Reinforcement Learning-based Position Optimization for Movable Antenna-Enabled Relay Communications

In this paper, we propose a movable antenna (MA)-enabled relay system, where the buffer is employed to enhance the transmission performance and the energy consumption for MA’s movement is considered. Specifically, we formulate an optimization problem to maximize the achievable rate from the source node to the destination node by dynamically optimizing the position of MA at the relay node and the transmit power of the source and relay nodes over time, under the constraints of antenna movement, limited cache capacity, and power supply. To solve this challenging problem, a reinforcement learning (RL)-based algorithm is proposed, where deep deterministic policy gradient (DDPG) is employed to obtain the solution for the MA’s movement and power allocation. Simulation results show that the MA-enabled relay system could obtain significant performance improvement over the conventional fixed position antenna (FPA)-aided relay system by full exploitation of the spatial degree of freedom (DoF) to balance the channel conditions of the two links. Moreover, with sufficient energy supply for antenna movement, a single MA can even achieve comparable performance to multi-antenna digital beamforming (DBF) system.