Decentralized processing for multitarget motion analysis

Track estimation of targets from passive-sensor data is one of the typical and hard applications in both distributed artificial intelligence and distributed sensor networks. Multitarget motion analysis, where there is more than one target, is to associate targets and sensor data, and estimate target tracks based on that association. This is an NP-hard problem in general, and solved using stepwise relaxation. However, it is hard to obtain the optimal solution, or in other words, to locate the global optimum out of many local optima in the search space. This paper proposes a new approach to improve estimation, decentralized cooperative search using several processors. Simulation shows this approach achieves almost the same estimation as a stochastic relaxation based on simulated annealing, and much better performance.