Skip to Main Content
Bearing estimation is a well-studied problem and maximum likelihood (ML) estimation provides the best solution in terms of performance. The difficulty with ML is the multi-modal nature of the likelihood cost function. Recently, the biologically inspired particle swarm optimisation (PSO) technique has been shown to provide a good solution to ML bearing estimation as it alleviates the effects of multi-modality. In this study, the ML bearing estimation in a distributed sensor network is addressed, where each sensor node has access only to data from its neighbours. Diffusion particle swarm optimisation (DPSO) is proposed to optimise the ML function in this context. During the optimisation process each associated node shares its best estimates of the source bearings with its neighbours. As each node only communicates its best estimates and its own data with its neighbours, the communication overhead is less than the existing centralised PSO method. Diffusion learning ensures robustness to changes in network topology. Simulation results compare the performance of DPSO, centralised PSO, the benchmark centralised MUltiple SIgnal Classification bearing estimation algorithm and the appropriate Cramer-Rao lower bounds. As might be expected, there is some degradation in performance of the DPSO with respect to centralised PSO.