Monday, February 28, 2022
Bipedal robots are designed to move like a human with two legs, and thus they have advantages over wheeled ones in traversing in human environments. The complexity of legs, however, imposes several challenges in the locomotion behavior realization. The motion of the internal degrees of freedom must be planned and controlled for balancing (in terms of not-falling); robustness must be realized to accommodate model discrepancy and external disturbances. Traditional approaches with ad-hoc style demonstrations still lack fundamental understanding, robustness, or efficiency for motion generations. I will argue what has been missing is the utilization of feedback structures in the hybrid dynamics of locomotion itself. To illustrate this, I will show how the investigation of feedback structures can effectively and elegantly provide solutions to the motion synthesis problem for realizing various dynamic underactuated bipedal walking behaviors such as periodic walking, path tracking, and push-recovery with significant computation-efficiency, versatility, and robustness. I will also touch upon some previous and ongoing works such as bipedal jumping, walking on granular terrain and stepping-stones, and blind traversing on stairs that share the same philosophy of using feedback structures of locomotion. The overarching goal is to thoroughly and rigorously understand and solve the bipedal mobility problem in structured and unstructured environments so that future bipedal robot-related applications such as package delivery (loco-manipulation) and robotic assistive mobility (via exoskeleton and prosthesis) can be safely, faithfully, and robustly realized in real life with guaranteed performances.