Multi-Temporal Resource Slicing and Scheduling Algorithm for 6G Networks Based on MPNN-DQN

This paper proposes a dynamic network slicing and resource scheduling algorithm for 6G networks, combining Message Passing Neural Networks (MPNN) and Deep Q-Learning (DQN). The algorithm uses MPNN to extract features from network nodes and links, updating their states through a message-passing mechanism. DQN then predicts the Q-value of each action based on these features, enabling fine-grained resource allocation. This approach dynamically adjusts both intra-slice and inter-slice resource allocations, significantly enhancing network resource utilization and service quality. Extensive simulations demonstrate the algorithm's ability to maintain stable resource allocation over longer time scales while improving real-time service performance on shorter time scales. Results indicate that the algorithm significantly outperforms existing methods in resource utilization, task completion rate, and service quality.