A Weight-Based Multi-Innovation Intelligent Tracking Algorithm with Maneuver Detection

Kalman filter is an unbiased minimum variance estimator under the linear, Gaussian circumstance. However, the robustness of Kalman filter is underdeveloped because it strongly depends on the system model and the performance may seriously deteriorate if the model turns erroneous or the target maneuvers unexpectedly. To solve this problem, a weight-based multi-innovation intelligent correction algorithm with maneuver detection is proposed. Firstly, an intelligent maneuver detection algorithm called double error ellipses is presented by setting a bilateral threshold to the error ellipse, which is able to detect the maneuver occasion by improving the sensitivity to the maneuver. Secondly, we compute the weighs of the multi-innovations through building a projection to decrease the maneuver deviations and the model errors. Particularly, the projection is constructed with respect to the Euclidean distance between the predicted and effective measurement in two-time steps. Finally, a new target tracking algorithm weight-based multi-innovation correction algorithm with maneuver detection is derived by combining the above maneuver detection approach with the multi-innovation correction method. The simulation results show the proposed algorithm is superior to the adaptive Kalman filter and the conventional Kalman filter in cost-effectiveness.