Toward Multi-Agent Reinforcement Learning for Distributed Event-Triggered Control

Lukas Kesper; Sebastian Trimpe; Dominik Baumann

Toward Multi-Agent Reinforcement Learning for Distributed Event-Triggered Control

Lukas Kesper, Sebastian Trimpe, Dominik Baumann

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

1 Citation (Scopus)

69 Downloads (Pure)

Abstract

Event-triggered communication and control provide high control performance in networked control systems without overloading the communication network. However, most approaches require precise mathematical models of the system dynamics, which may not always be available. Model-free learning of communication and control policies provides an alternative. Nevertheless, existing methods typically consider single-agent settings. This paper proposes a model-free reinforcement learning algorithm that jointly learns resource-aware communication and control policies for distributed multi-agent systems from data. We evaluate the algorithm in a high-dimensional and nonlinear simulation example and discuss promising avenues for further research.

Original language	English
Title of host publication	Proceedings of the Learning for Dynamics and Control Conference
Publisher	JMLR
Pages	1072-1085
Number of pages	14
Volume	211
Publication status	Published - 1 Jun 2023
MoE publication type	A4 Conference publication
Event	Learning for Dynamics and Control Conference - University of Pennsylvania, Philadelphia, United States Duration: 14 Jun 2023 → 16 Jun 2023 Conference number: 5 https://l4dc.seas.upenn.edu/

Publication series

Name	Proceedings of Machine Learning Research
ISSN (Electronic)	2640-3498

Conference

Conference	Learning for Dynamics and Control Conference
Abbreviated title	L4DC
Country/Territory	United States
City	Philadelphia
Period	14/06/2023 → 16/06/2023
Internet address	https://l4dc.seas.upenn.edu/

Keywords

Electrical Engineering and Systems Science - Systems and Control

Access to Document

pdfAccepted author manuscript, 622 KB

Cite this

@inproceedings{d7cea6607bd94e50a2a8eddf5b61f53d,

title = "Toward Multi-Agent Reinforcement Learning for Distributed Event-Triggered Control",

abstract = "Event-triggered communication and control provide high control performance in networked control systems without overloading the communication network. However, most approaches require precise mathematical models of the system dynamics, which may not always be available. Model-free learning of communication and control policies provides an alternative. Nevertheless, existing methods typically consider single-agent settings. This paper proposes a model-free reinforcement learning algorithm that jointly learns resource-aware communication and control policies for distributed multi-agent systems from data. We evaluate the algorithm in a high-dimensional and nonlinear simulation example and discuss promising avenues for further research.",

keywords = "Electrical Engineering and Systems Science - Systems and Control",

author = "Lukas Kesper and Sebastian Trimpe and Dominik Baumann",

year = "2023",

month = jun,

day = "1",

language = "English",

volume = "211",

series = "Proceedings of Machine Learning Research",

publisher = "JMLR",

pages = "1072--1085",

booktitle = "Proceedings of the Learning for Dynamics and Control Conference",

address = "United States",

note = "Learning for Dynamics and Control Conference, L4DC ; Conference date: 14-06-2023 Through 16-06-2023",

url = "https://l4dc.seas.upenn.edu/",

}

Kesper, L, Trimpe, S & Baumann, D 2023, Toward Multi-Agent Reinforcement Learning for Distributed Event-Triggered Control. in Proceedings of the Learning for Dynamics and Control Conference. vol. 211, Proceedings of Machine Learning Research, JMLR, pp. 1072-1085, Learning for Dynamics and Control Conference, Philadelphia, Pennsylvania, United States, 14/06/2023. <http://10.48550/arXiv.2305.08723>

Toward Multi-Agent Reinforcement Learning for Distributed Event-Triggered Control. / Kesper, Lukas; Trimpe, Sebastian; Baumann, Dominik.
Proceedings of the Learning for Dynamics and Control Conference. Vol. 211 JMLR, 2023. p. 1072-1085 (Proceedings of Machine Learning Research).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Toward Multi-Agent Reinforcement Learning for Distributed Event-Triggered Control

AU - Kesper, Lukas

AU - Trimpe, Sebastian

AU - Baumann, Dominik

N1 - Conference code: 5

PY - 2023/6/1

Y1 - 2023/6/1

N2 - Event-triggered communication and control provide high control performance in networked control systems without overloading the communication network. However, most approaches require precise mathematical models of the system dynamics, which may not always be available. Model-free learning of communication and control policies provides an alternative. Nevertheless, existing methods typically consider single-agent settings. This paper proposes a model-free reinforcement learning algorithm that jointly learns resource-aware communication and control policies for distributed multi-agent systems from data. We evaluate the algorithm in a high-dimensional and nonlinear simulation example and discuss promising avenues for further research.

AB - Event-triggered communication and control provide high control performance in networked control systems without overloading the communication network. However, most approaches require precise mathematical models of the system dynamics, which may not always be available. Model-free learning of communication and control policies provides an alternative. Nevertheless, existing methods typically consider single-agent settings. This paper proposes a model-free reinforcement learning algorithm that jointly learns resource-aware communication and control policies for distributed multi-agent systems from data. We evaluate the algorithm in a high-dimensional and nonlinear simulation example and discuss promising avenues for further research.

KW - Electrical Engineering and Systems Science - Systems and Control

UR - http://www.scopus.com/inward/record.url?scp=85172908259&partnerID=8YFLogxK

M3 - Conference article in proceedings

VL - 211

T3 - Proceedings of Machine Learning Research

SP - 1072

EP - 1085

BT - Proceedings of the Learning for Dynamics and Control Conference

PB - JMLR

T2 - Learning for Dynamics and Control Conference

Y2 - 14 June 2023 through 16 June 2023

ER -

Toward Multi-Agent Reinforcement Learning for Distributed Event-Triggered Control

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this