Application background：
Under the background of "Industry 4.0" and "Made in China 2025", in order to adapt to the fast changing characteristics of modern industry and meet the increasing complexity requirements, robots should not only be able to complete repetitive work stably for a long time, but also be intelligent, networked, open and human-computer friendly.
As an important aspect of the continued development and innovation of industrial robots, teaching technology is developing towards the direction of facilitating rapid teaching programming and enhancing human-machine collaboration capabilities. The traditional teaching box, which has the most practical applications, requires the operator to have a certain level of robot technology knowledge and experience, and the teaching efficiency is relatively low. Compared with the teaching box teaching method, the drag teaching method does not require the operator to master any robot knowledge and experience, and the operation is simple and fast, greatly improving the friendliness and efficiency of teaching.
Related concepts：
1. Drag teaching
It refers to the movement of the operating arm in the direction of human force under the traction of the operator (traction end or traction of a certain operating arm). This function can easily plan trajectories (for tasks with low process trajectory accuracy), allowing operators to record and reproduce trajectories without the need for manual programming, reducing the threshold for operators and improving efficiency.
2. Sensor based drag teaching
Traditional drag teaching relies on external multi-dimensional torque sensors (including base type, joint type, and end type) of the robot, which use the torque information obtained by the sensors to calculate the desired direction and speed of motion. Although this method can improve control accuracy, it also brings about increased costs and inconvenience in installation and maintenance. The cost of high-precision sensors is even higher than that of the machine itself.
3. Drag teaching based on torque control for zero force balance
For rigid industrial robots, without increasing manufacturing and maintenance costs, with the help of the robot's dynamic model, the controller can calculate the torque required for the robot to be dragged in real-time, and then provide this torque to the motor, enabling the robot to effectively assist the operator in dragging, meeting the requirements of good human-machine interaction.
Control methods：
There are various methods for controlling the motion of a series of robotic arms, among which there are three representative ones: independent joint nested double loop control, independent joint nested double loop plus gravity/friction compensation control, calculated torque control, and drag teaching control. Below is a brief comparison:
1. Independent joint nested dual loop control: refers to the use of two separate closed-loop controls for each joint, with the outer control loop being the joint angle control loop and the inner control loop being the joint angular velocity control loop. This method is the earliest robot control method, only starting from a simple motor control perspective, without considering the changes in motor load with motion, so this method has poor tracking accuracy.
2. Independent joint nested double loop with gravity/friction compensation control: On the basis of independent joint nested double loop control, the feedforward compensation of gravity and friction is directly applied to the torque output end. This algorithm takes into account the main factors of torque, gravity and friction, because these two torques account for a large proportion of all torques of the robot under normal working conditions. At higher speeds, acceleration torque, centrifugal force and Coriolis force torque can also be added (usually this is not done because the Angular acceleration noise obtained by sensor difference is too large). This control method is more commonly used in industrial robots, which is the actual method adopted in this case.
3. Computed torque control: this control mode is based on the premise that the dynamic model is very accurate. After gravity torque, Coriolis force, centrifugal force torque and Friction torque are added to the feedforward, the system can be simplified into a second-order system. Then the second-order system can be placed in a critical damping state by adjusting the coefficient of angle and angular velocity feedback, and the robot control system has good control performance. The difficulty of this control method lies in being able to establish the model accurately enough, which is one of the typical research directions.
4. Drag teaching: teaching is to compensate the heavy torque and Friction torque according to the current position and speed, and then the operating arm moves along the direction of the force exerted by the person.