Vibration actuator control device, imaging device, and automatic stage

Abstract

To provide a vibration actuator control apparatus which can obtain robust controllability to a drive condition and a temperature environment.SOLUTION: A vibration actuator control apparatus includes a control amount output unit. The control amount output unit includes a learned model which is machine learned to output a control amount to move a contact body relative to a vibrator when a target speed and a value based on a target position are input. The value based on the target position is a value based on the product of a first value and a second value. The first value is based on the difference between the target position and a detection position detected from a vibration actuator moved based on the control amount. The second value is based on the ratio of the control amount output from the control amount output unit to a value output from the learned model when the target speed and a predetermined value are input.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a control device for a vibration actuator, an imaging device, and an automatic stage. [Background technology]

[0002] As an example of a vibration type actuator, a vibration type motor will be described below. A vibration type motor is a non-electromagnetic drive motor configured to apply an AC voltage to an electro-mechanical energy conversion element such as a piezoelectric element coupled to an elastic body, generate high-frequency vibration in the element, and extract the vibration energy as continuous mechanical motion.

[0003] Vibration motors have excellent motor performance, such as being small and lightweight, highly accurate, and having high torque at low speeds, but they have nonlinear motor characteristics that make them difficult to model, and their controllability changes depending on the driving conditions and temperature environment, so they require ingenuity in the control system.In addition, there are many control parameters, such as frequency, phase difference, and voltage, and adjustments are also complicated.

[0004] FIG. 13(a) is a control block diagram of a vibration type control device using a conventional general PID control (see Patent Document 1). A two-phase (A-phase, B-phase) AC voltage (AC signal) is output from a drive circuit to which a control amount described later is input. The actual speed (detected speed) of the vibration type motor can be controlled by controlling the frequency (1 / cycle), phase difference, and voltage amplitude (hereinafter simply referred to as "voltage") of the two-phase AC voltage output from the drive circuit (FIG. 13(b)). The voltage amplitude can be varied by the pulse width input from a PID controller to the drive circuit, described later.

[0005] A position deviation, which is a difference (value based on the difference) between a target position of the vibration motor commanded by a position generating means (position command unit) and an actual position (relative position, detected position) of the vibration motor detected by a position detecting means (position detecting unit), is input to a PID controller. Then, a control amount (frequency, phase difference, and pulse width) PID-calculated according to the position deviation input to the PID controller is input to a drive circuit, and the control amount is sequentially output from the PID controller for each control sampling period. Then, a two-phase AC voltage is output from the drive circuit to which the control amount is input, and the speed of the vibration motor is controlled by the two-phase AC voltage output from the drive circuit. Then, position feedback control is performed by these. Hereinafter, the control sampling period is also simply referred to as a "sampling period."

[0006] Fig. 13(c) is a diagram showing the frequency-speed characteristics of a vibration motor. Specifically, Fig. 13(c) shows that at a frequency (f1) in the high-speed range (low frequency range), the speed is high and the slope of the frequency-speed characteristics is large. Also, at a frequency (f2) in the low-speed range (high frequency range), the speed is low and the slope of the frequency-speed characteristics is small. With a vibration motor, the control performance (frequency-speed characteristics, phase difference-speed characteristics) differs depending on the speed range used, making it difficult to adjust the PID control gain.

[0007] Fig. 13(d) is a diagram showing a schematic diagram of the phase difference-speed characteristic of a vibration type motor. Specifically, Fig. 13(d) shows that at a frequency (f1) in the high-speed region (low frequency range), the speed is high and the slope of the phase difference-speed characteristic is large. Also, at a frequency (f2) in the low-speed region (high frequency range), the speed is low and the slope of the phase difference-speed characteristic is small.

[0008] As shown in Figures 13(c) and 13(d), the frequency-speed characteristics and phase difference-speed characteristics of a vibration actuator vary depending on the speed range used, so the control performance changes depending on the drive frequency and phase difference.

[0009] In addition, when the environmental temperature changes, for example, when the temperature changes from room temperature to low temperature, the resonant frequency shifts from low to high frequency based on the temperature characteristics of the piezoelectric element. In this case, the speed corresponding to the drive frequency and the slope of the frequency-speed characteristic corresponding to the drive frequency are different before and after the resonant frequency shifts from low to high frequency, so the control performance also changes depending on the environmental temperature.

[0010] Furthermore, the speed and inclination differ depending on the individual vibration motor, so the control performance also varies depending on the individual motor.

[0011] In addition, control performance also changes over time. It is necessary to consider all of these factors of change and adjust the PID control gains (proportional gain, integral gain, and differential gain of PID control) to ensure a gain margin and phase margin. [Prior art documents] [Patent documents]

[0012] [Patent Document 1] JP 2016-144262 A Summary of the Invention [Problem to be solved by the invention]

[0013] Therefore, there has been a demand for a control device for a vibration type actuator having a control amount output unit different from the conventional PID controller as the main control amount output unit. The present invention aims to provide a control device for a vibration type actuator having a control amount output unit different from the conventional PID controller as the main control amount output unit. [Means for solving the problem]

[0014] 1. A control device for a vibration type actuator that moves a contact body that is in contact with a vibrator relative to the vibrator by vibration generated in the vibrator, comprising: a control unit that outputs a control amount for moving the contact body relative to the vibrator when a target speed and a target position for moving the contact body relative to the vibrator are input, The control unit is The control unit is machine-trained to output a control amount for moving the contact body relative to the transducer when a value based on the target velocity and the target position is input. 1. A trained model, a second trained model to which the target speed and a predetermined value are input; Equipped with the value based on the target position is a value based on a product of a first value and a second value; the first value is a value based on a difference between the target position and a detected position detected from the vibration actuator moved based on the control amount, The second value is First trained model The control amount output from the The above When a specified value is entered, Second It is characterized in that the value is based on the ratio of the value output from the trained model. Effect of the Invention

[0015] According to the present invention, it is possible to provide a control device for a vibration type actuator having a controlled variable output section different from a conventional PID controller as a main controlled variable output section. [Brief description of the drawings]

[0016] [Figure 1] FIG. 1 is a diagram (control block diagram) showing a vibration type driving device according to a first embodiment of the present invention. [Diagram 2] 1A and 1B are diagrams illustrating the driving principle of a linear drive vibration motor. [Diagram 3] FIG. 2 is a diagram illustrating a lens drive mechanism of the lens barrel. [Figure 4] FIG. 2 is a diagram showing a neural network configuration of a learning model according to the first embodiment of the present invention. [Diagram 5]1 shows a flowchart of machine learning and control using a trained model in a first embodiment of the present invention. [Figure 6] 13 shows a flowchart when Adam is used as an optimization calculation method (optimization algorithm) for neural network parameters. [Figure 7] The calculation results of Adam, RMSprop, Momentum, and SGD are compared using the learning model of the first embodiment of the present invention and actually measured learning data. [Figure 8] 11 is a timing chart illustrating batch learning and online learning in the machine learning unit. [Figure 9] 1 is a block diagram of an AGC circuit according to a first embodiment of the present invention. [Figure 10] FIG. 11 is a diagram (control block diagram) showing a vibration type driving device in a second embodiment of the present invention. [Figure 11] FIG. 11 is a diagram (control block diagram) showing a vibration type driving device in a third embodiment of the present invention. [Figure 12] FIG. 13 is a diagram (control block diagram) showing a vibration type driving device in a fourth embodiment of the present invention. [Figure 13] 1A to 1C are diagrams (control block diagrams) showing a conventional vibration type driving device using general PID control. [Figure 14] 1 shows the results of feedback control of a vibration type motor with a predetermined target position pattern in the first embodiment of the present invention. [Figure 15] 4 is a result showing the robustness of the control device in the first embodiment of the present invention. [Figure 16] FIG. 13 is a diagram (control block diagram) showing a vibration type driving device in a fifth embodiment of the present invention. [Figure 17] This shows the neural network structure of a learning model that outputs phase difference, frequency, and pulse width. [Figure 18] 1A and 1B are a plan view and a schematic diagram of an internal structure of an imaging device which is an application example of a control device of the present invention. [Figure 19]1 is a diagram showing the appearance of a microscope as an application example of a control device of the present invention; [Figure 20] FIG. 4 is a diagram showing the speed characteristic of a vibration actuator based on a control amount. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] (First embodiment) FIG. 1 is a diagram (control block diagram) showing a vibration type driving device in a first embodiment of the present invention.

[0018] The vibration type driving device 17 has a control unit 10 that controls a vibration type motor (vibration type actuator), a learning model generation unit 12, and a driving unit 11. The vibration type driving device 17 also has a vibration type motor 13 (vibration type actuator) having a vibrator 131 and a contact body 132, and a position detection means 14 (position detection unit) that detects the relative position (detected position) between the vibrator 131 and the contact body 132. The vibration type driving device 17 also has a speed detection means 16 (speed detection unit) that detects the relative speed (detected speed). In FIG. 1, the control device 15 is the vibration type driving device 17 excluding the vibration type motor 13 to be controlled.

[0019] The control unit 10 is configured to generate a signal that controls the drive of the vibrator by automatic gain control (AGC). That is, the target speed of the vibration type motor and a correction value of the position deviation (a value based on the target deviation) are input to a control amount output unit 103 equipped with a learned model for control, and the output phase difference and frequency are used as the control amount. Note that the pulse width for changing the voltage amplitude may be used as the control amount.

[0020] The control unit 10 has a speed generating means 101 (speed command unit), a position generating means 102 (position command unit), a controlled variable output unit 103 having a learned model for control, a controlled variable output unit 107 having a learned model for reference, and an AGC circuit 108. The drive unit 11 has an AC signal generating means 104 (AC signal generating unit) and a boost circuit 105.

[0021] A speed generating means 101 (speed command unit) generates a target speed for each time of the relative speed (detected speed) between the vibrator 131 and the contact body 132. A position generating means 102 (position command unit) generates a target position for each time of the relative position (detected position), and a difference (value based on the difference) between the relative position (detected position) detected by the position detecting means 14 (position detecting unit) and the target position is calculated as a position deviation (first value).

[0022] Here, for the target speed and target position, for example, one command value is output from each generating means (each command unit) for each control sampling period. Control sampling refers to one cycle from acquisition of the position deviation (first value) in Fig. 1 through output of the control amount, input of an AC signal to the vibrator, detection of the relative speed (detected speed) and relative position (detected position) between the vibrator and the contact body, to just before acquisition of the position deviation (first value) begins. In this cycle, the position or speed of the vibration type motor is feedback controlled.

[0023] The target speed is a speed given to make the vibration type motor follow a predetermined position, and may be generated by differentiating the target position over time. Conversely, the target position may be generated by integrating the target speed.

[0024] The AGC circuit 108, which is a feature of the present invention, will now be described in detail.

[0025] The target speed and a correction value (value based on the target position) of the position deviation (first value) are input to a controlled variable output unit 103 having a learned model for control, which outputs a phase difference and a frequency. Meanwhile, the target speed and a predetermined value (zero) are input to a controlled variable output unit 107 having a learned model for reference, which outputs a reference value of the phase difference. Therefore, the "phase difference as a controlled variable" and the "phase difference as a reference value" are input to the AGC circuit 108.

[0026] In addition, since the control amount output unit 107 having the learned model for reference calculates two reference values ​​(phase difference, frequency), the frequency reference value may be used for AGC. Also, both the phase difference and the frequency may be used as reference values. The AGC circuit 108 receives the control amount and the reference value as inputs and outputs a correction gain (second value). The correction gain (second value) indicates the relative ratio of the control amount used for actual driving to the reference value, which is the result of prior learning.

[0027] FIG. 9 is a block diagram of the AGC circuit according to the first embodiment of the present invention.

[0028] The control amount and the reference value are each subjected to an absolute value calculation 901, and then a predetermined value 902 is added to prevent division by zero. Then, a divider 903 calculates the relative ratio of the control amount to the reference value. The relative ratio is processed by a low-pass filter 904 to remove noise components, and is amplified by a predetermined set gain 905 and then output as a correction gain (second value). The set gain 905 is a reference gain that is set so that the vibration type motor can be controlled stably and accurately with the parameters of the NN obtained during learning.

[0029] In the AGC circuit, for example, in FIG. 13(d), when the phase difference-speed characteristic changes from the solid line to the dotted line, and the speed characteristic of the vibration type motor decreases from the time of learning, the same speed cannot be obtained unless the control amount is made larger than that of learning. Therefore, in this case, an increased correction gain (second value) is output. Conversely, in FIG. 13(d), when the phase difference-speed characteristic changes from the dotted line to the solid line, and the speed characteristic of the vibration type motor increases from the time of learning, the same speed can be obtained even if the control amount is made smaller than that of learning. Therefore, in this case, a decreased correction gain (second value) is output.

[0030] Therefore, a correction value (a value based on the target position, a value based on the product of the first value and the second value) obtained by multiplying the position deviation (first value) calculated during driving and the correction gain (second value) is input to a controlled variable output unit 103 equipped with a learned model for control. This makes it possible to compensate for the speed characteristics of the vibration type motor. As a result, even if the driving conditions or temperature environment change, the controlled variable is automatically corrected by auto gain control, and highly accurate and robust controllability can be obtained.

[0031] The operation of the AGC circuit according to the first embodiment of the present invention will be specifically described using control results in an actual device.

[0032] FIG. 14 shows the results of feedback control of a vibration type motor with a predetermined target position pattern in the control device of the present invention.

[0033] The target speed is trapezoidal drive with a maximum of 150 mm / s, and the pattern performs a reciprocating motion with a 12 mm stroke including positioning motion. The horizontal axis indicates time (sec), and the vertical axis indicates the target position (encoder pulse count: 8000 pls per mm) (left axis) and the position deviation (first value) in μm (right axis).

[0034] Figure 14(a) shows the control results at a startup frequency of 91 kHz, Figure 14(b) shows the control results at a startup frequency of 93 kHz, and Figure 14(c) shows the control results at a startup frequency of 95 kHz. Note that measurements were performed with a configuration in which the PID controllers shown in the examples described below were connected in parallel. The controlled variables are the phase difference and frequency. The trained model was generated using measurement data controlled at a startup frequency of 93 kHz. The position deviation (first value) tends to be large in the acceleration / deceleration region, and this is because an error occurs between the controlled variable output by the trained model and the actual controlled variable due to the influence of the inertia of the non-driven body.

[0035] The three figures in the lower row correspond to the figures in the upper row, and are log outputs of the correction gain (second value) output from the AGC circuit during control. The horizontal axis is time (sec), and the vertical axis is the correction gain (second value). For example, if the learned model and the actual control amount are completely consistent, the correction gain (second value) is output as 1. In the case of the start-up frequency of 91 kHz in (a), the speed characteristic of the vibration type motor increases because it is closer to the resonance frequency than during learning, and the same speed can be obtained with a smaller control amount than during learning. Therefore, as shown in the figure, the correction gain (second value) decreases to around 0.6 in the acceleration / deceleration region, and it operates in the direction of substantially lowering the gain related to the position deviation (first value). In addition, in the case of the start-up frequency of 95 kHz in (c), the speed characteristic of the vibration type motor decreases because it is farther from the resonance frequency than during learning, and the same speed cannot be obtained unless the control amount is larger than during learning. Therefore, as shown in the figure, the correction gain (second value) increases to around 1.6 in the acceleration / deceleration region, and it operates in the direction of substantially raising the gain related to the position deviation (first value). In this way, the control amount is automatically corrected by the auto gain control in response to changes in the characteristics of the vibration type motor, so that the position deviation (first value) during driving can be reduced and controllability can be improved.

[0036] FIG. 15 shows the results showing robustness in the control device in the first embodiment of the present invention.

[0037] Positioning was performed using trapezoidal drive at a maximum speed of 150 mm / s, and the position deviation (first value) during reciprocating movement with a 12 mm stroke was calculated at 3σ. The horizontal axis is the starting frequency, and the vertical axis is the position deviation (first value). Comparative Example 1 shows the results using conventional PID control. Compared to Comparative Example 1, control using a trained model that uses the auto gain control of the present invention significantly improves the position deviation (first value). It can be seen that even when the starting frequency is changed, there is little fluctuation in the position deviation (first value), and robustness has been improved.

[0038] Comparative Example 2 shows the control results using a neural network with auto gain control turned off. There is no significant difference at a startup frequency of 93 kHz, but the position deviation (first value) increases at a startup frequency of 95 kHz, demonstrating the effect of the present invention. As mentioned above, when control is performed at a different startup frequency, the speed slope changes due to the nonlinear characteristics of the vibration motor, which was difficult to achieve with conventional PID control. In the present invention, the control amount is automatically corrected by auto gain control, making it possible to obtain good controllability even at different startup frequencies.

[0039] The configuration of a learning model used in the present invention, a learning method, and a control method for a vibration type motor using the learned model will be described below.

[0040] FIG. 4 is a diagram showing a neural network configuration of a learning model according to the first embodiment of the present invention.

[0041] The control amount output unit 103 having a learned model for control and the control amount output unit 107 having a learned model for reference are configured with the following neural network (hereinafter referred to as "NN"). The NN is configured with an X layer as an input layer, an H layer as a hidden layer, and a Z layer as an output layer. In the first embodiment of the present invention, the target speed is set to x1 and the position deviation (first value) is set to x2 as input data, and the phase difference is set to z1 and the frequency is set to z2 as output data. The hidden layer is formed with seven neurons, and a general sigmoid function (FIG. 4(b)) is used as the activation function. The number of neurons in the hidden layer may be other than seven, and is preferably in the range of, for example, 3 to 20. The smaller the number of neurons, the lower the accuracy but the faster the learning converges, and the more the number of neurons, the higher the accuracy but the slower the learning. In addition, a sigmoid function or ReLU (ramp function) is generally used as the activation function of the output layer, but a linear function (FIG. 4(c)) is used to accommodate the negative sign of the phase difference, which is the control amount. The weight connecting each neuron in the input layer and hidden layer is wh, the threshold of the neuron in the hidden layer is θh, the weight connecting each neuron in the hidden layer and output layer is wo, and the threshold of the neuron in the output layer is θo. Values ​​learned by the machine learning unit 12 described below are applied to all weights and thresholds. The trained NN can be regarded as a collection of common feature patterns extracted from the time series data of the speed and control amount of the vibration type motor, and the output is a value obtained by a function with the weight and threshold as variables.

[0042] The control amount (phase difference, frequency) output from the control amount output unit 103 having a learned model for control is input to an AC signal generating means 104 (AC signal generating unit), which controls the speed and driving direction of the vibration type motor. The AC signal generating means 104 (AC signal generating unit) generates a two-phase AC signal based on the phase difference, frequency, and pulse width. The boost circuit 105 has, for example, a coil and a transformer, and the AC signal boosted to a desired driving voltage by the boost circuit 105 is applied to the piezoelectric element of the vibrator 131 and drives the contact body 132.

[0043] An example of a vibration type motor applicable to the present invention will be described with reference to the drawings. The vibration type motor according to the first embodiment of the present invention has a vibrator and a contact body.

[0044] FIG. 2 is a diagram for explaining the driving principle of a linear drive type vibration motor as an example of a vibration motor.

[0045] The vibration type motor shown in Fig. 2(a) has an oscillator 131 having an elastic body 203 and a piezoelectric element 204, which is an electrical-mechanical energy conversion element, bonded to the elastic body 203, and a contact body 132 driven by the oscillator 131. By applying an AC voltage to the piezoelectric element 204, two vibration modes are generated as shown in Fig. 2(c) and (d), and the contact body 132, which is in pressure contact with the protrusion 202, is moved in the direction of the arrow.

[0046] 2(b) is a diagram showing the electrode pattern of the piezoelectric element 204. For example, the piezoelectric element 204 of the vibrator 131 has electrode regions divided into two equal parts in the longitudinal direction. The polarization direction of each electrode region is the same (+). Of the two electrode regions of the piezoelectric element 204, an AC voltage (VB) is applied to the electrode region located on the right side of FIG. 2(b), and an AC voltage (VA) is applied to the electrode region located on the left side.

[0047] If VB and VA are AC voltages with frequencies near the resonance frequency of the first vibration mode and in phase with each other, the entire piezoelectric element 204 (the two electrode regions) will expand at one moment and contract at another moment. As a result, the vibrator 131 will generate vibrations in the first vibration mode (hereinafter, thrust vibrations) shown in Fig. 2(c). This causes a displacement in the thrust direction (Z direction) of the protrusion 202.

[0048] Furthermore, if VB and VA are AC voltages with frequencies near the resonant frequency of the second vibration mode and with a phase difference of 180°, at a certain moment, the electrode area on the right side of the piezoelectric element 204 contracts while the electrode area on the left side expands. At another moment, the opposite relationship occurs. As a result, the vibrator 131 generates vibrations in the second vibration mode (hereinafter, feed vibrations) shown in FIG. 2(d). This causes the protrusion 202 to be displaced in the drive direction (feed direction, X direction).

[0049] Therefore, by applying an AC voltage having a frequency near the resonant frequencies of the first and second vibration modes to the electrodes of the piezoelectric element 204, it is possible to excite a vibration in which the first and second vibration modes are combined.

[0050] In this way, by combining the two vibration modes, the protrusion 202 performs an elliptical motion in a cross section perpendicular to the Y direction (perpendicular to the X and Z directions) in Fig. 2(d). This elliptical motion drives the contact body 132 in the direction of the arrow in Fig. 2(a). The direction in which the contact body 132 and the vibrator 131 move relative to each other, that is, the direction in which the contact body 132 is driven by the vibrator 131 (here, the X direction) is referred to as the drive direction.

[0051] In addition, the amplitude ratio R (feed vibration amplitude / thrust vibration amplitude) of the second vibration mode to the first vibration mode can be changed by changing the phase difference of the two-phase AC voltage input to the two equally divided electrodes. In this vibration type motor, it is possible to change the speed of the contact body by changing the vibration amplitude ratio.

[0052] In the above description, the case where the vibrator 131 is stationary and the contact body 132 moves has been described as an example, but the present invention is not limited to this form. The positions of the contact parts of the contact body 132 and the vibrator 131 need only change relatively, and for example, the contact body 132 may be fixed and the vibrator 131 may move. That is, in the present invention, "driving" means changing the relative positions of the contact body and the vibrator, and does not require the position of the contact body (for example, the position of the contact body when the position of the housing containing the contact body and the vibrator is used as a reference) to change.

[0053] Vibration motors are used, for example, for driving the autofocus of cameras.

[0054] FIG. 3 is a diagram illustrating a lens drive mechanism of the lens barrel according to the first embodiment of the present invention.

[0055] The lens holder drive mechanism using a vibration type motor includes a vibrator, a lens holder, and a first guide bar and a second guide bar arranged in parallel to slidably hold the lens holder. In the first embodiment of the present invention, the second guide bar is a contact body, and the second guide bar is fixed, and the vibrator and the lens holder move together.

[0056] The vibrator generates a relative movement force between the vibrator and the second guide bar in contact with the protrusion of the elastic body by elliptical motion of the protrusion of the vibrator generated by application of a drive voltage to the electromechanical energy conversion element, thereby allowing the lens holder fixed integrally with the vibrator to move along the first and second guide bars.

[0057] Specifically, the contact body drive mechanism 300 mainly includes a lens holder 302 which is a lens holding member, a lens 306, a vibrator 131 to which a flexible printed circuit board is connected, a pressure magnet 305, two guide bars 303 and 304, and a base body (not shown). Here, the vibrator 131 will be described as an example of the vibrator.

[0058] Both ends of first guide bar 303 and second guide bar 304 are held and fixed by a base (not shown) so that they are arranged parallel to each other. Lens holder 302 has a cylindrical holder portion 302a, a holding portion 302b that holds and fixes transducer 131 and pressure magnet 305, and a first guide portion 302c that fits with first guide bar 303 to act as a guide.

[0059] Pressure magnet 305 for constituting pressure means (pressure unit) has a permanent magnet and two yokes arranged on both ends of the permanent magnet. A magnetic circuit is formed between pressure magnet 305 and second guide bar 304, and an attractive force is generated between these members. Pressure magnet 305 is arranged with a gap between it and second guide bar 304, and second guide bar 304 is arranged so as to contact vibrator 131.

[0060] The attractive force applies a pressure between second guide bar 304 and vibrator 131. Two protrusions of the elastic body come into pressure contact with second guide bar 304 to form a second guide section. The second guide section forms a guide mechanism using magnetic attractive force, and there may be a situation where vibrator 131 and second guide bar 304 are pulled apart due to external force or the like, but this is dealt with as follows.

[0061] That is, the lens holder 302 is adapted to return to a desired position when the fall prevention portion 302d provided on the lens holder 302 comes into contact with the second guide bar 304. By applying a desired AC voltage signal to the vibrator 131, a driving force is generated between the vibrator 131 and the second guide bar 304, and this driving force drives the lens holder.

[0062] A position sensor (not shown) attached to the contact body 132 or the vibrator 131 detects the relative position (detected position) and relative speed (detected speed) between the vibrator 131 and the contact body 132. The relative position (detected position) is fed back to the learned model control unit 10 (control unit) as a position deviation (first value), and the vibration type motor is feedback-controlled so as to follow the target position for each time. The relative speed (detected speed) is input to the machine learning unit 12 and used as learning data together with the control amount.

[0063] In addition, the first embodiment of the present invention will be described using as an example a two-phase drive control device that drives a piezoelectric element, which is an electro-mechanical energy conversion element, in two phases. However, the present invention is not limited to two-phase drive and can also be applied to a vibration type motor with two or more phases.

[0064] Next, the learning model generation unit 12 will be described.

[0065] The learning model is generated using a NN (see FIG. 4) that receives the relative speed (detected speed) and speed deviation from the speed detection means 16 (speed detection unit) as inputs and outputs the phase difference and frequency. The speed deviation is the deviation between the target speed and the relative speed (detected speed). Note that zero may be input as the target deviation instead of the speed deviation, and an offset value may be provided to compensate for the backlash of the mechanical system. The control amount (phase difference, frequency) output from the control amount output unit 103 having a learned model for control is used as the correct answer data, and is compared with the control amount output from the NN with the relative speed (detected speed) and speed deviation as inputs, and an error is calculated. Note that in this example, the phase difference and frequency are used as the control amount, but other combinations such as the pulse width and frequency, or the pulse width and phase difference can also be applied. Also, the output layer of the NN may have one neuron, or it may be designed to select one of the phase difference, frequency, and pulse width.

[0066] FIG. 5 shows a flowchart of machine learning and control using a trained model in the first embodiment of the present invention.

[0067] In step 1, the weights and thresholds of the controlled variable output unit 103 having a trained model for control and the controlled variable output unit 107 having a trained model for reference are set to initial values. The initial values ​​are set based on a random function (untrained state), but previously trained parameters may also be used.

[0068] In step 2, the vibration type motor is controlled using the unlearned model.

[0069] In step 3, the control amount output from the control amount output unit 103 having a learned model for control while the vibration type motor is being driven, and the time series data of the detected relative speed (detected speed) and speed deviation are acquired as learning data.

[0070] In step 4, the control amount of the learning data is treated as correct answer data, and an optimization calculation of the learning model by machine learning is performed. The weights and thresholds of the NN are optimized by machine learning, and the parameters of the control amount output unit 103 having a learned model for control and the control amount output unit 107 having a learned model for reference are updated.

[0071] In step 5, the vibration motor is controlled by auto gain control using the trained model with updated weights and thresholds. After control, the process returns to step 3 to acquire training data in order to respond to changes in driving conditions and temperature environment. Methods for acquiring training data include batch learning, which learns while the motor is stopped, and online learning, which learns sequentially while the motor is running.

[0072] FIG. 8 is a timing chart illustrating batch learning and online learning in the machine learning unit.

[0073] The horizontal axis indicates time, and the vertical axis indicates the target position pattern given as a command value for feedback control of the vibration motor. Figure 8(a) shows an example of batch learning in which learning is performed while the motor is stopped. Time series data of the speed and control amount detected during the driving period of the vibration motor is acquired as learning data, and the stopped period is used to perform machine learning and update the parameters of the NN (weights, thresholds).

[0074] It is not necessary to perform machine learning during each downtime; for example, it is possible to perform learning only when a change in the temperature environment or operating conditions is detected. Figure 8(b) shows an example of online learning, in which learning is performed sequentially during operation. In this example, machine learning is performed online in parallel with the operating period, and the parameters of the NN are updated during the operating period. By applying online learning, it becomes possible to respond to load fluctuations that occur during the operating period.

[0075] The machine learning in step 4 above will be further explained with reference to FIG.

[0076] FIG. 6(a) shows a flowchart when Adam is used as an optimization calculation method (optimization algorithm) for NN parameters.

[0077] Steps 1 and 2 are as described above in FIG. 5. In step 3, the learning data of the time series shown in FIG. 6(b) is acquired. The speed (n) and the control amount (n) are measurement data when controlled by an unlearned model, and the number of samples n of each of the speed and phase difference is 3400. This is actual measurement data when driven for 0.34 seconds at a control sampling rate of 10 kHz. Note that the learning data does not necessarily need to be acquired at the control sampling rate, and memory can be saved and learning time can be shortened by thinning it out. In the present invention, the speed (n) is used as the input of the learning model, and the output z(n) of the calculation result is compared with the control amount (n) of the correct data to calculate the error e(n). In step 4, the error E for 3400 pieces is calculated in the first loop, and the error gradient ∇E of the weights (wh, wo) and thresholds (θh, θo) is calculated. Next, the error gradient ∇E is used to optimize the parameters using Adam, which is one of the optimization calculation methods (optimization algorithms), as follows.

[0078]

number

[0079] w t is the parameter update amount, ∇E is the error gradient, and V t is the moving average of the error gradient, S t is the moving average of the squared error gradient, η is the learning rate, and ε is a constant to prevent division by zero. The parameters used are η = 0.001, β1 = 0.9, β2 = 0.999, and ε = 10e-12. Each time the optimization calculation is repeated, the weights and thresholds are updated, and the output z(n) of the learning model approaches the control amount (n) of the correct data, so the error E becomes smaller. Figure 6(c) shows the transition of the error E based on the number of calculation loops. Note that other optimization calculation methods may also be used.

[0080] FIG. 7(a) shows a comparison of the calculation results by Adam, RMSprop, Momentum, and SGD using the learning model of the first embodiment of the present invention and actually measured learning data.

[0081] From the viewpoints of the number of calculations, stability, and final error, Adam gave the best results. Figure 7(b) is an example of learning a control amount (phase difference) using Adam. It can be seen that the output z of the learning model in the first loop is significantly different from the correct data t. After repeated calculations, the output z of the learning model in the 5000th loop almost matches the correct data t. In this example, optimization was performed with the loop count set to 5000, but it is desirable to adjust the number of loops appropriately depending on the convergence rate. The above is the configuration of the control device of the present invention.

[0082] The control unit 10 and the learning model generation unit 12 are composed of elements such as digital devices, such as a CPU and a PLD (including ASIC), and an A / D converter. The AC signal generation means 104 (AC signal generation unit) of the drive unit 11 has, for example, a CPU, a function generator, and a switching circuit, and the boost circuit is composed of, for example, a coil, a transformer, and a capacitor. The control unit and the drive unit may be composed of not only one element or circuit, but also multiple elements or circuits. Each process may be executed by any element or circuit.

[0083] Second Embodiment A second embodiment of the present invention will now be described.

[0084] 10 is a diagram (control block diagram) showing a vibration type driving device in the second embodiment of the present invention. In FIG. 10, the control device 15 is a vibration type driving device 17 excluding the vibration type motor 13 to be controlled.

[0085] In the control block diagram shown in FIG. 10, machine learning is performed using pulse width and frequency as control variables, and control is performed using the learned model. In this control block, position feedback control of the vibration type motor 13 (vibration type actuator) is performed by the pulse width and frequency output from a control variable output unit 1003 having a learned model for control. In the learning model generation unit 12, the two control variables output from the control variable output unit 1003 having a learned model for control, and the relative speed (detected speed) detected by the speed detection means 16 (speed detection unit) are acquired as learning data, and machine learning is performed. The control variable and reference value used for the auto gain control are pulse widths, and are input to the AGC circuit 108, respectively. Note that a frequency may be used instead of the pulse width.

[0086] By using the second embodiment of the present invention, the pulse width and frequency can be automatically corrected by auto gain control even if the driving conditions or temperature environment change, thereby achieving highly accurate and robust controllability.

[0087] FIG. 20 is a diagram showing the speed characteristic of a vibration type motor based on a control amount.

[0088] FIG. 20(a) shows a case where control is performed using a phase difference and a frequency (see the first embodiment of the present invention). The horizontal axis shows frequency, and the vertical axis shows motor speed. As shown in the figure, the motor speed can be controlled by manipulating the phase difference and the frequency, respectively. For example, when control is performed in the gray area, the drive frequency is 88 to 93 kHz, and the phase difference is 0 to ±120°, and each control amount is output. The trained model of the present invention performs control by outputting two control amounts according to the input of the target speed. FIG. 20(b) shows a case where control is performed using a pulse width and a frequency according to the second embodiment of the present invention. Similarly, the motor speed can be controlled by manipulating the pulse width and the frequency, respectively. For example, when control is performed in the gray area, the drive frequency is 88 to 93 kHz, and the pulse width is 0 to 50%, and each control amount is output. The pulse width and the frequency are output according to the target speed input to the trained model, and control is performed.

[0089] (Third embodiment) A third embodiment of the present invention will now be described.

[0090] 11 is a diagram (control block diagram) showing a vibration type driving device in a third embodiment of the present invention. In FIG. 11, the control device 15 is a vibration type driving device 17 excluding the vibration type motor 13 to be controlled.

[0091] In the control block diagram shown in FIG. 11, machine learning is performed using the phase difference, frequency, and pulse width as control variables, and control is performed using the learned model.

[0092] In this control block, the position feedback control of the vibration type motor 13 (vibration type actuator) is performed by each control amount (phase difference, frequency, pulse width) output from the control amount output unit 1103 having a learned model. The target speed and a correction value (value based on the target position) of the position deviation (first value) are input to the control amount output unit 1103 having a learned model. Then, the phase difference, frequency, and pulse width calculated by the neural network are output to the drive unit 11 to control the vibration type motor 13 (vibration type actuator). The learning model generation unit 12 acquires the three control amounts output from the control amount output unit 1103 having a learned model, and the relative speed (detected speed) and speed deviation detected by the speed detection means 16 (speed detection unit) as learning data, and performs machine learning of the learning model. The control amount and reference value used for the auto gain control are any one or a combination of the phase difference, frequency, and pulse width, and are input to the AGC circuit 108, respectively.

[0093] By using the third embodiment of the present invention, automatic correction of the phase difference, frequency, and pulse width is performed by auto gain control even if the driving conditions or temperature environment change, thereby achieving highly accurate and robust controllability.

[0094] FIG. 17 shows the neural network structure of a learning model that outputs phase difference, frequency, and pulse width.

[0095] The controlled variable output unit 1103 having a learned model for control and the controlled variable output unit 1107 having a learned model for reference have an NN structure that receives speed and deviation as inputs and outputs three controlled variables. The learned data used for machine learning may be measurement data of control using a learned model. The learned data used for machine learning may be measurement data of control using an unlearned model with parameters set by a random function. The learned data used for machine learning may be measurement data by open drive that outputs a controlled variable with an arbitrarily set drive pattern, or time-series measurement data by PID control.

[0096] When determining the weights and thresholds of the NN, parameters with optimal conditions may be selected from multiple learning data in terms of the position deviation (first value), power consumption, etc. This is because there are an infinite number of combinations of conditions under which a predetermined speed of the vibration motor can be obtained, that is, phase difference, frequency, and pulse width.

[0097] By applying the third embodiment of the present invention, the number of parameters for operating the vibration type motor increases, so that it is possible to finely adjust the control performance by performing appropriate machine learning.

[0098] (Fourth embodiment) A fourth embodiment of the present invention will now be described.

[0099] 12 is a diagram (control block diagram) showing a vibration type driving device in a fourth embodiment of the present invention. In FIG. 12, the control device 15 is a vibration type driving device 17 excluding the vibration type motor 13 to be controlled.

[0100] In the control block diagram shown in FIG. 12, the position feedback control of the vibration type motor 13 (vibration type actuator) is performed by connecting a PID controller 109 in parallel to a control amount output unit 103 having a learned model for control. A position deviation (first value) is input to the PID controller 109, which performs PID calculation to output a control amount of a phase difference and a frequency. Note that a configuration other than the PID controller may be used, and for example, P control, PI control, PD control, etc. may be applied. A correction value (value based on a target position) obtained by multiplying a target speed, the first value, and a correction gain (second value) by an AGC circuit 108 is input to the control amount output unit 103 having a learned model for control. In other words, a target speed and a value based on a target position (value based on a product of a first value and a second value) are input to the control amount output unit 103 having a learned model for control.

[0101] The control amount output from the PID controller and the control amount output from a control amount output unit 103 having a learned model for control are added together and output to the drive unit 11. The added control amount is input to an AGC circuit 108 and compared with a reference value output from a control amount output unit 107 having a learned model for reference. In the learning model generation unit 12, machine learning is performed using the control amount after the addition and the relative speed (detected speed) and speed deviation detected by the speed detection means 16 (speed detection unit), and a learning model for the control unit 10 is generated.

[0102] By applying the fourth embodiment of the present invention, the control amount is automatically corrected by the auto gain control, so that highly accurate and robust controllability can be obtained regardless of the driving conditions or temperature environment. In addition, by using a PID controller in combination, the transfer characteristics of the control loop can be flexibly adjusted, and the positioning accuracy can be further improved.

[0103] Fifth embodiment A fifth embodiment of the present invention will now be described.

[0104] 16 is a diagram (control block diagram) showing a vibration type driving device in a fifth embodiment of the present invention. In FIG. 16, the control device 15 is a vibration type driving device 17 excluding the vibration type motor 13 to be controlled.

[0105] In this control block, the PID controller 109 is connected in parallel to the control amount output unit 103 having a learned model for control. Then, a first switch SW1 (switching unit) and a second switch SW2 (switching unit) selectively perform position feedback control of the vibration type motor 13 (vibration type actuator). A switch is provided in each output unit of the PID controller 109 and the control amount output unit 103 having a learned model for control, and it is possible to select control using only PID control or control using only the learned model depending on the driving conditions. In addition, both control amounts may be added together for control.

[0106] The same is true for machine learning; for example, if learning is performed using only PID control, a learning model can be generated with specified control parameters even if no learning has been performed at all.

[0107] By applying the fifth embodiment of the present invention, the controlled variable is automatically corrected by the auto gain control, so that highly accurate and robust controllability can be obtained regardless of the driving conditions and temperature environment. In addition, the stability of control and learning can be improved by selectively using a PID controller.

[0108] Sixth embodiment A sixth embodiment of the present invention will now be described.

[0109] In the first embodiment of the present invention, an example has been described in which the vibration type motor control device is used to drive a lens for autofocusing in an imaging device, but the application of the present invention is not limited to this. For example, as shown in FIG. 18, the control device can also be used to drive a lens or an imaging element during image stabilization. FIG. 18(a) is a plan view (top view) showing the exterior of an imaging device 60. Also, FIG. 18(b) is a schematic diagram of the internal structure of the imaging device 60.

[0110] The imaging device 60 is generally composed of a main body 61 and a lens barrel 62 that is detachable from the main body 61. The main body 61 includes an imaging element 63 such as a CCD sensor or a CMOS sensor that converts an optical image formed by light passing through the lens barrel 62 into an image signal, and a camera control microcomputer 64 that controls the overall operation of the imaging device 60. A plurality of lenses L such as a focus lens and a zoom lens are arranged at predetermined positions in the lens barrel 62. The image blur correction device 50 is also built into the lens barrel 62, and the image blur correction device 50 has a disk member 56 and an oscillator 131 provided on the disk member 56, and an image blur correction lens 65 is arranged in a hole formed in the center of the disk member 56. The image blur correction device 50 is arranged so that the image blur correction lens 65 can be moved within a plane perpendicular to the optical axis of the lens barrel 62. In this case, by driving the vibrator 131 using the control device 12 of the present invention, the vibrator 131 and the disk member 56 move relative to the contact body 132 fixed to the lens barrel, and the correction lens is driven.

[0111] The control device of the present invention can also be used to drive a lens holder for moving a zoom lens. Therefore, the control device of the present invention can be mounted on an interchangeable lens as well as an imaging device for lens drive.

[0112] The vibration motor control device according to the first embodiment of the present invention can also be used to drive an automatic stage. For example, as shown in FIG. 19, it can be used to drive an automatic stage of a microscope.

[0113] The microscope in Fig. 19 has an imaging section 70 incorporating an image sensor and an optical system, and an automatic stage 71 having a stage 72 that is mounted on a base and moved by a vibration type motor. An object to be observed is placed on the stage 72, and an enlarged image is captured by the imaging section 70. When the observation range is wide, the stage 72 is moved by driving the vibration type drive motor using the control device 12 of the first embodiment of the present invention or the control device 12 of the second embodiment of the present invention. This moves the object to be observed in the X direction and Y direction in the figure, and a large number of captured images are obtained. The captured images are combined by a computer (not shown) to obtain a single image with a wide observation range and high resolution.

[0114] In the first to fifth embodiments, the trained model includes two trained models. Specifically, the trained models include a first trained model to which a target speed and a value based on the target position are input, and a second trained model to which a target speed and a predetermined value are input.

[0115] However, the present invention is not limited to this, and only one trained model may be included as the trained model. Specifically, the trained model may be input with the target speed and the value based on the target position, and the target speed and the predetermined value input at different times.

[0116] The present invention has been described in detail above based on the preferred embodiments. However, the present invention is not limited to these specific embodiments, and various forms within the scope of the gist of the present invention are also included in the present invention. Furthermore, each of the above-mentioned embodiments merely shows one embodiment of the present invention, and each embodiment can be appropriately combined. [Explanation of symbols]

[0117] 10 Trained model control unit (control unit) 11 Drive unit 12 Learning model generation unit 13 Vibration motor (vibration actuator) 14 Position detection means (position detection unit) 15 Vibration type drive device 16 Speed ​​detection means (speed detection unit) 17 Vibration type drive device 101 Speed ​​generation means (speed command section) 102 Position generation means (position command unit) 103 Control output unit with a learned model for control 104 AC signal generation means (AC signal generation section) 105 Boost circuit 107 Control output unit with a learned model for reference 108 AGC circuit 131 Transducer 132 Contact body

Claims

1. 1. A control device for a vibration type actuator that moves a contact body that is in contact with a vibrator relative to the vibrator by vibration generated in the vibrator, comprising: a control unit that outputs a control amount for moving the contact body relative to the vibrator when a target speed and a target position for moving the contact body relative to the vibrator are input, The control unit is a first trained model that has been machine-learned to output a control amount for moving the contact body relative to the oscillator when a value based on the target velocity and the target position is input; A second trained model to which the target speed and a predetermined value are input, the value based on the target position is a value based on a product of a first value and a second value; the first value is a value based on a difference between the target position and a detected position detected from the vibration actuator moved based on the control amount, A control device for a vibration type actuator, characterized in that the second value is a value based on a ratio of a control amount output from the first trained model to a value output from the second trained model when the target speed and the specified value are input.

2. The control device for a vibration type actuator as described in claim 1, characterized in that the first learned model is machine-learned using learning data in which the detected speed of the vibration type actuator and the speed deviation, which is the difference between the target speed and the detected speed, are input, and the control amount is output.

3. 2. The control device for a vibration type actuator according to claim 1, characterized in that the second learned model is learned using learning data in which the target speed and a predetermined value are input and the control amount is output.

4. 4. The vibration actuator control device according to claim 3, wherein the predetermined value is zero.

5. The first trained model and the second trained model have a neural network configuration having an input layer having one or more first neurons, a hidden layer having a plurality of second neurons, and an output layer having one or more third neurons; A control device for a vibration type actuator as described in any one of claims 1 to 4, characterized in that the parameters in the neural network configuration include a plurality of first weights assigned to a plurality of outputs from the one or more first neurons to the plurality of second neurons, a plurality of second weights assigned to a plurality of outputs from the plurality of second neurons to the one or more third neurons, a threshold value of the second neuron, and a threshold value of the third neuron.

6. 6. The control device for a vibration actuator as described in claim 5, characterized in that the first learned model and the second learned model are optimized in terms of the first weight, the second weight, the threshold of the second neuron, and the threshold of the third neuron.

7. The control device for a vibration actuator as described in claim 6, characterized in that the first learned model and the second learned model are optimized by machine learning the first weight, the second weight, the threshold of the second neuron, and the threshold of the third neuron, which are set by a random function, based on an optimization algorithm.

8. The control device for a vibration actuator as described in claim 6, characterized in that the first learned model and the second learned model are optimized by machine learning the first weight, the second weight, the threshold of the second neuron, and the threshold of the third neuron based on an optimization algorithm.

9. 9. The vibration actuator control device according to claim 7, wherein the optimization algorithm is any one of Adam, Momentum, RMSprop, and SGD.

10. a PID controller that outputs a control amount for moving the contact body relative to the vibrator when the first value is input, A control device for a vibration type actuator as described in any one of claims 1 to 9, characterized in that the control amount output from the control unit is a value based on the sum of a first control amount as a control amount output from the first learned model and a second control amount as a control amount output from the PID controller.

11. The control device for a vibration type actuator according to claim 10, further comprising a switching unit capable of switching between the first learned model and the PID controller.

12. The control device for the vibration type actuator includes: a position command unit that outputs the target position; 12. The control device for a vibration-type actuator according to claim 1, further comprising a position detection unit that outputs the detected position.

13. 13. The control device for a vibration type actuator according to claim 1, wherein the controlled variable is at least one of a phase difference, a frequency, and a pulse width.

14. The trained model is 14. The control device for a vibration type actuator according to claim 1, wherein the target speed and the value based on the target position, and the target speed and the predetermined value are input at different times.

15. a vibration actuator that moves a contact body that is in contact with a vibrator relative to the vibrator by vibration generated in the vibrator; A vibration type driving device comprising: a control device for a vibration type actuator according to any one of claims 1 to 14.

16. The vibratory driving device according to claim 15 , a lens that is driven by moving the contact body relatively to the oscillator.

17. The vibratory driving device according to claim 15 , an imaging element that is driven by moving the contact body relatively to the transducer.

18. The vibratory driving device according to claim 15 , a stage that is driven by moving the contact body relative to the oscillator.

Images