Multi-User Dynamic Spectrum Access Based on Deep Reinforcement Learning
Yanwen Geng · 2024
Aiming at the multi-channel dynamic spectrum access problem, this paper proposes a distributed Dueling Double Deep Q Network based algorithm. The goal of the algorithm is to maximize the sum of user rates and uses a quality judgment algorithm to assess the quality of individual users, thereby improving the computation of the overall target value. And the distributed architecture is used to train the global model by a central unit, locally treating each user as an intelligence, and the individual intelligences learn using standard single-intelligence reinforcement learning methods. Simulation results show that the algorithm can reduce the collision rate and improve channel utilization by 2.78%.