Improvement and Stavilization of Multiagent Resource Share Problem by Machine Learning and Exhaustive Simulation

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K00328
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: National Institute of Advanced Industrial Science and Technology
• Principal Investigator: Noda Itsuki 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 総括研究主幹 (40357744)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: multiagent / resource shareing / learning / simulation / social simulation

Outline of Final Research Achievements

As the bottom-up approach, we focused on multi-agent resource sharing problem with dynamically changing resources. We proposed a method called "Win or Update Exploration-ratio (WoUE)", in which each agent adjusts its exploration ratio through learning independently.
As the top-down approach, we picked up large scale pedestrian control at big events. We evaluated two top-down control methods, that are, fixed and adaptive dividing guidance of peoples via a simulator, and compared with real data.

Free Research Field

人工知能