Optimization of Resource Allocation in Multi-Link P2P Communication Systems Based on Reinforcement Learning

The next-generation Wi-Fi 7 protocol supports multi-link operation and Peer-to-Peer (P2P) communication. Combining two of them can further improve spectrum resource utilization. However, P2P communication may cause severe co-channel interference. To address this issue, this paper proposes a resource allocation strategy based on the Double Deep Q Network-Twin Delayed Deep Deterministic policy gradient (DDQN-TD3) algorithm, utilizing a centralized training-distributed execution framework for multi-agent, multi-task reinforcement learning. This approach aims to ensure Quality of Service (QoS) for Wi-Fi users while facilitating reasonable Resource Unit (RU) reuse selection and transmission power control by P2P users across multiple links. It also seeks to mitigate the interference of P2P communication on direct Wi-Fi communication while maximizing system capacity. Simulation results indicate that the proposed algorithm achieves greater throughput performance gains compared to the Multi-Agent Deep Q Network (MA-DQN), Multi-Agent Greedy (MA-Greedy), and random selection algorithms mentioned in this paper, with the DDQN-TD3 algorithm exhibiting optimal convergence.