Beamforming and Resource Allocation in Multi-cell OFDMA Systems based on Deep Transfer Reinforcement Learning

In this paper, we study joint beamforming and resource allocation in downlink multi-cell orthogonal frequency division multiple access (OFDMA) systems. We design a multi-agent deep Q-network (MADQN) algorithm to solve this problem. Furthermore, in order to improve the adaptability of neural networks for different wireless environment, we propose a transfer learning framework based on MADQN called TL-MADQN to dynamically output optimal beamforming and resource allocation policy. Finally, we adjust the allocation policy to maximize the sum-rate of all users by updating the weights of each neural network. Simulation results illustrate that the proposed TL-MADQN algorithm has higher sum-rate and faster convergence speed compared with the baseline algorithms.