Reinforcement Learning for Multi-Agent Systems with Temporal Logic Specifications

Keita TERASHIMA; Koichi KOBAYASHI; Yuh YAMASHITA

doi:10.1587/transfun.2023KEP0016

Reinforcement Learning for Multi-Agent Systems with Temporal Logic Specifications

Keita TERASHIMA, Koichi KOBAYASHI, Yuh YAMASHITA

Full Text Views

0

Cite this

Summary :

In a multi-agent system, it is important to consider a design method of cooperative actions in order to achieve a common goal. In this paper, we propose two novel multi-agent reinforcement learning methods, where the control specification is described by linear temporal logic formulas, which represent a common goal. First, we propose a simple solution method, which is directly extended from the single-agent case. In this method, there are some technical issues caused by the increase in the number of agents. Next, to overcome these technical issues, we propose a new method in which an aggregator is introduced. Finally, these two methods are compared by numerical simulations, with a surveillance problem as an example.

Publication: IEICE TRANSACTIONS on Fundamentals Vol.E107-A No.1 pp.31-37

Publication Date: 2024/01/01

Publicized: 2023/07/19

Online ISSN: 1745-1337

DOI: 10.1587/transfun.2023KEP0016

Type of Manuscript: Special Section PAPER (Special Section on Circuits and Systems)

Category

Authors

Keita TERASHIMA
  Hokkaido University
Koichi KOBAYASHI
  Hokkaido University
Yuh YAMASHITA
  Hokkaido University

Keyword

multi-agent systems, reinforcement learning, linear temporal logic, aggregator, surveillance

Cite this

Copy

Keita TERASHIMA, Koichi KOBAYASHI, Yuh YAMASHITA, "Reinforcement Learning for Multi-Agent Systems with Temporal Logic Specifications" in IEICE TRANSACTIONS on Fundamentals, vol. E107-A, no. 1, pp. 31-37, January 2024, doi: 10.1587/transfun.2023KEP0016.
Abstract: In a multi-agent system, it is important to consider a design method of cooperative actions in order to achieve a common goal. In this paper, we propose two novel multi-agent reinforcement learning methods, where the control specification is described by linear temporal logic formulas, which represent a common goal. First, we propose a simple solution method, which is directly extended from the single-agent case. In this method, there are some technical issues caused by the increase in the number of agents. Next, to overcome these technical issues, we propose a new method in which an aggregator is introduced. Finally, these two methods are compared by numerical simulations, with a surveillance problem as an example.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2023KEP0016/_p

Copy

@ARTICLE{e107-a_1_31,
author={Keita TERASHIMA, Koichi KOBAYASHI, Yuh YAMASHITA, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Reinforcement Learning for Multi-Agent Systems with Temporal Logic Specifications},
year={2024},
volume={E107-A},
number={1},
pages={31-37},
abstract={In a multi-agent system, it is important to consider a design method of cooperative actions in order to achieve a common goal. In this paper, we propose two novel multi-agent reinforcement learning methods, where the control specification is described by linear temporal logic formulas, which represent a common goal. First, we propose a simple solution method, which is directly extended from the single-agent case. In this method, there are some technical issues caused by the increase in the number of agents. Next, to overcome these technical issues, we propose a new method in which an aggregator is introduced. Finally, these two methods are compared by numerical simulations, with a surveillance problem as an example.},
keywords={},
doi={10.1587/transfun.2023KEP0016},
ISSN={1745-1337},
month={January},}

Copy

TY - JOUR
TI - Reinforcement Learning for Multi-Agent Systems with Temporal Logic Specifications
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 31
EP - 37
AU - Keita TERASHIMA
AU - Koichi KOBAYASHI
AU - Yuh YAMASHITA
PY - 2024
DO - 10.1587/transfun.2023KEP0016
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E107-A
IS - 1
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - January 2024
AB - In a multi-agent system, it is important to consider a design method of cooperative actions in order to achieve a common goal. In this paper, we propose two novel multi-agent reinforcement learning methods, where the control specification is described by linear temporal logic formulas, which represent a common goal. First, we propose a simple solution method, which is directly extended from the single-agent case. In this method, there are some technical issues caused by the increase in the number of agents. Next, to overcome these technical issues, we propose a new method in which an aggregator is introduced. Finally, these two methods are compared by numerical simulations, with a surveillance problem as an example.
ER -