Distributed dynamic stochastic approximation algorithm over time-varying networks

Abstract In this paper, a distributed stochastic approximation algorithm is proposed to track the dynamic root of sum time-varying regression functions over network. Each agent updates its estimate by using local observation, information global root, and received from neighbors. Compared with similar works in optimization area, we allow observation be noise-corrupted, noise condition much weaker. Furthermore, instead upper bound error, present asymptotic convergence result algorithm. The consensus estimates are established. Finally, applied target tracking problem numerical example presented demonstrate performance