Learning controller, learning control method, and magnetic disc device

Abstract

To achieve accuracy improvement of an operation result state of a control target to a target state.SOLUTION: A learning controller includes a feedback control part 30 and a learning control part 40. The feedback control part 30 outputs a feedback signal for making an operation result state track a target state of a control target 34 operating according to an input control signal according to a feedback signal based on an input signal according to an error in the tracking between the operation result state and the target state. The learning control part 40 outputs, to a feedback route F for an input signal according to the tracking error to be inputted into the feedback control part 30, learning control input which has been updated according to the tracking error and is provided for bringing the tracking error closer to zero. An evaluation interval length of an evaluation interval of the tracking error made by the learning control part 40 is longer than an output interval length of an output interval of output of the learning control input to the feedback route F by the learning control part 40.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a learning control device, a learning control method, and a magnetic disk drive. [Background technology]

[0002] A known digital control device is a learning control device that repeatedly controls a controlled object in accordance with a learning control input stored in a learning memory, and sequentially updates the learning control input to be used in the next repeated learning using the tracking error between a target value and the output value of the controlled object, thereby improving control performance.

[0003] However, in conventional technology, while the learning control input and target value updated by learning control are applied to the feedback path, the tracking error is suppressed and the accuracy of the resulting operation state of the controlled object is improved, but the moment the learning control ends, a transient response occurs, and the accuracy of the resulting operation state of the controlled object relative to the target state may deteriorate. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] J. Ishikawa, et al, “A Robust Stability Analysis on Learning Control for Hard Disk Drives”, Advances in Information Storage Systems, pp.49-61 (1999) [Non-patent document 2] Benjamin T. Fine, et al, “Model Inverse Based Iterative Learning Control Using Finite Impulse Response Approximations”, 2009 American Control Conference, pp.931-936 Summary of the Invention [Problem to be solved by the invention]

[0005] An object of the present invention is to provide a learning control device, a learning control method, and a magnetic disk device that can improve the accuracy of the operation result state of a controlled object relative to a target state. [Means for solving the problem]

[0006] A learning control device according to an embodiment includes a feedback control unit and a learning control unit. The feedback control unit outputs a feedback signal based on an input signal corresponding to a tracking error between a target state and an operation result state of a controlled object that operates in response to an input control signal corresponding to the feedback signal, causing the operation result state of the controlled object to track the target state. The learning control unit outputs a learning control input, updated in accordance with the tracking error, to a feedback path that inputs the input signal corresponding to the tracking error to the feedback control unit, for making the tracking error asymptotically approach zero. An evaluation interval length of an evaluation interval for the tracking error evaluated by the learning control unit is longer than an output interval length of an output interval along which the learning control unit outputs the learning control input to the feedback path. [Brief explanation of the drawings]

[0007] [Figure 1] Block diagram of the main components of an HDD. [Figure 2] FIG. 2 is a block diagram showing an example of the configuration of a seek control unit. [Figure 3] FIG. 1 is an explanatory diagram of conventional learning control. [Figure 4] Schematic diagram of a seek control unit. [Figure 5] FIG. 1 is an explanatory diagram of conventional learning control. [Figure 6] FIG. 1 is an explanatory diagram of conventional learning control. [Figure 7] FIG. 10 is a schematic diagram showing the relationship between the evaluation interval length and the output interval length in the prior art. [Figure 8] FIG. 4 is a schematic diagram showing the relationship between the evaluation interval length and the output interval length according to the embodiment. [Figure 9]An explanatory diagram of the electronic control. [Figure 10] A diagram showing an impulse response. [Figure 11] An explanatory diagram of an FIR filter. [Figure 12] Schematic diagram of a learning control unit. [Figure 13] Illustrative diagram of the results of numerical simulation. [Figure 14] Illustrative diagram of the results of numerical simulation. [Figure 15] Illustrative diagram of the results of numerical simulation. [Figure 16] Illustrative diagram of the results of numerical simulation. [Figure 17] Illustrative diagram of the results of numerical simulation. [Figure 18] Illustrative diagram of the results of numerical simulation. [Figure 19] Illustrative diagram of the results of numerical simulation. DETAILED DESCRIPTION OF THE INVENTION

[0008] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a learning control device, a learning control method, and a magnetic disk device according to the present invention will be described in detail with reference to the accompanying drawings.

[0009] 1 is a block diagram showing an example of the configuration of the main part of an HDD (hard disk drive) 10 according to this embodiment. The HDD 10 is an example of a magnetic disk device and a learning control device.

[0010] The host 20 is a device that uses the HDD 10 as a storage device. The host 20 is connected to the HDD 10 via a host interface IF.

[0011] The HDD 10 includes one or more magnetic disks 11, multiple magnetic heads 12, a spindle motor (SPM) 13, a microactuator (MA) 14, a support member 14a, a voice coil motor (VCM) 15, an arm 15a, a driver IC (Integrated Circuit) 16, a head IC 17, and a system LSI (Large Scale Integration) 18.

[0012] The magnetic disk 11 is a magnetic recording medium, and is stacked at regular intervals. The magnetic disk 11 includes an upper disk surface and a lower disk surface. In this embodiment, both sides (upper disk surface and lower disk surface) of the magnetic disk 11 are recording surfaces on which data is magnetically recorded. The magnetic disk 11 is rotated at high speed by the SPM 13. The HDD 10 may include one magnetic disk 11 or multiple magnetic disks 11 (for example, 10 or more).

[0013] The magnetic heads 12 are arranged corresponding to the recording surfaces (upper disk surface and lower disk surface) on both sides of each magnetic disk 11, and are used to write data to and read data written on the recording surfaces of the magnetic disks 11. The magnetic heads 12 are attached to the tip of the support member 14a.

[0014] The SPM 13 is driven by a drive current (or drive voltage) supplied from a driver IC 16 .

[0015] The support member 14a is a member that supports the magnetic head 12, and is provided at the tip of the arm 15a. The support member 14a is, for example, a slider or a suspension.

[0016] The MA 14 is attached to the base or tip of the support member 14a and drives the support member 14a. The MA 14 is, for example, a piezoelectric element such as a piezo element. The MA 14 corresponds to an actuator that drives the tip side of a two-stage actuator. The MA 14 can be driven faster and more accurately than the VCM 15. Therefore, by operating the MA 14 prior to the VCM 15, seek control can be speeded up.

[0017] The VCM 15 rotates the arm 15a. The VCM 15 corresponds to an actuator that moves the head by a relatively large amount. On the other hand, the MA 14 corresponds to an actuator that moves the support member 14a provided at the tip of the arm 15a by a small amount (a relatively small amount).

[0018] The VCM 15 and MA 14 are driven by an input control signal supplied from the driver IC 16. This causes the magnetic head 12 to move radially across the magnetic disk 11. The input control signal is, for example, a drive current or a drive voltage. That is, the magnetic head 12 is an example of a controlled object. The VCM 15 and MA 14 are also examples of a controlled object.

[0019] The driver IC 16 drives the SPM 13, MA 14, and VCM 15 under the control of a CPU (Central Processing Unit) 186 (described later) in the system LSI 18.

[0020] The head IC 17 amplifies a signal (read signal) read by the magnetic head 12. The head IC 17 also converts write data transferred from a later-described R / W (read / write) channel 181 in the system LSI 18 into a write current and outputs it to the magnetic head 12.

[0021] The system LSI 18 is an LSI called an SoC (System On a Chip) in which multiple elements are integrated on a single chip. The system LSI 18 includes an R / W channel 181, a hard disk controller (HDC) 182, a buffer random access memory (RAM) 183, a flash memory 184, a program read only memory (ROM) 185, a CPU 186, and a RAM 187.

[0022] The R / W channel 181 is a signal processing device that processes signals related to reading and writing. The R / W channel 181 digitizes the read signal and decodes the read data from the digitized data. The R / W channel 181 also obtains servo data necessary for positioning the magnetic head 12 from the digital data. The R / W channel 181 also encodes the write data.

[0023] The HDC 182 is connected to the host 20 via the host interface IF. The HDC 182 receives commands (write commands, read commands, etc.) transferred from the host 20. The HDC 182 controls data transfer between the host 20 and the HDC 182. The HDC 182 also controls data transfer between the magnetic disk 11 and the HDC 182.

[0024] The buffer RAM 183 constitutes a buffer area for temporarily storing data to be written to the magnetic disk 11 via the head IC 17 and the R / W channel 181 and data read from the magnetic disk 11 .

[0025] The flash memory 184 is a rewritable non-volatile memory.

[0026] The program ROM 185 stores a control program (firmware). The control program may be stored in a partial area of ​​the flash memory 184. The control program is a program that is used after shipping.

[0027] The CPU 186 functions as a main controller for the HDD 10. The CPU 186 controls at least some of the elements in the HDD 10 in accordance with a control program or an adjustment program stored in a program ROM 185. The CPU 186 includes a seek control unit 19, which will be described later.

[0028] At least a portion of the RAM 187 is used as a working area for the CPU 186 .

[0029] The HDD 10 configured as described above includes a feedback control system that executes seek control of the magnetic head 12. Seek control refers to control for positioning the magnetic head 12 to a target track. The feedback control system is executed by, for example, the CPU 186 at regular intervals, i.e., each time servo data is acquired. Hereinafter, this regular time interval will be referred to as the sample time or sampling period. Furthermore, the step of sampling data at each sampling time will be referred to as the sampling step.

[0030] 2 is a block diagram showing an example of the configuration of the seek control unit 19. The seek control unit 19 executes seek control of the magnetic head 12.

[0031] The seek control unit 19 uses a feedback path F for each of the VCM 15 and the MA 14. In this embodiment, it is assumed that learning control acts on the movement of the VCM 15, and the configuration of the feedback path F for the VCM 15 will be described as an example.

[0032] The seek control unit 19 is a digital control device that repeatedly controls the control target 34 and sequentially updates the learning control input, thereby performing learning control that improves control performance with each repetition.

[0033] The control object 34 is an object controlled by the seek control unit 19 and has an actuator. The control object 34 is an object whose state is controlled by the seek control unit 19, and is, for example, the magnetic head 12 whose position is controlled by the VCM 15 or MA 14 of the HDD 10, semiconductor manufacturing equipment, or a robot. The position of the hand of the arm and the angle of each joint of the control object 34 can be controlled.

[0034] The state of the control object 34 is, for example, the head position of the magnetic head 12 on the magnetic disk 11 or the position of the robot. Note that the state of the control object 34 is not limited to a position. For example, the state of the control object 34 may be a combination of position and velocity, a combination of velocity and acceleration, a combination of position, velocity and acceleration, or a combination of position, velocity, acceleration and external force. It is preferable that the state of the control object 34 includes at least one of the position and velocity of the control object 34. The state of the control object 34 may also include an external force acting on the control object 34. The external force acting on the control object 34 is, for example, a bias force.

[0035] In this embodiment, the control object 34 is the magnetic head 12 whose position is controlled by the VCM 15, and the state of the control object 34 is the head position of the magnetic head 12. This will be described as an example.

[0036] The seek control unit 19 includes a learning control unit 40, a feedback control unit 30, a notch filter 31, an adder 33, a VCM 15, an error calculation unit 35, and an adder .

[0037] The VCM 15 operates in response to input control signals sequentially received from the feedback control unit 30 via the notch filter 31 and the adder 33, and sequentially outputs operation result states representing the state of the operation result. As described above, the input control signal is, for example, a drive current or a drive voltage.

[0038] In this embodiment, the magnetic head 12 sequentially outputs, as an operation result state, the head position of the magnetic head 12 on the magnetic disk 11, which is position-controlled by the VCM 15. The operation result state may be detected by a detection device such as a known sensor or based on servo data acquired via the magnetic head 12.

[0039] The error calculation unit 35 calculates a tracking error. The tracking error represents the error of the operation result state of the control object 34 with respect to the target state. In other words, the tracking error represents the error of the current state with respect to the target state of the control object 34. In this embodiment, the error calculation unit 35 calculates the position error between the head position of the magnetic head 12 and the target trajectory as the tracking error. The target trajectory is the target position of the magnetic head 12 and is an example of the target state.

[0040] The error calculation unit 35 outputs the calculated position error to the learning control unit 40 and the feedback control unit 30. That is, the error calculation unit 35 sequentially receives the head position, which is the operation result state output at each sampling period, and calculates the position error, which is the tracking error with respect to the target trajectory, each time it receives the head position, and outputs it to the learning control unit 40 and the feedback control unit 30.

[0041] The adder 36 adds the position error received from the error calculator 35 and the learning control input received from the learning controller 40 , and outputs the result to the feedback controller 30 .

[0042] The learning control input is a learning value learned by repeated learning by the learning control unit 40. The learning control input is used to correct a signal to be output to a controlled object 34 such as the VCM 15. In other words, the learning control input represents a correction amount used during learning control.

[0043] The feedback control unit 30 outputs a feedback signal for making the operation result state of the control object 34 follow the target state based on an input signal corresponding to the following error between the operation result state of the control object 34 and the target state. In this embodiment, the feedback control unit 30 outputs a feedback signal for making the head position follow the target trajectory based on the position error of the head position of the magnetic head 12, which moves based on the input control signal, with respect to the target trajectory on the magnetic disk 11.

[0044] The notch filter 31 is a filter for stabilizing the mechanical resonance of the VCM 15. The notch filter 31 removes the mechanical resonance frequency component of the VCM 15 from the feedback signal received from the feedback control unit 30, and outputs the signal to the adder 33.

[0045] The adder 33 adds the output signal of the notch filter 31 and the input FF (feedforward) input 32, and outputs the result of the calculation as the input control signal to the VCM 15. The FF input 32 represents an input for feedforward control of the VCM 15. The FF input 32 is expressed in the dimension of, for example, acceleration.

[0046] As described above, the seek control unit 19 is provided with a feedback path F. The feedback path F is a communication path through which an input control signal corresponding to a feedback signal output from the feedback control unit 30 is input to the VCM 15 (controlled object 34), and a tracking error (position error) corresponding to the operation result state (head position) of the controlled object 34 (magnetic head 12) corresponding to the input control signal is input to the feedback control unit 30. The feedback path F also operates to gradually bring the position error between the head position of the magnetic head 12, which is the output of the VCM 15, and the target trajectory of the seek control closer to zero.

[0047] The learning control unit 40 outputs a learning control input, updated according to the tracking error, for making the tracking error approach 0 to a feedback path F that inputs an input signal according to the tracking error to the feedback control unit 30. By providing the learning control unit 40 in the seek control unit 19, it is possible to improve seek settling.

[0048] In this embodiment, the learning control unit 40 outputs a learning control input, updated according to the position error, for making the position error approach zero to a feedback path F that inputs an input signal according to the position error to the feedback control unit 30. Also, in this embodiment, an example will be described in which the learning control unit 40 outputs the learning control input to the input terminal of the feedback control unit 30 on the feedback path F. Therefore, in this embodiment, the adder 36 outputs the sum of the position error and the learning control input to the feedback control unit 30 as an input signal.

[0049] The learning control unit 40 may output a learning control input to the input terminal of the VCM 15, which is the controlled object 34 in the feedback path F. In this case, an adder 36 may be provided at the input terminal of the VCM 15, and the adder 33 may add the output signal of the notch filter 31 and the FF input 32, to obtain a calculation result, to which the learning control input is added, and the result is output as an input control signal to the VCM 15. As described above, the controlled object 34 controlled by the seek control unit 19 may be a semiconductor manufacturing device, a robot, or the like. For example, if the controlled object 34 is a robot, the feedback path F may use the hand position or joint angle of the robot instead of the head position of the magnetic head 12.

[0050] Here, conventional learning control will be described.

[0051] FIG. 3 is an explanatory diagram of an example of conventional learning control. As shown in FIG. 3, conventional learning control improves the head positioning accuracy of the magnetic head 12 during the section (period) in which the learning control input and target trajectory generated by the learning control are applied to the feedback path F. However, conventional learning control can cause a transient response the moment the learning control ends, resulting in a deterioration in head positioning accuracy. In other words, conventional technology can sometimes reduce the accuracy of the operation result state of the controlled object 34 relative to the target state.

[0052] First, the reason why a transient response occurs when learning control is completed will be explained in detail.

[0053] FIG. 4 is a schematic diagram of an example of the seek control unit 19. As shown in FIG.

[0054] In order to analyze the operation of the learning control, it is assumed that each element of the seek control unit 19 described with reference to Fig. 2 is a linear system, and the time response of the position error during the seek control when the learning control is operating is assumed. This assumption is based on the impulse response sequence p1, p2,...p of the closed loop transfer characteristic from e to u that connects the learning control unit 40 with the feedback path F shown in Fig. 4. n-1 ,p n is used. e represents the tracking error. That is, in this embodiment, e represents the position error. u represents the learning control input. The impulse response sequence is a group of position errors e for each sampling time output from the controlled object 34 over time when one learning control input u is output from the learning control unit 40 to the feedback path F. n is an integer equal to or greater than 2. The numerical subscript under P represents the sampling time i=0, 1, 2,...n.

[0055] The vector in which the above position errors are arranged for each sampling time is e, and the impulse response sequence p1, p2, p n-1 ,p nLet P be the matrix in which the above is arranged in the lower triangular part of equation (2) described later, shifted by one sampling time, u be the vector in which the learning control inputs generated by the learning control by the learning control unit 40 are arranged for each sampling time, and d be the vector of position errors in seek control when learning control is disabled, then the following equation (1) is established.

[0056]

number

[0057] In equation (1), e represents the vector of position errors, and P represents the impulse response sequence p1, p2, p n-1 ,p n is a matrix in which the vectors are arranged while being shifted by one sampling time, u is a vector of learning control input, and d is a vector of position error when learning control is disabled. In the following explanation, the position error vector will be referred to as the position error vector, and the learning control input vector will be referred to as the learning control input vector.

[0058] For example, if the output section length of the output section plan of the learning control input output from the learning control unit 40 to the feedback path F is set to "7", and the content of the above equation (1) is expressed for each sampling time i=0, 1, 2,...n, the result is equation (2).

[0059]

number

[0060] In general learning control, the learning control unit 40 stores the learning control input in a memory provided in the learning control unit 40 for each seek control, and repeatedly updates the learning control input based on the position error in the next seek control.

[0061] 5 and 6 are explanatory diagrams of an example of conventional learning control.

[0062] In the prior art, when updating the learning control input stored in memory, position error signals e1 to e7 with an evaluation interval length that is the same as the output interval length of the output interval of the learning control input output from the learning control unit to the feedback path F are used. Therefore, when learning has sufficiently converged, as shown in Figure 5, the learning control input is divided into P and P2, where the learning control inputs u0 to u6 act on the matrix P and vector d, respectively. sq and d sq and converges to equation (3).

[0063]

number

[0064] The learning control input u after convergence in the conventional technology determined by equation (3) sq As shown in FIG. 6, the range of position errors e0 to e7 is 0, and the position error is completely suppressed in principle. However, in reality, the position error e after the position error e7 8··· does not necessarily become 0, and this appears as a transient response at the end of learning control between the position error e7 and the position error e8.

[0065] As described above, in the conventional technology, a transient response occurs the moment learning control ends, which can degrade head positioning accuracy. In other words, in the conventional technology, the accuracy of the operation result state of the control object 34 relative to the target state can decrease.

[0066] Therefore, in the learning control unit 40 of this embodiment, the evaluation interval length N of the evaluation interval for the position error, which is the tracking error, is controlled to be longer than the output interval length M of the output interval in which the learning control unit 40 outputs the learning control input to the feedback path F.

[0067] FIG. 7 is a schematic diagram showing an example of the relationship between the evaluation interval length N of the evaluation interval S1 and the output interval length M of the output interval S2 in the prior art. In FIG. 7, the black circles indicate the sampling points of the position error and the learning control input. As shown in FIG. 7, in the prior art, the evaluation interval length N and the output interval length M of the output interval S2 were the same. Also, in the prior art, the number of samples in the evaluation interval S1 and the number of outputs in the output interval S2 were the same.

[0068] FIG. 8 is a schematic diagram showing an example of the relationship between the evaluation interval length N of the evaluation interval S1 and the output interval length M of the output interval S2 in this embodiment. In FIG. 8, the dots indicated by black circles indicate the sampling points of the position error and the learning control input. As shown in FIG. 8, in this embodiment, the evaluation interval length N of the position error is longer than the output interval length M of the learning control input. In this embodiment, the number of outputs in the output interval S2 is less than the number of samplings in the evaluation interval S1. In this embodiment, the timing of the last sampling (e6) in the evaluation interval S1 is closer to the timing of the last output (u6) in the output interval S2. 10 ) is located later. In this embodiment, for example, the sampling time interval in the evaluation section S1 and the output time interval in the output section S2 match.

[0069] As shown in Figure 8, the evaluation section S1 and the output section S2 partially overlap. Furthermore, the evaluation section S1 includes a period that is later (later in time) than the output section S2. Specifically, as shown in Figure 8, the start timing of the evaluation section S1 coincides with the start timing of the output section S2, and the end timing of the evaluation section S1 is later than the end timing of the output section S2.

[0070] That is, in this embodiment, in order to mitigate the transient response, the learning control unit 40 adjusts the position errors e8 to e9 after the application of the learning control inputs u0 to u6 to the feedback path F is completed so as to satisfy the relationship of the evaluation interval length N>the output interval length M. n Repeated learning will be conducted, including the following.

[0071] Therefore, as shown in FIG. 9, the learning control unit 40 expands the portion where the learning control inputs u0 to u6 act on the matrix P and the vector d compared to the prior art (see FIG. 5). FIG. 9 is an explanatory diagram of an example of the learning control of the present application. Then, the learning control unit 40 converts the expanded matrix P and the expanded vector d into P ls and d ls By performing iterative learning to find a least squares solution to the following equation (4), the evaluation interval length N of the position error is controlled to be longer than the output interval length M of the learning control input.

[0072]

number

[0073]

number

[0074] In formula (4), u ls represents the vector of learning control inputs, and P ls represents a matrix of impulse response sequences arranged at intervals of one sampling time, and d ls represents the vector of the tracking error when learning control is disabled, and equation (4A) in equation (4) represents the pseudo-inverse matrix operation.

[0075] The learning control input u in equation (4) ls According to (2), the position error signal e does not, in principle, converge completely to 0. However, unlike conventional learning control, equation (4) also takes into account the position error after application of the learning control input to the feedback path F has finished. Therefore, by having the learning control unit 40 perform iterative learning to find a least-squares solution to equation (4), it is possible to control the evaluation interval length N of the position error to be longer than the output interval length M of the learning control input, thereby making it possible to mitigate the above-mentioned transient response.

[0076] Furthermore, it is considered that the learning control unit 40 repeatedly performs learning using equation (4) to repeatedly update the learning control input for each seek control. In this case, since equation (4) is a least squares solution, the repeated update of the learning control input is performed using the position error vector e ls Using the gradient method with the objective function of minimizing the square norm of ls can be approximately calculated by repeated calculations. The update formula for this repeated calculation is expressed by the following formula (5).

[0077]

number

[0078] In formula (5), u ls represents the vector of learning control inputs, and P ls represents a matrix in which the impulse response sequence is shifted by one sampling time, and e ls represents the vector of position error. In addition, in equation (5), j represents the number of learning iterations, and β represents the learning gain.

[0079] That is, the learning control unit 40 repeatedly performs learning using equation (5), thereby realizing control such that the evaluation interval length N of the position error is longer than the output interval length M of the learning control input.

[0080] Here, the matrix P composed of the impulse response sequence of the closed-loop transfer characteristic in equation (5) ls In practice, it may be difficult to accurately calculate the impulse response of the closed-loop transfer characteristic model. Therefore, some approximation is necessary. The learning control unit 40 calculates the closed-loop transfer characteristic model based on the model of the VCM 15 and obtains the impulse response.

[0081] FIG. 10 is a diagram showing an example of an impulse response when a closed-loop transfer characteristic model is calculated based on the model of the VCM 15. In FIG.

[0082] The learning control unit 40 truncates the obtained impulse response to a length m that can roughly approximate the waveform shape of the impulse response, and extracts it (see FIG. 10). Then, the learning control unit 40 extracts the extracted impulse response sequence p1, p2, . . . p m-1 ,p m From P ls The approximate matrix of is constructed by the following equation (6): Then, the learning control unit 40 performs repeated learning using equation (7), thereby realizing control so that the evaluation interval length N of the position error is longer than the output interval length M of the learning control input.

[0083]

number

[0084]

number

[0085] The left side of equation (6), equation (6A), is P in equation (5). ls In equation (6), p1 to p m represents a sequence of m impulse responses.

[0086] Also, P ls The learning update equation using the approximate matrix is ​​expressed by the following equation (7).

[0087]

number

[0088] In formula (7), u ls represents the vector of learning control inputs, and e ls represents the vector of position error. In equation (7), j represents the number of learning iterations, and β represents the learning gain. In equation (7), equation (6A) represents equation (6).

[0089]

number

[0090] Here, the part of equation (7) in equation (7) is a matrix x vector operation, which may be difficult to perform with the computing power of the CPU 186 (see FIG. 1) of the actual HDD 10.

[0091] Therefore, in the learning control unit 40, it is preferable to replace the calculation of the part of equation (7A) with an FIR (Finite Impulse Response) filter.

[0092] This FIR filter uses the impulse response sequence p1, p2, p m-1 ,p m It is sufficient to use a filter expressed by the following equation (8) having coefficients in reverse time order.

[0093]

number

[0094] In equation (8), F(z) represents a filter, and z represents a delay operator. m represents a sequence of m impulse responses.

[0095] The FIR filter expressed by equation (8) is applied with the position error vector e ls (j) If the output vector when input is y, then the remaining part of y after removing the delay m-1 of the FIR filter is equal to the vector obtained by calculating the above equation (7A) in the above equation (7), as shown in Figure 11. Figure 11 is an explanatory diagram of an FIR filter.

[0096] FIG. 12 is a schematic diagram of an example of the configuration of the learning control unit 40 of this embodiment, including an FIR filter.

[0097] The learning control unit 40 includes an FIR filter 41, a gain multiplication unit 42, an addition unit 43, and a memory 44.

[0098] The memory 44 is a memory for storing the learning control input for each sampling step i. The memory length of the memory 44 may be any length that matches the output interval length M.

[0099] The FIR filter 41 receives the position error e from the feedback path F at each sampling step i. ls (j) The position error signal obtained by filtering [i] is output to the gain multiplication unit 42. The gain multiplication unit 42 multiplies the position error signal received from the FIR filter 41 by a gain β. The addition unit 43 adds the multiplication result of the position error signal multiplied by the gain β to the learning control input read from the memory 44, and updates the learning control input stored in the memory at the sampling step im-d+1 among the memories included in the memory 44 with the addition result. m represents the FIR filter length. m-d+1 corresponds to both the phase lag of the controlled object 34 representing the closed-loop transfer characteristic shown in FIG. 4 and the phase lag of the FIR filter 41.

[0100] Therefore, the learning control input of the sampling step im-d+1 going back from the current time i, which is stored in the memory 44, is updated in accordance with the newly observed position error.

[0101] As described above, in this embodiment, in order to prevent learning from becoming unstable due to the influence of the phase delay of the controlled object 34, which represents the closed-loop transfer characteristic shown in Fig. 4, and the phase delay of the FIR filter 41, the learning control unit 40 performs processing to update the memory by the amount of both phase delays, going back from the current time i. In other words, the time at which the learning control unit 40 reads the value stored in the memory 44 from the memory 44 as a learning control input for outputting the value to the feedback path F is a time in the future with respect to the time at which the memory 44 is updated. For this reason, the learning control unit 40 can advance the phase and correct the phase delay by performing processing to update the memory by the amount of both phase delays, going back from the current time i.

[0102] As described above, the HDD 10 (learning control device, magnetic disk device) of this embodiment includes the feedback control unit 30 and the learning control unit 40.

[0103] The feedback control unit 30 outputs a feedback signal for causing the operation result state of the control object 34 to track the target state (target trajectory) based on an input signal corresponding to a tracking error (position error) between the operation result state (head position) of the control object 34 (magnetic head 12), which operates in response to an input control signal corresponding to the feedback signal, and a target state. The learning control unit 40 outputs a learning control input, updated in accordance with the tracking error, for making the tracking error approach zero to a feedback path F that inputs an input signal corresponding to the tracking error to the feedback control unit 30. The evaluation section length N of the evaluation section S1 of the tracking error by the learning control unit 40 is longer than the output section length M of the output section S2 in which the learning control unit 40 outputs the learning control input to the feedback path F (N>M).

[0104] Therefore, the learning control unit 40 of this embodiment can suppress the transient response that occurs the moment learning control is completed, which occurs in conventional technology, and can suppress deterioration in accuracy of the operation result state of the controlled object relative to the target state.

[0105] Therefore, the HDD 10 (learning control device, magnetic disk device) of this embodiment can improve the accuracy of the operation result state of the control target 34 relative to the target state.

[0106] Furthermore, when the HDD 10 of this embodiment is a magnetic disk drive, the positioning accuracy of the magnetic head 12 can be improved.

[0107] (effect) 13 to 19 are explanatory diagrams showing the results of numerical simulations of the HDD 10 of this embodiment and the conventional technology.

[0108] 13, 14, and 15 are diagrams respectively showing boat diagrams of the model of VCM 15, the notch filter 31, and the feedback control unit 30 shown in Fig. 2. Figs. 16 and 17 are diagrams showing the time history of the target trajectory and the FF input 32 in Fig. 2. For the model of VCM 15, the notch filter 31, and the feedback control unit 30, those exemplified in the reference material "http: / / www2.iee.or.jp / ~dmec / committee / DMEC1005 / dsa_HDD_bench_e.html" (HDD benchmark problem) were used.

[0109] 18 and 19 are diagrams comparing the results of a numerical simulation of the operation of conventional learning control and the operation of learning control of this embodiment using the above components. Fig. 18 is a diagram showing the results of a numerical simulation of the operation of conventional learning control. Fig. 19 is a diagram showing the results of a numerical simulation of the operation of learning control of this embodiment.

[0110] 18 and 19, the output interval length M of the output interval S2 of the learning control input is set to a length of 48 samples. Furthermore, the evaluation interval length N of the evaluation interval S1 of the position error in the conventional learning control shown in Fig. 18 is set to a length of 48 samples. The evaluation interval length N of the evaluation interval S1 of the position error in the learning control of this embodiment shown in Fig. 19 is set to a length of 78 samples. That is, in the learning control of this embodiment shown in Fig. 19, the length of the position error signal included in learning after the end of learning control to mitigate the transient response is set to 30 samples.

[0111] 18 and 19 show diagrams representing both the head position trajectory under seek control without learning control and the head position trajectory after learning has sufficiently converged. Due to the effects of learning control, in both cases, the head position closely follows the target trajectory indicated by the dashed line in the diagram during the section in which the learning control input is applied to feedback path F. However, with the conventional learning control shown in FIG. 18, a transient response occurs the moment the application of the learning control input ends, causing the head position to be offset. On the other hand, with the learning control of this embodiment shown in FIG. 19, it can be seen that this transient response is alleviated.

[0112] As described above, in this embodiment, it is possible to suppress the transient response that occurs the moment learning control ends, which occurs in conventional technology, and to suppress deterioration in the accuracy of the operation result state of the controlled object relative to the target state. Therefore, the HDD 10 (learning control device, magnetic disk drive) of this embodiment can improve the accuracy of the operation result state of the controlled object 34 relative to the target state. Furthermore, when the HDD 10 of this embodiment is a magnetic disk drive, it can improve the positioning accuracy of the magnetic head 12.

[0113] Each unit of the seek control unit 19 shown in Fig. 2 is realized by, for example, one or more processors. For example, each unit may be realized by having a processor such as the CPU 186 execute a program, i.e., by software. Each unit may be realized by a processor such as a dedicated IC (Integrated Circuit), i.e., by hardware. Each unit may be realized by a combination of software and hardware. When multiple processors are used, each processor may realize one of the units, or two or more of the units.

[0114] The programs executed by the magnetic disk device and the learning control device of the embodiment are provided in advance in a program ROM or the like.

[0115] The programs executed by the magnetic disk device and learning control device of the embodiments may be configured to be provided as a computer program product by being recorded in an installable or executable format on a computer-readable recording medium such as a CD-ROM (Compact Disk Read Only Memory), a flexible disk (FD), a CD-R (Compact Disk Recordable), or a DVD (Digital Versatile Disk).

[0116] Furthermore, the programs executed by the magnetic disk device and the learning control device according to the embodiments may be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. Also, the programs executed by the magnetic disk device and the learning control device according to the embodiments may be provided or distributed via a network such as the Internet.

[0117] The programs executed by the magnetic disk device and learning control device according to the embodiment can cause a computer to function as each part of the magnetic disk device described above. In this computer, the CPU 186 can read the programs from a computer-readable storage medium onto a main storage device and execute them.

[0118] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0119] 10 HDD 11 Magnetic Disk 12 Magnetic head 19 Seek control section 30 Feedback control section 34 Control Object 40 Learning control unit 41 FIR Filters 43 Addition section

Claims

1. a feedback control unit that outputs a feedback signal to cause a resultant state of an operation of a controlled object to follow a target state, based on an input signal corresponding to a tracking error between the resultant state of an operation of the controlled object, which operates in response to an input control signal corresponding to a feedback signal, and the target state; a learning control unit that outputs a learning control input, updated according to the tracking error, to a feedback path that inputs the input signal according to the tracking error to the feedback control unit, for making the tracking error gradually approach zero; Equipped with an evaluation interval length of an evaluation interval of the tracking error by the learning control unit is longer than an output interval length of an output interval in which the learning control unit outputs the learning control input to the feedback path; Learning control device.

2. The learning control unit Perform iterative learning to find the least squares solution of equation (4). The learning control device according to claim 1 . [Equation 1] [In formula (4), u ｌｓ represents the vector of learning control inputs, and P ｌｓ represents a matrix in which impulse response sequences are arranged with a shift of one sampling time, and d ｌｓ represents a vector of the tracking error when learning control is disabled, and equation (4A) in equation (4) represents a pseudo-inverse matrix operation.]

3. The learning control unit Iterative learning is performed using equation (5). The learning control device according to claim 2 . [Equation 2] [In formula (5), u ｌｓ represents the vector of learning control inputs, and P ｌｓ represents a matrix in which impulse response sequences are arranged with a shift of one sampling time, and e ｌｓ represents a vector of position errors. In addition, in equation (5), j represents the number of learning iterations, and β represents the learning gain.

4. The learning control unit Iterative learning is performed using equation (7). The learning control device according to claim 3 . [Equation 3] [The left side of the formula (6) is the formula (6A) in the formula (5). ｌｓ In equation (6), p １ ~p ｍ represents a sequence of m impulse responses. ｌｓ represents the vector of learning control inputs, and e ｌｓ represents a vector of position errors. In equation (7), j represents the number of learning iterations, and β represents the learning gain. In equation (7), equation (6A) represents equation (6).

5. The learning control unit the tracking error sampled sequentially from the feedback path is input to a filter represented by equation (8), and the learning control input is updated using the tracking error output from the filter; The learning control device according to claim 4 . [Equation 4] [In equation (8), F(z) represents a filter, z represents a delay operator, and p １ ~p ｍ represents a sequence of m impulse responses.]

6. The evaluation interval and the output interval partially overlap. The learning control device according to claim 1 .

7. the start timing of the evaluation interval is the same as the start timing of the output interval, the end timing of the evaluation interval is a timing after the end timing of the output interval, The learning control device according to claim 1 .

8. the object to be controlled is a magnetic head, the operation result state is a head position of the magnetic head on the magnetic disk, the target state is a target trajectory of the magnetic head, The tracking error is a position error. The learning control device according to claim 1 .

9. A learning control method executed by a learning control device, an output step of outputting a feedback signal for causing a resultant state of an operation of a controlled object to follow a target state, based on an input signal corresponding to a tracking error between the resultant state of an operation of the controlled object, which operates in response to an input control signal corresponding to a feedback signal, and the target state; a learning control step of outputting a learning control input, updated according to the tracking error, for making the tracking error approach zero, to a feedback path that inputs the input signal according to the tracking error to the output step; Including, an evaluation interval length of the evaluation interval of the tracking error in the learning control step is longer than an output interval length of an output interval in which the learning control step outputs the learning control input to the feedback path; Learning control method.

10. a feedback control unit that outputs a feedback signal for causing the head position to follow a target trajectory on a magnetic disk, based on an input signal corresponding to a position error of the head position relative to a target trajectory of the magnetic head, the head position of which moves in response to an input control signal corresponding to a feedback signal; a learning control unit that outputs a learning control input, updated in accordance with the position error, for making the position error gradually approach zero, to a feedback path that inputs the input signal in accordance with the position error to the feedback control unit; Equipped with an evaluation section length of an evaluation section of the position error by the learning control unit is longer than an output section length of an output section in which the learning control unit outputs the learning control input to the feedback path; Magnetic disk device.

Images