基于KUKA工业机器人的运动学标定及误差分析研究

发布时间：2018-03-04 13:48

本文选题：KUKA机器人　切入点：运动学标定　出处：《长春工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着人工智能的飞速发展和生产力水平的不断进步,工业机器人作为智能设备的代表被广泛应用到了生产生活的诸多领域,如餐饮服务、制造、医疗、航天等,并在一些高技术领域扮演着不可替代的角色。现如今,人们不仅需要机器人能完成一些简单的工作任务,更需要它们能够完成一些精细且复杂的工作。采用离线示教工作方式的机器人侧重于重复精度指标,难以满足高科技生产中对于绝对精度的要求。为了突破工业机器人绝对精度较低这一限制,因此进行机器人运动学标定显得日益重要。本文首先对刚体位姿和微分运动进行理论剖析,为后续KUKA机器人运动学模型的建立和运动学标定奠定理论基础。本文以工业机器人车身激光检测系统中最常用的KUKA工业机器人实际参数为基础,依据D-H方法建立运动学模型,然后对D-H模型在运动学标定中存在的不足和缺陷进行分析,采用修正后的5参数M-DH模型建立本文用于标定的运动学模型,并讨论运动学正、逆解问题。文章通过对机器人的误差进行综合分析,确定了误差主要来源为几何误差,并根据最终建立的M-DH模型建立机器人几何误差模型,即机器人定位误差与机器人连杆几何误差的关系式。鉴于代数法在求解机器人逆运动学问题上存在计算量大等缺点,本文选用神经网络来求解机器人逆解。该方法以运动学模型为基础计算出机器人正解,然后作为训练样本训练经微分进化算法优化的BP神经网络(DE-BP网络),实现机器人末端操作手位姿与各关节变量的非线性映射关系,从而避免了繁琐复杂的公式推导过程。本文讨论了点约束学标定方法及其原理。为辨识出更为准确的几何参数,在点约束标定方法的基础上提出了一种虚拟空间点约束的标定法,并详细分析和论证了该方法的原理。该方法通过激光束延长末端操作手的长度的方法将末端位置误差放大在观测平板上,以此能够获得更高精度的关节角的值。为了验证本文提出的虚拟空间点约束的标定方法的有效性和稳定性,本文以KUKA机器人为研究对象,利用有关数据进行了标定仿真和试验。
[Abstract]:With the rapid development of artificial intelligence and the continuous progress of productivity level, industrial robots, as representatives of intelligent devices, have been widely used in many fields of production and life, such as catering services, manufacturing, medical treatment, aerospace, and so on. And playing an irreplaceable role in some high-tech fields. Nowadays, people don't just need robots to accomplish some simple tasks. More importantly, they need to be able to do some fine and complex work. Robots that use off-line instruction work focus on repeating accuracy indicators. It is difficult to meet the requirement of absolute precision in high-tech production. In order to break through the limitation of low absolute precision of industrial robot, So it is more and more important to calibrate the robot kinematics. In this paper, the kinematics model of KUKA robot is established by D-H method, which is based on the practical parameters of the most commonly used KUKA industrial robot in the laser detection system of the body of the industrial robot. Then, the shortcomings and defects of D-H model in kinematics calibration are analyzed, and the kinematics model of this paper is established by using the modified 5-parameter M-DH model, and the kinematics positive is discussed. By synthetically analyzing the error of the robot, it is determined that the main source of the error is geometric error, and the geometric error model of the robot is established according to the final M-DH model. That is, the relationship between robot positioning error and robot connecting rod geometric error. In view of the disadvantages of algebraic method in solving robot inverse kinematics problems, In this paper, neural network is used to solve the inverse solution of the robot, and the forward solution of the robot is calculated based on the kinematics model. Then the DE-BP neural network optimized by differential evolution algorithm is trained as a training sample to realize the nonlinear mapping relationship between the position and pose of the robot end operator and each joint variable. In this paper, the calibration method of point constraint and its principle are discussed. In order to identify more accurate geometric parameters, the complicated formula derivation process is avoided. Based on the calibration method of point constraint, a calibration method of virtual space point constraint is proposed. The principle of the method is analyzed and proved in detail. The method amplifies the end position error on the observation plate by the method of extending the length of the end manipulator by laser beam. In order to verify the validity and stability of the calibration method proposed in this paper, the KUKA robot is taken as the research object, and the calibration simulation and experiment are carried out with the relevant data.
【学位授予单位】：长春工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242.2

【参考文献】