Collective-movement teams for cooperative problem solving

Self-organizing particle systems ("swarms") consist of numerous autonomous, reflexive agents ("particles") whose collective movements through space are determined primarily by local influences. Such systems have been traditionally used to simulate groups of animals and other biological phenomena. The simple nature of these systems limits their applications in other areas. We believe that by extending these systems and combining them with a top-down approach, they can be transformed into a general problem-solving technique. In this work we present an agent architecture derived by adding goal-directed control mechanisms to particles that allow them to switch between multiple dynamics according to different situations and to keep and propagate information relevant to solving a problem. Simulations of two different tasks related to search for and transportation of objects show that agents are able to not only effectively solve the assigned problems, but that they do so in a more efficient manner than similar but independently moving agents. Further, collectively moving agents display coordination emerging purely from their local interactions and this cooperation provides a clear advantage while solving a problem as a team. These results show that the self-organizing behavior of particle systems can be extended to support problem solving in various areas such as coordinated robotic teams.