Power Allocation for Full-Duplex Communication Systems Based on Deep Deterministic Policy Gradient

In full-duplex (FD) communications, the base station (BS) can communicate with an uplink user and a downlink user simultaneously on the same frequency band. Therefore, there exist two kinds of interference in FD communications, the self-interference and the co-channel interference. On the other hand, power allocation policy plays an important role in dealing with the interference. In this paper, a deep reinforcement learning method is applied to optimize the power allocation, aiming at maximizing the spectrum efficiency for FD systems. The optimization problem is a non-convex one. Then, a deep deterministic policy gradient (DDPG)-based power allocation algorithm is proposed to allocate downlink transmit power. Accordingly, the uplink transmit power can be adjusted adaptively. With the proposed DDPG-based algorithm, the BS can intelligently adjust its downlink transmit power by interacting with the environment. Numerical results show that the proposed algorithm can achieve a near optimal performance with much lower computational complexity than traditional approaches.