Autonomous decision on team roles for efficient team formation by parameter learning and its evaluation

D. Hamada, T. Sugawara*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)


We discuss a method of learning to determine appropriate roles for self-interested agents to efficiently form teams in task-oriented domains. A number of distributed applications are often expressed by task/resource allocation problems that can be modeled with a team formation problem in the multi-agent systems context so that tasks/resources are allocated to members of the team. Therefore, issues with efficient team formation have attracted our interest. The main feature of the proposed method is learning from two-sided viewpoints: team leaders who have the initiative to form teams or team members who work in one of the teams that are solicited. We introduce three parameters to agents for this purpose so that they can autonomously select their roles of being a leader or a member. Our experiments demonstrated that the amount of utility earned was considerably larger than that with conventional methods. We also conducted a number of experiments to investigate the characteristics of the proposed method. The results revealed that the proposed method could avoid excessive allocations to specific agents that had a large number of resources and disregarded agents that only had a few resources. Thus, teams could efficiently be formed.

Original languageEnglish
Pages (from-to)163-174
Number of pages12
JournalIntelligent Decision Technologies
Issue number3
Publication statusPublished - 2013


  • Team formation
  • assignment problem
  • autonomous decision
  • multi-agent systems
  • parameter learning
  • resource allocation
  • teamwork

ASJC Scopus subject areas

  • Software
  • Human-Computer Interaction
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence


Dive into the research topics of 'Autonomous decision on team roles for efficient team formation by parameter learning and its evaluation'. Together they form a unique fingerprint.

Cite this