Distributed sparsity-based bearing estimation with a swarm of cooperative agents

Abstract

The presented work discusses a distributed algorithm for solving a return-to-base problem in swarm robotics. A swarm of cooperative intelligent agents is used to span a phased array and cooperatively detect and estimate the bearing of a navigational beacon placed at an unknown location. Both signal detection and bearing estimation is solved jointly using sparse Bayesian learning with dictionary refinement. In the considered setting, Bayesian sparsity is used to detect the presence of the signal. Once signal is detected, its parameters are estimated using a gradient-based numerical technique, with both the gradient and the cost function value computed using classical average consensus over only 4 scalar values. As such, the scheme is independent of the network topology and is particularly useful for communication links with low communication rate. Synthetic simulations demonstrate the effectiveness of the algorithm.