Liquid-Graph Time-Constant Network for Multi-Agent Systems Control - Archive ouverte HAL
Conference Papers Year : 2024

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

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 the recent Liquid Time Constant (LTC) network. We analyse its stability leveraging contraction analysis and propose a closed- form model that preserves the model contraction rate and does not require solving an ODE at each iteration. Compared to discrete models like Graph Gated Neural Networks (GGNNs), the higher expressivity of the proposed model guarantees remarkable performance while reducing the large amount of communicated variables normally required by GNNs. We evaluate our model on a distributed multi-agent control case study (flocking) taking into account variable communication range and scalability under non-instantaneous communication
Fichier principal
Vignette du fichier
root.pdf (532.09 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04552893 , version 1 (19-04-2024)
hal-04552893 , version 2 (03-09-2024)

Licence

Identifiers

Cite

Antonio Marino, Claudio Pacchierotti, Paolo Robuffo Giordano. Liquid-Graph Time-Constant Network for Multi-Agent Systems Control. Conference on decision and control 2024, IEEE, Dec 2024, Milan (Italie), Italy. ⟨hal-04552893v2⟩
307 View
83 Download

Altmetric

Share

More