Liquid-Graph Time-Constant Network for Multi-Agent Systems Control

TL;DR

This paper introduces the Liquid-Graph Time-constant (LGTC) network, a continuous graph neural network model designed for multi-agent control, offering improved stability, expressivity, and communication efficiency over discrete GNN models.

Contribution

The paper presents a novel LGTC network that leverages contraction analysis for stability and provides a closed-form model that enhances expressivity and reduces communication in multi-agent systems.

Findings

LGTC guarantees stability via contraction analysis.

The model outperforms discrete GNNs in control tasks.

Reduces communication overhead in multi-agent systems.

Abstract

In this paper, we propose the Liquid-Graph Time-constant (LGTC) network, a continuous graph neural network(GNN) model for control of multi-agent systems based on therecent Liquid Time Constant (LTC) network. We analyse itsstability leveraging contraction analysis and propose a closed-form model that preserves the model contraction rate and doesnot require solving an ODE at each iteration. Compared todiscrete models like Graph Gated Neural Networks (GGNNs),the higher expressivity of the proposed model guaranteesremarkable performance while reducing the large amountof communicated variables normally required by GNNs. Weevaluate our model on a distributed multi-agent control casestudy (flocking) taking into account variable communicationrange and scalability under non-instantaneous communication