Digital and Hybrid Simulation of a Bayes-Optimal Nonlinear Filter

摘要

In the past several years there has been a considerable amount of research in the area of optimal nonlinear filtering. Most of this effort has been restricted to theoretical studies. Recently, however, several digital simulation studies have been reported. These studies deal primarily with explicit examples of optimal nonlinear filtering realizations. As pointed out by Bucy studies of this sort are of importance in obtaining insight into the characteristics of optimal nonlinear filters. In the paper one investigates a particular nonlinear filtering problem. One uses Bucy's representation theorem to synthesize the optimal filter. The plant, measurement process, and noise signals are simulated with the aid of a digital computer. The optimal filtering algorithm is implemented both digitally and via hybrid techniques. The hybrid system used in these studies includes an EAI 680 analog computer and an XDS-SIGMA 5 digital computer. These various filtering implementations are subjected to a Monte Carlo analysis, and are then compared on the bases of speed and accuracy. (Author)