Fluctuated peer selection policy and its performance in large-scale multi-agent systems

Toshiharu Sugawara, Kensuke Fukuda, Toshio Hirotsu, Shin Ya Sato, Osamu Akashi, Satoshi Kurihara

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

This paper describes how, in large-scale multi-agent systems, each agent's adaptive selection of peer agents for collaborative tasks affects the overall performance and how this performance varies with the workload of the system and with fluctuations in the agents' peer selection policies (PSP). An intelligent agent in a multi-agent system (MAS) often has to select appropriate agents to assign tasks that cannot be executed locally. These collaborating agents are usually chosen according to their skills. However, if multiple candidate peer agents still remain a more efficient agent is preferable. Of course, its efficiency is affected by the agent' workload and CPU performance and the available communication bandwidth. Unfortunately, as no agent in an open environment such as the Internet can obtain such data from any other agent, this selection must be done according to the available local information about the other known agents. However, this information is limited, usually uncertain and often obsolete. Agents' states may also change over time, so the PSP must be adaptive to some extent. We investigated how the overall performance of MAS would change under adaptive policies in which agents selects peer agents using statistical/reinforcement learning. We particularly focused on mutual interference for selection under different workloads, that is, underloaded, near-critical, and overloaded situations. This paper presents simulation results and shows that the overall performance of MAS highly depends on the workload. It is shown that when agents' workloads are near the limit of theoretical total capability, a greedy PSP degrades the overall performance, even after a sufficient learning time, but that a PSP with a little fluctuation, called fluctuated PSP, can considerably improve it.

Original languageEnglish
Pages (from-to)255-268
Number of pages14
JournalWeb Intelligence and Agent Systems
Volume8
Issue number3
DOIs
Publication statusPublished - 2010 Jul 30
Externally publishedYes

Fingerprint

Multi agent systems
Intelligent agents
Reinforcement learning
Program processors
Internet
Bandwidth
Communication

Keywords

  • cooperation
  • coordination
  • learning
  • load-balancing
  • Peer-to-peer computation
  • task allocation

ASJC Scopus subject areas

  • Software
  • Computer Networks and Communications
  • Artificial Intelligence

Cite this

Fluctuated peer selection policy and its performance in large-scale multi-agent systems. / Sugawara, Toshiharu; Fukuda, Kensuke; Hirotsu, Toshio; Sato, Shin Ya; Akashi, Osamu; Kurihara, Satoshi.

In: Web Intelligence and Agent Systems, Vol. 8, No. 3, 30.07.2010, p. 255-268.

Research output: Contribution to journalArticle

Sugawara, Toshiharu ; Fukuda, Kensuke ; Hirotsu, Toshio ; Sato, Shin Ya ; Akashi, Osamu ; Kurihara, Satoshi. / Fluctuated peer selection policy and its performance in large-scale multi-agent systems. In: Web Intelligence and Agent Systems. 2010 ; Vol. 8, No. 3. pp. 255-268.
@article{1ca554393c594934beacd3cea6b0de53,
title = "Fluctuated peer selection policy and its performance in large-scale multi-agent systems",
abstract = "This paper describes how, in large-scale multi-agent systems, each agent's adaptive selection of peer agents for collaborative tasks affects the overall performance and how this performance varies with the workload of the system and with fluctuations in the agents' peer selection policies (PSP). An intelligent agent in a multi-agent system (MAS) often has to select appropriate agents to assign tasks that cannot be executed locally. These collaborating agents are usually chosen according to their skills. However, if multiple candidate peer agents still remain a more efficient agent is preferable. Of course, its efficiency is affected by the agent' workload and CPU performance and the available communication bandwidth. Unfortunately, as no agent in an open environment such as the Internet can obtain such data from any other agent, this selection must be done according to the available local information about the other known agents. However, this information is limited, usually uncertain and often obsolete. Agents' states may also change over time, so the PSP must be adaptive to some extent. We investigated how the overall performance of MAS would change under adaptive policies in which agents selects peer agents using statistical/reinforcement learning. We particularly focused on mutual interference for selection under different workloads, that is, underloaded, near-critical, and overloaded situations. This paper presents simulation results and shows that the overall performance of MAS highly depends on the workload. It is shown that when agents' workloads are near the limit of theoretical total capability, a greedy PSP degrades the overall performance, even after a sufficient learning time, but that a PSP with a little fluctuation, called fluctuated PSP, can considerably improve it.",
keywords = "cooperation, coordination, learning, load-balancing, Peer-to-peer computation, task allocation",
author = "Toshiharu Sugawara and Kensuke Fukuda and Toshio Hirotsu and Sato, {Shin Ya} and Osamu Akashi and Satoshi Kurihara",
year = "2010",
month = "7",
day = "30",
doi = "10.3233/WIA-2010-0190",
language = "English",
volume = "8",
pages = "255--268",
journal = "Web Intelligence",
issn = "2405-6456",
publisher = "IOS Press",
number = "3",

}

TY - JOUR

T1 - Fluctuated peer selection policy and its performance in large-scale multi-agent systems

AU - Sugawara, Toshiharu

AU - Fukuda, Kensuke

AU - Hirotsu, Toshio

AU - Sato, Shin Ya

AU - Akashi, Osamu

AU - Kurihara, Satoshi

PY - 2010/7/30

Y1 - 2010/7/30

N2 - This paper describes how, in large-scale multi-agent systems, each agent's adaptive selection of peer agents for collaborative tasks affects the overall performance and how this performance varies with the workload of the system and with fluctuations in the agents' peer selection policies (PSP). An intelligent agent in a multi-agent system (MAS) often has to select appropriate agents to assign tasks that cannot be executed locally. These collaborating agents are usually chosen according to their skills. However, if multiple candidate peer agents still remain a more efficient agent is preferable. Of course, its efficiency is affected by the agent' workload and CPU performance and the available communication bandwidth. Unfortunately, as no agent in an open environment such as the Internet can obtain such data from any other agent, this selection must be done according to the available local information about the other known agents. However, this information is limited, usually uncertain and often obsolete. Agents' states may also change over time, so the PSP must be adaptive to some extent. We investigated how the overall performance of MAS would change under adaptive policies in which agents selects peer agents using statistical/reinforcement learning. We particularly focused on mutual interference for selection under different workloads, that is, underloaded, near-critical, and overloaded situations. This paper presents simulation results and shows that the overall performance of MAS highly depends on the workload. It is shown that when agents' workloads are near the limit of theoretical total capability, a greedy PSP degrades the overall performance, even after a sufficient learning time, but that a PSP with a little fluctuation, called fluctuated PSP, can considerably improve it.

AB - This paper describes how, in large-scale multi-agent systems, each agent's adaptive selection of peer agents for collaborative tasks affects the overall performance and how this performance varies with the workload of the system and with fluctuations in the agents' peer selection policies (PSP). An intelligent agent in a multi-agent system (MAS) often has to select appropriate agents to assign tasks that cannot be executed locally. These collaborating agents are usually chosen according to their skills. However, if multiple candidate peer agents still remain a more efficient agent is preferable. Of course, its efficiency is affected by the agent' workload and CPU performance and the available communication bandwidth. Unfortunately, as no agent in an open environment such as the Internet can obtain such data from any other agent, this selection must be done according to the available local information about the other known agents. However, this information is limited, usually uncertain and often obsolete. Agents' states may also change over time, so the PSP must be adaptive to some extent. We investigated how the overall performance of MAS would change under adaptive policies in which agents selects peer agents using statistical/reinforcement learning. We particularly focused on mutual interference for selection under different workloads, that is, underloaded, near-critical, and overloaded situations. This paper presents simulation results and shows that the overall performance of MAS highly depends on the workload. It is shown that when agents' workloads are near the limit of theoretical total capability, a greedy PSP degrades the overall performance, even after a sufficient learning time, but that a PSP with a little fluctuation, called fluctuated PSP, can considerably improve it.

KW - cooperation

KW - coordination

KW - learning

KW - load-balancing

KW - Peer-to-peer computation

KW - task allocation

UR - http://www.scopus.com/inward/record.url?scp=77954933605&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77954933605&partnerID=8YFLogxK

U2 - 10.3233/WIA-2010-0190

DO - 10.3233/WIA-2010-0190

M3 - Article

VL - 8

SP - 255

EP - 268

JO - Web Intelligence

JF - Web Intelligence

SN - 2405-6456

IS - 3

ER -