Contributed talk

On spatial pattern formation by autonomous mobile agents through language-like gossip communication

Masato Ishikawa

This study is concerned with behavior analysis of a group of autonomous mobile agents in a planar field, interacting with each other through language-like gossip communication protocol and learning. The proposed computation model is composed of a planar field and speaker agents moving around in the field. Each speaker is supposed to possess its own grammar, which has been randomly chosen at the initial state. Each speaker is able to (1) determine which direction to move in the next step, (2) encode the history of its recent actions with its own grammar, then tells it to the nearby agents , and (3) adjust the own grammar so that it better explains the other agents' words. Through numerical examination, we show that 'locomotion' is a moderate topic of conversation for grammar agreement, and also discuss influence of the individual behavior parameters on the global pattern formation.