Online Optimization of Wireless Sensors Selection over an Unknown Stochastic Environment

Wireless communication is considered to be more challenging than the typical wired communication due to unpredictable channel conditions. In this paper, we target coverage area problem, where a group of sensors is selected from a set of sensors placed in a particular area to maximize the coverage provided to that area. The constraints to this optimization are the battery power of the sensor and number of sensors that are active at a given time. We consider a variant of the coverage related to a particular sensor, where coverage is considered to be an unknown stochastic variable, and hence, we need to learn the best subset of sensors in real time. We propose an online combinatorial optimization algorithm based on multi-armed bandits framework that learns the expected best subset of sensors, and the regret of the proposed online algorithm is sub-linear in time. The achieved performance proves the robustness and effectiveness of the proposed online algorithm in wireless sensor selection over an unknown stochastic environment.