Implementation of a Disturbance Estimation and Compensation Model for a Dynamically Walking Bipedal Robot

It is the key requirement for bipedal robot to maintain balance and upright posture under unexpected external disturbances. Several control methods have been studied throughout the world over the last decades focusing on bipedal balancing in the scope of technical approaches. However, there are still limitations such as high energy consumption, high dependency on mathematical models, lack of postural control on unexpected disturbances. Consequently, locomotion of bipedal robot is far from achieving graceful motions compared to dynamic locomotion of human. This thesis presents a novel balancing concept based on a biologically motivated control method, which is proposed by Mergner et al. [Mergner 10]. The disturbances of gravity forces, supporting-surface tilt, external pushes and translational movement of plate are estimated respectively according to the outcome of sensory systems in human. Those disturbances are integrated and fed back to the local joints to generate torques and angular movement. The balancing in the sagittal plane and the frontal plane are implemented respectively with active control of ankle, hip, shoulder and knee joints. By implementing the DEC model into a hierarchical structured control system, the adaptability and robustness of DEC method are verified by various experiments.