DOA-Based 3D Tracking With Factor Graph Technique for a Multi-Sensor System

Abstract

In this paper, a direction-of-arrival (DOA)-based factor graph (FG) technique is proposed for three-dimensional (3D) tracking. Multiple sensors are utilized in this system, which could measure both azimuth and elevation DOAs emitted from an anonymous target. To realize non-linear tracking, a modified extended Kalman filter (EKF) is proposed. Specifically, on the one hand, the proposed EKF observer is no longer independent of the EKF predictor, but instead takes the predicted target location into its operation. With this technique, the accuracy of detection is improved while the computational complexity is dramatically reduced. On the other hand, the variance of the EKF observer error is estimated in real time, based on the predicted Cramer-Rao bound (PCRB). Therefore, the robustness of detection can be guaranteed even with an unstable sensing environment. In this sense, the EKF observer and the EKF predictor are operated in an integrated FG framework. The advantages of the proposed technique are verified by both complexity analyses and simulation results.