Reinforcement Learning for Resource Allocation in Multiuser OFDM Systems

Abstract In cellular mobile communications the subcarriers are repeatedly used for best utilizing the assigned frequency spectrum. The assignment of channels for users is complex and involves high computation time. Wong et al [15] proposed a heuristic algorithm to achieve the suboptimal solution for sub-carrier assignment. This proposal was based on constructive assignment in real time situation with prosperous results. However, the algorithm involves computational complexity. In this paper we proposes reinforcement learning algorithm for sub-carrier assignment to the users in a way that the total transmit power is minimized. Reinforcement learning algorithms are frequently used for optimization problems and are related to dynamic programming algorithms. Simulation results show that proposed reinforcement learning is robust and outperforms the heuristic algorithm proposed by Wong et al.