Adaptive Parameter Robust Estimation

In this paper, we describe an adaptive technique for states and parameter estimation involving a combination of two methods, namely the Variable Structure Filter (VSF) and the Extend Kalman Filters (EKF). The VSF concept is a model-based robust state/parameter estimation. It has a secondary set of uncertainties that provide a measure of uncertainties in the filter model. It is not however an optimal method. When combined with the Kalman Filter, it provides near optimal solution (further to the assumption pertaining to the Kalman Filter). The combined strategy would then also benefit from the robustness and the additional indicators of performance of the VSF. These features of the combined strategy used for removing uncertainties in the estimation process by dynamic adaptation of the filter model. The modeling uncertainties in the combined VSF/EKF method are removed by using two forms of Neural Networks adaptation. These adaptation methods are based on the Simultaneous Perturbation Stochastic Approximation (SPSA) and the Algorithm Of Pattern Extraction (ALOPEX). The use of dynamic adaption can significantly improve the performance of the estimation process. Other attractive features include computational simplicity, fast rate of convergence, robustness and stability.