A robust tracker with real-time input estimation

A modified input estimation approach for tracking a manoeuvering target is developed by Xinxi (1996). The equation of increment of acceleration is described as a first-order autoregressive process. The increment of acceleration is estimated in real time by the residual of a standard Kalman filter and then is used to codify the target states. It has a good tracking property for an abrupt and highly manoeuvering target. But for an abrupt manoeuvering target, because of the large residual variation, the acceleration estimate may be too large at the manoeuver spot. A robust tracker is developed to reduce the variation of estimated acceleration. It is proved efficient to use the approach to improve the tracking of abrupt and highly manoeuvering targets.