Monitoring the Kalman Gain Behavior for Maneuver Detection

Abstract

A different way to detect maneuvers in Kalman-filter-based target tracking is discussed. By examining the behavior of the Kalman gains, a target maneuver may be detected prior to the state change being detected. A first look at this approach is presented. The individual Kalman gains of the target tracking algorithm are monitored over time. Variations in the gains' behaviors such as zero crossings and short-term high frequency energy are seen to indicate the existence target maneuvers.