Simultaneous Track-to-Track Association and Bias Removal Using Multistart Local Search

A fundamental problem in multisensor data fusion is associating data from different sensor systems in the presence of sensor bias, random errors, false alarms, and missed detections. In this paper, we address the problem of simultaneously performing track-to-track association and bias estimation. We describe a polynomial-time multistart local search heuristic for quickly generating a number of high quality bias-assignment hypotheses. Computational results are presented to illustrate our findings and to compare our proposed algorithm to its competitors. Numerous test cases suggest that our multistart heuristic is suitable for real-time application and offers a practical and important "middle ground" for this problem by balancing performance and computation time.