Nonlinear filtering (5 credits)

lecturer: Professor Miroslav Simandl, University of West Bohemia in Pilsen,

schedule: Lectures and tutorials 3hrs/day (9am-noon) over 10 weekdays starting Monday 7 Aug and ending Friday 18 Aug 2006.

place: Tampere University of Technology, Tietotalo building, room Tb214.

Aug 7
Structural and probabilistic modelling, parameter estimation, state estimation, the Bayesian approach, statistical approach, the Bayesian recursive relation as a general solution of state estimation problem for stochastic system in discrete state space representation
Aug 8
State estimation of linear Gaussian system, derivation of the Kalman filter from the Bayesian recursive relation, linear filtering, what happens in case of nonlinear system or nongaussian random variables, classification of nonlinear filters, local and global filters
Aug 9
The Extended Kalman Filter, the Iteration Filter, the Second Order Filter
Aug 10
Derivative-free filters: the Unscented Kalman filter, the Difference filter
Aug 11
Aug 14
Analytical approach, the Gaussian sum filter, Gaussian sum representation, linear system with nongaussian noises, nonlinear system with Gaussian noises, special cases, abrupt changes of parameters, outliers
Aug 15
Simulation approach, simulation Monte Carlo methods, particle filters
Aug 16
Multi model approach and the Gaussian sum method, applications
Aug 17
Numerical approach, basic and advanced point mass methods
Aug 18
The Cramer Rao bounds as a tool for quality evaluation of the nonlinear filters
prerequisite: SGN-2606 or equivalent

literature: course notes, lecture audio files

credit requirements: written exam (Aug 24, 1pm, Tb214)

course registration: please contact the course coordinator Robert Piché <> .