Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 08, 2016 - June 11, 2016
Performance of swarm systems should be dependent on network structures resulting from various activities of agents. In the present study, we investigated properties of networks appeared through the activities of agents in a swarm system. The system called Ant System imitates a colony of ants in order to solve large scale problems with many agents acting based on local information. We observed the evolution of network structures constructed by the Ant System solving a Travelling Salesman Problem and investigated how they effect on the exploration for better solutions. In the early stage of exploration, nodes with relatively much links changed dynamically. In later stage, network structure showed relatively minor changes, suggesting gradual improvements of in the solution.
