Application of Genetic Algorithms to Observer Kalman Filter Identification

Abstract

In this paper several applications of genetic algorithm (GA) as an aid to the system identification process are presented. First, GAs are used in a set of covariance-based optimum input signal algorithms using a proposed architecture suitable for online system identification. The optimal signals are computed recursively using a predictive filter. The efficiency of these algorithms are compared based on a set of simulations. Second, a novel input design for a two-step identification scheme is presented. Constraint systems, such as commonly found in structural and biomedical engineering applications, are considered for the input design algorithm. This paper presents a novel approach that induces a learning scheme into the input design computation and allows for considerations of the given constraints of the system. The optimization of the new input signal is accomplished using a simple elitism based genetic algorithm. Simulation results indicate the proposed piecewise adaptive input design algorithm performs well compared to the general white-noise-based estimation results. In the third portion of this paper proof is given that no dynamic controller can reduce the noise influence in linear system identification. A new selection scheme of the corresponding singular values is proposed for the eigensystem realization portion of the Observer Kalman filter IDentification algorithm in noisy systems. The selection is done using a GA. Simulation results of the proposed algorithm in comparison with the traditional used method are presented. The results indicate an improved ability to extract system models from highly noise corrupted data.