文摘
Evolutionary swarm robotics (ESR) is an artificial approach for developing smart collective behavior in a system of homogenous autonomous robots. Robot behavior is generally controlled by evolving artificial neural networks. ESR has been considered a promising approach for swarm robotics systems (SRSs), because swarm behavior naturally emerges from numerous local interactions among the autonomous robots. In contrast, programming individual robots to display appropriate swarm behavior is extremely difficult. However, even in a simulated SRS, ESR is precluded by a very high computational cost. In this study, we introduce a novel implementation that overcomes the computational cost problem. The method employs parallel problem solving on a graphics processing unit (GPU) and OpenMP on a multicore CPU. To demonstrate the efficiency of the proposed method, we engage an evolving SRS in a food-foraging problem.