Spatiotemporal Adaptive Optimization of a Static-Sensor Network via a Non-Parametric Estimation of Target Location Likelihood

The search for a mobile target is a dynamic spatial and temporal problem. Information gathered during the search, regarding the potential whereabouts of the target, can be used to influence significantly the search strategy employed. In this paper, thus, a novel (wireless) static-sensor network deployment strategy is presented to detect efficiently and effectively a mobile target in an unbounded environment. The proposed strategy deploys the static sensors optimally and in a time-phased manner while adapting in real-time to the availability of new information during the search. An optimal network-deployment plan, herein, refers to a set of optimal sensor-deployment times and locations. The optimal sensor-deployment instances aim to achieve a uniform deployment of search effort over time. Optimal sensor-deployment locations, in turn, are determined according to the highest possible likelihood of detection of the mobile target. The proposed deployment strategy contains two novel contributions: it makes use of a non-parametric approach to estimate effectively the target location likelihood, and it performs an optimization of sensor-placement times to maximize the adaptive characteristic of the deployment plan. Several detailed experiments (in virtual and physical environments) of the proposed strategy for static-sensor network deployment in wilderness search and rescue applications are presented. Furthermore, a comparative study is included to highlight the advantages of our approach versus traditional methods that deploy sensors simultaneously for uniform coverage.