lk

© All Rights Reserved

747 views

lk

© All Rights Reserved

- SOP Robotics MIT
- Quiz 3 Solutions
- Quiz 5 Solutions
- Quiz 2 Solutions
- Quiz 1 Solutions
- bldc
- Quiz 6 Solutions
- Quiz 7 Solutions
- An Introduction to Mathematical Optimal Control Theory
- Inverted Pendulum State-Space Methods for Controller Design
- Homework Solution
- PID_Control & Ziegler-Nichols
- Modeling and Analysis of Model Refrence Adaptive Control.pptx
- Math 1ZC3 Custom Courseware 2012
- UT Dallas Syllabus for engr2300.001.11f taught by Jung Lee (jls032000)
- Lecture7 Design
- Sle Matlab
- Grimm - Linear LMI-Based External Anti-windup Augmentation for Stable Linear Systems
- Online Statement of Purpose Accepted ONLY
- 2002-07-04

You are on page 1of 4

1

Which of the following statements is correct?

SOLUTION

Lets go through the statements one by one:

When designing observer-based controllers, the observer must be slower than the controller. The statefeedback controller will only become effective once the estimate of the state is reasonably good, i.e., it

makes sense to make the observer faster than the controller. As such, the statement is false.

The separation principle tells us that the control design and the observer design cannot be done independently of each other. Actually, it tells us the complete opposite that control and observer design can

be decoupled. The statement is false.

The eigenvalues to a general system x = Ax + Bu can always be placed anywhere using state-feedback.

This one is also false. Arbitrary eigenvalue-placement is only possible when the system is completely

controllable.

If a linear system is completely controllable the state can be made to follow any trajectory. If it is

completely controllable then the system can move between any two states. But, it can not move (in

general) along any trajectory. So this one is also false.

If linear system is completely controllable the state can be made to go between any two initial and final

values. As stated above, this statement is true, and hence this is the correct answer.

2

Suppose you have a robot whose dynamics are

0

x =

0

1

0

x+

0

1

u,

and you go to the sensor store to buy a sensor. Each sensor comes with a corresponding output matrix C,

such that y = Cx. Which of the following sensors/C-matrices should you not buy? (Hint: Observability

seems to be a useful property in a sensor...)

SOLUTION

What we need to do is ensure that the sensor is good enough, i.e., that it allows us to reconstruct the

state. Mathematically speaking, this means that we need to check if the resulting system is completely

observable. The observability matrix is

C

=

CA

and this second order system is completely observable if and only if rank() = 2.

Lets go through the choices one by one:

C=

1 0

0 1

which has rank 2 (both columns are linearly independent), i.e., this sensor is good.

1 1

C= 1 1

=

,

0 1

which has rank 2, i.e., this sensor is also good.

C=

0 1

0 0

which has rank 1 since the first column can be obtained by multiplying the second column by zero. As a

consequence, this sensor is no good, and this C-matrix is the correct answer.

1 0

C = 1 0

=

,

0 1

which has rank 2, i.e., this sensor is good.

C=

1 1

1 1

0 1

3

Consider the scalar system

x = 2x + u.

We decide to try three different state feedback controllers u = ki x, i = 1, 2, 3, where

k1 = 1, k2 = 3, k3 = 5.

The different closed-loop system behaviors are shown in the figure below. Your job is to identify which

feedback controller was used to produce which plot. (The upper trajectory is referred to as solid, the

middle one as dotted, and the lower one as dashed.)

2.5

x(t)

1.5

0.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

SOLUTION

k1 = 1 gives the closed-loop system x = 2x x = x. Since this is a scalar system, the eigenvalue of the

A-matrix (which is now just a scalar) is the same as the value itself, i.e., the eigenvalue is 1. But, this

means that the system is unstable and, as such, the solid line corresponds to k1 .

k2 = 3 x = 2x3x = x, which is stable with eigenvalue 1. Similarly, k3 = 5 x = 2x5x = 3x,

which is stable with eigenvalue 3. As such, k3 gives a faster response than k2 (since the eigenvalue is

further away from the imaginary axis) and, as such, the dotted line corresponds to k2 and the dashed

one to k3 .

4

An attempt at a software implementation of an observer-based feedback controller is given below (in

some sort of pseudo-code):

% Given: dt (sample time), A,B,C (system matrics), K,L (gains)

% At each iteration, compute the following:

y=read_in_output_vaule;

u=-K*x_tilde;

x_tilde=update_estimator_value(x_tilde,u,y);

send_control_value(u);

Which of the following versions of update estimator value(x tilde,u,y) is correct?

SOLUTION

We know that the observer dynamics are given by

x

= A

x + Bu + L(y C x

),

3

so the update program must reflect this fact. Moreover, we need to implement this in discrete time so

the standard Euler approximation of this equation becomes

x

k+1 = x

k + t(A

xk + Buk + L(yk C x

k )),

where t is the sample time. As such, the correct answer is:

return x_tilde+dt*(A*x_tilde+B*u+L*(y-C*x_tilde));

5

We have seen that when using pole-placement, the larger (in magnitude) the closed-loop eigenvalues are,

the faster the response. So why not place all eigenvalues in 10, 000, 000, 000 and be done with it? Well,

there are problems with making the eigenvalues too large in magnitude.

Assume that we are designing a state-feedback, go-to-goal behavior for a mobile robot by stabilizing

the system around the goal point using pole-placement. Which of the following is a valid concern when

making the eigenvalues too large?

The actuators may saturate making the stability analysis invalid.

Fast and aggressive maneuvers lead to significant odometric drift.

Modeling errors get amplified by large control gains, which may deteriorate the performance.

Fast and aggressive maneuvers increase the risk of not being able to detect obstacles in the environment

quickly enough.

They all are valid concerns.

SOLUTION

Making the eigenvalues too large means that the system will move very quickly and use very large

actuation signals (since the control gains will be really large). This would certainly saturate the actuators.

Also, when moving and turning quickly, the robot will slip and skid and, as such, the odometric drift will

be quite bad. As a consequence of moving too quickly, the robot may also not be able to respond in time

to obstacles. And, finally, large gains means more amplified responses. And, since modeling errors can

be thought of as disturbances, the response to these disturbances will also be amplified.

Answer: they are all valid concerns.

- SOP Robotics MITUploaded bySoumya Chatterjee
- Quiz 3 SolutionsUploaded byyashar2500
- Quiz 5 SolutionsUploaded byyashar2500
- Quiz 2 SolutionsUploaded byIsaac Kuma Yeboah
- Quiz 1 SolutionsUploaded byErick Felipe
- bldcUploaded bythulasi_krishna
- Quiz 6 SolutionsUploaded byyashar2500
- Quiz 7 SolutionsUploaded byyashar2500
- An Introduction to Mathematical Optimal Control TheoryUploaded byAndré Vidal
- Inverted Pendulum State-Space Methods for Controller DesignUploaded byReyes_n
- Homework SolutionUploaded byyashar2500
- PID_Control & Ziegler-NicholsUploaded byHachemi Bhe
- Modeling and Analysis of Model Refrence Adaptive Control.pptxUploaded bydewe sun
- Math 1ZC3 Custom Courseware 2012Uploaded byVytautas Stasiulevicius
- UT Dallas Syllabus for engr2300.001.11f taught by Jung Lee (jls032000)Uploaded byUT Dallas Provost's Technology Group
- Lecture7 DesignUploaded bytasos7639
- Sle MatlabUploaded bymir emmett
- Grimm - Linear LMI-Based External Anti-windup Augmentation for Stable Linear SystemsUploaded byRiccardo
- Online Statement of Purpose Accepted ONLYUploaded byEthan Carlisle Couto Archibald
- 2002-07-04Uploaded byapi-168623254
- Control systems 2 take home assignmentUploaded byNipuna Thushara Wijesekara
- Custom 9Uploaded bykmd_venkatsubbu
- pset3-solnUploaded byMohamed Medhat
- Tensor Decomp PresentationUploaded byMahbod Matt Olfat
- Bms & Hvac SystemUploaded byMostafa Ali
- 296888784-Quiz-1-Solutions.pdfUploaded byMohammad Ghazi Vakili
- DocUploaded byArjun Hajong
- le2Uploaded bysatyajit
- Coursera-control-of-mobile-robots-quiz2-solutionsUploaded byetiknorc
- 368592327-Chapter-2-mathematical-models-of-systems-1-pptx.pptxUploaded byHoras Nies

- Quiz0b FeedbackUploaded byjcvoscrib
- 1.2eUploaded byjcvoscrib
- 1.2cUploaded byjcvoscrib
- 1.2dUploaded byjcvoscrib
- Lesson 10 ResourcesoUploaded byjcvoscrib
- miniproject1 rpslsUploaded byjcvoscrib
- 1.2c_solUploaded byjcvoscrib
- 1.2bUploaded byjcvoscrib
- Privacy PolicyfUploaded byjcvoscrib
- 1.2d_solUploaded byjcvoscrib
- 1.2a_mainUploaded byjcvoscrib
- 1.2fUploaded byjcvoscrib
- 1.2e_solUploaded byjcvoscrib
- 1.2gUploaded byjcvoscrib
- 1.2aUploaded byjcvoscrib
- miniproject 3Uploaded byjcvoscrib
- Source Videos Week3-Redo Canvas and DrawingUploaded byLuca Kab
- Quiz2b FeedbackUploaded byjcvoscrib
- Quiz2a FeedbackUploaded byjcvoscrib
- Programas de Las LeccionesUploaded byjcvoscrib
- Mini-project # 2 - Guess the NumberUploaded byjcvoscrib
- Quiz1b FeedbackUploaded byjcvoscrib
- Quiz1b Rograms for the QuizUploaded byjcvoscrib
- Ludo GraphyoUploaded byjcvoscrib
- Lesson 9 ResourcesoUploaded byjcvoscrib
- Lesson 7 ResourcesoUploaded byjcvoscrib

- Modern Control Engineering Problems Ch 8.pdfUploaded bymahendra shakya
- CLMS Cover Page GuideUploaded byLeinherij Xd
- Reengineering 1Uploaded byD Attitude Kid
- Chapter 10Uploaded byTeja Ravi
- ThermodynamicsUploaded byeaglebaby11
- BIOL353 Course Outline 2015 Easter-Break-AdjustedUploaded byImene Kouidmir
- Kaashoek - Robust Control of Linear Sistem and Nonlinear ControlUploaded byMuhidin Arifin
- Dependability DefnUploaded byDina Tauhida
- Chapter 1 QuizUploaded byMarZiYa
- B307 Lect9Uploaded byManuel Barrios
- Software_Quality_and_Assurance (1).pdfUploaded byobiad83
- Links for Production Planning and ControlUploaded bymehatajmosiq
- SAD AssignmentUploaded bystraticblaze
- PID ControllerUploaded byDuy Hai
- Motor Speed_ State-Space Methods for Controller DesignUploaded byParitosh Kumar Velalam
- 5. Process ControlUploaded bydhirajsingh_avit
- Understanding SimulationUploaded byArvin F. Villodres
- Scope ManagementUploaded byashok_k_alexander
- Robot Manipulators Trends and DevelopmentUploaded byJosé Ramírez
- 2015_An Estimation of Distribution Algorithm With Cheap and Expensive Local SearchUploaded bymabdulkadhar
- Sintonizacion IMC NO Integrantes IntegrantesUploaded byEstefany Salazar Vidal
- Systems DevtUploaded byjamisss
- Fundamental PragmaticsUploaded byhachuyen
- Control Systems SyllabusUploaded bydnow4p
- SOA Terminology OverviewUploaded byapi-3727274
- CH17Uploaded byMuchammad Fauzi
- V-Force in Action2Uploaded byron_ski
- Software Testing TutorialUploaded bykondalb
- Process FMEAUploaded byDevspring
- Steps to Developing a Project ScheduleUploaded byUjjal Regmi

## Much more than documents.

Discover everything Scribd has to offer, including books and audiobooks from major publishers.

Cancel anytime.