Optimization and Systems Theory Seminar
Friday, March 31, 2000, 11.00-12.00, Room 3721, Lindstedtsv. 25

Alessandro Chiuso
Division of Optimization and Systems Theory
Department of Mathematics
E-mail: alessandro.chiuso@math.kth.se

Geometric subspace identification and stochastic realization with inputs

In this talk geometric subspace methods for the identification of linear time invariant systems in the presence of exogenous inputs will be discussed. These methods may be seen, roughly speaking, as the translation into algorithms which works on data of ideas from stochastic realization theory. Therefore geometric stochastic realization for system with inputs will be discussed and some results from stochastic realization theory will be extended to this case. Since in practice only a finite amount of data is available one has to resort to finite interval realization, which will be discussed in some detail.

Some common subspace algorithms may become rather ill-conditioned in some situations, yet there is experimental evidence of poor behaviour in some situations. We shall therefore discuss to some extent which situations may lead to poor behaviour (in particular the link with the input process). We present a possible remedy which shows better performances in some ill conditioned cases and show simulation results.

Calendar of seminars
Last update: March 23, 2000 by Anders Forsgren, anders.forsgren@math.kth.se.