Self-Tuning Algorithms for Multisensor-Multitarget Tracking Using Belief Propagation

Situation-aware technologies enabled by multitarget tracking algorithms will create new services and applications in emerging fields such as autonomous navigation and maritime surveillance. The system models underlying multitarget tracking algorithms often involve unknown parameters that are potentially time-varying. A manual tuning of unknown model parameters by the user is prone to errors and can, thus, dramatically reduce target detection and tracking performance. We address this challenge by proposing a framework of “self-tuning” multisensor-multitarget tracking algorithms. These algorithms adapt in an online manner to time-varying system models by continuously inferring unknown model parameters along with the target states. We describe the evolution of the parameters by a Markov chain and incorporate them in a factor graph that represents the statistical structure of the tracking problem. We then use a belief propagation scheme to efficiently calculate the marginal posterior distributions of the targets and model parameters. As a concrete example, we develop a self-tuning tracking algorithm for maneuvering targets with multiple dynamic models and sensors with time-varying detection probabilities. The performance of the algorithm is validated for simulated scenarios and for a real scenario using measurements from two high-frequency surfacewave radars.