An online method for robust target tracking using a wireless sensor network

This work addresses the target tracking problem using a wireless sensor network (WSN). A set of sensors is randomly deployed to continuously monitor the moving targets, with their spatial trajectories already known but with estimated temporal trajectories and thus, possible delays and advances. The network has to transfer the data collected by the sensors to a base station, with hop-communication between the sensors. Previous works focused on the computation of a robust activation schedule which is passively tracking the targets (i.e., with no reaction to the actual behavior of the targets) by just activating sensors according to the pre-computed schedule. It has a maximized stability radius which is the guarantee on the maximal delay or advance of the targets it can cover, however it loses some targets whenever a delay or advance is beyond the value of its stability radius. Consequently, it is observed in many instances that a target is lost despite being in the range of a fully functioning WSN. A major novelty of the present work is to propose an online method that actively reacts to the delays and advances of the targets to extend the coverage offered by the robust schedule. The proposed method both detects the delays and advances beyond the stability radius, and constantly modifies the schedule to cover the targets as long as possible, at a computational cost that remains acceptable in a real-life application. It lets the WSN monitor the targets despite delays and advances that may not be covered by the guarantee of the stability radius. The second major novelty of our work is the introduction of the dynamic stability radius; it is a guarantee on the greatest delay or advance that can be covered by an online scheduling method. It is, by definition, greater than or equal to the classical stability radius but it is dynamically recomputed such that this guarantee increases over time. Our online method is dynamically optimizing the dynamic stability radius, meaning that each online decision is made so that the WSN has increasing capacities to cover the targets over time. Numerical experiments show that the online method proposed in this work offers more robustness than the previous works, with a dynamic stability radius that improves significantly over the stability radius from previous works, and that increases over time. Experiments also highlight the possibility to use such a method in a real-time context because of its fast computation time. •This work