Energy-efficient distributed state estimation via event-triggered consensus on exponential families

The paper addresses distributed state estimation over a peer-to-peer sensor network with an eye to communication/energy efficiency. In particular, consensus on exponential families of probability distributions is first introduced and shown to be equivalent to iteratively performing convex linear combinations on the natural parameters of such distributions. Then, an event-triggered consensus strategy is presented and exploited to derive a novel energy-efficient consensus Kalman filter algorithm for distributed state estimation. Simulation results are provided to demonstrate the effectiveness of the proposed algorithm.