Dynamic autonomous agent placement for target tracking based on target motion models

Tracking multiple navigating targets in a bounded region is a common problem that arises in many real-life applications, such as rescue operations, surveillance and reconnaissance. Placing a set of agents optimally to track targets, of interest is another problem associated with the tracking problem. This paper introduces a distributed stochastic approach to a well-defined agent placement problem, which can be shown to be NP-hard. First, a stochastic target motion model is introduced to enable agents to predict future target locations. Second, a model-based distributed algorithm is developed. Given the motion model, agents predict target location probabilities and compute their next best locations based on the predictions. The proposed approach involves coordination among mobile agents in order to achieve near-optimal global utilities. The approach has been evaluated through a set of simulation experiments. Simulation results reveal the superiority of the proposed model-based agent placement approach over existing approaches.