Radar Sensor Fusion via Federated Unscented Kalman Filter

Abstract

In this paper, a three-dimensional maneuvering target tracking algorithm in which radar sensors their tracks are fused at a federated Unscented Kalman Filter (UKF), has been presented. The federated filtering is composed of three levels namely sensors, local processors and a global processor. In the sensor-level, target range, azimuth and elevation are all measured by radar sensors in the Sphere Coordinate System (SCS). Each local processor uses UKF to proceed state estimation in the Reference Cartesian Coordinate System (RCCS). Meanwhile, the UKF processes the recursion and update of the state vector and the error covariance matrix through the unscented transformations. Finally, the state of each local processor is transmitted to the global processor for fusing as a final track for system output and information feedback. Two tracking schemes based on SCS (non-linear model) and RCCS (pseudo-linear model) measurements have been studied and their performance evaluated using simulation data. It is concluded that federated UKF processing in the nonlinear model has computational effectiveness and supplies almost the same tracking accuracy compared with federated UKF processing in the pseudo-linear model.