Decentralized Management of Sensors in a Multi-Attribute Environment Under Weak Network Congestion

We provide a game theoretic formulation for a sensor activation problem in a multi-attribute environment. Activated sensors randomly select one of M environmental attributes, and transmit data on that attribute to an end user. The goal is to maximize the number of attributes reported while minimizing redundant reports and packet collisions, which both increase with the number of active sensors. Sensor participation is optimized according to an adaptive scheme, in which sensors activate only when their expected utility, given by the number of unique attributes reported minus an energy cost, is positive. We formulate a Nash equilibrium policy that maximizes the expected performance from the perspective of each sensor when transmission is according to a one-shot frequency hopping scheme, and compare this to the global optimum