中国科学院自动化研究所机构知识库

本文针对四足机器人推动物体及主动撞击行为的运动规划及控制策略展开研究，包括四足机器人系统设计、步态控制及变换策略、推动物体行为的控制策略及主动撞击行为的运动规划及控制策略等。本文主要研究工作概要如下：

（1）基于仿生学原理，设计了一种具有低惯量腿足机构的四足机器人。对四足机器人正逆运动学及逆动力学进行了建模及分析，为四足机器人推撞物体行为的控制算法研究奠定了理论基础。

（2）提出了一种基于中枢模式发生器的四足机器人步态变换算法。该算法以相位差为耦合参数构建了基于Hopf振荡器的中枢模式发生器，生成四足机器人基本步态的控制信号，并进一步设计了相位差调节策略，通过调整Hopf振荡器的相位差及占空比，控制四足机器人实现了快速平滑的步态变换。

（3）提出了一种融合虚拟模型与中枢模式发生器的四足机器人推动物体行为的控制算法。该算法利用Hopf振荡器耦合构建的中枢模式发生器生成机器人腿足摆动相的控制信号。四足机器人腿足支撑相的控制信号由质心处构造的虚拟弹簧、阻尼器等虚拟构件生成的驱动力矩与中枢模式发生器的控制信号叠加构成。

（4）分析四足动物主动撞击行为的仿生机理，提出了一种四足机器人主动撞击行为的运动规划和控制策略。将四足机器人主动撞击过程划分为后摆阶段、前摆阶段和恢复阶段三个阶段，并对三个阶段的重心轨迹进行规划。采用虚拟模型控制机器人按照规划好的重心轨迹运动，实现主动撞击行为，同时控制四足机器人的俯仰姿态以维持机身的稳定性。

This thesis focuses on key problems of quadruped robot control, including design, gait control and transition, pushing operation and active impact motion of quadruped robots. The main work of this thesis is summarized as follows.

(1) A quadruped robot with low inertia legs is designed by referring to bionics principles. The forward and inverse kinematics as well as inverse dynamics of the quadruped robot are modeled and analyzed, which provides a theoretical basis for quadruped object pushing control algorithm.

(2) A gait transition algorithm for quadruped robots based on Central Pattern Generator (CPG) is proposed. A Hopf oscillator-based CPG is constructed using phase difference as coupling parameters to generate control signals of quadruped basic gaits. On this basis, a phase difference adjustment strategy is designed. Through adjusting phase difference and duty cycle of Hopf oscillators, fast and smooth gait transition is realized.

(3) This thesis proposes a control strategy aiming at generating and executing pushing operation for quadruped robots, which is based on a CPG network and Virtual Model Control (VMC). A Hopf oscillator-based CPG is used to generate control signals for swinging limbs. Control signals of supporting limbs are composed of driving torques generated by virtual components such as springs and dampers and control signals of CPG.

(4) This thesis presents a novel method of generating and executing active impact motion for quadruped robots by analyzing bionic mechanism of active collision behavior of quadruped animals. This research divides the whole active impact process into three phases: swing backward phase, swing forward phase and recovery phase, and presents a Center of Gravity (COG) trajectory planning method for these phases. Virtual models are used to control the quadruped robot to move according to the planned COG trajectory to execute active impact motion. Moreover, a pitch attitude adjustment controller is utilized to stabilize the quadruped robot.