Constrained random search for IIR adaptive and non-linear filters

A constrained random search algorithm is introduced to perform system identification for both linear and nonlinear unknown systems. This proposed method is a population based algorithm involving a structured yet randomised information exchange resulting in survival of fittest among the population. It has capabilities of reaching the global solution. It is demonstrated that the proposed algorithm works well on multimodal error surfaces that are characteristic of linear IIR and general nonlinear adaptive filters.