Energy Efficient Walking Control For Underactuated Biped Robots

Compared to the wheeled locomotion and the pedrail locomotion, the bipedal locomotion has a serial of advantages in the adaptability of the complex environment. Because the movement style of the bipedal locomotion is close to the one of human being, robots with bipedal locomotion are better candidates taking the place of humans working in dangerous environments. However, high energy consumption of biped robot walking is one of the critical factors, which impedes the development and application of biped robots. Hence, the study of efficient bipedal walking becomes very active in recent years in the field of robotics. This dissertation studies how to achieve high efficient and stable bipedal walking from the perspectives of designing the auxiliary devices on the joint and promising control schemes.Firstly, based on the model of compass-like biped robot with underactuated ankles, the efficiency and stability of bipedal walking affected by the constraint mechanism on the hip joint are studied. By comparing different scenarios of using a constraint mechanism, it is discovered that the retract constraint style is beneficial for walking in the level ground with little torque input. Then, contrasting to different methods of gait generation, retract constraint style with constant torque is helpful for achieving more efficient walking gaits. By analyzing the stability of the walking gaits with different step length and different torque inputs, it is found that bipedal walking changes from stable to unstable when the step length increases. For an unstable walking gait, the event-based control is applied to stabilize it, and then efficient and stable bipedal walking is achieved.Secondly, for the planar biped robot with underactuated ankles, a general control method to stabilize the walking gait to a reference one is studied. The most common planar biped robots with underactuated ankles are classified, and using the model of five-link biped robot, a unified feature that the angular momentum about the stance point has three degree of relationship with the inputs is summarized. Based on this feature, a method based transverse coordinate transform is proposed in the first time. With this method, the problem of reference trajectory tracking can then be transformed to a stabilization problem of a linear time invariant system with uncertainties in moving Poincare sections. Combining with the Poincare map, a feedback control design approach is introduced, and the state feedback controller used for stabilizing an unstable walking gait is obtained. From the simulation for a compass-like biped robot model, it is observed that the feedback controller based on transverse coordinate transformation is effective to stabilize the trajectory to the reference walking gait that is unstable in open-loop. Then, comparing to the event-based control, the feedback controller based on the transverse coordinate transformation exhibits advantages in both attraction basin and convergence rate.Finally, the high efficient walking style with a constraint mechanism is extended to a three-dimension(3D) compass-like biped robot with underactuated ankles. The robot has two underactuated degrees of freedom in the ankle joints and one actuated hip joint. As the high efficient walking gait in open-loop is unstable, a discrete transverse linearization method is proposed to stabilize the walking gait. With this method, the limit cycle stabilization problem is transformed to the point stabilization problem of an equivalent linear discrete periodic system. Then, using discrete-time linear quadratic regulator(DLQR) theory, a periodic feedback controller is obtained to stabilize the reference limit cycle. Simulation results show that the discrete transverse linearization control can be used to achieve an efficient and stable walking gait for the3D compass-like biped robot. With comparison to the conventional event-based control, the discrete transverse linearization controller results in a larger basin of attraction and faster convergence rate, which shows the advantage of this method.