Prognostic Target Tracking Accuracy of the Linearized Model Identified by the Neural Extended Kalman Filter

The neural extended Kalman filter is a technique that learns unmodeled dynamics while performing state estimation in the feedback loop of a control system. This coupled system performs the standard estimation of the states of the plant while estimating a function to approximate the difference between the given statecoupling function model and the behavior of the true plant dynamics. At each sample step, this new model is added to the existing model to improve the state estimate. To intercept a tracked target, the motion model of the target can be used to prognosticate the location of the target at given point in the future. In this paper, a investigation into the improved accuracy in the predicted target trajectory is investigated. The neural extended Kalman filter provides a motion model that will be linearized. This linearized model then is used to prognosticate the target motion to a specified time in the future. Three separate motion trajectories are used in this investigation. The results provide a quality measure of the predictive capabilities of the neural extended Kalman filter tracker compared to that of a standard Kalman filter tracker.