【作者】 杨永刚；

【导师】 赵杰；

【作者基本信息】 哈尔滨工业大学 ， 机械电子工程， 2008， 博士

【摘要】 并联机器人以其承载能力强、运动精度高、刚度大、运动惯量小等一系列的优点,在国民生产的各个领域得到越来越广泛应用。6-PRRS机构作为一种特殊构型的新型并联机构,其特点是连接杆长固定,依靠6个直线运动输入驱动,具有单方向较大工作空间的特性,改善了传统并联机构的工作空间小这一不足。本文在对国内外并联机器人理论、应用与发展分析的基础上,对其运动学、动力学、误差分析和标定方法以及相应的控制方法进行了深入的探讨和研究。并联机器人与串联机器人的一些特性存在对偶关系,其运动学逆解容易,正解困难。本文通过对现有的一些典型运动学正解方法进行分析,提出了一种神经网络结合拟牛顿迭代法的位置正解精确求解的方法。采用神经网络进行非线性映射,解决了拟牛顿迭代法的初值问题,两者的有效结合避免了以往并联机器人运动学求解需进行复杂运算的过程,并且能得到唯一的位置解。扩大了并联机器人运动学的应用范围。并联机器人的动力学模型是一个多变量、高度非线性,多参数耦合的复杂系统,本文采用虚功原理方法建立了并联机器人驱动电机到动平台的各构件随参数变化的完整动力学方程,其中包含了各支链质量与惯性力,考虑了所有结构因素对关节驱动力的影响,并根据实际的工作情况,对动力学模型进行了简化,给出了动力学方程线性形式,为并联机器人的动力学实时控制奠定了理论基础。机器人所能达到的精度受系统的机构误差影响很大,机构误差主要包括零部件的制造与装配误差、铰链和丝杠的间隙误差。对于机构误差对系统精度的影响,本文使用矢量链法对其进行了分析。通过基于连杆参数的D-H法对6-PRRS并联机器人进行了运动学建模,给出了参数识别方法,并提出了误差补偿的方案,通过实验完成了系统的标定工作。鉴于并联机器人的动力学计算复杂,不利于实时控制等特性,本文首先在不考虑并联机器人动力学的情况下,研究了并联机器人的控制器设计。基于分散控制策略,设计了可不依赖于并联机器人动力学模型的自抗扰控制器,将并联机器人的动力学耦合视为干扰,采用具有模型补偿功能的扩张状态观测器进行补偿,实现对6-PRRS并联机器人的高精度控制。仿真结果表明所设计的不依赖于动力学的自抗扰控制器可以实现系统的高速高精度控制,满足系统的性能要求,证明了所设计控制方法的有效性与可行性。不考虑动力学的控制方法虽然控制律简单,易于实现。但对于控制高速高精度机器人来说,这类方法有两个明显的缺点:一是难于保证受控机器人具有良好的动态和静态品质;二是需要较大的控制能量。因此本文根据机器人动力学模型的性质设计了一种利用机器人工作空间理想轨迹控制的自适应前馈控制方法,并对自适应控制下系统的稳定性进行了分析。其基本思想利用动力学模型的线性形式,在线实时的辨识对象特性的变化,使动力学耦合及外部干扰对系统的影响逐渐降到最低。实验结果验证了该方法有效性。最后,建立了6-PRRS并联机器人仿真与物理实验系统,对提出的自抗扰控制和自适应控制进行了实际验证。在实验中,6-PRRS并联机器人能够高精度的跟踪理想的运动轨迹、具有较强的抗干扰能力,表明对6-PRRS并联机器人控制方法的研究成果是正确的,可行的。更多还原

【Abstract】 Parallel robots are applied more and more widely in every field of national product because its merit of good carrying capacity, high movement precision, great rigidity and lower movement inertia and so on. 6-PRRS mechanism, as a new parallel mechanism with a special structure, has non-variable length of link and is drived by six linear straight motions, with a single-direction big workspace. It enlarges the workspace of parallel mechanism. The kinematics, dynamics, error analysis, calibration and control method based on the domestic and external theories of parallel robot are discussed and studied deeply.Parallel robot is opposite to series robot in some characteristic. Its inverse kinematics is easier to solve and forward kinematics is harder to solve.After some existent typical methods of solving forward kinematics being analyzed, a new method is brought forward, that adopts neural network combined with quasi-Newton iteration method. Efficiently combination of neural network and quasi-Newton iteration can avoid the complicated computing process of parallel robot forward kinematics solution and the solution is one and only. So enlarges the application range of parallel robot forward kinematics.Parallel robot’s dynamic equation is a system with multi variables, serious nonlinearity and multi parameter coupling. So principle of virtual work is used to set up the whole dynamics equation, including each branched chain’s inertia and quality, with the parameter varying form actuation motor to each component of the parallel robot. The affection of configuration on joint driving force is considered in this equation. After this, dynamics equation is simplified in terms of actual work condition and it is linearized. This work settles the basis of real-time control.The precision robot can reach is not only relative of controller but also of mechanism error. The errors mainly include components’machining and assembly error and clearance error between hinge and leading screw. In this paper, vector chain method is applied to analysis of the affection to mechanism error on system precision, D-H method based on links’parameters is used to model 6-PRRS parallel robot’s kinematics, the method of parameter identification is given and error compensation scheme is proposed, then system calibration is finished by experiments. Because the parallel robot is hard to real-time control in terms of the complication of its dynamics, controller without considering parallel robot’s dynamics is studied firstly. Based on discrete scheme, auto disturbance-rejection controller(ADRC) is designed, the coupling of robot dynamics is regarded as disturbance and extended state observer with model compensation function is used to compensate the disturbance. So ADRC on 6-PRRS parallel robot is realized. Simulation results show the ADRC can realize the robot high speed and high precision controlling, satisfy the demand of system performance and the validity and feasibility of this algorithm are proved.In spite of the control algorithm not considering dynamics is simple and easy realized, it has two obvious disadvantages for high speed and high precision control of robot. First, it is difficult to ensure the dynamic and static better quality of controlled robot. Second, the system needs larger control energy. So an adaptive feed forward control method is proposed, this method basing on robot dynamics adopts ideal trajectory on workspace to control robot, its stability is analyzed. The basis is using linear dynamics form, real-time identifying object property change online, then reduceing the affection of dynamics coupled and external disturbance on system to least. The validity of this algorithm is proved by experiments.Finally, simulation and hardware systems are built. The method of ADRC and NNPID is verified in practice. In experiments, 6-PRRS parallel robot can finish high-precision track given trajectory, the system has a strong disturbance- rejection capability. It is indicated from experiments that results about research of 6-PRRS parallel robot are correct and feasible.更多还原