Acquisition of adaptive behavior for virtual modular robot using evolutionary computation

Abstract

In areas such as evolutionary robotics and artificial life, simulating artificial robots and organisms are significant challenges to acquire their proper behaviors that achieve given tasks. This study proposes Animated Robot ("Anibot"), which can behave by obeying physical laws in a virtual 3D environment. Especially, we aim to obtain a control system in evolution which makes it possible to behave "Anibot" autonomously. This paper focuses on a virtual modular robot with a flexible structure and simulating it for learning and controlling. The experimental results show that the modular robot can move toward a light source as its goal in different circumstances. In addition, we discuss an adaptive ability in the different circumstances and a motion mechanism of an obtained behavior. © 2011 Springer-Verlag.