四足机器人环境适应行走的控制策略研究

CASIA OpenIR > 毕业生 > 博士学位论文

	四足机器人环境适应行走的控制策略研究
	李晓琪
	2017-05-27
学位类型	工学博士
中文摘要	四足机器人机动性高、环境适应性强，在灾难救援及特种作战等领域具有重要应用价值。其数学模型具有非线性强耦合的特点，建模及控制难度大，特别是在未知不确定环境中运动时，四足机器人的控制尤为复杂。本文针对高度受限环境及地面基质变化环境中四足机器人运动的控制策略展开研究，包括四足机器人快速步态变换控制、行走过程中的重心高度控制、地面基质分类方法与适应行走策略等，旨在提高四足机器人在复杂环境中的运动适应性及机动性。论文核心内容概要如下：（1）提出了一种四足机器人快速步态变换策略，实现了四足机器人在一个步态周期内快速平滑的步态变换，以适应地形或地质环境因素变化。该策略利用基于Hopf振荡器的中枢模式发生器（Central Pattern Generator, CPG），实现四足机器人基本运动步态；设计相位调节模块，改变节律信号相位；协调节律信号相位关系，实现四足机器人快速步态变换。该策略可使足端-地面作用力保持在一定范围内，并使机器人本体俯仰角及横滚角平滑柔顺。快速步态变换策略的实现为四足机器人复杂环境适应行走奠定了理论基础。（2）针对约束环境下的有效四足行走，提出了一种基于离散跟踪微分器（Tracking Differentiator, TD）的四足机器人重心高度控制方法，通过节律信号振荡中值的平滑过渡，实现机器人行走过程中的重心高度调节。该方法利用Hopf振荡器构成中枢模式发生器，以生成四足机器人节律运动控制信号；利用改进的离散跟踪微分器实现四足节律信号振荡中值的平滑过渡，且过渡时间可控。同时，分析了四足机器人重心高度调节过程中导致偏离行走方向的因素，设计了基于位置和姿态的侧向偏移控制策略，调节控制信号幅值，有效减小了行走方向的侧向偏移；分析了四足机器人重心高度与足端-地面作用力及各关节驱动力矩的内在关系，验证了四足机器人重心高度控制方法的有效性和可行性。（3）分析了四足机器人CPG变参数行走时足地接触力的动态特性，探索了不同地面基质情况下四足机器人足地接触力的变化。在四足机器人平台上实现了Walk及Trot节律运动控制实验，分析了CPG控制下Trot步态不同步长和步频行走时足地接触力的动态特性，为提高四足机器人动态性能奠定了基础。Walk步态行走实验探索了不同地面基质情况下四足机器人足地接触力的变化，分析了利用足地接触力进行地面基质识别和分类的有效性，为地面基质分类及适应行走的研究提供了理论依据。为进行实验研究，设计了基于ARM+DSP架构的电控系统，ARM控制器实现机器人运动规划及传感器信息处理，DSP控制器实现多关节驱动控制，该电控系统可使四足机器人实现自包含。（4）针对四足机器人在诸如摩擦系数、弹性模量等物理属性不同地面基质环境中的稳定行走问题，提出了基于支持向量机（Support Vector Machine, SVM）算法的地面基质分类方法。利用四足机器人在不同地面基质行走过程中的足地接触力及机器人本体PITCH-ROLL信息，提取特征向量，采用SVM算法进行地面基质分类。利用四足机器人Biodog在摩擦系数和弹性模量不同的五种地面基质上进行了实验研究，所提出的地面基质分类方法分类正确率为99.33%。（5）为了改善基于Hopf振荡器的四足机器人Walk步态在不同地面基质环境的行走性能，增强四足机器人在不同地面基质环境中的适应性，提出了四足机器人地面基质适应行走策略。利用重心位置调节方法，消除CPG输出节律信号相位与机器人腿足实际相位的偏差；利用CPG和足端轨迹规划结合的方法，实现对四足机器人腿足最大抬起高度的直接控制。地面基质适应行走策略有效改善了四足机器人Walk步态行走过程中的摆动腿足拖地现象，增强了四足机器人在不同地面基质行走的稳定性和适应性。
英文摘要	Quadruped robots have the ability of powerful maneuverability and great adaptability, and have significant application value in the fields such as disaster rescue and special operations. The mathematical model of quadruped robots is nonlinear and coupled, so it is complicated to model and control quadruped robots. Especially, when quadruped robots locomote in uncertain environments, control will be more challenging. Aiming at improving maneuverability and adaptability of quadruped robots in confined environments and terrains with different geologies, this thesis investigates several key problems of quadruped robot control, including fast gait transition, adjustment of the height of Center of Gravity (COG) during locomotion, ground substrate classification algorithm, and adaptive locomotion on terrains with different geologies. The main contributions of this thesis are summarized as follows. To adapt to terrain and geology changes, a fast gait transition approach for quadruped robots is presented. Smooth and fast gait transition can be obtained within a step cycle period of locomotion. Central Pattern Generators (CPGs), modeled by Hopf oscillators, are utilized to produce basic rhythmic gait pattern for a quadruped robot. Phase-adjustment control is employed to regulate the phase of rhythmic outputs generated by CPGs, and fast gait transition is realized by harmonizing the phase relationship of the rhythmic outputs. Simulation results illustrate that the Ground Reaction Force (GRF) is within a reasonable range; besides, the pitch angle and the roll angle during locomotion have no abrupt changes. The fast gait transition approach lays the basis for adaptive locomotion of quadruped robots in complex environments. Aiming at generating effective quadruped locomotion in confined environments, a COG-height control approach based on the discrete Tracking Differentiator (TD) is proposed. Smooth transition between two different rhythmic medium values is employed to realize the adjustment of the height of COG during locomotion. The CPGs modeled by Hopf oscillators are utilized to produce basic rhythmic motion for a quadruped robot. The modified discrete TD is creatively employed to implement the transition between two different rhythmic medium values of the CPGs, and can control the transition duration. Considering the deviation of COG in the transverse direction during the transition, we give a control strategy based on position and orientation to adjust the oscillation amplitude of the CPGs, so as to avoid the severe deviation of COG in the transverse direction. The relationship betweem the height of COG and the GRF as well as the height of COG and the joint torque is discussed, and the simulation results further verify the effectiveness and feasibility of the COG-height control approach. This thesis discusses the dynamic characteristic of foot-ground contact force resulting from CPGs’ parameters change, and explores the variation of foot-ground contact force when a quadruped robot locomotes on different ground substrates. Quadruped rhythmic control experiments, including walk and trot, are conducted on the quadruped robot Biodog, developed at our lab. In the trot experiment, we discuss the effect of different step lengths and step frequencies on the dynamic characteristic of foot-ground contact force, in order to improve the dynamic performance of quadruped robots. In the walk experiment, we explore the variation of foot-ground contact force when the quadruped robot locomotes on ground substrates with different properties. Experiment results show that it’s feasible to utilize foot-ground contact force for ground substrate recognition and classification, so this provides the theoretical basis for ground substrate classification and adaptive locomotion. For the experiments on the quadruped robot Biodog, an electronic control system based on ARM and DSP is developed. ARM controller is utilized to conduct sensor information processing and a quadruped robot’s motion planning, and the DSP controller is employed to implement the motor drive control. The designed electronic control system can be utilized to develop a self-contained quadruped robot. In order to achieve adaptive quadruped locomotion on terrains with different geological properties, such as surface friction or elasticity modulus, we propose a ground substrate classification approach using Support Vector Machine (SVM). The approach extracts the feature vector from the collected foot-ground contact force and torso pitch-roll information of a quadruped robot during locomotion, and classifies the terrains with different geologies by means of SVM algorithm. In the experiments, the quadruped robot Biodog locomotes in walk gait on five ground substrates with different surface friction and elasticity modulus, and about 99.33% of the five ground substrates can be classified correctly using the proposed approach. An adaptive locomotion approach is designed to improve the performance of the quadruped walk gait generated by CPGs on different ground substrates. We present a COG position adjustment method to eliminate the offset between the control signal generated by CPGs and the actual phase of the quadruped limb. On the other hand, CPGs are incorporated into a foot path planning method to make the lift height controllable. The adaptive locomotion approach guarantees that the limb in theoretical swing phase is able to lift off the ground, as well as enhances the stability and adaptability of the quadruped robot walking on terrains with different geologies.
关键词	四足机器人中枢模式发生器重心高度调节地面基质分类环境适应行走
文献类型	学位论文
条目标识符	http://ir.ia.ac.cn/handle/173211/14709
专题	毕业生_博士学位论文
推荐引用方式 GB/T 7714	李晓琪. 四足机器人环境适应行走的控制策略研究[D]. 北京. 中国科学院研究生院,2017.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
20170601_3--李晓琪学位论文.（6379KB）			限制开放	--