JPH05120253A - Controller - Google Patents

Abstract

PURPOSE:To accurately and flexibly control a nonlinear type controlled system in the multi-input, multi-output controller which is used to control the controlled system. CONSTITUTION:This controller is provided with a data processor 1 with a signal converting function, a virtual command setting device 5 which sets a transcendental knowledge between control states, and a calculation device 6 which calculates a correction quantity for a control manipulated variable and the data processor 1 consists of, for example, a neural network; and a control state quantity is inputted to the data processor 1, whose output is handled as the control manipulated variable and supplied to the controlled system 3 and corrected according to the current control quantity from the calculation device 6 to obtain a tutor signal, thus structuring the data processor 1 as a desired controller. The data processor 1 inputs the difference value between the control state quantity and a target control state quantity and the virtual command setting device 5 consists of a neural network and uses the difference value from the control state quantity as a control command as a parameter to obtain a virtual command curve showing the relation between the control state quantities by learning.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multi-input multi-output control device used for controlling a control target, and more particularly to a control device for accurately and flexibly controlling the non-linear control target.

A method of designing a control device by applying modern control theory because a classical PID correspondence cannot cope with a complicated controlled object such as a 1-input 2-output system of inverted pendulums. Has been taken. However, when the linear control theory of the modern control theory is used, the control performance is deteriorated outside the linear region because the controlled object model is constructed by linearizing the equation of motion of the controlled object.

Further, when the nonlinear control theory of the modern control theory is used, it is necessary to accurately describe the equation of motion, so that it is necessary to accurately identify the parameter of the controlled object, which is extremely difficult. It has the drawback of being work. Fuzzy control theory has also been proposed, which requires adjusting the membership function.
Against this background, a control device having a new configuration using a neural network is being proposed in recent years.

[0004]

2. Description of the Related Art A neural network has a property that, when a teacher signal group is given, the input / output characteristics of those teacher signal groups are provided by learning. Is given,
It has the property of realizing an adaptive data processing function of outputting an appropriate output signal.

When such a neural network is constructed as a control device, a sufficient number of control data are acquired from the controlled object and learning is executed by using the acquired control data as a teacher signal. By mapping the control rules for the controlled object on the neural network, a method of constructing as a control device will be adopted.

However, it is often impossible to obtain the control data of the controlled object in reality when the controlled object becomes complicated. So, as a way to deal with this,
Recently, a certain amount of a priori knowledge has been obtained qualitatively, but for a controlled object that has many unknown parts quantitatively, a teacher signal is obtained by trial and this is used to control the object. Proposal of Mapping Control Rules for a Neural Network on a Neural Network (Saito, Kitamura, "Stabilized Learning Control of Inverted Pendulum Using Multilayer Neural Network", Robotics and Mechatronics '90 Proceedings, p.
283-286, 1990).

This new method employs a configuration including a neural network, a virtual target generation section, and an evaluation section in order to stop the inverted pendulum on the carriage at the origin.

This newly provided virtual target value generator expresses a priori knowledge that "the further the position of the carriage is from the origin, the more the virtual target angle of the pendulum is tilted from the vertical direction toward the origin". When the position and speed of the dolly are given, a virtual target value of the pendulum angle and angular velocity for moving this to the origin is generated.

On the other hand, the newly provided evaluation unit evaluates how the difference between the generated virtual target value and the control output (angle / angular velocity of the pendulum) should be after one sampling. Finding the correction amount of the force to be applied to the dolly,
The force corrected by the correction amount is specified as a teacher signal.

Then, the neural network inputs the pendulum angle and each speed at each sampling time, and the position and speed of the carriage, and outputs the force to be applied to the carriage. The learning of the neural network is performed by the back propagation method based on the teacher signal generated by the evaluation unit.

By adopting this configuration, a certain amount of a priori knowledge is qualitatively obtained, but a teacher signal is obtained by trial for a control target that has many unknown quantitatively. Then, by using this to map the control rules for the controlled object on the neural network, it is possible to construct the neural network as a control device for the controlled object.

[0012]

However, in this method, (1) an intermediate state that can be taken by the controlled object can be expressed by only one kind of virtual target value. If it is expressed as a group of actions, it becomes impossible to deal with complicated changes in the system state. (2) It is difficult to express human empirical knowledge in virtual target values. Therefore, there is a problem that the control performance is deteriorated.

In view of the above, according to the present invention, (1) a plurality of operations constituting a series of operations, for example, an operation of continuously transiting on two curves having different characteristics, a virtual target setting section corresponding to each curve To learn a series of motions of the controlled object with multiple virtual target values, (2) some representative points from human empirical knowledge, or the state of the controlled object operated without control input. By learning the sampled points, we aim to obtain the function of the virtual target value.

[0014]

FIG. 1 is a block diagram showing the principle of the present invention. In the figure, 1 is a data processing device, 2 is a learning processing device, 3 is a control target, 4 is a target value setting device, 5 is a virtual target setting device, 6 is an operation correction amount calculation device, and 7 is a first differencer. , 8 is a second difference unit, and 10 is a control unit including a data processing device 1 and a learning processing device 2.

The data processing device 1 is provided with a variable signal conversion function, and when the teacher signal group is given, the signal conversion function can be set to one which realizes the input / output characteristics of the teacher signal group. Take. This data processing device 1 is
By being constructed as a control device, when the control state quantity of the controlled object 3 and its target value are given, it outputs a control operation amount for controlling the controlled object 3 to the target controlled state. I will go.

The data processing device 1 receives one or a plurality of inputs and internal state values to be multiplied by these inputs to obtain a product sum value, and the product sum value is obtained by a predetermined function. It may be composed of a network structure composed of the internal connection of the basic units for converting and obtaining the output value, and the qualitative data relationship between the control state quantity and the control operation quantity may be represented by IF-THEN. In addition to the rules, the fuzzy device may describe the qualitative attributes of the control state amount and the control operation amount described in the IF-THEN rule by the membership function.

The learning processing device 2 learns the signal conversion function of the data processing device 1 so as to realize the input / output characteristics of the teacher signal group when the teacher signal group is given. When the data processing device 1 is configured by the network structure unit, the learning processing device 2 executes a well-known learning algorithm such as a back propagation method.

The control target 3 is a control target controlled by the data processing device 1 constructed as a control device. The control target 3 is preferably an actual control target, but the control target model may be used instead of the actual control target.

The target value setting device 4 sets the target value of the control state quantity representing the desired control state of the controlled object 3. The virtual target setting device 5 is composed of a neural network, and is controlled so that the virtual target can be learned using the virtual target value learning control unit 12 that follows the law of back propagation, for example. The virtual target setting device 5 then controls the controlled object 3
Manages a priori knowledge of the data relationship between the control state quantities obtained in order to realize the desired control state, and the management data of this a priori knowledge is used as a parameter for the difference value from the target value of the control state quantity. And manage.

The selection unit 11 includes a plurality of virtual target curves f ₁ ,
A plurality of virtual target setting devices 5 that acquire f ₂ by learning and hold learning results as weight values are selected in accordance with the selection of the control unit 10. At this time, each virtual target curve f ₁ , f
₂ corresponds to each selected virtual target setting device 5.

The operation correction amount calculation device 6 has a control state amount of the controlled object 3 and a virtual target value of the control state amount specified from the management data of the virtual target setting device 5 corresponding to the control state amount. From this, the correction amount of the control operation amount for the controlled object 3 required to realize the target value of the control state amount is calculated. The operation correction amount calculation device 6 calculates the correction amount of the control operation amount by multiplying the difference value between the control state amount of the controlled object 3 and the virtual target value output by the virtual target setting device 5 by a proportional coefficient. The configuration may be adopted.

The first differentiator 7 calculates a difference value between the target value of the control state quantity set by the target value setting device 4 and the control state quantity of the controlled object 3, and the difference value is subjected to data processing. Input to the device 1 and the learning processing device 2. At this time, the output value of the first difference unit 7 may be multiplied by the proportional coefficient.

The second difference unit 8 calculates a difference value between the control operation amount output from the data processing device 1 and the correction amount of the control operation amount output from the operation correction amount calculation device 6, and the difference value. That is, the control operation amount output from the data processing device 1 is corrected by the calculated correction amount and input to the learning processing device 2.

[0024]

In the present invention, when the signal conversion function of the data processing device 1 is set to the initial state, for example, the control target 3
When the initial value of the control state quantity is output from, the first difference unit 7 calculates the difference value between the target value of the control state quantity set by the target value setting device 4 and the initial value of the control state quantity. Input to the data processing device 1. In response to this input, the data processing device 1 calculates the control operation amount defined by the initial state signal conversion function and outputs the control operation amount to the control target 3, and the control target 3 receives the output process of the control operation amount. Transition to a control state different from the initial state. Hereinafter, this process is repeated until the control state of the controlled object 3 reaches the specified limit.

In this process, when the virtual target setting device 5 receives the control state quantity from the controlled object 3, it specifies the virtual target value of the control state quantity according to the management data. According to the processing of the virtual target setting device 5, for example, the control target 3 is 1
In the case of the example of the control system of the input 2 output system, a virtual target value of one control state quantity output from the controlled object 3 for the other control state quantity is specified.

When the virtual target setting device 5 specifies a virtually constant target value of the control state quantity, the operation correction amount calculation device 6 uses this virtual target value to realize the target value of the control state quantity. The correction amount of the control operation amount for the controlled object 3 that is necessary for this is calculated. According to this correction amount calculation process,
In order to realize the target control state amount set by the target value setting device 4, it is decided how the control operation amount output by the data processing device 1 at the time of the processing should be corrected. Become.

In this way, when the teacher signal group consisting of the difference value of the control state quantity input to the data processing device 1 and the more preferable control operation quantity at the time of inputting the difference value is obtained, the learning processing is performed. The device 2 executes the learning process of the signal conversion function of the data processing device 1 and sets the signal conversion function to one suitable for realizing a more targeted control state.

Then, according to the signal conversion function of the newly set data processing apparatus 1, the same processing as described above is repeated so that the next teacher signal group is generated, and the learning processing apparatus 2 is used. The data processing device 1 is constructed as a control device by setting the signal conversion function of the data processing device 1 so as to realize a target control state.

In this way, a certain amount of a priori knowledge is qualitatively obtained, but a teacher signal is obtained by trial with respect to the controlled object 3 in which there are many unknown parts quantitatively. By using this to map the control rule for the controlled object 3 on the signal conversion function of the data processing device 1,
A configuration in which the data processing device 1 is constructed as a control device for the controlled object 3 and the construction process of the data processing device 1 is executed according to the difference value from the target value of the control state quantity. Therefore, even if the target value of the control state quantity is changed, it is not necessary to redo the learning.

Even if the target position changes, for example, the input to the controlled object for inversion does not change with respect to the difference value between the target position and the current position. That is, the input to the controlled object changes only by the difference value, and even if the target position changes, if the difference value does not change, the input to the same controlled object may be output. Therefore, if the input to the controlled object regarding the difference value is learned, the control is only performed in accordance with the difference value, and it is not necessary to relearn.

Even when the target value of the control state quantity is changed, the control state quantity is desired by learning in advance the relationship between the difference between the current value and the target value and the corresponding input value to the controlled object. Can be controlled to a desired control state at an arbitrary target position.

In the present invention, when a series of operations of one controlled object is composed of a plurality of operations, each operation is provided with a control unit 10, and those control units 10 are provided.
Virtual target value setting device 5 and operation correction amount calculation device 6 each of which includes a neural network for learning a virtual target curve corresponding to each control unit 10 in order to generate teacher data for learning. The virtual target curves f ₁ and f ₂ corresponding to each of the devices 5 are learned by the backpropagation method. It is assumed that the control unit 10 is trained to produce an output that approaches a given virtual target value according to the current operation purpose of the controlled object.

In this way, the selection unit 1 is selected according to the purpose of operation.
By selecting the control unit 10, the virtual target setting device 5, and the operation correction amount calculation device 6 in ₁ , the control rule for controlling the controlled object corresponds to the virtual target curves f ₁ and f ₂ that follow some simple control rules. Since it can be divided into two parts, it becomes easier to construct a control law. To control a series of actions of the controlled object using these, change them according to the state of the controlled object.

Further, by providing the virtual target setting device 5 with a learning function, the virtual target value learning control unit 12 is used to acquire a function of an appropriate empirical rule by using some representative virtual target values. Because it is possible, it becomes easy to express the rule of thumb.

[0035]

EXAMPLES The present invention will be described in detail below with reference to examples. FIG. 2 illustrates an inverted pendulum model assumed as a controlled object model in this embodiment.

As shown in FIG. 2, in the inverted pendulum model, the link L ₂ is connected to the base link L at the other end C of the base link L ₁ connected to the motor shaft (Z axis) at the origin 0.
It is constructed by connecting ₁ as a rotation axis. The rotation angles of the base link L ₁ and the link L ₂ are θ ₁ and θ ₂ , the masses are m ₁ and m ₂ , the lengths are 1 ₁ and l ₂ , respectively, and the gravity acceleration is g and the motor torque is If expressed by T, the equation of motion of this inverted pendulum model is as shown in FIG.

In this embodiment, with respect to the inverted pendulum model according to this equation of motion, the torque T generated by the motor is controlled to feed back the states of the links L ₁ and L ₂ to invert the pendulum. , Base link L ₁
The control target is to stop at an appropriate target position.

As described above, the inverted pendulum model has the control state quantity as

[0039]

[Equation 1]

Since there is one of the motor torque T as the control operation amount, the first
In the differencer 7 of, the corresponding target value set by the target value setting device 4 is

[0041]

[Equation 2]

If

[0043]

[Equation 3]

The difference value is calculated and input to the neural network. In the following, for convenience of description, when expressing the differential value of the angular velocity, (d / dt) may be added before the angle.

FIG. 4 shows an embodiment of the present invention. In the figure, a data processing device 1 is a neural network that functions as a control device, and 2 is a learning processing device that executes learning processing of the neural network. The learning processing device 2 constitutes the control unit 10. The difference value E _{1 (n)} , outer 1, E _{2 (n)} , outer 2 at time n is output to the control unit 10, and the torque T _(n) is output.

[0046]

[Outer 1]

[0047]

[Outside 2]

Is output. The controlled object 3 is an inverted pendulum model and forms a control system of a 1-input 2-output system which is a controlled object. The torque T _(n) is input and the control state quantity θ _{1 (n + 1} at the time n + 1. ₎ , Outer 3, θ _{2 (n + 1)} , outer 4
Is output. Target value setting device

[0049]

[Outside 3]

[0050]

[Outside 4]

Reference numeral 4 sets the target value θ _target (θ _t1 , θ _t2 ) of the control state quantity of the inverted pendulum model, and the virtual target setting device 5 sets the current position input θ _{1 (n)} and the outer 5 as target values. θ _t1 , θ _t2

[0052]

[Outside 5]

As a result of learning by inputting
Virtual target value θ of the control state quantity of the child model _{d1 (n)}, Outside 6
 , Θ_{d2 (n)}, Out 7 is output. Operation correction amount calculator
Place 6 is

[0054]

[Outside 6]

[0055]

[Outside 7]

The correction amount ΔT _(n) of the torque output from the Ural network is calculated by the following formula.

[0057]

[Equation 4]

However, K ₁ , K ₂ , and K ₃ are parameters. Reference numeral 17 is a delay device which delays the control state quantity output from the inverted pendulum model by one sampling time, and 7 is a first difference device, which is a target value of the control state quantity set by the target value setting device 4. And a difference value between the control state quantity output from the delay device 17 and the difference value is input to the neural network and the learning processing device 2. Reference numeral 8 denotes a second difference device, A difference value between the output torque T _(n) and the torque correction amount ΔT _(n) calculated by the operation correction amount calculation device 6 is calculated, and the difference value is input to the learning processing device 2 as a teacher signal. is there.

From the equation (4), the neural network is
Since it is necessary to prepare an input device having four units and an output layer having one unit for outputting the torque T, in this embodiment, as shown in FIG. 5, an input consisting of four input units 20 is provided. Between the input unit 20 of the input layer and the basic unit 21 of the intermediate layer, which is provided with a layer, an intermediate layer of one-stage structure composed of eight basic units 21, and an output layer composed of one basic unit 21. An inner join of
Basic unit 21 of the middle layer and basic unit 21 of the output layer
A neural network having a hierarchical network structure in which weight values are set for the internal connection between and is prepared.

The input unit 20 of the input layer distributes the input signal value as it is and outputs it to the basic unit 21 of the intermediate layer, and the basic units 21 of the intermediate layer and the output layer each internally couple a plurality of inputs. A multiplication processing unit that multiplies the weight values of, a cumulative processing unit that adds all the multiplication results, and a threshold processing unit that performs non-linear threshold processing on the cumulative value and outputs one final output. The learning processing device 2 is designed to realize the input / output characteristics of the teacher signal group,
The learning process of the weight value of these internal couplings will be performed.

The torque output from the motor causes the base link to
Ku works. Its base link L ₁At the end of the link L
₂(Pendulum) is attached. This control problem is
"Raise the pendulum and place the base link at any position.
To invert the pendulum. " The appearance is a series
It seems to work, but here, "swing up",
Includes two actions of "standing upside down"
ing. Therefore, the curves corresponding to these two movements individually
F₁, F₂Then, these curves f₁, F₂Satisfied
Neural network by learning the input-output relationship
It is set in the virtual goal setting device 5.

That is, it is necessary to learn the virtual target curve necessary for learning the control law for inverting the link L ₂ (pendulum) at a certain arbitrary position and the control rule for swinging up the link L ₂ (pendulum). The virtual target curve is learned by a virtual target setting device composed of a neural network. Specifically, the trajectory of the virtual target curve for inversion of the pendulum at an arbitrary position uses the empirical knowledge of "tilting the pendulum to the position where you want to invert the pendulum," The trajectory of the target curve "has a large velocity when the pendulum is at the bottom and zero velocity when the pendulum is at the top"
Function, that is, virtual target curve f
Obtained by ₁ and f ₂ .

These two virtual target curves f ₁ and f ₂ are selected by the selector 11 according to the current position of the pendulum to be controlled and the differentiation of that position. Also, for each
The control unit 10 is assigned. Then, data for actually operating the inverted pendulum is corrected so that the operation of the inverted pendulum approaches these virtual target values, and the control unit 10 acquires them by learning.

Next, when learning is performed using the virtual target value learning control unit 12 so that the input / output relationship of each virtual target setting device 5 formed of a neural network follows the virtual target curves f ₁ and f _2. The selection of the orbit will be described.

One is to obtain from empirical knowledge of human beings.
Select some of the inverted pendulum states that can be backed up
By learning using propagation, the
Number f ₁, F₂To win.

That is, a neural network whose input is the position of the pendulum and whose output is the speed of the pendulum is used as teacher data. In case of, the speed of the pendulum is 0, and some intermediate points are used for learning. Then, the neural network interpolates these values so that an appropriate function can be expressed. Similarly, various other empirical rules can be expressed by the neural network.

The other one uses the function obtained by sampling the process of operating the inverted pendulum from the target position without a control input and learning them and using it as a function of the virtual target value. This is because the target value is reached if the function moves.

That is, in the case of swinging up, the fact that the action of swinging up from just below to the action of swinging up from just above is very similar to each other Is trained by the virtual target value setting device 5 composed of a neural network as a teacher data, and a empirical rule function is expressed. Teacher data to be given to the data processing device 1 of the control unit 10 is created so as to approach this empirical rule.

Although the illustrated embodiment has been described above, the present invention is not limited to this. For example, in the embodiment, the one in which the control device is constructed by using the neural network is disclosed, but the present invention is not limited to this, and can be directly applied to all data processing devices capable of adjusting the signal conversion function according to the teacher signal. Of. Then, in the embodiment, the one in which the inverted pendulum is controlled is disclosed, but the invention is not limited to this, and can be applied as it is to all objects to be described.

In the embodiment, the control device is constructed not by using the actual controlled object but by using the controlled object model, but the present invention is not limited to this, and the actual controlled object itself is used. In this case, since accurate system identification including Coulomb friction and the like can be performed, a more appropriate control device can be constructed.

[0071]

As described above, according to the present invention, a series of operations of a controlled object are divided into several operations, virtual target values are learned for each, and a plurality of control rules learned by them are controlled. Depending on the state, it is possible to deal with a complicated change in the system state, and control performance is improved.

When learning the virtual target value, some representative points of the virtual target value given by a human are learned, or the control target is actually operated to learn the sampling point from the operating state. As a result, it becomes possible to obtain a function of the virtual target value, and control performance is improved.

[Brief description of drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is an explanatory diagram of an inverted pendulum used in an example.

FIG. 3 is an explanatory diagram of a motion equation of an inverted pendulum used in an example.

FIG. 4 is an example of the present invention.

FIG. 5 is a configuration diagram of a neural network.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Data processing device 2 Learning processing device 3 Control object 4 Target value setting device 5 Virtual target setting device 6 Operation correction amount calculation device 7 1st difference device 8 2nd difference device 11 Selection part 12 Virtual target value learning control part

Claims (7)

[Claims] 1. A virtual target setting unit having one or a plurality of learning functions for expressing a part or all of a control variable by a virtual target value, and calculating an error between the virtual target value and the state of a controlled object. , One or a plurality of operation correction amount calculation units for calculating a correction amount of an input to a control target that realizes the virtual target value, and a signal obtained by correcting the input to the control target by the correction amount as a teacher signal A signal obtained from the output of the controlled object is set as an input, the input to the controlled object is output, the input / output relationship is acquired by learning, and the controlled object is given based on the learned input / output relationship. A control device having one or more control units for controlling to a target value. 2. The control device according to claim 1, wherein a virtual target setting unit required from a virtual target setting unit having a plurality of learning functions is provided according to a value of one or a plurality of variables representing a state of a controlled object, , A required operation correction amount calculation unit is selected from a plurality of operation correction amount calculation units, and one or a plurality of control units is made to learn so as to realize the virtual target value calculated by the selected virtual target setting unit, A control device, wherein a required control unit is selected from a plurality of control units according to the values of one or more variables that represent the state of a control target. 3. The control device according to claim 1, wherein the virtual target setting unit sets a rule for calculating a virtual target value by learning from representative points of the virtual target value based on given human empirical knowledge. A control device characterized by acquiring. 4. The control apparatus according to claim 1, wherein the virtual target setting unit operates the control target from a given target value without a control input, and samples the state of the control target during that period by learning. A control device characterized by acquiring a rule for calculating a virtual target value. 5. The control device according to claim 3 or 4, wherein the virtual target setting unit is configured by a neural network. 6. A variable signal conversion function is provided, and when a teacher signal group is provided, the signal conversion function can be recognized as a device that realizes the input / output characteristics of the teacher signal group. A data processing device (1), a virtual target setting device (5) for managing a priori knowledge of the data relationship between control state quantities obtained to realize a desired control state of a controlled object, and a control operation When a quantity is given, the control state quantity of the control target or its control target model and the virtual control state quantity specified from the management data of the virtual target setting device (5) corresponding to the control state quantity An operation correction amount calculation device (6) for calculating a correction amount of the control operation amount required to realize a desired control state from the target value, and the data processing device (1) has a control state. As you enter the amount, The output corresponding to the force is treated as a control operation amount and given to the control target or its control target model, and the control operation amount is corrected according to the correction amount from the operation correction amount calculation device (6) at that time. The data processing device (1) is obtained by obtaining a teacher signal by going and setting the signal conversion function according to the obtained teacher signal.
Is a control device for realizing a desired control state, wherein the data processing device (1) is a difference between a control state amount output from a control target or its control target model and a control state amount serving as a control target. The virtual target setting device (5) is configured to input a value or a value corresponding to the value, and the virtual target setting device (5) uses the difference value from the control state amount serving as the control target as a parameter to set the data relationship between the control state amounts. Control characterized by being configured to manage experimental knowledge and acquire a rule for calculating a virtual target value by learning from a representative point of the virtual target value based on given human empirical knowledge apparatus. 7. A variable signal conversion function is provided, and when a teacher signal group is given, the signal conversion function can be recognized as a device that realizes the input / output characteristics of the teacher signal group. A data processing device (1), a virtual target setting device (5) for managing a priori knowledge of the data relationship between control state quantities obtained to realize a desired control state of a controlled object, and a control operation When a quantity is given, the control state quantity of the control target or its control target model and the virtual control state quantity specified from the management data of the virtual target setting device (5) corresponding to the control state quantity An operation correction amount calculation device (6) for calculating a correction amount of the control operation amount required to realize a desired control state from the target value, and the data processing device (1) has a control state. As you enter the amount, The output corresponding to the force is treated as a control operation amount and given to the control target or its control target model, and the control operation amount is corrected according to the correction amount from the operation correction amount calculation device (6) at that time. The data processing device (1) is obtained by obtaining a teacher signal by going and setting the signal conversion function according to the obtained teacher signal.
Is a control device construction system for constructing a control device for realizing a desired control state, wherein the data processing device (1) includes a control state quantity output from a control target or a control target model, and a control While adopting a configuration in which a difference value with respect to the target control state amount or a value corresponding thereto is input, the virtual target setting device (5) uses the difference value with the control state amount as the control target as a parameter. The a priori knowledge of the data relationship between the control state quantities is managed, the controlled object is operated from the given target value without control input, and the state of the controlled object during that period is sampled. A control device configured to obtain a rule for calculating.