Optimal Control Based Sensor Deployment Incorporating Centralized Data Fusion

Data fusion methods that involve combining information from multiple sensors can be used to satisfy detection and false alarm requirements in a distributed sensor network (DSN) based detection system. However, due to limited sensing and communication capabilities of sensors, the positions at which sensors are deployed play an important role in determining overall system performance. We study the problem of determining the positions where a fixed number of sensors are to be deployed in order to 1) meet false alarm requirements, and 2) minimize the squared error (SE) between achieved and required detection probabilities. Towards this end, we propose a novel treatment of the deployment problem within the optimal control theory framework. Specifically, we model the deployment problem as a linear quadratic regulator (LQR) problem, where the sensor positions serve as control vectors. Based on this formulation, we develop two sequential sensor deployment algorithms and study the impact of dynamically updating the collaboration radius. Simulation results show that, in comparison with existing heuristic algorithms, our proposed algorithms use up to 30% fewer number of sensors to satisfy detection and false alarm requirements when using a fixed collaboration radius. Additionally, we show that the use of a dynamic collaboration radius can save up to 45% in the number of sensors used relative to the fixed collaboration case.