prediction of plant behavior based on physics. The process is often problematic. control system design depends on a model of the plant. diﬃcult to design controllers for. and diﬃcult overall! Within the paradigm of linear systems. The reader interested in nonlinear control is referred to the book by Slotine and Li (1991).16. there are many sets of powerful tools available. plant models simply do not match real-world systems exactly. In any event. The main reason for this restriction is that nonlinear systems are diﬃcult to model. the process of system identiﬁcation generates a plant model from data. The transfer function description of linear systems has already been described in the discussion of the Laplace transform. A captures the natural dynamics of the system without external inputs.2. for example.1. and this is a limit to controller performance. Ziegler and Nichols. although it must be noted that many industrial controllers can be tuned up satisfactorily with no knowledge of the plant. 16.3 Nonlinear Control The bulk of this subject is taught using the tools of linear systems analysis. we can only hope to capture the basic components in the form of diﬀerential or integro diﬀerential equations.2 Representing Linear Systems Except for the most heuristic methods of tuning up simple systems. • A is a matrix mapping x to its derivative. At some frequency high enough.1 Standard State-Space Form We write a linear system in a state-space form as follows x = Ax + Bu + Gw ˙ y = Cx + Du + v where • x is a state vector.2 Representing Linear Systems 16. and opens the way to standard tools from linear algebra. developed a general recipe which we detail later. 16. since the measured response could be corrupted by sensor noise or physical disturbances in the system which cause it to behave in unpredictable ways.2 The Need for Modeling 77 Eﬀective control system design usually beneﬁts from an accurate model of the plant. most systems exhibit eﬀects that are diﬃcult to model or reproduce.1. with as many elements as there are orders in the governing diﬀer ential equations. (193) . 16. however. The state-space form is an entirely equivalent time-domain representation that makes a clean extension to systems with multiple inputs and multiple outputs.

For example. we obtain the system x .3 Converting a Transfer Function into a State-Space Model It may be possible to write the corresponding diﬀerential equation along one row of the state vector. y]T .2. which can create the same transfer function. comprised mainly of a linear combination of states Cx (where C is a matrix). w G + + + A x’ x + v + y + 1/s C u B D 16.2 Converting a State-Space Model into a Transfer Function There are a number of canonical state-space forms available. A similar equation holds for y(s)/w(s). and then cascade derivatives. w drives the state just like the control u. u(s) (194) where I is the identity matrix with the same size as A. 16. consider the following system: my ∗∗ (t) + by ∗ (t) + ky(t) = u∗ (t) + u(t) (mass-spring-dashpot) s+1 G(s) = 2 + bs + k ms Setting γ = [y ∗ . • G is a gain matrix for unknown disturbance w. • v is an unknown sensor noise which corrupts the measurement. • y is the observation vector. the transfer function (or transfer matrix) can be written as G(s) = y(s) = C(sI − A)−1 B + D.2. • Du is a direct map from input to output (usually zero for physical systems). and clearly y(s)/v(s) = I. In the case of no disturbances or noise.78 16 CONTROL FUNDAMENTALS • B is an input gain matrix for the control input u.

and so the control action from this part will continue to grow as long as an error exists. In words. . φi s where the last line is written using the conventions of one overall gain kp . the derivative action scales with the derivative of the error. If u is the output from the controller. but is also subject to an unknown disturbance d(t). P D. plus a time characteristic to the integral part (φi ) and and time characteristic to the derivative part (φd ).3 PID Controllers 79 dγ x −b/m −k/m = 1 0 dt y = [1 1] γ x ⎬ � γ+ x ⎬ 1/m 0 � u Note speciﬁcally that dx2 /dt = x1 . and it is perfectly measured. If the linear position of the mass is y(t).3 PID Controllers The most common type of industrial controller is the proportional-integral-derivative (PID) design. we have the plant my ∗∗ (t) = u(t) + d(t). It has a perfect thruster that generates force u(t). P ID. The integrator is accumulating the error signal over time. with initial conditions y(0) = yo and y ∗ (0) = 0. 16. the system response to u∗ is the same as the derivative of the system response to u. leading to an entry of 1 in the oﬀ-diagonal of the second row in A.16. The controller will retard motion toward zero error. (195) U (s) ki = kp + + kd s E(s) s � � 1 = kp 1 + + φd s . The common variations are: P . Finally. 16. and e is the error signal it receives. this control law has the form � t 0 u(t) = kp e(t) + ki C(s) = e(φ )dφ + kd e∗ (t). the proportional part of this control law will create a control action that scales linearly with the error – we often think of this as a spring-like action. Entries in the C-matrix are easy to write in this case because of linearity. P I. Suppose that the desired condition is simply y(t) = 0.4 Example: PID Control Consider the case of a mass (m) sliding on a frictionless table. which helps to reduce overshoot.

Note that if the mass had a very large amount of natural damping.4.1 Proportional Only 16 CONTROL FUNDAMENTALS A proportional controller alone invokes the control law u(t) = −kp y(t).80 16. easily found using the Laplace transform. the system would be overdamped and very slow to respond to any command. a small amount of overshoot is employed because the response time is shorter. Now consider what happens if d(t) has a constant bias do : it balances exactly the proportional control part. With an excessively large value for kd . In most applications. and kd the part of the dashpot.2 Proportional-Derivative Only � kp t. eventually settling out at y(t = ∗) = do /kp .4. The result. which is unacceptable. 16. is y(t) = yo e � −kd t 2m ⎬ kd cos �d t + sin �d t . m a marginally stable response that is Let u(t) = −kp y(t) − kd y ∗ (t). so that the closedloop dynamics follow my ∗∗ (t) = −kp y(t) + d(t). . 2m�d � 2 where �d = 4mkp − kd /2m. a negative kd could be used to cancel some of its eﬀect and speed up the system response. The control system has now created a third-order closed-loop response. In the absence of d(t). a time derivative leads to my ∗∗∗ (t) + kp y ∗ (t) + ki y(t) + kd y ∗∗ (t) = 0. we see that y(t) = yo cos undesirable.3 Proportional-Integral-Derivative �t 0 Now let u(t) = −kp y(t) − ki y(φ )dφ − kd y ∗ (t): we have � t 0 my ∗∗ (t) = −kp y(t) − ki y(φ )dφ − kd y ∗ (t) + d(t). very large values of kp will also drive the resonant frequency �d up.4. we would need to set kp very large. The system now resembles a second-order mass-spring-dashpot system where kp plays the part of the spring. and so the rule is kd < 2 mkp . If d(t) = d o . so that y(t = ∗) = 0. To achieve good rejection of do with a P D controller. This response is exponentially stable as desired. 16. as desired. provided the roots are stable. However. � � The kd value for critical damping in this example is 2 mkp . and it follows that my ∗∗ (t) = −kp y(t) − kd y ∗ (t) + d(t).

B´langer.1 Block Diagrams of Systems Fundamental Feedback Loop The topology of a feedback system can be represented graphically by considering each dy namical system element to reside within a box. For example. the input e is 1(t)) into a reasonable PID design. lightly-damped complex poles. This is especially pertinent for plants which are open-loop stable. position y(s). which relates the input. 16. which transforms the basic characteristics of a step response (e.27φ /T 2 ki = 0. having an input line and an output line.9φ /T kp = 1.2 Block Diagrams: General Case The simple feedback system above is augmented in practice by three external inputs.0φ /T kp = 0. it is additive with the control action. The feedback loop in block diagram form is shown below. force u(s). In the former case. The idea is to approximate the response curve by a ﬁrst-order lag (gain k and time constant φ ) and a pure delay T : ke−T s (196) φs + 1 The following rules apply only if the plant contains no dominating. One useful approach is due to Ziegler and Nichols (e.g..60φ /T 2 kd = 0.6..1995).5 Heuristic Tuning For many practical systems.g. which can be taken to act at the input of the physical plant.5 Heuristic Tuning 81 16. In turn. the PD-controller has transfer function C(s) = kp + kd s.2φ /T ki = 0. and its output is force u(s). The ﬁrst is a process disturbance.6. its input is the error signal E(s) = −y(s). and has no poles at the origin: G(s) � P PI PID kp = 1. and can be safely tested with varying controllers.60φ Note that if no pure time delay exists (T = 0). tuning of a PID controller may proceed without any system model.6 16.16. and so has . C(s) u(s) P(s) y(s) 16. this recipe suggests the proportional gain can become arbitrarily high! Any characteristic other than a true ﬁrst-order lag would therefore be expected to cause a measurable delay. or at the output. the plant used above (a simple mass) has transfer function P (s) = 1/ms2 . into the output.

causing some error in the evaluation of e(s). which usually corrupts the feedback signal y(s). and e du y du u du e dy −P 1 + PC P = 1 + PC −CP = 1 + CP −1 = = −S 1 + PC = . The ﬁnal input is sensor noise. We would generally like S to be small at low frequencies. y/r = T is called the complementary sensitivity function.3 Primary Transfer Functions Some algebra shows that e 1 = =S r 1 + PC PC y = =T r 1 + PC C u = = U. and so on. Another external input is the reference command or setpoint.82 16 CONTROL FUNDAMENTALS some physical meaning. Sensors with very poor noise properties can ruin the performance of a control system. in trying to force e(s) to zero. r 1 + CP e/r = S relates the reference input and noise to the error. and is known as the sensitivity function. the loop transfer function broken between C and P : CP .6. In the second case. Note that S + T = 1. Other systems we encounter again later are the (forward) loop transfer function P C. will necessarily make y(s) approximate r(s). implying that these two functions must always trade oﬀ. used to create a more general error signal e(s) = r(s) − y(s). so that the tracking error there is small. they cannot both be small or large at the same time. no matter how perfectly understood are the other components. du dy + P + + + n y r + - e C u + + 16. Note that the feedback loop. the disturbance has the same units as the plant output.

2 17. consider the system ˙ γ = Aγ . 17 17. The same notation holds for the case of a vector γ .2. . which can be driven by external inputs. If the disturbance is taken at the plant output. T . the response is x(t) = νeat . This will in fact be the procedure when we address loopshaping.1 MODAL ANALYSIS Introduction The evolution of states in a linear system occurs through independent modes. x x γ (t = 0) = ν. Throughout. x γ In the scalar case. and U (control action) completely describe the system. and matrix A: x γ (t) = eAt ν.83 y dy u dy e n y n u n = = = = = 1 1 + PC −C 1 + CP −1 1 + PC −P C 1 + PC −C 1 + CP =S = −U = −S = −T = −U. giving a decaying exponential if a < 0. and observed through plant output. where x γ (At)2 At e = I + At + +··· 2! eAt is usually called the matrix exponential. we use the state-space description of a system with D = 0: ˙ γ = Aγ + Bγ x x u γ = Cγ y x. 17. This section provides the basis for modal analysis of systems. then the three functions S.1 Matrix Exponential Deﬁnition In the instance of an unforced response to initial conditions.

γ γ Otherwise. γT γ If ν is perpendicular to wi .3 Forced Response and Controllability Now consider the system with an external input γ : u . since A = V ΦV −1 . The modal form is numerically robust for computations. they are in fact the poles of the equivalent transfer function description. x v For stability of the system. then the product is zero and the i’th mode does not respond. Namely. 17. Employing the above form for A. we have γ n ⎫ i=1 γ (t) = x e�i tγi (wi ν). the eigenvalues of A. v γT γ The product wi ν is a scalar. These states are uncoupled in Φ.2 Modal Canonical Form 17 MODAL ANALYSIS Introductory material on the eigenvalue problem and modal decomposition can be found in the MATH FACTS section. since the states are simply the modal amplitudes. we ﬁnd that (At)2 = I + At + +··· 2! � � (Φt)2 = V I + Φt + + ··· WT 2! = V e�t W T = n ⎫ i=1 � v γT e i tγi wi . γ = V γ. ηi . the i’th mode does participate in the response. The projection of the i’th mode onto the states γ is through the right eigenvector γ i . (197) This is called the modal canonical form.84 17.2. that is. This modal decomposition of A leads to a very useful state-space representation.3 Modal Decomposition of Response Now we are ready to look at the matrix exponential eAt in terms of its constituent modes. but may be coupled through the input (V −1 B) and output (CV ) mappings. a transformation of state variables can be made. the projection of the initial conditions onto the i’th mode. 17.2. must have negative real parts. e At In terms of the response to an initial condition ν. leading to x z ˙ γ = Φγ + V −1 Bγ z z u γ = CV γ y z.

represents the projection of the k’th control channel onto the γTb i’th mode. such that B = [γ 1 .4 Plant Output and Observability We now turn to a related question: can the complete state vector of the system be observed y. 17. x γ u Thus (sI − A)−1 can be recognized as the Laplace transform of the matrix exponential eAt . taking into account the initial condition for the derivative. γ 2 . Consider the part due to initial conditions ν.17. i.4 Plant Output and Observability 85 ˙ γ = Aγ + Bγ . a scalar. (198) This matrix has size n × (nm). then the mode is γTb uncontrollable.e. An−1 B]. then the system is controllable. x x u γ (t = 0) = ν. through the convolution integral. If Mc has rank n. γ m ]. the second term then has the form of a convolution of the matrix exponential and the net input Bγ : u � t 0 γ (t) = x = u eA(t−φ ) Bγ (φ )dφ t � e i (t−φ )γi wi Bγ (φ )dφ. given only the output measurements γ and the known control γ The response due to u? the external input is easy to compute deterministically. In normal applications. u The product wi γ k . · · · .. controllability for the entire system is checked using the following test: Construct the so-called controllability matrix: Mc = [B. where m is the number of input channels. · · · . we have sγ (s) − ν = Aγ (s) + Bγ (s) −∀ x γ x u −1 γ (s) = (sI − A) ν + (sI − A)−1 Bγ (s). Then some rearrange b b b ment will give n ⎫ i=1 m ⎫ γ (t) = x γi v (wi γ k ) γTb k=1 � t 0 e�i (t−φ ) γ k (φ )dφ. x γ Taking the Laplace transform of the system. We found above γ . We say that the i’th mode is controllable from the k’th input if the product is nonzero. all modes are controllable. AB. In the time domain. If a given mode i has wi γ k = 0 for all input channels k. v γT u n ⎫� i=1 0 Suppose now that there are m inputs.

In particular. The loopshaping approach is inherently two-fold. absolute values of transfer functions are replaced with the maximum singular values of transfer matrices. to be presented in the next lectures. or LQG/LTR. The sensitivity function is written as . AT C T . as given previously. single-output systems. the system is observable if Mo has rank n.86 18 CONTROL SYSTEMS – LOOPSHAPING γ (t) = x n ⎫ i=1 e�i tγi wi ν.2 Roots of Stability – Nyquist Criterion We consider the SISO feedback system with reference trajectory r(s) and plant output y(s). and writing y x γ1T c ⎞ ⎛ C = ⎝ · ⎠. Once this is done. v γT γ The observation is γ = Cγ (r channels of output). Design based on singular values is the idea of L2 -control. 18. The tracking error signal is deﬁned as e(s) = r(s) − y(s). Most of the analysis here is given for single-input. to meet performance and robustness speciﬁcations. and the nominal plant P (s). c v γT γ The i’th mode is observable in the k’th output if the product γkT γi ≥= 0. from from knowing the nominal product P (s)C(s). This matrix has size n × (rn). we shape the open-loop transfer function (or matrix) P (s)C(s). thus forming the negative feedback loop.1 CONTROL SYSTEMS – LOOPSHAPING Introduction This section formalizes the notion of loopshaping for linear control system design. (AT )n−1 C T . γrT c the k’th channel of the output is n ⎫ i=1 � ⎭ yk (t) = (γkT γi )e�i t (wi ν). but the link to multivariable control is not too diﬃcult. · · · . then the compensator must be computed. First. The usual test for system observability requires computation of the observability matrix: Mo = C T . � � (199) 18 18. We say that a c v system is observable if every mode can be seen in at least one output channel.

Let F (s) = 1 + P (s)C(s). It is a reasonable constraint because otherwise the loop transfer function could pass signals with inﬁnitely high frequency. the path encloses the entire right-half plane. where the underline and overline denote the denominator and numerator. The heart of the Nyquist analysis is the mapping theorem. which answers the following question: How do paths in the s-plane map into paths in the F -plane? We limit ourselves to closed. Namely. . 18. and the remarkable result of the mapping theorem is Every zero of F (s) enclosed in the s-plane generates exactly one CW encirclement of the origin in the F (s)-plane. This method for design involves plotting the complex loci of P (s)C(s) for the range � = [−∗. The Nyquist criterion allows us to assess the stability properties of a system based on P (s)C(s) only. and then proceeding to the right at an arbitrarily large radius. stability requires that the number of unstable poles in F (s) is equal to the number of CCW encirclements of the origin. the relation is often written Z − P = CW . i.e. (200) In words. whose roots are the poles of the closed-loop system. the equivalent of low-pass ﬁltering the error signal. or P (s)C(s). stability requires that there are no zeros of F (s) in the right-half s-plane. is 1 + P (s)C(s) = 0. back to the negative imaginary axis. every pole of F (s) enclosed in the s-plane generates exactly one CCW encirclement of the origin in the F (s)-plane. equivalent to P (s)C(s) + P (s)C(s) = 0. Since the zeros of F (s) are in fact the poles of the closed-loop transfer function.. and C(s) the compensator.g. e.2 Roots of Stability – Nyquist Criterion 87 S(s) = e(s) 1 = . ∗].18. r(s) 1 + P (s)C(s) where P (s) represents the plant transfer function. 18. this constraint can be easily met by adding a high-frequency rolloﬀ to the compensator. The closedloop characteristic equation. and in this sense the design approach is quite diﬀerent from the root-locus method (see Ogata). This leads to a slightly shorter form of the above relation: P = CCW..1 Mapping Theorem We impose a reasonable assumption from the outset: The number of poles in P (s)C(s) exceeds the number of zeros. S(s).2. Conversely. The trick now is to make the trajectory in the s-plane enclose all unstable poles.2. respectively. clockwise(CW) paths in the s-plane. as s sweeps around the entire right-half s-plane.2 Nyquist Criterion The Nyquist criterion now follows from one translation. moving up the imaginary axis. In the case of a PID controller (two zeros) and a second-order zero-less plant. Since CW and CCW encir clements of the origin may cancel. There is no explicit calculation of the closed-loop poles. encirclements of the origin by F (s) are equivalent to encirclements of the point (−1 + 0j) by F (s) − 1.

we should count the encirclements for the function [det(I + P (s)C(s)) − 1] instead of P (s)C(s). Rules of thumb are as follows. then the loci of P (s)C(s) must not encircle the critical point at all. kg √ 0. the procedure of looking at individual Nyquist plots for each element of a transfer matrix is unreliable and outdated. Referring to the multivariable deﬁnition of S(s). the deﬁnition of kg is diﬀerent for stable and unstable P (s)C(s). • Because the path taken in the s-plane includes negative frequencies (i.. for an unstable plant. the nega tive imaginary axis). . For a stable plant.5 and ρ → 30→ .3 Robustness on the Nyquist Plot The question of robustness in the presence of modelling errors is central to control system design. The loci need to stay away from the critical point. • For the multivariable (MIMO) case.e.2. These notions lead to deﬁnition of the gain margin kg and phase margin ρ for a design. how close the loci come to it can be expressed in terms of magnitude and angle. instead of those of F (s): P = CCW ⇒∀ closed-loop stability (201) This is in fact the complete Nyquist criterion for stability. its angle should not be −180→ . kg → 2 and ρ → 30→ .88 18 CONTROL SYSTEMS – LOOPSHAPING Then the stability criterion can be cast in terms of the unstable poles of P (s)C(s). each having a very clear graphical representation. P (s)C(s) always decays such that the loci go to the origin. The use of gain and phase margin in design is similar to the SISO case. There are several details to keep in mind when making Nyquist plots: • If neither the plant nor the controller have unstable modes. It is a necessary and suﬃcient condition that the number of unstable poles in the loop transfer function P (s)C(s) must be matched by an equal number of CCW encirclements of the critical point (−1 + 0j). the loci of P (s)C(s) occur as complex conjugates – the plot is symmetric about the real axis. 18. the magnitude |P (s)C(s)| should not be near one. • When the magnitude |P (s)C(s)| = 1. As the ﬁgure shows. • When the angle of P (s)C(s) is −180→ . There are two natural measures of robustness for the Nyquist plot. • The requirement that the number of poles in P (s)C(s) exceeds the number of zeros means that at high frequencies.

To force S(s) to be small at low �. 18.3 Design for Nominal Performance Performance requirements of a feedback controller. using the nominal plant model. but higher-frequency components in the response may be hard to capture or even excite repeatably. The requirement |W1 (s)S(s)| < 1 is equivalent to |W1 (s)| < |1 + P (s)C(s)|.01. For example. and this latter condition can be interpreted as: The loci of P (s)C(s) must stay outside the disk of radius W1 (s). ⊗� < η. W1 (s) should be large in the same range. One of the most useful descriptions of model uncertainty is the multiplicative uncertainty: ˜ P (s) = (1 + �(s)W2 (s))P (s). dy .4 Design for Robustness It is a ubiquitous observation that models of plants degrade with increasing frequency.18. Higher-frequency behavior may have more nonlinear properties as well. Since the sensitivity function maps reference input r(s) to tracking error e(s). possibly inﬁnitely large. The disk is to be quite large. can be cast in terms of the Nyquist plot. if one-percent tracking is to be maintained for all frequencies below � = η. we know that |S(s)| should be small at low frequencies. then |S(s)| < 0. the DC gain and slow. which is to be centered on the critical point (−1+0j). (202) where W1 (s) is a stable weighting function of frequency. For example. at the lower frequencies.3 Design for Nominal Performance 89 1/kg Im(s) P(s)C(s) Im(s) Re(s) Re(s) γ γ 1/kg P(s)C(s) Stable P(s)C(s) Unstable P(s)C(s) 18. The eﬀects of modeling uncertainty can be considered to enter the nominal feedback system as a disturbance at the plant output. (203) . We restrict the discussion to the scalar case in the following sections. This can be formalized by writing |W1 (s)S(s)| < 1. lightly-damped modes or zeros are easy to observe.

The path of the perturbed plant could be anywhere on a disk of radius |W2 (s)P (s)C(s)|. We must therefore have |P C | < 1/|W2 |. should not intersect the disk of radius |W2 P C |. suﬃce to let |P ˜ Example: Let P = k/(s − 1). We need to create a nominal model P = k0 /(s − 1).5 Robust Performance The condition for good performance with plant uncertainty is a combination of the above two conditions. which will not vary with frequency in this case. for robustness. where k is in the range 2–5. perturbed plant. Two equations can be written using the above estimate. since |P C | is large. W2 (s) should be growing with increasing frequency. For constructing the Nyquist plot. as follows: 1. The last inequality is thus a condition for robust stability in the presence of multiplicative uncertainty parametrized with W2 . where �(s) is an unknown but stable function of frequency for which |�(s)| √ 1. W2 (s) should not grow any faster than necessary. ˜ and P (s) is the actual. W2 T | � |W2 P C |. The robustness condition is that this disk should not intersect the critical point. and W2 = 3/7. yielding k0 = 7/2. with radius |W1 |.90 18 CONTROL SYSTEMS – LOOPSHAPING Here. the weight can be estimated as follows: since P /P − 1 = �W2 . since it will turn out to be at the cost of nominal performance. for the two extreme values of k. (205) The robust performance requirement is related to the magnitude |P C | at diﬀerent frequen cies. . (204) where T is the complementary sensitivity function. centered on the nominal loci P (s)C(s). This leads directly to the performance condition |P C| > |W1 | in this range. we observe that ˜ P (s)C(s) = (1 + �(s)W2 (s))P (s)C(s). At low frequency. ˜ In the scalar case. since |P C | is small. 2. P (s) represents the nominal plant model used in the design of the control loop. Graphically. However. centered on the nominal locus P C. At high frequency. The weighting function W2 (s) scales �(s) with frequency. �(s)W2 (s)P (s). with the smallest possible value of W2 . This is met if |W1 S | + |W2 T | < 1. it will ˜ /P − 1| < |W2 |. the disk at the critical point. This can be written as |1 + P C | > |W2 P C | ⇒∀ |W 2 P C | ⇒∀ 1 > |1 + P C | 1 > |W2 T |. The perturbation is of the multiplicative type. |W1 S | � |W1 /P C |. since the uncertainty grows. 18.

6 Implications of Bode’s Integral The loop transfer function P C cannot roll oﬀ too rapidly in the crossover region.7 The Recipe for Loopshaping In the above analysis. it suﬃces to just pick C = L/P. 18. n Example: Suppose H(s) = �0 /sn . and may be second-order (−40dB/decade) if an adequate phase angle can be maintained near crossover. (208) This extraordinarily simple step involves a plant inversion. In the general case. n and no zeros. equivalent to a single pole).6 Implications of Bode’s Integral 91 18. and ln|H | = −nln(�/�0 ). the critical component. Each zero at the origin adds 90→ phase lead. For clarity. The integral is hence taken over the log of a frequency normalized with �0 . It follows that |H | = �0 /� n . it is a simple function with n poles at the origin. if the slope is large. the angle is large. we have extensively described what the open loop transfer function P C should look like. It is not hard to see how the integral controls the angle: the function ln(coth(|λ |/2)) is nonzero only near λ = 0. 2 (207) This integral is trivial to look up or compute. to meet robustness and performance speciﬁcations. dλ (206) where λ = ln(�/�0 ). which in turn degrades the phase margin. In the immediate neighborhood of these poles and zeros. Thus. i. L = P C.e. We have said very little about how to get the compensator C. We will use concepts from optimal linear control for the MIMO case. .18. Then we have just angle(H) = − n β � √ −√ ln(coth(|λ|/2))dλ = − nβ . each pole not at the origin induces 90→ of phase loss for frequencies above the pole. and the phase margin can be easily lost if the slope of P C at crossover (where the magnitude is unity) is too steep. Each pole at the origin clearly induces 90→ of phase loss. the crucial relation is 1 angle(H(j�0 )) = β � √ −√ d (ln|H(j�)) · ln(coth(|λ |/2))dλ. and then divide through by the plant. The slope can safely be ﬁrst-order (−20dB/decade. so that d(ln|H |)/dλ = −n. but in the scalar case. implying that the angle depends only on the local slope d(ln|H |)/dλ. To see this requires a short foray into Bode’s integral. while zeros not at the origin at 90→ of phase lead for frequencies above the zero. The Nyquist loci are clearly susceptible to these variations is phase. the phase may vary signiﬁcantly with frequency.. �0 is a ﬁxed constant. The overall idea is to ﬁrst shape L as a stable transfer function meeting the requirements of stability and robustness. The simple reason is that a steep slope induces a large phase loss. For a transfer function H(s). let the designed loop transfer function be renamed.

optimal control solutions provide an automated design procedure – we have only to decide what ﬁgure of merit to use. • One caveat for the well-behaved plant is that lightly-damped poles or zeros should not be cancelled verbatim by the compensator. The linear quadratic regulator (LQR) is a well-known design technique that provides practical feedback gains. or modal. the division can be made directly. including a direct extension of our scalar condition θ(W1 S) + θ(W2 T ) < 1. The condition for MIMO robust performance can be written in many ways. i. In the case of unstable poles.92 19 LINEAR QUADRATIC REGULATOR • When the plant is stable and has stable zeros (minimum-phase). which are characterized by transfer matrices instead of transfer functions. In the case of unstable zeros. the troublesome zeros must be faster than the closed-loop frequency response.1 LINEAR QUADRATIC REGULATOR Introduction The simple form of loopshaping in scalar systems does not extend directly to multivariable (MIMO) plants. 19 19. through a higher damping ratio. according to some ﬁgure of merit. turn out to generate only stabilizing controllers for MIMO plants. The notion of optimality is closely tied to MIMO control system design. we ﬁnd that they impose an unavoidable frequency limit for the crossover.. In the following sections. We list below some basic properties of the singular value decomposition. Useful properties and relations for the singular value are found in the section MATH FACTS. In general. analysis. • Non-minimum phase or unstable behavior in the plant can usually be handled by performing the loopshaping for the closest stable model.e. controllers that are the best possible. we use the properties of optimal state estimation and control to perform the plant inversion for MIMO systems. Optimal controllers. because the closed-loop response will be sensitive to any slight change in the resonant frequency. In this sense. . the ideas from scalar loopshaping can be adapted using the singular value. When a control system involves multiple inputs and outputs. which is analogous to an eigenvector. and then explicitly considering the eﬀects of adding the unstable parts. (209) The open-loop transfer matrix L should be shaped accordingly. the converse is true: The feedback system must be faster than the corresponding frequency of the unstable mode. The usual procedure is to widen the notch or pole in the compensator.

˙ (210) xdesired represents the vector of desired states. The closed-loop system has exactly as many outputs as inputs: n. 19.3 The Maximum Principle First we develop a general procedure for solving optimal control problems. suitably-damped locations in the complex plane. implemented as u = −K(x−xdesired ). tf ) is the terminal cost. since it multiplies f − x = 0. the total cost J is a sum of the terminal cost and an integral along the way. The last term above can be evaluated using integration by parts as − and thus � tf to � Lx ζx + Lu ζu + ηT fx ζx + ηT fu ζu − ηT ζx dt. we assume the plant to be written in state-space form x = Ax + Bu. We write the variation as ¯ ζJ = ωx ζx(tf ) + � tf to where subscripts denote partial derivatives. t) ˙ x(to ) = xo where ω(x(tf ). The “A-matrix” of the closed loop system is (A − BK). t)dt (211) (212) (213) subject to x = f (x(t). u(t). The column dimension of B equals the number of channels available in u.2 Full-State Feedback For the derivation of the linear quadratic regulator.19. η(t) can be anything we choose. u(t).2 Full-State Feedback 93 19. and t. and that all of the n states x are available for the controller. We assume that L(x(t). This follows from the fact that J is trajectory. The ﬁrst step is to augment the cost using the costate vector η(t). ¯ J = ω(x(tf )) + � tf to (L + ηT (f − x))dt ˙ (214) Clearly. The system dynamics are then written as: x = (A − BK)x + BKxdesired . Pole-placement is the process of placing the poles of (A − BK) in stable. Along the optimum ˙ ¯ should vanish. variations in J and hence J chosen to be continuous in x. We begin with the statement of the problem for a ﬁxed end time tf : choose u(t) to minimize J = ω(x(tf )) + � tf to L(x(t). (216) . and must match the row dimension of K. ˙ The feedback gain is a matrix K. u(t). t) is nonnegative. ˙ � (215) ηT ζ xdt = −ηT (tf )ζx(tf ) + ηT (to )ζx(to ) + ˙ � tf to ˙ ηT ζxdt. u. and serves as the external input to the closedloop system. and the “B-matrix” of the closed-loop system is BK. using the calculus of variations.

5. and the propogate the costate backward in time from tf to to .it is a ﬁxed value. and the simplest approach is the gradient method. t) forward in time to create a state trajectory. 4. Now the last term is zero.94 19 LINEAR QUADRATIC REGULATOR ¯ ζJ = ωx (x(tf ))ζx(tf ) + � tf to � tf to (Lu + ηT fu )ζudt + (217) ˙ (Lx + ηT fx + ηT )ζxdt − ηT (tf )ζx(tf ) + ηT (to )ζx(to ). while the costate propogates backward. The second and third requirements are met by explicitly setting ˙ ηT = −Lx − ηT fx ηT (tf ) = ωx (x(tf )). Evaluate ωx (x(tf )). Go back to step 2 and repeat loop until solution has converged. where K is a positive scalar or a positive deﬁnite matrix in the case of multiple input channels. Propogate x = f (x. This way of writing J makes it clear that there are three components of the variation that must independently be zero (since we can vary any of x. It is outlined as follows: 1. u. since we cannot vary the initial condition of the state by changing ¯ something later in time . (221) (222) (218) (219) (220) The evolution of η is given in reverse time. 19. u. pick a control history u(t). or x(tf )): Lu + ηT fu = 0 ˙ Lx + ηT fx + ηT = 0 ωx (x(tf )) − ηT (tf ) = 0.4 Gradient Method Solution for the General Case Numerical solutions to the general problem are iterative. For a given xo . . The state and costate are coordinated through the above equations. Hence we see the primary diﬃculty of solving optimal control problems: the state propogates forward in time. using Equation 221. 6. 2. ˙ 3. from a ﬁnal state to the initial. Let u = u + ζu. choose ζu = −K(Lu + ηT fu ). At each time step.

19. (224) (225) (226) (227) x ˙ x(to ) ˙ η = −Qx − AT η η(tf ) Ru + B T η = Ax + Bu = xo = 0 = 0. The choice of ζu in step 4 then achieves ζJ < 0. we set ω = 0. The major diﬃculties are computational expense. and then using the x equation. and a substitution for u. (233) This has to hold for all x. In the case of linear plant dynamics also. 2 2 (223) where the requirement that L → 0 implies that both Q and R are positive deﬁnite. The gradient method is quite popular in applications and for complex problems may be the only way to eﬀectively develop a solution. and 1 1 L = xT Qx + uT Ru. Inserting this into the η equation. unless ζu = 0.5 LQR Solution In the case of the Linear Quadratic Regulator (with zero terminal cost). in which case the problem is solved. we have Lx Lu fx fu so that = = = = xT Q uT R A B. we try a connection η = P x. ¯ and η is computed directly from x and x(tf ). the matrix Riccati equation.19. (234) . we obtain ˙ ˙ P Ax + AT P x + Qx − P BR−1 B T P x + P = 0. All of ζJ in Equation 217 except the integral ¯ with ζu is therefore eliminated explicitly. (228) (229) (230) (231) (232) ˙ Since the systems are clearly linear.5 LQR Solution 95 The ﬁrst three steps are consistent in the sense that x is computed directly from x(to and u. and the requirement of having a reasonable control trajectory to begin. The steady-state solution is given satisﬁes P A + AT P + Q − P BR−1 B T P = 0. so in fact it is a matrix equation.

the order of the closed-loop system is the same as that of the plant. In this case. The equation Ru + B T η = 0 gives the feedback law: u = −R−1 B T P x. An angle 90→ /n separates the most lightly-damped poles from the imaginary axis. The MARE is easily solved by standard numerical tools in linear algebra. and there is no penalty for using large u. and the auxiliary matrix Q∗ weights the plant output. The angular separation of n closed-loop poles on the arc is constant. since y = Cx. however. but the output variables y. we set the weighting matrix Q = C T Q∗ C. whose solution P is needed to compute the optimal feedback gain K. and equal to 180 → /n. This is in good agreement with the practical guidelines for control system design. the gain will indeed go to zero. and at the mirror images of the unstable plant zeros. When R << C T Q∗ C. The LQR also guarantees phase margin ρ → 60 degrees. This part is akin to the high-gain limiting case of the root locus. The optimal control must always stabilize the closed-loop system. Behavior of Closed-Loop Poles: Cheap Control. (235) 19. The LQR achieves inﬁnite gain margin: kg = ∗. First. the cost function is dominated by the output errors y. The Butterworth pattern refers to an arc in the stable left-half plane. In many cases. implying that the loci of (P C) (scalar case) or (det(I +P C)−1) (MIMO case) approach the origin along the imaginary axis. There are two groups of closed-loop poles. so there should be some account made for unstable plant poles. The expensive control solution puts stable closed-loop poles at the mirror images of the unstable plant poles. whose radius increases to inﬁnity as R becomes smaller and smaller. In the case of a completely stable plant. which is not a dynamical system.6 Optimal Full-State Feedback This equation is the matrix algebraic Riccati equation (MARE). Behavior of Closed-Loop Poles: Expensive Control. The remaining poles assume a Butterworth pattern. as shown in the ﬁgure. When R >> C T Q∗ C. so that the closed-loop poles approach the open-loop plant poles in a manner consistent with the scalar root locus. Output Variables. Robustness.7 Properties and Use of the LQR Static Gain. The LQR generates a static gain matrix K. Hence. the cost function is dominated by the control eﬀort u. poles are placed at stable plant zeros.96 19 LINEAR QUADRATIC REGULATOR 19. . and so the controller minimizes the control action itself. it is not the states x which are to be minimized.

giving (with constant P based on the design system) ˙ V = 2xT P x ˙ T = 2x P (Ax − B ∗ R−1 B T Kx) = xT (−Q + P BR−1 B T P )x − 2xT KB ∗ R−1 B T P x. Ni. Since we require V < 0 for x ≥= 0. . Let B = BN . we see that the LQR provides for one-half gain reduction. then it will “settle.i > 1/2. The above equation becomes Q − P B(I − 2N )R−1 B T P > 0. it is like the energy of a physical system. we have actual gain B ∗ . continuous. 2 It can be shown that P is positive deﬁnite. (240) −1 T ∗ This is satisﬁed if. where N is a diagonal matrix of random.” It suﬃces to ﬁnd one Liapunov function to show that a system is stable. We need an intermediate tool (which is very useful in its own right) to prove it: the Liapunov function.8 Proof of the Gain and Phase Margins 97 X φ /2 φ n=4 φ = 45 degrees X φ φ X φ /2 X 19. pick V (x) = 1 xT P x. There is an intuitive feel to the Liapunov function. in the case of the LQR control. and inﬁnite gain ampliﬁcation. unknown gains on the control channels. then the origin is asymptotically stable. for all channels. Now. we must have Q + P BG − 2P (B − B ∗ )G > 0. Thus. If it can be shown that energy is leaving such a system. If this function is always decreasing. (239) where G = R B P . suppose that instead of the design gain B.19. in all channels. Consider a scalar. positive deﬁnite function of the state V (x). (236) (237) (238) ˙ where we used the Riccati equation in a substitution.8 Proof of the Gain and Phase Margins The above stated properties of the LQR control are not diﬃcult to prove. except at the origin.

2 Problem Statement We consider the state-space plant model given by: x = Ax + Bu + W1 ˙ y = Cx + W2 . (242) There are n states.1 KALMAN FILTER Introduction In the previous section. B is n × m.i = ejδi . the Kalman Filter designs an optimal H. This is true for all |δi | < 60→ . y C Bu + + A y + H + x 1/s x 20. since C x = y . The plant subject to two random input signals. 20 20. H operates on the estimation error mapped to the plant output y.98 20 KALMAN FILTER The phase margin is seen (roughly) by allowing N to be a diagonal matrix of transfer functions Ni. and may not allow reconstruction of all the states. In real applications. This state estimation is the task of a model-based estimator having the form: ˙ x = Aˆ + Bu + H(y − C x) ˆ x ˆ (241) ˆ The vector x represents the state estimate. Given ˆ ˆ statistical properties of real plant disturbances and sensor noise. and hence sixty degrees of phase margin is provided in all channels. and C is l × n. W1 represents . and a correction term with the estimator gain matrix H. W1 and W2 . whose evolution is governed by the nominal A and B matrices of the plant. so that A has dimension n × n. The inherent assumption was that each state was known perfectly. m inputs. we require that the real part of Ni. and l outputs. the measurements are subject to disturbances. In this case.i is greater than onehalf. we derived the linear quadratic regulator as an optimal solution for the full-state feedback control problem.

The Kalman ﬁlter design provides H that minimizes the scalar cost function J = E(eT W e).˙ 20. The second term above.3 ˙ Step 1: An Equation for � The evolution of Γ follows from some algebra and the convolution form of e(t). The estimation error may be deﬁned as e = x − x. which corrupts ˙ the measurement y. V1 is an n × n diagonal matrix of intensities. We begin with ˙ Γ = E(eeT + eeT ) ˙ ˙ = E[(A − HC)eeT + (W1 − HW2 )eT + eeT (AT − C T H T ) + T T e(W1 − W2 H T )]. An important assumption of the Kalman Filter is that W1 and W2 are each vectors of unbiased. the symmetric error covariance. (243) (244) (245) Here ζ(t) represents the impulse (or delta) function. A related matrix. is deﬁned as Γ = E(eeT ). The last term above can be expanded. independent white noise. if E(·) denotes the expected value. W1 + HW2 is considered an external input. since it drives x directly. W2 denotes sensor noise. (246) The eigenvalues of the matrix A − HC thus determine the stability properties of the es timation error dynamics. and V2 is an l × l diagonal matrix of intensities. Hence. There are two main steps for deriving the optimal gain H. E(W1 (t1 )W1 (t2 )T ) = V1 ζ(t1 − t2 ) E(W2 (t1 )W2 (t2 )T ) = V2 ζ(t1 − t2 ) E(W1 (t)W2 (t)T ) = 0n×l . using the property that (249) . (247) where W is an unspeciﬁed symmetric. and that all the n + l channels are uncorrelated. It can then be veriﬁed that ˆ e = [Ax + Bu + W1 ] − [Aˆ + Bu + H(y − C x)] ˙ x ˆ = (A − HC)e + (W1 − HW2 ). (248) 20. positive deﬁnite weighting matrix.3 Step 1: An Equation for Γ 99 disturbances to the plant.

20. It is independent of the initial condition e(0). 2 i. and the ﬁnal result follows from � t 0 1 ζ(t − φ )dφ = .4 Step 2: H as a Function of � We now make the connection between Γ = E(eeT ) (a matrix) and J = E(eT W e) (a scalar). the written integral includes only half of the impulse. and it can be veriﬁed that J = E(eT W e) = trace(ΓW ). (250) This equation governs propagation of the error covariance.e. and therefore appears in Equa tion 249 twice. and depends on the (as yet) unknown estimator gain matrix H. (252) (251) .. The fact that W1 and HW2 are uncorrelated leads to the third line. leading to ˙ Γ = (A − HC)Γ + Γ(AT − C T H T ) + V1 + HV2 H T . T T The ﬁnal expression for E(e(W1 − W2 H T )) is symmetric. The trace of a matrix is the sum of its diagonal elements. 2 2 e(A−HC)(t−φ ) (V1 ζ(t − φ ) + HV2 H T ζ(t − φ )) dφ To get from the ﬁrst right-hand side to the second. we note that the initial condition e(0) is T T uncorrelated with W1 − W2 H T .100 20 KALMAN FILTER e(t) = e(A−HC)t e(0) + We have � t 0 e(A−HC)(t−φ ) (W1 (φ ) − HW2 (φ )) dφ. T T T T E(e(W1 − W2 H T )) = e(A−HC)t E(e(0)(W1 − W2 H T )) + � t 0 t 0 t 0 e(A−HC)(t−φ ) E((W1 (φ ) − HW2 (φ )) T T (W1 (t) − W2 (t)H T )) dφ = � � e(A−HC)(t−φ ) E((W1 (φ ) − HW2 (φ )) T T (W1 (t) − W2 (t)H T )) dφ = = 1 1 V1 + HV2 H T . We now introduce an auxiliary cost function deﬁned as J ∗ = trace(ΓW + ΦF ).

Equation 250: ˙ J ∗ = trace ΓW + Φ(−Γ + AΓ − HCΓ + ΓAT − ΓC T H T + V1 + HV2 H T ) � � (253) If J ∗ is an optimal cost. V1 .5 Properties of the Solution 101 where F is an n×n matrix of zeros. The evolution of Γ is always stable. Lagrange multipliers provide an ingenious mechanism for drawing constraints into the optimization.. (254) Hence the estimator gain matrix H can be written as a function of Γ. leading to a more common MARE form of Equation 255: 0 = AΓ + ΓAT + V1 − ΓC T V2−1 CΓ. �H Proofs of the ﬁrst two are given at the end of this section. the last lemma uses the chain rule. the correct choice of H achieves an extremal value. Notice also that the result is independent of the weighting matrix W . so minimizing J ∗ solves the same problem. and Φ is an n×n matrix of unknown Lagrange multipliers. which give the derivative of a trace with respect to a constituent matrix: � trace(−ΦHCΓ) = −ΦT ΓC T �H � trace(−ΦΓC T H T ) = −ΦΓC T �H � trace(ΦHV2 H T ) = ΦT HV2 + ΦHV2 . which minimizes the squared-norm of the estimation error.e. 20. (256) . (255) Equations 254 and 255 represent the practical solution to the Kalman ﬁltering problem. and the previous two lemmas. like the LQR. it follows that �J ∗ /�H = 0. satisﬁed if H = ΓC T V2−1 . Next. This is in fact the case. the constraint we invoke is the evolution of Γ. We need the following lemmas. and the same analysis and numerical tools can be applied to both methodologies. V2 ]. we enforce Φ = ΦT . suggesting a duality with the LQR problem. J ∗ = J. and depends only on the constant matrices [A.20. Inserting this back into Equation 250. Note that since F is zero. Then the condition �J ∗ /�H = 0 leads to 0 = 2Φ(−ΓC T + HV2 ). since the values are arbitrary. The steady-state solution for Γ is valid for time-invariant systems. we obtain ˙ Γ = AΓ + ΓAT + V1 − ΓC T V2−1 CΓ. C. i. which might as well be the identity.5 Properties of the Solution The solution involves a matrix Riccati equation.

Some closed-loop poles go to the stable plant zeros.6 Combination of LQR and KF An optimal output feedback controller is created through the use of a Kalman ﬁlter coupled with an LQR full-state feedback gain. weighting the model dynamics Aˆ + Bu more. • V2 >> V1 : poor sensors. and Γ is the error covariance matrix. V2 > 0 Q → 0. placing emphasis on the corrective action of the ﬁlter. The remaining poles follow a Butterworth pattern whose radius increases with increasing V1 /V2 . x • A large disturbance V1 . The qualitative dependence of the estimator gain H = ΓC T V2−1 on the other parameters can be easily seen. 20. Recall that V1 is the intensity matrix of the plant disturbance. and inﬁnite gain margin. Closed-loop poles go to the stable plant poles. or the mirror image of unstable plant zeros. small disturbance. The limiting closed-loop poles of the Kalman ﬁlter are similar. This combination is usually known as the Linear Quadratic Gaussian design. when the plant is state-controllable. as long as the plant is fully state-observable. and the mirror images of the unstable plant poles. like the LQR. so that the ﬁlter’s correction is dominant. for all the channels together or independently. and large sensor noise V2 creates a small H. dual to expensive-control problem. • A small disturbance V1 .102 20 KALMAN FILTER Duality of Linear Quadratic Regulator and Kalman Filter Linear Quadratic Regulator Kalman Filter x = Ax + Bu ˙ x = Ax + Bu + W1 ˙ y = Cx + W2 ˙ u = −Kx x = Aˆ + Bu + H(y − C x) ˆ x ˆ �√ T T 2J = 0 (x Qx + u Ru)dt J = E(eT W e) V1 → 0. For the plant given as . H small. Furthermore. large disturbance. H >> 1. and dual to those of the LQR: • V2 << V1 : good sensors. dual to cheap-control problem. R > 0 −1 T H = ΓC T V2−1 K=R B P P A + AT P + Q − P BR−1 B T P = 0 ΓAT + AΓ + V1 − ΓC T V2−1 CΓ = 0 The Kalman Filter is guaranteed to create a stable nominal dynamics A − HC. This is dual to the stability guarantee of the LQR loop. or LQG. V2 is the intensity of the sensor noise. • A large uncertainty Γ creates large H. the KF loop achieves 60 → phase margin. and small sensor noise V2 creates a large H.

We use the output error as a control input in the next section. .7. we put the Kalman Filter and controller gain G together as follows: ˙ x = Aˆ + Bu + H(y − C x) ˆ x ˆ ˆ u = −K x. In this case. x no longer represents an estimated state. See proof below. Output Tracking: This compensator is a stand-alone system that. tries to drive its input y to zero.7 20. It can be hooked up to receive tracking error e(s) = r(s)−y(s) as an input instead. on loopshaping via loop transfer recovery.1 Proofs of the Intermediate Results Proof that E(eT W e) = trace(ΓW ) E(eT W e) = E � = � n ⎫ i=1 n ⎫ ⎫ n i=1 j=1 ei � � j=1 n ⎫ Wij ej ⎪⎪ ⎩⎩ Γij Wji . 2. as written. Separation Principle: The eigenvalues of the nominal closed-loop system are made of up the eigenvalues of (A − HC) and the eigenvalues of (A − BK). separately. but rather an estimated state tracking ˆ error. so that it is not limited to the regulation problem alone. 20. (257) (258) C(s) B y + H + + C φ = (sI-A) -1 φ -K u There are two central points to this construction: 1.20.7 Proofs of the Intermediate Results 103 x = Ax + Bu + W1 ˙ y = Cx + W2 .

7. the il∗ th element trace(AH T ) = trace ⎝ = trace ⎝ = n n ⎫⎫ i=1 j=1 � j=1 n ⎫ � n ⎫ ⎭ j=1 Aij Hlj ⎠ ⎭ Aij Hij . k=1 where the second form is a sum over i of the ii’th elements. the il∗ th element ⎭ Aij Hjk Bki . Now consider the diagonal elements of the product ΓW : Γ11 W11 + Γ12 W21 + · · · · · ⎛ · Γ21 W12 + Γ22 W22 + · · · · ⎞ ∀ ΓW = ⎝ ⎠ · · ··· trace(ΓW ) = n n ⎫⎫ � ⎭ Γij Wji . � i=1 j=1 20.3 Proof that � trace(−ΦΓC T H T ) �H n ⎫ i=1 Aijo Bko i (BA)ko jo (BA)To ko −∀ j (BA)T AT B T .104 20 KALMAN FILTER the transpose of W being valid since it is symmetric.2 � Proof that �H trace(−ΦHCΓ) = −ΦT ΓC T trace(AHB) = trace ⎝ = n n ⎫⎫ i=1 j=1 � j=1 l ⎫ n ⎫ Aij k=1 l ⎫ Hjk Bkl ⎠ . � = −ΦΓC T T Aij Hjl ⎠ . where the last form is a sum over the ii’th elements.7. Now � trace(AHB) = �Hjo ko = = � trace(AHB) = �H = 20. It follows that .

the closed-loop system evolves according to d dt � x x ˆ � = ⎬ −BK A HC A − BK − HC �� x ˆ x � . may yield good results. combining the LQR and KF into an output feedback compensator.105 � trace(AH T ) = Aio jo −∀ �Hio jo � trace(AH T ) = A. we can write the above in another form: ˆ d dt � x e � = ⎬ BK A − BK 0 A − HC �� x e � . There are two reasons to avoid this kind of direct LQG design procedure. although the LQR and KF each possess good robustness properties. First. there do exist plants for which there is no robustness guarantee for an LQG compensator. the determinant of an upper triangular block matrix is the product of the determinants of each block on the diagonal: det(sI − A∗ ) = det(sI − (A − BK))det(sI − (A − HC)). and hence the separation of eigenvalues follows. 21 21. which are at the root of feedback control. If the sensor noise and disturbance properties of the plant are indeed well-known. then an LQG design approach.4 Proof of the Separation Principle Without external inputs W1 and W2 . where δ(s) = (sI − A)−1 . however.1 LOOP TRANSFER RECOVERY Introduction The Linear Quadratic Regulator(LQR) and Kalman Filter (KF) provide practical solutions to the full-state feedback and state estimation problems. a second problem is that this design technique has no clear equivalent in frequency space.7. that is. Even if one could steer clear of such pathological cases. However. The KF has open-loop transfer function L(s) = Cδ(s)H. The LQR tuning matrices Q and R would be picked heuristically to give a reasonable closed-loop response. We now reconsider just the feedback loop of the Kalman ﬁlter. respectively. Using the deﬁnition of estimation error e = x − x. then its eigenvalues are the roots of det(sI − A ∗ ) = 0. � �H 20. This follows from the estimator evolution equation . If A∗ represents this compound A-matrix. It cannot be directly mapped to the intuitive ideas of loopshaping and the Nyquist plot.

First. for which W T W = I. speciﬁcally to perturbations at the output y . the KF loop is an ˆ ideal candidate for a loopshaping design. Supposing that we have an estimator gain H which creates an attractive loop function L(s).2 A Special Property of the LQR Solution Letting Q = C T C and R = πI. Secondly.106 21 LOOP TRANSFER RECOVERY y H φ (s) ^ x C ˙ x = Aˆ + Bu + H(y − C x) ˆ x ˆ and the ﬁgure. Note that we have not included the factor Bu as part of the ﬁgure. The condition θ(W1 (s)S(s)) + θ(W2 (s)T (s)) < 1 is suﬃcient for robust performance with multiplicative plant uncertainty at the output. we will show (roughly) that ≈ lim( πK) = W C. The procedure is termed Loop Transfer Recovery (LTR). so that the Kalman ﬁlter loop itself has good performance and robustness properties. one carries out a KF design for H. the KF loop has good robustness properties. . (259) It will turn out that the LQR can be set up so that the an LQG-type compensator achieves exactly this result. In short. LTR is useful as a SISO control technique. the KF loop has sensitivity func tion S(s) = (I + Cδ(s)H)−1 and complementary sensitivity T (s) = (I + Cδ(s)H)−1 Cδ(s)H. where I is the identity matrix. 21. or Cδ(s)BC(s) ⇐ Cδ(s)H. we would like to ﬁnd the compensator C(s) that establishes P (s)C(s) ⇐ Cδ(s)H. As noted previously. since it does not aﬀect the error dynamics of the ﬁlter. so that the LQG compensator satisﬁes the approximation of Equation 259. and has two main parts. but has a much larger role in multivariable control. First recall the gain and Riccati equations for the LQR: K = R−1 B T P 0 = Q + P A + AT P − P BR−1 B T P. and W is an orthonormal matrix. π�0 where K is the LQR gain matrix. In this regard. and further is amenable to output tracking. we pick suitable parameters of the LQR design.

This matrix may be absorbed into W through a matrix inverse.e.3 The Loop Transfer Recovery Result Now Q = C T C = C T W T W C = (W C)T W C. Finally. • The design plant has no unstable zeros. it must be the case that P ∀ 0 also. rows of K) is equal to the number of outputs (i. � Note that another orthonormal matrix W ∗ could be used in separating K T from K in the last line. and so does not need to be written. The Riccati equation becomes 107 0 = π(W C)T W C + πP A + πAT P − P BB T P = 0. rows of C). The proof of the LTR result depends on some easy lemmas..e. The LTR method is limited by two conditions: • The plant has an equal number of inputs and outputs. Instead. with the deﬁnitions δ(s) = (sI − A)−1 and X(s) = (δ−1 (s)+HC)−1 = (sI − A + HC)−1 . with W an orthonormal matrix.21. See Athans for more details and references on this topic.e. and so in this limit π(W C)T W C ⇐ = = = WC ⇐ P BB T P (B T P )T B T P (R−1 B T P )T RR(R−1 B T P ) π2 K T K −∀ ≈ πK. i. The LTR method can be in fact be applied in the presence of unstable plant zeros. but the recovery is not to the Kalman ﬁlter loop transfer function. however. .. we develop C(s). then the limiting LQG controller C(s) satisﬁes π�0 lim P (s)C(s) = Cδ(s)H. First. we note that the above property is true only for LQR designs with minimum-phase plants. 21.. The result of the last line establishes that the plant must be square: the number of inputs (i. the recovered function will exhibit reasonable limitations inherent to unstable zeros. given at the end of this section. those with only stable zeros (Kwakernaak and Sivan). In the limit as π ∀ 0.3 The Loop Transfer Recovery Result ≈ The theorem is stated as: If limπ�0 ( πK) = W C (the above result).

π�0 21. Finally it follows that limπ�0 P (s)C(s) = Cδ(s)H. Now we look at the product CX(s): C(SI − A + HC)−1 C(δ−1 (s) + HC)−1 . both prop erties of orthonormal matrices. ≈ πK : Next we invoke the result from the LQR design. as desired. Some speciﬁc techniques are useful. reintroduced into the limiting compensator. . we used the assumption that W is square and invertible. gives lim C(s) = ((I + Cδ(s)H)−1 Cδ(s)B)−1 (I + Cδ(s)H)−1 Cδ(s)H = (Cδ(s)B)−1 Cδ(s)H = P −1 (s)Cδ(s)H. by using the limiting cheap-control LQR design. then use Lemma 2 ∀ C(δ(s) − δ(s)H(I + Cδ(s)H)−1 Cδ(s)) (I − Cδ(s)H(I + Cδ(s)H)−1 )Cδ(s). then use Lemma 3 ∀ (I + Cδ(s)H)−1 Cδ(s). then use Lemma 3 ∀ (I + KX(s)B)−1 KX(s)H ≈ ≈ ≈ ( πI + πKX (s)B)−1 πKX (s)H. to eliminate lim C(s) = (W CX(s)B)−1 W CX(s)H = (CX(s)B)−1 CX(s)H. with π ∀ 0.108 21 LOOP TRANSFER RECOVERY C(s) = = = = = = = K(sI − A + BK + HC)−1 H K(X −1 (s) + BK)−1 H. then use Lemma 2 ∀ K(X(s) − X(s)B(I + KX(s)B)−1 KX(s))H KX(s)H − KX(s)B(I + KX(s)B)−1 KX(s)H (I − KX(s)B(I + KX(s)B)−1 )KX(s)H. CX(s) = = = = = This result. with Q = C T C and R = πI. The LQR design step is thus trivial. π�0 In the last expression.4 Usage of the Loop Transfer Recovery The idea of LTR is to “recover” a Kalman ﬁlter loop transfer function L(s) = Cδ(s)H.

and the compensator C(s) is constructed.5 Three Lemmas 109 • Scale the plant outputs (and references). For example. 21. Then. so that one unit of error in one channel is as undesirable as one unit of error in another channel. • Scale the plant inputs in the same way. • Design for crossover frequency. once the KF design is completed. for example. One Newton of propeller thrust cannot be compared with one radian of rudder angle.21. The bandwidth of the controller is roughly equal to the frequency at which the (recovered) loop transfer function crosses over 0dB. the plant input channels must be augmented with the necessary additional poles (at the origin). • Integrators should be part of the KF loop transfer function.5 Three Lemmas Lemma 1: Matrix Inversion (A + BCD)−1 = A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 . the integrators are moved from the plant over to the input side of the compensator. Often. Since the Kalman ﬁlter loop has only as many poles as the plant. the bandwidth of is a more intuitive design parameter than is. Proof: (A + BCD)(A + BCD)−1 = I A(A + BCD)−1 = I − BCD(A + BCD)−1 (A + BCD)−1 = A−1 − A−1 BCD(A + BCD)−1 = A−1 − A−1 BCD(I + A−1 BCD)−1 A−1 = A−1 − A−1 B(D −1 C −1 + A−1 B)−1 A−1 = A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 . Lemma 3 � . in depth and pitch control of a large submarine. the highfrequency multiplicative weighting W2 . Quantitative uncertainty models are usually at the cost of a lengthy identiﬁcation eﬀort. Lemma 2: Short Form of Lemma 1 (X −1 + BD)−1 = X − XB(I + DXB)−1 DX Proof: substitute A = X −1 and C = I into Lemma 1. one meter of depth error cannot be compared directly with one radian of pitch error. if no steady-state error is to be allowed. The tracking errors will accrue as desired.

in three-space. Here we use the second notation.1 22.1 APPENDIX 1: MATH FACTS Vectors Deﬁnition A vector has a dual deﬁnition: It is a segment of a a line with direction.1.e. a linear segment with a direction. The ﬁrst form is independent of any reference system. x is meant as a column vector. For example. a ⎧ ⎧ ⎣ 7 ⎨ z ⎡ ⎧ ⎢ ⎤ The elements of a vector have a graphical interpretation.. i.110 22 APPENDIX 1: MATH FACTS I − A(I + A)−1 = (I + A)−1 Proof: I − A(I + A)−1 = (I + A)(I + A)−1 − A(I + A)−1 = (I + A − A)(I + A)−1 = (I + A)−1 . which is particularly easy to see in two or three dimensions. � 22 22. or it consists of its projection on a reference system 0xyz.e. i. we write a vector in terms of its components with respect to a reference system as 2 ⎧ ⎥ γ = 1 . whose components are found as projections of an (invariant) directed segment on a speciﬁc reference system. 7 a 1 2 x y . usually orthogonal and right handed. whereas the second (in terms of its components) depends directly on the coordinate system. We use the overhead arrow to denote a column vector..

The total length of a vector of dimension m. i i=1 ||γ || = x This scalar is commonly used to normalize a vector to length one. Scalar multiplication: ⎡ ⎧ ⎢ ⎤ ⎧ ⎥ ⎡ ⎧ ⎢ ⎤ ⎧ ⎥ ⎧ ⎨ 2 −2 × ⎧ 1 ⎣ 7 22. 2. γ · γ = ||γ ||||γ cos χ. 1 + 3 = ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎣ ⎣ 9 ⎨ 7 ⎨ ⎣ 2 ⎨ ⎤ ⎡ ⎤ ⎡ ⎤ Graphically. x y x y|| where χ is the angle between the vectors.1. its Euclidean norm.2 Vector Magnitude −4 = ⎧ −2 ⎧ ⎨ ⎣ −14 . 22. addition is stringing the vectors together head to tail.3 Vector Dot or Inner Product The dot product of two vectors is a scalar equal to the sum of the products of the corre sponding components: m ⎫ i=1 γ ·γ =γ γ = x y x y The dot product also satisﬁes T xi y i . .1 Vectors 1. Vector addition: 111 γ +γ = γ a b c ⎡ ⎧ ⎢ ⎧ 5 ⎧ 2 ⎧ ⎧ 3 ⎧ ⎥ ⎢ ⎥ ⎥ ⎢ 4 .1. is given by ⎦ �m �⎫ � x2 .22.

(signed) magnitude is equal to ||γ y|| sin χ. rotating from γ to γ. j=1 where n is the number of columns in A. Matrices. � � � � = � � � γ ×γ x y 22. Note c that this multiplication is deﬁned only if γ has as many rows as A has columns. x y 3.1 Matrices Deﬁnition ˆ ˆ k � ⎧ (x2 y3 − x3 y2 )ˆ i i j ˆ � ⎧ ⎢ � � = (x3 y1 − x1 y3 )ˆ j x1 x2 x3 � ⎧ ⎧ ˆ y1 y2 y3 � ⎣ (x1 y2 − x2 y1 )k � ⎡ ⎤ ⎧ ⎧ ⎥ ⎧ ⎧ ⎨ . γ × γ = γ.2.1.4 Vector Cross Product 22 APPENDIX 1: MATH FACTS The cross product of two three-dimensional vectors γ and γ is another vector γ.2 Multiplying a Vector by a Matrix If Aij denotes the element of matrix A in the i’th row and the j’th column. A matrix. permit addition and scalar multiplication. a vector is a special case of a matrix with one column. a b 7 2 � ⎭ This matrix has three rows (m = 3) and two columns (n = 2). like vectors. where χ is the angle between the two x||||γ vectors.2. then the multi plication γ = Aγ is constructed as: c v n ⎫ ci = Ai1 v1 + Ai2 v2 + · · · + Ain vn = Aij vj . say 2 3 ⎛ ⎞ A = [γ γ ] = ⎝ 1 3 ⎠ . direction is normal to the plane formed by the other two vectors. We usually use an upper-case symbol to denote a matrix. magnitude is the area of the parallelogram formed by the two vectors – the cross product of two parallel vectors is zero – and 4.112 22. 22. x y In terms of their components. measured from γ to γ. or array. γ will have as many rows as A has rows (m).2 22. they have v . arranged side by side. is equivalent to a set of column vectors of the same dimension. direction is given by the right-hand rule. x y z x y z whose 1. 2.

e. NOT Commutative. so that each A:j is the j’th column vector. 22. 0 0 3 � ⎭ The product of a diagonal matrix with another diagonal matrix is diagonal. c v We are multiplying column vectors of A by the scalar elements of γ .22. A diagonal matrix is square. v and the proper number of columns. 1 0 0 ⎛ ⎞ = ⎝ 0 1 0 ⎠ .. Diagonal Matrices.2. The product γ A would be well-posed only if A had one row.2 Matrices 113 consistent inner dimension n. and comprises a square matrix with ones on the diagonal. 0 0 1 � ⎭ I3×3 The identity always satisﬁes AIn×n = Im×m A = A. Associative. where k is the number of columns in matrix B. v 22. (B + C)A = BA + CA. Matrix multiplication is: 1. 3.2. except in special cases. A(B + C) = AB + AC. (AB)C = A(BC). and zeros elsewhere. and in this case the operation is commutative.3 Multiplying a Matrix by a Matrix The multiplication C = AB is equivalent to a side-by-side arrangement of column vectors C:j = AB:j . There is another important interpretation of this vector multiplication: Let the subscript : indicate all rows. so that C = AB = [AB:1 AB:2 · · · AB:k ]. . Distributive. 2. and has all zeros oﬀ the diagonal. The same inner dimension condition applies as noted above: the number of columns in A must equal the number of rows in B. AB ≥= BA.4 Common Matrices Identity. the following is a diagonal matrix: 4 0 0 ⎞ ⎛ A = ⎝ 0 −2 0 ⎠ . For instance. The identity matrix is usually denoted I. Then γ = Aγ = A:1 v1 + A:2 v2 + · · · + A:n vn .g.

22. it is equivalent to “ﬂipping” the vector or matrix around the diagonal line. For example. The two-dimensional case is particularly easy to re member. where �kj is the determinant of the sub-matrix formed by neglecting the k’th row and the j’th column. A = ⎝ 4 5 ⎠ −∀ A = 2 5 9 8 9 1 ⎧ ⎥ 2 −∀ γ T = a ⎧ ⎧ ⎣ ⎨ 3 ⎭ ⎤ {1 2 3} (AB)T = B T AT .114 22.6 Determinant The determinant of a square matrix A is a scalar equal to the volume of the parallelepiped enclosed by the constituent vectors. in the sense that one could also target the k’th column: . indicated by a T superscript results from simply swap ping the row-column indices of each entry. and illustrates the principle of volume: det(A) = A11 A22 − A21 A12 �� 1 −1 = 1 + 1 = 2. The general formula allows one to pick a row k. perhaps the one containing the most zeros. and apply j=n ⎫ j=1 det(A) = Akj (−1)k+j �kj . The formula is symmetric. 1 1 y Area = det(A) = 2 A:2 -1 A:1 x 1 det �⎬ In higher dimensions.2. ⎡ ⎧ ⎢ � γ= a A very useful property of the transpose is � ⎬ 1 2 1 4 8 ⎞ ⎛ T .5 Transpose 22 APPENDIX 1: MATH FACTS The transpose of a vector or matrix.2. the determinant is more complicated to compute.

then the matrix is said to be nonsingular. then the inverse does not exist. Its computation requires the determinant above. in three dimensions.2.22. and this results from the fact that the constituent vectors do not span the matrix dimension.2. The inverse ﬁnds common application in solving systems of linear equations such as Aγ = γ −∀ γ = A−1γ x b x b. denoted A−1 . in two dimensions. Once this computation is made. a singular matrix has the vectors colinear. satisﬁes AA−1 = A−1 A = I.2. Note also that det(A) = det(AT ). x x . a singular matrix has all its vectors lying in a (two-dimensional) plane. and the following deﬁnition of the n × n adjoint matrix: (−1)1+1 �11 · · · (−1)1+n �1n ⎞ ⎛ ··· ··· ··· adj(A) = ⎝ ⎠ . If the determinant of a matrix is zero. For instance.. 22. det(A) � ⎭T A−1 = If A is singular. A typical eigenvalue problem is stated as Aγ = ηγ . 22.2 Matrices 115 det(A) = j=n ⎫ j=1 Ajk (−1)k+j �jk .e. If det(A) ≥= 0. the inverse follows from adj(A) .9 Eigenvalues and Eigenvectors Aii .8 Trace The trace of a square matrix is the sum of the diagonals: n ⎫ i=1 tr(A) = 22. det(A) = 0.7 Inverse The inverse of a square matrix A. then the matrix is said to be singular – there is no volume. n+1 n+n (−1) �n1 · · · (−1) �nn . i.

and scalars η (eigenvalues) will the equation x be satisﬁed. 4 −5 2 −3 � −∀ � Thus. η1 = −1 and η2 = 2. here is an example for the two-dimensional case: ⎬ ⎬ A = A − ηI = det(A − ηI) = = = 4−η −5 −∀ 2 −3 − η (4 − η)(−3 − η) + 10 η2 − η − 2 (η + 1)(η − 2). x 0. 2 29/29 . are also useful for many problems. Each is associated with a right eigenvector γ . Example: (AT − η1 I)γ1 = γ −∀ y 0 . They are often normalized to have unity magnitude. a unique direction for the eigenvector. The condition that rank(A − ηi I) = rank(A) − 1 indicates that there is only one eigenvector for the eigenvalue ηi . The above discussion relates only the right eigenvectors.. A and AT share the same eigenvalues η. and η is a scalar. x Eigenvectors are deﬁned only within an arbitrary constant. Since the above is equivalent to (A − ηI)γ = γ it is clear that det(A − ηI) = 0. Left eigenvectors. and positive ﬁrst element (as above). then there are multiple eigenvectors that go with ηi .e. If the left-hand side rank is less than this. deﬁned as γ T A = ηγ T . and can x y y be deﬁned simply as the right eigenvectors of AT . since they share the same determinant. x (A − η1 I)γ 1 = γ −∀ x 0 � 5 −5 γ 1 = γ −∀ x 0 2 −2 �≈ ≈ �T 2/2 2/2. if γ is an eigenvector then cγ x x is also an eigenvector for any c ≥= 0. This observation leads to the solutions for η. We ask x for what nonzero vectors γ (right eigenvectors). since the magnitude can be arbitrary. In this example. i. more precisely. γ is a column vector with n elements. A has two eigenvalues.116 22 APPENDIX 1: MATH FACTS where A is an n × n matrix. generated from the equation Aγ = x ηγ . γ1 = x ⎬ ⎬ (A − η2 I)γ 2 = γ −∀ x 0 � 2 −5 γ 2 = γ −∀ x 0 2 −5 � ≈ �T ≈ γ 2 = 5 29/29.

or W T = V −1 . we consider matrices that have unique eigenvectors for each eigenvalue. AV = V Φ −∀ A = V ΦW T = n ⎫ i=1 ηi γ i w i . construct a diagonal matrix containing the eigenvalues: � η1 0 · 0 ηn it follows that Φ=⎛ ⎝ ⎭ ⎞ ⎠.2. γ2 = y 22.2 Matrices ⎬ 117 5 2 −5 −2 � γ1 = γ −∀ y 0 γ1 = y � �T ≈ ≈ 2 29/29. − 2/2 .10 Modal Decomposition For simplicity. it can be shown that Ak has its eigenvalues at �k . is V = [γ 1 · · · γ n ] . The dot products xi y of a left eigenvector with the right eigenvectors corresponding to diﬀerent eigenvalues are zero. that is γ T γi = 1. respectively. y y then we have W T V = I. The right and left eigenvectors corresponding to a particular eigenvalue η can be deﬁned to have unity dot product. V and W .22. and keeps i the same eigenvectors as A. if the set of right and left eigenvectors. −5 29/29 ⎬ (AT − η2 I)γ2 = γ −∀ y 0 � 2 2 γ2 = γ −∀ y 0 −5 −5 �≈ ≈ �T 2/2. 2 . v γT Hence A can be written as a sum of modal components. Thus. Next.2 By carrying out successive multiplications. with the normalization noted above. and x x W = [γ1 · · · γn ] .

The singular value decomposition (SVD) computes three matrices satisfying G = U ΓV � .11 Singular Value 22 APPENDIX 1: MATH FACTS Let G be an m×n real or complex matrix. They are deﬁned explicitly: U is the matrix of right eigenvectors of GGH . ηmin (A� A).. The notation is common that θ1 = θ. Each θi is an eigenvalue of GH G (or GGH ). � 22. θi represents the Euclidean size of the matrix G along the i’th singular vector: θ = max||x||=1 ||Gx|| θ = min||x||=1 ||Gx||. and V is the matrix of right eigenvectors of GH G. • θ(A) = • θ(A) = � ηmax (A� A). the maximum singular value. where Γ1 = ⎛ 0 · 0 ⎞ . Other properties of the singular value include: • θ(AB) √ θ(A)θ(B).3 22. Like eigenvalues. positive 2 singular value. The star notation indicates a complex-conjugate transpose (the Hermitian of the matrix). they satisfy X � = X −1 . The auxiliary matrices U and V are unitary. Γ is m×n. where U is m×m. The signal y(t) has transform Y (s) deﬁned as follows: . • θ(A) = 1/θ(A−1 ).e.3.2. The matrix Γ has the form ⎬ � Γ= Γ1 0 0 0 θ1 0 0 . the singular values of G are related to projections. and θp = θ. n). ⎠ ⎝ 0 0 θp � ⎭ and p = min(m.1 Laplace Transform Deﬁnition The Laplace transform projects time-domain signals into a complex frequency-domain equiv alent. ordered such that θ1 > θ2 > · · · θp . i. and V is n×n. Each nonzero entry on the diagonal of matrix Γ1 is a real.118 22. the minimum singular value. • θ(A) = 1/θ(A−1 ).

The following table gives a list of some useful transform pairs and other properties. where s is a complex variable. This result from complex analysis holds that if two complex functions are equal on some arc (or line) in the complex plane. if y(t) = et . that the Laplace transform is deﬁned only within the region of convergence. the output is y(t) = g(t) ← x(t). then they are equivalent everywhere. properly constrained within a region so that the integral converges. 22. The last two properties are of special importance: for control system design. For instance. the convolution of two signals in the time domain corresponds with the multiplication of signals in the frequency domain.3. since the integrand is e1−s in this case. the function Y (s) is deﬁned for all s through analytic continuation. The other terms that arise will cancel if y(0) = 0. integration of a signal is equivalent to division by s.2 Convergence We note ﬁrst that the value of s aﬀects the convergence of the integral. Consider a system whose impulse response is g(t).3 Convolution Theorem One of the main points of the Laplace transform is the ease of dealing with dynamic systems.3 Laplace Transform 119 Y (s) = L(y(t)) = � √ 0 y(φ )e−sφ dφ. It should be noted however. Although the integral converges within a well-deﬁned region in the complex plane. the diﬀerenti ation of a signal is equivalent to multiplication of its Laplace transform by s. being driven by an input signal x(t).22. or if y(0) is ﬁnite. The Convolution Theorem is � t 0 y(t) = g(t − φ )x(φ )dφ �� Y (s) = G(s)X(s).3. Note that the Laplace transform is linear. for reference. As with the Fourier transform. 22. then the integral converges only for Re(s) > 1. and so it is distributive: L(x(t) + y(t)) = L(x(t)) + L(y(t)). Y (s) is a complex function as a result. Here’s the proof given by Siebert: Y (s) = = = = � √ 0 √ 0 √ 0 √ 0 y(t)e−st dt �� t � � � ��0√ 0 g(t − φ ) x(φ ) dφ e−st dt �� √ � g(t − φ ) h(t − φ ) x(φ ) dφ e−st dt 0 � x(φ ) g(t − φ ) h(t − φ ) e−st dt dφ � .

120 22 APPENDIX 1: MATH FACTS 1 1 1+ (be−at − ae−bt ) ab a−b ≈ 1− ≈ � y(t) ⇒∀ Y (s) (Impulse) ζ(t) ⇒∀ 1 1 (Unit Step) 1(t) ⇒∀ s 1 (Unit Ramp) t ⇒∀ 2 s 1 e−∂t ⇒∀ s+∂ � sin �t ⇒∀ 2 s + �2 s cos �t ⇒∀ 2 s + �2 � e−∂t sin �t ⇒∀ (s + ∂)2 + � 2 s+∂ e−∂t cos �t ⇒∀ (s + ∂)2 + � 2 1 1 (e−at − e−bt ) ⇒∀ b−a (s + a)(s + b) � ⇒∀ 1 s(s + a)(s + b) � 2 �n �n e−α�n t sin �n 1 − ψ 2 t ⇒∀ 2 2 s + 2ψ�n s + �n 1 − ψ2 � � � 2 �n 1 −α�n t 2t + δ e sin �n 1 − ψ ⇒∀ 2 s(s2 + 2ψ�n s + �n ) 1 − ψ2 � � ≈ 1 − ψ2 −1 δ = tan ψ (Pure Delay) y(t − φ )1(t − φ ) ⇒∀ Y (s)e−sφ dy(t) (Time Derivative) ⇒∀ sY (s) − y(0) dt � 0+ � t y(t)dt Y (s) (Time Integral) y(φ )dφ ⇒∀ + 0− s s 0 .

denoted F (s) exists. The usual rules of integration apply in complex analysis. 3. consider a function of the complex variable s = u + iv: f (s). speciﬁcally Cauchy’s Theorem and the Residue Theorem.4 Background for the Mapping Theorem The mapping theorem uses concepts from complex analysis. are not analytic functions.22. insofar as the antiderivative of f (s). but more than one point in the s plane can map to a single f (s)-plane point. break Y (s) into partial fractions with no powers of s greater than 2 in the denominator.4 Background for the Mapping Theorem = � √ 0 121 x(φ ) G(s)e−sφ dφ = G(s)X(s). as are many functions that can be written as a Taylor or Laurent series. We say f (s) is analytic in a region S if it has ﬁnite derivative and takes ≈ one value at each point s in S. Ignoring the eﬀects of pure time delays. 22. The integral of interest here is � f (s)ds taken on a path in the s-plane. References for this section include Ogata and Hildebrand. and undeﬁned for other values of s. using the transform pairs.4 Solution of Diﬀerential Equations by Laplace Transform The Convolution Theorem allows one to solve (linear time-invariant) diﬀerential equations in the following way: 1. the roots of the numerator. the roots of the numerator. Speciﬁc examples of this procedure are given in a later section on transfer functions. Therefore only discontinuous or multi-valued functions. s.g. so the number of complete loops may not be the same. First. where h(t) is the unit step function. a polynomial in s divided by another polynomial in s. 22. Rational functions are consequently zero for certain values of s. A closed path in the complex s-plane leads to a closed path in the f (s) plane. and the input signal x(t) into X(s). Polynomials in s are analytic. 4. and f (s) is analytic on the path. Perform the multiplication in the Laplace domain to ﬁnd Y (s). we have .3. so that. Apply time delay as necessary. Transform the system impulse response g(t) into G(s). When g(t) is the impulse response of a dynamic system. then y(t) represents the output of this system when it is driven by the external signal x(t).. Generate the time-domain response from the simple transform pairs. e. 2. An especially important class for control system design is the rational function. also called the poles.

since s1 = s2 . provided that f (s) is analytic on the path. and thus � � 2ψ 0 2ψ 0 f (s)ds = i∂r n+1 = i∂r n+1 ei(n+1)ζ dχ [cos(n + 1)χ + i sin(n + 1)χ]dχ. Indeed. Note that it is only the terms with (s − ai )−1 . the integral inside the closed path is zero because it excludes the three simple poles. for which we obtain � f (s)ds = i∂2β. the integral over the outermost curve would be equal to the summation of residues for the poles within (multiplied by i2β). � The second integral is clearly zero for all n. The residue theorem is an extension to an arbitrary number of simple poles which are enclosed by a path: � f (s)ds = i2β ⎫ ∂i . we have s = reiζ . the polynomial order in the numerator is less than that of the denominator) in s can be put into this form. Without the cuts however.122 22 APPENDIX 1: MATH FACTS � s2 s1 f (s)ds = F (s2 ) − F (s1 ). on a circular path of radius r. Consider the function f (s) = ∂sn . This result does not depend on r. ∂2 ∂1 + +. and everywhere within the region enclosed. The connection between Cauchy’s theorem and the residue theorem is indicated in the ﬁgure below. i. i. . The constants ∂i are the residues associated with the poles..e..e. This latter condition results from the following observation. simple poles at ai .. It can be shown also that the result holds for any closed path around the simple pole at s = 0.. whereas the ﬁrst is zero except in the case of n = −1. which characterizes the function. Here. and so applies to a vanishingly small circle around the point s = 0. s − p1 s − p2 f (s) = We show in another section how any strictly proper rational function (that is. It appears that this integral is zero for a closed path. that generate nonzero components. Cauchy’s theorem states that it is.

.4 Background for the Mapping Theorem 123 Im(s) .. we take advantage of the fact that exact diﬀerentials d log |f (s)| and dχ(s) have been found. The ﬁrst integral is zero. f (s) = Working through some algebra. Both terms pertain to the f (s) plane... . . consider the function (s − z1 )m1 (s − z2 )m2 . ds ds Considering the integral of f ∗ (s)/f (s) over a closed contour in the s-plane. . . The mapping theorem comes from now looking in detail at the integral above: f (s) = |f (s)|eiζ(s) dχ(s) d|f (s)| iζ(s) + i|f (s)|eiζ(s) f ∗ (s) = e ds ds 1 d|f (s)| dχ(s) f ∗ (s)/f (s) = +i |f (s)| ds ds d log |f (s)| dχ(s) = +i . is the number of zeros in f . (s − p1 )n1 (s − p2 )n2 . it is straightforward that f ∗ (s)/f (s) = resulting in � m2 n1 n2 m1 + +. . . .22. Re(s) Looking toward the mapping theorem now.− − . s − z1 s − z2 s − p1 s − p2 f ∗ (s)/f (s)ds = i2β(Z − P ). . . . and P = n1 + n2 + . . not the f ∗ (s)/f (s) plane. where Z = m1 + m2 + . is the number of zeros..

23 APPENDIX 2: ADDED MASS VIA LAGRANGIAN DYNAMICS The development of rigid body inertial dynamics presented in a previous section depends on the rates of change of vectors expressed in a moving frame. In words. ⎞ ⎞ ⎞ ⎠ (260) . which is the mapping theorem. 23. Since the body dynamics were already developed. An alternative approach is to use the lagrangian. using the analogous example of added mass terms. the end result is exactly the same.1 Kinetic Energy of the Fluid The added mass matrix for a body in six degrees of freedom is expressed as the matrix Ma . a single circuit in the s-plane may or may not map to a single circuit in the f (s)-plane. speciﬁcally that of the vehicle. As noted above. whose negative is equal to: ⎛ ⎛ ⎛ ⎛ = ⎛ ⎛ ⎛ ⎛ ⎝ � −Ma Xu ˙ Yu ˙ Zu ˙ Ku ˙ Mu ˙ Nu ˙ Xv ˙ Yv ˙ Zv ˙ Kv ˙ Mv ˙ Nv ˙ Xw ˙ Yw ˙ Zw ˙ Kw ˙ Mw ˙ Nw ˙ Xp ˙ Yp ˙ Zp ˙ Kp ˙ Mp ˙ Np ˙ Xq˙ Yq˙ Zq˙ Kq˙ Mq˙ Nq˙ Xr ˙ Yr ˙ Zr ˙ Kr ˙ Mr ˙ Nr ˙ ⎭ ⎞ ⎞ ⎞ ⎞ ⎞ . so n depends directly on the function f (s). it states that the number of zeros minus the number of poles enclosed in a contour in the s-plane is equal to the number of counterclockwise encirclements of the origin. we here develop the lagrangian technique. since log |f (s)| has to be the same at the start and end of the closed contour. the equations elicit the origins of the Munk moment. Among other eﬀects. and is not necessarily one.124 23 APPENDIX 2: ADDED MASS VIA LAGRANGIAN DYNAMICS � s2 s1 d log |f | = 0. wherein the dynamic behavior follows directly from consideration of the kinetic co-energy of the vehicle. in the f (s) plane. Taking χ(s) counterclockwise through angle n2β results in the second term � n2ψ 0 idχ = in2β. Assembling the results. we have i2β(Z − P ) = in2β −∀ Z − P = n.

yG .1 Kinetic Energy of the Fluid 125 where (X. ˙ ˙ ˙ ˙ ˙ ⎞ ⎞ ⎞ ⎞ ⎞. w) denotes the velocity and (p. then.2. or. q. M. nonsingular. we will assume from here on that Ma is symmetric. The sense of Ma is that the ﬂuid forces due to added mass are given by ⎡ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎢ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎣ Xam Yam Zam Kam Mam Nam ⎤ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎥ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎨ = −Ma ⎧ d dt ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎣ ⎡ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎧ ⎢ The added mass matrix Ma is completely analagous to the actual mass matrix of the vehicle. The kinetic co-energy of the ﬂuid Ek is found as: 1 E k = − q T Ma q 2 where q T = (u. we ﬁnd: . Refering to the above equation. r) the angular velocity. v. q. N ) the moment. p ⎧ ⎧ ⎧ ⎧ ⎧ q ⎧ ⎧ ⎧ ⎨ r ⎤ (261) M = m 0 0 0 zG −yG 0 m 0 −zG 0 xG 0 0 m yG −xG 0 0 −zG yG Ixx Ixy Ixz zG 0 −xG Ixy Iyy Iyz −yG xG 0 Ixz Iyz Izz ⎭ where [xG . w.6]. We expand to ﬁnd in the non-symmetric case: −2Ek = Xu u2 + Xv uv + Xw uw + Xp up + Xq˙ uq + Xr ur + ˙ ˙ ˙ ˙ ˙ 2 Yu uv + Yv v + Yw vw + Yp vp + Yq˙ vq + Yr vr + ˙ ˙ ˙ ˙ ˙ 2 Zu uw + Zv vw + Zw w + Zp wp + Zq˙ wq + Zr wr + ˙ ˙ ˙ ˙ ˙ Ku up + Kv vp + Kw wp + Kp p2 + Kq˙ pq + Kr pr + ˙ ˙ ˙ ˙ ˙ 2 Mu uq + Mv vq + Mw wq + Mp pq + Mq˙ q + Mr qr + ˙ ˙ ˙ ˙ ˙ 2 Nu ur + Nv vr + Nw wr + Np pr + Nq˙ qr + Nr r . Z) denotes the force. (K. this implies that the i’th force due to ˙ ˙ the j’th acceleration is equal to the j’th force due to the i’th acceleration. The mass matrix is symmetric.5. ⎞ ⎞ ⎞ ⎠ (262) (263) (264) For the purposes of this discussion. for example. Y. (u. In general. that T is Ma = Ma . p.4.3. for i and j = [1. v. zG ] are the (vessel frame) coordinates of the center of gravity. By restricting the added mass to be symmetric. and positive deﬁnite. r). these terms occur as pairs. Yu = Xv . These properties are also true for the added mass matrix Ma .23. so the asymmetric case could be reconstructed easily from what follows. � ⎛ ⎛ ⎛ ⎛ ⎛ ⎛ ⎛ ⎛ ⎝ u ⎧ ⎧ ⎧ ⎧ v ⎧ ⎧ ⎧ ⎧ w ⎥ . although symmetry fails when there is a constant forward speed.

Y. w] denotes the velocity vector and � = [p. then the inertia terms expressed in axes aﬃxed to a moving vehicle are � �Ek γ γ −�× F = − �t �γ v � � � �Ek γ −�× γ Q = − �t �γ � � � �Ek . in the form: (268) Y = +Xv (u + ur) − Xw )up ˙ ˙ ˙ +Yv (v + vr) + Yw (w − vp + wr) + Yp (p + pr) + Yq˙ (q + qr) + Yr (r + r 2 ) ˙ ˙ ˙ ˙ ˙ ˙ ˙ ˙ ˙ 2 −Zw wp − Zp p − Zq˙ pq − Zr pr (269) ˙ ˙ ˙ . and × denotes the cross product. q. � �γ v �γ (266) (267) γ γ where F = [X.126 23 APPENDIX 2: ADDED MASS VIA LAGRANGIAN DYNAMICS −2Ek = Xu u2 + Yv v 2 + Zw w 2 + Kp p2 + Mq˙ q 2 + Nr r 2 + ˙ ˙ ˙ ˙ ˙ 2Xv uv + 2Xw uw + 2Xp up + 2Xq˙ uq + 2Xr ur + ˙ ˙ ˙ ˙ 2Yw vw + 2Yp vp + 2Yq˙ vq + 2Yr vr + ˙ ˙ ˙ 2Zp wp + 2Zq˙ wq + 2Zr wr + ˙ ˙ 2Kq˙ pq + 2Kr pr + ˙ 2Mr qr. �γ v �Ek �Ek −γ × v . N ] the moment vector. see the derivation below. ˙ (265) 23. through rotational symmetry. M. or Milne-Thomson (1960). 23. These relations state that if γ = [u. we derive the following terms containing the ﬂuid inertia forces: X = +Xu u + Xv (v − ur) + Xw (w + uq) + Xp p + Xq˙ q + Xr r ˙ ˙ ˙ ˙ ˙ ˙ ˙ ˙˙ ˙˙ 2 −Yv vr + Yw (vq − wr) − Yp pr − Yq˙ qr − Yr r ˙ ˙ ˙ ˙ 2 +Zw wq + Zp pq + Zq˙ q + Zr qr. r] the v γ angular velocity vector. v. we use Kirchhoﬀ’s relations with the co-energy Ek .3 Fluid Inertia Terms Applying Kirchhoﬀ’s relations to the expression for the kinetic co-energy with a symmetric added mass matrix. Z] express the force vector. ˙ ˙ ˙ The Y and Z forces can be obtained. Q = [K.2 Kirchhoﬀ ’s Relations To derive the ﬂuid inertia terms in the body-referenced equations of motion.

i. and N are obtained in a similar manner as: K = −Xv wu + Xw uv + Xr uq + Xp u − Xq˙ ur ˙ ˙ ˙ ˙˙ 2 2 −Yv vw + Yw (v − w ) + Yp (v − wp) − Yq˙ (vr + wq) + Yr (vq − wr) ˙ ˙ ˙ ˙ ˙ +Zw vw + Zp (w + vp) + Zq˙ (vq − wr) + Zr (wq + vr) ˙ ˙ ˙ ˙ +Kp p + Kq˙ (q − rp) + Kr (r + pq) ˙ ˙˙ ˙ ˙ 2 2 +Mr (q − r ) − Mq˙ qr ˙ +Nr qr ˙ (271) M = +Xu uw + Xv vw + Xw (w 2 − u2 ) + Xp (ur + wp) + Xq˙ (u + wq) + Xr (wr − up) ˙ ˙ ˙ ˙ ˙ ˙ −Yw uv + Yp vr + Yq˙ v − Yr vp ˙ ˙ ˙ ˙ −Zw uw + Zp (wr − up) + Zq˙ (w − uq) − Zr (ur + wp) ˙ ˙ ˙ ˙ 2 2 +Kp pr + Kq˙ (p + qr) + Kr (r − p ) ˙ ˙ ˙ ˙ +Mq˙ q − Mr pq + Mr r ˙ ˙ ˙ −Nr pr (272) ˙ N = −Xu uv + Xv (u2 − v 2 ) − Xw vw − Xp (uq + vp) + Xq˙ (up − vq) + Xr (u − vr) ˙ ˙ ˙ ˙ ˙ ˙ +Yv uv + Yw uw + Yp (up − vq) + Yq˙ (uq + vp) + Yr (v + ur) ˙ ˙ ˙ ˙ ˙ −Zp wq + Zq˙ wp + Zr w ˙ ˙ ˙ 2 2 −Kp pq + Kq˙ (p − q ) + Kr (p − qr) ˙ ˙ ˙ +Mq˙ pq + Mr (q + pr) ˙ ˙ +Nr r (273) ˙˙ 23. Mv = Yq˙ .e.4 Derivation of Kirchhoﬀ ’s Relations We can derive Kirchhoﬀ’s relation for a lagrangian L(γ . to minimize the integral I (Crandall et al. involving the velocity γ and v γ v angular velocity � .. which allows us to use only the 21 upper-right elements of the added mass matrix in Equation 260. �. 1968): . i. e.e.. in a reference system ﬁxed on the body. whose components will be expressed in a local coordinate system rotating γ with the angular velocity � .23.g. t). M . ˙ The moments K.4 Derivation of Kirchhoﬀ’s Relations 127 Z = −Xu uq + Xv (up − vq) + Xw (u − wq) − Xp pq − Xq˙ q 2 − Xr qr ˙ ˙ ˙ ˙ ˙ ˙ 2 +Yv vp + Yw (v + wp) + Yp p + Yq˙ pq + Yr pr ˙ ˙ ˙ ˙ ˙ ˙ ˙ ˙ ˙ +Zw w + Zp p + Zq˙ q + Zr r ˙ ˙ ˙ (270) The apparent imbalance of coeﬃcients comes from symmetry. The principle to γ satisfy is that of least action.

in view of equations 277 and 278. (280) . homogeneous ﬂuid. while ζ� and ζδ are the � projections of the free vectors onto a given reference frame. �. can be written as v � ˆ ˆ ζγ = ζv T x + {v T ζ x} v = � �ζ� + � � ζ� �t �T γ ˆ v x + γ × ζ δ. ζ δ v γ � � � �� dt. � t2 t1 ζI = � γ γ γ � ζL(γ . (275) γ Hamilton’s principle in its general form also accounts for applied forces and moments Σ. ζ� �γ v �γ � � � � (279) The variations ζγ and ζγ . inﬁnitesimal linear and angular displacements ζγ � γ leading to and ζ δ. ζγ and ζ δ are interpreted as free vectors. ζv + . ζI. The following relationships link the displacements with the body-referenced rates: �� + � � �. there is no potential energy. and �t �δ � = . Using the notation detailed in the Nomen γ clature section below. � (277) (278) ζL(γ . They are deﬁned to align with the generalized. ζγ + Σp. �t v = A variation of the lagrangian at a given time t is.q. Since we are considering the motion of a body in an unbounded. t) = v γ �L γ �L γ . the lagrangian is invariant under coordinate transformation. to ﬁrst order.r . (276) Now. which in the more general case is the kinetic co-energy minus the potential energy of the system. so it is a function of the free vectors of velocity and angular velocity. so the lagrangian is exactly the kinetic energy: L = Ek .w . as written.128 23 APPENDIX 2: ADDED MASS VIA LAGRANGIAN DYNAMICS I= � t2 t1 L(γ . involves only the lagrangian.the admissible condition . t) + Σu. and hence of L with the velocity γ and angular rate � vanishes. t)dt v γ (274) At the minimum value of I .v. �. v γ Our condition ζI = 0.the variation of I. �.

γ × γ >=< γ γ × γ > . c a Hence. We now return to the evaluation of the lagrangian L. these are the forces that the ﬂuid exerts on the body. the variation of the action I2 . in ζI2 . speciﬁcally: how the free ˆ ˆ vectors ζγ and ζ w change as the triad rotates. v γ With this in mind. one can easily derive the proper interpretations. but the projections v and w remain constant. ζγ + � . The kinetic co-energy we developed is that of the ﬂuid.23. In evaluating the remaining terms. γ � �γ v �γ v �γ � t1 t2 (283) � ⎬� � � �� (284) Finally. ζ δ dt. leading for example to v T (ζ δ × x). .4 Derivation of Kirchhoﬀ’s Relations 129 ˆ ˆ ζγ = ζ� T x + {� T ζ x} � = � �ζδ �t �T γ ˆ γ x + � × ζ δ. with a sign change.r . This form. ζγ + γ × v +�× γ . ζγ + γ p. ζ δ � Σ � � �� dt. �γ � �t � �� � � � �T ⎬� � �L � �L γ . and so the generalized forces indicate what forces the body would exert on the ﬂuid. as written above. we have from an integration by parts � t2 t1 ζI1 = �� t2 = − � �L �ζ� ⎝ . � × ζ δ dt γ = �γ v �γ � t1 � � �� � t2 ⎬� �L �L �L γ = − �× . It can be shown that ζ x. There are no terms remaining at the time boundaries because the lagrangian is zero at these points. one needs to use the following property of triple vector products: < γ .v. is ζI2 �L �L γ γ . γ × γ >=< γ. �γ v �t t1 � �T � �L �ζδ x + ˆ . (285) The Kirchoﬀ relations follow directly from combining the three action variations. we see that there will be ﬁve inner products to consider. ˆ however. Combining terms. (281) ˆ ˆ The terms {v T ζ x} and {w T ζ x} above represent the eﬀects of the variation of body orienta tion. � × ζγ + γ × ζ δ + γ � v . and lead to one of the more subtle points of the derivation. Writing the terms involving the time derivatives as ζL1 and ζI1 .q. a b c c b a b.w . we write the part of ζI due to the generalized forces: ζI3 = � t2 t1 �� γ γ Σu. expressed in terms of vectors. ˆ γ ˆ γ ˆ the displacement of the unit triad x is ζ δ × x. fails to capture what is meant by {v T ζ x} and {w T ζ x}. ζδ �t �γ v �t �γ � ˆ x ⎠ dt (282) �⎭ dt.

5 23. ˆ The notation is hybrid. x = (ˆ ˆ k)T ˆ i. j.: containing as elements the unit vectors of the chosen Cartesian system.5. i. whose element i is the derivative of L with x respect to the ith element of γ . and a column vector f which denotes the components of the vector f in a given coordinate system.5. . j. due to R. g x >. i.1 Nomenclature Free versus Column Vector γ We make the distinction between a free vector f . ˆ ˆ ˆ ˆ k. The product between the row vector ˆ f T and the column vector x is in the usual matrix multiplication sense. γ >.e. denoted in terms of column vectors as z = f � g. The key observation is best given through a loose example. z 24 APPENDIX 3: LQR VIA DYNAMIC PROGRAM MING There are at least two conventional derivations for the LQR. i. but convenient since it allows us to write: γ ˆ f = f T x (287) (286) where f T denotes the transpose of f .2 Derivative of a Scalar with Respect to a Vector The derivative of a scalar L with respect to a vector γ is denoted as: x �L �γ x (288) and is a vector with the same dimension as γ .130 24 APPENDIX 3: LQR VIA DYNAMIC PROGRAMMING 23. and in the same direction: x � �L �γ x � = i �L �γ i x (289) 23. The connection between the two concepts is given via the free triad x.3 Dot and Cross Product γ γ g The dot product of two free vectors f and γ is denoted as < f . 23. Bellman. The cross product of two vectors f and γ is ˆ ˆ g γ g γ = f × γ . a row vector. we present here one based on dynamic programming.e.5. and we use the following g T T γ notation for column vectors: < f x. which is an element of a linear vector γ ﬁeld.

Example: Nodal Travel. and the C-nodes to the terminal point D. Suppose there are only three places to cross the Rocky Mountains. B2 D |opt . for example. There are three possible paths from B1 to D. AB2 + B2 D |opt . Suppose that all of the paths (and distances) from A to the B-nodes are known. this same arbitrary point X would be along the new optimal path as well. and nine paths total from the B-level to D. compute the distance for each of the possible paths from A to D. as are those from the B-nodes to the C-nodes. The comparison of numbers is relatively cheap. this example gives only a mild advantage to the dynamic programming 3 Apologies to readers not familiar with American geography. Store the best paths as B1 D |opt . C3 .3 By way of notation. This operation involves nine additions of two numbers. The combined path with the shortest distance is the total solution. C2 . The dynamic programming approach has two steps. B1 . and three places to cross the Mississippi River. there are many paths to choose from! Assume that one way or another we have found the best path. and the idea is to use this property backwards through time to evolve the optimal path. Let X be an arbitrary point east of Las Vegas. Consider now traveling from point A (Los Angeles) to Point D (Boston). There are nine unique paths from A to D. say Las Vegas. from each B-node. B3 .24. and picks the shortest one. and we ask what is the shortest path in miles. this second step involves three sums of two numbers. constrained to the optimal paths from the B-nodes onward: AB1 + B1 D |opt .1 Example in the Case of Discrete States Suppose that we are driving from Point A to Point C. and that a Point B lies along this path. we say that the path from A to B1 is AB1 . If A and C represent Los Angeles and Boston. pick the best path to D. Needless to say. B2 . Second. . C1 . In terms of computations. First.1 Example in the Case of Discrete States 131 B1 A B2 B3 C1 C2 C3 D n=3 s=2 24. we could summarize that this method requires nine additions of three numbers. equivalent to eighteen additions of two numbers. and total optimization is done in twelve additions of two numbers. for example. If we were to now solve the optimization problem for getting from only Las Vegas to Boston. B3 D |opt . The point is a subtle one: the optimization problem from Las Vegas to Boston is easier than that from Los Angeles to Boston. beginning in Boston. We now add some structure to the above experiment. or AB3 + B3 D |opt . A brute-force approach sums up the total distance for all the possible paths.

u(t + ζt))} . u(t))ζt + V (x(t + ζt). the integral over time of the instantaneous penalty. u �x We next make the assumption that V (x. and each has width n (e. u) = xT P x. u(t)) + (Ax(t) + Bu(t))ζt. u(t)) + (Ax(t) + Bu(t)) . If V is smooth and has no explicit dependence on t. if there are s layers of nodes (e. dynamic programming needs only 105. t + ζt). n river crossing points).o. �x V (x(t + ζt). (290) i. u(t)) is made by considering all the possible control inputs u in the time interval (t. As suggested by dynamic programming. the optimal return is the cost of the optimal trajectory remaining after time t: V (x(t).t.132 24 APPENDIX 3: LQR VIA DYNAMIC PROGRAMMING approach over brute force. t + ζt) cannot aﬀect V (x(t). We have directly from the dynamic programming principle V (x(t). as one increases the dimensions of the solution space. In the case of n = 5.g. First we deﬁne an instantaneous penalty function l(x(t). of keeping xdesired = 0.. u(φ ))dφ.g. so inserting the above and making a cancellation gives �V 0 = min l(x(t). then �V dx ζt + h. 24. u (292) The minimization of V (x(t). u(t)) + (293) Now control input u in the interval (t. The gap widens vastly. u) has the following form: 1 V (x. however. which is to be greater than zero for all nonzero x and u. The cost associated with this penalty.e. −∀ �x dt �V = V (x(t). the return at time t is constructed from the return at t + ζt.. u(t)) = min {l(x(t). Finally. as written. u(t)). u(t)).2 Dynamic Programming and Full-State Feedback We consider here the regulation problem. In general. the brute force approach will take (sns ) additions. u(t + ζt)) = V (x(t). u(t)) = � √ t l(x(φ ). u(t))dt. u). s = 5. that is. brute force requires 6250 additions. is J= � √ 0 l(x(t). and the diﬀerential component due to l(x. The closed-loop system thus is intended to reject disturbances and recover from initial conditions. 2 (295) � � (294) . (291) . while the dynamic programming procedure involves only (n2 (s − 1) + n) additions. There are several necessary deﬁnitions. along an optimal trajectory. rivers or mountain ranges).. but not necessarily follow y-trajectories.

133 where P is a symmetric matrix. u (296) We ﬁnally specify the instantaneous penalty function. an inﬁnite gain ampliﬁcation in any input channel. (299) 25 Further Robustness of the LQR The most common robustness measures attributed to the LQR are a one-half gain reduction in any input channel. The matrices Q and R are to be set by the user. and represent the main “tuning knobs” for the LQR. While these are general properties 4 5 Positive deﬁniteness means that xT P x > 0 for all nonzero x. . 2 2 leading to the ﬁnal matrix-only result 0 = Q + P A + AT P − P BR−1 B T P. Substitution of this form into the above equation. u) = xT Qx + uT Ru. we have 1 1 0 = xT Qx − xT P BR−1 B T P + xT P Ax. and eliminating u in favor of x. 2 2 All the matrices here are symmetric except for P A. or a phase error of plus or minus sixty degrees in any input channel. (297) 2 2 where Q and R are both symmetric and positive deﬁnite. (298) The gain matrix for the feedback control is thus K = R−1 B T P . we can make its eﬀect symmetric by letting 1 1 xT P Ax = xT P Ax + xT AT P x. and setting the derivative with respect to u to zero gives 0 = uT R + xT P B uT = −xT P BR−1 u = −R−1 B T P x.45 It follows that �V = xT P −∀ �x � � 0 = min l(x. and xT P x = 0 if x = 0. Inserting this solution back into equation 297. u) + xT P (Ax + Bu) . The LQR employs the special quadratic form 1 1 l(x. since xT P Ax = xT AT P x. and positive deﬁnite. This suggested form for the optimal return can be conﬁrmed after the optimal controller is derived.

or in a passive electrical system through resistance. dissipated in x a passive mechanical system by damping. These include robustness to uncertainty in the real coeﬃcients of the model (e. moderate robustness in saturation follows from the basic analysis. and certain nonlinearities. x A < B means γ T Aγ < γ T Bγ . ρ (302) The scalar ρ is a free parameter we can specify. rate and position limits in control surface actuators. for all real ρ > 0. coeﬃcients in the A matrix). A useful analogy for the Lyapunov function V (γ ) is energy. Positive deﬁnite A means γ T Aγ > 0 for all γ .2 Matrix Inequality Deﬁnition Inequalities in the case of matrices are in the sense of positive and negative (semi) deﬁnite ness.3 Franklin Inequality (301) (300) A theorem we can use to assist in the Lyapunov analysis is the following. With A and B square and of the same dimension.. 25.1. Saturation nonlinearities in particular are ubiquitous in control system application. attributed to Franklin (1969). including control switching and saturation. these robustness measures hold.1 25.1. we say for the case of a scalar ρ. 25.134 25 FURTHER ROBUSTNESS OF THE LQR that have a clear graphical implication on the Bode or Nyquist plot. x the system is stable. positive semideﬁnite A means γ T Aγ → 0 x x x x x for all γ . x x x x x Also.1. It is assumed that the matrices A and B are of consistent dimensions. we ﬁnd them in ”railed” conditions of ampliﬁers.g. other useful robustness conditions can be developed. with V (γ ) = 0 only when γ = γ and if dV (γ )/dt < 0 for all time. We will use the Lyapunov stability and the LMI formulation of matrix problems in this section to expand these ideas.1 Tools Lyapunov’s Second Method The idea of Lyapunov’s Second Method is the following: if a positive deﬁnite function of the state γ can be found. and in well-meaning but ad hoc software limits. . 25. then x x x 0. As shown below. but much stronger results can be obtained with new tools. for all γ . When the LQR is used to deﬁne the loop shape in the loop transfer recovery method (as opposed to the Kalman ﬁlter in the more common case). 1 AT B + B T A √ ρAT A + B T B. A < ρ means A − ρI < 0.

(309) Again. is obvious by considering γ that involve only A or D.1.1 Tools 25. x x 0 where nA and nD are the dimensions of A and D. consider now the symmetric matrix A B C ⎛ T ⎞ M = ⎝ B D 0 ⎠. x (307) and thus M > 0 is equivalent to A > 0 and �A > 0.g. .25. e. 25. 25.. The fact that both A and D have to be positive deﬁnite. positive deﬁniteness of a large matrix is reﬂected in a smaller matrix inequality. independent of the Schur γ complements.5 Proof of Schur Complement Sign It is easy to verify that M= ⎬ A B BT D � = ⎬ I 0 B T A−1 I ⎬ �⎬ The outer matrices are nonsingular so there is a one-to-one transformation γ= z I A−1 B 0 I � A 0 0 D − B T A−1 B �⎬ I A−1 B 0 I � (306) γ.1. The proof for �D is completely analogous. as �A = D − B T A−1 B �D = A − BD −1 B T . respectively. and an important property is that M > 0 ⇒∀ �A > 0 and �D > 0 and A > 0 and D > 0. (303) The Schur complements of blocks A and D are deﬁned.6 Schur Complement of a Nine-Block Matrix For the purposes of derivation below.1. but in a smaller dimensioned matrix equation. γ = [11×nA γ 1×nD ]T . we have M > 0 ⇒∀ A − BD −1 B T − CE −1 C T > 0 and D > 0 and E > 0.4 Schur Complement 135 Consider the symmetric block matrix deﬁned as follows: M= ⎬ A B BT D � . (305) (304) The Schur complements thus capture the sign of M . CT 0 E � ⎭ (308) Using the Schur complement of the diagonal block including D and E.

136 25.1.7

25 FURTHER ROBUSTNESS OF THE LQR Quadratic Optimization with a Linear Constraint

Consider the quadratic function V = γ T P γ , P symmetric and positive deﬁnite, and the x x minimization of V subject to a constraint γ T γ = a, where γ T is a row vector, and a is a c x c scalar. The solution is: a2 . (310) γ T P −1γ c c To show that this is true, augment the objective with a LaGrange multiplier to become min V = ˜ V = V + η(γ T γ − a). c x (311)

˜ x Setting � V /�γ = 0 gives 2γ T P + ηγ T = 0. Post-multiply this equation by P −1γ, and solve x c c to give η = −2a/γ T P −1γ. Then, using 2γ T P + ηγ T = 0 again, post-multiply by γ /2, and c c x c x insert η, concluding the proof.

25.2

Comments on Linear Matrix Inequalities (LMI’s)

The reason that solving M > 0 is simpler than the equivalent Schur complement inequality y) is that, as long as M (γ is aﬃne in the unknowns γ the solution set is convex. Aﬃne means y, that M (γ = Mo + K[y γ ... γ]N, y) γy y (312)

i.e., that M is a constant matrix plus terms linear in each element of γ. Note the matrices y K and N here, as well as the columnwise expansion of γ are needed to put elements of γ in y y, arbitrary locations of M . Convex means that if γ1 and γ2 are feasible solutions to the inequality, then so is γ3 = y y y ∂γ1 + (1 − ∂)γ2 , for any ∂ between zero and one. A convex solution set has no hidden y y corners or shadows, because every point on a straight line between two solutions is also a solution. Computer programs solving convex optimization problems can always get the global optimum eﬃciently, or determine that no solution exists. From the Schur complements above, clearly we can transform a (suitable) matrix equation that is quadratic or higher in the unknowns into an aﬃne M (γ ), and therefore make an y eﬃcient solution. The term linear matrix inequality refers of course to the fact that M is aﬃne and hence linear in the unknown variables, but also to the methods of analysis and numerical solution, wherein the idea is to recognize M (γ ) = Mo + M1 y1 + M2 y2 + ...Mn yn , y where n is the dimension of the unknown vector γ. y Example. Consider the LMI M (γ) = y

⎬

(313)

y1 − 1 2y2 − 1 y2 + 1 y 2 − y 1

�

> 0.

(314)

25.3 Parametric Uncertainty in A and B Matrices

137

We use the theorem that M > 0 is achieved if each of the submatrices cornered at the (1,1) element has positive determinant. Then M > 0 is equivalent to

y1 > 1 (y1 − 1)(y2 − y1 ) − (2y2 − 1)(y2 + 1) > 0.

(315)

y The solution set is shown as S in the ﬁgure below; M is aﬃne in γ and the solution set is convex.

1.5

1

0.5

y2 0

−0.5

−1

−1.5

−0.5 0 0.5 y

1

S

1

1.5

25.3

25.3.1

**Parametric Uncertainty in A and B Matrices
**

General Case

Let S be the solution to the matrix Riccati equation AT S + SA + Q − P BR−1 B T P = 0. (316)

˙ We consider here perturbations to the design plant matrices A and B of the form γ = x −1 T (A + �A)γ + (B + �B)γ . The control in the LQR is γ = −R B P γ . Select a Lyapunov x u u x function of the form V (γ ) = γ T Sγ . Then x x x ˙ V ˙T ˙

= γ Sγ + γ T Sγ x �

x x x = γ T AT S + SA + �AT S + S�A − 2SBR−1 B T S− x SBR−1 �B T S − S�BR−1 BS γ . x

�

(317)

We give the following special structures to the perturbations:

138

25 FURTHER ROBUSTNESS OF THE LQR

�A = DA �A EA �B = DB �B EB .

(318)

The idea is to capture all of the uncertainty in �A and �B and capture the structure of the perturbations using the other four matrices. We have, using the Franklin inequality, ˙ V

T T = γ T AT S + SA − 2SBR−1 B T S + EA �T DA S + SDA �A EA x A

�

(319)

√ γ x

T

⎬

T T −SBR−1 EB �T DB S − SDB �B EB R−1 BS γ x B

T AT S + SA − 2SBR−1 B T S + ρA EA �T �A EA + A −1 T EB �T �B EB R−1 B T S B

�

1 T SDA DA S ρA

�

1 T +ρB SBR + SDB DB S γ x ρB ⎬ 1 T T T SDA DA S √ γ AT S + SA − 2SBR−1 B T S + ρA EA EA + x ρA � 1 T −1 T −1 T +ρB SBR EB EB R B S + x SDB DB S γ , ρB where the last inequality holds if �T �A √ I, A �T �B √ I. B

(320)

(321) (322)

For a given perturbation model with matrices DA , EA , DB , and EB , we thus obtain a matrix inequality which, if met, guarantees stability of the system for all γ : x

T AT S + SA − 2SBR−1 B T S + ρA EA EA +

1 T SDA DA S ρA 1 T T +ρB SBR−1 EB EB R−1 B T S + SDB DB S < 0. ρB

(323)

Since scalars ρA and ρB can take on any positive value, they remain to be picked, and in fact judicious choices are needed to maximize the robustness guarantees – i.e., to minimize the conservativeness. 25.3.2 Uncertainty in B

Putting aside �A for the moment, and substituting in the Ricatti equation into the above leads to

This �B is a very speciﬁc structure. this analysis conﬁrms the one-half gain reduction property derived earlier. but the full. standard result on a second look.i | ρB + ρB Solving for ρB gives 1± � 2 1 − 4Ni. Usage of the Franklin inequality to symmetrize the components involving �B is the cause of this. In summary. On the other hand. in the special case of diagonal N in the model of �B.i � � > 0. we crafted DB and EB so as to align terms of the form SBR−1 B T S.3 Parametric Uncertainty in A and B Matrices 139 T −Q − SBR−1 B T S + ρB SBR−1 EB EB R−1 B T S + 1 T SDB DB S < 0. which is overly conservative. ρB (326) � R|N | = DB �B EB . (325) because Q can be arbitrarily small in the LQR. simplifying the analysis. ρB (327) 1 SBR−1 |N |B T S < 0. or 1 + 2Ni. (329) showing that real.25. This leads to a conservative result using the Franklin inequality. ρB (324) Following the presentation of gain and phase margins of the LQR in a previous section. . positive ρB are attained for |Ni.i | . we here consider gain margins through the diagonal matrix N . Hence. (330) showing both the one-half reduction and inﬁnite upward gain margins. such that �B = BN = B � R−1 |N | ×I × |N | denotes the matrix made up of the absolute values of the elements of N . The stability inequality becomes −Q − SBR−1 B T S + ρB SB |N |R−1 B T S + or. but probably with greater eﬀort. Since N is diagonal. for example). equivalently.i | < 1/2. it confers only a one-half gain ampliﬁcation robustness. whose rationale becomes evident below.i > 0. certainly a large value of Q increases the robustness (see the case of �A below. I − ρB |N | − 1 |N | > 0. (328) ρB = 2|Ni. it is suﬃcient to keep all 1 1 − |Ni. since it can be veriﬁed that using Equation 319 directly gives −Q − SBR−1 (I + N + N )B T S < 0. Better results could be obtained for speciﬁc cases.

and requires that elements of �A would have to change together.5 1 0 � . for example . B = [1 0]T Q = diag(1. 1 2 r Now let. and very substantial perturba 2 1 tions in A are allowed.140 25. which is primarily an issue of gain and does not aﬀect the open-loop stability. setting aside �B now: T −Q − SBR−1 B T S + ρA EA EA + 1 T SDA DA S < 0. Pertur bations to the design matrix A cause the open.and closed-loop poles to move. or 2 1 equivalently.i . Consider the main in equality for stability again. and we set Q diagonal. ⎬ (335) A = −0. A more useful result involves a diagonal �A . and for the purposes of illustration. we consider a second order system as follows: ⎬ � �A = D A �A E A = = ⎬ d1 d2 0 0 . The resultant inequality is �⎬ −Q + ρA I − S b2 /r 0 0 0 � − ⎬ (d2 + d2 )/ρA 0 1 2 0 0 �� S < 0. setting ρA = min Qi. (334) The condition for stability will be met if Q − ρA I > 0 and b2 /r − (d2 + d2 )/ρA > 0. �⎬ ζ1 0 0 ζ2 � I (332) (333) d 1 ζ1 d 2 ζ2 0 0 In addition.3 Uncertainty in A 25 FURTHER ROBUSTNESS OF THE LQR Uncertainty in the system matrix A has a diﬀerent role in robustness than does uncertainty in B.i > d2 + d2 . 1) R = 1.3.and closed-loop systems.5 −0. potentially destabilizing either or both of the open. ρA (331) Aligning the uncertain terms with SBR−1 B T S (as with �B above) proves to be impractical here because it imposes a speciﬁc structure on �A. we set B = [b 0]T . for example. This informs us that the robustness condition is d2 + d2 < 2. if b2 min Qi. so that R is a scalar r.

T −Q − SBR−1 B T S + ρA EA EA + 1 T SDA DA S ρA 1 T T +ρB SBR−1 EB EB R−1 B T S + SDB DB S < 0 ρB (336) is equivalent to � T T −Q − SBR−1 B T S + ρA EA EA + ρB SBR−1 EB EB R−1 B T S SDA SDB ⎛ T −ρA I 0 ⎞ < 0.4 A and B Perturbations as an LMI Using the triple LMI of Equation 309.3 Parametric Uncertainty in A and B Matrices 141 A + �A = ⎬ 0. or EB . if it exists. Alternatively. one could lower the robustness by altering DA . 25. is convex. where possible may increase robustness. the state penalty. (right) closed-loop pole locations with LQR.3.25.5 0. EA . for the combined A and B structured uncertainty models. increasing Q. If a solution does not exist. . 2 2 1 Imag(s) Imag(s) 1 0 0 −1 −1 −2 −3 −2 −1 Re(s) 0 1 −2 −3 −2 −1 Re(s) 0 1 Figure 9: (left) open-loop pole locations along the boundary of the stability region in Equation 335. Computed results that conﬁrm the bound are shown in Figure 9. (337) DA S ⎠ ⎝ T 0 −ρB I DB S ⎭ The solution set [ρA ρB ].5 1 0 � . An interesting note is that the gain reduction margin in the presence of this �A suﬀers dramatically.93. we have. with failure of the robustness if b < 0. the uncertainty in A has used up almost all of the available safety margin. DB . Intuitively.

and for certain vectors γ. switch(u. Through scaling the B matrix. the following inequality holds: z 2γT f (R−1γ) → γT W R−1γ.U) Our stability analysis so far is one wherein we seek to guarantee a negative rate in V (γ ) for x all time. we can set Ui = 1. U sign(u). |u| > U.i (341) . We say that f (γ ) is the diagonal function taking control commands from the LQR (γ ) to u u the input to the plant. ∗). and hence f ’s extreme outputs are ±1.i � � 2 → zi Wi.4 Input Nonlinearities Control systems in practice often make use of switched or saturated action in the control channels. zi 2zi sat Ri. This section details the approach of Huang & Lam (2002). it is clear that any N can be tolerated within the range (−. which uses LMI’s and the structured uncertainty models above. Ri.5. elements of f correspond to the saturated operator above. Switched and saturated control channels are deﬁned respectively as switch(u.i . If the second argument of sat(·) is omitted. In practice. this condition is met if. |u| < U . U ) = U sign(u). that is. It follows directly then that switching and saturated control channels cannot destabilize the LQR system until the gain reduction is more than one-half. 1) = u. we assert that under the proper choice of positive deﬁnite matrix W . for each control channel. and this is the aspect of saturated control in particular that we wish to pursue now. and � (338) sat(u. (339) where the scalar U is positive. U ) = U sat(u/U. z z z z (340) If the control penalty matrix R is diagonal. systems routinely operate beyond this point.142 25 FURTHER ROBUSTNESS OF THE LQR 25. even if it is time varying.U) U -U u -U -U U u U sat(u. and we constrain W also to be diagonal. |u| > 2U . it is taken to be one. Referencing Equation 330. however. First.

and considering uncertainty �B only.i � � � 2 � � zi → Wi. As we show ˙ x below.i .i . namely |eT B T Sγ | √ 2Ri. the optimization then is to maximize the state space satisfying Equation 342. . giving T SA + AT S − SBW R−1 B T S + ρB SDB DB S + (346) 1 T SBR−1 EB EB R−1 B T S < 0 ρB (347) as the suﬃcient condition for stability. Because for the LQR we have zi = eT B T Sγ . the matrix W appears in the only negative term of V (γ ). the last term is bounded. since in fact |fi | √ 1. (344) x x x x x x Note that �B never appears inside f because f operates on the control u. however. we obtain ˙ x V (γ ) √ γ T (SA + AT S)γ − γ T SBW R−1 B T Sγ x x x x 1 T −1 T T +ρB γ T SDB DB Sγ + x x f (R B Sγ )EB EB f (R−1 B T Sγ )). With V (γ ) = γ T Sγ . (342) We will use the inequality of Equation 340 in our Lyapunov analysis. If the channel is saturated. x i (343) Viewed this way. Huang and Lam address this point. then it is required instead that � � � Ri. with special attention to these conditions. where ei is the unit vector x i in the i’th direction. under the proper constraints on W .25.4 Input Nonlinearities 143 If the channel is not saturated. calculated with B alone.i /Wi. we have x x x ˙ x V (γ ) = γ T (SA + AT S)γ − 2γ T SBf (R−1 B T Sγ ) − 2γ T S�Bf (R−1 B T Sγ ). This last step is extremely conservative. Equation 342 eﬀectively places a bound on the state under saturated control. consider the remaining asymmetric term. while keeping W big enough to guarantee stability.i as small as possible.i √ 2. x T x ρB Finally. then we require Wi. clearly we would like to make each Wi. Applying the Franklin inequality to the last term yields ˙ x V (γ ) √ γ T (SA + AT S)γ − 2γ T SBf (R−1 B T Sγ ) x x x x 1 T −1 T T x T x +ρB γ T SDB DB Sγ + x x f (R B Sγ )EB EB f (R−1 B T Sγ )). Applying Equation 340. ρB (345) Next.

< k. x x In summary. < W. and Equation 349 for maximizing the level of V . As noted above. 0 √ Wi. This itself can be posed as an optimization in a new scalar k: 1 (349) minimize k such that kI − R−1 W � > 0. the range of states leading to guaranteed stability is limited to those that are known not to leave Ω.i 2 Wi. As in the case of no input saturation. The problem now is that although the Lyapuonov function derivative is clearly less than zero for states within Ω. we have to settle for the largest ellipse completely contained in Ω. to yield Vmax = mini 2 4Ri. This is entirely likely since V is a quadratic function. which is equivalent to maximizing the minimum value of all the possible right hand sides of the above equation. the state trajectory could leave Ω. these translate to linear constraints on γ .i must be as in Equation 342. this can be posed and solved as set of LMI’s (which itself can be posed as a single LMI): ⎬ T T SA + AT S − SBW R−1 B T S + ρB SDB DB S SBR−1 EB EB R−1 B T S ρB I 1 −1 R W� 2 W 0 � < 0. (350) (351) (352) (353) .i √ 2 for cases where channels are not saturated. we have three constraints and one optimization: Equation 347 for Lyapunov stability.144 25 FURTHER ROBUSTNESS OF THE LQR The supporting conditions on elements of W are linear: Wi.i eT B T SBei i (348) We would like to maximize Vmax . In other words. x2 V contours Vmax Ψ x1 As a result. < 2.an elliptical region in the state space . referencing Equation 343 as the set of linear constraints. It follows that any γ is stable when V (γ ) = x x i T 2 γ Sγ < 1/k .for which stability is guaranteed.i = eT B T SBei . We now invoke Equation 310. with �i. Call the set x of states that satisfy these conditions Ω. 2 � where � is diagonal. leading to elliptical contours as in the ﬁgure below. and less than two.

4 Input Nonlinearities 145 To solve.25. Decrease k as far as possible. the feasible solution achieves the maximum V from which the range of state space can be found from V = 1/k 2 . pick a value for k and ﬁnd a feasible solution. .

- Jurnal two port
- 093977366XStrength of Ships
- Acoustic analogy
- Ref Exact Stiffness Matrix for Beams on Elastic Foundation
- Analysis of Yagi
- Gauge Fields Supercomputers
- Talk
- GATE Syllabus
- Chapter 4
- Beam
- A Continuum Mechanics Premier
- Art417
- Natural Frequency of Beam
- AN OVERVIEW OF NUMERICAL METHODS FOR ACOUSTIC WAVE PROPAGATION
- Fdflow a Fortran 77 Solver for 2d Incomp Flow
- 5 Convolution
- Vibrations Session 11
- Two Ports
- Jun-Bao Wu and Chuan-Jie Zhu-MHV Vertices and Scattering Amplitudes in Gauge Theory
- The PISO Algorithm-1
- Chpt 3 Lecture Powerpoint
- EE 370, Lecture 11
- Chapter 9
- fem bits for mid 2
- 0705.3868v1
- 2_Introduction to Modern Solid State Physics
- Chapter 3 Lecture
- Wave Equation
- lamp_tu-graz_ac_at_~hadley_ss1_KronigPenney_KronigPenney_php
- Khalil - Nonlinear Systems Slides

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd