Automated control method and system based on optimal iterative learning control and feedforward control

Abstract

The application belongs to the technical field of industrial automation control, and provides an automation control method and system based on optimal iterative learning control and feedforward control. The automation control method comprises the following steps: constructing an automation control model by using an optimal iterative learning control algorithm and a model-based feedforward control algorithm; optimizing the automation control model; and implementing automation control based on the automation control model. The automation control method based on optimal iterative learning control and feedforward control solves the problems of performance decline of the optimal iterative learning control algorithm when the trajectory changes and possible degradation of the model-based feedforward control algorithm under actuator constraints, and can realize high-precision tracking and reliable control under repeated and non-repeated tasks, and is suitable for high-precision trajectory tracking and actuator constraint management of industrial motion systems.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation control, and in particular to an automation control method and system based on optimal iterative learning control and feedforward control. Background Art

[0002] In modern industrial automation, robotic navigation, aerospace and other fields, the demand for high-precision positioning of unknown targets is increasing. Traditional target positioning methods mainly rely on a single control strategy, such as model-based feedforward control (MFC) or iterative learning control (ILC). However, these methods have many limitations when facing complex environments and unknown targets. MFC requires an accurate system model and its performance will significantly degrade when the trajectory changes. Although iterative learning control can optimize the control input by learning historical trajectory errors, thereby achieving better trajectory tracking performance in repetitive tasks, it has a strong dependence on initial conditions and its performance will significantly deteriorate when the task trajectory changes. In addition, existing systems generally lack effective handling of actuator saturation constraints, which makes it difficult to achieve high-precision positioning in practical applications.

[0003] In recent years, the optimal iterative learning control (OILC) algorithm has attracted attention due to its excellent trajectory tracking performance in repetitive tasks. The optimal iterative learning control algorithm can effectively compensate for tracking errors by continuously optimizing the control input during the iteration process, while taking into account the size and changes of the system control force, thereby avoiding actuator saturation; however, the optimal iterative learning control algorithm still has the problem of performance degradation when processing non-repetitive tasks. At the same time, the model-based feedforward control method provides a new idea for solving the feedforward parameter adjustment problem in non-repetitive tasks. Through online bootstrapping and recursive update algorithms, unbiased parameter estimation can be achieved in a closed-loop setting, thereby achieving good tracking performance in non-repetitive tasks. This provides the possibility of developing a feedforward control algorithm that can adapt to both repetitive and non-repetitive tasks.

[0004] Therefore, an automated control method and system based on optimal iterative learning control and feedforward control is needed to achieve high-precision positioning of unknown targets, while meeting the actuator saturation constraints and trajectory change adaptability in complex environments, providing strong support for achieving high-precision and high-reliability target positioning. Summary of the Invention

[0005] The present invention provides an automated control method and system based on optimal iterative learning control and feedforward control, which solves the problems of performance degradation of the optimal iterative learning control algorithm when the trajectory changes and the possible degradation of the model-based feedforward control algorithm under actuator constraints. It can achieve high-precision tracking and reliable control under repetitive and non-repetitive tasks, and is suitable for industrial motion systems with high-precision trajectory tracking and actuator constraint management.

[0006] The present invention provides an automation control method based on optimal iterative learning control and feedforward control, which is applied to automation control equipment and includes the following steps:

[0007] Step 1: Build an automated control model using the optimal iterative learning control algorithm and the model-based feedforward control algorithm; including:

[0008] S201: Constructing a first automated control sub-model based on an optimal iterative learning control algorithm; the first automated control sub-model is configured to implement the following functions: initializing and adjusting the amplitude and rate of change of a control signal; limiting the magnitude and change of a control force; and optimizing the control signal through iterative learning of historical tracking errors extracted from historical data.

[0009] Limits on the size and variability of control forces, including:

[0010] A weight matrix is ​​introduced to limit the size and rate of change of the control signal. Specifically, a mathematical representation of the physical limits of the actuator is established to define the control signal amplitude constraint and the control signal change rate constraint. A positive definite weight matrix is ​​introduced into the objective function to construct a quadratic cost function. The quadratic cost function is:

[0011]

[0012] in, represents the tracking error, represents the square of the Euclidean norm of the tracking error, represents the control signal vector; A positive definite weight matrix representing the control signal vector; represents the control signal change rate vector; A positive definite weight matrix representing the control signal rate of change vector; represents the quadratic cost function; Represents the total execution time; t represents the specific time in the time interval [0, T]; and is a diagonal matrix used to adjust the size and change rate of the control signal;

[0013] The control signal vector is constrained between the maximum output force of the actuator and the lower limit of the output force of the actuator, and the control signal change rate vector is constrained between the maximum change rate allowed by the actuator and the lower limit of the change rate of the actuator to prevent the actuator from saturating;

[0014] S202: Constructing a second automation control sub-model using a model-based feedforward control algorithm;

[0015] S203: Connect the first automation control sub-model and the second automation control sub-model to construct an automation control model;

[0016] Step 2: Optimize the automation control model;

[0017] Step 3: Implement automated control based on the automated control model.

[0018] Furthermore, the adjustment of the amplitude and rate of change of the control signal is achieved through an iterative learning gain matrix, which satisfies the following formula:

[0019]

[0020] Where k represents the number of iterations, represents the control signal vector for the k+1th iteration, represents the control signal vector for the kth iteration, represents the tracking error of the kth iteration; Represents the proportional gain matrix, which is used to weight the correction of the tracking error amplitude value; represents the differential gain matrix, which is used to weight the correction of the rate of change of the tracking error; Represents the rate of change of tracking error at the kth iteration.

[0021] Furthermore, the second automation control sub-model is used to implement the following functions:

[0022] A feedforward controller designed based on a target mathematical model is used to calculate the control signal compensation when the tracked trajectory changes. The target mathematical model is constructed as follows: based on the basic mathematical model, basis functions and parameter vectors are introduced. The basic mathematical model includes one or more combinations of transfer function models, state space models or neural network models. The basis functions are one or more combinations of polynomial basis functions, Fourier basis functions or wavelet basis functions. The parameter vector is updated by the least squares method or the maximum likelihood estimation method. The control parameters of the target mathematical model are dynamically adjusted according to the task requirements and system characteristics.

[0023] Furthermore, the automation control model is optimized, including:

[0024] A feedback controller is introduced to obtain feedback signals of the automated control model during the training and execution process; the feedback controller is one of a proportional-integral-derivative controller, a robust controller, or a sliding mode controller;

[0025] An adaptive search algorithm is introduced to dynamically adjust the search strategy according to the feedback signal to search for the optimal control parameters of the feedforward controller, and the optimal control parameters are used as the control parameters in the actual use of the automation control model.

[0026] Further, the adaptive search algorithm is one or a combination of recursive least squares, particle swarm optimization or gradient descent algorithm.

[0027] Further, the optimization of the automatic control model further comprises: according to the set event trigger condition of performing iterative update, implementing the optimal control of iterative update of the control signal, and the specific steps are:

[0028] The event trigger condition of performing iterative update is set, the trigger threshold of tracking error change is defined first, and then the trigger condition is set according to the fluctuation of the control signal; the trigger threshold is:

[0029]

[0030] Wherein, is the trigger threshold at the kth iteration and time t; k is the iteration number; is the initial threshold, is the adaptive coefficient, is the maximum output control signal of the actuator; is the control signal at the kth iteration and time t; is the norm of the control signal at the kth iteration and time t; is the norm of the maximum output control signal of the actuator;

[0031] The trigger condition is: when or is satisfied, the iterative update is performed; wherein, is the control signal change weight coefficient; is the tracking error at the kth iteration and time t; is the tracking error at the k-1th iteration and time t; represents and the norm of the difference; is the control signal at the kth iteration and time t; is the control signal at the k-1th iteration and time t; represents and the norm of the difference; is the trigger threshold at the kth iteration and time t;

[0032] According to the event trigger condition, the iterative update is performed, specifically: if the time of iteration is in the non-trigger time interval, the control signal is obtained based on the iteration result of the previous time; if the time of iteration is in the trigger time interval, the control signal is incrementally updated according to the set increment amplitude;

[0033] A multi-objective cost function is constructed. Within the constraints of the control signal vector and the control signal change rate vector, the multi-objective cost function is evaluated for minimum cost. The optimal target of the iterative update of the control signal is obtained based on the evaluation result. The iterative update is stopped when the optimal target is reached. The multi-objective cost function is expressed as:

[0034]

[0035] in, represents the multi-objective cost function, is the weight coefficient of tracking error, is the square of the norm of the tracking error; Express Integral from 0 to T; is the total execution time; is the weight coefficient of the control signal, is the square of the norm of the control signal; Express From 0 to 's points; is the weight coefficient of the trigger frequency; is the trigger moment indicator function. At the trigger moment, is 1, at the non-triggering moment, is 0; Represents the total number of trigger events; Represents the index value of the iterative trigger event, The value range is 1,2,..., ; Indicates the The time when the trigger event occurs.

[0036] An automated control system based on optimal iterative learning control and feedforward control is applied to automated control equipment and is used to implement automated control using an automated control method based on optimal iterative learning control and feedforward control, including:

[0037] An automated control model building module, used to build an automated control model using an optimal iterative learning control algorithm and a model-based feedforward control algorithm;

[0038] Automation control model optimization module, used to optimize the automation control model;

[0039] The automation control implementation module is used to implement automation control based on the automation control model.

[0040] Compared with the prior art, the present application has the following advantages and beneficial effects: by combining the advantages of optimal iterative learning control and model-based feedforward control, the tracking error is significantly reduced, the control accuracy and response speed are improved, and stable control performance can be maintained in the case of external disturbance or system parameter change; the change process of the control signal is optimized, the saturation phenomenon of the actuator caused by excessive signal fluctuation is avoided, thereby prolonging the service life of the equipment; an event triggering mechanism is introduced, the frequency of control signal update is reduced, the consumption of computing resources is reduced, while the dynamic performance of the system is ensured; multi-objective optimization is realized, a comprehensive cost function is constructed to balance the tracking error, control signal amplitude and triggering frequency, and the overall control effect is improved; it has strong adaptability and can dynamically adjust the control parameters according to the specific task requirements and system characteristics, and is suitable for automatic control tasks in various complex scenes.

[0041] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained by the structure particularly pointed out in the written description and the accompanying drawings.

[0042] The technical solutions of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0043] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0044] Figure 1 A flowchart of the steps of the automatic control method based on optimal iterative learning control and feedforward control;

[0045] Figure 2 A flowchart of the steps of the method for optimizing the automatic control model;

[0046] Figure 3 A schematic diagram of the structure of the automatic control system based on optimal iterative learning control and feedforward control. DETAILED DESCRIPTION

[0047] The preferred embodiments of the present application will be described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not limit the present application.

[0048] The present application provides an automatic control method based on optimal iterative learning control and feedforward control, as shown in Figure 1 applied to an automatic control device, comprising the following steps:

[0049] Step 1: Build an automated control model using the optimal iterative learning control algorithm and the model-based feedforward control algorithm; including:

[0050] S201: Constructing a first automated control sub-model based on an optimal iterative learning control algorithm; the first automated control sub-model is configured to implement the following functions: initializing and adjusting the amplitude and rate of change of a control signal; limiting the magnitude and change of a control force; and optimizing the control signal through iterative learning of historical tracking errors extracted from historical data.

[0051] Limits on the size and variability of control forces, including:

[0052] A weight matrix is ​​introduced to limit the size and rate of change of the control signal. Specifically, a mathematical representation of the physical limits of the actuator is established to define the control signal amplitude constraint and the control signal change rate constraint. A positive definite weight matrix is ​​introduced into the objective function to construct a quadratic cost function. The quadratic cost function is:

[0053]

[0054] in, represents the tracking error, represents the square of the Euclidean norm of the tracking error, represents the control signal vector; A positive definite weight matrix representing the control signal vector; represents the control signal change rate vector; A positive definite weight matrix representing the control signal rate of change vector; represents the quadratic cost function; Represents the total execution time; t represents the specific time in the time interval [0, T]; and is a diagonal matrix used to adjust the size and change rate of the control signal;

[0055] The control signal vector is constrained between the maximum output force of the actuator and the lower limit of the output force of the actuator, and the control signal change rate vector is constrained between the maximum change rate allowed by the actuator and the lower limit of the change rate of the actuator to prevent the actuator from saturating;

[0056] S202: Constructing a second automation control sub-model using a model-based feedforward control algorithm;

[0057] S203: Connect the first automation control sub-model and the second automation control sub-model to construct an automation control model;

[0058] Step 2: Optimize the automation control model;

[0059] Step three: implementing automatic control based on the automatic control model.

[0060] The working principle of the technical solution is as follows: in order to realize the automatic control method based on optimal iterative learning control and feedforward control, the method is applied to an automatic control device, and the automatic control model is constructed by using an optimal iterative learning control algorithm and a feedforward control algorithm based on a model; the optimal iterative learning control algorithm is a control algorithm for improving the next iteration by using historical operation data, and the historical operation data can be error or control input; for example, for high-precision positioning of a mechanical arm and a conveyor belt in industrial automation, the historical operation data includes a desired trajectory of the last iteration, specifically a spatial path coordinate sequence that a mechanical arm end effector needs to accurately follow, and a precise position sequence that a certain point on a conveyor belt needs to reach at a specific time; actual output of the last iteration, specifically actual position coordinates of the mechanical arm end effector at each sampling time in k runs, and actual positions of a conveyor belt marker point at each sampling time in k runs; k represents the number of iterative runs; control input of the last iteration, specifically a torque instruction sequence for driving a mechanical arm joint motor, and a speed or position instruction sequence for driving a conveyor belt motor; tracking error of the last iteration, specifically a position deviation vector of the mechanical arm end effector at each time in k runs, a position deviation of the conveyor belt marker point in k runs, and an iteration number identifier; k represents the number of iterative runs; this algorithm continuously adjusts and optimizes the control signal through historical data and iterative learning, so that it gradually approaches the ideal output; the feedforward control algorithm based on the model calculates the influence of disturbance on the output by using a mathematical model, and compensates for the influence directly to offset the influence before the disturbance actually affects the output; the automatic control model constructed by combining the two algorithms can not only use the mathematical model for accurate prediction, but also dynamically adapt to complex environmental changes through iterative learning.

[0061] In order to realize the construction of the automatic control model, the first automatic control sub-model is first constructed based on the optimal iterative learning control algorithm. The optimal iterative learning control algorithm gradually optimizes the generation process of the control signal by continuously accumulating historical operation data, thereby improving the dynamic response capability of the system; then the second automatic control sub-model is constructed using the model-based feedforward control algorithm, and the model-based feedforward control algorithm uses precise mathematical modeling to predict possible disturbances in advance and implement compensation; finally, the first automatic control sub-model is connected with the second automatic control sub-model. The connection is based on the use of the first automatic control sub-model to perform control, and then the first automatic control sub-model is used to perform control. The first automatic control sub-model is to control the reduction of the trajectory error, and the second automatic control sub-model is to respond in a very short time when the trajectory changes, so as to achieve accurate and rapid tracking of the trajectory. Then the automatic control model is generated. The constructed automatic control model can fully utilize the advantages of the two algorithms and lay a solid foundation for subsequent optimization and implementation.

[0062] During the initialization phase, the first automated control sub-model initializes and adjusts the amplitude and rate of change of the control signal. Simultaneously, during operation, through in-depth analysis of historical data, it extracts key tracking error information and uses this information for iterative optimization, thereby gradually reducing deviations. By limiting the magnitude and rate of change of the control force, the stability and security of the system are further enhanced, avoiding unnecessary fluctuations caused by sudden changes or over-adjustments.

[0063] To limit the magnitude and variation of the control force and prevent actuator saturation, the present invention first introduces a weight matrix and constructs a quadratic cost function. Through the regulation of the diagonal matrices Q and R, the relationship between the magnitude and rate of change of the control signal can be flexibly balanced to meet the needs of different application scenarios. At the same time, the control signal is constrained within the physical limits of the actuator. By embedding the constraints into the optimization process, the adaptability and robustness of the algorithm are further enhanced, enabling it to maintain high performance in complex dynamic environments.

[0064] Then, the automation control model is optimized. The optimization includes: introducing a feedback controller and an adaptive search algorithm, using the obtained optimal control parameters as the control parameters in the actual use of the automation control model; and implementing optimal control of iterative updates of control signals based on set event trigger conditions for executing iterative updates;

[0065] Finally, based on the automation control model, automation control is implemented, such as the robot arm in industrial automation when repeating the welding path, the joint friction will cause the trajectory to have an error >0.5mm, to solve this problem, first use the optimal iterative learning control algorithm in the automation control model to correct the control signal through the first 3 times of iterative learning, so that the tracking error is reduced to below 0.1mm; when the welding path changes, the model-based feedforward control algorithm in the automation control model is used again, and the feedforward signal is calculated in real time based on the inverse dynamics model of the robot arm, so as to quickly reduce the trajectory switching time, thereby realizing precise automation control.

[0066] The beneficial effects of the above technical solutions are: by using the scheme provided in the embodiment, the optimal iterative learning control algorithm and the model-based feedforward control algorithm are used to construct the automation control model, and are applied to automation control in automation control equipment, which can reduce control error, improve control precision, stability and adaptability, and provide reliable technical support for actual high-precision control applications; by combining the optimal iterative learning control algorithm and the model-based feedforward control algorithm to construct the automation control model, the performance of the automation control model can be improved, and reliable technical support is provided for high-precision control scenarios; the first automation control sub-model can significantly improve the control precision and response speed of the system, while reducing the instability risk caused by external interference or internal fluctuations; by limiting the size and change of the control force, the performance degradation problem caused by overloading or excessive change rate of the actuator can be effectively avoided, thereby prolonging the service life of the equipment and improving the stability of the system; by introducing the diagonal matrix adjustment mechanism, not only the flexibility of the control signal is improved, but also accurate regulation and control can be realized in variable application scenarios; embedding the constraint condition into the optimization process design further enhances the practicality of the algorithm.

[0067] In one embodiment, the adjustment of the amplitude and the change rate of the control signal is realized through an iterative learning gain matrix, and the iterative learning gain matrix satisfies the following formula:

[0068]

[0069] wherein k represents the iteration number, represents the control signal vector of the k+1th iteration, represents the control signal vector of the kth iteration, represents the tracking error of the kth iteration; represents a proportional gain matrix for weighted correction of the amplitude value of the tracking error; represents a differential gain matrix for weighted correction of the change rate of the tracking error; represents the change rate of the tracking error of the kth iteration.

[0070] The working principle of the technical solution is that the amplitude of the control signal refers to the numerical value of the control signal, such as the amplitude of the motor driving force, which directly affects the dynamic range of the system output; the rate of change of the control signal refers to the speed of change of the control signal with time, such as the acceleration and deceleration rate of the driving force, which affects the smoothness of the system response and mechanical stress; the adjustment of the amplitude and rate of change of the control signal is to make the error between the system output trajectory and the expected trajectory converge to zero, i.e. complete tracking, within a limited time, such as a single operation cycle of a mechanical arm; the present application can gradually optimize the control signal according to the error information in the historical data by introducing an iterative learning gain matrix, and in each iteration process, the proportional gain matrix and the differential gain matrix respectively weight and correct the amplitude and rate of change of the tracking error, thereby realizing dynamic adjustment of the control signal, for example: in the application scenario of precise positioning of a semiconductor wafer table, the implementation process is as follows: in the first iteration, the amplitude of the control signal is set according to the empirical value, and the rate of change of the control signal is limited according to the maximum acceleration of the motor, and in the subsequent iteration process, the amplitude of the driving force is increased by the proportional gain matrix to reduce the positioning deviation, and the rate of change in the acceleration stage is reduced by the differential gain matrix to suppress mechanical vibration; after 3 iterations, the nanoscale positioning error can converge from ±50nm to ±10nm.

[0071] The beneficial effects of the above technical solution are that the scheme provided by the embodiment not only improves the self-adaptive ability of the system, but also enhances its robustness in complex environments, and can quickly converge to the target value within a limited time interval, while avoiding the instability caused by excessive or insufficient control force.

[0072] In one embodiment, the second automated control sub-model is used to implement the following functions:

[0073] The feedforward controller designed based on the target mathematical model is used to calculate the control signal compensation when the tracked trajectory changes; wherein the target mathematical model is constructed by introducing a basis function and a parameter vector based on a basic mathematical model; the basic mathematical model includes one or more combinations of a transfer function model, a state space model or a neural network model; the basis function is one or more combinations of a polynomial basis function, a Fourier basis function or a wavelet basis function; the parameter vector is updated by least squares method or maximum likelihood estimation method; and the control parameters of the target mathematical model are dynamically adjusted according to task requirements and system characteristics.

[0074] The working principle of the technical solution is that: in order to realize accurate control of a complex system, the design of the feedforward controller combined with the target mathematical model in the embodiment can effectively cope with the challenges brought by trajectory changes; in the specific implementation process, the selection of the basis function and the updating mechanism of the parameter vector work together to ensure the adaptability of the model to the dynamic environment; by introducing polynomial basis functions, Fourier basis functions or wavelet basis functions, the nonlinear relationship between system input and output can be flexibly captured; at the same time, the least square method or the maximum likelihood estimation method is used to iteratively optimize the parameter vector, further improving the prediction accuracy and robustness of the model; this design not only can dynamically adjust the control parameters according to the task requirements, but also can maintain stable performance in various application scenarios.

[0075] The beneficial effects of the above technical solution are that: by using the scheme provided in the embodiment, the response speed and accuracy of the control system can be significantly improved, and by dynamically adjusting the control parameters, the system can maintain efficient operation under different working conditions, thereby reducing the need for manual intervention.

[0076] In one embodiment, the automation control model is optimized, as shown in Figure 2 , including:

[0077] A feedback controller is introduced to obtain the feedback signal of the automation control model during the training execution process; the feedback controller is one of a proportional-integral-derivative controller, a robust controller or a sliding mode controller;

[0078] An adaptive search algorithm is introduced to dynamically adjust the search strategy according to the feedback signal, search for the optimal control parameters of the feedforward controller, and use the optimal control parameters as the control parameters in the actual use process of the automation control model.

[0079] The working principle of the above technical solution is that: first, the feedback controller collects system running data in real time, compares these data with target values, and generates feedback signals; based on the feedback signals, the adaptive search algorithm analyzes the performance of the current control parameters and dynamically adjusts the search direction and step size to more quickly approach the optimal solution; this process ensures that the feedforward controller can always maintain optimal performance under different working conditions by continuously evaluating the impact of the control parameters on the system output; at the same time, the adaptive search algorithm has strong robustness and can effectively cope with the challenges brought by system nonlinearities, time-varying characteristics and external disturbances, thereby further improving the stability and accuracy of the overall control system.

[0080] The beneficial effects of the above technical solution are that: by using the scheme provided in the embodiment, through the synergistic effect of the feedback controller and the adaptive search algorithm, the system can achieve fast response and accurate control under complex and variable working conditions.

[0081] In one embodiment, the adaptive search algorithm is one or a combination of recursive least squares, particle swarm optimization, or gradient descent.

[0082] The working principle of the above technical solution is that the recursive least squares iteratively optimizes by constantly updating parameter estimates using historical data and current feedback signals to quickly approach the optimal solution; the particle swarm optimization simulates group behavior and searches for the best control parameter combination in the solution space through collaboration and information sharing between individuals, with the characteristic of strong global search capability; the gradient descent algorithm adjusts parameters step by step along the descending direction based on the gradient information of the objective function to achieve a local optimal solution; when these algorithms are used alone or in combination, their respective advantages can be fully utilized to adapt to different working conditions and further improve the performance of the control system.

[0083] The beneficial effects of the above technical solution are that by flexibly combining recursive least squares, particle swarm optimization, and gradient descent, the system can dynamically switch or integrate algorithm strategies according to different working conditions, thereby maintaining efficient operation in complex environments.

[0084] In one embodiment, optimizing the automated control model further includes implementing optimal control of iterative updates of the control signal according to a set event trigger condition for performing iterative updates, and the specific steps are:

[0085] The event trigger condition for performing iterative updates is first defined by setting a trigger threshold for tracking error changes, and then setting a trigger condition according to fluctuations in the control signal; the trigger threshold is:

[0086]

[0087] wherein, is the trigger threshold at time t for the kth iteration; k is the number of iterations; is the initial threshold, is the adaptive coefficient, is the maximum output control signal of the actuator; is the control signal at time t for the kth iteration; is the norm of the control signal at time t for the kth iteration; is the norm of the maximum output control signal of the actuator;

[0088] The trigger condition is that when or is met, the iterative update is performed; wherein, is the control signal change weight coefficient; is the tracking error at time t for the kth iteration; is the tracking error at time t for the (k-1)th iteration; represent and The norm of the difference; is the control signal at the kth iteration and time t; is the control signal at the k-1th iteration and time t; represent and The norm of the difference; is the trigger threshold at the kth iteration and time t;

[0089] Iterative updates are performed based on event trigger conditions. Specifically, if the iteration time is within the non-triggering time interval, the control signal is obtained based on the previous iteration result. If the iteration time is within the triggering time interval, the control signal is incrementally updated according to the set increment amplitude.

[0090] A multi-objective cost function is constructed. Within the constraints of the control signal vector and the control signal change rate vector, the multi-objective cost function is evaluated for minimum cost. The optimal target of the iterative update of the control signal is obtained based on the evaluation result. The iterative update is stopped when the optimal target is reached. The multi-objective cost function is expressed as:

[0091]

[0092] in, represents the multi-objective cost function, is the weight coefficient of tracking error, is the square of the norm of the tracking error; Express Integral from 0 to T; is the total execution time; is the weight coefficient of the control signal, is the square of the norm of the control signal; Express From 0 to 's points; is the weight coefficient of the trigger frequency; is the trigger moment indicator function. At the trigger moment, is 1, at the non-triggering moment, is 0; Represents the total number of trigger events; Represents the index value of the iterative trigger event, The value range is 1,2,..., ; Indicates the The time when the trigger event occurs.

[0093] The working principle of the technical solution is as follows: the traditional time-driven control method often updates the control signal at each fixed time interval, which may cause unnecessary iteration operation when the system state changes smoothly, thereby wasting computing resources; the event triggering mechanism adopted by the application can avoid this problem, and the control signal is only iteratively updated when the specific triggering condition is met, thereby greatly reducing the unnecessary iteration times; the adaptive threshold is used to dynamically adjust the triggering condition according to the actual running state; in actual application, the running state of the system is affected by various factors, such as changes in the external environment, aging of internal components, etc.; the adaptive threshold can be adjusted in real time according to these changes, thereby ensuring that the system can stably run under various complex conditions, and the robustness and adaptability of the system can be improved.

[0094] In the specific implementation process, the incremental update strategy of the control signal updates the control signal according to the set increment amplitude at the triggering moment, fully utilizes the advantages of the feedforward control, and enables the desired control effect to be achieved faster; meanwhile, the tracking error, control signal energy and triggering frequency in the multi-objective cost function are used as three optimization targets to measure the update target of the control signal; in actual control implementation, different performance indicators often restrict each other; for example, in order to reduce the tracking error, the energy of the control signal may need to be increased; and too much control signal update will increase the triggering frequency, thereby increasing the computing burden of the system; the multi-objective cost function balances and optimizes different performance indicators by setting weight coefficients; by adjusting the weight coefficients of the tracking error, the weight coefficients of the control signal and the weight coefficients of the triggering frequency, the target of the control signal update can be flexibly adjusted according to the specific application requirements.

[0095] The beneficial effects of the above technical solution are that: by adopting the scheme provided in the embodiment, the event triggering condition and the multi-objective cost function are set, the consumption of computing resources can be significantly reduced while the control precision is ensured; by flexibly configuring the triggering condition and the cost function parameters, the application scenarios can be customized and designed.

[0096] The automatic control system based on optimal iterative learning control and feedforward control, as shown in Figure 3 is applied to an automatic control device, and is used to implement automatic control by using an automatic control method based on optimal iterative learning control and feedforward control, and includes:

[0097] An automatic control model construction module is used to construct an automatic control model by using an optimal iterative learning control algorithm and a model-based feedforward control algorithm.

[0098] An automatic control model optimization module is used to optimize the automatic control model.

[0099] an automation control implementation module for implementing automation control based on the automation control model;

[0100] The automation control system based on optimal iterative learning control and feedforward control is applied to automation control devices including but not limited to a mechanical arm in industrial automation, high-precision positioning of a conveying belt, path tracking in robot navigation, and aircraft attitude control in the aerospace field.

[0101] Specifically, the automation control is applied in a semiconductor wafer handling robot system:

[0102] In the repeated handling task performed by the automation control device of the semiconductor wafer handling robot, the optimal iterative learning control algorithm in the automation control model is first used to correct the control signal through the first three iterations, so that the tracking error is reduced from the initial 0.1 mm to 0.02 mm. When the wafer placement position is changed, the model-based feedforward control algorithm in the automation control model is used to calculate the feedforward signal in real time based on the inverse model of the robot dynamics, so that the trajectory switching time is shortened from 1.2 s to 0.6 s. The weight matrix is set to limit the motor driving force amplitude ≤ 50 N and the change rate ≤ 10 N / s to avoid motor overload alarm. Finally, the feedforward parameters are optimized by online recursive least squares to improve the positioning accuracy.

[0103] The automation control is applied in the path planning of a UAV:

[0104] First, in the repeated flight path task of the UAV, the optimal iterative learning control algorithm in the automation control model is used to make the flight path tracking error less than 0.5 m. When encountering a sudden obstacle, the model-based feedforward control algorithm in the automation control model is used to quickly generate an obstacle avoidance feedforward signal based on the UAV dynamics model, which can make the response time < 200 ms. At the same time, the weight matrix is set to constrain the motor output power within the rated range, which can improve the continuous operation stability of the UAV path planning system by 75%.

[0105] The working principle of the above technical solution is as follows: In order to realize the automation control system based on optimal iterative learning control and feedforward control, the automation control method based on optimal iterative learning control and feedforward control is used to implement automation control. The automation control model construction module is proposed to model the automation control model by combining the optimal iterative learning control algorithm and the model-based feedforward control algorithm.

[0106] Then, the automation control model optimization module optimizes the automation control model.

[0107] Finally, the automation control implementation module implements automation control according to the optimized control model.

[0108] The technical scheme has the beneficial effects that: the scheme provided by the embodiment can effectively improve the precision and efficiency of automatic control, is suitable for different types of automatic control equipment, and provides a more flexible and reliable solution for the industrial automation field.

[0109] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the application. Accordingly, it is intended that all such modifications and changes be included within the scope of the application as defined by the following claims and their equivalents.

Claims

1. An automated control method based on optimal iterative learning control and feedforward control, characterized in that: Applied to automated control equipment, the method comprises the following steps: Step 1: Build an automated control model using the optimal iterative learning control algorithm and the model-based feedforward control algorithm; including: S201: Constructing a first automated control sub-model based on an optimal iterative learning control algorithm; the first automated control sub-model is configured to implement the following functions: initializing and adjusting the amplitude and rate of change of a control signal; limiting the magnitude and change of a control force; and optimizing the control signal through iterative learning of historical tracking errors extracted from historical data. Limits on the size and variability of control forces, including: A weight matrix is ​​introduced to limit the size and rate of change of the control signal. Specifically, a mathematical representation of the physical limits of the actuator is established to define the control signal amplitude constraint and the control signal change rate constraint. A positive definite weight matrix is ​​introduced into the objective function to construct a quadratic cost function. The quadratic cost function is: in, represents the tracking error, represents the square of the Euclidean norm of the tracking error, represents the control signal vector; A positive definite weight matrix representing the control signal vector; represents the control signal change rate vector; A positive definite weight matrix representing the control signal change rate vector; represents the quadratic cost function; Represents the total execution time; t represents the specific time in the time interval [0, T]; and is a diagonal matrix used to adjust the size and change rate of the control signal; The control signal vector is constrained between the maximum output force of the actuator and the lower limit of the output force of the actuator, and the control signal change rate vector is constrained between the maximum change rate allowed by the actuator and the lower limit of the change rate of the actuator to prevent the actuator from saturating; S202: Constructing a second automation control sub-model using a model-based feedforward control algorithm; S203: Connect the first automation control sub-model and the second automation control sub-model to construct an automation control model; Step 2: Optimize the automation control model; Step 3: Implement automated control based on the automated control model; Optimize the automation control model, including: A feedback controller is introduced to obtain feedback signals of the automated control model during the training and execution process; the feedback controller is one of a proportional-integral-derivative controller, a robust controller, or a sliding mode controller; An adaptive search algorithm is introduced to dynamically adjust the search strategy according to the feedback signal to search for the optimal control parameters of the feedforward controller, and the optimal control parameters are used as the control parameters in the actual use of the automation control model; Optimizing the automation control model also includes: implementing optimal control of the iterative update of the control signal according to the set event triggering conditions for executing the iterative update. The specific steps are as follows: Set the event trigger conditions for iterative updates. First, define the trigger threshold for tracking error changes, and then set the trigger conditions based on the fluctuation of the control signal. The trigger threshold is: in, is the trigger threshold at the kth iteration and time t; k is the number of iterations; is the initial threshold, is the adaptive coefficient, is the maximum output control signal of the actuator; is the control signal at the kth iteration and time t; is the norm of the control signal at the kth iteration and time t; is the norm of the maximum output control signal of the actuator; The trigger condition is: when or When , iterative update is performed; among them, is the weight coefficient of control signal change; is the tracking error at the kth iteration and time t; is the tracking error at the k-1th iteration and time t; represent and The norm of the difference; is the control signal at the kth iteration and time t; is the control signal at the k-1th iteration and time t; represent and The norm of the difference; is the trigger threshold at the kth iteration and time t; Iterative updates are performed based on event trigger conditions. Specifically, if the iteration time is within the non-triggering time interval, the control signal is obtained based on the previous iteration result. If the iteration time is within the triggering time interval, the control signal is incrementally updated according to the set increment amplitude. A multi-objective cost function is constructed. Within the constraints of the control signal vector and the control signal change rate vector, the multi-objective cost function is evaluated for minimum cost. The optimal target of the iterative update of the control signal is obtained based on the evaluation result. The iterative update is stopped when the optimal target is reached. The multi-objective cost function is expressed as: in, represents the multi-objective cost function, is the weight coefficient of tracking error, is the square of the norm of the tracking error; Express Integral from 0 to T; is the total execution time; is the weight coefficient of the control signal, is the square of the norm of the control signal; Express From 0 to 's points; is the weight coefficient of the trigger frequency; is the trigger moment indicator function. At the trigger moment, is 1, at the non-triggering moment, is 0; Represents the total number of trigger events; Represents the index value of the iterative trigger event, The value range is 1,2,..., ; Indicates the The time when the trigger event occurs.

2. The automated control method based on optimal iterative learning control and feedforward control according to claim 1, characterized in that: The adjustment of the amplitude and rate of change of the control signal is achieved through the iterative learning gain matrix, which satisfies the following formula: Where k represents the number of iterations, represents the control signal vector for the k+1th iteration, represents the control signal vector for the kth iteration, represents the tracking error of the kth iteration; Represents the proportional gain matrix, which is used to weight the correction of the tracking error amplitude value; represents the differential gain matrix, which is used to weight the correction of the rate of change of the tracking error; Represents the rate of change of tracking error at the kth iteration.

3. The automated control method based on optimal iterative learning control and feedforward control according to claim 1, characterized in that: The second automation control sub-model is used to implement the following functions: A feedforward controller designed based on a target mathematical model is used to calculate the control signal compensation when the tracked trajectory changes. The target mathematical model is constructed as follows: based on the basic mathematical model, basis functions and parameter vectors are introduced. The basic mathematical model includes one or more combinations of transfer function models, state space models or neural network models. The basis functions are one or more combinations of polynomial basis functions, Fourier basis functions or wavelet basis functions. The parameter vector is updated by the least squares method or the maximum likelihood estimation method. The control parameters of the target mathematical model are dynamically adjusted according to the task requirements and system characteristics.

4. The automated control method based on optimal iterative learning control and feedforward control according to claim 1, characterized in that: The adaptive search algorithm is one or more combinations of recursive least squares method, particle swarm optimization algorithm or gradient descent algorithm.

5. An automated control system based on optimal iterative learning control and feedforward control, characterized in that: Applied to an automation control device, for implementing automation control using the automation control method based on optimal iterative learning control and feedforward control as claimed in any one of claims 1 to 4, comprising: An automated control model building module, used to build an automated control model using an optimal iterative learning control algorithm and a model-based feedforward control algorithm; Automation control model optimization module, used to optimize the automation control model; The automation control implementation module is used to implement automation control based on the automation control model.

Images