Disturbance rejection of Central Pattern Generator based torque-stiffness-controlled dynamic walking

The concept of torque-stiffness-controlled dynamic walking expands the applicability of passivity-based bipeds while preserving energetic efficiency due to the addition of controllable stiffness. Central Pattern Generator (CPG) based approach introduces limb coordination and simplifies the control structure. In this paper, we study the disturbance rejection of torque-stiffness-controlled bipedal walking with CPGs. To evaluate the performance of rejecting disturbance, we study the locomotion under a perturbation torque and the walking performance on uneven terrains. In simulations, the performance of CPG-controlled walkers is compared with that of walkers without CPG, to study the impacts of CPG-based control. We also analyze the performance under different actuation parameters to clarify the effects of joint torque and joint stiffness on rejecting disturbances. The results may help better understand the flexible human gaits and build stable walking robots adapt to unpredictable environments.