An Efficient Computational Approach for Multitarget Tracking from Bearings-only Measurements by Decentralized Cooperative Processing

This paper firstly has proved that the well known Hungarian type assignment algorithms [14, 4] embedded in the relaxation-based maximum-likelihood (ML) solution for a bearings-only passive multitarget-multisensor tracking problem can be replaced by a much simpler sorting algorithm of O(N logN) complexity, provided that the sensor system is ideal such that the system has no cluttering points nor missing data. A new computationally efficient ML-based relaxation method for multitarget motion analysis under a fixed networked multisensor environment is then developed by exploiting a new cooperative decentralized processing scheme of [15]. Embedding locally an optimal data association algorithm of O(N logN) into each of Gauss-Newton’s downhill iteration loops, our simulations show that we are able to track multiple targets with improved accuracy and efficiency, where all targets are allowed to move in variable directions at varying speeds. The solution we have developed constitutes a suboptimal solution in the sense of [14, 8] because an optimal solution is embedded within part of the entire optimization problem. keywords: Hungarian algorithm, Data associate problem, Multitarget motion tracking, Bearings-only measurement, Distributed passive sensor network, Cooperative processing