With the deep penetration of automatic control technology into various fields of industrial automation, conventional control strategies based on precise mathematical models are insufficient to meet the control performance requirements of various systems, including servo systems. The literature discusses intelligent servo control strategies based on fuzzy logic and neural networks, and also briefly mentions other forms of intelligent servo control strategies. To study the problem more deeply, it is necessary to conduct an in-depth analysis of the conventional PID control algorithm to identify the key issues; secondly, the characteristics of the controlled object should be studied to solve the problem of matching the control strategy with the characteristics of the controlled object. The following is a brief discussion of related issues. Problems in the Application of Conventional PID In computer control systems, conventional PID control algorithms generally adopt incremental control algorithms. Their advantages include: relatively easy to obtain good control effects through weighting; because the computer output is incremental, the impact of malfunctions is small, and logical judgment methods can be used to eliminate them if necessary; the impact during manual/automatic switching is small, facilitating disturbance-free switching; when the computer malfunctions, the original value can still be maintained because the output channel or actuator has signal latching function; and no accumulation is required in the formula. The determination of the control increment Δu(k) is only related to the most recent samplings. The incremental PID control algorithm is: ±Δu(k) = KP[e(k) - e(k-1)] + KIe(k) + KD[e(k) - 2e(k-1) + e(k-2)] Where: T, sampling period; k, sampling number; u(k), output value at sampling time k; e(k), deviation value at sampling time k; e(k-1), deviation value at sampling time k-1. Since general computer control systems use a constant sampling period T, once the values ​​of KP, KI, and KD are determined, the control increment can be calculated using the deviation of the first three measurements. Once the sampling period is selected, it generally does not change. For PID parameters, a suitable set of parameters KP, KI, and KD can be found offline to make the system basically close to the optimal working state. The influence of each parameter of the PID controller on the control effect: KP—proportional control. The characteristics of proportional control are simplicity and speed. The disadvantages are: steady-state error for self-balancing systems (self-balancing means the final value of the system's step response is a finite value); and potential oscillations and poor dynamic characteristics for systems with hysteresis. Increasing the proportional gain KP can speed up the response, reduce the steady-state error, and improve control accuracy. However, excessively large KP can lead to significant overshoot and even system instability; conversely, excessively small KP can reduce overshoot and increase stability margin, but it will reduce the system's regulation accuracy and prolong the transient response time. KI-integral control can eliminate the system's static error and is suitable for self-balancing systems. Increasing the integral gain KI (reducing TI) helps reduce the system's steady-state error, but excessive integral action can exacerbate overshoot and even cause oscillations; decreasing the integral gain KI, while beneficial for system stability, avoiding oscillations, and reducing overshoot, is detrimental to eliminating static error. KD (Kind-Difference) – Differential control primarily addresses the inertia of the controlled object to improve its dynamic characteristics. It provides a deceleration signal to brake the response process in advance, helping to reduce overshoot, overcome oscillations, and stabilize the system. Simultaneously, it accelerates the system's response speed and reduces settling time, thus improving the system's dynamic characteristics. The value of KD has a significant impact on the response process. Increasing the derivative action KD helps accelerate the system response, reduce overshoot, and increase stability, but it also increases sensitivity to disturbances and weakens the ability to suppress interference. If KD is too large, the response process will brake excessively in advance, thus prolonging the settling time; conversely, if KD is too small, the deceleration during the control process will lag, overshoot will increase, the system response will slow down, and stability will deteriorate. Therefore, for time-varying and uncertain systems, KD should not be a fixed value but should be randomly changed to adapt to the time constant of the controlled object. Problems with PID Control Algorithms As can be seen from the above analysis: the optimal PID in a strict sense is a mathematical solution, and the optimal PID parameters are actually just a compromise. For example, some working conditions require no overshoot, which PID cannot achieve; it requires establishing a strict mathematical model of the object, which is difficult to achieve in many cases; once the PID parameters are tuned, there are no more means of intervention to control the parameters and the performance of the control process; it is difficult to effectively control complex objects with uncertainties using PID control strategies because it is difficult to mathematically model them. Characteristics of Complex Controlled Objects and Humanoid Intelligent Control Technology Characteristics of Complex Controlled Objects PID control is difficult to control complex objects with uncertainties. The characteristics of such complex objects can be summarized as follows [3] [4]: ​​the unknown, time-varying, random and dispersed nature of system parameters; the unknown and time-varying nature of system time delay; severe nonlinearity of the system; the correlation between system variables; the unknown, diverse and random nature of environmental disturbances. Since the characteristics of such controlled objects are difficult to describe with mathematical models, it is very difficult to achieve high dynamic and static quality control of controlled objects using traditional PID control based on classical control theory and modern control theory methods based on state space description. Therefore, the black box method, i.e., the input-output description method, is generally used to analyze its control system. A large amount of human wisdom, experience and skills are introduced. The controller is designed based on a generalized model that combines mathematical models and knowledge systems. In other words, the control of such systems generally adopts intelligent control strategies, while traditional control is powerless. Characteristics of Human Simulated Intelligent Controller Systems constructed using human simulated intelligent controller (HSIC) technology have the following characteristics [2>: They have sufficient knowledge about human control strategies, controlled objects and the environment, as well as the knowledge of applying this knowledge; they are a hybrid process of non-mathematical generalized models that can be represented by knowledge and models that can be represented by mathematics; the system can be controlled by a multimodal control method that combines open-loop control and qualitative and quantitative control; HSIC has variable structure characteristics, can perform overall self-optimization, and has self-adaptive, self-organizing, self-learning and self-coordinating capabilities; it has compensation and self-repair capabilities, judgment and decision-making capabilities and high reliability. Humanoid intelligent control posits that the system output is a comprehensive reflection of the control action and the internal characteristics of the controlled object. HSIC can identify unstable trend characteristics in the system output response, make online predictions and judgments, and eliminate such unstable trends with corresponding control strategies, ensuring system stability online. These characteristics guarantee the high dynamic and static performance requirements of HSIC in high-precision servo control systems. [b]Humanoid Intelligent Control Algorithm [/b] Static Characteristics of Humanoid Intelligent Controllers The static characteristics of humanoid intelligent control are shown in Figure 1, which to some extent mimics human intelligent control characteristics. Figure 1 illustrates the static characteristics of the working process of humanoid intelligent control. In the analysis, it is assumed that: e represents the system error, e* represents the first derivative of the system error, and U represents the controller output. [IMG=Static Characteristics of Humanoid Intelligent Control]/uploadpic/THESIS/2007/12/2007121415224196850T.jpg[/IMG] Figure 1. Static Characteristics of Humanoid Intelligent Control. OA Segment—Proportional Control Mode: When the system experiences an error and the error trend increases, i.e., when e×e＊>0, the humanoid intelligent controller generates a proportional output U=Kpe, where Kp is the proportional gain, which can greatly exceed the value allowed by traditional proportional controllers. This mode operates in the range from e=0 to e=em1. When e reaches the first error extreme value em1, the mode immediately ends and enters the AB suppression stage. AB Segment—Gain Suppression Control Mode: This is a process of multiplying the originally excessively high proportional gain KP by a factor k less than 1, thereby reducing its gain. Therefore, at point B, the output has dropped to the value corresponding to U01=kKpem1. Suppression control helps improve system quality and increase stability margin. BC segment—Open-loop hold mode: In this stage, the error decreases from its extreme value and can only approach the origin. Therefore, the hold process in segment BC is a line parallel to the e-axis. CD segment → DE segment → EF segment: The second control cycle is still a combination of the three modes, but the direction of action is opposite to the previous cycle. CD segment is a reverse proportional control. When the e-value crosses the U-axis and becomes negative, the system, under the action of the reverse proportional closed-loop control, causes the error to generate another extreme value, -em2. For a stable control system, generally |em2| < |em1|. In the second control cycle, k and K can take different values ​​than in the previous cycle, thus increasing the flexibility of the humanoid intelligent control. FH segment → HG segment → GI segment—The third control cycle: This cycle is in the same direction as the first cycle. After several cycles, the system is controlled to a desired stable state. Dynamic characteristics of the humanoid intelligent controller The dynamic characteristics are shown in Figure 2, which analyzes the intelligent control characteristics in the time domain. From the dynamic characteristics, it can be seen that: in segment OA, the error satisfies the condition e×e＊＞0∪e＝0∩e＊≠0, so proportional control is used, and the control quantity changes with the error proportion; in segment AB, the error satisfies the condition e×e＊＜0∪e＊＝0, so hold control is used, and the control quantity u holds the cumulative sum of the extreme values ​​of error e; in segment BC, the error satisfies the condition e×e＊＞0∪e＝0∩e＊≠0, so proportional control is used, and the control quantity u changes with the error proportion; in segment CD, the error... When the error satisfies the condition e×e＊＜0∪e＊＝0, hold control is adopted, and the control quantity u holds the cumulative sum of extreme values ​​of error e, and is less than the previous hold value. When the error satisfies the condition e×e＊＞0∪e＝0∩e＊≠0, proportional control is adopted, and the control quantity u changes proportionally with the error. When the error satisfies the condition e×e＊＜0∪e＊＝0, hold control is adopted, and the control quantity u holds the cumulative sum of extreme values ​​of error e, and is greater than the previous hold value. Finally, the result of the control is that the error e converges to zero, and the control u converges to a constant value. [IMG=Dynamic Model of Humanoid Intelligent Control]/uploadpic/THESIS/2007/12/2007121412035584860B.jpg[/IMG] Figure 2 Dynamic Model of Humanoid Intelligent Control From the prototype algorithm and characteristics of the humanoid intelligent control's operation control level, it can be seen that the controller identifies and judges the system to be in two different motion states by two simple relationship features of error and its rate of change, and adopts two different control modes respectively. The quantitative control operation output is determined by the memory of error peak features and prior knowledge, such as the relationship between the proportional coefficient K, the inhibition coefficient k and the current error magnitude. These can all be regarded as a simple imitation of human kinesthetic intelligence. From the above dynamic and static characteristics, it can be seen that the operation mechanism of humanoid intelligent control is achieved through correction: dynamic correction is a method of dynamically predicting the motion trend of the system based on the dynamic performance index and the dynamic characteristics of the system during the transition process. If necessary, it is necessary not only to correct the current control mode parameters, but also to temporarily change the current control mode strategy to force the correction of the controller parameters so that the system can operate normally. Static correction is a controller parameter correction method that, after the transient process has ended, evaluates the deviation between the previous dynamic process and the static state based on comprehensive performance indicators, as well as the effectiveness of the dynamic correction, and uniformly weighs the correction of all modes and the modification of the characteristic space partition. Humanoid Intelligent Controller and its Control Algorithm The humanoid intelligent control system is a control system composed of a humanoid intelligent controller, as shown in Figure 3. [IMG=Humanoid Intelligent Controller Block Diagram]/uploadpic/THESIS/2007/12/20071214120401609925.jpg[/IMG] Figure 3 Humanoid Intelligent Controller Block Diagram For some controlled objects, although simple proportional control can ensure their stability, there is often a large steady-state error, which cannot meet the requirements of steady-state accuracy. By using a microcomputer to mimic human operation and continuously adjust the setpoint, the system output can be made to continuously approach the desired value, thereby improving the steady-state accuracy. This is the basic principle of humanoid intelligence. The basic algorithm of an anthropomorphic intelligent controller closely resembles human thought processes: When the system error tends to increase or remains constant, the anthropomorphic intelligent controller adopts proportional control mode, thus generating a strong closed-loop control effect to quickly stop the error from increasing; while when the system error changes towards decreasing or reaches zero, the anthropomorphic intelligent controller adopts hold control mode, canceling the strong control effect, keeping the control quantity at a constant value, and waiting and observing the dynamic process of the system until overshoot occurs again, at which point the controller switches back to proportional control mode. The decision-making process of an anthropomorphic intelligent controller is based on human intelligent behaviors such as observation, memory, and decision-making regarding the controlled object. It uses the deviation of the adjusted quantity and the rate of change of the deviation as characteristic information to select a mode, thereby determining the controller's output. The algorithm is described as follows: Mode 1: if en × Δen > 0 or Δen = 0, |en| > 0, then un = un-1 + k + ×en; Mode 2: if en × Δen < 0 and |en| ≥ M, then un = un-1 + k - ×en; Mode 3: if en × Δen < 0 and |en| ≥ M, then un = un-1 + k × em × n. Where en is the nth error; Δen = en - en-1; K+ is the acceleration coefficient, K+ > 1; K- is the suppression coefficient, 0. Engineering Application Example The servo system is an important component of the radar system. It directly drives the antenna, enabling the radar to quickly and accurately aim at the target. The performance of the entire servo system directly affects the measurement accuracy of the radar system. Traditional radar servo systems employ PID control, with their tracking position loop using phase lag-lead compensation control. Phase lag-lead compensation, through the configuration of zeros and poles, can improve the transient response performance of the system by utilizing the leading part of the compensation network, while the phase lag part can significantly improve the steady-state accuracy. Therefore, it is widely used in high-precision measurement radars. A simplified block diagram of a radar position control loop is shown in Figure 4. [IMG=Simplified Block Diagram of Radar Position Loop]/uploadpic/THESIS/2007/12/2007121412041448281H.jpg[/IMG] Figure 4 Simplified Block Diagram of Radar Position Loop However, PID control parameters in control systems are generally manually tuned. The PID parameters obtained through a one-time tuning are unlikely to guarantee that the control effect is always optimal. A method must be found to adjust the parameters online at any time to improve the dynamic and static performance of the system and ensure that the system always has good adaptive performance. To address the existing problems, a humanoid intelligent control strategy was adopted, replacing the original PID control with an HSIC controller. Preliminary tests showed its response as curve y1 in Figure 5, while y2 represents the response curve of the original PID controller. Clearly, the HSIC controller outperforms the PID controller. [IMG=System Response Performance Comparison]/uploadpic/THESIS/2007/12/2007121412042260712U.jpg[/IMG] Figure 5 System Response Performance Comparison Conclusion This paper mainly discusses the humanoid intelligent control strategy and its prototype control algorithm. Finally, simulation and engineering application examples illustrate the advantages of this strategy. It is worth noting that the dynamic and static performance indicators of the servo control system are closely related to the selection of the control strategy. The specific characteristics of the controlled object should be analyzed, and the control strategy and algorithm should be determined based on the degree to which the control indicators are met, rather than simply rejecting or affirming them. More complex control problems often require the integration of multiple control strategies.