Graph Attention Inference of Network Topology in Multi-Agent Systems

TL;DR

This paper presents a machine learning approach using graph attention mechanisms to infer network topology in multi-agent systems, effective even without prior knowledge of the underlying dynamics.

Contribution

Introduces a novel graph attention-based method for inferring network structure from multi-agent system data, applicable to both linear and non-linear dynamics.

Findings

Successfully predicts network links with high F1 scores

Applicable to systems with unknown dynamic models

Effective for both linear consensus and Kuramoto oscillator dynamics

Abstract

Accurately identifying the underlying graph structures of multi-agent systems remains a difficult challenge. Our work introduces a novel machine learning-based solution that leverages the attention mechanism to predict future states of multi-agent systems by learning node representations. The graph structure is then inferred from the strength of the attention values. This approach is applied to both linear consensus dynamics and the non-linear dynamics of Kuramoto oscillators, resulting in implicit learning of the graph by learning good agent representations. Our results demonstrate that the presented data-driven graph attention machine learning model can identify the network topology in multi-agent systems, even when the underlying dynamic model is not known, as evidenced by the F1 scores achieved in the link prediction.