Introduction Permanent magnet AC servo technology is a key technology for developing various advanced mechatronic equipment, such as industrial robots, CNC machine tools, and machining centers. Currently, the motor servo systems used in high-performance CNC machine tools and industrial robots still mainly rely on imports, which restricts the development of China's high-tech industries. Therefore, by learning from the advanced experience of foreign research, starting from a high level, and developing a high-performance, practical AC servo system with current international standards as soon as possible is of great significance for promoting the development of China's aviation, aerospace, defense, and industrial automation fields, and for catching up with and surpassing the world's advanced levels. With the remarkable development of power electronics, microelectronics, sensing technology, permanent magnet technology, and control theory, especially the successful application of advanced control strategies, the research and application of AC servo systems have achieved remarkable development in just over two decades since the late 1980s. They now possess excellent technical performance, including a wide speed range, high speed stability, fast dynamic response, and four-quadrant operation. Their dynamic and static characteristics are now fully comparable to DC servo systems, and the long-held aspiration of "AC servo replacing DC servo" is gradually becoming a reality. It is foreseeable that research on AC servo systems will continue to be a research hotspot in the field of electrical drives, driving the rapid development of related industries. Therefore, it is necessary to have a comprehensive understanding of the development of AC servo systems and their advanced control strategies. This paper aims to provide a comprehensive overview and comparison of AC servo systems and their control strategies, reflecting their latest research progress in recent years. Stages of Servo System Development The development of servo systems is closely linked to the different stages of servo motor development. Servo motors have a history of over fifty years, experiencing three main development stages: The first stage (before the 1960s) was centered on hydraulic servo motors driven by stepper motors or direct drive by power stepper motors, where the position control of the servo system was an open-loop system. The second stage (1960s-1970s) was the era of the birth and golden age of DC servo motors. Due to the excellent speed regulation performance of DC motors, many high-performance drive devices adopted DC motors, and the position control of servo systems evolved from an open-loop system to a closed-loop system. The third development stage (1980s to present) is set against the backdrop of the mechatronics era. Breakthroughs in servo motor structure, permanent magnet materials, and control technology led to the emergence of various new types of motors, such as brushless DC servo motors (square wave driven) and AC servo motors (sine wave driven). Entering the 1980s, the rapid development of microelectronics technology and the increasing integration of circuits significantly impacted servo systems. The control methods of AC servo systems rapidly evolved towards microcomputer control, shifting from hardware servo to software servo. Intelligent software servo became a major trend in servo control. The implementation of servo system controllers in digital control also evolved from hardware to software; within software, the development progressed from the outer loop of the servo system to the inner loop, and further to a deeper level approaching the motor loop. Servo System Control Strategies In applications with low dynamic performance requirements, open-loop variable frequency control (VFD) is still widely used in general speed control systems due to its simple control logic. Constant Voltage-Frequency Ratio Control Constant voltage-frequency ratio control is an open-loop control method. Based on the given system parameters, it uses space vector pulse width modulation to convert the output Uout into the desired output for control, allowing the motor to operate at a certain speed. However, it relies on the motor's steady-state model, thus failing to achieve ideal dynamic control performance. To obtain high dynamic performance, a dynamic mathematical model of the motor is necessary. The dynamic mathematical model of a permanent magnet synchronous motor is nonlinear and multivariable, containing a product term of angular velocity ω and current id or iq. Therefore, to obtain accurate control performance, angular velocity and current must be decoupled. In recent years, various nonlinear controllers have been studied to address the nonlinear characteristics of permanent magnet synchronous motors. Vector Control: In 1971, F. Blaschke of Siemens in Germany proposed vector control theory, which significantly advanced the development of AC drives, marking the first qualitative leap in AC motor control theory. Its basic principle is: taking the rotational space vector of the rotor flux linkage as the reference coordinate, the stator current is decomposed into two mutually orthogonal components. One is in the same direction as the flux linkage, representing the stator current excitation component, and the other is orthogonal to the direction of the flux linkage, representing the stator current torque component. Then, they are controlled independently to obtain good dynamic characteristics like a DC motor. The torque equation of the permanent magnet synchronous motor dq model is: Te=P[λfiq+（Ld-Lq）idiq] (1) Vector control is actually controlling the phase and amplitude of the stator voltage or current vector of the motor at the same time. It can be seen from equation (1) that when the excitation flux linkage of the permanent magnet and the direct-quadrature axis inductance are determined, the torque of the motor depends on the stator current space vector is=id+jiq. That is, controlling id and iq can control the torque, thereby controlling the speed. However, the vector control method requires complex coordinate transformation when implemented, and it is highly dependent on the parameters of the motor, making it difficult to guarantee complete decoupling, which reduces the control effect. Direct torque control In the 1980s, Professor Depenbrock proposed the direct torque control method for asynchronous motors. This method abandons the decoupling concept of vector control, implements stator field orientation, avoids the complex coordinate transformations in vector control, and the estimation of stator flux linkage only involves stator resistance. The deceleration reduces the dependence on motor parameters. This control method is simple, has a fast torque response, and good dynamic performance. Currently, some scholars have devoted themselves to applying this control method to permanent magnet synchronous motors, and significant progress has been made. The rotational coordinates on the stator are x, y; the rotational coordinates on the rotor are d, q; they are all 90 degrees apart. The angle between the x-axis and the d-axis is the load angle δ. It can be proven that under the condition that the stator flux linkage amplitude |λs| remains constant, the electromagnetic torque is proportional to the rotor flux linkage λr and the sine of the angle δ between it and λs. In steady state, δ is constant, corresponding to the corresponding torque, and the stator and rotor flux linkages rotate at synchronous speeds; in the dynamic process, δ changes with the load, and the instantaneous speeds of the stator and rotor flux linkages will differ to match the change in δ. Therefore, the stator flux amplitude can be kept constant by selecting a suitable voltage space vector U, and the speed and direction of the stator flux can be adjusted while the load angle δ is adjusted to control the electromagnetic torque. Feedback linearization control The above three control strategies have been maturely applied. However, they only constitute an approximate decoupling control of torque and flux from a physical relationship perspective, without or with little application of control theory. Permanent magnet synchronous motor is essentially a nonlinear, multivariable system. Feedback linear control is an effective method for studying nonlinear control systems. It achieves dynamic decoupling and global linearization of the system through nonlinear state feedback and nonlinear transformation, thereby using linear control theory to design the system to achieve the expected performance indicators. Feedback linearization control is generally divided into two categories: (1) Differential geometric feedback linearization method, which gives the decoupling structure through differential homeomorphic coordinates and a nonlinear state feedback. It requires transforming the problem to the "geometric domain", so the method is abstract and not conducive to engineering applications. However, it considers the problem from a higher mathematical perspective and is relatively easy to develop in theory. (2) Dynamic inverse control, which uses nonlinear inverse system theory to design control laws. It is also called direct feedback linearization method. The physical concept of this method is clear and the mathematics used is simple. Inverse control In 1992, KOKOTOVIC P proposed inverse control. It is an effective nonlinear control. It is designed in the following steps: (1) Select a state of the system to form a subsystem, construct the Lyapunov function of the subsystem, and design the assumed control function to make the subsystem stable; (2) Based on the assumed control in (1), design the error variable, and form a new subsystem with the error variable and the previous subsystem. Construct the Lyapunov function of the new subsystem, and design the assumed control to make the new subsystem stable; (3) If the actual control of the system has not yet been obtained, return to (1) to continue the design. If the actual control of the system is obtained, design downward; (4) Design the actual control of the system to ensure the stability of the entire system. Inverse control can achieve complete decoupling of the permanent magnet synchronous motor system. Moreover, the design method is relatively simple and can ensure the stability of the system. Sliding Mode Variable Structure Control Sliding mode variable structure control is a control strategy within variable structure control. Its fundamental difference from conventional control lies in the discontinuity of control, i.e., a switching characteristic that allows the system's "structure" to change constantly. It also falls under the category of Bang-Bang control. Its main characteristic is that, based on the deviation of the controlled variable and its derivative, the system is intentionally guided to move along a pre-designed "sliding mode" trajectory, independent of the parameters and disturbances of the controlled object, thus giving the system strong robustness. However, it inevitably introduces "juddering" problems into the system. Adaptive Control In the control strategies mentioned above, changes in motor model parameters more or less degrade the system's control performance. Therefore, adaptive control was proposed. It continuously extracts information about the model during system operation, allowing the control strategy to adjust according to the new information. It is a powerful means of overcoming the influence of parameter changes. Currently applied adaptive methods in control include model reference adaptation, parameter identification self-correction control, and various newly developed nonlinear adaptive controls. However, all these methods have problems: first, the mathematical models and calculations are cumbersome, complicating the control system; second, identification and correction both require a process, which may not be sufficient for systems with rapidly changing parameters, resulting in poor performance. With the advent of DSP controllers, high-speed computation mitigates the drawback of slow calculation. Intelligent control, which has received significant attention in the control field in recent years, can break free from dependence on mathematical models of the controlled object and has become a prominent method for solving robustness problems. Currently, fuzzy control and neural network control are the most mature intelligent control methods in AC servo systems, and most of them add intelligent control techniques to model control to eliminate the effects of parameter changes and disturbances. Fuzzy control uses fuzzy sets to characterize the fuzziness in everyday concepts, enabling the controller to more realistically mimic the control experience and methods of skilled operators and experts. It includes three parts: fuzzification of precise quantities, fuzzy inference, and fuzzy decision. Early fuzzy controllers were only intended to replace traditional PID controllers, and while robustness was improved, fuzzy controllers generally lacked integral action, leading to steady-state error when the servo system experiences load disturbances. While the fuzzy controller with added integral effect is equivalent to a variable coefficient PID controller and can achieve zero steady-state error control, simply using a simple traditional fuzzy controller in a high-precision motor servo system does not yield satisfactory performance. Fuzzy control systems can only achieve excellent performance when combined with other control methods, such as fuzzy PID. Neural Network Control The application of neural network control in AC servo systems mainly includes the following aspects: (1) replacing traditional PID control; (2) since the actual vector control effect is very sensitive to the servo system parameters, neural networks are used for online identification and tracking of motor parameters, and adaptive adjustment of flux and speed controllers; (3) sensorless vector control requires knowing the speed, and neural networks are used to accurately estimate the position and speed; (4) combined with model reference adaptive control, neural network controllers are used for adaptive speed controllers. Although some progress has been made in the research on intelligent control for AC servo systems, many problems remain to be solved. For example, intelligent controllers are mainly designed based on experience, lacking objective theoretical foresight of system performance (such as stability and robustness), and designing a system requires acquiring a large amount of data, making the designed system prone to oscillation. Current Development Trends of AC Servo Systems Based on the overall development and current status of AC servo systems, their development trends are clearly visible, mainly in the following aspects: 1. Continuous Improvement of Theoretical Research: Although many methods exist for implementing AC servo systems, many problems remain to be solved, such as improving system accuracy, reliability, and low-speed performance. Additionally, sensorless control is another research direction: improving speed estimation accuracy while enhancing control performance and strengthening system resistance to parameter changes. 2. Practical Application: Currently, high-performance AC servo systems mainly rely on imports, which restricts the development of China's high-tech industry. Therefore, by learning from the advanced experience of foreign research, we should start from a high starting point and develop high-performance, commercially viable AC servo systems with international standards as soon as possible. 3. Networking: Abroad, industrial automation based on Ethernet has developed significantly. To adapt to this trend, the latest servo systems are equipped with standard serial interfaces and dedicated LAN interfaces, enhancing the interconnectivity between servo units and other control devices, thus simplifying connections with CNC systems.