Resource Allocation Algorithm Based on D3QN for Ultra-Large-Scale Connection Scenarios in the 6G Context

In the ultra-large-scale connectivity scenario of 6G, traditional resource allocation is limited by finite communication resources and complex propagation environments, leading to inefficient wireless resource allocation. To address this issue, this paper utilizes hypergraph theory to construct a conflict-free resource allocation model, combining deep reinforcement learning (DRL) with the conflict-free resource allocation model to propose a resource allocation algorithm based on Dueling Double Deep Q Network (D3QN). This algorithm aims to maximize the system's cumulative reward, using deep neural networks to select resource allocation strategies that maximize the expected long-term return, achieving conflict-free resource allocation in the 6G ultra-large-scale connectivity scenario. Simulation experiments show that the proposed algorithm improves network throughput by over 70% compared to random algorithms and over 56% compared to Q-Learning algorithms, increasing resource efficiency by up to 33.17% and 41.55%, effectively enhancing network performance.