Jump-Markov process estimation based on distributed Rao-Blackwellized particle filter

The particle filter is widely employed for estimation of non-Gaussian problems. A disadvantage of the particle filter is long computation time when a large number of particles is used. In this paper, distributed Rao-Blackwellized particle filter is explored in order to produce a common estimate by distributing the particles in several agents. By implementing a consensus algorithm of the estimated global statistics it is shown that, in each agent, the estimated global statistics is more accurate than the local, per-agent statistics. For verification, the numerical example of a vertical channel of an inertial navigation system with barometric altimeter aiding is given, which confirms that DRBPF reduces errors effectively.