GENNA - A system for modelling adaptive systems with genetic and constrained identificaton

In the paper some new concepts concerning identification in adaptive control systems are discussed. A tool for modelling adaptive control systems with several modern identification and control options - called GENNA - is described in the paper. The tool enables to simulate standard modern estimation and adaptive control algorithms as well as some interesting alternatives - like neural networks and fuzzy rules based systems for control, and genetic algorithms and constrained identification algorithms for parameter estimation. Genetic algorithms are used for structure identification as well as for parameter estimation, especially in the case of a'priori knowledge about the adaptive control systems behaviour. The constrained identification is based on some concepts from Constraint Logic Programming methods and is used in the GENNA tool for determining the identified system structure and parameters using the a'priori knowledge as constraints. Typical predictive control algorithms were used in the simulation experiments, in the SISO as well as in multivariable framework. The adaptive predictive control algorithms with genetic and constrained identification were tested using typical single input - single output and multivariable plant models.