Evolutional interactivity in a swarm of robots

In this paper we consider the multi-robot system that collects target objects spread in the unexplored environment. Each single robot does not have sufficient sensors or manipulators to work. Therefore, the robots cooperates each other to improve the capability and the efficiency. The robots attract or intimidate each other, as behaviors of quorum sensing bacterial swarms or particles with electrical moments. In some environments, the interactions would increase the working efficiency, but in some others environments it would not. Therefore, the system has to decide how much interactions are effective in the given environment. The degree of interactions are expressed as a set of gene codes that means the weights of each kind of vectors, and adjusted by genetic algorithm to the optimized value adaptive to the working environment. The proposed approach has been validated by simulation. The result of this paper demonstrates that our multi-robot interacting system and optimizing algorithm reduces the working time to collect target objects, adaptively to the spreading characteristics of target objects.