Active RIS-Assisted Wireless Communications: A DDPG-Based Anti-Jamming Approach

In order to address the high-power wideband jamming in open channel environments, an active reconfigurable intelligent surface (RIS)-assisted multiple-input single-output (MISO) network is first established. On this basis, an optimization problem on maximizing the signal-to-interference-plus-noise ratio (SINR) is formulated, which requires a joint design of the adaptive beamforming vector, the phase shift and amplitude adjustment of the active RIS, subject to the power allocation of the whole system. In order to handle the non-convex characteristic of the proposed optimization problem and the unpredictable dynamics of jamming channels, a deep deterministic policy gradient (DDPG)-based algorithm is devised due to its superior convergence speed in dynamic environments. Simulation results show that deploying active RIS significantly enhances the anti-jamming capability of the system especially when the direct link is not blocked, with an improvement of SINR by 4 dB compared to conventional passive RIS schemes. Compared to convex optimization and genetic algorithms (GA), by leveraging historical information from time-varying channels, the real-time performance is improved for the proposed approach with faster convergence speed.