抄録
For generating motions of a robot, a simple method is needed. We use a human motion movie taken from one direction and made a simulator of the robot that displays its motion. Comparing both outline which divided for several poses, the robot imitates a human motion. There are some optimization problems for these poses. We propose the method using single population genetic algorithms to search for multi-problems. In GA each individual has several fitness functions. Therefore one GA process generates whole motion data. Experimental results show that proposed method generates a motion with decreasing the cost and the limitation of human.
