Dynamics Based Control of Humanoid Robot

在本研究中，我们从以下几个角度考虑步行控制。也就是说，行走控制的基本原理是被动动态行走。因此，我们一直在研究利用准被动行走机器人实现准被动行走，这种准被动行走机器人设置在缓坡上。具体地说，我们开发了一种由直驱电机组成的关节行走机器人。主要研究成果如下：(1)设计并开发了两足机器人PB-I。这个机器人有两条腿，每条腿都有树状关节。关节执行器采用直接驱动电机。(2)建立了PB-I的动力学模型。并将机器人作为一个非线性动力学系统进行了分析。然后研究了机器人被动行走的可实现性，并进行了稳定性分析。(3)提出了准被动行走的控制律。在此基础上提出了一种学习控制方法。(4)针对行走过程中的一个碰撞点，利用庞加莱地图，试图理解该地图的结构。(5)提出了一种连续型延迟反馈控制。

In this research, we consider the walking control from the following point of view. That is, the essential principle in walking control is a passive dynamic walking. Therefore, we have been researching a Quasi-passive walking by using a quasi-passive walking robot which is set on a slight slope. Concretely speaking, we developed a walking robot whose joint consists of direct-drive motor. The following are the results.(1)We designed and develop a biped robot PB-I. This robot has two legs and the each leg has tree joints. Direct drive motor is used for the joint actuator.(2)We obtained a dynamical model of PB-I. And we analyzed the robot as a non-linear dynamical system. Then, we researched a realizability of passive walking of the robot and a stability analysis.(3)We propose a control law for Quasi-passive walking. At that time, a kind of learning control method was proposed.(4)We focused on a impact point in walking process, and using a Poincare map, we tried to understand the structure of the map.(5)We proposed a kind of continuous type delayed feedback control.