Distributed particle filtering for large scale dynamical systems

The paper proposes a near-optimal, distributed implementation of the particle filter for large scale dynamical systems with sparse measurements, as in wireless sensor networks (WSN) and power distribution systems. Compared to the centralized approach, the distributed implementation of the particle filter is computationally efficient and provides considerable transmission bandwidth savings due to a significant decrease in information exchange between the neighbouring nodes. In our test simulations, the performance of the distributed implementation is almost indistinguishable from the centralized particle filter.