Adaptive Multi-Mode DRL Framework for Intelligent RIS-Assisted Anti-Jamming in 6G Wireless Networks

Abstract

—This paper proposes an adaptive multi-mode Deep Reinforcement Learning (DRL) framework for intelligent RIS-assisted anti-jamming communication in dynamic 6G wireless networks. The proposed Framework jointly integrates RIS beamforming, channel hopping, and transmit power adaptation through a DRL-Driven decision engine capable of dynamically responding to varying interference conditions and channel fluctuations. To improve deployment realism, practical constraints including imperfect Channel State Information (CSI), finite-resolution RIS phase quantization, reflection loss, control delay, and user mobility are incorporated into the system model. The anti-jamming problem is formulated as a Markov decision process and solved using DQN, PPO, and SAC algorithms. Extensive simulations are conducted using MATLAB-based wireless channel modeling and Python-based DRL training platforms. Simulation results demonstrate that the proposed framework achieves approximately 25%–40% higher throughput and 18%– 35% SINR improvement compared with conventional anti-jamming approaches. Moreover, the proposed scheme maintains stable communication performance under strong jamming power, CSI uncertainty, and high-mobility scenarios. Statistical evaluations over 20 independent random seeds further confirm the robustness and reproducibility of the proposed framework.