A motion model for tracking highly maneuvering targets

Mehrotra developed a jerk model for tracking highly maneuvering targets in 1997, which include terms at the most up to the third order derivatives of target position. The model is investigated in this paper. By theoretical analysis, it is shown that the filter, which based on the jerk model, may suffer from deterministic steady state estimation deviations. To find a way out of this question, a current statistic jerk model, for short cs-jerk, is developed, in which the jerk maneuvering is assumed to be an exponential correlated stochastic process with non-zero mean. It consists of the cs-jerk model of target motion, and a tracking filter with compatible order. The stable performance of the cs-jerk model is also analyzed and the result indicates that the cs-jerk model eliminates performance limitation of the jerk model. The improved performance of the cs-jerk model over the jerk model is illustrated through simulation.