应用扩展零力矩点预观控制和分解动量控制的仿人机器人全身运动规划方法  

A Whole-Body Motion Planner for Humanoid Robots Using EZMP Preview Control and Resolved Momentum Control

作　　者：周雪峰[1] 孙广彬[2] 刘晓光[1] 黄丹[1] 蒋晓明[1] 朱富贵[1] 

机构地区：[1]广东省科学院广东省智能制造研究所,广州510070 [2]南洋理工大学机器人研究中心,新加坡639798

出　　处：《西安交通大学学报》2016年第12期58-63,共6页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(51405091);广州市珠江科技新星专项资助项目(201504300920029)

摘　　要：针对简化动力学模型对运动规划造成较大误差的问题,提出一种仿人机器人全身运动生成方法,它结合扩展零力矩点(extended zero moment point,EZMP)的预观控制和分解动量控制,实现了含多种约束的高效运动规划。该方法给出了非线性EZMP方程在三维空间中的封闭形式,使得机器人稳定步态生成更加方便,同时将之前文献中忽略的角动量等因素也考虑在内,另外可以对EZMP、四肢运动和角动量等自由添加约束。仿真结果表明,该方法将EZMP平均误差减小到单一预观控制方法的2/5,并可以方便地实现含复杂约束的运动规划,为仿人机器人提供了一种形式简单且灵活、高效的运动规划方法。A whole-body motion planning method for humanoid robots is proposed to tackle motion planning errors caused by simplified dynamic model. The method combines the EZMP （extended zero moment point） control and the resolved momentum control to realize high efficiency motion planning with multiple constraints. A close form of the nonlinear EZMP equation in 3D space is presented to facilitate the stable motion pattern generation of humanoid robots. The angular momentum neglected in most previous literatures is also taken into account. Furthermore, some constraints such as EZMP, limb motion and angular momentum can be added. Simulation results show that the proposed method reduces the average EZMP error to around 2/5 of that using preview control alone and is able to realize motion planning with complex constraints. Therefore, the method provides a compact-formed, flexible and efficient motion planning solution for humanoid robots.

关 键 词：仿人机器人 扩展零力矩点 全身运动规划 角动量 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]