A heuristic search algorithm for on-line system identification

The application of heuristic search techniques to system identification is described. The search algorithm is used for the development of ARMAX models of physical systems, and, in addition to heuristics used to search for appropriate polynomial orders, also contains heuristics for sampling rate selection, delay estimation, and model validation. When coded in rulebase form and coupled with an expert-supervised control architecture, a fully automated approach to modeling and control can be achieved. The development of the heuristic system identification algorithm is detailed, and a demonstration of the algorithm in the development of an ARMAX model of a sampled analog system is provided