Performance Comparison of the Distributed Extended Kalman Filter and Markov Chain Distributed Particle Filter (MCDPF)

We compare the performance of two distributed nonlinear estimators for a multi-vehicle flocking system using range measurements only. The estimators are the Distributed Extended Kalman Filter (DEKF) and the Markov Chain Distributed Particle Filter (MCDPF), where the distributed implementation in both cases is done using consensus-type algorithms. The performance of the estimators is compared as the system complexity (number of vehicles) and measurement frequency are varied. It is shown that for simple systems (few vehicles) or high measurement frequency the DEKF method has lower expected error than MCDPF, while for complex systems (many vehicles) or low measurement frequency the MCDPF method is both more robust and more accurate.