带有机械臂的球形机器人若干问题的研究

【作者】 庄未；

【导师】 刘晓平；

【作者基本信息】 北京邮电大学 ， 机械电子工程， 2009， 博士

【摘要】 带有机械臂的球形机器人是在现有球形机器人的基础上,融合多臂机器人特点的一种移动机器人。这种机器人一方面能够进行全方位行走运动,同时又具有对外界环境进行灵活操作的能力,具有广泛的应用前景。本论文以一种新型的带有机械臂球形机器人为主要研究对象,并借助1个移动关节5个旋转关节(1P5R)和3个旋转关节(3R)的柔性关节机械臂模型对该球形机器人的若干问题进行较为深入的研究。论文建立了一种带有机械臂球形机器人的动力学模型,提出了欠驱动球形机器人惯性参数的辨识方法,设计了球形机器人球壳静止时的柔性关节机械臂的动力学控制策略,并提出了一种运动状态下的关节面物理参数辨识方法。主要研究工作如下：1.分析了一种具有13个刚体、15个自由度的带有机械臂球形机器人系统的运动特性。以球壳为基座,球壳内部的驱动装置、两侧的球冠以及球冠内部的臂杆为连杆,分析了系统各构件的运动规律；充分考虑非完整约束条件,得到了各构件的偏速度、偏角速度、广义主动力和广义惯性力的表达式。利用Kane方程建立了系统质量分布均匀时的动力学模型。对球壳沿直线及S曲线行走时各关节力／力矩以及球壳欧拉角分别进行了仿真分析。2.针对欠驱动球形机器人动力学模型的特点,提出了基于惯性测量单元(IMU)传感器的惯性参数辨识方法。根据Kane方法建立了系统质量分布不均匀时的动力学模型,利用模型中的三个欠驱动方程推导出了球形机器人惯性参数辨识模型；以一种新型的带有机械臂球形机器人为实验平台,根据安装在球形机器人内框上的惯性测量单元实测信息,对所提辨识方法进行了实验验证。实验结果表明该方法简单可行,同时为其它欠驱动机器人惯性参数辨识提供了一种新的思路。3.将球形机器人球壳静止时的臂杆操作系统假想为带有移动关节和旋转关节的机械臂系统,建立了6自由度柔性关节机械臂的刚柔耦合动力学模型,并在该模型的基础上,详细讨论了1P5R(1个移动关节、5个旋转关节)柔性关节机械臂的反馈线性化控制方法。在系统工作空间内设计轨迹生成器,借助MATLAB软件工具,采用数值方法分别对同一路径下两种不同运动状态的轨迹进行了跟踪仿真控制。该研究对于合理规划机械臂末端操作轨迹有很好的指导意义。4.针对带有机械臂球形机器人的多连杆柔性关节机械臂,设计了基于高斯径向基函数(RBF)神经网络的滑模控制器,该控制器利用神经网络的逼近能力,将各关节的切换函数作为网络的输入,系统的控制量作为网络的输出,控制器完全由连续的RBF神经网络实现。利用该控制器与线性二次型跟踪器以及传统滑模控制器对三连杆柔性关节机械臂进行了轨迹跟踪控制仿真,并对结果进行了对比分析。5.针对柔性关节机械臂关节面参数的时变特性,提出了一种多体机械系统运动状态下关节面参数辨识的新方法。将机械臂柔性关节等效为弹性扭转轴,并将应用于结构中的行波分析方法与机器人关节旋转变换矩阵相结合,建立了运动机构系统的波导方程。根据系统各结点力平衡与位移边界条件,建立了运动状态下系统激励预测模型,根据预测出的系统激励以及实验测量得到的关节扭转角等因素,建立了系统振动方程。分别采用最小二乘法和神经网络方法对上述方程进行求解,推导出了关节动态刚度和阻尼的辨识模型。以两种不同的运动轨迹驱动3自由度机械臂,对其进行了辨识实验研究。更多还原

【Abstract】 The spherical robot with arms is different from the present spherical robot. It is a novel structure of robot, which combines the advantages of general spherical robot with traditional manipulator robot. This type of robot has excellent performance, including high maneuverability to the environment and omni-direction movement, so it will have far-ranging application perspective in various fields. In this dissertation, all research work is mainly on a newtype of spherical robot with arms, but for the purpose of convenient, the platforms of 1 prismatic joint and 5 revolute joints (1P5R) flexible manipulator and 3 revolute joints (3R) flexible manipulator are under consideration as well. Some critical technique about this spherical robot is discussed deeply and detailly in this dissertation. The innovational work includes deriving the dynamic model of the spherical robot, presenting an approach to identify inertial parameters of the underactuated spherical robot, designing an intelligent controller for flexible joint manipulator, and proposing an approach to identify joint parameters of the flexible joint robot under movement. The detail specification of this work will be shown as follow:At the beginning of the dissertation, a novel structure of spherical robot with arms was introduced. This robot has 13 rigid bodies and 15 degrees of freedom (DOF). Kinemetic characteristic of the robot was detailly analyzed. Regarding the sphere shell of the robot as the base and the driving installations inside the shell plus the sphere caps and the arms inside the caps as the links, law of motion of each link was analyzed. Considering the condition of nonholonomic constraints fully, partial velocity, partial angle velocity, generalized active force and generalized inertia force of each link were derived respectively. Based on Kane’s equation, dynamic model of the spherical robot was derived ultimately. Joints forces/torques and Euler angles of the robot were analyzed respectively after rectilinear motion and sigmoid curve motion simulation.Secondly, according to the characteristic of the dynamic model of spherical robot, an approach based on inertial measurement unit (IMU) to identify inertial parameters of the underactuated spherical robot was presented. Considering the condition of asymmetrical mass distribution in each link of the spherical robot, dynamical model was constructed with Kane’s method, and accordingly, the inertial parameters identification model was derived. Aming at a newtype of spherical robot, the proposal identification method was examplified by an identification experiment according to the information from IMU fixed on the frame of the robot. The experimental result shows that the identification datum are reasonable, and the proposal approach is simple and feasibly, which is a good guidline for inertial parameters identification of other underactuated robots.Thirdly, based on the assumption that the arms system of the spherical robot can be taken as the manipulator system of prismatic joints and revolute joints when the sphere shell was immobile, the dynamic model of rigid-flexible coupling system of a 6DOF flexible joint manipulator was established.According to the dynamic model, a feedback linearization controller of 1P5R manipulator was discussed detailedly. Furthermore, the trajectory was designed in workspace of the system, and tracking controls under the condition of two different motions at the same route were simulated numerically with MATLAB. This research is a good guidance for rationally planning the trajectory of manipulator endpoint.Fourthly, aiming at the multiple-links flexible joint arms of the spherical robot, a sliding mode controller for manipulator was proposed based on Gaussian Radial Basis Function Neural Networks (GRBFNN). Regarding the switching function of each joint as the input and the controlling variables as ouput of RBFNN, the proposed controller was realized by continuous RBF neural network using the approximation ability of the network. The simulations of tracking control of three-links flexible joint manipulator was done with the proposed controller, and the linear quadratic regulator (LQR) tracker and traditional sliding mode controller (SMC) comparely.Finally, focusing on the time-variant joint surface parameters of the flexible joint manipulator, a new approach to identify joint parameters of the flexible joint manipulator under the condition of moving states was presented. Regarding a flexible joint as an elastic twist axis, and combining the wave theory used in structures analysis with rotation transform matrix of joints, the wave equations of mechanism were established. Considering the condition of generalized forces balance and displacement boundary of each joint, the stimulator prediction model of the system under the condition of moving states was derived. Vibration equations were derived according to the predicted stimulator and the measuring twist angles of joints. Joint dynamic stiffness and damping were identified via solving the former identification model by use of least square method (LSM) or neural network method. Experiments on a 3-DOF manipulator were performed when driving the manipulator under two different trajectories.更多还原