A variational Bayesian approach for formation target tracking

This paper is concerned with the problem of formation target tracking, where target-originated measurements are modeled as spatially structured multiple detections of the formation center due to multi-mode propagation, and each mode corresponds to a target member. Such modeling transformation brings a set of unknown inputs with an equality constraint in the resultant multi-mode measurement model. Based on variational Bayesian, a joint tracking and identification algorithm that incorporates state estimation and parameter (including unknown inputs and measurement-to-mode association) identification is developed in a unified Bayesian framework, and further optimized in a closed-form iterative manner, which is effective for minimizing the performance deterioration caused by the coupling between estimation errors and identification risks. Finally, the performance of the proposed algorithm is evaluated on non-maneuvering and maneuvering formation tracking scenarios, and simulation results demonstrate its superiority in terms of estimation accuracy, identification effectiveness, and computational complexity.