Control device

Abstract

The present invention provides a control device that executes model predictive control related to a predetermined control target corresponding to an actual target device which is an actual target of servo control in order to cause an output of the actual target device to track a predetermined command. The control device includes: an integrator which receives input of a deviation between the predetermined command and an output of the predetermined control target; and a model predictive control unit which has a prediction model that defines a correlation between a predetermined state variable and an input to the predetermined control target in a form of a predetermined state equation, and which performs model predictive control based on the prediction model according to a predetermined evaluation function in a prediction section having a predetermined time width using the output of the integrator as an input.

Claims

1. A control device that executes model predictive control related to a predetermined control target corresponding to an actual target device which is an actual target of servo control in order to cause an output of the actual target device to track a predetermined command, the control device comprising: an integrator which receives input of a deviation between the predetermined command and an output of the predetermined control target; a state acquisition unit that acquires a value of a predetermined state variable related to the predetermined control target; a model predictive control unit which has a prediction model that defines a correlation between the predetermined state variable and an input to the predetermined control target in a form of a predetermined state equation, and which performs the model predictive control based on the prediction model according to a predetermined evaluation function in a prediction section having a predetermined time width using the output of the integrator as an input, and outputs a value of the input at least at an initial time point of the prediction section as an input to the predetermined control target corresponding to the predetermined command, and a gain adjustment unit that adjusts a predetermined integral gain so that the predetermined integral gain increases as a magnitude of the deviation decreases, wherein the prediction model includes a predetermined integral term represented by a product of the predetermined integral gain and the deviation between the output of the predetermined control target and the predetermined command.

2. The control device according to claim 1, wherein the predetermined control target comprises the actual target device itself, the control device further comprises a feedback system that feeds back the output of the actual target device, the integrator further receives the input of the deviation between the predetermined command and the output of the actual target device fed back by the feedback system, the state acquisition unit further acquires a value of the predetermined state variable related to the actual target device, and the prediction model comprises a model that defines a correlation between the predetermined state variable and the input to the actual target device, and the predetermined integral term included in the prediction model is represented by a product of the predetermined integral gain and a deviation between the output of the actual target device and the predetermined command.

3. The control device according to claim 1, further comprising: an actual target model control unit which includes an actual target model which is a model that models the actual target model and which is the predetermined control target and which simulates and outputs an output of the actual target device, wherein the output of the actual target model control unit is configured to be supplied to the actual target device, a deviation between the predetermined command and the output of the actual target model control unit is input to the integrator, the state acquisition unit further acquires a value of the predetermined state variable related to the actual target model included in the actual target model control unit, and the prediction model comprises a model that defines a correlation between the predetermined state variable and the input to the actual target model control unit, and the predetermined integral term included in the prediction model is represented by a product of the predetermined integral gain and a deviation between the output of the actual target model and the predetermined command.

4. The control device according to claim 1, wherein the gain adjustment unit further adjusts the predetermined integral gain such that the predetermined integral gain increases as the magnitude of the deviation decreases when the value of the deviation belongs to a predetermined first range including zero and sets the predetermined integral gain to zero when the value of the deviation does not belong to the predetermined first range.

5. The control device according to claim 4, wherein the predetermined control target has a two-dimensional output, the predetermined first range is defined by a downwardly convex function f(x), and the predetermined integral gain is represented by a function of α(|f(x)|+f(x)), where α is a predetermined coefficient.

6. The control device according to claim 4, wherein the predetermined control target has a two-dimensional output, the predetermined first range is defined by an upwardly convex function f(x), and the predetermined integral gain is represented by a function of a α(|f(x)|+f(x)) where α is a predetermined coefficient.

7. The control device according to claim 1, wherein the predetermined control target has a plurality of control axes, a command to the predetermined control target, an input to the predetermined control target, and an output to the predetermined control target are correlated with the plurality of control axes, and the prediction model is defined by the predetermined state equation so as to correspond to each of the plurality of control axes and includes a plurality of the predetermined integral terms corresponding to the plurality of control axes.

8. The control device according to claim 7, wherein the gain adjustment unit further adjusts the predetermined integral gain corresponding to each of the plurality of control axes according to the magnitude of the deviation corresponding to each of the plurality of control axes, and the gain adjustment unit further adjusts the predetermined integral gain corresponding to each of the plurality of control axes so as to increase as a relative magnitude of the deviation corresponding to each of the plurality of control axes increases.

9. The control device according to claim 7, wherein a predetermined working coordinate system is set in the predetermined control target on the basis of the plurality of control axes, the gain adjustment unit further adjusts the predetermined integral gain corresponding to each of the plurality of control axes according to a magnitude in the predetermined working coordinate system, of the deviation corresponding to each of the plurality of control axes, and the gain adjustment unit further adjusts the predetermined integral gain corresponding to each of the plurality of control axes so as to increase as a relative magnitude of the deviation converted to the predetermined working coordinate system, corresponding to each of the plurality of control axes increases.

10. The control device according to claim 1, wherein the gain adjustment unit further calculates the predetermined integral gain according to a gain setting function that is differentiable on the basis of the deviation.

11. The control device according to claim 1, wherein the gain adjustment unit includes a gain setting function related to the deviation, for calculating the predetermined integral gain, and is further configured to calculate, when the gain setting function is differentiable on the basis of the deviation, the predetermined integral gain according to the gain setting function, and perform, when the gain setting function is not differentiable on the basis of the deviation, a predetermined stabilization process for stabilizing an arithmetic process of the model predictive control by the model predictive control unit and calculates the predetermined integral gain according to the gain setting function.

12. The control device according to claim 1, wherein the prediction model includes a high pass filter processing term represented by a product of the output of the integrator and a predetermined filtering gain correlated with a high-pass filtering process on the deviation in addition to the predetermined integral term wherein the predetermined integral gain is one, when the value of the deviation is outside a predetermined second range including zero, the predetermined filtering gain is set such that a cutoff frequency decreases as the magnitude of the deviation approaches a boundary of the predetermined second range in the high-pass filtering process, and when the value of the deviation belongs to the predetermined second range, the predetermined filtering gain is set to zero.

Description

BRIEF DESCRIPTION OF DRAWINGS

(1) FIG. 1 is a diagram illustrating a schematic configuration of a control system configured to include a control device according to the present invention.

(2) FIG. 2 is a first diagram illustrating a control structure of a servo driver included in the control system illustrated in FIG. 1.

(3) FIG. 3 is a first diagram illustrating a correlation between a deviation and a predetermined integral gain.

(4) FIG. 4 is a second diagram illustrating a correlation between a deviation and a predetermined integral gain.

(5) FIGS. 5A and 5B are third diagrams illustrating a correlation between a deviation and a predetermined integral gain.

(6) FIG. 6 is a diagram illustrating a state in which a working coordinate system is set in a plant.

(7) FIGS. 7A and 7B are first diagrams illustrating the results of trackability when two control axes are servo-controlled by a servo driver included in the control system illustrated in FIG. 1.

(8) FIGS. 8A and 8B are second diagrams illustrating the results of trackability when two control axes are servo-controlled by a servo driver included in the control system illustrated in FIG. 1.

(9) FIGS. 9A and 9B are third diagrams illustrating the results of trackability when two control axes are servo-controlled by a servo driver included in the control system illustrated in FIG. 1.

(10) FIGS. 10A and 10B are fourth diagrams illustrating the results of trackability when two control axes are servo-controlled by a servo driver included in the control system illustrated in FIG. 1.

(11) FIG. 11 is a second diagram illustrating a control structure of a servo driver included in the control system illustrated in FIG. 1.

(12) FIGS. 12A and 12B are diagrams illustrating the results of trackability when two control axes are servo-controlled by a servo driver illustrated in FIG. 11.

(13) FIG. 13 is a diagram illustrating a control structure of a standard PLC included in a control system illustrated in FIG. 1.

DESCRIPTION OF EMBODIMENTS

First Embodiment

(14) FIG. 1 is a diagram illustrating a schematic configuration of a control system including a servo driver 4 which is a control device according to an embodiment of the present invention. The control system includes a network 1, a servo driver 4, and a standard PLC (Programmable Logic Controller) 5. The servo driver 4 is a control device for servo-controlling a plant 6 including a motor 2 and a load device 3. In the control system, the servo driver 4 servo-controls the plant 6 so that the plant 6 tracks a command signal generated by the standard PLC 5. In this case, examples of the load device 3 that constitutes the plant 6 include various mechanisms (e.g., arms of industrial robots and conveyance devices), and the motor 2 is incorporated in the load device 3 as an actuator that drives the load device 3. For example, the motor 2 is an AC servo motor. An encoder (not illustrated) is attached to the motor 2, and a parameter signal (a position signal, a velocity signal, and the like) related to an operation of the motor 2 is transmitted to the servo driver 4 as a feedback signal by the encoder.

(15) The standard PLC 5 generates a command signal related to an operation (motion) of the plant 6 and transmits the same to the servo driver 4. The servo driver 4 receives the command signal from the standard PLC 5 via the network 1 and receives the feedback signal output from the encoder connected to the motor 2. The servo driver 4 supplies a driving current to the motor 2 on the basis of the command signal and the feedback signal from the encoder so that servo control related to driving of the plant 6 is performed (that is, the output of the plant 6 tracks the command). The supplied current is an AC power delivered from an AC power supply 7 to the servo driver 4. In the present embodiment, the servo driver 4 is a type that receives a three-phase AC power but may be a type that receives a single-phase AC power. As the servo control of the servo driver 4, model predictive control of a model predictive control unit 43 is executed as illustrated in FIG. 2.

(16) A control structure in the servo driver 4 will be described with reference to FIG. 2. The servo driver 4 includes an integrator 41, a state acquisition unit 42, a model predictive control unit 43, and a gain adjustment unit 44. A deviation e (e=r−y) between a command r transmitted from the standard PLC 5 and an output y of the plant 6 fed back by a feedback system 45 is input to the integrator 41. An output z of the integrator 41 is input to the model predictive control unit 43 through the state acquisition unit 42.

(17) In this case, the state acquisition unit 42 acquires the value of a state variable included in a state x related to the plant 6, provided to model predictive control performed by the model predictive control unit 43. For example, the state acquisition unit 42 can acquire predetermined information generated from an output signal of the encoder connected to the motor 2 included in the plant 6 as the state variable included in the state x. Moreover, a predetermined parameter (for example, the position of an output unit of the load device 3 or the like) related to the load device 3 included in the plant 6 may be acquired as a state variable included in the state x. Furthermore, in the present embodiment, the state acquisition unit 42 also acquires the output z of the integrator 41 as the state variable included in the state x. The model predictive control unit 43 executes model predictive control (Receding Horizon control) using the state x related to the plant 6 acquired by the state acquisition unit 42 and an input u to the plant 6 input by the model predictive control unit 43.

(18) Specifically, the model predictive control unit 43 has a prediction model that defines a correlation between the state x related to the plant 6 and the input u to the plant 6 using a state equation (Equation 1) below. Equation 1 below is a nonlinear state equation.
[Math. 1]
{dot over (x)}(t)=P(x(t),u(t))   (Equation 1)

(19) Here, the model predictive control unit 43 performs model predictive control based on the prediction model represented by Equation 1 according to an evaluation function represented by Equation 2 below in a prediction section having a predetermined time width T using the state x related to the plant 6 and the input u to the plant 6 as an input.
[Math. 2]
J=φ(x(t+T))+∫.sub.t.sup.t+TL(x(τ),u(τ))dτ   (Equation 2)

(20) The value of the input u at an initial time point t of the prediction section, calculated in the model predictive control is output as the input u to the plant 6 corresponding to the command r at that time point t. In model predictive control, a prediction section having a predetermined time width T is set at each control timing thereof, and the input u to the plant 6 at that control timing is calculated according to the evaluation function of Equation 2 and is transmitted to the plant 6. A problem of calculating an operation amount that optimizes a value of an evaluation function J having such a form as in Equation 2 is a problem widely known as an optimal control problem, and an algorithm for calculating a numerical solution thereof is disclosed as a known technique. A continuous deformation method can be exemplified as such a technique, and the details thereof are disclosed, for example, in “A continuation/GMRES method for fast computation of nonlinear receding horizon control”, T. Ohtsuka, Automatica, Vol. 40, pp. 563-574 (2004), for example.

(21) In a continuous deformation method, an input U(t) to model predictive control is calculated by solving a simultaneous linear equation related to the input U(t) illustrated in Equation 3 below. Specifically, Equation 3 is solved and dU/dt is numerically integrated to update the input U(t). In this manner, since the continuous deformation method does not perform repeated computation, it is possible to suppress a computational load for calculating the input U(t) at each time point as much as possible.

(22) $\begin{array}{cc}\left[\mathrm{Math}.3\right]& \\ \frac{\partial F}{\partial U}\stackrel{.}{U}=-\zeta F-\frac{\partial F}{\partial x}\stackrel{.}{x}-\frac{\partial F}{\partial t}& \left(\mathrm{Equation}3\right)\end{array}$

(23) Where F and U(t) are represented by Equation 4 below.

(24) $\begin{array}{cc}\left[\mathrm{Math}.4\right]& \\ F\left(U\left(t\right),x\left(t\right),t\right)=\left[\begin{array}{c}\frac{\partial H}{\partial u}\left(x_0^{*}\left(t\right),u_0^{*}\left(t\right),{\lambda }_1^{*}\left(r\right),{\mu }_{0^{*}}\left(t\right)\right)\\ C\left(x_0^{*}\left(t\right),u_0^{*}\left(t\right)\right)\\ \mathrm{.Math.}\\ \frac{\partial H}{\partial u}\left(x_{N\_1}^{*}\left(t\right),u_{N\_1}^{*}\left(t\right),{\lambda }_N^{*}\left(t\right),{\mu }_{N\_1}^{*}\left(t\right)\right)\\ \begin{array}{cc}& C\left(x_{N\_1}^{*}\left(t\right),u_{N\_1}^{*}\left(t\right)\right)\end{array}\end{array}\right]U\left(t\right)=\left[u_0^{*T}\left(t\right),{\mu }_0^{*T}\left(t\right),\mathrm{.Math.},u_{N\_1}^{*T}\left(t\right),u_{N\_1}^{*T}\left(t\right)\right]& \left(\mathrm{Equation}4\right)\end{array}$

(25) Where H is a Hamiltonian, λ is a costate, and μ is a Lagrange multiplier of constraints C=0.

(26) Here, in the present embodiment, as illustrated in FIG. 2, a deviation e between the command r and the output y of the plant 6 is input to the integrator 41, and the output z of the integrator 41 is input to the model predictive control unit 43 through the state acquisition unit 42, whereby the model predictive control is performed. A prediction model included in the model predictive control unit 43 based on the control structure including the integrator 41 in this manner is formed as illustrated in Equation 5 below, for example. Predetermined physical characteristics of the plant 6 are applied to Equation 5.

(27) $\begin{array}{cc}\left[\mathrm{Math}.5\right]& \\ \begin{array}{c}\stackrel{.}{x}\left(t\right)=P\left(x\left(t\right),u\left(t\right)\right)\\ =\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ {\stackrel{\mathrm{.Math.}}{x}}_1\\ {\stackrel{.}{z}}_1\end{array}\right]=\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ \frac{u_1}{M_1}\\ \left(x_{f1}-x_1\right).Math.K_{i1}\end{array}\right]\end{array}& \left(\mathrm{Equation}5\right)\end{array}$

(28) A subscript “1” in Equation 5 indicates the number of control axes controlled by the servo driver 4, and in the present embodiment, since the number of control axes is one, the subscripts of the respective variables of the prediction model illustrated in Equation 5 are “1”. A variable x1 of a state x indicates an output position of the plant 6 and is predetermined fed back as the output y of the plant 6. Moreover, x.sub.f1 indicates a position command r to the control axis. Therefore, (x.sub.f1−x.sub.1) in the prediction model indicates the deviation e. It can be understood that the prediction model includes an integral term represented by a product of the deviation e (=x.sub.f1−x.sub.1) and a predetermined integral gain K.sub.i1. When the integral term is included in the prediction model in this manner, a control structure in which another virtual integrator separate from the integrator 41 is connected in parallel can be obtained in the servo driver 4. In this way, it is easy to adjust an integration amount serving as a driving source of the servo control performed by the servo driver 4 which uses model predictive control, and it is easy to realize servo control suppressing an overshoot without using a disturbance observer which requires difficult adjustment such as extension of a disturbance model or design of an observer gain as in the conventional technique.

(29) Furthermore, the control structure of the servo driver 4 includes a gain adjustment unit 44. The gain adjustment unit 44 is a functional unit that adjusts a predetermined integral gain K.sub.i1 of an integral term included in the prediction model illustrated in Equation 5 on the basis of the deviation e. Specifically, as illustrated in FIG. 3, the gain adjustment unit 44 adjusts the predetermined integral gain K.sub.i1 so that the value of the predetermined integral gain K.sub.i1 increases as the magnitude of the deviation e decreases. Particularly, in FIG. 3, when the magnitude of the deviation e is equal to or larger than e0, the predetermined integral gain K.sub.i1 is zero, and a value larger than zero and equal to or smaller than one is set as the predetermined integral gain K.sub.i1 in a range in which the magnitude of the deviation e is smaller than e0. Moreover, the value of the predetermined integral gain K.sub.i1 approaches abruptly one as the magnitude of the deviation e approaches zero, and the transition of the predetermined integral gain K.sub.i1 is set so that the predetermined integral gain K.sub.i1 is one when the magnitude of the deviation e is zero. In this manner, since the gain adjustment unit 44 can adjust the predetermined integral gain K.sub.i1 on the basis of the magnitude of the deviation e, when an output y(x.sub.1) of the plant 6 deviates far from a command x.sub.f1, the value of the predetermined integral gain K.sub.i1 is adjusted to be small, and an integration amount for servo control is adjusted so as not to accumulate. Moreover, since the value of the predetermined integral gain K.sub.i1 is adjusted to be large by the gain adjustment unit 44 when the amount of deviation of the output y(x.sub.1) of the plant 6 from the command x.sub.f1 decreases (that is, the magnitude of the deviation e decreases), it is possible to effectively enhance the trackability of servo control. By varying the value of the predetermined integral gain K.sub.i1 in this manner, it is possible to achieve both suppression of an overshoot and improvement in the trackability of servo control.

(30) In adjustment of the predetermined integral gain K.sub.i1 by the gain adjustment unit 44, data related to correlation between the deviation e and the predetermined integral gain K.sub.i1 illustrated in FIG. 3 may be stored in the memory of the servo driver 4, and in that case, the gain adjustment unit 44 adjust the predetermined integral gain K.sub.i1 by accessing the data. Moreover, as another method, the gain adjustment unit 44 may calculate the predetermined integral gain K.sub.i1 according to a function for calculating the predetermined integral gain K.sub.i1 illustrated in Equation 6 below.
[Math. 6]
K.sub.i1=1/exp(e.sup.2×C1) or
K.sub.i1=1/exp(|e|)×C1)   (Equation 6)

(31) In Equation 6, a constant C1 is a coefficient for setting calculation sensitivity of the predetermined integral gain K.sub.i1 with respect to the deviation e. For example, as illustrated in FIG. 4, when the value of the constant C1 decreases, change in the value of the predetermined integral gain K.sub.i1 with respect to change in the magnitude of the deviation e becomes gentle. This means that, as the value of the constant C1 decreases, the predetermined integral gain K.sub.i1 is set so as to correspond to a wider range of the deviation e. The constant C1 can be set appropriately on the basis of the control characteristics or the like of the plant 6, and for example, when an overshoot is likely to occur in the plant 6, it is preferable that the value of the constant C1 is set to be larger in order to make an integral term difficult to accumulate so as to suppress the overshoot. Moreover, the calculation function illustrated in Equation 6 is a function that is differentiable on the basis of the deviation e, and therefore, it is possible to stably perform model predictive control which uses the prediction model including the predetermined integral gain K.sub.i1.

(32) As still another method, in adjustment of the predetermined integral gain K.sub.i1 by the gain adjustment unit 44, correlation between the deviation e and the predetermined integral gain K.sub.i1 may be set as illustrated in FIGS. 5A and 5B. According to correlation illustrated in FIG. 5A, the predetermined integral gain K.sub.i1 is one when the magnitude of the deviation e is zero, the predetermined integral gain K.sub.i1 decreases linearly up to zero until the deviation e reaches e0, and the predetermined integral gain K.sub.i1 is adjusted to zero when the deviation e becomes equal to or larger than e0. Moreover, according to correlation illustrated in FIG. 5B, the predetermined integral gain K.sub.i1 is one when the magnitude of the deviation e is zero, the predetermined integral gain K.sub.i1 decreases inproportionally up to zero until the deviation e reaches e0, and the predetermined integral gain K.sub.i1 is adjusted to zero when the deviation e becomes equal to or larger than e0. In the transition of the predetermined integral gain K.sub.i1 illustrated in both correlations illustrated in FIGS. 5A and 5B, the predetermined integral gain K.sub.i1 is not differentiable when the deviation e is near zero. In such a case, it is preferable to take known countermeasures for preventing an arithmetic process of the model predictive control unit 43 from becoming unstable (that is, countermeasures for preventing abrupt change in a partial differential value of the Hamiltonian H such as for example, setting the value of the predetermined integral gain K.sub.i1 to one and δH/δx=0).

(33) <Modification>

(34) Here, in the control system illustrated in FIG. 1, a direct control target of the servo driver 4 is the motor 2. When the output of the plant 6 that is to track a command is not a rotation angle of the motor 2 but is the position of an output unit of the load device 3 driven by the motor 2, a coordinate system related to a rotation axis of the motor 2 is not identical to a coordinate system (that is, a working coordinate system) related to the output unit of the load device 3. In such a case, it is preferable to perform adjustment for servo control of the output of the plant 6 immediately (particularly, adjustment of the predetermined integral gain included in a prediction model) by defining the prediction model of the model predictive control unit 43 on the basis of the working coordinate system.

(35) For example, as illustrated in FIG. 6, when a rod-shaped load device 3 is rotated by rotation of the motor 2 and a position in a horizontal direction of a tip thereof is the output of the plant 6, a motion equation related to the load device 3 can be represented by Equation 7 below. It is assumed that a tip position of the load device 3 is fed back to the model predictive control unit 43 as a state variable included in the state x. In this modification, this tip position is represented as “x1” for the sake of convenience.
[Math. 7]
M{umlaut over (θ)}+C{dot over (θ)}=U.sub.q   (Equation 7)

(36) Where Uq is a driving torque for rotating the load device 3. Equation 8 below represents correlation between a rotation angle θ of the motor 2 and the tip position x1 of the load device 3.
[Math. 8]
{umlaut over (x)}.sub.1={dot over (J)}{dot over (θ)}+J{umlaut over (θ)}
{dot over (θ)}=J.sup.−1{dot over (x)}.sub.1   (Equation 8)

(37) Where J represents a Jacobian matrix. In the case illustrated in FIG. 6, the Jacobian matrix is represented by Equation 9 below.

(38) $\begin{array}{cc}\left[\mathrm{Math}.9\right]& \\ J=-l_1\sin \left({\theta }_1\right)=-\sqrt{l_1^2-x_1^2}\stackrel{.}{J}=\frac{x_1}{\sqrt{l_1^2-x_1^2}}& \left(\mathrm{Equation}9\right)\end{array}$

(39) Equation 10 below is obtained on the basis of Equations 7 to 9.

(40) $\begin{array}{cc}\left[\mathrm{Math}.10\right]& \\ {\stackrel{\mathrm{.Math.}}{x}}_1=-\frac{x_1{\stackrel{.}{x}}_1}{\sqrt{l_1^2-x_1^2}}+M^{-1}\left(-C{\stackrel{.}{x}}_1+p^{-1}u\right)M=\frac{m_1{l}_1^2}{3}& \left(\mathrm{Equation}10\right)\end{array}$

(41) From the above, a prediction model of this modification can be represented by Equation 11 below by taking Equation 10 and the integral term represented by the product of a deviation and a predetermined integral gain into consideration.

(42) $\begin{array}{cc}\left[\mathrm{Math}.11\right]& \\ \begin{array}{c}\stackrel{.}{x}\left(t\right)=P\left(x\left(t\right),u\left(t\right)\right)\\ =\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ {\stackrel{\mathrm{.Math.}}{x}}_1\\ {\stackrel{.}{z}}_1\end{array}\right]=\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ -\frac{x_1{\stackrel{.}{x}}_1}{l_1^2-x_1^2}+M^{-1}\left(-C{\stackrel{.}{x}}_1+u\right)\\ \frac{x_{f1}-x_1}{\exp \left(.Math.x_{f1}-x_1.Math.*K_r\right)}\end{array}\right]\end{array}& \left(\mathrm{Equation}11\right)\end{array}$

(43) Where Kr is a constant corresponding to the constant C1 in Equation 6.

Second Embodiment

(44) Although one control axis is included in the plant 6 in the control system illustrated in FIG. 1, still another control axis may be included and the servo driver 4 may perform servo control on the plant 6 including a plurality of control axes. In such a case, a prediction model included in the model predictive control unit 43 can be represented by Equation 12 below. The prediction model represented by Equation 12 corresponds to a case in which two control axes are included in the plant 6.

(45) $\begin{array}{cc}\left[\mathrm{Math}.12\right]& \\ \begin{array}{c}\stackrel{.}{x}\left(t\right)=P\left(x\left(t\right),u\left(t\right)\right)\\ =\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ {\stackrel{.}{x}}_2\\ {\stackrel{\mathrm{.Math.}}{x}}_1\\ {\stackrel{\mathrm{.Math.}}{x}}_2\\ {\stackrel{.}{z}}_1\\ {\stackrel{.}{z}}_2\end{array}\right]=\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ {\stackrel{.}{x}}_2\\ \frac{u_1}{M_1}\\ \frac{u_2}{M_2}\\ \left(x_{f1}-x_1\right).Math.K_{i1}\\ \left(x_{f2}-x_2\right).Math.K_{i2}\end{array}\right]\end{array}& \left(\mathrm{Equation}12\right)\end{array}$

(46) Subscripts in Equation 12 correspond to control axes similarly to Equation 5. Variables x1 and x2 in the state x represent the output positions of the first and second control axes of the plant 6 and are parameters fed back as the output y of the plant 6. Moreover, x.sub.f1 and x.sub.f2 represent the position commands r to the control axes. Therefore, (x.sub.f1−x.sub.1) and (x.sub.f2−x.sub.2) in the prediction model represent the deviations e of the control axes. It can be understood that the prediction model includes integral terms represented by the products between the deviations e and the predetermined integral gains K.sub.i1 and K.sub.i2 corresponding to the respective control axes. When the integral terms are included in the prediction model, control structures in which another virtual integrator separate from the integrator 41 is connected in parallel, corresponding to the respective control axes can be obtained in the servo driver 4. In this way, it is easy to adjust an integration amount serving as a driving source of the servo control performed by the servo driver 4 which uses model predictive control, and it is easy to realize servo control suppressing an overshoot.

(47) FIGS. 7A and 7B illustrate a diagram that traces the trajectory of a robot arm made up of two control axes when the servo driver 4 having the above-described configuration servo-controls an operation of bending the arm at 90 degrees. FIG. 7A illustrates the results of a control operation of bending a robot arm according to the conventional technique, and model predictive control is performed according to the conventional technique, a servo driver is generally configured as a regulator in order to avoid deterioration of a transient response such as an overshoot. Therefore, if a continuous disturbance occurs, a steady-state deviation is likely to occur due to the influence of the disturbance. In contrast, in the servo control performed by the servo driver 4 of the present embodiment illustrated in FIG. 7B, a steady-state deviation is eliminated effectively, and trackability to commands is improved without causing any overshoot.

(48) Here, when the servo driver 4 performs servo control of a plurality of control axes, the predetermined integral gains adjusted by the gain adjustment unit 44 in the model predictive control corresponding to the respective control axes are preferably adjusted by taking the trackability to the commands at the respective control axes into consideration. The overall output of the plant 6 can be optimized by taking a balance of the trackabilities between the control axes into consideration. Therefore, the following two modes will be illustrated as a method for adjusting the predetermined integral gain by taking a balance of the trackabilities between the control axes into consideration. In the present embodiment, the number of control axes serving as the target of the servo control of the servo driver 4 is two.

(49) <First Gain Adjustment Mode>

(50) When the prediction model of the model predictive control unit 43 is represented by Equation 12, a predetermined integral gain K.sub.i1 corresponding to a first control axis and a predetermined integral gain Kit corresponding to a second control axis are calculated on the basis of Equation 13 below.

(51) $\begin{array}{cc}\left[\mathrm{Math}.13\right]& \\ K_{i1}=\frac{.Math.x_{f1}-x_1.Math.}{.Math.x_{f1}-x_1.Math.+.Math.x_{f2}-x_2.Math.}{K}_{i2}=\frac{.Math.x_{f2}-x_2.Math.}{.Math.x_{f1}-x_1.Math.+.Math.x_{f2}-x_2.Math.}\mathrm{where}{K}_i=1\mathrm{for}.Math.x_{f1}-x_1.Math.+.Math.x_{f2}-x_2.Math.=0& \left(\mathrm{Equation}13\right)\end{array}$

(52) The adjustment of the predetermined integral gains K.sub.i1 and Kia according to Equation 13 means adjustment according to the magnitude of the deviation e corresponding to each control axis. Specifically, when a deviation in one control axis is relatively larger than a deviation in the other control axis, the values of the predetermined integral gains K.sub.i1 and Kit are determined according to the relative proportion. That is, the predetermined integral gains are adjusted such that the value of the corresponding integral gain increases as the magnitude of the relative deviation increases. This means that, the larger the deviation of a control axis, the larger becomes the predetermined integral gain, and therefore, the trackabilities to commands can be enhanced by accumulating an integration amount. As a result, it is possible to equalize the final trackabilities to commands between control axes.

(53) As another method, the predetermined integral gain K.sub.i1 corresponding to the first control axis and the predetermined integral gain K.sub.i2 corresponding to the second control axis can be calculated on the basis of Equation 14 below.

(54) $\begin{array}{cc}\left[\mathrm{Math}.14\right]& \\ K_{i1}=\frac{.Math.\frac{x_{f1}-x_1}{x_{f1}-x_{s1}}.Math.}{.Math.\frac{x_{f1}-x_1}{x_{f1}-x_{s1}}.Math.+.Math.\frac{x_{f2}-x_2}{x_{f2}-x_{s2}}.Math.}{K}_{i1}=\frac{.Math.\frac{x_{f2}-x_2}{x_{f2}-x_{s2}}.Math.}{.Math.\frac{x_{f1}-x_1}{x_{f1}-x_{s1}}.Math.+.Math.\frac{x_{f2}-x_2}{x_{f2}-x_{s2}}.Math.}\mathrm{where}{K}_i=1\mathrm{for}.Math.\frac{x_{f1}-x_1}{x_{f1}-x_{s1}}.Math.+.Math.\frac{x_{f2}-x_2}{x_{f2}-x_{s2}}.Math.=0& \left(\mathrm{Equation}14\right)\end{array}$

(55) x.sub.s1 and x.sub.s2 in Equation 14 are target positions of the control axes, and Equation 14 is a calculation formula for adjusting the predetermined integral gains K.sub.i1 and K.sub.i2 according to the magnitude of a deviation percentage to targets corresponding to the respective control axes. Therefore, according to Equation 14, in each control axis, the value of the corresponding predetermined integral gain is adjusted to be large as the magnitude of the relative deviation increases, and therefore, it is possible to equalize the final trackabilities to commands between control axes.

(56) <Second Gain Adjustment Mode>

(57) A prediction model of the model predictive control unit 43 is as illustrated in Equation 12. In the second gain adjustment mode, a predetermined working coordinate system based on the first and second control axes is set in the plant 6. The predetermined integral gains K.sub.i1 and K.sub.i2 corresponding to the respective control axes are adjusted on the basis of deviations of the respective control axes in the working coordinate system. Specifically, the predetermined integral gain K.sub.i1 corresponding to the first control axis and the predetermined integral gain K.sub.i2 corresponding to the second control axis are calculated on the basis of Equation 15 below.

(58) 0$\begin{array}{cc}\left[\mathrm{Math}.15\right]& \\ E_{wrk}=\sqrt{{\left(x_{f1}-x_1\right)}^2+{\left(x_{f2}-x_2\right)}^2}\mathrm{Eratio}1=\frac{.Math.\left(x_{f1}-x_1\right).Math.}{E_{wrk}}\mathrm{Eratio}2=1-\mathrm{Eratio}1K_{i1}=\frac{1}{\exp \left(\mathrm{Eratio}1\times E_{wrk}\times K_r\right)}{K}_{i2}=\frac{1}{\exp \left(\mathrm{Eratio}2\times E_{wrk}\times K_r\right)}& \left(\mathrm{Equation}15\right)\end{array}$

(59) Where Ewrk represents a deviation in the working coordinate system, Eratio1 and Eratio2 represent relative error ratios of the respective control axes, and Kr is a constant corresponding to the constant C1 in Equation 6. Calculation itself of the predetermined integral gains in Equation 15 is similar to Equation 6 described above.

(60) In FIGS. 8A and 8B, trajectories of commands in a working coordinate system in which the output of the first control axis set as a horizontal axis and the output of the second control axis is set as a vertical axis are represented by lines L1 and L3, and the trajectories of the outputs of the plant 6 which are the results of the servo control performed by the servo driver 4 are represented by lines L2 and L4. FIG. 8A corresponds to a case in which servo control is performed using the predetermined integral gains of the respective control axes calculated according to Equation 15, and FIG. 8B corresponds to a case in which servo control is performed using the predetermined integral gains of the respective control axes calculated according to Equation 16 without taking the deviations of the respective control axes of the working coordinate system into consideration. The line L1 in FIG. 8A and the line L3 in FIG. 8B are the same trajectory.

(61) $\begin{array}{cc}\left[\mathrm{Math}.16\right]& \\ E_{\mathrm{gen}1}=.Math.x_{f1}-x_1.Math.E_{\mathrm{gen}2}=.Math.x_{f2}-x_2.Math.K_{i1}=\frac{1}{\exp \left(E_{gen1}\times K_r\right)}{K}_{i2}=\frac{1}{\exp \left(E_{gen2}\times K_r\right)}& \left(\mathrm{Equation}16\right)\end{array}$

(62) As illustrated in FIG. 8A, it is possible to allow the output (an output position) of the plant 6 in the working coordinate system to track the command ideally as compared to the case illustrated in FIG. 8B by adjusting the predetermined integral gains of the respective control axes by taking correlation between the deviations of the respective control axes in the working coordinate system into consideration. As a result, in the case illustrated in FIG. 8A, the output of the plant 6 in the working coordinate system becomes identical to the command in a relative early stage.

Third Embodiment

(63) Another mode of adjustment of the predetermined integral gain Ki by the gain adjustment unit 44 included in the servo driver 4 will be described. In the present embodiment, similarly to the above-described embodiment, the gain adjustment unit 44 adjusts the predetermined integral gain K.sub.i so that the value of the predetermined integral gain K.sub.i increases as the magnitude of the deviation e decreases. However, an integration process using the predetermined integral gain K.sub.i is performed only when a predetermined condition is satisfied.

(64) In the present embodiment, the predetermined condition is that the gain adjustment unit 44 performs adjustment of the predetermined integral gain K.sub.i when the value of the deviation e belongs to a predetermined first range including zero and an integration process is performed according to the adjusted predetermined integral gain K.sub.i. In contrast, when the value of the deviation e does not belong to the predetermined first range, the predetermined integral gain K.sub.i is set to zero so that an integration process is not performed substantially. As an example, when the output of the plant 6 is two-dimensional, the predetermined first range can be defined by a downwardly convex function f(x). In such a case, the predetermined integral gain K.sub.i is set in the following manner whereby the adjustment of the predetermined integral gain K.sub.i by the gain adjustment unit 44 of the present embodiment is realized.
K.sub.i=α(|f(x)|−f(x))

(65) Where α is a predetermined coefficient.

(66) By representing the predetermined integral gain K.sub.i using a function in this manner, in a program process of model predictive control based on the continuous deformation method, it is possible to adjust the value of the predetermined integral gain K.sub.i (for example, K.sub.i=0) without performing a conditional determination process and it is easy to generate a program for model predictive control based on Equations 3 and 4 described above.

(67) Moreover, the predetermined first range may be defined as an upwardly convex function f(x). In such a case, the predetermined integral gain K.sub.i can be set as follows.
K.sub.i=α(|f(x)|+f(x))

(68) Here, when the output of the plant 6 is two-dimensional, if the predetermined first range is set as a circle having a predetermined radius (r) about a target position of the output of the plant 6, the predetermined integral gain K.sub.i can be set according to Equation 17 below as an example.

(69) $\begin{array}{cc}\left[\mathrm{Math}.17\right]& \\ K_i=\frac{1}{2r^2}\left(.Math.{\left(x_{f1}-x_1\right)}^2+{\left(x_{f2}-x_2\right)}^2-r^2.Math.-\left({\left(x_{f1}-x_1\right)}^2+{\left(x_{f2}-x_2\right)}^2-r^2\right)\right)& \left(\mathrm{Equation}17\right)\end{array}$

(70) According to Equation 17, the predetermined integral gain K.sub.i is zero when the output (x1,x2) of the plant 6 is within a circle having a predetermined radius r about a target position (xf1,xf2) (that is, the deviation e is within the predetermined first range) and when the predetermined integral gain K.sub.i is set and the output of the plant 6 is outside the circle (that is, the deviation e is outside the predetermined first range). As a result, an integration process of model predictive control is performed in a limited region.

(71) FIG. 9A illustrates the trajectory of an output of the plant 6 when the prediction model of the model predictive control unit 43 is represented by Equation 12 described above. In this diagram, a circle having a predetermined radius r about a target position is represented as the predetermined first range in which an integration process is performed. Moreover, FIG. 9B illustrates the trajectory of an output of the plant 6 in a mode (a mode where the predetermined integral gain is set to K.sub.i=1 in an entire region) in which a region where an integration process is performed is not limited but an integration process is performed in an entire region as a comparative example. As is obvious from comparison between both cases, when a region where an integration process is performed is not limited, a relatively large overshoot occurs due to the influence of a deviation accumulated up to that time point before a target position is reached. In contrast, when the region where an integration process in model predictive control is performed is limited to the predetermined first range, excessive accumulation of deviations is suppressed, and occurrence of an overshoot is obviated ideally.

(72) <Modification>

(73) If an obstacle is present around the plant 6 controlled to track a command r, the plant 6 is requested to avoid collision with the obstacle. Whether the obstacle is moving or not does not matter. Therefore, in order to avoid collision between the plant 6 and the obstacle, a probabilistic potential field indicating the probability that an obstacle is present around the plant 6 is calculated and is applied to model predictive control. Calculation of the probabilistic potential field itself is a known technique, and the probabilistic potential field can be calculated using a technique disclosed in Japanese Patent Application Publication No. 2003-241836, for example. Specifically, the probabilistic potential field is applied to a stage cost (the second term on the right side of Equation 2 is a stage cost) of model predictive control according to Equation 18 below.

(74) $\begin{array}{cc}\left[\mathrm{Math}.18\right]& \\ OD=\sqrt{{\left(x_{d1}-x_1\right)}^2+{\left(x_{d2}-x_2\right)}^2}OP=\frac{0.2}{\exp \left(OD\right)}L=\frac{1}{2}\left({\left(x_f-x\right)}^{T}Q\left(x_f-x\right)+u^T\mathrm{Ru}+\mathrm{OP}\right)& \left(\mathrm{Equation}18\right)\end{array}$

(75) OD indicates the distance between the position (xd1,xd2) of an obstacle and the position (x1,x2) of the plant 6, OP indicates a probabilistic potential of an obstacle, and L indicates a stage cost. In the stage cost L, Q and R are coefficients (weighting coefficients) indicating the weighting factor of a state quantity of the stage cost and the weighting factor of a control input.

(76) FIG. 10A illustrates the trajectory of the output of the plant 6 when the prediction model of the model predictive control unit 43 is represented by Equation 12 described above and the stage cost of the model predictive control is calculated as Equation 18. In this diagram, a circle having a predetermined radius r about a target position is illustrated as the predetermined first range in which an integration process is performed. Moreover, as a comparative example, FIG. 10B illustrates the trajectory of the output of the plant 6 in a mode (a mode where the predetermined integral gain is set to K.sub.i=1 in an entire region) in which a region where an integration process is performed is not limited but an integration process is performed in an entire region. As is obvious from comparison between both cases, when a region where an integration process is performed is not limited, a relatively large overshoot occurs due to the influence of a repulsive force received from an obstacle or a deviation accumulated up to that time point before a target position is reached. In contrast, when the region where an integration process in model predictive control is performed is limited to the predetermined first range, accumulation of deviations can be suppressed to be as small as possible.

Fourth Embodiment

(77) In the present embodiment, another mode of the control structure of the servo driver 4 will be described with reference to FIG. 11. A difference of the control structure illustrated in FIG. 11 from the control structure illustrated in FIG. 2 is that the gain adjustment unit 44 is not included and that the deviation e is input to the integrator 41 through an HPF processing unit 48. Moreover, in an integration process by the integrator 41, the predetermined integral gain K.sub.i is set to one.

(78) The HPF processing unit 48 performs a high-pass filtering process on the deviation e. In the high-pass filtering process, when the value of the deviation e belongs to a predetermined second range including zero, a cutoff frequency thereof is set to zero, and the deviation e is input to the integrator 41 substantially as it is. In contrast, when the value of the deviation e is outside the predetermined second range, the cutoff frequency is adjusted according to the deviation e. Specifically, when the value of the deviation e is outside the predetermined second range, the cutoff frequency is set to be low as the magnitude of the deviation e approaches a boundary of the predetermined second range and is set to zero when the magnitude reaches the boundary.

(79) A transfer function of the HPF processing unit having the above-described configuration is represented by Equation 19 below.

(80) $\begin{array}{cc}\left[\mathrm{Math}.19\right]& \\ \frac{\stackrel{.}{z}}{e}=\frac{s}{s+K_j}& \left(\mathrm{Equation}19\right)\end{array}$

(81) Kj is a predetermined filtering gain correlated with the high-pass filtering process. Equation 20 below is derived on the basis of the transfer function.
[Math. 20]
{umlaut over (z)}=se−K.sub.jż=s(e−K.sub.jz)
ż=(e−K.sub.jz)   (Equation 20)

(82) By applying Equation 20, a prediction model of the model predictive control unit 43 related to the plant 6 having a two-dimensional output (two control axes) can be represented by Equation 21 below.

(83) $\begin{array}{cc}\left[\mathrm{Math}.21\right]& \\ \begin{array}{c}\stackrel{.}{x}\left(t\right)=P\left(x\left(t\right),u\left(t\right)\right)\\ =\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ {\stackrel{.}{x}}_2\\ {\stackrel{\mathrm{.Math.}}{x}}_1\\ {\stackrel{\mathrm{.Math.}}{x}}_2\\ {\stackrel{.}{z}}_1\\ {\stackrel{.}{z}}_2\end{array}\right]=\left[\begin{array}{c}{\stackrel{.}{x}}_1\\ x_2\\ \frac{u_1}{M_1}\\ \frac{u_2}{M_2}\\ \left(x_{f1}-x_1\right)-K_{j1}.Math.x_5\\ \left(x_{f2}-x_2\right)-K_{j2}.Math.x_6\end{array}\right]\end{array}& \left(\mathrm{Equation}21\right)\end{array}$

(84) x5 is the output of the integrator 41 in a first control axis, x6 is the output of the integrator 41 in a second control axis, and (xf1−x1)−Kj1.Math.x5 and (xf2−x2)−Kj2.Math.x6 in Equation 21 correspond to HPF processing terms correspond to the first and second control axes, respectively.

(85) Here, when the output of the plant 6 is two-dimensional, the predetermined second range can be defined as a downwardly convex function f(x). In such a case, the predetermined filtering gain K.sub.j is set as below, whereby the high-pass filtering process is realized.
K.sub.j=α(|f(x)|+f(x))

(86) α is a predetermined coefficient.

(87) By representing the predetermined high-pass filtering gain K.sub.j using a function in this manner, in a program process of model predictive control based on the continuous deformation method, it is possible to adjust the value of the predetermined integral gain K.sub.i (for example, K.sub.i=0) without performing a conditional determination process and it is easy to generate a program for model predictive control based on Equations 3 and 4 described above.

(88) Moreover, the predetermined second range may be defined as an upwardly convex function f(x). In such a case, the predetermined filtering gain K.sub.j can be set as follows.
K.sub.j=α(|f(x)|−f(x))

(89) Here, when the output of the plant 6 is two-dimensional, if the predetermined second range is set as a circle having a predetermined radius (r) about a target position of the output of the plant 6, the predetermined filtering gain K.sub.j can be set according to Equation 22 below as an example.

(90) $\begin{array}{cc}\left[\mathrm{Math}.22\right]& \\ K_j=\frac{1}{2r^2}\left(.Math.{\left(x_{f1}-x_1\right)}^2+{\left(x_{f2}-x_2\right)}^2-r^2.Math.+\left({\left(x_{f1}-x_1\right)}^2+{\left(x_{f2}-x_2\right)}^2-r^2\right)\right)& \left(\mathrm{Equation}22\right)\end{array}$

(91) According to Equation 22, the predetermined filtering gain K.sub.j is set to zero and the integration process of the integrator 41 is performed when the output (x1,x2) of the plant 6 is within a circle having a predetermined radius r about a target position (xf1,xf2) (that is, the deviation e is within the predetermined second range). In contrast, when the output of the plant 6 is outside the circle (that is, the deviation e is outside the predetermined second range), the predetermined filtering gain K.sub.j is adjusted so that the high-pass filtering process is applied. The high-pass filtering process is performed on the deviation e whereby the influence of a relatively old deviation e is alleviated.

(92) As described above, according to the present embodiment, an overshoot resulting from accumulation of an integration amount of deviations when the deviation e is relatively large is suppressed. Moreover, when the filtering gain K.sub.j correlated with the high-pass filtering process is correlated with a distance to the predetermined second range in which the integration process of the integrator 41 is performed by setting the cutoff frequency of the high-pass filtering process in the above-described manner, it is possible to suppress accumulation of an integration amount as the value of the deviation e deviates from the region and to contribute to suppression of an overshoot. The control structure of the servo driver 4 may include a control element (for example, a predetermined gain or the like) other than the integrator 41 and the HPF processing unit 48.

(93) FIG. 12A illustrates the trajectory of the output of the plant 6 when the prediction model of the model predictive control unit 43 is represented by Equation 21 described above. In this diagram, a circle having a predetermined radius r about a target position is illustrated as a “complete integration region” as a predetermined second range of the integration process (a predetermined integral gain K.sub.i=1) by the integrator 41. Moreover, FIG. 12B illustrates the trajectory of the output of the plant 6 in a mode in which the region where the integration process (the predetermined integral gain K.sub.i=1) by the integrator 41 is the same as that of FIG. 12A but the HPF processing unit 48 is not included in the control structure as a comparative example. In calculation of these trajectories, servo control to the target position is performed while avoiding collision with an obstacle using the stage cost illustrated in Equation 18. As is obvious from comparison between both cases, when the HPF processing unit 48 is not included, the target position is not reached due to the influence of a repulsive force received from an obstacle. In contrast, when the HPF processing unit 48 is included, the target position is reached while suppressing an overshoot.

Fifth Embodiment

(94) In the above-described embodiment, the control structure including the model predictive control unit 43 formed in the servo driver 4 has been described. However, in the present embodiment, a control structure which includes a model predictive control unit 53 corresponding to the model predictive control unit 43 and is formed in the standard PLC 5 will be described with reference to FIG. 13. The standard PLC 5 includes a command generation unit 50, an integrator 51, a state acquisition unit 52, a model predictive control unit 53, a gain adjustment unit 54, and a plant model 56. Since the integrator 51, the state acquisition unit 52, the model predictive control unit 53, and the gain adjustment unit 54 substantially correspond to the integrator 41, the state acquisition unit 42, the model predictive control unit 43, and the gain adjustment unit 44 illustrated in FIG. 2, respectively, the detailed description thereof will be omitted.

(95) The command generation unit 50 generates a command r for instructing the output of the plant 6. In the present embodiment, the command r is provided to model predictive control of the model predictive control unit 53 rather than being supplied from the standard PLC 5 directly to the servo driver 4. Moreover, the plant model 56 has a plant model (corresponding to an actual target model of the present application) that models the plant 6 and simulates the output of the plant 6 using the plant model. The simulation result is used as an output y of the plant model 56. The output y of the plant model 56 is fed back to an input side of the integrator 51 by a feedback system 55.

(96) Here, in the present embodiment, a deviation e (e=r−y) between the command r generated by the command generation unit 50 and the output y of the plant model 56 fed back by the feedback system 55 is input to the integrator 51. An output z of the integrator 51 is input to the model predictive control unit 53 through the state acquisition unit 52.

(97) Here, the state acquisition unit 52 acquires a value x of a state variable included in a state x related to the plant model that models the plant 6, provided to the model predictive control performed by the model predictive control unit 53. For example, a value of a predetermined parameter obtained in the simulation process of the plant model 56 can be acquired as the state variable included in the state x. Furthermore, in the present embodiment, the state acquisition unit 52 acquires z which is the output of the integrator 51 as the state variable included in the state x. The model predictive control unit 53 executes model predictive control using the state x related to the plant model acquired by the state acquisition unit 52 and the input u to the plant model 56 output by the model predictive control unit 53. In the model predictive control performed by the model predictive control unit 53, an integral term represented by the product of the deviation e and the predetermined integral gain is included in the prediction model used in the model predictive control similarly to the model predictive control performed by the model predictive control unit 43 illustrated in First and Second Embodiments.

(98) Due to the control structure formed in this manner, the standard PLC 5 illustrated in FIG. 13 can supply the output y which is the simulation result of the plant model 56 to the servo driver 4 as a command for ideally tracking the command r while suppressing the output of the plant 6 from entering an overshoot state. That is, the standard PLC 5 can suppress occurrence of a steady-state deviation in the output of the plant 6 while performing model predictive control even when the target trajectory changes frequently according to the command r although the control amount of the plant 6 which is the actual control target is not fed back directly.

REFERENCE SIGNS LIST

(99) 1 Network 2 Motor 3 Load device 4 Servo driver 5 Standard PLC 6 Plant