Reinforcement Learning Based Adaptive Resource Allocation for Wireless Powered Communication Systems

Wireless powered communication (WPC) is one of the promising techniques for future energy-constrained wireless networks. In this letter, we consider a WPC system composed of a hybrid access point and an energy harvesting node (EHN). In this system, we propose a reinforcement learning based adaptive resource allocation scheme that dynamically assigns the channel resources to minimize the outage probability of information transfer while satisfying the average power constraint at the EHN, which is formulated as a constrained Markov decision process (MDP) problem. To solve this challenging problem, we first transform the originally formulated problem into its equivalent unconstrained MDP with multi-objective. Then, to find the resource allocation policy, we propose a novel Q-learning algorithm. Numerical results demonstrate the superior performance and effectiveness of the proposed scheme.