Abstract
This paper describes a coordinative behavior of virtual robots. Virtual robots move on a two-dimensional plane by the law based on the Newton's equations. Motions of virtual robots are controlled by an artificial neural network (ANN). Parameters of an ANN are optimized by a real-coded genetic algorithm so that virtual robots cooperate on catching targets. Targets escape from virtual robots in the opposite direction of the nearest virtual robot. The experimental result shows that two virtual robots are located in front and the rear of a target like a straight line to catch it. It is also shown that the role of a waiting virtual robot changes dynamically to surround a target when the target breaks through the besieging two virtual robots.
