可移动阵元STAR-RIS辅助抗干扰传输:基于DDPG算法的双尺度优化OA

For the problem of excessive channel estimation overhead in the anti-jamming system assisted by movable elements based simultaneous transmitting and reflecting reconfigurable intelligent surface(ME-STAR-RIS),a two-timescale collaborative optimization method for anti-jamming transmission is proposed by leveraging the deep deterministic policy gradient(DDPG)algorithm.First,long-term element positions are optimized using statistical channel state information(CSI),then the instantaneous CSI is estimated based on the optimized element positions,which further enables the optimization of short-term beamforming.To address the optimization challenges posed by high-dimensional continuous state spaces and continuous action spaces(such as element positions and phase shift coefficients),the DDPG algorithm is introduced to achieve dynamic policy learning.Simulation results demonstrate that although this method incurs a performance loss of approximately 1.5 b/s/Hz compared with the joint optimization scheme based on instantaneous CSI,it significantly reduces the channel estimation overhead.