ABSTRACT
Estimation of contact motion is based on observations of some aspects of that motion. The prerequisite maneuver introduces added complexity to the problem, makes the filter prone to divergence, and renders the estimation process sensitive to the choice of coordinate system. The Kalman filter has been designed under the assumption that the contact maintains a constant course and speed. Thus, when the contact maneuvers, a system modeling bias error develops which invariably causes filter divergence. A maneuver causes a decorrelation between observations and hypothesized parameters in the dynamic model. The chapter emphasizes the application of the extended Kalman filter using the modified polar coordinate. Batch algorithms that derive from a Gauss procedure have been effective when dealing with the long-range problem where the bearing rate is very low. Then, system observability is low and linearization errors in the Kalman filter take on added significance.
