Variable Dimension Filter for Maneuvering Target Tracking

Abstract

A novel approach to tracking a maneuvering target is developed. This approach does not rely on a statistical description of the maneuver as a random process. Instead, the state model for the target is changed by introducing extra state components when a maneuver is detected. The maneuver, modeled as an acceleration, is estimated recursively. The performance of this estimator is shown to be superior to a recent algorithm presented by Chan et al. that handles the maneuver by estimating it as an unknown input. A significant departure from the current practice of comparison of algorithms is made: a recently introduced rigorous statistical methodology is used in the comparison of these estimators.