A Study on Intelligent Dynamic Walking Control of Biped Robots Based on Sensor Information

本研究旨在实现机器人的三维动态往复运动。在人体动态行走中，不稳定的运动就像一个倒立摆，通过反复的运动形成一个稳定的极限环。本文提出了一种利用多个传感器信息的分层控制策略，并在爬楼梯的环境下，利用仿人机器人进行了实验，验证了该策略的有效性。（1）利用双支撑阶段的动力学方程，确定了在有楼梯的情况下的最佳力矩分配。在该计算中，关节扭矩的总和在鞋底等不发生滑动的限度内被最小化。该扭矩分布存储在控制计算机中，并用于行走实验。（2）从机器人系统的动力学方程出发，推导出了在有楼梯的情况下机器人系统角动量的最优变化规律。（3）利用上述力矩分配和角动量的最优变化进行了连续上升楼梯的行走实验。

This study study aims at realizing 3D dynamic biped locomotion in a robot. In human dynamic walking, a stable limit-cycle is formed as a whole by repeating the unstable motions as an inverted pendulum. In this study, a hierarchical control strategy in which the information from many sensors was utilized was proposed, and its effectiveness was confirmed by experiments under the circumstamces with stairs using a biped robot.(1)The optimal torque distribution under the circumstance with stairs was decided by using the dynamic equations of motion during the double support phase. In this calculation, the summation of the joint torques is minimized within the limits in which the slippage of the sole and like do not happen. This torque distribution is stored in the control computer, and is used in the walking experiments.(2)The optimal variation of the angular momentum of the robot system under the circumstance with stairs was also derived from the dynamic equations of motion.(3)The walking experiments on the continuous ascending stairs were conducted by using the above torque distribution and the optimal variation of the angular momentum.