A robust walking controller based on online step location and duration optimization for bipedal locomotion.

arXiv: Robotics(2017)

引用 5|浏览67
Step adjustment for humanoid robots has been shown to improve gait robustness, while timing adjustment is often neglected in control strategies. In this paper, a new walking controller is proposed that combines both step location and timing adjustment for generating robust gaits. In this approach, step location and timing are decided, based on the Divergent Component of Motion (DCM) measurement. We define the DCM offset as the offset between the DCM and landing location of the swing foot at landing time, and employ it to split state space into viable/non-viable parts. Constructing our walking controller based on the DCM offset, we can exploit the whole capability of a biped robot in terms of stepping to recover from disturbances. The proposed approach is comprised of two stages. In the first stage, the nominal step location and step duration for the next step(s) are decided. In this stage, the main goal is to schedule the gait variables far from constraint boundaries for a desired walking velocity. The second stage adapts at each control cycle the landing position and time of the swing foot. By using the DCM offset and a change of variable for the step timing, we can formulate the second stage of our controller as a small sized quadratic program without the need to preview several steps ahead. To map the adapted gait variables to the full robot, a hierarchical inverse dynamics is employed. Interestingly, our approach does not require precise control of the center of pressure and can also be used on robots with passive ankles or point feet. Simulation experiments show a significant improvement in robustness to various types of external disturbances, such as pushes and slippage, compared to state of the art preview controllers where step timing is not adjusted. In particular, we demonstrate robust walking behavior for a simulated robot with passive ankles. Keywords
AI 理解论文