Dynamic Activation Policies for Event Capture with Rechargeable Sensors

摘要: We consider the problem of event capture by a rechargeable sensor network. assume that events interest follow renewal process whose inter-arrival times are drawn from general probability distribution, and stochastic recharge is used to provide energy for sensors' operation. Dynamics processes make optimal activation highly challenging. In this paper we first single-sensor problem. Using dynamic control theory, full-information model in which, independent its schedule, will know whether an has occurred last time slot or not. case, framed as Markov decision (MDP), develop simple policy solution. then further partial-information where knows about occurrence only when it active. This falls into class partially observable (POMDP). Since POMDP's exponential computational complexity intrinsically hard solve, propose efficient heuristic clustering evaluate performance. Finally, our solutions extended handle network setting which multiple sensors collaborate events. extensive simulation results performance solutions.