Trajectory Sampling and Fitting Restoration Based on Machine Vision for Robot Fast Teaching
Feng Zheng, Fang Fang, Xudong Ma · 2020
Industrial robot teaching movement methods can be roughly divided into two types: direct teaching and offline teaching. Because the offline teaching method requires modeling of the robot and its surroundings, the technical requirements of the operator are high and the process is time-consuming and energy-consuming, the direct teaching method is commonly used. However, when the trajectory of the robot movement is complicated, a large number of points need to be taught to ensure the accuracy of the movement. In order to improve the teaching efficiency, a rapid teaching method for industrial robots based on machine vision technology is proposed. This method requires the robot to move on a fixed plane and achieves rapid teaching by sampling and fitting the robot's end trajectory. Firstly, a robot visual teaching system composed of a robot movement control subsystem and a visual image acquisition and processing subsystem is set up, which solved the communication problem between B & R real-time operating system and Windows operating system. It realizes direct observation of movement trajectory by method of "eye-in-hand". Then, using the data points obtained by manual sampling, a cubic B-spline curve fitting algorithm is used to fit and restore a continuous and smooth trajectory curve. Using the camera's internal and external parameters and the hand-eye calibration results, the target set point sequence to which the robot end should move can be inversely calculated. Finally, the experiment discusses the influence of different sampling methods on the trajectory restoration result and obtains the optimal sampling method to improve the accuracy of trajectory restoration. Although this method cannot be applied to scenarios with strict location accuracy requirements, it can be applied to most welding and spraying tasks with general location accuracy requirements. At the same time, the binocular camera or RGB-D depth camera can be used to replace the monocular camera in this paper, so that the method can be extended to three-dimensional stereo application scenarios.