CASIA OpenIR  > 毕业生  > 硕士学位论文
双足机器人综合稳定判据研究
Alternative TitleResearch on Synthetical Stability criterion of Biped Robot
图博
Subtype工程硕士
Thesis Advisor台宪青
2013-05-31
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline控制工程
Keyword双足机器人 稳定判据 机器学习 随机森林算法 Webots Biped Robot Stable Criterion Machine Learning Random Forest Algorithm Webots
Abstract双足机器人的研究始于上世纪七十年代,并于近几年进入了研究的高潮期,国内外先进的原型机更是各显神通,不断刷新着人类对双足机器人的固有认识。双足机器人高阶、非线性的特点吸引着不少控制理论与机械电子领域的科研人员从稳定判据、控制策略以及机械设计三个层面对其展开研究。其中,稳定判据作为控制与结构的基础,对双足机器人的稳定行走与操作控制起到了至关重要的作用。 本文主要研究双足机器人的综合稳定判据。研究了传统的基于特殊点位置的稳定判据以及基于庞加莱映射的稳定判据,并针对传统判据的特点提出了综合稳定判据的概念。基于ZMP稳定判据,规划了机器人的空间步行轨迹,并利用Webots机器人仿真软件搭建了机器人虚拟环境。引入了机器学习分类算法,针对特定数据集,通过对比得到了最优的分类算法。求得消耗函数的最小值,实现了成本与效率的折衷,完成了综合稳定判据的设计。具体内容包括以下几个方面: 1、对传统双足机器人稳定判据展开研究。首先,介绍了理想稳定判据应具备的特点。其次,对传统双足机器人稳定判据的原理进行推导,总结其特点并与理想判据进行对比。最后,结合传统与理想判据之间的差异,提出了包含传统判据优点并且可操作性能良好的综合稳定判据的概念。 2、规划机器人平衡步态,构建虚拟仿真环境。首先,基于零力矩点(ZMP)稳定判据,利用三次样条插值法规划得到光滑稳定的双足机器人步态轨迹。其次,使用Webots机器人仿真软件,搭建双足机器人本体模型,构造虚拟交互环境。最后,加载步态轨迹至机器人本体模型,控制其按规划行走。设计虚拟环境交互构架,完成轨迹、本体以及环境的实时交互。 3、生成数据集,确定基分类器。首先,阐明机器学习缘何引入,建立稳定判据与分类算法之间的映射,把研究角度转换到分类器的构造上来。其次,基于虚拟环境交互架构,引入随机扰动,生成正负样本实例。利用生成的样本实例训练分类器,并测试模型性能。最后,通过分析准确率、kappa统计参数以及ROC曲线等性能指标,确定随机森林模型为综合稳定判据的基分类器。 4、根据消耗函数,确定基分类器的工作点。结合理论背景对随机森林算法的理论基础进行推导,为后续工作中的算法改进奠定基础。接下来,结合消耗函数的性质,确定TOR指标与FPR指标的折衷点。最终,合适的工作点与基分类器共同构建了双足机器人综合稳定判据。
Other AbstractThe biped robot research began in the early 1970s, great progresses have been made in the recent several years. Incredible prototypes are constructed constantly, our original cognition are being subverted by the fabulous performances. Because the biped robot platform has attractive preferences of high order and nonlinear, many researches devote themselves to the development of stability criterion, strategy of control and mechanical designing. Stability criterion considered as the base of structure and control method is playing an important role in stable walking. This thesis aspires to propose a synthesis biped robot stability criterion. In the thesis, the principle of criterion based on feature points and Poincare projection are discussed, a concept of synthesis stability criterion is proposed based on the tradition ones. Smooth and stable joints trajectories are planned by the method of ZMP stability criterion and a virtual environment were constructed with the utility of Webots software which is concentrating on the robot simulation. Introducing a Machine Learning classify algorithm, the best classifier are applied after having trained and tested by the certain database. Estimation function is used to determine the trade off of the efficiency and cost. Finally, the basic classifier and the proper operating point will construct the frame of the synthesis stability criterion. Concretely, it includes several aspects, such as: 1. The study of traditional stability criterion. First of all, the thesis introduces what natures should an ideal stability criterion has. Secondly, the principle of traditional criterion is deduced, a comparison is made between the traditional criterion and the ideal one. Finally, considering the different between the tradition and the ideal, the thesis proposes an idea which may involve all the advances of traditional criterion. 2. Generate the stable gait, construct the virtual environment. Firstly, smooth and stable joints trajectories will be generated based on the ZMP criterion using the cubic spline interpolation. Secondly, a robot model and its virtual world file will be constructed using Webots software. Finally, gait which has been generated will be loaded on the robot model controlling the robot walk with the trajectory we set. 3. Build data set, obtain basic classifier. Firstly, the reason of why should we introduce machine learning is discussed; meanwhile, a projection between stability criterion and classifier is ...
shelfnumXWLW1922
Other Identifier2010e8014668004
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/7690
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
图博. 双足机器人综合稳定判据研究[D]. 中国科学院自动化研究所. 中国科学院大学,2013.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_2010e801466800(2780KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[图博]'s Articles
Baidu academic
Similar articles in Baidu academic
[图博]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[图博]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.