Multi-Bernoulli sensor-selection for multi-target tracking with unknown clutter and detection profiles

A new sensor-selection solution within a Multi-Bernoulli-based multi-target tracking framework is presented. The proposed method is especially designed for the general multi-target tracking case with no prior knowledge of the clutter distribution or the probability of detection, and uses a new task-driven objective function for this purpose. Step-by-step sequential Monte Carlo implementation of the method is presented along with a similar sensor-selection solution formulated using an information-driven objective function (Rényi divergence). The two solutions are compared in a challenging scenario and the results show that while both methods perform similarly in terms of accuracy of cardinality and state estimates, the taskdriven sensor-selection method is substantially faster.