Simulation of object transportation by a robotic swarm with learning ability

Abstract

A transportation simulation of multiple objects to the appropriate goals by a robotic swarm has been performed with learning ability of the swarm. It is often observed that each insect in a colony seems to have its own agenda, and the group as a whole appears to be highly organized. By swarming, social insects achieve difficult tasks that an insect cannot manage alone. This collective behavior emerged from a group of social insects has been called “swarm intelligence”. Applications of swarm intelligence show high robustness and scalability, but, when it comes to applying to some intelligent task, the adjusting process of the parameters for a desired cooperative behavior may be a difficult, time-consuming task for a human designer. On the other hand, the artificial neural networks are known to have the learning ability. In this study, the swarm behavior in nature was simulated by “Boids” model. Boids model was known to be able to simulate the motion of a flock of birds by simple rules. Using this model, a cooperative transportation task has been simulated. The cooperative transportation task is to transport some objects by a swarm. In addition, the task of decision-making problem about transportation goal has been added. In this problem, the swarm of Boids makes a decision about the goal from the size of an object, considering each Boids as a unit of artificial neural networks, the network of Boids maps to the transportation destinations. Additionally, the back-propagation method has been applied to learn the appropriate goals. As a result, it is shown that the artificial swarm acquires the learning ability how to decide an appropriate goal for given objects.