Learning Ad Hoc Network Dynamics via Graph-Structured World Models

arXiv cs.LG / 4/17/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces G-RSSM, a graph-structured recurrent state space model designed to capture the coupled dynamics of ad hoc wireless networks, including mobility, energy depletion, and topology changes.
Unlike flat-state model-based methods, G-RSSM maintains per-node latent states and uses cross-node multi-head attention to learn network dynamics jointly from offline trajectories.
The approach is evaluated on a clustering downstream task, where a cluster head selection policy is trained entirely via imagined rollouts inside the learned world model.
Across 27 scenarios spanning MANET, VANET, FANET, WSN, and tactical networks (N=30–1000), the learned policy preserves high connectivity even when only trained for N=50, indicating strong scalability.
The authors claim this is the first multi-physics graph-structured world model targeting size-agnostic wireless ad hoc networks and combinatorial per-node decision making.
Point 2
Point 3

Abstract

Ad hoc wireless networks exhibit complex, innate and coupled dynamics: node mobility, energy depletion and topology change that are difficult to model analytically. Model-free deep reinforcement learning requires sustained online interaction whereas existing model based approaches use flat state representations that lose per node structure. Therefore we propose G-RSSM, a graph structured recurrent state space model that maintains per node latent states with cross node multi head attention to learn the dynamics jointly from offline trajectories. We apply the proposed method to the downstream task clustering where a cluster head selection policy trains entirely through imagined rollouts in the learned world model. Across 27 evaluation scenarios spanning MANET, VANET, FANET, WSN and tactical networks with N=30 to 1000 nodes, the learned policy maintains high connectivity with only trained for N=50. Herein, we propose the first multi physics graph structured world model applied to combinatorial per node decision making in size agnostic wireless ad hoc networks.