CORC  > 自动化研究所  > 中国科学院自动化研究所  > 毕业生  > 博士学位论文
题名四足机器人环境适应行走的算法及实验研究
作者邵雪松
学位类别工学博士
答辩日期2013-05-29
授予单位中国科学院大学
授予地点中国科学院自动化研究所
导师杨一平 ; 王伟
关键词四足机器人 环境适应行走 步态规划 步态控制 地质检测与分类 耦合动力学 Quadruped Robot Environment-Aware Adaptive Locomotion Gait Planning Gait Control Geology Detection and Classification Couple Dynamics
其他题名Algorithms and Experiments of Environment-Aware Adaptive Locomotion for Quadruped Robots
学位专业计算机应用技术
中文摘要四足机器人因具有较高的动态性能和较强的环境适应能力而备受关注。作为一种强耦合非线性复杂动力学系统,四足机器人涉及学科知识多,模型结构复杂,尚有许多基础理论与关键技术有待进一步研究。本文围绕四足机器人节律柔顺行走、复杂地形适应行走和不同地质环境适应行走等方面展开研究,并在四足机器人仿真模型和物理平台上进行分析验证。本文完成的主要工作和贡献如下: 1. 针对节律运动突变碰撞力大和柔顺性低的问题,改进了基于Hopf振荡器的中枢模式发生器模型,提出了一种节律柔顺行走控制方法。分析Hopf振荡器输出信号与关节运动之间的关系,整合膝关节变量,改变神经元之间的作用关系,实现了对称步态和非对称步态行走;分析节律运动碰撞力突变产生的负面影响,提出了基于碰撞力大小和四足机器人身体姿态的柔顺性评估方法;连续调整碰撞阶段大腿的摆动幅度,增大摆动周期,从而减小碰撞阶段的关节运动速度,形成机器人本体与地面之间的缓冲,在一定程度上实现了节律柔顺行走。 2. 在基于站立相和摆动相逆动力学模型的基础上,设计了一种耦合逆动力学建模方法。采用将四足机器人四条腿视为四个单独串联机构的思想,分别建立站立相和摆动相阶段的逆动力学模型,并对站立相和摆动相之间的过渡过程建模,实现模型的平滑切换;加入串联机构之间的耦合作用力因素,在减少模型计算复杂度的同时增加了模型的可靠性。 3. 为了实现障碍环境下的稳定柔顺行走,提出了一种足端轨迹优化方法。针对已知环境信息,设计混合抛物线路径规划算法,规划柔顺高效的足端无碰路径,根据速度、加速度约束,利用非线性优化方法,实现了时间最优的关节轨迹规划;设计了障碍环境下基于翻转势能和稳定裕度的重心轨迹调整方法,保证了越障的稳定性。 4. 为了提高未知复杂地形条件下的自主运动能力,设计了基于双目视觉的环境适应行走步态规划方法。利用双目摄像机感知环境地形,采用绝对误差和的立体匹配方法估算视差值,构建基于统计和几何关系的视差图滤波处理算法,利用随机采样一致性算法提取摄像机视角,并根据四足机器人运动学模型构建栅格地形图;针对存在凸起障碍和禁行区域的复杂地形,根据稳定性、柔顺性和高效性原则设计步态规划算法,通过多步态生成、步态变换、混合抛物线足端轨迹规划和关节轨迹插值实现复杂地形下的适应行走。 5. 针对环境地质属性难以检测且不适合腿足机器人行走的情况,建立了机器人与地面之间的耦合动力学模型,提出了针对不同地质环境的分类检测方法和适应行走策略。将机器人与地面环境视为一个整体,构建了闭环系统模型、柔顺碰撞模型和摩擦力模型,分析行走过程中路面基底对机器人产生的动力学影响;根据传感-电机一致性思想,将各关节电机的电流信号和位置误差信号引入检测系统,与足底碰撞力一起检测机器人身体动力学参数变化,设计了一种多级加权kNN分类算法用于消除检测变量作用不同和采样不均带来的影响;通过分析足-地碰撞力和能量耗散特性,提出了基于触地角调整的地质适应策略,实现了不同地质环境下的适应行走。
英文摘要Quadruped robots have drawn much attention because of their higher dynamic performance and better environment adaptability. As a heavily coupled and strong nonlinear dynamics system, the quadruped model is complicated and the research involves multi-subject knowledge, so many key issues remain to be further discussed. This research focuses on rhythmic compliant locomotion, complicated terrain locomotion, and geological adaptive locomotion based on the central pattern generator and the perception-modeling-planning technique. The control methods are verified on the simulation quadruped model and the real quadruped platform. The main contributions of this thesis are summarized as follows. 1. A rhythmic compliant control method is proposed through improving the central pattern generator modeled by Hopf oscillators. Considering the relationship between the Hopf oscillator outputs and joint motion, the quadruped robot can locomote on symmetric gaits and asymmetric gaits by integrating knee joint variables and changing the interaction of neuron oscillators. Based on the analysis of the side effects of the mutation compact force, this research develops the compliance evaluation approach in accordance with the size of the compact force and the quadruped robot pose. The rhythmic control method, which implements the adjusting of the swing amplitude of the thigh and the swing period of the leg, not only can reduce the collision speed but also can form a cushion between the quadruped body and the ground to accomplish the rhythmic compliant locomotion. 2. This research improves the couple inverse dynamics model, motivated by different dynamics characteristics during the swing phase and the stance phase. The quadruped robot is considered as four independent series mechanisms. Different inverse dynamics models are built for the swing phase and the stance phase respectively for the purpose of low computational complexity. In order to switch the models smoothly, the transition stage between the swing phase and the stance phase is modeled. Meanwhile, coupling factors of different series mechanisms are added into the dynamics models to improve the integrity of the whole quadruped robot model. 3. A foot trajectory optimization method is developed for obstacle overstriding. To get the smooth and efficient collision-free path for known obstacle models, this research proposes a mixed parabola path planning algorithm. The time optimal joint trajectories can be obtained through...
语种中文
其他标识符201018014629090
内容类型学位论文
源URL[http://ir.ia.ac.cn/handle/173211/6535]  
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
邵雪松. 四足机器人环境适应行走的算法及实验研究[D]. 中国科学院自动化研究所. 中国科学院大学. 2013.
个性服务
查看访问统计
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。


©版权所有 ©2017 CSpace - Powered by CSpace