Abstract
Recent expectations of what computational neuroscience contributes to the development of robotic technologies are remarkable. In dynamic environments, robots are required to work as an online and on-demand system, like humans do. However a simple conjunction of neural network models merely provides a pre-ordered circuit to control the robot, and then such a constructive approach casts doubt on the strength of the reality of brain-based robots. Nowadays, neuroscientists know the limitation of the traditional analytic science and start to explore a kind of the synthetic approach as the integration of past knowledge that have been investigated analytically. In this study, we discuss on the direction to the synthetic approach through the neuro-robotics.