Optimization of Multitarget Tracking Within a Sensor Network via Information-Guided Clustering

This paper presents a new algorithm for rapid and efficient clustering of sensing nodes within a heterogeneous wireless sensor network. The objective is to enable optimal sensor allocation for localization uncertainty reduction in multitarget tracking. The proposed algorithm is built on three metrics: 1) sensing feasibility, 2) measurement quality to maximize information utility, and 3) communication cost to minimize data routing time. The derived cluster serves as the search space for the optimal sensor(s) to be tasked with the measurement, via optimization of differential entropy, that effectively minimizes uncertainty in the targets predicted posterior distribution. Theoretical analysis is used to show the advantage of the proposed method in terms of information utility over the Euclidean distance-based clustering approach. The analysis is verified via simulated target-tracking examples, in terms of metrics of information utility and computational expenditure. Simulations also reveal relationships between sensor field density and the extent of information gain over competing methods.