Abstract : To deeply study the impact of delay on control performance and delay compensation methods in closed-loop network control systems, a Controller Area Network (CAN) bus was applied to a servo control system to form a distributed control system with a bus-type network topology. The overall composition and implementation of the motion control system experimental platform were investigated. A controller was designed to address the practical problems existing in this network control system, effectively improving system performance. Experimental results verified its effectiveness. Keywords : CAN bus; Smith predictor; servo control system; delay compensation. Fieldbus is an application and development of computer networks in the field of modern control technology. It is a bus-type topology network applied at the lowest level of production and can serve as a communication network for field control systems, directly connecting serially with all controlled (device) nodes. Traditional control systems struggle to achieve information exchange between devices and between the system and the outside world, forming an information silo. Fieldbus control systems are both open communication networks and fully distributed control systems. CAN (Controller Area Network) is also known as a controller local area network. The authors applied the CAN bus to a servo control system to form a distributed control system with a bus-type network topology, constructed a motion control system experimental platform based on the CAN bus, and applied the CAN bus to the servo control system. 1. Construction of the Motion Control System Experimental Platform 1.1 System Overall Composition The CAN bus system is designed using a host computer and node-based approach, with communication between the host computer and nodes via a CAN bus. Shielded twisted-pair cable is used as the transmission medium. The host computer consists of an industrial computer and a CAN adapter card plugged into the ISA bus, monitoring the status of each node and sending control commands. Each node consists of a microcontroller, a CAN controller, and other peripheral circuits, receiving and executing commands sent by the host computer, and simultaneously returning status information. The bus network topology of the motion control system is shown in Figure 1. The main function of the system is to collect real-time information on motor motion and send control commands based on the obtained information to control the motor motion. The host computer (PC or industrial computer) communicates with each node through a CAN interface adapter card, plans gait, starts and stops the motor, applies closed-loop control algorithms to calculate and send control signals, and completes overall decision-making and control. The intelligent control node receives the control signals sent by the host computer, converts them into analog signals to drive the speed unit, and realizes the control of the speed unit. The intelligent sensor node periodically encodes the signals transmitted by the coarse and fine synchro using timed interrupts to obtain the load's shaft angle signal and sends it to the host computer through the CAN bus for control decision-making. This forms a distributed motion control system based on the CAN bus. In the intelligent node, the microcontroller is responsible for the initialization of the CAN controller and realizes communication tasks such as data reception and transmission by controlling the CAN controller. The CAN bus driver provides the interface between the CAN controller and the physical bus, providing the function of sending and receiving data on the bus. The D/A converter and the synchro-to-digital converter and their peripheral circuits constitute the analog output channel and analog input channel of the control system, respectively. 1.2 System Implementation The controlled object of the system is a PWM power amplifier speed unit produced by FANUC, as shown in Figure 2. The design applies CAN bus technology to the control system, constructing a CAN bus-based motion control system experimental platform by developing CAN adapter cards, intelligent control nodes, and intelligent sensor nodes. A CAN communication adapter card is inserted into a PC bus. The intelligent controller node and intelligent sensor node consist of a microcontroller, CAN communication circuit, signal detection, A/D, D/A converters, and their interface circuits. All parts are connected via the CAN bus to form the experimental system. This allows for modular separation and design of system functions and facilitates network system construction. An industrial PC is selected as the main controller to handle gait planning, motor start/stop, and closed-loop control algorithms. An 80C196 microprocessor is used as the intelligent node's microprocessor. The intelligent control node receives control signals from the host computer via the CAN bus, performs D/A conversion using a DAC1210, and drives the motor. The intelligent sensor node performs A/D conversion using an electronic Scott transformer and RDC to obtain shaft angle signals, which are then transmitted to the host computer via the CAN bus. Thus, the CAN bus-based motor position closed-loop motion control system is completed. 2. Problem Description and Control Algorithm 2.1 Problem Description Compared with traditional control systems, distributed control systems have many advantages such as simplicity, speed, fewer connections, and ease of installation and maintenance. However, in distributed control systems with a bus-type network topology, data transmission delay is unavoidable, which to some extent affects the performance of the control system and may even cause system instability. Traditional control systems are point-to-point control systems. Data collected by sensors is directly fed back to the controller, which outputs the calculated control quantity directly to the D/A converter. The resulting voltage control signal immediately acts on the controlled object to complete closed-loop control. The delay in the system mainly comes from the calculation time of the control algorithm and the delay time of the hardware circuit. After introducing the CAN bus into the closed-loop control system, the change in system structure brings about a huge difference in control behavior. Data collected by sensor nodes is transmitted to the controller node through the bus. The data transmission delay in the feedback loop prevents the controller from obtaining the status information of the controlled object in real time. Similarly, the control signals generated by the controller node must be transmitted to the actuator node through the bus. The existence of transmission delay also prevents the control signals from acting on the controlled object in a timely manner. In this case, data transmission delay becomes the main factor affecting system performance and destabilizing the system. Therefore, a control algorithm that can effectively compensate for transmission delay must be adopted. 2.2 Controller Design The Smith predictor method is another effective control method to overcome pure time delay. Its basic principle is to predict the dynamic characteristics of the object and use a predictor model to compensate for the time delay. The compensator and the controlled object together constitute a generalized controlled object without time delay. In this way, the controller is equivalent to controlling a system without time delay, thus effectively overcoming the influence of pure time delay. In this system, the baud rate and the number of bytes of data transmitted on the CAN bus are fixed values. Without considering the algorithm calculation time and hardware circuit delay time, the delay in the system can be approximately simulated by pure time delay. Due to the compensation through the Smith predictor, it can be approximately assumed that the generalized controlled object no longer contains a time delay component. The controller employs a three-stage control algorithm for large error, medium error, and small deviation. During large error, the digital controller outputs a saturation value (i.e., D/A saturation output), causing the servo system's speed loop to start with maximum acceleration until reaching maximum speed, and then maintain this constant speed. After reaching medium error, the controller uses the maximum deceleration law ω = (2eε)¹/² to guide the servo system to brake at maximum deceleration, smoothly reaching the coordination point. Where ε is the deceleration acceleration; e is the error. During small deviation, the control algorithm uses a PID position algorithm with feedforward and integral separation to achieve closed-loop position control of the controlled object. Since the Smith predictive controller is based on a precise mathematical model of the controlled object, accurate identification of the controlled object's mathematical model is necessary. 2.3 Identification of the Controlled Object Model The FANUC DC PWM driver model is relatively complex; it is combined with the motor to form a second-order system (inner loop), as shown in Figure 3. In the figure: the reduction ratio of the reducer i = 69.47; k, k1, k2 are the parameters to be identified. The continuous state equation of the controlled object is given above, discretized with a sampling period Ts = 5 ms, and then identified using the least squares method. Finally, k = 69354, k1 = 35.10, k2 = 4254.10, and the transfer function of the controlled object is 998.33 / (s3 + 35.1s2 + 4254.1s). 3 Experiment and Results In the experiment, the intelligent sensor node samples with a period of 5 ms, the intelligent controller and actuator node adopt the event-driven mode, the CAN bus baud rate is set to 1 Mbit/s, and both the feedback data and control data are 6 B. The system is controlled by a three-segment control algorithm with Smith predictor compensation and a three-segment control algorithm without Smith predictor compensation. The system response curve and error curve are shown in Figures 4 to 6 when the input is a step, constant velocity, and sinusoidal signal. In Figures 4-6, the vertical axis unit is "code" (1 code = 360° / 65536). In Figure 4, the step input setpoint is 16.5°. When the three-segment control algorithm is directly applied, the system exhibits overshoot and significant oscillation during dynamic processes. After adopting the Smith pre-estimation method, the system stability is enhanced, there is no overshoot during dynamic processes, the dynamic performance of the step response is significantly improved, and the absolute value of the steady-state error is less than 1 mrad, meeting the accuracy requirements. Figure 5 shows the constant velocity tracking situation. The angular velocity setpoint is 30°/s. Due to the influence of data transmission delay in the system, the system tracking error is relatively large, and the tracking accuracy is reduced. Using the Smith pre-estimation method can significantly reduce the system tracking error, improve tracking accuracy, and ensure the accuracy and stability of the tracking process. Figure 6 shows a sinusoidal signal with a maximum angular velocity of 30°/s and a maximum angular acceleration of 30°/s². Without the Smith predictor method, the system error value and error variation are large due to the bus transmission delay. The Smith predictor method effectively compensates for the delay to a certain extent, thus significantly reducing the system error, reducing the vibration during motor movement, and greatly improving system performance. 4. Conclusion A motion control system experimental platform based on the CAN bus was constructed. Based on this, the impact of data transmission delay on the bus on system control performance was investigated, and control algorithms were researched and experimentally verified. Experimental results show that the controller composed of the Smith predictor and the three-segment control algorithm can effectively compensate for system delay, reduce system error, and ensure the stability and control performance of the system.