An Event-Based Diffusion LMS Strategy

Yuan Wang, Wee Peng Tay, Wuhua Hu

Abstract: We consider an event-based communication mechanism for diffusion least mean-squares estimation in a sensor network, in which an intermediate estimate from a sensor is communicated to its neighbors only when a triggering criterion is satisfied. We provide a sufficient condition for the mean error stability of our proposed event-based diffusion strategy, and derive an upper bound of its steady-state network mean-square deviation (MSD). Simulations demonstrate that our event-based strategy can achieve similar steady-state network MSD as the adapt-then-combine diffusion strategy but at a significantly lower communication rate.