Bipedal walking control in dynamic environment using data mining techniques

In order to design stable walking for a bipedal robot over uneven terrain, advanced control methods such as nonlinear control and receding-horizon control, and exact hybrid dynamics are needed. They are too complicated to be used in the many applications. In this paper, we use data mining techniques, locally weighted learning, principal component regression and regression clustering, and combine with the classical proportional-integral-derivative control. The biped model also uses the observation of human walking. The model structure consists of locally linear modules and principal component regression groups. Experiments and analysis are given to evaluate the effectiveness of our novel method.