Final Year Units

geographertonguesAI and Robotics

Nov 30, 2013 (3 years and 11 months ago)

53 views

© University of Reading 2012

www.reading.ac.uk

School of Systems Engineering

Matlab Assessment for
Final Year Units

Victor M. Becerra

School of Systems Engineering

Matlab

Assessment of Final Year Units

Outline


Context


Contents


Software tools employed


Materials


Tutorials


Advantages


Disadvantages / Risks


Assessment


Concluding remarks


2

Matlab

Assessment of Final Year Units

Context



Advanced Control
is an M
-
level
module at Reading.


Compulsory in MEng Cybernetics,
year 4


Optional in MSc Cybernetics and
for the MEng Robotics.

3

Matlab

Assessment of Final Year Units

Context

Prior control knowledge by students includes:

-
Basic analysis in the Laplace domain

-
Basic frequency design methods

-
State space analysis and design

-
Basic knowledge of Matlab/Simulink


4

Matlab

Assessment of Final Year Units

Context


Prior control knowledge often includes:

-
Sampled data systems

-
Introductory optimal and nonlinear control

-
Introductory system identification

5

Matlab

Assessment of Final Year Units

Contents


20 credit module (40 contact hours)


Divided into four subtopics
:


Nonlinear programming


Robust
control


Nonlinear control


System identification (until 2011)


Currently 30 hours of lectures and 10 hours of
Matlab

based tutorials.



6

Matlab

Assessment of Final Year Units

Contents


Nonlinear programming

-
Unconstrained local
optimisation

-
Line Search

-
Quasi
-
Newton methods

-
Constrained local optimisation (SQP)

-
Evolutionary algorithms

-
Control applications

-
Controller tuning

-
Nonlinear optimal control

7

Matlab

Assessment of Final Year Units

Contents


Robust control:


Mixed sensitivity design


MIMO systems analysis


Uncertainty


Robustness


Structured singular value


H


control design


Robust PID control



8

Matlab

Assessment of Final Year Units

Contents


Nonlinear control


Lyapunov

stability concepts


La Salle invariant set theory


Model reference adaptive control


Feedback
linearisation


Sliding mode
control


Control
Lyapunov

functions


Backstepping


Written exam answers for this
part

9

Matlab

Assessment of Final Year Units

Software tools employed


Matlab


Simulink


Control systems toolbox


Optimisation toolbox


Robust control toolbox


Limited third party Matlab code (e.g.
differential evolution)

10

Matlab

Assessment of Final Year Units

Materials


Materials employed includes:


Lecture notes


Tutorial questions and solutions


Matlab/Simulink code


Past exam questions and solutions

11

Matlab

Assessment of Final Year Units

Tutorials



10 hours of computer based tutorials


Each student is assigned a PC during the tutorial.


Typically one design question is given for each
tutorial.


The lecturer is available to answer questions and
to provide guidance to the students.

12

Matlab

Assessment of Final Year Units

Assessment


3 hour exam consists of 6
questions, 2 for each section


One question from each
section must be answered


Four of the questions require
the use of Matlab.


Four Matlab based questions
are worth 30 marks each, the
other two are worth 20 marks
each.


13

Matlab

Assessment of Final Year Units

Assessment


Open book


Network connection is mostly disabled during
the exam.


Only connections to the
Matlab

license server and Blackboard
are allowed.


Code reuse is possible


If reusing code, student must modify the code to attempt the
answer, otherwise no marks.



14

Matlab

Assessment of Final Year Units

Assessment


Answers are returned in the form of


Matlab

scripts for each question,


MS
-
Word
documents with
Matlab

plots and other output,


written
comment on
answer books
.


Digital files produced by the students are uploaded using
Blackboard.


These files are later printed by the examiner and joined
with the answer books prior
to marking
.


Printing may be tedious but avoids disruption during the exam.



15

Matlab

Assessment of Final Year Units

Advantages


The students appreciate the challenge.


It is possible to ask realistic design questions which
would be impossible to consider otherwise.


It is still possible to assess analytical skills.


Students are forced to learn the use of software tools
that are relevant to industry.



16

Matlab

Assessment of Final Year Units

Disadvantages / Risks


It usually takes additional time /
effort to prepare the questions
and model answers.


Sufficient
number
Matlab

and
toolbox
licenses is required.


Risk of power failure during exam


Risk of software bugs
affecting
the
students


Either
Matlab

or the operating system may
crash or become irresponsive.


17

Matlab

Assessment of Final Year Units

Disadvantages / Risks


Extra PC’s should be on standby


Work space of PC labs not always ideal for
written answers


More difficult with large groups. So far groups of
up to 18 students have been handled without
problems

18

Matlab

Assessment of Final Year Units

Concluding remarks


Matlab based exams running for
11 years


Overall the experience has been
positive, which is reflected by
feedback from the students


External examiners have
consistently praised the use of
computer based assessment and
have suggested its use in other
modules


19

Matlab

Assessment of Final Year Units

Concluding remarks


Three other modules taught at the School now
employ
Matlab

based assessment:


Nonlinear and Optimal Control,


Signal Processing,


Medical Image Processing.


A new MSc module “
Fundamentals of Control
Systems
” will use
Matlab

based assessment from
2013.


20