CPG based control for generating stable bipedal trajectories under external perturbation

This paper presents a system of central pattern generators (CPGs) made of a network of mutually coupled Matsuoka nonlinear neural oscillators. The CPG based controller is used to generate human like rhythmic motion for biped robots. The new design has been investigated through interaction between the dynamics of the coupled CPGs and the dynamics of a five-link musculoskeletal robot model with six DOFs. The scaled outputs of the mutually inhibited oscillators are utilized as torques to actuate the relevant joints of the robot during walking. The effectiveness of the proposed approach is examined through simulations and the results are compared to those of Taga´s CPG model, where better and efficient generation of stable rhythmic walking trajectories is shown under external perturbation to disturbances.