Solution to a multisensor tracking problem with sensor registration errors

An on-line solution to the aircraft tracking problem with a distributed network of 3D sensors that are imperfectly registered is presented. Errors in the relative positions and orientations of the sensors are modelled and estimated by an extended Kalman filter [EKF], along with the track variables. This optimal solution, tractable for a limited number of targets, is compared through Monte-Carlo simulation with a suboptimal filter that neglects the sensor systematic errors. The registration errors substantially degrade the accuracy of the aircraft position estimates. The performance of the optimal filter depends on whether the registration errors are to be determined in an absolute or relative setting. When the sensor systematic errors are estimated relative to one of the sensors, the optimal filter performance is excellent. The simulation results are confirmed by a covariance analysis of the optimal filters