GNN-DRL-Based Intelligent Routing and Resource Allocation Algorithms for Multi-Layer Wireless Mesh Network.

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Engineering

Background:

• Dynamic wireless mesh networks face challenges in emergency communications due to changing topologies and diverse AI service demands.
• Existing routing and resource allocation methods struggle with the complexity of real-time network adjustments and varied Quality of Service (QoS) requirements.

Purpose of the Study:

• To propose an intelligent routing and resource allocation algorithm for dynamic wireless mesh networks.
• To address the challenges of AI model training services and dynamic node capabilities in emergency communication scenarios.

Main Methods:

• Developed Graph Sample and Aggregate-Multi-Agent Proximal Policy Optimization (GraphSAGE-MAPPO), integrating Graph Neural Networks (GNN) and Deep Reinforcement Learning (DRL).
• Utilized GNN for network feature extraction (node capabilities, link status) to generate hidden feature vectors.
• Employed these feature vectors and service flow data as state input for a distributed multi-agent DRL model to optimize routing and resource allocation.

Main Results:

• GraphSAGE-MAPPO demonstrated flexible adaptation to dynamic network environments and user needs.
• The algorithm effectively met diverse service QoS requirements.
• Simulation results indicated good generalization performance across network topology and resource variations.

Conclusions:

• GraphSAGE-MAPPO offers a scalable and flexible solution for intelligent routing and resource allocation in large-scale wireless mesh networks.
• The proposed algorithm shows significant promise for enhancing emergency communication systems.
• The integration of GNN and DRL provides a robust framework for complex network management.