JPH08339207A - Controller - Google Patents

Abstract

PURPOSE: To provide a controller which can substantially eliminate the influence of the dead time fluctuation in a state prediction mode. CONSTITUTION: This device A1 is provided with an observer 3' which estimates the real data on the output based on the input (u) of a controlled system 1 and the sporadic pulse outputs (y) sent from an encoder 2 that is attached to the system 1 and then predicts the state of a control system including the system 1 based on the estimated real data. Then the dead time L that causes the deviation between the real data on the output and its estimated value is corrected based on the state prediction result of the observer 3'. In such a constitution, the fluctuation of the dead time L can be substantially eliminated in a state prediction mode.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control device, and more particularly to a control device equipped with an observer for predicting the state of a control system.

[0002]

2. Description of the Related Art When position and speed data of a controlled object is sampled by an encoder attached to the controlled object, these data become stepwise data depending on the resolution of the encoder. Therefore, if these data are directly used as the input of the controller, the controller performance is limited and the desired control performance cannot be obtained. Therefore, an observer is installed between the encoder and the controller to estimate the actual position and speed data of the controlled object. In this case, if the model is perfect, the observer can obtain an estimated value that is almost the same as the actual data, so that data with higher accuracy than the actual resolution can be sent to the controller. However, in general, the model has parameter errors and parameter fluctuations, and disturbances also enter, so there is a problem that the estimation accuracy by the observer deteriorates. On the other hand, conventionally, the above problem is avoided by increasing the convergence gain or incorporating a disturbance estimation observer (Analysis on estimation error of disturbance torque / speed estimation observer: Akira Shimada, T. Japan Japan, Vol.
113-D, No. 7, '93). FIG. 5 shows a block diagram of such a conventional controller A ₀ . In the figure,
A, B, and C are state coefficients of the controlled object 1, and F is a gain of the observer 3. Observer 3 controls 1
The input value u of
The difference between the output value y from y and the output value of the model is calculated and fed back via the gain F. By that signal,
The model is modified so that the observer 3 has the same dynamic characteristics as the controlled object 1.

[0003]

The conventional control device A ₀ as described above has the following problems. (1) When the model is incomplete and moving at a low speed, the pulses from the encoder 2 become sporadic, and sufficient position and speed data are not input to the observer 3 The estimated value of fluctuates. This is due to fluctuations in dead time, which causes vibration when moving at low speed in motor control, for example. Although the observer 3 considers the dead time, the dead time in the state prediction is constant. Therefore, when the dead time fluctuates, it cannot be dealt with. (2) Further, the speed data obtained from the encoder 2 is calculated from the amount of change in the position data, and therefore the dead time between the position data and the speed data is different,
That is also impossible. In order to solve such a problem in the related art, the first invention is improved in the control device, and even when the dead time changes, the influence is almost always exerted in the state prediction of the control system. The purpose (first purpose) is to provide a control device which does not have such a control device.

Further, when the disturbance estimation observer is built in the conventional control unit A ₀ , there are the following problems. (1) Since the disturbance, the position, and the velocity are estimated simultaneously by the same observer 3, the observer gains in the respective states interfere with each other and cannot be set independently. (2) In addition, the observer 3 is in the transient state immediately after the start of the estimation, and the encoder 2
Disturbance estimation is not performed until the pulse from.
Therefore, the estimated values of position and velocity are affected by the disturbance,
At the initial stage of estimation, the value is far from the true value, and a good estimation result cannot be obtained. In order to solve such a problem in the prior art, a second invention is to improve the control device and to provide a control device which is hardly affected by disturbance in the state prediction of the control system (second Purpose)).

[0005]

In order to achieve the above-mentioned first object, a first invention is to realize the output of an output based on an input to a controlled object and a sporadic pulse output from the controlled object. In a control device equipped with an observer that estimates data and uses the estimated value to predict the state of the control system including the controlled object, the dead time that causes the deviation between the actual data of the output and the estimated value is The control device is configured to perform a correction calculation based on a result of state prediction by the observer. Furthermore, the control device corrects and calculates the dead time by feeding back the deviation between the value obtained by adding the correction term of the dead time to the sporadic pulse output and the estimated value thereof through the gain of the observer. Further, the control device compensates the state coefficient and the gain by incorporating the dead time into the state coefficient and the gain of the observer and correcting and calculating the dead time for each state prediction by the observer. Furthermore,
This is a control device for performing the correction calculation of the dead time for each state variable of the observer. Furthermore, it is a control device that changes the state coefficient of the observer so as to keep the poles of the observer constant.

Further, it is a control device in which a disturbance estimating mechanism is added to the observer. In order to achieve the second object, a second invention is to estimate actual data of the output based on an input to the controlled object and a sporadic pulse output from the controlled object, and calculate the estimated value. In a control device equipped with an observer for predicting the state of a control system including the controlled object, a disturbance estimation mechanism is added to the observer, and the dead time that causes the deviation between the actual data of the output and the estimated value. When the correction calculation is performed on the basis of the state prediction result by the observer, the disturbance estimation value by the disturbance estimation mechanism is input to the state prediction mechanism by the observer. . Furthermore, the control device is characterized in that the state prediction mechanism initializes or corrects the state of the mechanism each time the output is output. Furthermore, the state prediction mechanism is a control device that estimates the position and speed of the controlled object. Furthermore, the state prediction mechanism is a control device that estimates the position and the velocity of the controlled object by maximizing a preset range for each of the position and the velocity of the controlled object.

[0007]

According to the first aspect of the invention, the actual data of the output is estimated by the observer based on the input to the controlled object and the sporadic pulse output from the controlled object. When the state of the control system including the controlled object is predicted, the dead time that causes the deviation between the actual output data and the estimated value is corrected and calculated based on the state prediction result by the observer. Therefore, even when the dead time varies, the influence of the dead time on the state prediction result can be reduced. Further, the deviation between the value obtained by adding the correction term of the dead time to the sporadic pulse output and its estimated value is fed back via the gain of the observer, whereby the dead time is corrected and calculated. In this way, the dead time can be captured in the observer, so that the state can be predicted accurately. Further, the dead time is incorporated into the state coefficient and gain of the observer, and the dead time is corrected and calculated for each state prediction by the observer, whereby the state coefficient and gain are compensated. Therefore, the dead time correction calculation can be performed in a timely manner. Furthermore, if the dead time correction calculation is performed for each state variable of the observer, a more accurate estimation calculation of actual data can be performed. Further, if the state coefficient of the observer is changed so as to keep the pole of the observer constant, the stability of the control system can be secured. Furthermore, if a disturbance estimation mechanism is added to the observer, even if there is a parameter fluctuation of the controlled object, the influence can be absorbed by the disturbance estimation mechanism and more accurate state prediction can be performed. As a result, even if the dead time fluctuates, its influence can be almost eliminated in the state prediction of the control system.

According to the second aspect of the present invention, the disturbance estimating mechanism is added to the observer, and the dead time which causes the deviation between the actual data of the output and the estimated value is calculated based on the result of the state prediction by the observer. When performing correction calculation with
First, the estimated disturbance value is independently calculated by the disturbance estimation mechanism. Then, this calculated value is input to the state prediction mechanism of the observer, and the state estimation calculation other than the disturbance is performed here. Therefore, mutual interference between the disturbance estimation mechanism and the state prediction mechanism can be eliminated. Further, if the state predicting mechanism initializes or corrects the state of the mechanism each time the output is present, the state estimation error at the time when the output is taken into the state predicting mechanism is small. Therefore, the convergence speed in state estimation can be increased. Furthermore, if the state prediction mechanism estimates the position and the velocity by maximizing the preset ranges for the position and the velocity of the controlled object, respectively, the state immediately after the start of the estimation, which is the transient state, is estimated. Position and velocity estimation errors are reduced. Further, it is possible to prevent the estimated value and the true value from being values far apart from each other with respect to unsteady disturbance such as parameter fluctuation. As a result, the influence of disturbance can be almost eliminated in predicting the state of the control system.

[0009]

Embodiments of the present invention will be described with reference to the accompanying drawings.
Examples will be described to provide an understanding of the present invention. In addition,
The following example is an example embodying the present invention.
It does not limit the technical scope of Ming. here
In addition, FIG. 1 relates to an embodiment (first embodiment) of the first invention.
Control device A₁2 is a block diagram showing the schematic configuration of
Control device A₁When a disturbance estimation observer is added to (A
₁′) Is a block diagram showing the schematic configuration, and FIG.
Control device A according to one embodiment (second embodiment) of the present invention₂Of
FIG. 4 is a block diagram showing a schematic configuration, and FIG.₂smell
FIG. 6 is a flowchart showing an operation procedure for initializing a state by
It Incidentally, the conventional control device A shown in FIG.₀In one example
Elements common to the block diagram showing the schematic configuration in
Use the same symbols. <First Invention> One embodiment of the first invention as shown in FIG.
Control device A according to example (first embodiment) ₁Is the controlled object 1
Input u to the control object 1 and the encoder 2 attached to the controlled object 1
Actual data of output based on sporadic pulse output y
Is estimated and the estimated value is used to control the system including the controlled object 1.
Is equipped with an observer 3'that predicts the state of
It is similar to the conventional example. However, in the first embodiment, the above
Dead time that causes the deviation between the actual data of the output y and the estimated value
Based on the result of state prediction by the observer 3 '.
This is different from the conventional example in that a positive operation is performed. Hereafter, this device A₁of
The basic principle of operation is described and the configuration is further
Materialize.

Here, it is assumed that the present apparatus A ₁ is constructed in the control of a motor, for example. The equation of motion of the motor is expressed by the following equation.

[Equation 1] Where m is mass, b is viscosity, u is input, w is disturbance, x
₁ (and x ₂ described later) indicates the position. Also, t is time, dot (.) Is the first-order time derivative, and two dots (.
・) Indicates the second-order time derivative (hereinafter the same). The state space representation of such a system (corresponding to the control system) is as follows.

[Equation 2] Here, A, B, and C are state coefficients, and x (t) is a state. Of these, the state x (t) is unknown, so the state x (t) is estimated using the observer 3 '. However, since the output y (t) cannot be observed, a normal observer cannot be constructed. Therefore, the state after L time z (t) = x (t
Considering + L), it becomes as follows.

(Equation 3)

From this, the output y (t) is given by the following equation.

[Equation 4] Here, consider a value y _L (t) obtained by adding a correction term to the output y (t).

(Equation 5) Then, the following new system is obtained.

(Equation 6)

Therefore, Ce- ^AL becomes a new state coefficient here. In this device A ₁ , as shown in FIG. 1, a value y _L (t) obtained by adding a correction term for the dead time L to the sporadic pulse output actual data y (t) and the value y _L (t) The dead time L is corrected and calculated by feeding back the deviation from the estimated value of 1 through the gain E of the observer 3 '. In this way, a so-called dead time adaptive observer is configured, and the dead time L can be taken into the observer 3 ', so that the state prediction (estimation) can be performed accurately. If the dead time L is sufficiently small, the correction term can be omitted, but in that case, the observer 3 similar to the conventional device A ₀ is constructed. Further, the dead time L is incorporated into the state coefficient Ce- ^AL and the gain E of the observer 3 ', and the dead time L is corrected and calculated for each state prediction by the observer 3', so that the state coefficient and the gain are compensated. In the observer 3 of the conventional apparatus A ₀ , the observer gain F is fixed, but in the present apparatus A ₁ , the dead time L is changed and the observer gain E is changed every time the observer calculation is performed. Therefore, even if the dead time changes, the dead time at that time is compensated, and the fluctuation of the estimated value due to the dead time is reduced.

Further, in the present apparatus A ₁ , if the dead time L is corrected for each state variable z (t) of the observer 3 ', even if there is a dead time determined for each of a plurality of state variables. ， The dead time can be calculated accurately for each. For example, encoder 2
Strictly speaking, since the speed data obtained is calculated from the deviation of the position, the dead time between the position data and the speed data is different. Even if these data come in as inputs to the observer 3 ', it is possible to perform state prediction while correcting and calculating each dead time, and more accurate position and velocity data estimation values can be obtained. Of.

Here, the dead time of the position is L₁， Speed is waste
Time to L₂Then, the state coefficient Ce ^-ALIs as follows
It

(Equation 7) Therefore, the observer 3'may be constructed as follows.

(Equation 8) Here, D is a new state coefficient, and the hat symbol (^) is an estimated value. Further, here, the gain E must be selected so that the state coefficient D becomes asymptotically stable regardless of the variation of the dead time. For example, the eigenvalues of the state coefficient D are λ ₁ , λ ₂ ,
The case where λ ₃ is desired can be achieved by selecting the gain E as follows.

[Equation 9] Although the observer 3'has the position data and the velocity data as inputs, it is also possible to configure an observer having only the position data as an input. In this case, the observer may be configured as follows.

[Equation 10] If the eigenvalues of the state coefficient D at this time are λ ₁ , λ ₂ , and λ ₃ , the observer poles can be arbitrarily arranged by setting the gain E as follows.

[Equation 11] Therefore, if the state coefficient D of the observer is changed so as to keep the pole of the observer 3'constant, the stability of the observer can be secured.

Further, it is conceivable to incorporate a disturbance estimation observer similar to the conventional example (addition of a disturbance estimation mechanism) to the observer 3'of the apparatus A ₁ . This is because in an observer that estimates speed and position, it is generally impossible to accurately estimate speed and position when there are parameter fluctuations (such as inertial fluctuations) to be controlled, so it is necessary to deal with such cases. Because there is. Therefore, by incorporating a disturbance estimation observer in the observer 3'to provide versatility, it is possible to absorb the influence of these fluctuations and perform more accurate state prediction. As a result, in the first embodiment, even if the dead time changes, it is possible to obtain the control device that can almost eliminate the influence in the state prediction.

<Second Invention> FIG. 2 shows a case in which a disturbance estimation observer is built in the observer 3'of the device A ₁ (A
₁ ') is shown. In the device A ₁ ′, the encoder 2
In a low speed range in which there is a dead time L between the output y and the actual data, the input u required for driving the controlled object 1 is
Very small compared to the disturbance. Since 'the observer 3 of' the apparatus A ₁ is observer dead time model including exponential function as shown in the first embodiment, the estimated value is sensitive to disturbances. Moreover, observer 3 '
Since the disturbance, the position, and the velocity are estimated at the same time as in the conventional example, if the disturbance cannot be estimated accurately, the estimated values of the position and the velocity fluctuate due to the influence of the disturbance. If the estimation error is large, the estimated value may oscillate.

In order to avoid such a situation,
It is a second invention shown in. As shown in FIG. 3, the second invention
Control device A according to one embodiment (second embodiment) ₂Is
The input u to the controlled object 1 and the encoder attached to the controlled object 1
Output based on sporadic pulse output y from coder 2
Of the actual data of the control target 1
With an observer 3'for predicting the state of the control system including
This is the same as the conventional example and the first embodiment described above. Shi
However, in the second embodiment, the disturbance estimation is performed in the observer 3 '.
Observer 3_aIf ′ (corresponding to the disturbance estimation mechanism) is added,
Both cause the deviation between the actual data of the output y and the estimated value.
The dead time is used as the result of state prediction by the observer 3 '.
The disturbance estimation observer 3_a′
The disturbance estimation value by the observer 3'state estimation observer
Ba 3_b′ (Corresponding to the state prediction mechanism)
Different from the conventional example. Hereafter, this device A₂To the basic principle of operation
In addition to describing it, the configuration will be further specified.

Here, it is assumed that the device A ₂ is constructed in the control of a motor, for example. The equation of motion of the motor is expressed by the following equation.

(Equation 12) Where m is mass, b is viscosity, u is input, w is disturbance, x
₁ (and x ₂ described later) indicates the position. Also, t is time, dot (.) Is the first-order time derivative, and two dots (.
・) Indicates the second-order time derivative (hereinafter the same). The state space representation of such a system (corresponding to the control system) is as follows.

(Equation 13) Here, A, B, and C are state coefficients, and x (t) is a state. Although the state x (t) is unknown, first, the disturbance estimation observer 3 _a ′ is used to estimate the disturbance w (t). However, the disturbance estimation observer 3 _a ′ does not compensate the dead time.

[Equation 14] Here, D is a new state coefficient, and the hat symbol (^) is an estimated value. Further, here, the gain F must be selected so that the state coefficient D becomes asymptotically stable. Next, disturbance w
Using the estimated value of (t), the input to the state estimation observer 3 _b ′ that estimates the position and speed of the motor is given by the following equation.

(Equation 15) Here, the waveform symbol (~) indicates a conditioned value. The state space expression of the system at this time is given by the following equation.

[Equation 16]

If a dead time adaptive observer is constructed for this system by the same method as in the first embodiment, this becomes the state estimation observer 3 _b ′.

[Equation 17] Here, the gain E must be selected so that the conditional value of the state coefficient D becomes asymptotically stable regardless of the variation of the dead time L. For example, when it is desired to set the eigenvalues of the conditional value of the state coefficient D to λ ₁ and λ ₂ , the poles of the observer can be arbitrarily arranged by setting the gain E as follows.

(Equation 18) By thus constructing the disturbance estimation observer 3 _a ′ and the state estimation observer 3 _b ′, mutual interference between the two can be completely eliminated. Therefore, the apparatus A ₂ can estimate the disturbance from the start of the estimation, and the disturbance estimation value can be used to accurately estimate the state other than the disturbance such as the position and speed of the controlled object. Furthermore, this device A ₂
The state estimation observer 3 _b ′ may be initialized or corrected each time the output encoder pulse enters.

[0020] Here, for example, an observer 3 'state estimation observer whenever the incoming encoder pulses 3 _b'
The case of initializing the state will be described in order of steps S1, S2, ... With reference to FIG. When the observer computation is started, first, the disturbance is estimated by the disturbance estimation observer 3 _a '(S1). Next, it is determined whether or not an encoder pulse has entered the observer 3 '(S
2). When the encoder pulse enters the observer 3 ', the position estimation observer 3 is used as the position and the velocity as the value calculated from the difference between this encoder value and the encoder value when the pulse came last time.
_By substituting for the state of _b ′, the observer 3 _b ′
The state of is initialized (S3). 'If that has not entered the position, the state estimation observer 3 _b for estimating a velocity as it is' encoder pulses observer 3 performs the calculation of (S4). As a result, the state estimation error at the time when the encoder pulse is captured by the state estimation observer 3 _b ′ becomes small, so that the convergence speed in state estimation can be increased. Needless to say, the same effect can be obtained even if appropriate correction is performed instead of initializing the state of the state estimation observer 3 _b ′.

Further, the state estimation observer 3 _b ′ is provided with a so-called limiter function for estimating the position and the velocity of the controlled object by maximizing preset ranges of the position and the velocity of the controlled object, respectively. You can also An example of this limiter function is shown below. (1) Forward rotation of motor to be controlled Estimated position> Encoder value + resolution Estimated position = Encoder value + resolution Estimated speed = Difference between encoder values when a pulse came in last time Estimated position <Encoder value Estimated position = Encoder Value Estimated speed = 0 Estimated speed <0 Estimated speed = 0 (2) Reverse rotation of motor to be controlled Estimated position> Encoder value Estimated position = Encoder value Estimated speed = 0 Estimated position <Encoder value-Resolution Estimated position = Encoder value -Resolution Estimated velocity = Difference between encoder values when the previous pulse came in Estimated velocity> 0 Estimated velocity = 0 This reduces the estimation error of the position and velocity of the control target immediately after the start of estimation in the transient state. Further, it is possible to prevent the estimated value and the true value from being values far apart from each other with respect to unsteady disturbance such as parameter fluctuation. As a result, in the second embodiment, the influence of disturbance can be almost eliminated in the state prediction of the control system.

[0022]

Since the control device according to the first aspect of the invention is configured as described above, it is possible to reduce the influence of the dead time on the state prediction result even when the dead time changes. Further, the deviation between the value obtained by adding the correction term of the dead time to the sporadic pulse output and its estimated value is fed back via the gain of the observer, whereby the dead time is corrected and calculated. In this way, the dead time can be captured in the observer, so that the state can be predicted accurately. Further, the dead time is incorporated into the state coefficient and gain of the observer, and the dead time is corrected and calculated for each state prediction by the observer, whereby the state coefficient and gain are compensated. Therefore, the dead time correction calculation can be performed in a timely manner. Furthermore, if the dead time correction calculation is performed for each state variable of the observer, a more accurate estimation calculation of actual data can be performed. Further, if the state coefficient of the observer is changed so as to keep the pole of the observer constant, the stability of the control system can be secured. Furthermore, if a disturbance estimation mechanism is added to the observer, even if there is a parameter fluctuation of the controlled object, the influence can be absorbed by the disturbance estimation mechanism and more accurate state prediction can be performed. As a result, even if the dead time fluctuates, its influence can be almost eliminated in the state prediction of the control system.

Further, in the control device according to the second aspect of the present invention, mutual interference between the disturbance estimating mechanism and the state predicting mechanism can be eliminated. Further, if the state predicting mechanism initializes or corrects the state of the mechanism each time the output is present, the state estimation error at the time when the output is taken into the state predicting mechanism is small. Therefore, the convergence speed in state estimation can be increased. Furthermore, if the state prediction mechanism estimates the position and the velocity by maximizing the preset ranges for the position and the velocity of the controlled object, respectively, the state immediately after the start of the estimation, which is the transient state, is estimated. Position and velocity estimation errors are reduced. Further, it is possible to prevent the estimated value and the true value from being values far apart from each other with respect to unsteady disturbance such as parameter fluctuation. As a result, the influence of disturbance can be almost eliminated in predicting the state of the control system.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a control device A ₁ according to an embodiment (first embodiment) of the first invention.

FIG. 2 is a block diagram showing a schematic configuration in the case where a disturbance estimation observer is added to the control device A ₁ (A ₁ ′).

FIG. 3 is a block diagram showing a schematic configuration of a control device A ₂ according to an embodiment (second embodiment) of the second invention.

FIG. 4 is a flowchart showing an operation procedure when the control device A ₂ initializes a state.

FIG. 5 is a block diagram showing a schematic configuration of an example of a conventional control device A ₀ .

[Explanation of symbols]

A _1, A ₂ ... controller 1 ... controlled target 2 ... encoder 3 '... observer 3 _a' ... (to the disturbance estimator) the disturbance estimation observer 3 _b '... state estimation observer (corresponds to the state predictor)

Claims (10)

[Claims] 1. The state of a control system including the controlled object is estimated by estimating actual data of the output based on an input to the controlled object and a sporadic pulse output from the controlled object. In a controller equipped with an observer that predicts
A control device, wherein a dead time, which is a cause of a deviation between the actual data of the output and an estimated value, is corrected and calculated based on a result of state prediction by the observer. 2. The dead time is corrected and calculated by feeding back a deviation between a value obtained by adding a correction term for the dead time to the sporadic pulse output and an estimated value thereof through a gain of the observer. The control device described. 3. The state coefficient and gain are compensated by incorporating the dead time into the state coefficient and gain of the observer, and correcting the dead time for each state prediction by the observer to correct the state coefficient and gain. Control device. 4. The control device according to claim 1, wherein the dead time correction calculation is performed for each state variable of the observer. 5. The control device according to claim 1, wherein the state coefficient of the observer is changed so as to keep the poles of the observer constant. 6. The control device according to claim 1, wherein a disturbance estimation mechanism is added to the observer. 7. A state of a control system including the controlled object is estimated by estimating actual data of the output based on an input to the controlled object and a sporadic pulse output from the controlled object. In a controller equipped with an observer that predicts
A disturbance estimation mechanism is added to the observer, and when the dead time that causes the deviation between the actual output data and the estimated value is corrected and calculated based on the state prediction result by the observer, the disturbance estimation mechanism is used. A control device characterized in that the estimated disturbance value is input to a state prediction mechanism of the observer. 8. The control device according to claim 7, wherein the state predicting mechanism initializes or corrects the state of the mechanism each time the output is output. 9. The control device according to claim 7, wherein the state prediction mechanism estimates the position and speed of the controlled object. 10. The state predicting mechanism estimates a position and a velocity of the controlled object by maximizing a preset range for the position and the velocity of the controlled object, respectively. Control device.