Tracking of targets of interest using labeled multi-Bernoulli filter with multi-sensor control

Abstract

This paper proposes a novel multi-sensor control scheme for tracking of targets of interest in the presence of clutter and detection uncertainty, using labeled multi-Bernoulli filters. The focus is on multi-target tracking applications where multiple centrally connected and controllable sensors are used for tracking of particular targets of interest. We derive an efficient analytical approximation for a task-driven objective function for multi-sensor selective control, that can be computed immediately after the prediction step. Compared to the state-of-the-art, drastic reduction in computation is achieved. Numerical experiments, involving challenging multi-sensor control scenarios, demonstrate how the proposed method can lead to significant improvements in the tracking accuracy of the targets of interest, in comparison to the generic non-selective sensor control methods. It is also shown that while our proposed method has comparable performance to the state-of-art selective sensor control method (selective-PEECS) in terms of the mean-square-error (MSE) of tracking for targets of interest, it runs significantly faster than the state-of-art (both selective and non-selective) multi-sensor control algorithms. Indeed, the resulting reduction in computation time is shown to be in the order of tens to hundreds time, depending on the number of sensors to be controlled.