Distributed multiple model joint probabilistic data association with Gibbs sampling-aided implementation

Abstract

This paper proposes a new distributed multiple model multiple manoeuvring target tracking algorithm. The proposed tracker is derived by combining joint probabilistic data association (JPDA) with consensus-based distributed filtering. Exact implementation of the JPDA involves enumerating all possible joint association events and thus often becomes computationally intractable in practice. We propose a computationally tractable approximation of calculating the marginal association probabilities for measurement-target mappings based on stochastic Gibbs sampling. In order to achieve scalability for a large number of sensors and high tolerance to sensor failure, a simple average consensus algorithm-based information JPDA filter is proposed for distributed tracking of multiple manoeuvring targets. In the proposed framework, the state of each target is updated using consensus-based information fusion while the manoeuvre mode probability of each target is corrected with measurement probability fusion. Simulations clearly demonstrate the effectiveness and characteristics of the proposed algorithm. The results reveal that the proposed formulation is scalable and much more efficient than classical JPDA without sacrificing tracking accuracy.