仿人双足机器人多地形步态规划和稳定控制方法研究

发布时间：2018-01-09 04:13

本文关键词：仿人双足机器人多地形步态规划和稳定控制方法研究　出处：《中国科学技术大学》2016年博士论文　论文类型：学位论文

【摘要】：仿人双足机器人具有人类的形状并能像人一样活动,能在人所处的现实环境中工作,并且能够使用人所用的工具,有关它的研究和发展代表了机器人研究的顶尖水平。它集多学科交叉应用,代表了一个国家在机器人领域的最高成就。稳定步行是仿人双足机器人在多任务地形的环境中作业的基础,也是仿人机器人研究的难点和热点。本文以美国国防部举办的机器人挑战赛为背景,在总结国内外研究成果的基础上,着重围绕仿人机器人多种地形的步态规划和稳定控制方法展开研究。本文的主要研究内容和创新点在如下几个方面：1.基于三质心模型的髋关节高度和姿态补偿步态生成方法：目前仿人机器人所实现的步行方式与人类自然步态相比,在步态平滑自然性、单位移动距离能耗以及步行频率和速度方面具有较大差距。以人体为参照,利用Vicon Nexus运动捕捉系统搭建人体自然步行运动研究平台,通过捕捉人体各关节运动轨迹,提取人类行走的特征数据,提出了基于三质心模型的髋关节高度和姿态补偿的步态生成方法,该方法可以有效解决单质心模型不精确的问题,并通过地毯和柏油路步行实验证明了该方法的有效性和稳健性。2.仿人机器人连续障碍物跨越方法：首先提出仿人机器人的障碍物连续跨越问题,定义了两种障碍物连续跨越类型—稀疏障碍物跨越和紧密障碍物跨越；然后提出了连续障碍物跨越的可行性分析方法,使用腿轨迹规划调整与上半身质心运动轨迹补偿的方法修正最终全身运动轨迹；最后,通过OpenHRP仿真平台和XT实验平台分别开展两种障碍物连续跨越的实验,验证了该方法的有效性和可行性。3.仿人机器人通用斜面步行模式生成方法：提出了一种考虑仿人机器人腰部姿态和低落地冲击的斜面步行模式生成方法—不对称倒立摆斜面步行模式生成方法。首先提出了基于虚拟地面方法的机器人单质心模型的二维和三维运动方程,获得了斜面连续直观的ZMP方程。然后分析了腰部姿态对斜面行走的影响,提出并使用不对称倒立摆模型实现仿人机器人斜面稳定动态行走,该方法可以有效解决落地冲击问题。最后使用OpenHRP模拟仿真实验和XT机器人平台斜面动态行走实验验证了上述新方法的有效性和可行性。4.仿人机器人两级稳定步行控制器：设计的稳定器由两级稳定器组成,前级稳定器包含髋关节力矩补偿器、落地控制器和ZMP补偿控制器,其主要作用是令机器人当前ZMP向目标ZMP移动,减小这两点之间的距离；后级稳定器包含上半身躯体控制和落地位置修正控制器：其主要作用是当在机器人目标ZMP和当前ZMP点几乎重合的情况下仍有跌倒趋势时(上半身倾斜过大),通过突发式的调整上半身加速度重新规划目标ZMP以恢复机器人姿态并修正摆动腿的着地位置。两级稳定步行控制器保证了仿人机器人在多种地面的稳定步行。
[Abstract]:Humanoid biped robot has human shape and human like to play, can work in the real environment, and can use the tool to use, research and development about it represents the top level of the robot research. It sets the interdisciplinary, representing the highest achievement of a country in the field of robotics. Walking stability is the foundation of humanoid biped robot operating in multi task terrain environment, and is also a difficult and hot research of humanoid robot. The robot based on the challenges the U.S. Department of defense held the tournament as the background, based on summarizing the domestic and foreign research results, focusing on the humanoid robot variety of terrain the gait planning and stability control method is researched in this paper. The main research contents and innovations in the following aspects: 1. based on the hip height and gait generation three centroid attitude compensation model Method: the humanoid robot the way of walking and human natural gait, gait in smooth nature, a wide gap between the unit energy consumption and the moving distance of walking speed and frequency. To the human body as the reference, capture system set up human natural walking motion research platform using Vicon Nexus movement, through the capture of human joint motion trajectory the characteristics of human walking, data extraction, proposed the gait generation method of hip joint height and attitude compensation three centroid based on model, the method can effectively solve the problem of inaccurate elemental heart model, and through the carpet and asphalt road walking experiments demonstrate the effectiveness and robustness of the.2. humanoid robot the method of continuous obstacle crossing methods: firstly, the obstacles of humanoid robot continuous crossing problem, defines two types of sparse obstacles straight across the obstacle crossing The more closely and obstacle crossing; and then put forward the feasibility analysis of continuous obstacle crossing method, method of using leg trajectory planning and adjustment of the upper body centroid trajectory correction of the final compensation body movement trajectory; finally, through OpenHRP simulation platform and XT experimental platform were carried out two kinds of obstacle crossing continuous experiments validate the method the validity and feasibility of the.3. humanoid robot walking pattern generation method: general slope presents a slope considering the waist of humanoid robot posture and low landing impact the walking pattern generation method of asymmetric inverted pendulum inclined walking pattern generation method is proposed. Firstly, equation of two-dimensional robot elemental heart model and 3D virtual ground method based on motion, ZMP equation was obtained. Then the analysis of inclined continuous visual effects on the waist posture cant walk, put forward and used Model of humanoid robot walking stable slope dynamic asymmetric inverted pendulum, the method can effectively solve the problem of landing impact simulation experiment. Finally, using OpenHRP and XT robot platform slope dynamic walking experiments show the new method is effective and feasible.4. humanoid robot walking stability controller design: grade two from grade two stabilizer stabilizer before the class, including stabilizer hip joint torque compensator, landing controller and ZMP controller, its main function is to make the current ZMP mobile robot to the target ZMP, which reduces the distance between two points; after class contains the upper body stabilizer control and landing position correction controller: its main function is when the robot target ZMP and current ZMP almost the same situation still fall trend (upper body incline too much, through the adjustment of the burst mode) and the upper part of the body The target ZMP is reprogrammed to restore the robot's posture and correct the landing position of the swinging leg. The two level stable walking controller ensures the stable walking of the humanoid robot on various ground.

【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP242

【相似文献】