Domestic industrial robots are used more and more widely: they can only replace simple manual work, such as assembly and handling, replace some jobs that require workers to have corresponding skills, or replace processing that originally required expensive special equipment, such as cutting, grinding and polishing, deburring, spraying, dispensing, quenching, scanning, welding, 3D printing and so on. Compared with point-to-point robot applications such as assembly and handling, robots used in cutting, polishing and other applications have become more complex, facing problems such as complex trajectories and high precision requirements. In this case, the traditional manual teaching programming method is very inefficient, which can not guarantee the accuracy of the working trajectory, affect the product quality and cause the loss of robot downtime. So it is inevitable to use off-line programming software to plan the robot's working trajectory, and it is becoming more and more popular. However, friends who use off-line programming software for robots for the first time will always find that there is an error between the actual machining trajectory and the robot working trajectory generated in the software, and sometimes this error is even greater. As a practitioner, I try to analyze the causes of such mistakes and propose solutions.

Manual teaching method is to guide the robot from one point to another point through the naked eye observation of the operator, and synthesize the whole robot working trajectory point by point (except straight line and arc trajectory).

The offline programming method is to construct a set of corresponding virtual three-dimensional scene according to the real robot workstation in the software, and configure the production factors such as robot, cutter, workpiece and external shaft in this scene, and generate the robot working trajectory for production through this series of operations.