An optimal two-stage algorithm for highly maneuvering targets tracking

A new general two-stage algorithm which extends the target dynamic model to higher order time-derivatives of acceleration is proposed. The conventional input estimation techniques assume constant acceleration level and there are uncovered a generalized acceleration modeling. In contrast, the augmented algorithms, which are based on the jerk modeling, are computationally expensive. In this paper, an innovative scheme is developed to overcome these drawbacks by using a new generalized dynamic modeling of acceleration and an optimal two-stage modified Kalman filter. The proposed scheme is based on the basic Fisher and Bayesian uncertainty models. The optimality of the proposed two-stage modified Kalman filter is proved. Comparisons with two well known approaches using Monte Carlo simulation show that the proposed scheme has a computational advantage over the augmented algorithms and also more significant improvement than the input estimation techniques.