Matlab Invariant Set Toolbox

Current version: 0.10.5
Date: 19 April 2005

Download: zip or gzip file.

This toolbox has been tested with Matlab Version 5.3 and the Optimization Toolbox Version 2.0. It should, however, still be compatible with the latest versions.

Version History

0.10.5, 19 April 2005. Bug fix.
0.10.4, 10 August 2001. Bug fix.
0.10.3, 25 July 2001. Bug fix.
0.10.2, 20 July 2001. Bug fix.
0.10.1, 18 July 2001. New function and bug fix.
0.10, 23 March 2001. First public release.
0.01, 29 November 2000. Used in preparation of PhD thesis.

New Features

Version 0.10.1

isbounded.m - Quick check to determine whether a given polyhedron is bounded.

Bug Fixes

Version 0.10.5

pdiff.m If Pontryagin difference is empty, the function now returns [] and displays a warning.

Version 0.10.4

isrobposinv.m Corrected a typo in line 43 - 1 on rhs of inequality was replaced with r(i).

Version 0.10.3

isrobposinv.m - A cosmetic change to the help information and added a small tolerance for checking whether a constraint is redundant.
oinfdistcl.m - Corrected problem which occurs when C is not the identity matrix.

Version 0.10.2

kinfset.m - Lines 71-76 put back in from v0.10. This was not the source of a previously reported bug.
k1set.m - Line 81 changed from Q = polymap(inv(A),polysum(T,polymap(-B,U))) to Q = polysum(T,polymap(-B,U)). This is because the mapping inv(A) on the Minkowski sum was implemented again two lines later on in v0.10. This caused problems if k1set was used to recursively compute the sets, rather than kinfset/sinfset/cinfset. This implementation is preferred because it does not require A to be invertible.

Version 0.10.1

kinfset.m - Lines 71-76 commented out from v0.10.
oinfdistcl.m - Argument tmax left out in call to oinfdist.m in v0.10.

Notes and Help

See Appendix E of my PhD thesis for an introduction to the use of the toolbox.

NB: In order to simplify implementation, the software makes the assumption throughout that all the sets in the arguments are bounded and non-empty. If, for example, you do not want to have constraints on all of the states, define a very large bounded set X = symineq(1e6*ones(1,n)). Similarly, where necessary the software assumes that A, B, E etc. are defined and non-empty. Correct behaviour of the code is not guaranteed if these assumptions do not hold.

Disclaimer

Please note that this software is experimental and non-commercial. The only reason I am making it available is for other researchers to try out some of the set invariance ideas presented in my thesis. It is the user's responsibilty in assessing the correctness of the theory and software implementation before putting it to use in their own research or exploiting the results commercially.

I am, however, very happy to answer any questions and investigate any bug reports.

[ Cambridge University | CUED | Control Engineering Group ]

Eric Kerrigan- eck21@cam.ac.uk