Graph neural networks improve resilient 5g wireless network planning
Resilient Integrated Access and Backhaul Network Planning via Graph Neural Network Curriculum Learning
Networking and Internet Architecture
Summary
Planning wireless networks that connect many devices cheaply and reliably is hard, especially when the fiber connections to the main network points vary a lot. The authors look at how to build these networks so they cover almost all users and keep working even when some connections break. They use a special type of AI called graph neural networks combined with a step-by-step training method to create plans that adapt well to different city layouts. Their method uses fewer network nodes and keeps users connected better than older approaches, without needing to be retrained for each new setup.
What this means in practice
- •For wireless network engineers: Design 5G mmWave networks that maintain coverage and handle backhaul failures efficiently across different city layouts.
- •For urban infrastructure planners: Plan urban communication infrastructure by optimizing placement of wireless nodes to ensure resilient connectivity without excessive hardware.
Authors
Jie Zhang, Haochen Sun, Mingzhang Zhou, Paul D. Mitchell, Hamed Ahmadi, Qiang Ni
Abstract
Integrated Access and Backhaul (IAB) is a key enabler for cost-effective millimeter-wave (mmWave) densification, yet practical deployments are strongly constrained by the spatial availability of fiber-connected IAB donors. Variations in donor topologies reshape both the initial coverage footprint and the feasible multi-hop backhaul expansion patterns, causing planning policies trained on a single topology to generalize poorly. This paper studies cross-topology generalization for resilience-aware IAB deployment under coverage and transport feasibility requirements. We formulate IAB deployment as a constrained Markov decision process that couples access coverage satisfaction with multi-hop transport feasibility, including hop-by-hop rate-budget degradation and a minimum parent-connectivity requirement for fault tolerance. To solve the resulting large discrete planning problem, we propose a graph-based actor-critic approach that uses a Graph Attention Network v2 (GATv2) encoder with Proximal Policy Optimization (PPO), and we develop a multi-stage curriculum learning strategy that progressively increases constraint difficulty and scenario diversity for stable training and transferable policies. In a 60 GHz urban simulation over a 1000 m x 1000 m area with multiple donor layouts, the trained policy satisfies the minimum coverage requirement of 98% with fewer deployed IAB nodes than learning and heuristic baselines, and maintains higher coverage under progressive backhaul failures. The results demonstrate that curriculum-trained graph-based reinforcement learning can improve deployment efficiency and resilience while transferring across heterogeneous IAB donor topologies without retraining.