Particle weight approximation with clustering for gossip-based distributed particle filters

Distributed particle filters are appealing for cooperative tracking in distributed systems. However, a non-negligible amount of communication overhead can be required to synchronize the particle weights between agents. This paper proposes a particle weight approximation method, based on clustering and a smoothness assumption on the particle cloud distribution, to reduce the communication overhead. The proposed algorithm is evaluated on both simulated data and data from an at-sea trial involving bearings-only tracking. The results demonstrate that the proposed approach achieves state-of-the-art accuracy, especially in cases with a limited communication budget.