Abstract
The purpose of this research is generation of stepping points for walking robot in obstacle environment based on the optimality of motion. A walking robot is assumed as a quadrupedal locomotion type and obstacles in a route are recognized to the extent necessary. There are obstacles with various sizes and configurations in human activity space. It is difficult to generate optimum stepping points for these patterns. The neural network learns stepping point parameters and corresponding obstacle environment to output the stepping points for the obstacle environment recognized by the robot vision. Stepping points are optimized by genetic algorithm and learned by neural network. The input data is the obstacle environment parameters that are generated by the genetic algorithm. The output data of the neural network is the stepping points. Workable stepping points are obtained by using proposed method.
