Decentralized Online Direction-of-Arrival Estimation and Tracking

The problem of the decentralized Direction-of-Arrival (DoA) estimation and tracking is addressed, where the sample covariance matrix is represented as a recursive update of rank-one components. Due to the fact that the conventional decentralized power method is a batch approach and involves high communication costs, we propose a distributed DoA estimation method combining the decentralized online eigenvalue decomposition and the ESPRIT algorithm. The Push-Sum protocol is applied to realize the local interactions among neighboring subarrays. The simulation results show that the Root Mean Square Error (RMSE) performance of our distributed algorithm surpasses that of the power method based d-ESPRIT algorithm in the DoA estimation with lower total communication cost, and that of the decentralized Normalized Oja (d-NOja) method in DoA tracking applications with faster convergence speed.