You are on page 1of 5

Hydrodynamic/Autopilot Design of an AUV

Michael L. Graves II
Engineer, Vehicle Control Technologies, Inc.
Dr. Douglas E. Humphreys
President, Vehicle Control Technologies, Inc.

Abstract – The Navy is currently developing a small AUV


that consists of a hull with counter rotating propellers on the
front, a wing, and control fins arranged in an “X”
configuration at the rear. This paper addresses the
hydrodynamics as well as the autopilot design for this unique
underwater vehicle. A high fidelity, six-degree-of-freedom
(6-DoF) hydrodynamic and dynamics model was developed
for use in the hydrodynamic and autopilot design. The
performance analysis required frequency domain, linear Fig. 1. Isometric view of vehicle.
time-domain, and nonlinear time-domain models. The vehicle
model is a rigid body model consisting of individual geometry II. HYDRODYNAMIC DESIGN
components that react to the local angle of attack/sideslip at
each geometry component during nonlinear time domain VCT developed a high fidelity six-degree-of-freedom
simulations. All hydrodynamic models required for this hydrodynamic and dynamics model for this vehicle. The
simulation were developed by VCT. VCT utilized its generic vehicle model consists of individual geometry components
vehicle modeling environment, referred to as VCT Tools™, to that react to the local angle of attack/sideslip at each
develop and integrate the hydrodynamics of the vehicle, IMU, geometry component during nonlinear time domain
actuators and control system models that were required for simulations. Fourteen baseline configurations were
the 6-DoF vehicle simulation. Production cost requirements established during the design spiral.
limited the cost of the actuators and CPU to low frequency and Fig. 2 and Fig. 3 are the open-loop root locus diagrams
low update rate COTS components. Since the speed of the for the lateral and longitudinal planes, respectively. They
AUV is equivalent to that of a 100 knot submarine, the are plotted as a baseline for the hydrodynamic design
challenge with the autopilot design was to control the vehicle presentation that follows.
high frequency dynamics with these low cost and low
frequency components. The design approach used by VCT
allowed the vehicle to be controlled with an off the shelf 4 Hz
actuator and a 10 Hz update rate autopilot.
Imaginary Axis (rad/sec)

I. INTRODUCTION

The Navy is currently developing a small AUV that


consists of a rigid hull with counter rotating propellers on
the front and control fins arranged in an “X” configuration
at the rear. In order to meet the aggressive speed and
endurance requirements imposed by the Navy, the vehicle
consists largely of batteries. Because of the length and
diameter constraints, the vehicle is heavy in water. A wing
was incorporated into the design in order to support the
weight of the vehicle. This paper addresses the
hydrodynamics as well as the autopilot design for this
unique underwater vehicle. Fig. 1 depicts the vehicle in Real Axis (rad/sec)
3-D.
Fig. 2. Lateral root locus.

A. Drag-Brake Control Surfaces


An initial concept utilized four individual “drag-brake”
flaps for control. Twenty different drag-brake sizes were
investigated. The aerodynamics for a drag-type of control
effector differs from lifting fins in two important ways.
First, the deflection required for a drag-brake control
effector to produce a control moment is 60o to 90o, causing
the vehicle drag to be increased sigificantly. Second, the
control moment from a drag-brake is much smaller than that

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:48:44 UTC from IEEE Xplore. Restrictions apply.
produced by conventional lifting control surfaces. It was The vehicle is open-loop unstable in pitch for span
found that the drag-brake was ineffective except at angles lengths over intermediate ranges. Fin span has a significant
creating unacceptable levels of drag. For this reason, effect on the two real modes at -1 and -9 RAD/SEC. As the
conventional fins were used. span increases, these two roots initially couple into an
oscillatory pair, but then return to the real axis. For the
lateral plane, increasing the control fin span increases the
high frequency pole (the primary driver for the actuator
requirements) from -32 to -65 RAD/SEC requiring a trade off
between open-loop stability (large fins) and actuator
Imaginary Axis (rad/sec)

selection (small fins). Trading open-loop stability for


reduction in actuator requirements was the approach taken
for this design.
The final design parameter investigated was the fin
chord. As the chord is increased, the fin area increases, but
the fin aspect ratio decreases, causing some configurations
to become less effective as the chord length is increased. As
the fin span is increased the stern plane trim angle is
reduced. Increasing the fin chord length will also decrease
the trim angle. Since it is desirable to keep the stern plane
trim angle small so that the autopilot will have the
maximum amount of fin deflection available for disturbance
Real Axis (rad/sec)
rejection, the tradeoff between fin span and chord has a
large impact on the actuator requirements. However,
Fig. 3. Longitudinal root locus. further increases in the chord length cause the trim stern
plane angle to become larger. Increasing the span will shift
B. Wing Dihederal Angle the crossover point where increasing the chord length
An initial positive wing dihedral angle was used (wing actually makes the trim angle worse. The fin chord was
tips up). Parametrics on dihedral angle showed that the found to have a small effect on the roots while the fin span
vehicle Dutch roll mode becomes more lightly damped as has a large effect on the fastest pole. These trade-offs lead
the dihedral angle is increased (follows the same trend as to a final fin planform span and root chord design.
aircraft), while the roll mode becomes much faster. The fast For the longitudinal plane, these same fin parameters do
roll mode at ~100 rad/sec was expected to be problematic not affect the actuator requirements and have only a small
given the selected actuator. Reducing the dihedral angle effect on open-loop stability.
reduces this mode and, in combination with span design
yields acceptable rates. F. Weight and CG
The weight and Xcg of the vehicle were selected at the
C. Wing Span objective speed of the vehicle to optimize the control system
Longitudinal stability increases with wing span design.
achieving stability well within the imposed dimensional In order to stabilize the vehicle in the lateral plane,
constraints. Interestingly, larger spans lead to a much faster detailed design iterations on weight and Xcg were
roll root, which adversely impacts the interaction of the performed at the objective speed. Analyzing the open-loop
actuator with the rest of the control system. Decreasing the yaw-rate per rudder deflection revealed a large peak
wing span from the maximum acceptable dimension response for small Xcg values. This is shown in Fig. 4 for a
significantly reduces this mode although further reduction fixed weight minus displacement of the nominal design.
leads to an increase in trim angles in pitch and stern plane,
thereby increasing drag.

D. Wing Incidence Angle


The baseline wing incidence angle was investigated
over a range of positive and negative values to assess the
r per δr (dB)

impact on static and dynamic performance. The trim pitch


angle at the objective speed is increased as the incidence
angle becomes more negative. This in turn increases the
trim stern plane angle and drag. Increasing the incidence
angle can reduce the trim stern plane angle and reduces
damping of the 3.5 RAD/SEC oscillatory pitch mode as wing
incidence angle is increased.

E. Control Fin Span and Chord


With the wing span size set, parametrics were run to size the
control fins. Reducing the control fin span over the range Fig. 4. Bode plot of yaw-rate per rudder deflection
investigated increases the trim stern plane angle while the pitch
angle is not significantly affected. This reduction in fin span Fig. 5 shows the open-root loops that correspond with
also reduces the drag of the vehicle but has little effect on the Fig. 4. A final combination of net buoyancy and Xcg was
fastest pitch mode. selected as the baseline configuration for autopilot design.

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:48:44 UTC from IEEE Xplore. Restrictions apply.
the crossover frequency is 0.413 RAD/SEC.

Magnitude (dB)
Imaginary Lat (rad/sec)

Phase (deg)
Real Lat (rad/sec)

Fig. 5. Root locus diagram Frequency (RAD/SEC)


Fig. 7. Bode plot of the heading loop.

III. AUTOPILOT DESIGN Fig. 8 is the bode diagram of the roll loop. The gain
margin is 16 dB, the phase margin is 66 degrees, and the
The design of the autopilot for the vehicle is now crossover frequency is 0.158 RAD/SEC.
discussed. The primary functions of the autopilot are depth
keeping and heading control. The autopilot will run at 10
Hz on a small form-factor computer board. Off-the-shelf
actuators are utilized with a bandwidth of 4 Hz. An attitude
Magnitude (dB)

computation algorithm runs at 20 Hz using a low cost IMU.


A. Linear Design
The process to developing a successful autopilot was not
easy. First, the lateral root presented in Fig. 4 had to be
dealt with. Another hurdle to overcome was that the target
computer board required changing the autopilot code from
C++ to C.
Phase (deg)

The major blocks of the autopilot are the depth and pitch
controllers for the longitudinal plane, and the roll and
heading controllers for the lateral plane. There are also a
speed controller and fin mixing blocks.
Using VCT Tools™, the hydrodynamic configuration of
Frequency (RAD/SEC)
the vehicle was used to design the control system gains.
The next three figures depict the bode diagrams for the Fig. 8. Bode plot of the roll loop.
lateral plane.
1). Lateral Plane Bode Diagrams 2). Longitudinal Plane Bode Diagrams
Fig. 6 shows the bode diagram for the rudder loop. The The next three bode diagrams are for the longitudinal
gain margin is 12 dB, the phase margin is 89.5 degrees, and plane. Fig. 9 is for the stern plane loop. The gain margin is
the crossover frequency is 1.07 RAD/SEC. 11 dB, the phase margin is 61 degrees, and the crossover
frequency is 2.11 RAD/SEC.
Magnitude (dB)

Magnitude (dB)
Phase (deg)

Phase (deg)

Frequency (RAD/SEC)
Frequency (RAD/SEC)
Fig. 6. Bode plot of the rudder loop.
Fig. 9. Bode plot of the stern plane loop.
Fig. 7 is the bode diagram for the heading loop. The
gain margin is 17 dB, the phase margin is 59 degrees, and

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:48:44 UTC from IEEE Xplore. Restrictions apply.
Fig. 10 is for the pitch loop. The gain margin is 10 dB, 4). Longitudinal Plane Step Response
the phase margin is 55 degrees, and the crossover frequency Fig. 13 is a commanded step response for depth. The
is 1.52 RAD/SEC. vehicle is commanded to a depth excursion of 10 FT in black.
The depth response of the vehicle is shown in red.
Magnitude (dB)

Depth (ft)
Phase (deg)

Frequency (RAD/SEC)
Fig. 10. Bode plot of the pitch loop.
Time (sec)
Fig.11 is for the depth-pitch loop. The gain margin is 17 Fig. 13. Depth step response.
dB, the phase margin is 60 degrees, and the crossover
frequency is 0.342 RAD/SEC.

B. Non Linear Simulation


Fig. 14 is a simulated race track pattern that the vehicle
Magnitude (dB)

has flown at constant depth command. The vehicle is


initially commanded at a 0 DEG heading, and then 90, 180,
and 270 DEG. The black triangle in Fig. 14 depicts the
starting location of the vehicle as it proceeds on a heading of
0 DEG. The red dots mark the transition to the next
commanded heading. The drift in the heading is due to the
low cost IMU that is employed.
Phase (deg)

Both the heading (in black) and roll (in red) are plotted
in Fig. 15. Note that as the vehicle completes its clockwise
heading transitions, the vehicle rolls into the turn about 9
DEG.
Fig. 16 details the pitch during the simulation. There is
Frequency (RAD/SEC)
minimal pitch disturbance as the vehicle transitions from
Fig. 11. Bode plot of the depth-pitch loop. the heading commands. The time history of depth in Fig. 17
also reflects steadiness in the longitudinal plane during the
3). Lateral Plane Step Response turns.
Fig.12 is a commanded step response of the heading of
10 DEG, shown in blue. The vehicle’s heading is in red, and
its corresponding roll is in black.
Heading and Roll (deg)

Time (sec) Y

Fig. 12. Heading step response. Fig. 14. Simulated race track pattern

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:48:44 UTC from IEEE Xplore. Restrictions apply.
Heading and roll (deg)

Stern plane (deg)


Time (sec) Time (sec)
Fig. 15. Vehicle heading. Fig. 18. Vehicle stern plane.

Rudder and Aileron (deg)


Pitch (deg)

Time (sec) Time (sec)


Fig. 16. Vehicle pitch Fig. 19. Vehicle rudder and aileron.

The stern plane is shown in Fig. 18. Both the rudder (in IV. CONCLUSIONS
black) and aileron (in red) are plotted in Fig. 19. Only -0.3
DEG of rudder is required to turn the vehicle. A value of The hydrodynamic and autopilot design were presented
-0.55 DEG was used by the aileron to control the roll in the for an AUV. The challenge of fine tuning the open-loop
turn. dynamics of the vehicle was shown to overcome a large
frequency peak in the lateral plane. Gains were succesfully
designed to give the vehicle at least an 8 dB gain margin and
a 45 DEG phase margin. Finally, a non-linear simulation of
the vehicle proved its design.

© Copyright 2005 IEEE


Depth (FT)

Time (sec)
Fig. 17. Vehicle depth.

Authorized licensed use limited to: Central Mechanical Engineering Research Institute. Downloaded on October 21,2022 at 07:48:44 UTC from IEEE Xplore. Restrictions apply.

You might also like