Control device and program

Abstract

The present invention provides a control device and program that suppress increases in computational load, reduce the impact of model errors (the difference between the characteristics of the controlled object and the characteristics of the nominal model), and maintain good control performance by suppressing the occurrence of malfunctions in the controlled object even when the model error becomes large. [Solution] The control device 1 includes a nominal model 12 representing the characteristics of the controlled object 20, a controller 11 that performs model predictive control based on the nominal model 12, and a control input u to the controlled object 20 based on the difference between the output of the controlled object 20 and the output of the nominal model 12. c Compensation input u for adjusting d A model error suppression compensator 13 outputs a compensation input u d The system includes an identification unit 14 that, when the value exceeds a predetermined threshold, identifies the controlled object 20, modifies the nominal model 12, and applies the modified nominal model 12 to the internal model of the controller 11.

Description

[Technical Field]

[0001] The present invention relates to a control device and program using a model error suppression compensator and a model prediction controller. [Background technology]

[0002] In recent years, the development of advanced control systems using numerous sensors and actuators has progressed in a wide range of fields, including industrial machinery and autonomous driving in automobiles. Model-based development (MBD) is used as an efficient development method for such complex control systems. In model-based development, the control system is designed against an ideal model (nominal model) constructed from the physical phenomena of the controlled object, and the desired control performance can be obtained by implementing the designed controller. However, in reality, a discrepancy (model error) occurs between the controlled object and the nominal model, and this model error is thought to affect the control performance.

[0003] To suppress the effects of such model errors, a method has been developed to re-identify the model at predetermined intervals or at timings such as changes in operating conditions (for example, Patent Document 1). However, if model re-identification is repeated at predetermined intervals, the computational load for identification may become excessive. Furthermore, since a certain error (identification error) occurs in the identification process, repeating the identification process in a short period may actually degrade control performance. On the other hand, if the re-identification period is set to be long, the effects of model errors become larger before re-identification can be performed, which may also degrade control performance.

[0004] As a control system design method to suppress the effects of model errors without re-identifying the controlled object, a Model Error Compensator (MEC) (e.g., Non-Patent Document 1) has been proposed, which incorporates a nominal model of the controlled object into the control system and utilizes the difference between the output of the controlled object (control output) and the output of the nominal model (model output). By performing such compensation and suppressing the effects of model errors, the response characteristics of the controlled object can be brought closer to the response characteristics of the nominal model, making it possible to directly apply a controller designed for the nominal model to the actual controlled object. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2011-198327 [Non-patent literature]

[0006] [Non-Patent Document 1] H.Okajima, H.Umei, N.Matsunaga, T.Asai, “A Design Method of Compensator Minimize Model Error”, SICE Journal of Control, Measurement, and System Integration, Vol.6, No.4, pp.267-275, 2013 [Overview of the Initiative] [Problems that the invention aims to solve]

[0007] A model error suppression compensator is configured to adjust the control input (manipulated variable) by feeding back the effect of the difference between the control output and the model output. Therefore, if there is a large discrepancy between the characteristics of the actual controlled object and the characteristics of the nominal model, an excessive manipulated variable may be input to the controlled object, potentially causing failures, accidents, or shortening the product's lifespan.

[0008] The present invention has been made in view of the above circumstances, and while suppressing an increase in the calculation load, it reduces the influence of a model error, which is the difference between the characteristics of the control target and the characteristics of the nominal model, and even when the model error becomes large, it suppresses the occurrence of malfunctions in the control target and aims to provide a control device and a program capable of maintaining good control performance.

Means for Solving the Problems

[0009] In order to achieve the above object, a control device according to a first aspect of this invention includes a nominal model representing the characteristics of a control target, a controller that performs model predictive control based on the nominal model, a model error suppression compensator that outputs a compensation input for adjusting a control input to the control target based on a difference between an output of the control target and an output of the nominal model, and an identification unit that identifies the control target, corrects the nominal model, and applies the corrected nominal model to an internal model of the controller when the compensation input exceeds a predetermined threshold value.

[0010] A control device according to a second aspect of this invention includes a nominal model representing the characteristics of an actuator among control targets including an actuator that controls the operation of a machine, a controller that performs model predictive control based on the nominal model, a model error suppression compensator that outputs a compensation input for adjusting a control input to the actuator based on a difference between an output of the actuator and an output of the nominal model, and an identification unit that identifies the actuator, corrects the nominal model, and applies the corrected nominal model to an internal model of the controller when the compensation input exceeds a predetermined threshold value.

[0011] Furthermore, the controlled object is a redundant input system including a plurality of actuators. Each actuator is provided with the nominal model and the model error suppression compensator. The identification unit identifies the actuator for which the compensation input exceeds the threshold, modifies the nominal model, and applies the modified nominal model to the internal model of the controller. It would be acceptable to do so.

[0012] Furthermore, if the identification unit performs identification and modifies the nominal model, it will not perform identification until the compensation input falls below the threshold. It would be acceptable to do so.

[0013] A program relating to a third aspect of this invention is: Computers, A controller that performs model predictive control based on a nominal model that represents the characteristics of the controlled object. A model error suppression compensator that outputs a compensation input for adjusting the control input to the controlled object based on the difference between the output of the controlled object and the output of the nominal model. An identification unit that, when the compensation input exceeds a predetermined threshold, identifies the controlled object, modifies the nominal model, and applies the modified nominal model to the internal model of the controller. To operate as such. [Effects of the Invention]

[0014] According to the control device and program of the present invention, since a model error suppression compensator is included, it is possible to reduce the effect of errors between the controlled object and the nominal model while suppressing an increase in computational load. Furthermore, since the controlled object is re-identified when the compensation input becomes large due to model errors, it is possible to suppress the occurrence of malfunctions in the controlled object and maintain good control performance. [Brief explanation of the drawing]

[0015] [Figure 1]This is a block diagram showing the configuration of a control system including a control device according to an embodiment of the present invention. [Figure 2] This is a block diagram showing the basic structure of a model error suppression compensator according to an embodiment. [Figure 3] This is a block diagram showing the configuration of a model error suppression compensator based on generalized minimum variance control. [Figure 4] This is a block diagram showing the configuration of the control system related to the numerical example. [Figure 5] This is a block diagram showing the configuration of the subsystem related to the numerical example. [Figure 6] This figure shows the results of the control output and control input related to the comparative example. [Figure 7] This figure shows the power consumption results for the actuator in the comparative example. [Figure 8] This figure shows the output and input results of the control according to a numerical example of the present invention. [Figure 9] This figure shows the power consumption results of an actuator according to a numerical example of the present invention. [Modes for carrying out the invention]

[0016] The control device according to an embodiment of the present invention will be described below with reference to the figures.

[0017] The control device 1 according to this embodiment comprises a controller 11, a nominal model 12, a model error suppression compensator 13, and an identification unit 14, as shown in the block diagram of Figure 1. The control device 1 is configured using, for example, a computer device, and elements such as the controller 11 and the nominal model 12 are implemented as software. The control device 1 controls the output y of the controlled object 20 and the output y of the nominal model 12. n Based on the difference, the control input u input to the controlled object 20 is adjusted. This makes it possible to obtain control performance close to the desired performance using the controller 11 designed based on the nominal model 12, even if there is a model error between the controlled object 20 and the nominal model 12.

[0018] The controller 11, based on the target input r, control output y, etc., sets the control input u to the controlled object 20. c The output is as follows: The controller 11 according to this embodiment is a controller that performs Model Predictive Control (MPC) based on a nominal model 12 that represents the characteristics of the controlled object 20. In Model Predictive Control, the model of the controlled object 20 is kept as an internal model, and the optimization problem is calculated at each control cycle based on the internal model to perform control, so that high-performance control can be performed for complex control systems.

[0019] The controlled object 20 is an object controlled by the controller 11. The controlled object 20 in this embodiment includes an actuator 21 that controls the operation of the machine. As shown in Figure 1, the controller 11 in this embodiment receives a control input u which is the amount manipulated by the actuator 21. c By outputting this, the controlled object 20 is controlled.

[0020] The nominal model 12 is a mathematical model that represents the control characteristics of the controlled object 20. The nominal model 12 is established by pre-identifying the characteristics of the controlled object 20.

[0021] The model error suppression compensator 13 uses the output y of the controlled object 20 and the output y of the nominal model 12. n Based on this, compensate input u to suppress model errors. d It is a compensation element that provides feedback. More specifically, the control input u, which is the output of the controller 11. c is the compensation input u d The adjustment is made, and the adjusted control input u is input to the controlled object 20.

[0022] The model error suppression compensator 13 suppresses the influence of model errors in the controlled object 20, thereby bringing the response characteristics of the controlled object 20 closer to those of the nominal model 12. This makes it possible to obtain high control performance without periodically performing computationally intensive processes such as re-identifying the controlled object 20 and constructing the nominal model 12.

[0023] The model error suppression compensator 13 is H ∞ Based on control methods (Hiroshi Okajima, "Robustification of Existing Control Systems Using Model Error Suppression Compensators", Measurement and Control, Vol. 62, No. 3, pp. 168-175, 2023), methods based on adaptive control techniques (Shin Wakitani, et al., "Design of a Hierarchical-Type Control System Based on Smart MBD Approach and its Application to Hydraulic Excavator", Journal of Robotics and Mechatronics, Vol. 36, No. 4, pp. 909-917, 2024), etc., can be designed using known methods. In this embodiment, by applying the design method used in Generalized Minimum Variance Control (GMVC), which is a control system design method based on minimizing evaluation criteria, to the design of the model error suppression compensator 13, a model error suppression compensation rule is derived. Generalized Minimum Variance Control is known to be effective for non-minimum phase systems having dead time and unstable zeros, and is considered to be applicable to a wide range of control targets as a model error suppression compensator.

[0024] The identification unit 14 determines whether the compensation input u, which is the output of the model error suppression compensator 13 d is less than or equal to a predetermined threshold value. When the compensation input u d exceeds the threshold value, it is considered that the model error, which is the difference between the control characteristics of the control target 20 and the control characteristics of the nominal model 12, is large, that is, the characteristics of the control target 20 have changed significantly since the start of control.

[0025] When the compensation input u d exceeds the threshold value, the identification unit 14 identifies (re-identifies) the control target 20 and updates the nominal model 12. Thereby, it is possible to improve the control performance by performing re-identification at an appropriate timing while suppressing an increase in the calculation load.

[0026] Furthermore, the identification unit 14 modifies the controller 11, which is a model predictive controller, based on the updated nominal model 12. Specifically, it applies the updated nominal model 12 to the internal model of the controller 11. As a result, the controller 11 can perform model predictive control based on the updated nominal model 12, and thus can provide a control input u corresponding to the characteristics of the controlled object 20 at that time. c It can output the following. Therefore, the excessive compensation input u caused by the increase in model error d This makes it possible to suppress the occurrence of malfunctions in the controlled object 20.

[0027] Furthermore, the identification unit 14 receives the compensation input u d When it is detected that the threshold is exceeded, the control target 20 is re-identified and the nominal model 12 is modified, the compensation input u d Re-identification may be withheld until the value falls below a threshold or a predetermined time has elapsed. This avoids unnecessary re-identification processing, reduces the computational load, and minimizes the impact of identification errors.

[0028] In the control system described above, the controlled object 20 controlled by the controller 11 includes the actuator 21, and the nominal model 12 represents the control characteristics of the entire controlled object 20, but is not limited to this. In cases where the controlled object is large in scale and has a complex structure, the control device 1, which includes the nominal model 12 and a model error suppression compensator 13, may be applied to a part of the controlled object 20, for example, the actuator 21. This makes it possible to apply the control device 1 to parts where system fluctuations are expected in large-scale control systems, thereby simplifying the configuration of the control system. Furthermore, it is possible to reduce the computational load related to re-identification and modification of the nominal model 12.

[0029] The model error suppression compensator 13, the controller 11 that performs model predictive control, and the identification unit 14 according to this embodiment will be described in detail below.

[0030] (Model error suppression compensator) Figure 2 shows the basic structure of the model error suppression compensator, and is a structure that excludes the controller 11 and identification unit 14 from the control system according to this embodiment shown in Figure 1. The model error suppression compensator 13 receives input u c This compensator aims to bring the input / output characteristics from to output y closer to the input / output characteristics of the nominal model 12. As described above, the design method of the model error suppression compensator 13 is not particularly limited. The model error suppression compensator 13 according to this embodiment is designed based on generalized minimum variance control, which is expected to be applicable to a wide range of controlled objects.

[0031] Figure 3 is a block diagram of the control system of the model error suppression compensator 13 based on generalized minimum variance control. The controlled object 20 is a CARIMA (Controlled Auto-Regressive Integrated Moving Average) model described by the following equations (1) and (2), and A(z -1 ), B(z -1 Let the degrees of the terms be n and m, respectively.

number

[0032] F(z -1 ) is derived from the Diophantine equations shown in the following equations (3) and (4).

number

[0033] In this embodiment, since we treat the dead time as 0, E(z -1 )=1, and F(z -1 The degree of ) is A(z -1 It is set to the same order as ). Also, G(z -1 ) is defined by the following equation (5).

number

[0034] The evaluation standard J1(k) is expressed by the following formulas (6) and (7).

number

[0035] Based on minimizing the evaluation standard J1(k), the model error suppression compensation law is expressed by the following equations (8) and (9).

number

[0036] In equation (8) above, λ is a weighting coefficient for the change in the compensation input. By adjusting the weighting coefficient λ as appropriate, the user can adjust the gain of the model error suppression compensator 13 to stabilize the control system.

[0037] (Model Predictive Controller) In the model predictive control of the controller 11 according to this embodiment, the evaluation function J2(k), expressed by the following equation, is minimized.

number

[0038] Model predictive control has features such as easy extension to multi-input / output systems and the ability to explicitly consider constraints. In model predictive control, constraint equations are designed for the control input u, the input change Δu, and the output y, and control is performed by solving the minimization problem of the evaluation function J2(k) based on these constraint equations.

[0039] (Identification section) The control device 1 according to this embodiment includes an identification unit 14 as a mechanism for evaluating the performance (control performance) of the control system (Figure 1). The evaluation index for control performance is the compensation input u of the model error suppression compensator 13. d Use the compensation input u. d As shown in Figure 2, this can be expressed by the following equation (15), and is considered to be an index of the magnitude of the model error.

number

[0040] Compensation input u d By performing system identification when the threshold is exceeded, unnecessary identification can be avoided, and the risk of degrading control performance due to identification errors can be reduced. In addition, by setting the model error suppression compensator 13, the effect of model errors can be suppressed even if identification errors exist, and the control performance of the control system can be maintained. Furthermore, the compensation input u d Since the control target 20 is re-identified using this as an evaluation index, malfunctions such as failures of the control target 20 caused by excessive control input u can be suppressed.

[0041] (Numerical example) The following describes an example of the application of the control device 1 according to this embodiment. This example shows an application to a control system having redundant inputs of 2 inputs and 1 output.

[0042] As shown in Figure 4, in this example, we consider a system in which the controlled object 20 includes two motors M1 and M2, which are actuators 21, and the angular velocity of the rotating body 22 is controlled by the operation of the two motors M1 and M2. r is the target value of the angular velocity of the rotating body, V MPC1 ,VMPC2 τ1 and τ2 are the manipulated values ​​of the voltage applied to motors M1 and M2, respectively, calculated by controller 11, which is a model predictive controller; τ1 and τ2 are the torques output from motors M1 and M2, respectively; and ω is the angular velocity of the rotating body 22.

[0043] Furthermore, the control system in this example includes subsystems S1 and S2 (Figure 5) for each circuit section of motors M1 and M2, each having a nominal model 12 and a model error suppression compensator 13. Figure 5 is a block diagram of subsystem S1 related to motor M1, but subsystem S2 related to motor M2 has the same configuration as subsystem S1. The actual voltages V1 and V2 applied to motors M1 and M2 are the output voltage V of the controller 11. MPC1 ,V MPC2 And the voltage V is the output of the model error suppression compensator 13. MEC1 ,V MEC2 The sum of (V1=V MPC1 +V MEC1 ,V2=V MPC2 +V MEC2 ) is. Also, the current i is in the motor circuit section M 1E This is the output current.

[0044] The mathematical models of motors M1 and M2, and the mathematical model of the rotating body 22, are expressed by the following differential equations (16) and (17), respectively.

number

[0045] The system parameters for each mathematical model in equations (16) and (17) above are shown in Table 1 below. [Table 1]

[0046] In this example, verification was performed using a simulation with MATLAB® R2024a (MathWorks). The Model Predictive Control Toolbox version 24.1 add-on was used for the controller 11. The system's default settings were used for other configuration conditions. The target values ​​were a rotating body angular velocity of 12.57 rad / s, current values ​​of 0.0105 A for motors M1 and M2, and a 5-second periodic square wave, which were used as input to the Model Predictive Control. Furthermore, the armature resistance of motor M1 was maintained at 5.7 Ω according to the nominal model 12, while the armature resistance of motor M2 changed to 11 Ω 5 seconds after the start of control. This simulates a case where the control characteristics of actuator 21, motor M2, change over time (system fluctuation), resulting in a model error compared to the nominal model 12.

[0047] Furthermore, the identification unit 14 receives the compensation input u of the model error suppression compensator 13. d When the value exceeds 0.5, a system fluctuation is detected, and the armature resistance R, which is the parameter causing the fluctuation, is re-identified. In this example, the armature resistance R is identified based on the successive least squares method. Specifically, with respect to the identification parameter, the armature resistance R, equation (16) above is approximated as equation (18) below.

number

[0048] In this example's control system, when a system fluctuation is detected, the estimated armature resistance R^ is identified using the weighted least squares method shown in equation (19) below, with data vectors Φ and Y and a weight matrix W constructed from data going back N steps from the current time step.

number

[0049] Furthermore, assuming the energy load on the motor was the magnitude of the power consumption in the armature resistance, the design was constrained to be 1.425W or less. The model error suppression compensator 13 was designed based on Figure 3, with λ = 0.1. The sampling time was set to 0.01 seconds. Numerical simulations were performed under these conditions.

[0050] For comparison, Figures 6 and 7 show the simulation results when the system is equipped with a model error suppression compensator 13 but lacks an identification unit 14 and does not perform system re-identification. Figure 6 shows the manipulated variables V1 and V2 and the controlled variable ω, and Figure 7 shows the power consumption W1 and W2 generated by the armature resistances of motors M1 and M2, respectively. As shown in Figure 6, even when system fluctuations occurred, the model error suppression compensator 13 functioned to maintain control performance equivalent to that when there were no system fluctuations. However, in the comparative example control system, the only objective was to maintain control performance without detecting the occurrence of system fluctuations, so as shown in Figure 7, the power consumption W2 increased after the system fluctuations.

[0051] Next, Figures 8 and 9 show the simulation results using the control system according to this embodiment. In this example, by re-identifying the system after detecting the system fluctuation and updating the nominal model, the increase in the power consumption W2 of the fluctuating system can be suppressed. Furthermore, to make the effectiveness of the model error suppression compensator 13 easier to understand, Table 2 below shows the average error and variance values ​​of the control variable for each step with and without the model error suppression compensator 13 (MEC) when the output value includes Gaussian white noise with a mean of 0 and a variance of 0.111.

[0052] [Table 2] As shown in Table 2, both the control error and variance were smaller when the model error suppression compensator 13 was present. Therefore, it can be seen that the control device 1 including the model error suppression compensator 13 and the identification unit 14 according to this embodiment is effective in maintaining control performance.

[0053] As described above, the control device 1 according to this embodiment is equipped with a model error suppression compensator 13, which makes it possible to reduce the influence of errors between the controlled object 20 and the nominal model 12. In addition, the control device 1 is equipped with an identification unit 14, which compensates the input u based on the model error. d Since the controlled object is re-identified when the value increases, it is possible to suppress the increase in computational load, suppress the occurrence of malfunctions in the controlled object, and maintain good control performance.

[0054] Furthermore, as described above, the nominal model 12 of the control device 1 according to the present invention is not limited to representing the control characteristics of the entire controlled object 20. Specifically, the control device 1, which includes the nominal model 12 and a model error suppression compensator 13, may be applied to a part of the controlled object 20, for example, an actuator 21 that controls the operation of a machine. This makes it possible to apply the control device 1 to a part of the large and complex controlled object 20, such as an actuator 21, where characteristic fluctuations are a particular concern, thereby simplifying the configuration of the control system. Moreover, even with a simple control system configuration, it is possible to reduce the effects of model errors and suppress the occurrence of malfunctions in the controlled object through timely re-identification, thereby maintaining good control performance.

[0055] Alternatively, the controlled object 20 may be configured as a redundant input system including multiple actuators 21, and each actuator 21 may be provided with a nominal model 12 and a model error suppression compensator 13. In this case, the identification unit 14 receives the compensation input u d For actuators 21 whose threshold value exceeds the limit, re-identification is performed, the nominal model 12 is modified, and the modified nominal model 12 is applied to the internal model of the controller 11, which is a model predictive controller. As shown in the numerical example above, for multiple actuators 21, the compensation input ud Since the occurrence of characteristic fluctuations can be monitored based on this, it is possible to prevent excessive manipulated inputs (control inputs) from being input to the controlled object 20, including the actuator 21, and to maintain good control performance.

[0056] The control device 1 according to the above embodiment can be implemented using a normal computer system. For example, a computer program for executing the processing of the control device 1 according to the above embodiment can be distributed via a network, and by installing the computer program on a computer, the computer device can be made to function as the control device 1. [Industrial applicability]

[0057] The present invention is suitable for model predictive control of a controlled object having model errors. In particular, it is suitable for controlling a redundant input system in which the controlled object includes multiple actuators. [Explanation of Symbols]

[0058] 1 Control device, 11 Controller, 12 Nominal model, 13 Model error suppression compensator, 14 Identification unit, 20 Controlled object, 21 Actuator, 22 Rotating body, S1, S2 Subsystems

Claims

1. A nominal model that represents the characteristics of the controlled object, A controller that performs model predictive control based on the aforementioned nominal model, A model error suppression compensator outputs a compensation input for adjusting the control input to the controlled object based on the difference between the output of the controlled object and the output of the nominal model, The system includes an identification unit that, when the compensation input exceeds a predetermined threshold, identifies the controlled object, modifies the nominal model, and applies the modified nominal model to the internal model of the controller. A control device characterized by the following features.

2. Among the controlled objects, which include actuators that control the operation of a machine, a nominal model representing the characteristics of the actuators, A controller that performs model predictive control based on the aforementioned nominal model, A model error suppression compensator outputs a compensation input for adjusting the control input to the actuator based on the difference between the output of the actuator and the output of the nominal model, The system includes an identification unit that, when the compensation input exceeds a predetermined threshold, identifies the actuator, modifies the nominal model, and applies the modified nominal model to the internal model of the controller. A control device characterized by the following features.

3. The controlled object is a redundant input system including a plurality of actuators, Each actuator is provided with the nominal model and the model error suppression compensator. The identification unit identifies the actuator for which the compensation input exceeds the threshold, modifies the nominal model, and applies the modified nominal model to the internal model of the controller. The control device according to claim 2.

4. The identification unit, after performing identification and modifying the nominal model, refrains from performing identification until the compensation input falls below the threshold. The control device according to any one of claims 1 to 3.

5. Computers, A controller that performs model predictive control based on a nominal model that represents the characteristics of the controlled object. A model error suppression compensator that outputs a compensation input for adjusting the control input to the controlled object based on the difference between the output of the controlled object and the output of the nominal model. An identification unit that, when the compensation input exceeds a predetermined threshold, identifies the controlled object, modifies the nominal model, and applies the modified nominal model to the internal model of the controller. A program that operates as such.

Images