Unsupervised learning of two legged robot

A simple bipedal locomotion is considered to study the learning processes in a self-organized fashion. The author introduces two different concepts, self-learning and self-training. The former is a step by step learning to a better performance starting from a bad moving style. The latter is concerned with how to keep the performance after arriving at the best one. The author also discusses how to manipulate two different moving modes, walk and rotation.<
>