Study on principle of optimal control for energy-efficient walking systems

点击翻译按钮获取中文摘要

We have studied robotic biped gait generation based on its dynamics and improvement of the gait efficiency theoretically and experimentally. It is empirically known that the convex curve of the foot, which characterizes passive-dynamic walkers, has important effects on increasing the walking speed. We mathematically clarified the effects of semicircular feet; acting as the virtual ankle-joint torque during the stance phase and reducing the energy dissipation caused by heel strikes, and confirmed that they dramatically improve the gait efficiency through numerical simulations. Parametrically excited dynamic bipedal walking is also our main research subject. We finally completed an experimental walking machine and succeeded its stable level walking by adjusting the physical and control parameters. The experimental result was highly-appraised internationally. In the theoretical investigation, we have considered to achieve parametrically excitation by actuating the knee joint, and successfully applied the effect to level gait generation. We further proposed several advanced methods; parametric excitation based on up-and-down motion of robot's center of mass and ornithoid walking by inverse knee-joint bending, and compared their gait efficiency. Furthermore, we have investigated the gait stability of limit cycle walkers from the viewpoint of mechanical energy balance. We then identified the conditions necessary to systematically achieve this energy balance and clarified their physical meanings. We also considered applying a bisecting hip mechanism (BHM) for adding an upper body to passive-dynamic walkers without destroying their natural dynamics. By clarifying the robot's driving mechanism incorporating the BHM, we successfully achieved its efficient level dynamic walking. Throughout our investigations, we provided novel insights and deep understandings into how energy-efficient and ZMP-free robots can generate a dynamic bipedal gait.