Investigation of moving-bank multiple model adaptive algorithms

The feasibility of a moving-bank multiple model adaptive estimator/controller is examined. Compared to a conventional full-bank multiple model adaptive algorithm, a significant reduction in the number of required elemental filters is accomplished through a dynamic re-declaration of the positions in parameter space that the elemental filters occupy. Critical to the performance of the moving-bank estimator is the decision method that governs movement of the bank of elemental filters. Five such methods are developed and their performances are compared to each other and to a benchmark filter with artificial knowledge of the true parameter value. Three adaptive controller algorithms are also generated and evaluated. A simple but physically motivated example is used to gain insights into relative performance potential of the proposed algorithms.