A robust fusion algorithm for multi-sensor tracking

This paper presents a technique for the online adaptive weighted fusion algorithm for multi-sensor tracking. A suitable method for estimating measurement noise variance is developed, and the formulas of fuzzy logic-based fusion and parallel adaptive tracking are derived. The algorithm consists of three steps: (i) estimation of the sensor's measurement noise variance using the statistical theory; (ii) adjustment of the fused sensor's weight coefficient according to the sensor's noise variance change; (iii) prediction of the target position using the "current" statistical model and Kalman filter method. The algorithm is able to adapt itself to the changes of sensor's noise, and its estimation error is of least mean square. Computer simulation results are presented to demonstrate the robust performance of this algorithm.