Control device, control method, and program

Abstract

A control device (1) determines an operation amount for controlling a control target (2), and the control device (1) is configured to have an operation amount determination section (12) that determines the operation amount based on a function that evaluates a difference between a predicted value of a control amount of the control target (2) after the control by the operation amount and a target value of the control amount.

Description

Technical Field

[0001] This invention relates to control devices, control methods, and procedures. Background Technology

[0002] There exists a control device that determines the operational quantities used to control the controlled object.

[0003] Conventional control devices exist that perform model predictive control based on the operational quantity used to control the controlled object and the control quantity of the controlled object, thereby determining the operational quantity for the next control cycle based on the prediction of the control quantity. However, to improve the prediction accuracy of the control quantity, the control device needs to set a long control cycle. However, with a long control cycle, the control accuracy deteriorates when changes occur in the motion characteristics of the controlled object between control timing intervals.

[0004] As a control device that can suppress the degradation of control accuracy even when the control cycle of model predictive control is set to a long control cycle, for example, Patent Document 1 discloses the following control device, which includes: a main control unit that performs model predictive control based on the operation quantity for controlling the controlled object and the control quantity of the controlled object in units of a first control cycle, thereby predicting the operation quantity (hereinafter referred to as "the first operation quantity") in the next control cycle; and a high-speed compensation control unit that performs linear predictive control based on the first operation quantity and the target value in units of a second control cycle with a control cycle shorter than the first control cycle, thereby predicting the operation quantity (hereinafter referred to as "the second operation quantity") in the next control cycle, and calculating the operation quantity for controlling the controlled object based on the first operation quantity and the second operation quantity.

[0005] Patent Document 1: Japanese Patent No. 7283646 Summary of the Invention

[0006] As shown above, a control device that helps to calculate the operation quantity at high speed has been disclosed, but there is a problem that the operation quantity cannot be determined stably sometimes, depending on the method of calculating the operation quantity.

[0007] The present invention was proposed to solve the above-mentioned problems, and its purpose is to obtain a control device that helps to stably determine the operating quantity.

[0008] The control device of the present invention determines the operational quantity for controlling a controlled object. The control device includes an operational quantity determination unit that determines the operational quantity based on a function that evaluates the difference between a predicted value of the control quantity for the controlled object after control is performed via the operational quantity and a target value of the control quantity.

[0009] The effects of the invention

[0010] According to the present invention, it is helpful to stably determine the amount of operation. Attached Figure Description

[0011] Figure 1 This is a structural diagram showing the control device 1 involved in Embodiment 1.

[0012] Figure 2 This is a hardware structure diagram showing the hardware of the control device 1 involved in Embodiment 1.

[0013] Figure 3 This is a hardware structure diagram of a computer when the control device 1 is implemented through software or firmware.

[0014] Figure 4 This is a flowchart representing the processing flow, or control method, of control device 1.

[0015] Figure 5 This is an illustrative diagram representing an example of the cost function F when using the bisection method.

[0016] Figure 6 This is an illustrative diagram representing an example of the cost function F when using the steepest descent method.

[0017] Figure 7 This is a structural diagram showing the control device 1 involved in Embodiment 2.

[0018] Figure 8 This is a hardware structure diagram showing the hardware of the control device 1 involved in Embodiment 2.

[0019] Figure 9 This is a flowchart representing the processing flow, or control method, of control device 1.

[0020] Figure 10 This is an explanatory diagram showing an example when the number of steps n set by the step setting unit 13 is 1.

[0021] Figure 11 This is an explanatory diagram showing an example when the number of steps n is set to 2 by the step setting unit 13.

[0022] Figure 12 This is a structural diagram of the control device 1 involved in Embodiment 3.

[0023] Figure 13 This is a hardware structure diagram showing the hardware of the control device 1 involved in Embodiment 3.

[0024] Figure 14 This is an explanatory diagram showing the relationship between the number of steps n and the overshoot G(n).

[0025] Figure 15 This is a structural diagram of the control device 1 involved in embodiment 4.

[0026] Figure 16 This is a hardware structure diagram showing the hardware of the control device 1 involved in Embodiment 4.

[0027] Figure 17 This is a structural diagram showing the control device 1 involved in embodiment 5.

[0028] Figure 18 This is a hardware structure diagram showing the hardware of the control device 1 involved in Embodiment 5.

[0029] Figure 19 This is a flowchart representing the processing flow, or control method, of control device 1.

[0030] Figure 20 This is an explanatory diagram showing the relationship between the number of steps n and the overshoot G(n).

[0031] Figure 21 This is an illustration of the bisection method for efficiently finding the optimal value of n. Detailed Implementation

[0032] Hereinafter, in order to illustrate the present invention in more detail, the embodiments for carrying out the present invention will be described with reference to the accompanying drawings.

[0033] Implementation Method 1

[0034] Figure 1 This is a structural diagram showing the control device 1 involved in Embodiment 1.

[0035] Figure 2 This is a hardware structure diagram showing the hardware of the control device 1 involved in Embodiment 1.

[0036] Figure 1 The details of the control device 1 shown will be described later. A function that evaluates the difference between the predicted value and the target value of the control quantity is disclosed. As a non-limiting example, a bisection method is used to determine the structure of the operand. Furthermore, while a bisection method is used in the following explanation, it is not a limitation; algorithms such as genetic algorithms or particle swarm optimization can also be used to determine the operand. That is, any algorithm can be used, as long as it explores solutions existing between positive and negative values ​​by using the function that evaluates the difference between the predicted value and the target value of the control quantity as shown below.

[0037] Figure 1 The control device 1 shown has a prediction unit 11 and an operation quantity determination unit 12.

[0038] Control device 1 determines the operational quantity u used to control the controlled object 2. k , will operate on u k Output to control object 2. k is an integer greater than or equal to 1, which is an identifier representing the k-th control action taken on control object 2.

[0039] The controlled object 2 is, for example, a device controlled by the control device 1, or a system controlled by the control device 1.

[0040] For example, prediction unit 11 is composed of Figure 2 The prediction circuit 31 shown is implemented.

[0041] The prediction unit 11, based on model 11a of the controlled object 2, controls the control quantity y of the controlled object 2. k＋1 Make predictions. Control quantity y k＋1 It is the control result of controlled object 2.

[0042] Specifically, the prediction unit 11 obtains the operation quantity u from the operation quantity determination unit 12. k Then, the prediction unit 11 will operate on the amount u. k Assign the control quantity y to model 11a, and obtain the control quantity y of the controlled object 2 from model 11a. k＋1 The predicted value y k＋1 '.

[0043] Prediction unit 11 will control quantity y k＋1 The predicted value y k＋1 Output to the operation quantity determination unit 12.

[0044] Model 11a mathematically describes the operand u assigned to the controlled object 2. k and control quantity y k＋1 The predicted value y k＋1 The relationship between '. Or, model 11a learns the operand u assigned to the controlled object 2 through learning. k and control quantity y k＋1 The predicted value y k＋1 The relationship between '.

[0045] If model 11a is assigned an operation amount u by the prediction unit 11 k Then the control quantity y of the controlled object 2 will be... k＋1 The predicted value y k＋1 Output to prediction section 11.

[0046] Operation quantity determination unit 12, for example, by Figure 2 The operation quantity is determined by circuit 32.

[0047] Operation quantity determination unit 12 obtains control quantity y from prediction unit 11. k＋1 The predicted value y k＋1For example, obtaining control variables y from external sources. k Target value y d .

[0048] Operation quantity determination unit 12 is based on the operation quantity u k The control quantity y of the controlled object 2 after execution of control k＋1 The predicted value y k＋1 'and target value y d The function that evaluates the difference determines the amount of operation u. k .

[0049] Specifically, the operation quantity determination unit 12 is based on the predicted value y k＋1 'and target value y d The amount of operation u is determined by the cost function F related to the difference. k The determination of the operation quantity uk is, for example, done using a dichotomy.

[0050] Operation quantity determination unit 12 will determine the operation quantity u k The outputs are sent to the prediction unit 11 and the control object 2, respectively.

[0051] exist Figure 1 In this design, the structural elements of the control device 1, namely the prediction unit 11 and the operation quantity determination unit 12, are each independent. Figure 2 The case shown is a dedicated hardware implementation. That is, it is envisioned that the control device 1 is implemented by the prediction circuit 31 and the operation quantity determination circuit 32.

[0052] The prediction circuit 31 and the operation determination circuit 32 are each, for example, a single circuit, a composite circuit, a processor that performs processing based on instructions generated by a program, a processor that performs processing based on instructions generated by a parallel program, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.

[0053] The structural elements of the control device 1 are not limited to being implemented by dedicated hardware; the control device 1 can also be implemented by software, firmware, or a combination of software and firmware.

[0054] Software or firmware is stored as a program in a computer's memory. A computer refers to the hardware that executes programs, such as a CPU (Central Processing Unit), GPU (Graphics Processing Unit), central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, or DSP (Digital Signal Processor).

[0055] Figure 3 This is a hardware structure diagram of a computer when the control device 1 is implemented by software or firmware.

[0056] When the control device 1 is implemented by software or firmware, the program for causing the computer to execute the processing flow of the prediction unit 11 and the operation quantity determination unit 12 is stored in the memory 51. Furthermore, the computer's processor 52 executes the program stored in the memory 51.

[0057] In addition, Figure 2 The diagram shows an example of the dedicated hardware implementation for each structural element of the control device 1. Figure 3 The diagram shows an example of control device 1 implemented by software or firmware. However, this is just one example; it is also possible that some structural elements of control device 1 are implemented by dedicated hardware, while the remaining structural elements are implemented by software or firmware.

[0058] Next, for Figure 1 The operation of the control device 1 shown will be explained.

[0059] Figure 4 This is a flowchart representing the processing flow, or control method, of control device 1.

[0060] First, the prediction unit 11 obtains the operation amount u from the operation amount determination unit 12. k The amount of operation u at that moment k Operations that have not yet been optimized by the operation quantity determination unit 12 are not assigned to the control object 2.

[0061] Forecasting Department 11 will operate the amount u k Assigned to model 11a.

[0062] If model 11a is assigned an operation amount u by the prediction unit 11 k Then the control quantity y of the controlled object 2 will be... k＋1 The predicted value y k＋1 The output is sent to the prediction unit 11. The prediction processing of model 11a itself is a well-known technique, so detailed explanation is omitted.

[0063] Prediction unit 11 obtains the control quantity y of the controlled object 2 from model 11a. k＋1 The predicted value y k＋1 '( Figure 4 Step ST1).

[0064] Predicted value y k＋1 'is through the operation quantity u k The control quantity y of the controlled object 2 when control was performed k＋1 The predicted value. Below, sometimes the operation quantity is u. k The predicted value y at that timek＋1 'Written as "y k＋1 (u) k )'".

[0065] Prediction unit 11 will control quantity y k＋1 The predicted value y k＋1 Output to the operation quantity determination unit 12.

[0066] Operation quantity determination unit 12 obtains control quantity y from prediction unit 11. k＋1 The predicted value y k＋1 For example, obtaining control variables y from external sources. k Target value y d .

[0067] The operation quantity determination unit 12 is set as shown in equation (1) below, and the predicted value y is... k＋1 (u) k )' and target value y d The cost function F(the difference) Figure 4 Step ST2).

[0068]

[0069] Figure 5 This is an illustrative diagram representing an example of the cost function F when using the bisection method. Here, consider the case where friction with the ground exists, in order to increase the position y of the object... k Move to target position y d Then it comes to rest, and a force u, which is the operational quantity, is applied. k .

[0070] exist Figure 5 In the middle, the horizontal axis shows the operating quantity u. k The vertical axis represents the cost function F.

[0071] exist Figure 5 The example illustrates the effect of static friction at the operating amount u k When the static friction is less than or equal to the static friction value, the controlled object 2 does not move. When the controlled object 2 does not move, the cost function F remains unchanged. Figure 5 In the example, the number of operations u k The larger the value of u, the smaller the cost function F. If the number of operations u... k If the value increases to a certain extent, the cost function F becomes negative.

[0072] Operation determination unit 12 uses a bisection method to determine the operation u that makes the cost function F equal to 0. k ( Figure 4 Step ST3).

[0073] The bisection method is an iterative method that solves an equation by repeatedly finding the midpoint of the interval containing the solution. It involves calculating the operation u to determine F=0 through iterative computation. k An algorithm for approximate solutions.

[0074] Specifically, the bisection method calculates the operation quantity u for F=0 through the following four steps (1) to (4). k An algorithm for approximate solutions.

[0075] Step (1)

[0076] In the cost function F, let F(u) be set. k1 The symbol of ) and F(u) k2 The lower limit value u of the interval with different signs) k1 and the upper limit of the interval u k2 .

[0077] Step (2)

[0078] For the lower limit value u k1 and upper limit value u k2 The median value u kM Perform the calculation.

[0079] Step (3)

[0080] If F(u) kM The symbol of ) and F(u) k1 If the signs of the values ​​are the same, then the lower limit value u will be set to... k1 Change to u kM If F(u) kM The symbol of ) and F(u) k2 If the signs of the values ​​are the same, then the upper limit value u will be... k2 Change to u kM .

[0081] Step (4)

[0082] Returning to step (2), the operation quantity u for F=0 is obtained by repeating the process. k An approximate solution.

[0083] exist Figure 1 In the control device 1 shown, the operation quantity determination unit 12 uses a binary search method to determine the operation quantity u that makes the cost function F become 0. k .

[0084] u is the operation that determines whether the cost function F becomes 0. k Methods, such as Figure 6 As shown, there exists a method that uses the steepest descent method.

[0085] The cost function in this case is, for example, via F = (yd -y k＋1 (u) k )') 2 express.

[0086] In operation quantity u k Within the range of less than or equal to static friction (i.e., less than or equal to maximum friction), even if the operating amount u is changed... k The object does not move, therefore y is the control variable. k＋1 (u) k )' remains unchanged. In y, which is the control variable... k＋1 (u) k Within the invariant range, the value of the cost function F is constant, and the cost function F has no gradient. Therefore, in the operation u... k When the solution is less than or equal to static friction, the steepest descent method based on gradient exploration cannot determine whether to update in the direction of increasing the solution or in the direction of decreasing the solution. As a result, sometimes the solution is not updated and the calculation becomes unstable.

[0087] Figure 6 This is an illustrative diagram representing an example of the cost function F when using the steepest descent method.

[0088] exist Figure 6 In the middle, the horizontal axis shows the operating quantity u. k The vertical axis represents the cost function F.

[0089] In contrast, when using Figure 5 The cost function F, as shown, determines the number of operations u that makes the cost function F equal to 0. k In such cases, the solution can be updated based on whether the cost function F is positive or negative. Therefore, for example, in the control of a force applied to move an object placed on the ground, even at the level of the operational quantity u... k Even within a range less than or equal to static friction, the operational quantity u that makes the cost function F equal to 0 can be stably calculated. k .

[0090] Furthermore, the operation u that determines the value of the cost function F to be 0 k In this case, the cost function is, for example, through F = y d -y k＋1 (u) k )'express.

[0091] Furthermore, the operation u that determines the value of the cost function F to be 0. k In this case, the operation quantity determination unit 12 can use the above-described dichotomy method. For example, the operation quantity determination unit 12 uses the dichotomy method to determine the operation quantity u. kIn this case, compared to other optimization methods that use gradients that do not refer to the cost function F, such as genetic algorithms or particle swarm optimization, the amount of operation u is determined. k Compared to the situation where the amount of operation u is typically used to determine the amount of operation u k The number of calculations is reduced, and the calculation time is shortened.

[0092] exist Figure 1 In the control device 1 shown, the operation quantity determination unit 12 uses a cost function F as the basis for determining the predicted value y. k＋1 'and target value y d The function used by the operation quantity determination unit 12 to evaluate the difference is not limited to the cost function F; the operation quantity determination unit 12 may also use, for example, a loss function as the function to evaluate the difference.

[0093] Operation quantity determination unit 12 will determine the operation quantity u k Output to control object 2 ( Figure 4 Step ST4).

[0094] If the controlled object 2 is given an operational quantity u by the controlled device 1 k Then through the operation quantity u k It is under control.

[0095] Therefore, the control variable of controlled object 2 becomes y. k .

[0096] In Embodiment 1 described above, the control device 1 is configured with an operation amount determination unit 12. This operation amount determination unit 12 determines the operation amount for controlling the controlled object 2. The operation amount is determined based on a function that evaluates the difference between the predicted value of the control amount of the controlled object 2 after control is performed by the operation amount and the target value of the control amount. Therefore, since the operation amount is determined based on the function that evaluates the difference between the predicted value and the target value of the control amount, the control device 1 updates the solution, i.e., the operation amount, according to whether the value of the function is positive or negative. For example, it can stably perform calculations even in regions where the cost function, such as static friction, does not produce a gradient, even if the operation amount does not exceed the static friction force.

[0097] In Embodiment 1, the control device 1 is configured such that the operation amount determination unit 12 determines the operation amount based on a cost function related to the difference, which is the function used to evaluate the difference. Therefore, since the operation amount is determined based on the cost function related to the difference, the solution is updated according to whether the function value is positive or negative. Thus, the control device 1 can perform calculations stably even in regions where, for example, the operation amount does not exceed the static friction force, the cost function does not produce a gradient.

[0098] In Embodiment 1, the control device 1 is configured such that the operation amount determination unit 12 uses a bisection method to determine the operation amount. Therefore, even in regions where the cost function, such as the static friction force, does not produce a gradient, the calculation of the control device 1 does not become unstable, and the operation amount can be determined. Furthermore, by using the bisection method, the convergence of the solution can generally be improved compared to genetic algorithms or particle swarm optimization.

[0099] In Embodiment 1, the control device 1 is configured such that the operation amount determination unit 12 determines the operation amount using a genetic algorithm or particle swarm optimization, etc. Therefore, even in regions where the cost function, such as static friction, does not produce a gradient, the control device 1 can determine the operation amount without becoming unstable during computation. Furthermore, by using these methods, solutions can generally be explored more stably compared to the bisection method.

[0100] In Embodiment 1, the control device 1 is configured with a prediction unit 11, which predicts a control quantity based on a model 11a of the controlled object 2 and outputs the predicted value of the control quantity to the operation quantity determination unit 12. Therefore, the control device 1 can predict the control quantity with minimal computation. Furthermore, since the prediction unit 11 incorporates the model 11a, the input-output between the prediction unit 11 and the model 11a can be completed in a short time, resulting in high-speed prediction of the control quantity.

[0101] exist Figure 1 In the control device 1 shown, the prediction unit 11 has a built-in model 11a. However, this is just an example, and the model 11a can also be disposed outside the prediction unit 11.

[0102] In the case where the external model 11a is set up, for example in the cloud, the prediction unit 11 will obtain the operation amount u from the operation amount determination unit 12. k For example, model 11a is sent to the cloud via communication lines such as the internet. If the model 11a in the cloud is assigned an operational variable u by the prediction unit 11... k Then the control quantity y of the controlled object 2 will be... k＋1 The predicted value y k＋1 For example, it is sent to the prediction unit 11 via a communication line.

[0103] Prediction unit 11 obtains the control quantity y of controlled object 2 from model 11a in the cloud. k＋1 The predicted value y k＋1 Then, the prediction unit 11 will use the obtained control quantity y k＋1 The predicted value y k＋1 The output is sent to the operation quantity determination unit 12, whereby the operation quantity determination unit 12 can use the predicted value y obtained from the prediction unit 11. k＋1 'And that determines the amount of work to be done.'

[0104] Implementation Method 2

[0105] In Embodiment 2, the control device 1 will be described. In this control device 1, the prediction unit 14 predicts the control quantity when multiple controls are performed on the controlled object 2 and uses it as the control quantity of the controlled object 2.

[0106] Figure 7 This is a structural diagram showing the control device 1 according to Embodiment 2. Figure 7 In, with Figure 1 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0107] Figure 8 This is a hardware structure diagram illustrating the hardware of the control device 1 according to Embodiment 2. Figure 8 In, with Figure 2 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0108] Figure 7 The control device 1 shown has a step setting unit 13, a prediction unit 14, and an operation amount determination unit 12.

[0109] The step setting unit 13 sets the number of times the control of the controlled object 2 is executed, i.e., the number of steps n, based on the control result of the model 15 when the model 15 is controlled.

[0110] The step setting unit 13 outputs the step number n to the prediction unit 14.

[0111] The prediction unit 14 predicts the control quantity when multiple controls are performed on the controlled object 2 and uses it as the control quantity of the controlled object 2. That is, the prediction unit 14 predicts the control quantity at the time after multiple controls are performed on the controlled object 2 from the current time and uses it as the control quantity of the controlled object 2.

[0112] Specifically, the prediction unit 14, based on the model 15 of the controlled object 2, determines the control quantity y for the time elapsed after the number of steps n calculated by the step setting unit 13 from the current time. k＋n It is used as a control variable for prediction. Furthermore, when the start time of control is set to 0, the control variable y... k＋n It is the control quantity at time (k+n)×Δt, where Δt represents the control period.

[0113] Prediction unit 14 will control quantity y k＋n The predicted value y k＋n Output to the operation quantity determination unit 12.

[0114] Model 15 mathematically describes the operand u assigned to the controlled object 2. kThe control quantity y at the time n steps from the current time. k＋n The predicted value y k＋n The relationship between '. Or, Model 15 learns and masters the operands u assigned to the controlled object 2. k The control quantity y at the time n steps from the current time. k＋n The predicted value y k＋n The relationship between '.

[0115] If model 15 is assigned an operation amount u by the prediction unit 14 k Given the number of steps n, the control quantity y of the controlled object 2 will be... k＋n The predicted value y k＋n Output to prediction section 14.

[0116] Step setting unit 13, for example, is made by Figure 8 The step setting circuit 33 shown is implemented. The prediction unit 14 is, for example, implemented by... Figure 8 The prediction circuit 34 shown is implemented.

[0117] exist Figure 7 In the design, the structural elements of the control device 1, namely the step setting unit 13, the prediction unit 14, and the operation quantity determination unit 12, are each independently configured. Figure 8 The case shown is a dedicated hardware implementation. That is, it is envisioned that the control device 1 is implemented by a step setting circuit 33, a prediction circuit 34, and an operation quantity determination circuit 32.

[0118] The step setting circuit 33, the prediction circuit 34, and the operation quantity determination circuit 32 are each, for example, a single circuit, a composite circuit, a processor that performs processing based on instructions generated by a program, a processor that performs processing based on instructions generated by a parallel program, an ASIC, an FPGA, or a combination thereof.

[0119] The structural elements of the control device 1 are not limited to being implemented by dedicated hardware; the control device 1 can also be implemented by software, firmware, or a combination of software and firmware.

[0120] When the control device 1 is implemented by software or firmware, the program for causing the computer to execute the processing flow of the step setting unit 13, the prediction unit 14, and the operation amount determination unit 12 is stored in Figure 3 The memory 51 shown. Furthermore, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0121] In addition, Figure 8 The diagram shows an example of the dedicated hardware implementation for each structural element of the control device 1. Figure 3The diagram shows an example of control device 1 implemented by software or firmware. However, this is just one example; it is also possible that some structural elements of control device 1 are implemented by dedicated hardware, while the remaining structural elements are implemented by software or firmware.

[0122] Next, for Figure 7 The operation of the control device 1 shown will be explained.

[0123] Figure 9 This is a flowchart representing the processing flow, or control method, of control device 1.

[0124] Step setting unit 13, for example, obtains the control quantity y from an external source. k Target value y d .

[0125] Step setting unit 13 obtains operation quantity u from operation quantity determination unit 12 k The amount of operation u at that moment k Operations that have not yet been optimized by the operation quantity determination unit 12 are not assigned to the control object 2.

[0126] The step setting unit 13 sets the number of steps n when the overshoot is approximately 0 as the optimal number of steps. Figure 9 Step ST11).

[0127] The following is an example of the step number n setting process implemented by the step number setting unit 13.

[0128] First, the step setting unit 13 sets an arbitrary number of steps n.

[0129] Step setting unit 13 will set the operation amount u k Assign an arbitrary number of steps n to model 15, and obtain the control quantity y from model 15 after any number of steps n from the current time. k＋n The predicted value y k＋n '.

[0130] Step setting unit 13 for predicted value y k＋n 'and target value y d The difference is the overshoot G. k＋n Perform the calculation.

[0131] The overshoot G implemented by the step setting unit 13 k＋n The calculation methods include, for example, the following two methods (1) and (2).

[0132] (1) To actually control the controlled object and control the overshoot G k＋n Methods of measurement

[0133] (2) Virtually control the model and control the overshoot Gk＋n Methods of measurement

[0134] Furthermore, in the case of method (1), the actual control of the controlled object is performed while changing the number of steps, so in addition to the possibility of becoming an unstable state, it sometimes requires a lot of time. Therefore, method (2) is mostly used.

[0135] Step setting unit 13 will set the overshoot G k＋n Compare with the threshold Th, if the overshoot G k＋n If the number of steps n is less than or equal to the threshold Th, then any number of steps n is set as the optimal number of steps. The threshold Th can be stored in the internal memory of the step setting unit 13, or it can be assigned from an external source.

[0136] If the overshoot G k＋n If the number of steps exceeds the threshold Th, the step setting unit 13 changes the step number n to a number that has not been set before. For example, the step setting unit 13 changes the step number n by increasing by 1 or by decreasing by 1.

[0137] Step setting unit 13 will set the operation amount u k The modified step number n is assigned to model 15, and the control quantity y is obtained from model 15 after the modified step number n from the current time. k＋n The predicted value y k＋n '.

[0138] Step setting unit 13 for predicted value y k＋n 'and target value y d The difference is the overshoot G. k＋n Perform the calculation.

[0139] Step setting unit 13 will set the overshoot G k＋n Compare with the threshold Th, if the overshoot G k＋n If the number of steps is less than or equal to the threshold Th, then the modified number of steps n is set as the optimal number of steps.

[0140] If the overshoot G k＋n If the number of steps exceeds the threshold Th, the step setting unit 13 will further change the changed step number n to a number of steps that has not been set before.

[0141] Below, up to the overshoot G k＋n Until the value is less than or equal to the threshold Th, the step setting unit 13 repeatedly performs the process of changing the step number n, and adjusts the overshoot G. k＋n The number of steps n when it is less than or equal to the threshold Th is set as the optimal number of steps.

[0142] Figure 10 This is an explanatory diagram showing an example when the number of steps n set by the step setting unit 13 is 1.

[0143] exist Figure 10 In the middle, at time step k+1, the control quantity y k＋1 The predicted value y k＋1 'and target value y d The difference is the overshoot G. k＋1 It did not become approximately zero. This is because, after one step, the control variable y and its target value y must be equalized. d Since they are consistent, the operation quantity u increases, and the control quantity y exceeds the target value y. d .

[0144] Figure 11 This is an explanatory diagram showing an example when the number of steps n is set to 2 by the step setting unit 13.

[0145] exist Figure 11 In the middle, at time step k+2, the control quantity y k＋2 The predicted value y k＋2 'and target value y d The difference is the overshoot G. k＋2 It becomes approximately 0. This is because, after two steps, the control variable y and its target value y0 are... d As long as the values ​​are consistent, the operation amount u decreases, causing the control amount y to gradually approach the target value y. d .

[0146] exist Figure 10 and Figure 11 In the diagram, the horizontal axis represents the time step, and the vertical axis represents the control variable. The thick arrows in the diagram indicate the orientation and magnitude of the control variable u at each time step. Here, it shows the target value y assigned at time 0. d And the situation where control measures are implemented.

[0147] The prediction unit 14 obtains the number of steps n from the step setting unit 13.

[0148] The prediction unit 14, based on model 15 of the controlled object 2, determines the control quantity y at a time n steps from the current time. k＋n Make predictions ( Figure 9 Step ST12). Furthermore, if the start time of control is set to 0, then the control quantity y k＋n It is the control quantity at time (k+n)×Δt, where Δt represents the control period.

[0149] Specifically, the forecasting unit 14 will operate the amount u k The number of steps, n, is assigned to model 15. Model 15 is then assigned an operation amount, u, by the prediction unit 14. k Given the number of steps n, the control quantity y will be the value n steps from the current time. k＋n The predicted value y k＋nThe output is sent to the prediction unit 14. The prediction unit 14 obtains the control quantity y from the model 15 after n steps from the current time. k＋n The predicted value y k＋n '.

[0150] Prediction unit 14 will control quantity y k＋n The predicted value y k＋n 'Output to the operation quantity determination unit 12. Below, sometimes the predicted value y...' k＋n 'Written as "y k＋n (u) k )'".

[0151] Operation quantity determination unit 12 obtains control quantity y from prediction unit 14. k＋n The predicted value y k＋n For example, obtaining control variables y from external sources. k Target value y d .

[0152] The operation quantity determination unit 12 is set as shown in equation (2) below, and the predicted value y is... k＋n (u) k )' and target value y d The cost function F(the difference) Figure 9 Step ST13).

[0153]

[0154] Operation determination unit 12, for example, uses a bisection method to determine the operation u that makes the cost function F equal to 0. k ( Figure 8 Step ST14).

[0155] Operation quantity determination unit 12 will determine the operation quantity u k Output to control object 2 ( Figure 9 Step ST15).

[0156] In the above-described embodiment 2, it is configured as follows: Figure 7 The control device 1 shown, namely the prediction unit 14, predicts the control quantity when multiple controls are performed on the controlled object 2, and uses this prediction quantity as the control quantity for the controlled object 2. Therefore, Figure 7 In addition to the effects described above, the control device 1 shown also has the following effects: Figure 1 Compared to the control device 1 shown, it can also suppress overshoot.

[0157] The control quantity used when controlling object 2 multiple times is the control quantity of the pre-read state. Since it is only necessary to make the above control quantity consistent with the target value, there is no need to perform drastic operations, and as a result, overshoot is suppressed. The larger the number of steps n set by the step setting unit 13, the higher the overshoot suppression effect.

[0158] exist Figure 7 In the control device 1 shown, model 15 is built into the control device 1. However, this is just an example, and model 15 can also be set outside the control device 1.

[0159] In the case where the external model 15 is set up in the cloud, for example, the prediction unit 14 will operate on the amount u. k The number of steps, n, is sent to the model 15 in the cloud via communication lines such as the internet. The model 15 in the cloud is assigned an operational amount, u, by the prediction unit 14. k Given the number of steps n, the control quantity y will be the value n steps from the current time. k＋n The predicted value y k＋n For example, it is sent to the prediction unit 14 via a communication line.

[0160] Prediction unit 14 obtains the control quantity y of controlled object 2 from model 15 in the cloud. k＋1 The predicted value y k＋1 Then, the prediction unit 14 will use the obtained control quantity y k＋1 The predicted value y k＋1 The output is sent to the operation quantity determination unit 12, whereby the operation quantity determination unit 12 can use the predicted value y obtained from the prediction unit 14. k＋1 'And that determines the amount of work to be done.'

[0161] Implementation Method 3

[0162] In Embodiment 3, the control device 1 will be described. In the control device 1, the step setting unit 16 sets the allowable overshoot amount as a function of the step number and calculates the step number based on the function of the step number.

[0163] Figure 12 This is a structural diagram showing the control device 1 involved in Embodiment 3. Figure 12 In, with Figure 1 and Figure 7 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0164] Figure 13 This is a hardware structure diagram illustrating the hardware of the control device 1 according to Embodiment 3. Figure 13 In, with Figure 2 and Figure 8 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0165] Figure 12 The control device 1 shown has a step setting unit 16, a prediction unit 14, and an operation amount determination unit 12.

[0166] The step setting unit 16 sets the allowable overshoot G as a function G(n) of the step number n, and calculates the step number n based on the function G(n) of the step number n.

[0167] The step setting unit 16 outputs the step number n to the prediction unit 14.

[0168] Step setting unit 16, for example, is composed of Figure 13 The step setting circuit 36 ​​shown is implemented.

[0169] exist Figure 12 In the design, the structural elements of the control device 1, namely the step setting unit 16, the prediction unit 14, and the operation quantity determination unit 12, are each independently configured. Figure 13 The case shown is a dedicated hardware implementation. That is, it is envisioned that the control device 1 is implemented by a step setting circuit 36, a prediction circuit 34, and an operation quantity determination circuit 32.

[0170] The step setting circuit 36, the prediction circuit 34, and the operation quantity determination circuit 32 are each, for example, a single circuit, a composite circuit, a processor that performs processing based on instructions generated by a program, a processor that performs processing based on instructions generated by a parallel program, an ASIC, an FPGA, or a combination thereof.

[0171] The structural elements of the control device 1 are not limited to being implemented by dedicated hardware; the control device 1 can also be implemented by software, firmware, or a combination of software and firmware.

[0172] When the control device 1 is implemented by software or firmware, the program for causing the computer to execute the processing flow of the step setting unit 16, the prediction unit 14, and the operation amount determination unit 12 is stored in Figure 3 The memory 51 shown. Furthermore, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0173] In addition, Figure 13 The diagram shows an example of the dedicated hardware implementation for each structural element of the control device 1. Figure 3 The diagram shows an example of control device 1 implemented by software or firmware. However, this is just one example; it is also possible that some structural elements of control device 1 are implemented by dedicated hardware, while the remaining structural elements are implemented by software or firmware.

[0174] Next, for Figure 12 The operation of the control device 1 shown will be explained.

[0175] In addition to the step setting unit 16 and Figure 7 Since the control device 1 shown is the same, only the operation of the step setting unit 16 will be described here.

[0176] The function G(n) for the number of steps n is set in the step setting section 16.

[0177] The step setting unit 16 calculates the number of steps n that the following formula (3) holds, for example, using the bisection method.

[0178]

[0179] In equation (3), Mg is the margin of the allowable overshoot G.

[0180] The step setting unit 16 outputs the step number n to the prediction unit 14.

[0181] Figure 14 This is an explanatory diagram showing the relationship between the number of steps n and the overshoot G(n).

[0182] Figure 14 This example shows that when the set number of steps n is less than the optimal number of steps, the overshoot G(n) becomes a positive value, and when the set number of steps n is greater than the optimal number of steps, the overshoot G(n) becomes a negative value.

[0183] In the above-described embodiment 3, it is configured in the following manner. Figure 12 The control device 1 shown, namely the step setting unit 16, sets the allowable overshoot as a function of the step count and calculates the step count based on this function. Therefore, Figure 12 In addition to the effects described above, the control device 1 shown can also be used with... Figure 7 Compared to the control device 1 shown, it calculates the optimal number of steps with less computation.

[0184] Implementation Method 4

[0185] In Embodiment 4, the control device 1 will be described. In the control device 1, the step setting unit 17 updates the model 15 and the step number n after the control of the controlled object 2 has started and a set time has elapsed or the set conditions have been met.

[0186] Figure 15 This is a structural diagram showing the control device 1 involved in embodiment 4. Figure 15 In, with Figure 1 , Figure 7 and Figure 12 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0187] Figure 16 This is a hardware structure diagram illustrating the hardware of the control device 1 according to embodiment 4. Figure 16 In, with Figure 2 , Figure 8 and Figure 13 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0188] Figure 15 The control device 1 shown has a step setting unit 17, a prediction unit 14, and an operation amount determination unit 12.

[0189] After the step setting unit 17 starts controlling the controlled object 2 and a set time has elapsed or the set conditions have been met, it updates the model 15 and the step number n.

[0190] The step setting unit 17 outputs the updated step number n to the prediction unit 14.

[0191] Step setting unit 17, for example, is composed of Figure 16 The step setting circuit 37 shown is implemented.

[0192] exist Figure 15 In the design, the structural elements of the control device 1, namely the step setting unit 17, the prediction unit 14, and the operation quantity determination unit 12, are each independently configured. Figure 16 The case shown is a dedicated hardware implementation. That is, it is envisioned that the control device 1 is implemented by a step setting circuit 37, a prediction circuit 34, and an operation quantity determination circuit 32.

[0193] The step setting circuit 37, the prediction circuit 34, and the operation quantity determination circuit 32 are each, for example, a single circuit, a composite circuit, a processor that performs processing based on instructions generated by a program, a processor that performs processing based on instructions generated by a parallel program, an ASIC, an FPGA, or a combination thereof.

[0194] The structural elements of the control device 1 are not limited to being implemented by dedicated hardware; the control device 1 can also be implemented by software, firmware, or a combination of software and firmware.

[0195] When the control device 1 is implemented by software or firmware, the program for causing the computer to execute the processing flow of the step setting unit 17, the prediction unit 14, and the operation amount determination unit 12 is stored in Figure 3 The memory 51 shown. Furthermore, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0196] In addition, Figure 16 The diagram shows an example of the dedicated hardware implementation for each structural element of the control device 1. Figure 3 The diagram shows an example of control device 1 implemented by software or firmware. However, this is just one example; it is also possible that some structural elements of control device 1 are implemented by dedicated hardware, while the remaining structural elements are implemented by software or firmware.

[0197] Next, for Figure 15 The operation of the control device 1 shown will be explained.

[0198] In addition to the step setting unit 17, and Figure 7 Since the control device 1 shown is the same, only the operation of the step setting unit 17 will be described here.

[0199] After the step setting unit 17 starts controlling the controlled object 2 and a set time has elapsed or the set conditions have been met, it updates the model 15 and the step number n.

[0200] The step setting unit 17 outputs the updated step number n to the prediction unit 14.

[0201] The set time can be stored, for example, in the internal memory of the step setting unit 17, or it can be retrieved from... Figure 15 The control device 1 shown is externally supplied.

[0202] For example, there is a time when the operation of the controlled object 2 is completed, or when the operation of the controlled object 2 is 50% complete.

[0203] In the case where model 15 is implemented, for example, by a neural network, model 15 is given the operational quantity u that is assigned to the controlled object 2 during operation. k and the control quantity y of the operating controlled object 2 k＋n It then performs backpropagation to update the data.

[0204] In the case where model 15 is implemented, for example, by a Gaussian process regression model, model 15 uses the operational quantity u assigned to the controlled object 2 during operation. k and the control quantity y of the operating controlled object 2 k＋n The kernel function is then recalculated, and thus updated.

[0205] The step number n update process implemented by the step number setting unit 17 and the process through Figure 7 The step number setting unit 13 shown implements the step number n update process, or through... Figure 12 The step number setting unit 16 shown implements the same step number n update process.

[0206] Furthermore, before updating model 15 and step number n respectively, an initial value for step number n is set in step number setting unit 17. The initial value for step number n in the step number n update process is, for example, a value obtained by multiplying the time constant of the controlled object 2 by an arbitrary coefficient. Using the above initial value enables more stable operation of the controlled object 2 compared to using a smaller initial value. A small initial value is, for example, "1".

[0207] In the above-described embodiment 4, it is configured in the following manner. Figure 15The control device 1 shown, namely the step setting unit 17, updates the model 15 and the step count after a set time has elapsed since control of the controlled object 2 begins or after a set condition is met. Therefore, Figure 15 In addition to the effects described above, the control device 1 shown also has the following effects: Figure 7 Compared to the control device 1 shown, it can improve the accuracy of operation quantity calculation.

[0208] Implementation Method 5

[0209] exist Figure 1 In the control device 1 shown, the operation quantity determination unit 12 determines the operation quantity u based on the operation quantity u. k The control quantity y of the controlled object 2 after execution of control k＋1 The predicted value y k＋1 'and target value y d The function that evaluates the difference determines the amount of operation u. k .

[0210] In embodiment 5, a control device 1 having an operation quantity determination unit 20 will be described. This operation quantity determination unit 20 is based on a predicted value y based on a time t2 after the first time t1. k＋n 'and target value y d Determine the amount of operation u k The predicted value y at time t2. k＋n 'Is control object 2 being controlled by operation variable u k The control quantity y of the controlled object 2 at time t2 when control was performed k＋n The predicted value.

[0211] Figure 17 This is a structural diagram showing the control device 1 involved in embodiment 5.

[0212] Figure 18 This is a hardware structure diagram showing the hardware of the control device 1 involved in Embodiment 5.

[0213] Figure 17 The control device 1 shown has a prediction unit 18, a prediction value calculation unit 19, and an operation quantity determination unit 20.

[0214] For example, prediction section 18 is composed of Figure 18 The prediction circuit 38 shown is implemented.

[0215] The prediction unit 18 obtains the throughput u at time t1 based on model 18a of the controlled object 2. k The control quantity y of the controlled object 2 after control is performed k The predicted value is the first predicted value y. k '.

[0216] The prediction department 18 will predict the first value y k Output to the prediction calculation unit 19.

[0217] Model 18a mathematically describes the operand u assigned to the controlled object 2. k and the first predicted value y k The relationship between '. Or, Model 18a learns and masters the operands u assigned to the controlled object 2. k and the first predicted value y k The relationship between '.

[0218] If model 18a is assigned an operation amount u by the prediction unit 11 k Then the first predicted value y k Output to prediction section 18.

[0219] For example, the prediction calculation unit 19 is composed of... Figure 18 The predicted value calculation circuit 39 shown is implemented.

[0220] The prediction calculation unit 19 obtains the first predicted value y at time t1 from the prediction unit 18. k '.

[0221] Prediction calculation unit 19 is based on the first predicted value y k 'Calculate the throughput u of the time after time t1, i.e., time t2. k The control quantity y of the controlled object 2 after control is performed k＋n The predicted value is the second predicted value y. k＋n '.

[0222] The prediction calculation unit 19 will calculate the second prediction value y. k＋n Output to the operation quantity determination unit 20.

[0223] Operation quantity determination unit 20, for example, by Figure 18 The operation quantity shown is determined by circuit 40.

[0224] Operation quantity determination unit 20 obtains the second predicted value y from predicted value calculation unit 19. k＋n For example, obtaining control variables y from external sources. k Target value y d .

[0225] Operation quantity determination unit 20 is based on the second predicted value y k＋n 'and target value y d Determine the amount of operation u k .

[0226] Operation quantity determination unit 20 will determine the operation quantity u k The outputs are sent to the prediction unit 18 and the control object 2, respectively.

[0227] exist Figure 17 In this design, the structural elements of the control device 1, namely the prediction unit 18, the prediction value calculation unit 19, and the operation quantity determination unit 20, are each independent. Figure 18 The case shown is a dedicated hardware implementation. That is, it is envisioned that the control device 1 is implemented by a prediction circuit 38, a prediction value calculation circuit 39, and an operation quantity determination circuit 40.

[0228] The prediction circuit 38, the prediction value calculation circuit 39, and the operation quantity determination circuit 40 are each, for example, a single circuit, a composite circuit, a processor that performs processing based on instructions generated by a program, a processor that performs processing based on instructions generated by a parallel program, an ASIC, an FPGA, or a combination thereof.

[0229] The structural elements of the control device 1 are not limited to being implemented by dedicated hardware; the control device 1 can also be implemented by software, firmware, or a combination of software and firmware.

[0230] When the control device 1 is implemented by software or firmware, the program for causing the computer to execute the processing flow of the prediction unit 18, the prediction value calculation unit 19, and the operation quantity determination unit 20 is stored in Figure 3 The memory 51 shown. Furthermore, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0231] In addition, Figure 18 The diagram shows an example of the dedicated hardware implementation for each structural element of the control device 1. Figure 3 The diagram shows an example of control device 1 implemented by software or firmware. However, this is just one example; it is also possible that some structural elements of control device 1 are implemented by dedicated hardware, while the remaining structural elements are implemented by software or firmware.

[0232] Next, for Figure 17 The operation of the control device 1 shown will be explained.

[0233] Figure 19 This is a flowchart representing the processing flow, or control method, of control device 1.

[0234] First, the prediction unit 18 obtains the operation quantity u from the operation quantity determination unit 20. k The amount of operation u at that moment k Operations that have not yet been optimized by the operation quantity determination department 20 are not assigned to the control object 2.

[0235] The prediction unit 18 obtains the throughput u at time t1 based on model 18a of the controlled object 2. k The control quantity y of the controlled object 2 after control is performed k The first predicted value yk '.

[0236] That is, the forecasting unit 18 will operate the amount u k Assign the first predicted value y at time t1 to model 18a, and obtain the first predicted value y from model 18a. k '( Figure 19 Step ST21).

[0237] If model 18a is assigned an operation amount u by the prediction unit 11 k Then the first predicted value y k The output is sent to the prediction unit 18. The prediction processing of model 18a is a well-known technique, so detailed explanation is omitted.

[0238] The prediction department 18 will predict the first value y k Output to the prediction calculation unit 19.

[0239] The prediction calculation unit 19 obtains the first predicted value y from the prediction unit 18. k '.

[0240] The prediction calculation unit 19 is shown in equation (4) below, based on the first prediction value y. k 'For the second predicted value y k＋n 'Perform calculations ( Figure 19 Step ST22).

[0241]

[0242] The prediction calculation unit 19 will calculate the second prediction value y. k＋n Output to the operation quantity determination unit 20.

[0243] Operation quantity determination unit 20 obtains the second predicted value y from predicted value calculation unit 19. k＋n For example, obtaining control variables y from external sources. k Target value y d .

[0244] Operation quantity determination unit 20 is based on the second predicted value y k＋n 'and target value y d Determine the amount of operation u k ( Figure 19 Step ST23).

[0245] As the operation quantity u implemented by the operation quantity determination unit 20 k The decision processing, for example, can determine the relationship with the second predicted value y. k＋n 'and target value y d The difference is proportional to the amount of operation u k It can also be with Figure 1Similarly, the operation quantity determination unit 12 shown determines the operation quantity u that makes the cost function F become 0. k .

[0246] Furthermore, the operation quantity u implemented by the operation quantity determination unit 20 k The decision to process the data is based on the second predicted value y. k＋n 'and target value y d And determines the amount of operation u k That is, it can be any processing method, for example, it can be the second predicted value y. k＋n 'and target value y d The larger the difference, the greater the impact on the operation amount u. k Perform calibration, and determine the calibrated operation quantity as the operation quantity u. k .

[0247] Operation quantity determination unit 20 will determine the operation quantity u k Output to control object 2 ( Figure 19 Step ST24).

[0248] In the above-described embodiment 5, the control device 1 is configured as follows: it includes a prediction value calculation unit 19, which calculates a first prediction value (based on a predicted value of the control quantity of the controlled object 2 after control by an operation quantity at a first time point) and a second prediction value (based on a predicted value of the control quantity of the controlled object 2 after control by an operation quantity at a time after the first time point, i.e., a second time point); and an operation quantity determination unit 20, which determines the operation quantity based on a target value of the control quantity and the second prediction value calculated by the prediction value calculation unit 19. Therefore, the control device 1 can not only stably determine the operation quantity, but also determine the operation quantity at high speed.

[0249] In embodiment 5, the control device 1 includes a prediction value calculation unit 19, which calculates a second prediction value based on a first prediction value. Therefore, the control device 1 is capable of predicting the control quantity of the controlled object 2 at a time after the first time, i.e., at the second time.

[0250] In implementation method 5, it is configured in the following manner. Figure 17 The control device 1 shown, namely the prediction value calculation unit 19, predicts the control quantity of the controlled object 2 at the second time, which is after the first time, and uses it as the control quantity of the controlled object 2. Therefore, Figure 17 The control device 1 shown is Figure 1 Compared to the control device 1 shown, it can suppress overshoot.

[0251] The control quantity for controlled object 2 at time 2 is the control quantity of the pre-read state. Since it is sufficient to make the above control quantity consistent with the target value, no abrupt operation is performed, thereby suppressing overshoot.

[0252] In embodiment 5, the control device 1 is configured with a prediction unit 18, which predicts the control quantity at time t1 based on the model 18a of the controlled object 2, and sets the predicted value of the control quantity at time t1 as the first predicted value y. k The output is sent to the prediction calculation unit 19. Therefore, the control device 1 can predict the control quantity of the controlled object 2 at the second time with less computation. In addition, the prediction unit 18 has a built-in model 18a, so the input and output between the prediction unit 18 and the model 18a can be completed in a short time, resulting in high-speed prediction of the control quantity.

[0253] exist Figure 17 In the control device 1 shown, the prediction unit 18 has a built-in model 18a. However, this is just an example, and the model 18a can also be provided outside the prediction unit 18.

[0254] In the case where the external model 18a is set up, for example in the cloud, the prediction unit 18 will obtain the operation amount u from the operation amount determination unit 20. k For example, model 18a is sent to the cloud via a communication line. If the model 18a in the cloud is assigned an operational variable u by the prediction unit 18... k Then the first predicted value y of the first time t1 of the controlled object 2 will be... k For example, it is sent to the prediction unit 11 via a communication line.

[0255] Prediction unit 18 obtains the first predicted value y at time t1 from model 18a in the cloud. k The prediction unit 18 will predict the first value y at time t1. k Output to the prediction calculation unit 19.

[0256] In the implementation method 2 Figure 7 In the control device 1 shown, the prediction unit 14 predicts the control quantity for the controlled object 2 when multiple controls are performed on the controlled object 2 from the current moment, and uses this prediction quantity as the control quantity for the controlled object 2. However, this is only an example; the prediction unit 14 and... Figure 17 Similarly, the prediction unit 18 shown also obtains, for example, the control quantity y at time t1. k The predicted value y k That is, the prediction unit 14 will operate on the amount u. k Assign the value to model 15, and obtain the predicted value y at time t1 from model 15. k Furthermore, the prediction unit 14 can also be connected with... Figure 17 Similarly, the prediction calculation unit 19 shown calculates the predicted value y based on time t1. k ', the predicted value y of the control quantity at time t2, which is the time after time t1. k＋nThe calculation is performed to determine the control quantity at the moment after multiple controls have been performed on the controlled object 2 starting from time t1. The prediction unit 14 calculates the predicted value y of the control quantity at the second time t2. k＋n Output to the operation quantity determination unit 12.

[0257] In this case, the implementation method 2 involves Figure 7 Model 15 shown describes the operand u assigned to the controlled object 2 in mathematical form. k and predicted value y k The relationship between '. Or, Model 15 learns and masters the operands u assigned to the controlled object 2. k and predicted value y k The relationship between '.

[0258] If model 15 is assigned an operation amount u by the prediction unit 14 k Then the predicted value y k Output to prediction section 15.

[0259] In implementation methods 1-5, backlash was not specifically mentioned. Backlash refers to the time it takes for the target value y to be within a certain range. d and control quantity y k When there is a tiny difference, even if the tiny operation amount u corresponding to that difference is... k The output is sent to control object 2, and control object 2 will not affect the operation amount u. k Respond.

[0260] If an attempt is made to control object 2 while an idle period exists, control object 2 will not respond to the operation variable u. k Make a response, therefore relative to the target value y d and control quantity y k The difference will result in a larger operation volume u k The output is sent to control object 2, and the result is sometimes the control quantity y of control object 2. k It produces vibration.

[0261] As a countermeasure to the aforementioned empty schedule, the following countermeasures exist, for example.

[0262] The operation quantity determination unit 12, etc., takes the following countermeasures, namely, with the target value y d and control quantity y k The difference Δy (=|y d -y k＋n '|) Correspondingly, the operation quantity u k Adjust the size.

[0263] Specifically, if the difference Δy is less than a specified value, the operation quantity determination unit 12, etc., will adjust the operation quantity u. k Adjustment set to 0.

[0264] In implementation 3, the relationship between the number of steps n and the overshoot G(n) is shown. Figure 14 Examples of relationships are shown. In Figure 14 In the example, there is 1 n that satisfies G(n) = 0. However, the relationship between the number of steps n and the overshoot G(n) is... Figure 20 In the case of the relationship shown, there are multiple n that satisfy G(n) = 0.

[0265] Figure 20 This is an explanatory diagram showing the relationship between the number of steps n and the overshoot G(n).

[0266] When there are multiple n that satisfy G(n) = 0, the smaller the n, the shorter the pre-read time, and therefore the faster the response. Therefore, the optimal n that satisfies G(n) = 0 and min(n) can also be efficiently explored using a binary search method.

[0267] Specifically, such as Figure 21 As shown, the upper bound of the solutions for n in the bisection method is n. upper The lower bound of the solution for n in the bisection method is n lower The midpoint of the solution for n in the bisection method is n middle In this case, by processing it as follows, the optimal n can be found efficiently.

[0268] Figure 21 This is an illustration of the bisection method for efficiently finding the optimal value of n.

[0269] In G(n) middle When n < 0, as shown below, n upper Change to n middle .

[0270] n upper ←n middle In G(n) middle When n = 0, as shown below, n upper Change to n middle .

[0271] n upper ←n middle In G(n) middle When n > 0, as shown below, n lower Change to n middle .

[0272] n lower ←n middle

[0273] Furthermore, the present invention allows for free combination of various embodiments, modification of any structural elements in each embodiment, or omission of any structural elements in each embodiment.

[0274] Industrial applicability

[0275] This invention applies to control devices, control methods, and programs.

[0276] Explanation of the label

[0277] 1 Control device, 2 Controlled object, 11 Prediction unit, 11a Model, 12 Operation amount determination unit, 13 Step setting unit, 14 Prediction unit, 15 Model, 16, 17 Step setting units, 18 Prediction unit, 18a Model, 19 Prediction value calculation unit, 20 Operation amount determination unit, 31 Prediction circuit, 32 Operation amount determination circuit, 33 Step setting circuit, 34 Prediction circuit, 36, 37 Step setting circuit, 38 Prediction circuit, 39 Prediction value calculation circuit, 40 Operation amount determination circuit, 51 Memory, 52 Processor.

Claims

1. A control device that determines the operational quantity used to control a controlled object. The control device is characterized by, The device has an operation quantity determination unit that determines the operation quantity based on a function that evaluates the difference between the predicted value of the control quantity of the controlled object after control is performed by the operation quantity and the target value of the control quantity.

2. The control device according to claim 1, characterized in that, As a function for evaluating the difference, the operation amount determination unit determines the operation amount based on a cost function associated with the difference.

3. The control device according to claim 1, characterized in that, The operation quantity determination unit uses a binary search method to determine the operation quantity.

4. The control device according to claim 1, characterized in that, It has a prediction unit that predicts the control quantity based on the model of the controlled object and outputs the predicted value of the control quantity to the operation quantity determination unit.

5. The control device according to claim 4, characterized in that, The prediction unit predicts the control quantity when multiple controls are performed on the controlled object, and uses this prediction quantity as the control quantity.

6. The control device according to claim 5, characterized in that, The prediction unit predicts the control quantity at the time after multiple controls have been performed on the controlled object from the current time, and uses this prediction quantity as the control quantity.

7. The control device according to claim 6, characterized in that, It includes a step setting unit that sets the number of times control is executed for the controlled object, i.e., the number of steps, based on the control result of the model when it is controlled. The prediction unit predicts the control quantity as the time elapsed after the number of steps set by the step setting unit from the current time.

8. The control device according to claim 7, characterized in that, The step setting unit sets the allowable overshoot as a function of the step count, and calculates the step count based on the function of the step count.

9. The control device according to claim 8, characterized in that, The step setting unit uses a binary search method to calculate the number of steps.

10. The control device according to claim 7, characterized in that, The step setting unit updates the model and the number of steps after starting control of the controlled object and after a set time has elapsed or after a set condition has been met.

11. The control device according to claim 7, characterized in that, Before updating the model and the number of steps, an initial value for the number of steps is set in the step setting unit.

12. A control method that determines the operational quantity used to control a controlled object. The characteristic of this control method is that, The operation quantity determination unit determines the operation quantity based on a function that evaluates the difference between the predicted value of the control quantity of the controlled object after control is performed through the operation quantity and the target value of the control quantity.

13. A program for execution by a computer of a control device, the control device determining the operational quantities used to control a controlled object. in, The program is used to cause the computer to perform the following process: determining the operation quantity based on a function that evaluates the difference between the predicted value of the control quantity of the controlled object after control is performed through the operation quantity and the target value of the control quantity.

14. A control device that determines an operational quantity for controlling a controlled object. The control device is characterized by having: The prediction value calculation unit calculates, based on the predicted value of the control quantity of the controlled object after being controlled by the operation quantity at time 1 (i.e., the first predicted value), the predicted value of the control quantity of the controlled object after being controlled by the operation quantity at time 2 (i.e., the second predicted value); and The operation quantity determination unit determines the operation quantity based on the target value of the control quantity and the second predicted value calculated by the predicted value calculation unit.

15. A control method that determines the operational quantity used to control a controlled object. The characteristic of this control method is that, The prediction value calculation unit calculates the predicted value of the control quantity of the controlled object after being controlled by the operation quantity at time 1, i.e., the first predicted value, and the predicted value of the control quantity of the controlled object after being controlled by the operation quantity at time 2, i.e., the second predicted value. The operation quantity determination unit determines the operation quantity based on the target value of the control quantity and the second predicted value calculated by the predicted value calculation unit.

16. A program for execution by a computer of a control device, the control device determining the operational quantities for controlling a controlled object. in, This program is used to cause the computer to perform the following processes: Based on the predicted value of the control quantity of the controlled object after being controlled by the operation quantity at time 1 (i.e., the first predicted value), the predicted value of the control quantity of the controlled object after being controlled by the operation quantity at time 2 (i.e., the second predicted value) after being controlled by the operation quantity at time 1; and The operation quantity is determined based on the second predicted value and the target value of the control quantity.

Images