BASC, a bottom-up approach to automated design of spatial coordination

The design of spatial coordination mechanisms for a group of agents is a challenging problem to which the top-down approach of traditional Artificial Intelligence often fails to find convincing solutions. Bottom-up approaches have given more promising results, but they often rely on a tedious manual hand-crafting that raises scalability issues. In this paper, we show that combining a bottom-up spatial coordination mechanisms with specially designed evolutionary methods to search the space of solutions is an efficient approach to such problems. More precisely, we show the benefits of our platform, BASC, through a quantitative comparison with previous work published by Balch and Hybinette and we conclude on the methodological issues raised by our work.