A Gaussian Process Regression Approach for Point Target Tracking

Abstract

Target tracking performance relies on the match between the tracker motion model and the unknown target dynamics. The performance of these model-based trackers degrades when there is a mismatch between the model and the target motion. In this paper, a Gaussian process based approach, namely, Gaussian process motion tracker (GPMT) is proposed. The Gaussian process framework is flexible and can represent an infinite number of motion modes. The evaluation of the proposed approach is performed on challenging scenarios and is compared with popular single and multiple-model based approaches. The results show high accuracy of the predicted and estimated target position and velocity over challenging maneuver scenarios.