Abstract: This paper introduces the control system structure and control strategy of an industrial material handling robot, as well as the secondary teaching mechanism between the upper and lower computers implemented based on this control strategy. This control strategy improves the real-time communication between the upper and lower computers, avoids various system failures caused by communication interference, and greatly improves the system reliability. Keywords: robot; secondary teaching; control system; distributed control 1 Introduction With the rapid development of the processing industry in China, the level of automation equipment in various industries is getting higher and higher. Modern processing workshops are often equipped with robots to improve production efficiency and complete tasks that are difficult or dangerous for workers. At present, robots are commonly used in China for the following tasks: in the injection molding industry, they are used to quickly grab products from the mold and transfer them to the next production process; in the machining industry, they are used for material handling and feeding; in the casting industry, they are used to extract high-temperature molten liquid, etc. This paper takes the material handling robot that can perform these tasks as the research object, and studies the control strategy and secondary teaching method of the industrial material handling rectangular coordinate robot control system in which the author participated. At present, the control methods of the main motion arm of the industrial material handling rectangular coordinate robot commonly used in the market are mainly hydraulic or pneumatic drive. The advantage of this control strategy is its simple structure and ease of system control. However, its disadvantage lies in the fact that system positioning relies on setting the position of proximity switches. If the user requests a change in the material handling operation, the positioning switches of each hydraulic or pneumatic cylinder must be readjusted to adapt to the new task, which is detrimental to the automation of the production process. The industrial material handling Cartesian coordinate robot designed by the authors of this paper uses a fully electric control method for the main motion. The system uses an AC motor as the drive source, a dedicated AC motor servo controller as the lower-level machine, and an industrial control computer as the upper-level machine. The overall system adopts a distributed control strategy. This design scheme provides high system precision, a compact system structure, and ease of user operation, and has good market prospects. 2 System Structure and Control Strategy The industrial material handling robot operates according to the concept of a "teach/reproduce robot," where a human teaches the robot's behavior using a teach pendant. During the teaching process, the starting positions of each joint are recorded, and this action is repeated during production at the user-set speed and acceleration. Based on the above analysis, we designed a type of industrial material handling robot with a fully electric main motion. The robotic arm control system adopts a distributed control structure, where one host computer controls five slave servo controllers. Each slave servo controller independently drives an AC asynchronous motor, which in turn drives the robotic arm's actuators to complete the movements. The host computer in this system is an industrial control computer, and the slave servo controllers are dedicated AC asynchronous motor servo controllers. The host computer performs two functions: first, it provides a human-machine interface for user operation; second, it controls the slave servo controllers via a serial port according to the RS-485 protocol standard. The slave controllers receive commands from the host computer and control the motor rotation accordingly. Simultaneously, the slave servo controllers also detect the asynchronous motor's motion status via an encoder on the AC motor and feed it back to the slave computer, enabling real-time control of the motor's motion. The control system structure is shown in Figure 1. [align=center] Figure 1: Robotic Arm Control System Structure Diagram[/align] The AC asynchronous motor servo controller is a device specifically designed for controlling AC asynchronous motors. The servo controller is equipped with an RS-485 interface and I/O ports for communication with the host computer. Using a dedicated language for servo controllers, we designed a program that runs on the servo controller. The lower-level servo controller automatically executes this program stored in the controller after power-on. Its main function is to receive instructions from the upper-level computer and control the motor's motion state according to the upper-level computer's requirements. The specific real-time control of the motor motion is implemented by the lower-level servo controller's internal control algorithm. 3. Secondary Teaching The robotic arm adopts a distributed control structure, so the control network for information transmission between the upper and lower-level computers becomes a key factor in system control. Both the AC motor servo controller and the industrial control computer have an RS-485 port, so we chose the RS-485 serial port of the upper and lower-level computers as the communication method between them. In industrial settings, the upper and lower-level computers do not frequently communicate via RS-485 because RS-485 communication is easily interfered with in environments with interference, leading to communication failures. Although we designed fault-tolerant functions in the communication program, such failures, along with the time consumed by RS-485 communication itself, will significantly reduce the real-time performance of the system. To address this issue, we adopted a two-stage teaching control strategy. The implementation method of this strategy is briefly described below. During the teaching phase, the host computer records the workflow required by the user, i.e., the sequence of motion for each degree of freedom, as well as the motion distance, speed, and acceleration of each motion stage, forming a teaching data queue. This teaching data queue records the complete information required for the robot's operation, thus completing one stage of teaching the system by a human. Figure 2 illustrates a data queue for a teaching process, where A, B, C, and D represent lower-level devices. The sequence number after the device name in each box node indicates the sequence number in which the node starts the device within a workflow. For example, in "B2", "B" indicates that this node records information about the lower-level device "B", and "2" indicates that this node contains the motion information of device "B" when it is started for the second time in a workflow. Note that the data recording content of the same device in different sequences is often different. [align=center]Figure 2[/align] During the initial automatic operation of the robotic arm, in the first automatic work cycle, the host computer sends relevant motion data information, namely displacement, velocity, and acceleration, to the corresponding controllers of the degrees of freedom that should be moving. The lower-level controllers then save this data to their own memory to form data records. Next, the host computer initiates the I/O of the relevant lower-level controllers to notify them to begin their actions. The actions performed by the lower-level controllers are based on this newly saved information. After completing the action, the lower-level controllers notify the host computer, which then notifies the next controller requiring action. After completing a workflow, each controller has sequentially formed its own data records according to its actions within the workflow and stored them in its own memory. These records document the velocity, distance, and acceleration of each action, thus completing the secondary teaching from the host computer to the lower-level controllers. At this point, the data queue becomes as shown in Figure 3, which represents the initial working state of a workflow. [align=center]Figure 3[/align] In this way, each controller knows the information of each action in a workflow when the host computer notifies it to work, and also knows the order of its own actions after being notified. However, each controller does not record the order of its own actions in the overall workflow; this order is coordinated by the main controller. After secondary teaching, all the information of a workflow is decomposed, that is, the host computer records the overall work sequence, and the lower controller records its own internal work sequence and work details. In subsequent workflows, the host controller only sends control I/O to the controllers that should take action, and no longer sends specific control details, because these details have already been formed into data records in the lower controllers in the first workflow. The lower controllers retrieve these data records in sequence according to the host computer's I/O to complete the actions. Figure 4 illustrates the change in the above figure when the host data queue works to B3. [align=center]Figure 4[/align] Through secondary teaching, the real-time performance of the system is enhanced, many faults caused by communication interference are avoided, and the overall performance of the system is greatly improved.