Multiple-model estimation with variable structure: some theoretical considerations

Existing multiple-model (MM) estimation algorithms have a fixed structure-they use a fixed set of models. An important fact that has been overlooked is that the performance of these algorithms depends largely on the set of models used. The limitations with the existing fixed structure MM algorithms are first addressed. In particular, it is shown theoretically that use of more models does not guarantee better performance (actually, it may yield even poorer results), apart from the increase in computation. This paper then presents theoretical results pertaining to the two ways of overcoming the limitations of the fixed structure algorithms: selection/construction of a better set of models and adoption of a variable set of models in contrast to the past and current efforts of developing better implementable fixed structure estimators