【作者】 陈强；

【导师】 闵华清；

【作者基本信息】 华南理工大学 ， 计算机应用技术， 2011， 硕士

【摘要】 仿人机器人是当前机器人技术研究领域最为活跃的方向之一,其中,仿人机器人直立行走的步态规划与控制是仿人机器人技术的基础和重点。针对仿人机器人的关节具有较大灵活性等特点,如何实现对仿人机器人步态规划和控制的优化和学习,确保步态行走的稳定、迅速和高效,成为仿人机器人步态规划和控制研究中的新课题。本文以仿人机器人的步态行走的多目标优化为研究目标,提出了一种基于改进粒子群多目标优化的方法对仿人机器人的步态进行多目标优化。文中步态规划和控制作为主线,以双足机器人、倒立摆模型和传感信息的反馈控制模型为研究对象进行了如下工作与分析。首先,本文以华南理工大学第一代仿人机器人SCUT-I为研究对象,通过对其运动体系结构的分析,建立了仿人机器人简化模型,并在此基础上,通过研究仿人机器人行走理论,对其进行了运动学和动力学分析。通过动量、角动量以及零点力矩点的定义,推导出SCUT-I的ZMP计算方法及稳定方程,为SCUT-I机器人实现稳定、高效和快速行走提供了运动学和动力学基础。其次,介绍了三维线性倒立摆的特性以及基于三维线性倒立摆模型的仿人机器人的步态模式生成,引进了双足支撑阶段,得出了影响机器人基于三维倒立摆步行单元的控制参数。同时,在仿人机器人身上安装了加速度传感器、角速度传感器和压力传感器,并对这些传感器的进行了卡尔曼滤波处理,给出了机器人身体姿态和ZMP的计算方法,把这些信息作为机器人反馈控制的信息来源,提出了反馈控制的总体框架及具体控制方法实现。最后,提出了改进的粒子群多目标优化方法,研究了改进策略,并对仿人机器人离线步态规划和在线行走反馈控制建立了多目标优化模型,提出目标适度函数及约束条件,并在Matlab和Webots中建立相应的模型,利用Matlab作为机器人的控制器,在Webots进行仿真实现,验证了仿人机器人离线步态规划和在线行走反馈控制有有效性,对多目标优化模型和改进粒子群多目标优化算法的可靠性也进行了验证。更多还原

【Abstract】 Humanoid robot technology research field is the current direction of one of the most active, in which humanoid robots walk gait planning and control of humanoid robot technology is the foundation and focus. For humanoid robot joints have greater flexibility and other characteristics, if the realization of humanoid robot optimal gait planning and control and learning, to ensure stable walking gait, rapid and efficient, a humanoid robot gait planning and control of the new topics.In this paper, a humanoid robot walking gait of multi-objective optimization for target, proposed based on multi-objective particle swarm optimization method for humanoid robot gait multi-objective optimization.Gait planning and control of the text as the main line, the biped robot, inverted pendulum model and sensor information, the feedback control model for the study and analysis of the work carried out as follows.First, this paper first generation of South China University of Technology humanoid robot SCUT-I as the research object, its movement through architecture analysis, the simplified model of the humanoid robot, and on this basis, through the study of humanoid robot Walk theory, its kinematics and dynamics analysis. By momentum, angular momentum and the definition of zero moment point derived SCUT-I of the ZMP and stability equation method for SCUT-I robot to achieve stable, efficient and fast walking provides a basis for kinematics and dynamics.Secondly, the introduction of the three-dimensional characteristics of the linear inverted pendulum and the linear inverted pendulum based on three-dimensional model of a humanoid robot’s gait pattern generation, the introduction of the feet supporting phase, obtained based on three-dimensional inverted pendulum of walking robot unit Control parameters. Meanwhile, in humanoid robots installed accelerometer, angular rate sensors and pressure sensors, these sensors and Kalman filtering were given ZMP of the robot body posture and the method for calculating this information as the robot Sources of information feedback control, feedback control is proposed a general framework and specific control method.Finally, an improved multi-objective particle swarm optimization method to study the improvement strategy, and the humanoid robot off-line walking gait planning and feedback control to establish a multi-objective optimization model, the objective function and constraints appropriate and Created in Matlab and Webots corresponding model, the use of Matlab as a robot controller, in the Webots simulation implementation, verification of the humanoid robot off-line walking gait planning and feedback control with the effectiveness of multi-objective optimization model And improve the multi-objective particle swarm optimization algorithm reliability were verified.更多还原