The Research & Development Of Robot Measurement System Based On VXtrack Dynamic Tracking

With the development of industrial technology, robots as humanoid automated equipments, especially the serial versatile industrial robots with six degrees of freedom used more and more widely, have completed many high-risk, high-pressure and high-intensity tasks instead of manual labor. However, due to the presence of the external environment and the geometry errors and other factors, when the robot end-effector is at work, its repeat positioning accuracy is high but absolute positioning accuracy is low. So it is often difficult to meet the needs in the industrial production of high precision. Therefore, it's significant to achieve the robot measuring, calibration, accuracy assessment and error correction to improve its absolute accuracy, which can increase robot's life and broaden its application scope. This paper aims to study the robot kinematic parameters measuring and calibration methods and develop a robot measurement system based on VXtrack dynamic tracking for successfully and efficiently improving positioning and motion accuracy of our self-developed robot.Firstly, this paper studies and summarizes the domestic and foreign theories of industrial robot measuring, calibration and error compensation, establishes the mathematical model of “PingFang” robot, and carries out its kinematic analysis and inference as well as error modeling and analysis, which lay foundation for robot calibration and visual control optimization.Then this paper builds the robot vision system combined the robot platform and the vision equipment, with an object of “PingFang” to the actual measurements, a certain evaluation including the robot's static parameters and dynamic trajectory and repeat positioning accuracy is made. Also, the robot calibration is carried out by joint circumference point and kinematic loop method based on space geometry. The former can measure and construct kinematic parameters of the model by separately moving each rotary joint, then obtain and modify part of the link parameters, which can initially improve the positioning accuracy. And, the latter can use a modified five-parameter model to set up the pose error model of the robot-end, make the solution of all the geometric parameters by combining the spherical numbered targets with the least square iterative method, then amend the measured parameters to the robot control system which further improve the positioning accuracy of the robot. Meanwhile, this paper proposes the joint variable compensation method based on optimal control of robot vision. This propsed method deals with the tracked target trajectory and reproduced trajectory as data samples in the robot vision control experiment, and constructs the optimization function with the introduction of genetic algorithm to correct the trajectory deviation, which can achieve visual control optimization as well as high-precision reproduction. Finally, this paper summarizes the various measuring and calibration methods and conducts the redevelopment using the API application interface provided by VXtrack binocular vision dynamic measurement system, then integrates the parameter measurement and error calibration methods into the software and compiles the pertinent calibration software by the modularization, and finally verifies the feasibility of the measurement software through examples. Experimental results show that the robot vision measurement calibration system built can easily and quickly complete the self-developed robot parameter measurement and compensation as well as optimized motion control, which can achieve fool-operation and is highly practical.This paper studies and proposes the measurement method of serial robot based on spherical numbered targets.The proposed method tracks the standard target ball rigidly connected to the end flange by binocular vision, which can complete pose measurement when the manipulator is moved arbitrarily in the working space. Then, a wide range of highly accurate actual pose data can be obtained for identifying the accurate parameters of robot model.The experiments show the accurate results, the simple, reliable method and solve the problem that binocular cameras' vision field is narrow when traditional vision system is tracking and reduces cost of position and orientation measurements of robot-end, which has some reference value.