Graph Neural Network-Driven Deep Reinforcement Learning for Scalable RIS Allocation

2026-10-05

AI Summary

This study addresses the NP-hardness and scalability bottlenecks associated with the dynamic allocation of shared reconfigurable intelligent surfaces (RIS) in 6G multi-cell networks. To this end, it proposes an RIS orchestration framework that integrates graph neural networks with deep reinforcement learning. By exploiting physical topology sparsification, the method transforms dense channel matrices into sparse graphs and designs a local graph structure with linear complexity, enabling the model to generalize to arbitrary network scales without retraining. Experimental results demonstrate that, under strict budget constraints, the proposed approach improves coverage by approximately 12% over baseline methods while significantly reducing computational complexity, thereby achieving efficient and scalable network optimization.

Abstract

Reconfigurable Intelligent Surfaces (RISs) offer a promising paradigm to mitigate blockage and extend millimeter-wave coverage in 6G multi-cell networks. However, dynamically allocating shared RIS infrastructure across competing base stations is an NP-hard problem posing severe scalability bottlenecks. In this paper, we propose a scalable framework combining Graph Neural Networks (GNNs) with Deep Reinforcement Learning (DRL) for dynamic shared RIS orchestration. By formulating allocation as a Markov Decision Process, we introduce a physical topology sparsification strategy that prunes dense channel matrices into a sparse tripartite graph. This pruning reduces edge density by 77% and removes representation noise, thereby improving global coverage probability while reducing computational complexity. Our relational message-passing architecture naturally generalizes to arbitrary network dimensions without model retraining. Furthermore, structural ablation studies reveal that physical path loss localizes surface dependencies, enabling a highly efficient localized graph design with linear computational scaling. Extensive simulations in dense urban environments demonstrate that under strict infrastructure budget constraints, the proposed GNN-DRL framework consistently outperforms greedy heuristic baselines by up to ~12% in coverage while delivering faster inference speed via GPU acceleration.

Problem

• Reconfigurable Intelligent Surfaces
• Dynamic RIS Allocation
• Multi-cell Networks
• Scalability
• NP-hard

Innovation

• Graph Neural Networks
• Deep Reinforcement Learning
• Reconfigurable Intelligent Surfaces
• Topology Sparsification
• Scalable RIS Allocation