Information processing method, information optimization method, information processing device, and program
By gradually selecting and adding representative actions and updating control parameters based on measurement data, the problem of difficult to efficiently generate control parameters in the prior art is solved, and more efficient control parameter optimization is achieved.
Patent Information
- Application Number
- CN202380034475.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-21
- Filing Date
- 2023-03-20
- Publication Date
- 2024-12-06
AI Technical Summary
When generating control parameters, the prior art is difficult to efficiently search for appropriate control parameters under huge combinations due to the large search range, and may even be unable to search for appropriate parameters.
By selecting representative actions among all actions that can be performed by the production device, the number and complexity of the actions are gradually increased, the evaluation indicators are calculated based on the measurement data, the control parameters are updated, and the processing is repeatedly optimized to shorten the setting time or reduce the extent of the excess.
It realizes efficient and reliable generation of appropriate control parameters under a large number of combinations, shortens the setting time or reduces the degree of excess, and improves the optimization efficiency of control parameters.
Smart Images

Figure CN119096203A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to an information processing method, an information optimization method, an information processing device, and a program. Background Art
[0002] Conventionally, there is known a control parameter generation method for generating a control parameter (for example, refer to Patent Document 1).
[0003] Conventionally, it has been desired to efficiently generate appropriate control parameters.
[0004] Prior art literature
[0005] Patent Literature
[0006] Patent Document 1: International Patent Publication No. 2018 / 151215 Summary of the invention
[0007] An object of the present disclosure is to provide an information processing method, an information optimization method, an information processing device, and a program that can efficiently generate appropriate control parameters.
[0008] Solutions to the problem
[0009] An information processing method involved in one embodiment of the present invention is used to optimize the control parameters in a device that performs multiple actions based on multiple control parameters, allowing the information processing device to select a representative action that includes at least one action among all the actions that the device can perform during the control parameter optimization process, and allowing the device to perform the selected representative action, obtain measurement data related to the action of the device measured during the execution of the representative action, calculate an evaluation value of a specified evaluation indicator based on the obtained measurement data, update the control parameters based on the calculated evaluation value, and perform the control parameter optimization process multiple times while changing the evaluation benchmark of the control parameter optimization process.
[0010] In an information optimization method involved in one embodiment of the present invention, an information processing device uses a first evaluation criterion to perform a first optimization process of information, sets a search range of the information based on multiple evaluation values calculated in the first optimization process, and performs a second optimization process of the information using a second evaluation criterion for the set search range.
[0011] An information processing device involved in one embodiment of the present invention is used to optimize the control parameters in a device that performs multiple actions based on multiple control parameters, and includes a selection unit, a control unit, an acquisition unit, a calculation unit and an update unit. In the control parameter optimization process, the selection unit selects a representative action that includes at least one action among all the actions that the device can perform, the control unit instructs the device to perform the representative action selected by the selection unit, the acquisition unit acquires measurement data related to the action of the device measured during the execution of the representative action, the calculation unit calculates an evaluation value of a specified evaluation indicator based on the measurement data acquired by the acquisition unit, and the update unit updates the control parameters based on the evaluation value calculated by the calculation unit, and the control parameter optimization process is performed multiple times while changing the evaluation benchmark of the control parameter optimization process.
[0012] A program involved in one embodiment of the present disclosure is used to enable an information processing device to function as a selection unit, a control unit, an acquisition unit, a calculation unit and an update unit. The information processing device is used to optimize the control parameters in a device that performs multiple actions based on multiple control parameters. In the control parameter optimization process, the selection unit selects a representative action that includes at least one action among all the actions that the device can perform, the control unit causes the device to perform the representative action selected by the selection unit, the acquisition unit acquires measurement data related to the action of the device measured during the execution of the representative action, the calculation unit calculates an evaluation value of a specified evaluation indicator based on the measurement data acquired by the acquisition unit, and the update unit updates the control parameters based on the evaluation value calculated by the calculation unit, and the control parameter optimization process is performed multiple times while changing the evaluation criteria of the control parameter optimization process.
[0013] Effects of the Invention
[0014] According to the present disclosure, appropriate control parameters can be efficiently generated. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 This is a schematic diagram showing an overview of the control parameter generation system according to the first embodiment.
[0016] Figure 2 This is a block diagram showing the structure of the control parameter generation system involved in the first embodiment.
[0017] Figure 3 This is a perspective view showing an example of a production apparatus according to the first embodiment.
[0018] Figure 4 This is a data structure diagram showing an example of control parameters stored in the memory involved in the first embodiment.
[0019] Figure 5 This is a schematic diagram showing an example of the transition of the position deviation of the driven object with respect to the target position.
[0020] Figure 6 This is a data structure diagram showing an example of measurement data output by the sensor according to the first embodiment.
[0021] Figure 7 This is a flowchart of the first control parameter generation process.
[0022] Figure 8 This is a timing chart of the first control parameter generation process.
[0023] Fig. 9 This is a schematic diagram showing an example of an image displayed on the display unit according to the first embodiment.
[0024] Fig.10 This is a flowchart of the condition acquisition process.
[0025] Fig.11 This is a schematic diagram showing an example of an image displayed on the display unit according to the first embodiment.
[0026] Fig.12 This is a flowchart representing the action designation process.
[0027] Fig.13 This is a schematic diagram showing an example of an image displayed on the display unit according to the first embodiment.
[0028] Fig.14 It is a schematic diagram showing the experimental results of the experiments conducted by the inventors.
[0029] Fig.15 This is a block diagram showing the structure of the control parameter generation system involved in the second embodiment.
[0030] Fig.16A This is a schematic diagram showing an example of the transition of the position deviation of the driven object with respect to the target position.
[0031] Fig. 16B This is a schematic diagram showing an example of the transition of the position deviation of the driven object with respect to the target position.
[0032] Fig.17 This is a flowchart of the second control parameter generation process.
[0033] Fig.18 This is a flowchart of the third control parameter generation process.
[0034] Fig.19This is a schematic diagram showing an example of a case where the determination unit according to the third embodiment selects X representative actions from Y actions.
[0035] Fig. 20 This is a flowchart of the parameter update process using all actions.
[0036] Fig.21 This is a flowchart of the first control parameter generation process.
[0037] Fig. 22 This is a flowchart showing the flow of processing executed by the information processing unit.
[0038] Fig.23 This is a diagram schematically showing a method for setting a search range by a generation unit.
[0039] Fig.24 This is a diagram showing in simplified form a portion of the operation condition information describing the execution conditions of each operation executed by the production device.
[0040] Fig.25 This is a flowchart showing the flow of processing executed by the information processing unit in association with notifying the user of the status information.
[0041] Fig.26 This is a diagram showing an example of a status notification image. DETAILED DESCRIPTION
[0042] (How one method of the present disclosure was obtained)
[0043] In recent years, a method has been proposed as a method for generating appropriate control parameters for a driving source in a production device (e.g., an installation device) that includes a driving source (e.g., a servo motor) that drives a driven object, wherein appropriate control parameters are searched for using a machine learning model, etc., thereby generating appropriate control parameters (e.g., refer to Patent Document 1).
[0044] Generally speaking, the number of control parameters for a driving source of a production device may reach 50 or more, and the number of adjustment levels of the control parameters may reach 100 or more.
[0045] For example, when the production device performs 80 actions, the number of control parameters of the drive source is 50, and the adjustment level of the control parameters is 100, the number of these combinations reaches 100. 50 ×80 pcs.
[0046] The inventors have discovered the following problem: When generating control parameters for such a large number of combinations, when using methods such as machine learning models to search for appropriate control parameters, due to the large search range, it sometimes takes a lot of time to search for appropriate control parameters, or it sometimes fails to search for appropriate control parameters no matter how much time is spent.
[0047] Therefore, the inventors have intensively conducted experiments and studies repeatedly in order to realize a parameter generation method that can efficiently generate appropriate control parameters even when generating control parameters for a huge number of combinations.
[0048] Through the above experiments and studies, the inventors have obtained the following insights: instead of searching for appropriate control parameters from the very beginning by targeting all the huge combinations, for example, when the production device performs 80 actions, as the initial step, only the combinations related to several actions among the 80 actions, such as one action, are targeted, that is, only the combinations of a part of the huge combinations are targeted to search for appropriate control parameters; as the next step, the number of actions performed by the production device is increased, for example, combinations related to two actions are targeted, and appropriate control parameters are searched based on the search results in the initial step; as the next step, the number of actions performed by the production device is further increased, for example, combinations related to four actions are targeted, and appropriate control parameters are searched based on the search results in the previous step; these steps are repeated, and finally appropriate control parameters are searched based on the combinations related to all 80 actions. As a result, appropriate control parameters can be generated more efficiently and reliably by targeting the combinations related to all 80 actions.
[0049] Next, the inventors further repeated experiments and studies based on this knowledge, and conceived the control parameter generation method and the like according to the present disclosure described below.
[0050] The information processing method involved in the first mode of the present disclosure is used to optimize the control parameters in a device that performs multiple actions based on multiple control parameters, allowing the information processing device to select a representative action that includes at least one action among all the actions that the device can perform during the control parameter optimization process, and the device executes the selected representative action, obtains measurement data related to the action of the device measured during the execution of the representative action, calculates an evaluation value of a specified evaluation indicator based on the obtained measurement data, updates the control parameters based on the calculated evaluation value, and executes the control parameter optimization process multiple times while changing the evaluation benchmark of the control parameter optimization process.
[0051] According to the first aspect, the control parameter optimization process is executed multiple times while changing the evaluation criterion of the control parameter optimization process. Thus, by repeatedly executing the control parameter optimization process while changing the evaluation criterion, appropriate control parameters can be efficiently generated.
[0052] The information processing method according to the second aspect of the present disclosure may be such that, in the information processing method according to the first aspect, changing the evaluation criterion includes changing the representative action.
[0053] According to the second aspect, the evaluation criteria can be changed by changing the representative action.
[0054] The information processing method according to the third aspect of the present disclosure may be such that, in the information processing method according to the second aspect, changing the representative action includes changing the number of actions included in the representative action.
[0055] According to the third aspect, the evaluation criteria can be changed by changing the number of actions included in the representative action.
[0056] The information processing method according to a fourth aspect of the present disclosure may be: in the information processing method according to the third aspect, changing the number of actions includes increasing the number of actions.
[0057] According to the fourth aspect, the number of actions included in the representative action is gradually increased, and the control parameter optimization process is repeatedly performed, thereby enabling efficient generation of appropriate control parameters.
[0058] The information processing method according to the fifth aspect of the present disclosure may be such that, in the information processing method according to the second aspect, changing the representative action includes changing the content of an action included in the representative action.
[0059] According to the fifth aspect, the evaluation criteria can be changed by changing the content of the actions included in the representative action.
[0060] The information processing method according to a sixth aspect of the present disclosure may be such that, in the information processing method according to the fifth aspect, changing the content of the action includes complicating the content of the action.
[0061] According to the sixth aspect, appropriate control parameters can be efficiently generated by repeatedly executing the control parameter optimization process while gradually making the contents of the motions included in the representative motion more complex.
[0062] The information processing method involved in the seventh embodiment of the present invention may be: in the information processing method of any one of the second to sixth embodiments, in the change of the representative action, all the actions are classified into multiple groups based on similarity, and based on the evaluation value or the past average of the evaluation value, the action that needs to be improved is determined from the multiple actions contained in each of the multiple groups, and the multiple actions that need to be improved determined for the multiple groups are selected as the representative actions in the next control parameter optimization processing.
[0063] According to the seventh aspect, a plurality of actions requiring improvement determined for a plurality of groups are selected as representative actions in the next control parameter optimization process, thereby enabling efficient generation of appropriate control parameters.
[0064] The information processing method according to the eighth aspect of the present disclosure may be the information processing method according to any one of the first to seventh aspects, wherein changing the evaluation criterion includes changing the evaluation index.
[0065] According to the eighth aspect, the evaluation criteria can be changed by changing the evaluation index.
[0066] The information processing method involved in the ninth aspect of the present disclosure may be: in the information processing method of the eighth aspect, each of the multiple actions is an action in which the device moves the object to the target position, and the evaluation index includes a settling time, which is the time taken for the position deviation between the position of the object and the target position to converge to an allowable range.
[0067] According to the ninth aspect, the settling time is used as an evaluation index, thereby making it possible to appropriately evaluate the control parameter.
[0068] The information processing method involved in the tenth embodiment of the present disclosure may be: in the information processing method of the eighth embodiment, each of the multiple actions is an action in which the device moves the object to the target position, and the evaluation index includes the excess time or the cumulative value of the position deviation amount between the position of the object and the target position that exceeds the allowable range.
[0069] According to the tenth aspect, the control parameter can be appropriately evaluated by using the accumulated value of the excess time or the position deviation amount as the evaluation index.
[0070] The information processing method involved in the eleventh embodiment of the present disclosure may be: in any one of the information processing methods from the first embodiment to the tenth embodiment, the search range of the control parameters for updating candidates in the next control parameter optimization process is set based on the multiple evaluation values calculated in the current control parameter optimization process.
[0071] According to the eleventh aspect, the setting information of the search range of the control parameter is inherited from the current control parameter optimization process to the next control parameter optimization process, thereby making it possible to efficiently search for appropriate control parameters in the next control parameter optimization process.
[0072] The information processing method involved in the twelfth embodiment of the present disclosure may be: in the information processing method of any one of the first to eleventh embodiments, in the comprehensive evaluation processing, the device is instructed to execute a prescribed evaluation action for comprehensively evaluating the action of the device, measurement data related to the action of the device measured during the execution of the evaluation action is obtained, an evaluation value of the evaluation indicator is calculated based on the obtained measurement data, and the comprehensive evaluation is performed based on the calculated evaluation value.
[0073] According to the twelfth aspect, the progress of optimization of the control parameters in the overall operation of the device can be evaluated through the comprehensive evaluation process.
[0074] The information processing method according to the thirteenth aspect of the present disclosure may be such that, in the information processing method according to the twelfth aspect, the evaluation operation is all the operations.
[0075] According to the thirteenth aspect, by causing the device to execute all actions as evaluation actions, accurate comprehensive evaluation can be performed.
[0076] The information processing method according to the fourteenth aspect of the present disclosure may be: in the information processing method according to the twelfth aspect, the evaluation action is a difference action between the entire action and the representative action closest to the evaluation action.
[0077] According to the fourteenth aspect, by causing the device to execute a difference action between all actions and the nearest representative action of the evaluation action as the evaluation action, it is possible to maintain the accuracy of the comprehensive evaluation and realize the efficiency of the evaluation action.
[0078] The information processing method according to the fifteenth aspect of the present disclosure may be, in the information processing method according to the twelfth aspect, wherein the evaluation operation is a predetermined operation predetermined from among all the operations.
[0079] According to the fifteenth aspect, by causing the device to execute a predetermined operation that is predetermined from among all operations as the evaluation operation, it is possible to improve the efficiency of the evaluation operation.
[0080] The information processing method involved in the sixteenth mode of the present disclosure may be: in the information processing method of any one of the twelfth to fifteenth modes, it is also determined whether the evaluation result of the comprehensive evaluation satisfies the prescribed termination condition, and when the evaluation result satisfies the termination condition, the control parameter optimization processing is terminated.
[0081] According to the sixteenth aspect, when the evaluation result of the comprehensive evaluation satisfies the predetermined end condition, the control parameter optimization process is ended even before the control parameter optimization process reaches the final stage, so that the time required for the process can be shortened.
[0082] The information processing method involved in the seventeenth embodiment of the present disclosure may be: in the information processing method of any one of the first to sixteenth embodiments, each of the multiple actions is an action of the device to move the object to the target position, and status information is also output to notify the user, and the status information is related to at least one of the following information: the elapsed time from the start of the control parameter optimization processing, the remaining time until the maximum time to continue the control parameter optimization processing, identification information related to the currently executed action of at least one action contained in the representative action, an action condition of at least one of the movement amount, movement speed, acceleration time and deceleration time, the number of actions contained in the representative action, the number of updates of the control parameter, and the remaining number of times until the maximum number of updates of the control parameter.
[0083] According to the seventeenth aspect, status information is output to notify the user, whereby the user can easily confirm the progress status or operation conditions of the control parameter optimization process based on the notification, thereby improving the convenience for the user.
[0084] The information processing method according to the eighteenth aspect of the present disclosure may be such that, in the information processing method according to the seventeenth aspect, the status information is also updated every time the evaluation criterion is changed.
[0085] According to the eighteenth aspect, the status information is updated every time the evaluation criteria are changed, thereby making it possible to notify the user of the latest status information.
[0086] The information processing method involved in the nineteenth aspect of the present disclosure may be: in the information processing method of the seventeenth aspect, the maximum time, the action condition and the maximum number of times are set based on condition setting information input by the user.
[0087] According to the nineteenth aspect, the user can arbitrarily set the maximum time, the operation conditions and the maximum number of times.
[0088] The information processing method involved in the twentieth embodiment of the present disclosure may be: in the information processing method of the seventeenth embodiment, the action condition is pre-set for the device, and the action condition corresponding to the selected device is also set based on the device selection information input by the user.
[0089] According to the twentieth aspect, appropriate operating conditions can be automatically set according to the device selected by the user.
[0090] The information processing method involved in the twenty-first mode of the present disclosure may be: in any one of the first to twentieth modes, for each action performed by the device in the control parameter optimization process, action history information including execution time and action conditions is recorded.
[0091] According to the twenty-first aspect, action history information including execution time and action conditions is recorded for each action executed by the device, thereby making it possible to provide the action history information to the user.
[0092] In the information optimization method involved in the twenty-second mode of the present disclosure, the information processing device uses the first evaluation criteria to perform the first optimization processing of the information, sets the search range of the information based on multiple evaluation values calculated in the first optimization processing, and performs the second optimization processing of the information using the second evaluation criteria for the set search range.
[0093] According to the twenty-second aspect, the setting information of the search range is inherited from the first optimization process to the second optimization process, whereby the second optimization process can be efficiently executed.
[0094] The information processing device involved in the twenty-third mode of the present disclosure is used to optimize the control parameters in a device that performs multiple actions based on multiple control parameters, and includes a selection unit, a control unit, an acquisition unit, a calculation unit and an update unit. In the control parameter optimization process, the selection unit selects a representative action that includes at least one action among all the actions that the device can perform, the control unit instructs the device to perform the representative action selected by the selection unit, the acquisition unit acquires measurement data related to the action of the device measured during the execution of the representative action, the calculation unit calculates an evaluation value of a specified evaluation indicator based on the measurement data acquired by the acquisition unit, and the update unit updates the control parameters based on the evaluation value calculated by the calculation unit, and the control parameter optimization process is performed multiple times while changing the evaluation basis of the control parameter optimization process.
[0095] According to the twenty-third aspect, the control parameter optimization process is executed multiple times while changing the evaluation criteria of the control parameter optimization process. Thus, the control parameter optimization process is repeatedly executed while evaluating the criteria, thereby efficiently generating appropriate control parameters.
[0096] The program involved in the twenty-fourth mode of the present disclosure is used to enable an information processing device to function as a selection unit, a control unit, an acquisition unit, a calculation unit and an update unit. The information processing device is used to optimize the control parameters in a device that performs multiple actions based on multiple control parameters. In the control parameter optimization process, the selection unit selects a representative action that includes at least one action among all the actions that the device can perform, the control unit causes the device to perform the representative action selected by the selection unit, the acquisition unit acquires measurement data related to the action of the device measured during the execution of the representative action, the calculation unit calculates an evaluation value of a specified evaluation indicator based on the measurement data acquired by the acquisition unit, and the update unit updates the control parameters based on the evaluation value calculated by the calculation unit, and the control parameter optimization process is performed multiple times while changing the evaluation criteria of the control parameter optimization process.
[0097] According to the twenty-fourth aspect, the control parameter optimization process is executed multiple times while changing the evaluation criterion of the control parameter optimization process. Thus, by repeatedly executing the control parameter optimization process while changing the evaluation criterion, appropriate control parameters can be efficiently generated.
[0098] Hereinafter, a specific example of a control parameter generation system involved in one embodiment of the present disclosure will be described with reference to the accompanying drawings. The embodiments shown here are all embodiments of a specific example of the present disclosure. Therefore, the numerical values, shapes, constituent elements, configurations and connection methods of constituent elements, and steps (processes) and the order of steps shown in the following embodiments are examples and are not intended to limit the present disclosure. In addition, each figure is a schematic diagram and is not necessarily a strictly represented figure. In each figure, the same symbols are attached to substantially the same structures, and repeated descriptions are omitted or simplified.
[0099] (Implementation Method 1)
[0100] Hereinafter, a control parameter generation system according to Embodiment 1 will be described. The control parameter generation system is a system for generating control parameters for a production device including a drive source for driving a driven object.
[0101] <Structure>
[0102] Figure 1 This is a schematic diagram showing an overview of the control parameter generation system 1 according to the first embodiment.
[0103] Figure 2 It is a block diagram showing the structure of the control parameter generation system 1.
[0104] like Figure 1As shown, the control parameter generating system 1 includes a control parameter generating device 10 , a production device 20 and a sensor 30 .
[0105] The production device 20 is a device used to produce equipment, and performs installation, processing, manufacturing, and transportation of equipment. The production device 20 is, for example, installed in a production line of a factory. Specifically, the production device 20 is, for example, an installation device, a processing device, a manufacturing device, and a transportation device.
[0106] The production device 20 performs N (N is an integer greater than or equal to 3) operations. N is 80, for example.
[0107] like Figure 2 As shown, the production device 20 includes a memory 21 , a control circuit 22 , a driving source 23 , and a driven object 24 .
[0108] The operation of the driving source 23 is controlled by the control circuit 22 , and the driving source 23 drives a driven object 24 .
[0109] Specifically, the driving source 23 is, for example, a servomotor, a directional flow control valve for controlling the fluid of a pneumatic artificial muscle arm, or a directional flow control valve for controlling the fluid of a hydraulic arm. The servomotor may be, for example, a rotary motor or a linear motor.
[0110] The driven object 24 is an object driven by the driving source 23. For example, when the driving source 23 is a servo motor, the driven object 24 is a head for conveying a processing target component, a nozzle mounted on the head for adsorbing the processing target component, etc. In addition, when the driving source 23 is a directional flow control valve, for example, the driven object 24 is a pneumatic artificial muscle arm, a hydraulic arm, etc.
[0111] As an example, Figure 3 1 is a perspective view of a production device 20 having a structure in which a driving source 23 is a servo motor and a driven object 24 is a nozzle provided in a head.
[0112] like Figure 3 As shown, as an example, the production device 20 may be a mounting device that mounts components on a substrate 120 placed on a stage 110 .
[0113] As an example, the production device 20 includes: a nozzle 81, which is a suction component; a head 80, on which the nozzle 81 is installed; a servo motor 23A that functions as a driving source 23 and is used to move the head 80 along the X-axis direction in a plan view of the machine 110; and a servo motor 23B that functions as a driving source 23 and is used to move the head 80 along the Y-axis direction in a plan view of the machine 110. Here, the head 80 is connected to the servo motor 23A via the arm 72 and the servo motor 23B.
[0114] Return again Figure 1 and Figure 2 , continue to explain the control parameter generating system 1.
[0115] The control circuit 22 controls the drive source 23 by outputting a command for setting the position of the drive object 24 to a predetermined target position to the drive source 23. The command output by the control circuit 22 to the drive source 23 may be, for example, a position command indicating the position of the drive source 23 or the drive object 24, or a torque command indicating the torque of the drive source 23.
[0116] The control circuit 22 controls the driving source 23 based on the control parameters stored in the memory 21. In other words, the control circuit 22 uses the control parameters stored in the memory 21 when controlling the driving source 23. There are 50 control parameters, for example.
[0117] The memory 21 stores control parameters used when the control circuit 22 controls the driving source 23. The control parameters stored in the memory 21 are control parameters output from the control parameter generating device 10.
[0118] When the control parameter is output from the control parameter generating device 10 , the memory 21 acquires the output control parameter and updates the stored control parameter with the acquired control parameter, that is, overwrites and stores the control parameter.
[0119] Figure 4 It is a data structure diagram showing an example of the control parameters stored in the memory 21 .
[0120] like Figure 4 As shown, the control parameters stored in the memory 21 include, for example, parameters a1 and a2 for adjusting the vibration frequency of the driven object 24, parameters b1 and b2 for adjusting the speed of the driven object 24, parameters c1 and c2 for adjusting the depth of the singular point in the vibration characteristics of the driven object 24, parameters d1 and d2 for adjusting the vibration amplitude of the driven object 24, etc.
[0121] In general, control parameters include parameters that have a trade-off relationship with each other, such as parameters b1 and b2 for adjusting speed, parameters c1 and c2 for adjusting the depth of a singular point in vibration characteristics, and parameters d1 and d2 for adjusting vibration amplitude.
[0122] Return again Figure 1 and Figure 2 , continue to explain the control parameter generating system 1.
[0123] The sensor 30 measures the position of the driven object 24 in the production device 20 performing at least one of the N operations in time series. Then, the sensor 30 outputs measurement data indicating the measured position corresponding to each of the at least one operation to the control parameter generation device 10 .
[0124] Figure 5 This is a schematic diagram showing an example of changes in the positional deviation of the driven object 24 relative to the target position when the production device 20 drives the driven object 24 to the target position.
[0125] exist Figure 5 In FIG. 1 , the horizontal axis represents time, and the vertical axis represents the position deviation of the driven object 24 relative to the target position.
[0126] like Figure 5 As shown, in this specification, the allowable range refers to a range in which the position deviation from the target position is within the required accuracy.
[0127] In addition, if Figure 5 As shown, in this specification, the time when the allowed position is reached which can be evaluated as having reached the target position (hereinafter also referred to as the "setting time") refers to the time when the driven object 24 finally reaches the allowed range after reaching the allowed range and does not exceed the allowed range again.
[0128] In addition, if Figure 5 As shown, in this specification, the settling time refers to the time from the stop of the driven object 24 based on the instruction for setting the position of the driven object 24 to the target position to the time when the position of the driven object 24 reaches the allowable position that can be evaluated as having reached the target position, that is, the time from the start of the stop to the settling time, or refers to the time from the start of the movement of the driven object 24 based on the instruction for setting the position of the driven object 24 to the target position to the time when the position of the driven object 24 reaches the allowable position that can be evaluated as having reached the target position, that is, the time from the start of the movement to the settling time.
[0129] Figure 6 2 is a data structure diagram showing an example of measurement data output by the sensor 30 .
[0130] like Figure 6 As shown in FIG. 1 , as an example, the measurement data is data that makes a one-to-one correspondence between the elapsed time [ms] after the reference time and the deviation [mm] from the target position. Here, the reference time is the stop start time of the driven object 24 based on the instruction for setting the position of the driven object 24 to the target position, or the movement start time of the driven object 24 based on the instruction for setting the position of the driven object 24 to the target position.
[0131] Return again Figure 1 and Figure 2 , continue to explain the control parameter generating system 1.
[0132] The control parameter generating device 10 generates control parameters for the production device 20 .
[0133] For example, in a computer device including a processor, a memory, and an input / output interface, the processor executes a program stored in the memory, thereby realizing the control parameter generating device 10. Such a computer device is, for example, a personal computer.
[0134] like Figure 2 As shown, the control parameter generating device 10 includes an information processing unit 41 such as a processor, a storage unit 42 such as a memory, an input unit 44 such as a mouse or keyboard, a display unit 17 such as a liquid crystal display or an organic EL (Electroluminescence) display, and a communication unit 43 such as a communication module.
[0135] The information processing unit 41 includes a judgment unit 13, a selection unit 19, a control unit 18, an acquisition unit 11, an output unit 12, an operation accepting unit 15, an image generating unit 16, and a generating unit 14 as functions realized by a processor executing a program read from a storage medium such as a computer-readable ROM (Read Only Memory). That is, the above-mentioned program is a program for causing the information processing unit 41, which is an information processing device mounted on the control parameter generating device 10, to function as the judgment unit 13 (judgment means), the selection unit 19 (selection means), the control unit 18 (control unit), the acquisition unit 11 (acquisition means), the output unit 12 (output means), the operation accepting unit 15 (operation accepting means), the image generating unit 16 (image generating means), and the generating unit 14 (generating means).
[0136] The selection unit 19 selects a representative action including at least one action among all actions executable by the production device 20 .
[0137] The control unit 18 instructs the production device 20 to execute the representative operation or all operations selected by the selection unit 19 .
[0138] The acquisition unit 11 acquires each piece of measurement data output from the sensor 30 and corresponding to one or more operations performed by the production device 20 .
[0139] The determination unit 13 determines, for each of the one or more actions acquired by the acquisition unit 11, a settling time for the action, which is a time from the start of stopping or moving the driven object 24 based on the instruction for setting the position of the driven object 24 to the target position until the driven object reaches a permissible position that can be evaluated as having reached the target position, based on the measurement data corresponding to the action. That is, the determination unit 13 functions as a calculation unit that calculates an evaluation value of an evaluation index (the settling time in the example of the present embodiment) based on the measurement data acquired by the acquisition unit 11.
[0140] The generator 14 updates the control parameter by an optimization process that shortens at least the longest setting time among the one or more setting times corresponding to the one or more actions acquired by the acquirer 11, that is, the one or more setting times determined by the determiner 13, and generates an updated control parameter. That is, the generator 14 functions as an updater that updates the control parameter based on the evaluation value calculated by the determiner 13 as a calculator.
[0141] That is, the generator 14 optimizes at least the longest setting time among the one or more setting times as an evaluation value to be optimized.
[0142] Alternatively, the average value of the settling times or the average value of the longest P settling times (P is a value smaller than L) among the L settling times may be used as the evaluation value.
[0143] like Figure 2 As shown, the generator 14 has an optimization algorithm 140 for optimizing control parameters so as to shorten the settling time. The generator 14 uses the optimization algorithm 140 to perform an optimization process to shorten at least the longest settling time among the one or more settling times determined by the determination unit 13.
[0144] The optimization algorithm 140 may be a known algorithm such as a Bayesian optimization algorithm, an evolutionary strategy algorithm (CMA-ES), or a genetic algorithm (GA). The optimization process for shortening the settling time using the optimization algorithm 140 may be a known process using the known algorithms.
[0145] The generator 14 may generate the updated control parameters by, for example, performing the optimization process once, or may generate the updated control parameters by repeatedly performing the optimization process until the optimization process end condition is satisfied.
[0146] Here, the optimization process end condition is, for example, a period during which the optimization process is repeatedly performed. In this case, the generator 14 repeatedly performs the optimization process during a predetermined period.
[0147] The optimization process end condition is, for example, the number of times the optimization process is repeated. In this case, the generator 14 repeats the optimization process a predetermined number of times.
[0148] The optimization process end condition is, for example, a time that one or more settling times should satisfy. In this case, the generator 14 repeatedly performs the optimization process until one or more settling times reaches or exceed a predetermined time.
[0149] The output unit 12 outputs the control parameter generated by the generating unit 14 to the production device 20 so that the control parameter is stored in the memory 21 .
[0150] The operation accepting unit 15 accepts an input operation of inputting setting information and the like into the control parameter generating device 10 from the input unit 44 by a user who uses the control parameter generating system 1 .
[0151] The display unit 17 displays an image provided to the user who utilizes the control parameter generation system 1. The method of notifying the user of information is not limited to image display, and may be audio output or the like.
[0152] The image generating unit 16 generates an image to be displayed on the display unit 17 .
[0153] <Action>
[0154] The following describes the operation performed by the control parameter generation system 1 having the above configuration.
[0155] The control parameter generation system 1 performs a first control parameter generation process, a condition acquisition process, and a representative action designation process.
[0156] Hereinafter, these processes will be described in sequence with reference to the drawings.
[0157] First, the first control parameter generation process will be described.
[0158] The first control parameter generation process is a process of generating control parameters for the production device 20 .
[0159] For example, when a user who uses the control parameter generation system 1 performs an operation on the operation accepting unit 15 to start the first control parameter generation process, the first control parameter generation process starts.
[0160] Figure 7 is a flowchart of the first control parameter generation process, Figure 8 This is a timing chart of the first control parameter generation process.
[0161] like Figure 7 and Figure 8 As shown, when the first control parameter generation process starts, the operation accepting unit 15 starts a predetermined program for executing the first control parameter generation process (step S5).
[0162] When the prescribed program is started, the output unit 12 outputs the initial value of the control parameter to the memory 21 (step S10). At this time, the output unit 12 may output the initial value of the control parameter including a predetermined value, the initial value of the control parameter including a value specified by the user, or the initial value of the control parameter including a value calculated by a calculation method specified by the user.
[0163] When the control parameter is output from the output unit 12 , the memory 21 stores the control parameter output from the output unit 12 (step S15 ).
[0164] Next, the control parameter generating device 10 substitutes an initial value 1 into the integer variable X (step S20 ).
[0165] In addition, although the initial value substituted into the integer variable X is described here as 1, the initial value may be an integer greater than 1 and less than N-1, and is not necessarily limited to 1. For example, the initial value substituted into the integer variable X may be specified by the user.
[0166] Next, after the selection unit 19 selects a representative action including X actions, the production device 20 performs X actions out of N actions under the control of the control unit 18 (step S25). Here, the X actions selected by the selection unit 19 may be, for example, actions specified by the user, predetermined prescribed actions, or actions actively determined according to a determination method specified by the user. When the user specifies the X actions to be performed by the production device 20, this is achieved, for example, by executing a representative action designation process described later.
[0167] If the production device 20 performs X operations, the sensor 30 outputs X measurement data corresponding to the X operations, respectively, to the control parameter generation device 10 (step S30 ).
[0168] Next, the acquisition unit 11 acquires X pieces of measurement data corresponding to the X actions, respectively, from the sensor 30 (step S35 ).
[0169] When the acquisition unit 11 acquires X pieces of measurement data, the determination unit 13 determines the settling time for each of the X pieces of measurement data (step S40 ).
[0170] If the determination unit 13 determines that the setting time is X, the generation unit 14 checks whether the optimization process end condition is satisfied (step S45).
[0171] When the optimization process end condition is not satisfied in the process of step S45 (step S45: No), the generator 14 performs the optimization process so as to shorten at least the longest setting time among the X setting times (step S50). Preferably, the optimization process of the control parameter is performed so as to make all the X setting times equal to or less than the target value. Next, the generator 14 generates the optimized updated control parameter by the optimization process so as to shorten at least the longest setting time among the X setting times (step S55).
[0172] Then, the output unit 12 outputs the generated updated control parameter to the production device 20 so that the updated control parameter is stored in the memory 21 (step S60).
[0173] Next, the memory 21 acquires the output control parameter, and updates the stored control parameter using the acquired control parameter (step S65).
[0174] When the process of step S65 is completed, the first control parameter generation process proceeds to the process of step S25 again.
[0175] When the optimization process end condition is satisfied in the process of step S45 (step S45 : Yes), the control parameter generating device 10 checks whether the value of the integer variable X is equal to N (step S70 ).
[0176] When the value of the integer variable X is not equal to N during the processing of step S70 (step S70: No), that is, when the value of the integer variable X is less than N, the control parameter generating device 10 substitutes an integer greater than X+1 and less than N into the integer variable X (step S75).
[0177] When the process of step S75 is completed, the first control parameter generation process proceeds to the process of step S25 again.
[0178] When the value of the integer variable X is equal to N in the process of step S70 (step S70: Yes), the first control parameter generation process ends.
[0179] Furthermore, it is also assumed that when the determination unit 13 determines that X setting times are reached in the process of step S40 , the image generation unit 16 generates an image showing the status of the optimization process at the current time point, and the display unit 17 outputs the image generated by the image generation unit 16 .
[0180] Fig. 9 1 is a schematic diagram showing an example of the above-mentioned image displayed on the display unit 17 .
[0181] exist Fig. 9In the figure, the display "5" in the column to the right of "Optimization Stage" indicates that the number of times an integer is substituted into the integer variable X in the first control parameter generation process currently being executed is 5 times, that is, the first control parameter generation process currently being executed is a loop process that repeats the process from step S25 through the process of "Yes" in step S45 and the process of "No" in step S70 and then returns to the process of step S25 four times.
[0182] The display “1” in the column “Number of representative actions” of “First stage” indicates that the integer substituted into the integer variable X for the first time, that is, the integer substituted in the process of step S20 is “1”.
[0183] The display "2" in the "Representative Action Number" column of "Second Stage" indicates that the integer substituted into the integer variable X for the second time, that is, the integer substituted in the process of step S75 executed for the first time in the above loop process is "2".
[0184] The display "8" in the "Representative Action Number" column of "Third Stage" indicates that the integer substituted into the integer variable X for the third time, that is, the integer substituted into the process of step S75 executed for the second time in the above loop process is "8".
[0185] The display "32" in the "Representative Action Number" column of "Fourth Stage" indicates that the integer substituted into the integer variable X for the fourth time, that is, the integer substituted in the process of step S75 executed for the third time in the above loop process is "32".
[0186] The display "80" in the "Representative Action Number" column of "Fifth Stage" indicates the integer substituted into the integer variable X for the fifth time, that is, the integer substituted in the processing of step S75 executed for the fourth time (i.e., last time) in the above-mentioned loop processing is "80".
[0187] The display of "2" in the column of "Maximum Optimization Time (h)" of "First Stage" to "Fifth Stage" indicates that the conditions for ending the optimization processing in each stage include the condition that the optimization processing (i.e., the processing of step S50) is repeated for a specified period of 2 hours, that is, the upper limit of the number of hours for which the optimization processing in each stage is repeated is 2 hours.
[0188] The display "12" in the column of "Settling time for stopping optimization (ms)" of "First stage" indicates that the end condition of the optimization process in the first stage includes the condition that the settling time for which the optimization process is repeatedly performed is 12ms, that is, the upper limit of the optimization process in the first stage being repeated is until the settling time determined in the processing of step S40 reaches 12ms.
[0189] The display "23" in the column of "Settling time for stopping optimization (ms)" of "Second stage" to "Fifth stage" indicates that the end condition of the optimization process of the second stage to the fifth stage includes the condition that the setting time for which the optimization process is repeatedly performed is 23 ms, that is, the upper limit of the optimization process in the second stage to the fifth stage is the condition that the X setting times determined in the process of step S40 are less than 23 ms.
[0190] The display "12" in the column of "Minimum Settling Time" of "First Stage" indicates that the minimum value of the settling time determined in the first stage is 12 ms.
[0191] The display "23" in the column of "Minimum Settling Time" of "Second Stage" indicates that the minimum value of the settling time determined in the second stage is 23 ms.
[0192] The display "45" in the column of "Minimum Settling Time" of "Third Stage" indicates that the minimum value of the settling time determined in the third stage is 45 ms at the current time point.
[0193] The blank display of the "minimum settling time" column of the "fourth stage" and the "fifth stage" indicates that the settling time has not yet been determined in the fourth stage and the fifth stage.
[0194] The broken line graph 210 in the right figure shows the time transition of the settling time determined in the first stage, the broken line graph 220 shows the time transition of the settling time determined in the second stage, and the broken line graph 230 shows the time transition of the settling time determined in the third stage.
[0195] Next, the condition acquisition process will be described.
[0196] The condition acquisition process is a process for acquiring a condition for executing the first control parameter generation process.
[0197] Here, the condition acquisition processing is described as a processing for acquiring the number of times an integer is substituted into the integer variable X in the first control parameter generation processing, the value when the integer is substituted into the integer variable X, and the optimization processing end condition used in the processing of step S65. The optimization processing end condition is a case where the optimization processing is executed within a specified period, or a case where the X settling times determined in the processing of step S40 reach less than a specified time.
[0198] For example, a user who uses the control parameter generating system 1 performs an operation on the operation accepting unit 15 to start the condition acquisition process, whereby the condition acquisition process starts.
[0199] Fig.10 This is a flowchart of the condition acquisition process.
[0200] like Fig.10 As shown, when the condition acquisition process starts, the operation accepting unit 15 starts a predetermined program for executing the condition acquisition process (step S110).
[0201] When the predetermined program is activated, the image generation unit 16 generates an image that prompts the user who uses the control parameter generation system 1 to input conditions for executing the first control parameter generation process. Next, the display unit 17 displays the generated image (step S120).
[0202] When the image is displayed, the operation accepting unit 15 waits until the user completes inputting the conditions for executing the first control parameter generating process (repeating the loop of “step S130 : No” and “step S120 ”).
[0203] Fig.11 It is a schematic diagram showing an example of the above-mentioned drawings displayed on the display unit 17. Fig.11 This is an example of the above-mentioned drawing in a state where the conditions for executing the first control parameter generation process are input by the user.
[0204] exist Fig.11 In FIG. 5 , the display “5” in the column to the right of “Optimization stage” indicates that the number of times an integer is substituted into the integer variable X in the first control parameter generation process input by the user is 5.
[0205] The display "1" in the column "Number of representative actions" of "First stage" indicates that the integer input by the user or set by default for the first time to be substituted into the integer variable X is "1". In addition, here, as described above, the initial value substituted into the integer variable X is set to the default "1", and therefore, it does not need to be input by the user.
[0206] The display of “2” in the column of “Number of Representative Actions” of “Second Stage” indicates that the integer input by the user and substituted into the integer variable X for the second time is “2”.
[0207] The display of “8” in the column of “Number of Representative Actions” of “Third Stage” indicates that the integer input by the user and substituted into the integer variable X for the third time is “8”.
[0208] The display “32” in the column “Number of representative actions” of “Fourth stage” indicates that the fourth integer input by the user and substituted into the integer variable X is “32”.
[0209] The display “80” in the column “Number of representative actions” of “Fifth stage” indicates that the integer input by the user and substituted into the integer variable X for the fifth time is “80”.
[0210] The display "2" in the column "Maximum Optimization Time (h)" of "First Stage" to "Fifth Stage" indicates that the specified period in the optimization condition of each stage input by the user is 2 hours. In other words, it indicates that the user has input the condition that the upper limit of the optimization process in each stage is repeated for 2 hours as the end condition of the optimization process in each stage.
[0211] The display "12" in the column "Settling time to stop optimization (ms)" of "First stage" indicates that the prescribed time in the optimization process termination condition of the first stage input by the user is 12 ms. That is, it indicates that the user has input, as the optimization process termination condition of the first stage, the condition that the upper limit of the optimization process in the first stage is repeated until the settling time determined in the process of step S40 reaches 12 ms.
[0212] The display "23" in the column of "Settling time to stop optimization (ms)" of "Second stage" to "Fifth stage" indicates that the prescribed period in the optimization process termination condition of the second stage to the fifth stage input by the user is 23ms. That is, it indicates that the user has input the condition that the upper limit of the optimization process in the second stage to the fifth stage is repeated until the settling time determined in the process of step S40 reaches 23ms as the optimization process termination condition of the first stage to the fifth stage.
[0213] Return again Fig.10 , continue to explain the condition acquisition processing.
[0214] When the user completes inputting the conditions for executing the first control parameter generating process in the process of step S130 (step S130 : Yes), the operation accepting unit 15 acquires the conditions for executing the first control parameter generating process input by the user (step S140 ).
[0215] When the process of step S140 is completed, the condition acquisition process is completed.
[0216] Next, the representative action designation process will be described.
[0217] The representative action designation process is a process for designating each of the X actions to be executed in each stage when the first control parameter generation process is executed. Hereinafter, the X actions to be executed in each stage are also referred to as "representative actions" in each stage.
[0218] For example, when a user who uses the control parameter generating system 1 performs an operation on the operation accepting unit 15 to start the representative action specifying process, the representative action specifying process starts.
[0219] Fig.12 This is a flowchart representing the action designation process.
[0220] like Fig.12 As shown in FIG. 2 , when the representative action designation process starts, the operation accepting unit 15 starts a predetermined program for executing the representative action designation process (step S210 ).
[0221] When the predetermined program is activated, the image generator 16 generates an image that prompts the user of the control parameter generator 1 to designate X actions to be performed in each stage when the first control parameter generator process is executed. Next, the display unit 17 displays the generated image (step S220).
[0222] If the image is displayed, the operation accepting unit 15 waits until the user specifies that X actions to be performed in each stage when the first control parameter generation process is automatically executed, or the user manually specifies that X actions to be performed in each stage when the first control parameter generation process is executed (repeating the cycle of "step S230: No", "step S220").
[0223] Fig.13 1 is a schematic diagram showing an example of the above-mentioned image displayed on the display unit 17 . Fig.13 The above-mentioned figure shows an example of a state in which the user (1) manually designates X actions to be executed in each stage, and (2) manually designates X actions to be executed in each stage.
[0224] exist Fig.13 In the figure, the black circle in the column on the left side of "Manual representative selection" indicates that the user has manually specified X actions to be performed in each stage. On the other hand, when the user has specified that X actions to be performed in each stage are automatically specified, a black circle is displayed in the column on the left side of "Automatic representative selection".
[0225] The display of white circles in the column of “Representative actions of the first stage” indicates that the corresponding actions (actions identified by the action names on the left side of the table, the same applies hereinafter) are designated by the user as X actions to be executed in the first stage.
[0226] Similarly, the display of white circles in the columns of “representative actions of the second stage” to “representative actions of the fifth stage” indicates that the corresponding actions are designated by the user as X actions to be executed in the second stage to the fifth stage, respectively.
[0227] Return again Fig.12 , continue to explain the condition acquisition processing.
[0228] In the processing of step S230, if the user specifies that X actions to be performed in each stage when the first control parameter generation process is automatically performed, or the user manually specifies that X actions to be performed in each stage when the first control parameter generation process is manually performed (step S230: yes), the operation acceptance unit 15 checks whether the user specifies that X actions to be performed in each stage when the first control parameter generation process is automatically performed (step S240).
[0229] In the process of step S240, when the user does not specify that the X actions to be performed in each stage when the first control parameter generation process is executed are automatically performed (step S240: No), that is, when the X actions to be performed in each stage when the first control parameter generation process is executed are manually specified, the operation accepting unit 15 designates the X actions to be performed in each stage specified by the user as X representative actions in each stage (step S250).
[0230] In the processing of step S240, when the user specifies that X actions to be performed in each stage when the first control parameter generation process is executed automatically (step S240: Yes), the operation acceptance unit 15 determines the X actions to be performed in each stage when the first control parameter generation process is executed based on a prescribed algorithm (step S260).
[0231] For example, the operation accepting unit 15 may determine each predetermined action as X actions to be executed in each stage, or may determine each action satisfying a predetermined condition as X actions to be executed in each stage.
[0232] When the process of step S260 is completed, the operation accepting unit 15 designates the X actions to be executed in each stage determined in the process of step S260 as X representative actions in each stage (step S270 ).
[0233] When the process of step S250 is completed, and when the process of step S270 is completed, the representative action designation process is completed.
[0234] <Research>
[0235] According to the control parameter generating device 10 of the above structure, the appropriate control parameters are not searched for all N (for example, 80) actions performed by the production device 20 at once, but initially, the appropriate control parameters are searched for a part of the N actions (for example, one) as the object, and the number of actions as the object is gradually increased while the appropriate control parameters are searched based on the previous search results. Finally, the appropriate control parameters are searched for all N actions as the object, thereby generating appropriate control parameters.
[0236] Therefore, the control parameter generating device 10 having the above configuration can generate control parameters more efficiently and reliably than searching for appropriate control parameters and generating appropriate control parameters for all N operations performed by the production device 20 at once.
[0237] As described above, according to the control parameter generation device 10 having the above-described configuration, appropriate control parameters can be generated efficiently.
[0238] In addition, the inventors conducted an experiment using a production device simulation model, in which the setting time achieved when control parameters were generated according to a conventional method was compared with the setting time achieved when control parameters for a production device 20 were generated using a control parameter generating device 10. The conventional method generated appropriate control parameters by searching for appropriate control parameters for all N actions performed by the production device 20 at once.
[0239] Fig.14 It is a schematic diagram showing the experimental results.
[0240] like Fig.14 As shown, when the control parameters are generated by the conventional method, the settling time achieved is 36 [ms], and when the control parameters are generated by the control parameter generating device 10 , the settling time achieved is 23 [ms].
[0241] Thus, the inventors have verified through this experiment that, by using the control parameter generating device 10 to generate control parameters, the achieved setting time can be reduced by approximately 35% compared to the conventional method.
[0242] (Implementation Method 2)
[0243] Hereinafter, a control parameter generation system according to a second embodiment which is configured by changing a part of the configuration of the control parameter generation system 1 according to the first embodiment will be described.
[0244] The control parameter generation system 1 according to the first embodiment is an example of a structure in which the optimization process performed by the control parameter generation device 10 is a process for shortening the settling time. In contrast, the control parameter generation system according to the second embodiment is an example of a structure in which the optimization process performed by the control parameter generation device according to the second embodiment is a process for reducing the excess degree described later.
[0245] Here, regarding the control parameter generation system involved in Embodiment 2, since the same components as those of the control parameter generation system 1 have been described, the same reference numerals are attached and their detailed description is omitted, and the description will focus on the differences from the control parameter generation system 1.
[0246] <Structure>
[0247] Fig.15 This is a block diagram showing the structure of a control parameter generating system 1A according to the second embodiment.
[0248] like Fig.15 As shown, the control parameter generation device 10 of the control parameter generation system 1 according to the first embodiment is changed to a control parameter generation device 10A, thereby forming a control parameter generation system 1A. In addition, the determination unit 13 of the control parameter generation device 10 is changed to a determination unit 13A, the generation unit 14 is changed to a generation unit 14A, and the optimization algorithm 140 is changed to an optimization algorithm 140A, thereby forming the control parameter generation device 10A.
[0249] The determination unit 13A determines the degree of excess for each of the one or more actions acquired by the acquisition unit 11 based on the measurement data corresponding to the action, which is the degree to which the object 24 does not converge to the allowable range in the period when the position of the driven object 24 is set to the target position, when a period of time during which the driven object 24 does not converge to the allowable range of deviation from the target position occurs. That is, the evaluation index in the first embodiment is the settling time taken for the position deviation of the position of the driven object 24 and the target position to converge to the allowable range, while the evaluation index in the second embodiment is the cumulative value of the excess time or the position deviation amount during which the position deviation of the position of the driven object 24 and the target position exceeds the allowable range.
[0250] Hereinafter, a specific example of the degree of excess determined by the determination unit 13A will be described with reference to the drawings.
[0251] Fig.16A , Fig. 16B This is a schematic diagram showing an example of changes in the positional deviation of the driven object 24 relative to the target position when the production device 20 drives the driven object 24 to the target position.
[0252] exist Fig.16A , Fig. 16B In FIG. 1 , the horizontal axis represents time, and the vertical axis represents the position deviation of the driven object 24 relative to the target position.
[0253] like Fig.16A , Fig. 16B As shown, when the position of the driven object 24 is set to the target position, there is a case where the driven object 24 exceeds the allowable range after the driven object 24 temporarily reaches the allowable range.
[0254] In this case, for example, Fig.16AAs shown, the determination unit 13A determines the total of the time during which the driven object 24 does not converge to the allowable range after the driven object 24 first reaches the allowable range (hereinafter also referred to as “first total”) as the excess degree.
[0255] In addition, the determination unit 13A may also calculate the first sum by weighting and adding the elapsed time from the stop or movement of the driven object 24. That is, the longer the elapsed time from the stop or movement of the driven object 24 is, the greater the weight (penalty) added by the determination unit 13A may be, and the calculation of the first sum may be performed. Thus, it is possible to suppress large fluctuations of the driven object 24 after a long time from the stop or movement of the driven object 24.
[0256] Or, in this case, for example, Fig. 16B As shown, the determination unit 13A determines the total of integrated values of position deviations during the time when the driven object 24 does not converge to the allowable range after the driven object 24 first reaches the allowable range (hereinafter also referred to as “second total”) as the excess degree.
[0257] Furthermore, when the length of time during which the driven object 24 does not converge to the allowable range is less than a threshold, the determination unit 13A may exclude the driven object 24 from being added to the second sum. This allows for tolerance to sensor noise generated in a short period of time.
[0258] Return again Fig.15 , the control parameter generating system 1A is further described.
[0259] The generating unit 14A updates the control parameters and generates updated control parameters through optimization processing, wherein the optimization processing reduces at least the maximum degree of excess among one or more degrees of excess corresponding to one or more actions acquired by the acquiring unit 11, i.e., one or more degrees of excess determined by the determining unit 13A.
[0260] like Fig.15 As shown, the generator 14A has an optimization algorithm 140A for optimizing the control parameters so as to reduce the overrun time. The generator 14A uses the optimization algorithm 140A to perform an optimization process that reduces at least the largest overrun degree among the one or more overrun degrees determined by the determination unit 13A.
[0261] The optimization algorithm 140A may be a known algorithm such as a Bayesian optimization algorithm, an evolutionary strategy algorithm (CMA-ES), or a genetic algorithm (GA), similar to the optimization algorithm 140 involved in Embodiment 1. In addition, the optimization process for reducing the degree of excess using the optimization algorithm 140A may be a known process performed using the above-mentioned known algorithms.
[0262] Similar to the generator 14 according to the first embodiment, the generator 14A may generate the updated control parameters by performing the optimization process once, or may generate the updated control parameters by repeatedly performing the optimization process until the optimization process end condition is satisfied.
[0263] Here, the optimization process end condition is, for example, a period during which the optimization process is repeatedly performed. In this case, the generator 14A repeatedly performs the optimization process during a predetermined period.
[0264] The optimization process end condition is, for example, the number of times the optimization process is repeated. In this case, the generator 14A repeats the optimization process a predetermined number of times.
[0265] The optimization process end condition is, for example, a degree that one or more excess degrees should satisfy. In this case, the generator 14A repeatedly performs the optimization process until one or more excess degrees reach or fall below a predetermined degree.
[0266] <Action>
[0267] The following describes the operation performed by the control parameter generation system 1A having the above configuration.
[0268] The control parameter generation system 1A executes a second control parameter generation process obtained by modifying a part of the first control parameter generation process, instead of the first control parameter generation process according to the first embodiment.
[0269] The second control parameter generation process is a process for generating control parameters for the production device 20 similarly to the first control parameter generation process.
[0270] Fig.17 This is a flowchart of the second control parameter generation process.
[0271] In the second control parameter generation process, the processing of step S305 to step S335, the processing of step S345 and the processing of step S360 to step S375 are respectively the same as the following processing, which is the processing obtained by reading the control parameter generation device 10, the judgment unit 13 and the generation unit 14 as the control parameter generation device 10A, the judgment unit 13A and the generation unit 14A respectively, and reading the first control parameter generation process as the second control parameter generation process.
[0272] Therefore, the second control parameter generation process will be described here mainly focusing on the process of step S340 , the process of step S350 , and the process of step S355 .
[0273] When the acquisition unit 11 acquires X pieces of measurement data in the process of step S335 , the determination unit 13A determines the degree of excess for each of the X pieces of measurement data (step S340 ). Next, the process proceeds to the process of step S345 .
[0274] When the optimization process end condition is not satisfied in the process of step S345 (step S345: No), the generator 14A performs the optimization process in a manner that at least the largest degree of excess among the X degrees of excess is reduced (step S350). It is preferable to perform the optimization process of the control parameter in a manner that all the X degrees of excess are below the target value. Then, the generator 14A generates the updated control parameter optimized in a manner that at least the largest degree of excess among the X degrees of excess is reduced through the optimization process (step S355). Then, the process proceeds to the process of step S345.
[0275] <Research>
[0276] According to the control parameter generating device 10A of the above structure, similar to the control parameter generating device 10, appropriate control parameters are not searched for all N (for example, 80) actions performed by the production device 20 at once, but initially, appropriate control parameters are searched for a part of the N actions (for example, one) as objects, and the number of actions as objects is gradually increased while the appropriate control parameters are searched based on the previous search results. Finally, appropriate control parameters are searched for all N actions, thereby generating appropriate control parameters.
[0277] Therefore, according to the control parameter generating device 10A of the above structure, like the control parameter generating device 10, it is possible to generate control parameters more efficiently and reliably than searching for appropriate control parameters and generating appropriate control parameters for all N actions performed by the production device 20 at once.
[0278] As described above, according to the control parameter generation device 10A having the above configuration, similarly to the control parameter generation device 10 , it is possible to efficiently generate appropriate control parameters.
[0279] (Implementation method 3)
[0280] Hereinafter, a control parameter generation system according to a third embodiment which is configured by changing a part of the operation of the control parameter generation system 1 according to the first embodiment will be described.
[0281] <Structure>
[0282] The hardware configuration of the control parameter generation system according to the third embodiment is the same as that of the control parameter generation system according to the first embodiment, but a part of the software to be executed is changed.
[0283] Therefore, regarding the control parameter generation system involved in Embodiment 3, since each component has been described, the same reference numerals are attached and the detailed description thereof is omitted, and the description will focus on the differences from the operation of the control parameter generation system 1.
[0284] <Action>
[0285] The control parameter generation system according to the third embodiment executes a third control parameter generation process obtained by modifying a portion of the first control parameter generation process, instead of the first control parameter generation process and the representative action designation process according to the first embodiment.
[0286] The third control parameter generation process is a process for generating control parameters for the production device 20 similarly to the first control parameter generation process.
[0287] Fig.18 This is a flowchart of the third control parameter generation process.
[0288] In the third control parameter generation process, the processing of step S405 to step S465 are respectively the same as the following processing, which is the processing of step S5 to step S65 in the first control parameter generation process, and the first control parameter generation process is read as the third control parameter generation process.
[0289] Therefore, the third control parameter generation process will be described here mainly focusing on the processes of step S525 to step S545 and the processes of step S570 to step S580.
[0290] When the process of step S420 is completed, the production device 20 performs Y operations (step S525). Here, Y is an integer greater than or equal to X+1 and less than or equal to N.
[0291] If the production device 20 performs Y actions, the sensor 30 outputs Y measurement data corresponding to the Y actions, respectively, to the control parameter generating device 10 (step S530 ).
[0292] Next, the acquisition unit 11 acquires Y pieces of measurement data corresponding to the Y actions, respectively, from the sensor 30 (step S535 ).
[0293] When the acquisition unit 11 acquires Y pieces of measurement data, the determination unit 13 determines the settling time for each of the Y pieces of measurement data (step S540 ).
[0294] When the settling time is determined for each of the Y pieces of measurement data, the determination unit 13 selects X representative actions from the Y actions based on a predetermined algorithm (step S545 ).
[0295] Fig.19 1 is a schematic diagram showing an example of a case where the determination unit 13 selects X representative actions from Y actions. Fig.19 The example shown is an example in the case where Y is 40 and X is 2.
[0296] like Fig.19 As shown, the determination unit 13 may also determine the settling time of the Y actions determined in the process of step S540 (see Fig.19 (a)) Sort the determined Y setting times in order from short to long setting times (refer to Fig.19 (b)) and select X actions in the specified order as representative actions (refer to Fig.19 (c)).
[0297] Return again Fig.18 , continue to explain the third control parameter generation process.
[0298] When the process of step S545 is completed, the third control parameter generation process proceeds to the process of step S445.
[0299] When the optimization process termination condition is satisfied in the process of step S445 (step S445: Yes), the control parameter generation device 10 substitutes an integer from X+1 to N into the integer variable X (step S570). Next, the control parameter generation device 10 checks whether the value of the integer variable X is equal to N (step S575).
[0300] When the value of the integer variable X is not equal to N in the process of step S575 (step S575: No), that is, when the value of the integer variable X is smaller than N, the third control parameter generation process proceeds to the process of step S525 again.
[0301] When the value of the integer variable X is equal to N in the process of step S575 (step S575: Yes), the control parameter generation system according to the third embodiment executes parameter update processing using all actions (step S580).
[0302] Fig. 20 This is a flowchart of the parameter update process using all actions.
[0303] like Fig. 20 As shown, when the parameter update process using all actions starts, the production device 20 performs N actions under the control of the control unit 18 (step S581).
[0304] If the production device 20 performs N actions, the sensor 30 outputs N measurement data corresponding to the N actions, respectively, to the control parameter generating device 10 (step S582 ).
[0305] Next, the acquisition unit 11 acquires N measurement data corresponding to the N actions, respectively, from the sensor 30 (step S583 ).
[0306] When the acquisition unit 11 acquires N measurement data, the determination unit 13 determines the settling time for each of the N measurement data (step S584 ).
[0307] If the determination unit 13 determines that N settling times are required, the generation unit 14 checks whether the optimization process end condition is satisfied (step S585 ).
[0308] When the optimization process end condition is not satisfied in the process of step S585 (step S585: No), the generator 14 performs the optimization process so as to shorten at least the longest setting time among the N setting times (step S586). Next, the generator 14 generates the updated control parameters optimized so as to shorten at least the longest setting time among the N setting times through the optimization process (step S587).
[0309] Then, the output unit 12 outputs the generated updated control parameter to the production device 20 so that the updated control parameter is stored in the memory 21 (step S588).
[0310] Next, the memory 21 obtains the output control parameter, and updates the stored control parameter using the obtained control parameter (step S589).
[0311] When the process of step S589 is completed, the parameter update process performed using all actions proceeds to the process of step S581 again.
[0312] When the optimization process end condition is satisfied in the process of step S585 (step S585: Yes), the parameter update process using all actions ends.
[0313] Return again Fig.18 , continue to explain the third control parameter generation process.
[0314] When the process of step S580, that is, the parameter generation process using all actions is completed, the third control parameter generation process is completed.
[0315] In addition, although it is described here that in the process of step S545, the determination unit 13 sorts the determined Y settling times in the order of the settling time from short to long, and selects X actions in the prescribed order as representative actions, it is not necessarily limited to the example in which the determination unit 13 selects X actions as representative actions using this method. For example, the determination unit 13 may divide the Y actions into X action groups including actions similar to each other, and select the action that reaches the most characteristic settling time (for example, the longest settling time) in each action group as the X representative actions.
[0316] <Research>
[0317] The control parameter generating device 10 involved in embodiment 3 for performing the above-mentioned actions is similar to the control parameter generating device 10 involved in embodiment 1. Instead of searching for appropriate control parameters for all N (for example, 80) actions performed by the production device 20 at once, the appropriate control parameters are initially searched for a portion of the N actions (for example, one) and then the number of actions as objects is gradually increased while searching for appropriate control parameters based on the last search result. Finally, appropriate control parameters are searched for all N actions, thereby generating appropriate control parameters.
[0318] Therefore, the control parameter generating device 10 involved in embodiment 3 for performing the above-mentioned actions can generate control parameters more efficiently and reliably than searching for appropriate control parameters for all N actions performed by the production device 20 at once to generate appropriate control parameters, similar to the control parameter generating device 10 involved in embodiment 1.
[0319] As described above, according to the control parameter generation device 10 according to the third embodiment that performs the above-described operation, appropriate control parameters can be efficiently generated, similarly to the control parameter generation device 10 .
[0320] (Implementation 4)
[0321] In Embodiment 3, the parameter update process performed using all N actions is implemented in the final stage of the multi-stage control parameter optimization process as an evaluation action for comprehensively evaluating the actions of the production device 20. However, it may be implemented at least once in the middle stage of the multiple stages instead of in the final stage, or in addition to being implemented in the final stage, it may be implemented at least once in the middle stage of the multiple stages. The evaluation action is performed during the period when the representative action of each stage is executed and during at least one period of transition from a certain stage to the next stage. As in Embodiment 3, the production device 20 is made to execute all actions as evaluation actions, thereby enabling accurate comprehensive evaluation.
[0322] However, the evaluation action is not limited to all actions, but may be a differential action between all actions and the nearest representative action of the evaluation action. The differential action is the remaining action after excluding the action contained in the nearest representative action from all N actions. In the evaluation action, the control unit 18 instructs the production device 2 to perform only the differential action. By instructing the production device 20 to perform the differential action between all actions and the nearest representative action of the evaluation action as the evaluation action, the accuracy of the comprehensive evaluation can be maintained and the efficiency of the evaluation action can be achieved. Alternatively, the evaluation action is not limited to all actions, but may be a predetermined action pre-determined from all actions. For example, a plurality of actions in all actions that are specified in order from large to small in terms of the impact on the action performance of the production device 20 are defined as predetermined actions. In the evaluation action, the control unit 18 instructs the production device 2 to perform only the predetermined action. By instructing the production device 20 to perform the predetermined action pre-determined from all actions as the evaluation action, the efficiency of the evaluation action can be achieved.
[0323] The control unit 18 determines whether the evaluation result of the comprehensive evaluation satisfies the specified optimization termination condition, and terminates the control parameter optimization process when the evaluation result satisfies the optimization termination condition. The target value of the evaluation value related to the evaluation action can be set as the optimization termination condition, the upper limit of the total value of the number of repetitions of the optimization process related to all stages up to the current time point can be set as the optimization termination condition, and the upper limit of the execution time of the optimization process related to all stages up to the current time point can be set as the optimization termination condition. For example, when the target value of the evaluation value is set as the optimization termination condition, the control unit 18 determines that the optimization termination condition is not satisfied when the evaluation value of the most recent evaluation action exceeds the target value, and determines that the optimization termination condition is satisfied when the evaluation value of the most recent evaluation action is below the target value. When the evaluation result of the comprehensive evaluation obtained by executing the evaluation action satisfies the optimization termination condition, the optimization process is terminated even before the representative action reaches the final stage, thereby shortening the time required for the optimization process.
[0324] (Implementation method 5)
[0325] In the above-mentioned embodiments, the number of representative actions (X) is gradually increased as the stages of the control parameter optimization process progress, but the content of the actions may be changed without increasing the number of actions.
[0326] For example, the number of representative actions may be fixed to 4 (X=4) in all the multiple stages, and the selection unit 19 may reselect 4 representative actions each time the optimization process end condition is satisfied and the stage advances.
[0327] An example of a method for selecting X representative actions in each stage by the selection unit 19 is as follows. The selection method described below is not limited to the fifth embodiment in which the number of representative actions is fixed in all stages, but can also be applied to the first to fourth embodiments in which the number of representative actions is changed in each stage.
[0328] As a first example, the selection unit 19 randomly selects X representative actions from N total actions.
[0329] As a second example, the selection unit 19 selects X representative actions according to a preset selection rule. For example, a rule may be set as a selection rule, where the simpler the action, the earlier it is selected, and the more complex the action, the later it is selected. That is, a rule may be set such that the content of the representative action becomes more complex as the stage of the control parameter optimization process progresses.
[0330] As a third example, the selection unit 19 randomly arranges all N actions and selects X actions as representative actions from the front end of the arrangement. When the selection position reaches the end of the arrangement, the selection unit 19 returns to the front end of the same arrangement and repeats the selection.
[0331] As a fourth example, the selection unit 19 randomly arranges all N actions and selects X actions as representative actions from the front end of the arrangement. If the selection position reaches the end of the arrangement, the selection unit 19 randomly arranges all N actions again and selects from the front end.
[0332] As a fifth example, the selection unit 19 classifies all N actions into X groups, and randomly selects representative actions from each group one by one.
[0333] As a sixth example, the selection unit 19 classifies all N actions into X groups including actions similar to each other based on the similarity of the waveform or evaluation value of the measurement data, randomly arranges multiple actions belonging to each group, and selects one action as a representative action in turn from the front end of the arrangement in each group.
[0334] As a seventh example, the selection unit 19 classifies all N actions into X groups based on the similarity, determines the action with the largest evaluation value (e.g., the latest evaluation value) or the largest past average of the evaluation values from the multiple actions belonging to each group as the action to be improved, and selects the X actions to be improved determined for the X groups as representative actions in the next control parameter optimization process. The past average value may be the average of the evaluation values of all periods, or the average of the most recent specified number of evaluation values.
[0335] As an eighth example, the selection unit 19 classifies all N actions into X groups based on the correlation similarity indicating the change tendency of multiple evaluation values, determines the action with the largest evaluation value (e.g., the latest evaluation value) or the largest past average of the evaluation values from the multiple actions belonging to each group as the action to be improved, and selects the X actions to be improved determined for the X groups as representative actions in the next control parameter optimization process. The past average value may be the average value of the evaluation values of all periods or the average value of the most recent specified number of evaluation values.
[0336] In addition, the selection unit 19 may perform the classification of the groups in the fifth to eighth examples multiple times. The selection unit 19 may perform the classification of the groups again each time the evaluation action (step S71) described later is performed, may perform the classification of the groups again at a certain time interval, and may perform the classification of the groups again each time the evaluation result of the representative action or the evaluation action increases above a certain threshold.
[0337] As a ninth example, the selection unit 19 randomly selects X representative actions from all N actions, and randomly selects X representative actions from unselected actions excluding the selected actions the next time and thereafter. If the number of unselected actions is less than X, X representative actions are randomly selected again from all N actions.
[0338] As a tenth example, the selection unit 19 randomly selects X representative actions from all N actions, and randomly selects X-1 representative actions from unselected actions excluding the action with the worst evaluation value among the X representative actions and the selected action next time and thereafter. If the number of unselected actions is less than X, X representative actions are randomly selected again from all N actions.
[0339] Furthermore, in each stage of the multi-stage control parameter optimization process, the X representative actions may be executed only once each, or the X representative actions may be set as one group and executed in multiple groups until the optimization process end condition is satisfied.
[0340] In addition, similarly to the fourth embodiment, the evaluation operation may be performed at least once in the middle of a plurality of stages. Fig.21 1 is a flowchart of the first control parameter generation process. Fig.21 As shown, after each stage of the control parameter optimization process of multiple stages, an evaluation action (step S71) may be performed, and when the evaluation result satisfies the optimization end condition (step S72: yes), the optimization process is terminated; when the evaluation result does not satisfy the optimization end condition (step S72: no), the next X representative actions are selected (step S24), and the control parameter optimization process of the next stage is performed.
[0341] (Implementation method 6)
[0342] In Embodiment 6, the details of the optimization algorithm 140 used by the generator 14 are described. The following algorithm is used as the optimization algorithm 140, which is based on the evolutionary strategy algorithm (CMA-ES), sets the search range of the control parameters of the update candidates in the control parameter optimization process of the next stage based on the multiple evaluation values calculated in the control parameter optimization process of the current stage, and further changes the evaluation criteria of the control parameter optimization process in the current stage and the next stage. That is, the generator 14 uses the optimization algorithm 140 to perform the first optimization process of the information (control parameters in this example) using the first evaluation criteria, sets the search range of the information based on the multiple evaluation values calculated in the first optimization process, and performs the second optimization process of the information using the second evaluation criteria different from the first evaluation criteria for the set search range.
[0343] Here, changing the evaluation criteria includes, for example, changing the representative action as in Implementation Examples 1 and 5. Also, changing the representative action includes, for example, changing the number of actions included in the representative action in each of the multiple stages as in Implementation Example 1. Also, changing the number of actions includes, for example, increasing the number of actions as the stage progresses as in Implementation Example 1. Also, changing the representative action includes, for example, fixing the number of actions included in the representative action in each of the multiple stages and changing the content of the action as in Implementation Example 5. Also, changing the content of the action includes, as in Implementation Example 5, making the content of the action more complicated as the stage progresses.
[0344] In addition, changing the evaluation criteria includes changing the evaluation index in each of the multiple stages. For example, when the settling time is used as the evaluation index in the current stage, the degree of excess (the cumulative value of the excess time or the position deviation amount) is used as the evaluation index in the next stage. Other evaluation indicators different from the settling time and the degree of excess may be used. Alternatively, a weighted sum of the two evaluation indicators of the settling time and the degree of excess may be used, and the weight value multiplied by each evaluation indicator may be changed in each of the multiple stages.
[0345] In the optimization process of information, sometimes, compared with the use of the evaluation indicator as the final improvement object, when other evaluation indicators are used, or when the weighted sum of other multiple evaluation indicators is used, or when the weighted sum of the evaluation indicator as the improvement object and other evaluation indicators is used, the optimization will proceed faster. However, it is not clear which evaluation indicator is effective, or it is not clear what the effective weight value of each evaluation indicator is. In this method, the following structure is adopted, that is, the generation unit 14 can also cope with the situation where the above-mentioned evaluation indicators or weight values are not clear by dynamically changing the evaluation indicators or weight values used. In addition, it is not limited to the weighted sum, and any evaluation operation (for example, logarithmic sum) using multiple evaluation values of multiple evaluation indicators can be used. In addition, a structure that can select the evaluation operation desired by the user from multiple evaluation operations pre-stored in the storage unit 42 can be adopted, and a structure that dynamically changes the evaluation operation by the generation unit 14 in each stage of the optimization process can also be adopted.
[0346] In addition, the generator 14 may also change the evaluation index or weight value used in probability, and repeatedly perform the optimization process of information, and in the storage unit 42 or other storage units, store and accumulate the corresponding relationship between the evaluation index or weight value used and the evaluation value of the evaluation index as the final improvement object. The probability is increased in a way that the evaluation index or weight value that has obtained a good evaluation value is easy to be selected again, thereby promoting the advancement of optimization. In addition, the generator 14 may perform the change of the evaluation index or weight value used according to the frequency of each stage of the optimization process of multiple stages, or may perform the change of the evaluation index or weight value used according to the frequency of once per multiple stages. In addition, the generator 14 may randomly change the evaluation index or weight value used in each stage, may use the same evaluation index or weight value in multiple continuous stages, and may use the same evaluation index or weight value in multiple discontinuous stages.
[0347] Since effective settings of the evaluation index or weight value used vary in each stage of the optimization process in multiple stages, the progress of the optimization can be accelerated by dynamically changing the evaluation index or weight value as described above.
[0348] In addition, when it is desired to improve multiple evaluation indicators at the same time, the generator 14 may also change the multiple evaluation indicators alternately and randomly, or change the evaluation indicators or weight values used according to a predetermined rule. Here, when there is an evaluation indicator that does not reach the target among the multiple evaluation indicators, the generator 14 may also increase the frequency of using the evaluation indicator.
[0349] Furthermore, the generator 14 may also change the evaluation index by switching the sensor used to measure the measurement object during the execution of the optimization process. As an example, the generator 14 may first perform a rough optimization by using a first sensor, which has low measurement accuracy but high noise tolerance and can perform measurement in a short time, and then perform a fine optimization by using a second sensor, which has high measurement accuracy but low noise tolerance and requires multiple measurements, so that the measurement requires a long time.
[0350] Furthermore, the generator 14 may change the evaluation index by switching the content of the preprocessing applied during the execution of the optimization process. For example, it may switch between using the measurement data in the time domain obtained from the measured waveform and the measurement data obtained by converting the measurement data in the time domain to the frequency domain by performing frequency analysis.
[0351] Furthermore, the generator 14 may change the parameters to be adjusted during the execution of the optimization process.
[0352] Fig. 22 It is a flowchart showing the flow of processing executed by the information processing unit 41 .
[0353] First, in step S100, the generator 14 and Figure 7 Similarly, in steps S5 to S20, initial values of the control parameters are set.
[0354] Next, in step S101 , the selection unit 19 selects a representative action to be performed in the first phase from among all N actions.
[0355] Next, in step S102, the control unit 18 and Figure 7 Similarly, in step S25 , the production device 20 is instructed to execute the representative action selected by the selection unit 19 .
[0356] Next, in step S103, the acquisition unit 11 and Figure 7 Similarly, in steps S30 to S35 , measurement data related to the operation of the production device 20 measured by the sensor 30 during the execution of the representative operation is obtained.
[0357] Next, in step S104, the determination unit 13 and Figure 7 Similarly, in step S40 , the evaluation value of the evaluation index is calculated based on the measurement data acquired by the acquisition unit 11 .
[0358] Next, in step S105 , the generator 14 determines whether the optimization process end condition is satisfied.
[0359] If the optimization process end condition is not satisfied (step S105: No), then in step S106, the generator 14 and Figure 7 Similarly, in steps S50 to S65, the control parameter is updated based on the evaluation value calculated by the determination unit 13, and the processing after step S102 is repeated.
[0360] When the optimization process end condition is satisfied (step S105: Yes), then in step S107, the control unit 18 and Fig.21 Similarly, in step S71, the production device 20 is instructed to perform an evaluation operation.
[0361] Next, in step S108, the generator 14 and Fig.21 Similarly, in step S72, it is determined whether the evaluation result satisfies the optimization end condition.
[0362] When the optimization end condition is satisfied (step S108 : Yes), the optimization process of the control parameters is ended.
[0363] If the optimization end condition is not met (step S108: No), then in step S109, the generation unit 14 sets the search range of the update candidate control parameters in the next stage of the control parameter optimization process based on the multiple evaluation values calculated in the current stage of the control parameter optimization process using the optimization algorithm 140.
[0364] Next, in step S110, the generator 14 generates Figure 7 Similarly, in step S75, the evaluation criteria are changed (in Figure 7 In the example of , the number of representative actions). In the control parameter optimization process at the next stage, the control parameter optimization process is executed using the search range set in step S109 and the evaluation criterion changed in step S110.
[0365] In addition, in the sixth embodiment, the image generating unit 16 may generate the same Fig. 9 The same image (i.e., an image showing the status of the optimization process) is displayed by the display unit 17. In this case, the total count of the applied evaluation criteria may be displayed instead of Fig. 9 The “optimization stage” of the image shown. In addition, the content of the evaluation criteria applied in each stage may be displayed instead of the “representative action number”. In addition, the maximum execution time of the optimization process using the evaluation criteria applied in each stage may be displayed instead of the “maximum optimization time”. In addition, the target performance (for example, target evaluation value) of the optimization process using the evaluation criteria applied in each stage may be displayed instead of the “setting time to stop optimization”. In addition, the best evaluation value of the optimization process using the evaluation criteria applied in each stage may be displayed instead of the “minimum setting time”.
[0366] Fig.23 Schematically shows a method of setting a search range by the generator 14. Here, for simplicity, only two control parameters P1 and P2 are focused on. The horizontal axis of the coordinate represents the value of the control parameter P1, and the vertical axis represents the value of the control parameter P2.
[0367] In the first stage, the control parameter optimization process is performed using the evaluation reference V1, and a plurality of (six in this example) evaluation values are plotted on the coordinates. Each of the plotted evaluation values is the worst value among the X evaluation values obtained by the X representative actions. The generator 14 determines the upper evaluation values from the plotted evaluation values. Fig.23In the figure, multiple evaluation values of the upper level are depicted with white circles, and other evaluation values are depicted with black circles. The generation unit 14 uses the optimization algorithm 140 of CMA-ES to calculate the center coordinates and covariance matrix of multiple evaluation values of the upper level, thereby setting the search range W1 of the next stage. When multiple evaluation values of the upper level are dispersed in a wide area, a wide search range W1 is set, and when multiple evaluation values of the upper level are concentrated in a narrow area, a narrow search range W1 is set.
[0368] In the second stage, the control parameter optimization process is performed for the search range W1 using an evaluation criterion V2 different from the evaluation criterion V1, and a plurality of (six in this example) evaluation values are plotted on the coordinates. The generation unit 14 determines a plurality of upper evaluation values from the plurality of plotted evaluation values in the same manner as described above, and calculates the center coordinates and covariance matrix of the plurality of upper evaluation values using the optimization algorithm 140 of CMA-ES, thereby setting the search range W2 of the next stage.
[0369] In the third stage, the control parameter optimization process is performed for the search range W2 using an evaluation benchmark V3 different from the evaluation benchmark V2, and multiple (six in this example) evaluation values are plotted on the coordinates. The generation unit 14 determines multiple evaluation values of the upper level from the multiple evaluation values plotted, and uses the optimization algorithm 140 of CMA-ES to calculate the center coordinates and covariance matrix of the multiple evaluation values of the upper level, thereby setting the search range W3 of the next stage. Hereinafter, the same process as above is also repeatedly performed for the fourth stage and beyond, and the values of the control parameters P1 and P2 that finally give the best evaluation value are adopted.
[0370] According to Embodiment 6, the control parameter optimization process is executed multiple times while changing the evaluation criterion of the control parameter optimization process. Thus, by changing the evaluation criterion and repeatedly executing the control parameter optimization process, appropriate control parameters can be efficiently generated.
[0371] In addition, the setting information of the control parameter search range W is inherited from this (current stage) control parameter optimization processing to the next (next stage) control parameter optimization processing, thereby enabling efficient search for appropriate control parameters in the next control parameter optimization processing.
[0372] Furthermore, in Embodiment 6, a simple implementation example of CMA-ES is described as an example of an evolutionary strategy, but a more complex implementation or other evolutionary strategy methods may also be used.
[0373] (Implementation 7)
[0374] In the seventh embodiment, a method of notifying the user of status information indicating the progress of the control parameter optimization process and the operating conditions will be described. Each operation performed by the production device 20 is an operation in which the production device 20 moves the driven object 24 to a target position.
[0375] Fig.24 The figure is a simplified diagram showing a part of the action condition information 400 describing the execution conditions of each action executed by the production device 20. The action condition information 400 is stored in the storage unit 42. The action condition information 400 includes an ID (Identifier) as identification information related to each action executed by the production device 20. In addition, the action condition information 400 includes user setting items related to the moving speed, acceleration and deceleration time (acceleration time and deceleration time), number of repetitions, waiting time, and stop position corresponding to each ID. In addition, the action condition is not limited to Fig.24 The example shown only needs to include at least one of ID, movement amount, movement speed, acceleration time, and deceleration time. Fig.24 Other items not shown in the figure. The user operates the input unit 44 to input the condition setting information, thereby, the action condition of each action can be arbitrarily set. In addition, the user operates the input unit 44 to input the condition setting information, thereby, the maximum time for continuing the control parameter optimization process and the maximum number of times the control parameter is updated can be arbitrarily set. In addition, the action condition information 400 includes default setting items related to the moving speed and acceleration and deceleration time corresponding to each ID. In the case where the user does not input the condition setting information from the input unit 44, the moving speed and acceleration and deceleration time use the respective default setting values. In addition, in the case where different action conditions are set by default for the device as the adjustment object, the following structure can also be adopted, that is, the user inputs the device selection information for selecting the device as the adjustment object from the input unit 44, thereby using the default setting value corresponding to the selected device.
[0376] Fig.25 1 is a flowchart showing the flow of processing executed by the information processing unit 41 in connection with notifying the user of the status information.
[0377] First, in step S200 , the acquisition unit 11 acquires the operating condition information 400 from the storage unit 42 .
[0378] Next, in step S201, the image generation unit 16 obtains status information based on the action condition information 400 and the progress information of the control parameter optimization process obtained by the acquisition unit 11. The image generation unit 16 obtains the progress information from the control unit 18. In addition, the control unit 18 forms a database of action history information including execution time and action conditions for each action performed by the production device 20 in the control parameter optimization process and records it in the storage unit 42. The output unit 12 can display the action history information read from the storage unit 42 on the display unit 17 or output it to the outside.
[0379] The action condition information 400 includes information about the maximum time for the control parameter optimization process to continue and the maximum number of times the control parameters are updated, which are set according to the condition setting information. The progress information includes information indicating the stage currently being executed among the multiple stages of the control parameter optimization process and information indicating the action currently being adjusted among the multiple representative actions. In addition, the progress information includes information about the elapsed time from the start of the control parameter optimization process of the first stage to the present (the elapsed time after the start of adjustment) and the number of times the control parameters are updated (the number of parameter updates). The image generation unit 16 calculates the remaining time (adjustment remaining time) to the maximum time by subtracting the elapsed time after the start of adjustment to the present, and calculates the remaining number of times (update remaining number) to the maximum number by subtracting the number of parameter updates to the present from the maximum number.
[0380] Next, in step S202 , the image generation unit 16 generates a situation notification image for notifying the user of the situation information.
[0381] Fig.26 The figure is an example of a status notification image. The status notification image includes an item 300 indicating the elapsed time (the elapsed time after the adjustment started) and the remaining time (the remaining adjustment time), an item 301 indicating the number of updates (the number of parameter updates) and the remaining number (the number of remaining updates), an item 302 indicating the current number of adjustment target actions (that is, the current number of representative actions), and an item 303 indicating the action condition related to the action currently being adjusted in the representative action. In addition, the status notification image is not limited to Fig.26 The example shown, Fig.26 Part of the information shown may be omitted, and may also include Fig.26 Other information not shown in the table. In addition, a progress bar can be used to display the information so that the user can intuitively understand the time elapsed after the start of adjustment and the remaining adjustment time, and the ratio of the number of parameter updates and the number of remaining updates.
[0382] Next, in step S203, the image generation unit 16 inputs the image data of the generated status notification image to the display unit 17. The display unit 17 displays the status notification image based on the input image data. Fig.26 The status notification image shown. In addition, the method of notifying the user of the status information is not limited to the display of the image, and it may also be output of sound from a speaker or the like.
[0383] According to the seventh embodiment, status information is output to notify the user, so that the user can easily confirm the progress status or operation conditions of the control parameter optimization process based on the notification, thereby improving the convenience for the user.
[0384] (Replenish)
[0385] As described above, examples of the technology disclosed in this application are described based on embodiments 1 to 7. However, the present disclosure is not limited to these embodiments 1 to 7. As long as it does not depart from the purpose of the present disclosure, various modifications of this embodiment or this modification that are thought of by those skilled in the art, or a method of combining constituent elements in different embodiments or modifications to construct the same may also be included in the scope of one or more embodiments of the present disclosure.
[0386] (1) In Embodiments 1 to 7, the control parameter generating devices 10 and 10A are implemented by a single computer device as an example. However, the control parameter generating devices 10 and 10A are not necessarily limited to the example of being implemented by a single computer device as long as they can achieve the same function. The control parameter generating devices 10 and 10A may be implemented by, for example, a plurality of computer devices that can communicate with each other.
[0387] (2) The general or specific aspects of the present disclosure may also be implemented by a system, device, method, integrated circuit, program, or non-temporary storage medium such as a computer-readable CD-ROM (Compact Disc Read Only Memory). In addition, the present disclosure may also be implemented by any combination of a system, device, method, integrated circuit, program, and non-temporary storage medium. For example, the present disclosure may also be implemented as a program for causing a computer device to execute the processing performed by the control parameter generation device.
[0388] Industrial Applicability
[0389] The present disclosure can be widely applied to systems for generating control parameters, etc.
Claims
1. An information processing method for optimizing control parameters in a device for performing multiple actions based on multiple control parameters, characterized in that: Let the information processing device, In the control parameter optimization process, a representative action including at least one action among all actions that the device can perform is selected, the device is caused to perform the selected representative action, measurement data related to the action of the device measured during the execution of the representative action is obtained, an evaluation value of a prescribed evaluation index is calculated based on the obtained measurement data, the control parameter is updated based on the calculated evaluation value, and, The control parameter optimization process is executed multiple times while changing an evaluation criterion of the control parameter optimization process.
2. The information processing method according to claim 1, characterized in that: Changing the evaluation criterion includes changing the representative action.
3. The information processing method according to claim 2, characterized in that: Changing the representative action includes changing the number of actions included in the representative action.
4. The information processing method according to claim 3, characterized in that: Changing the number of the actions includes increasing the number of the actions.
5. The information processing method according to claim 2, characterized in that: Changing the representative action includes changing the content of the action included in the representative action.
6. The information processing method according to claim 5, characterized in that: Changing the content of the action includes complicating the content of the action.
7. The information processing method according to claim 2, characterized in that: In the change of the representative action, all the actions are classified into multiple groups based on similarity, and the actions that need to be improved are determined from the multiple actions contained in each of the multiple groups based on the evaluation value or the past average value of the evaluation value, and the multiple actions that need to be improved determined for the multiple groups are selected as the representative actions in the next control parameter optimization processing.
8. The information processing method according to claim 1, characterized in that: Changing the evaluation criteria includes changing the evaluation index.
9. The information processing method according to claim 8, characterized in that: Each of the plurality of actions is an action in which the device moves the object to a target position, The evaluation index includes a settling time, which is a time taken for a positional deviation between the position of the object and the target position to converge within an allowable range.
10. The information processing method according to claim 8, characterized in that: Each of the plurality of actions is an action in which the device moves the object to a target position, The evaluation index includes an excess time or a cumulative value of a position deviation amount during which a position deviation between the position of the object and the target position exceeds an allowable range.
11. The information processing method according to claim 1, characterized in that: Furthermore, based on the plurality of evaluation values calculated in the current control parameter optimization process, a search range for the control parameters of update candidates in the next control parameter optimization process is set.
12. The information processing method according to claim 1, characterized in that: Also in the comprehensive evaluation process, the device is instructed to execute a prescribed evaluation action for comprehensively evaluating the action of the device, measurement data related to the action of the device measured during the execution of the evaluation action is obtained, an evaluation value of the evaluation indicator is calculated based on the obtained measurement data, and the comprehensive evaluation is performed based on the calculated evaluation value.
13. The information processing method according to claim 12, characterized in that: The evaluation action is the total action.
14. The information processing method according to claim 12, characterized in that: The evaluation action is a difference action between the entire action and the representative action closest to the evaluation action.
15. The information processing method according to claim 12, characterized in that: The evaluation action is a predetermined action that is predetermined from among all the actions.
16. The information processing method according to claim 12, characterized in that: It is also determined whether the evaluation result of the comprehensive evaluation satisfies a prescribed end condition, and if the evaluation result satisfies the end condition, the control parameter optimization process is ended.
17. The information processing method according to claim 1, characterized in that: Each of the plurality of actions is an action in which the device moves the object to a target position, The status information is also output to notify the user, the status information being related to at least one of the following information: the elapsed time from the start of the control parameter optimization process, the remaining time until the maximum time for continuing the control parameter optimization process, an action condition including identification information related to a currently executed action among the at least one action included in the representative action, and at least one of a movement amount, a movement speed, an acceleration time, and a deceleration time, The number of actions contained in the representative action, The number of updates of the control parameters, and The number of times remaining until the maximum number of times the control parameter is updated.
18. The information processing method according to claim 17, characterized in that: Whenever the evaluation criteria are changed, the status information is also updated.
19. The information processing method according to claim 17, characterized in that: The maximum time, the operation condition, and the maximum number of times are also set based on condition setting information input by the user.
20. The information processing method according to claim 17, characterized in that: The operating conditions are preset for the device, and Furthermore, based on device selection information input by a user, the operating condition corresponding to the selected device is set.
21. The information processing method according to claim 1, characterized in that: Furthermore, for each action performed by the device in the control parameter optimization process, action history information including execution time and action conditions is recorded.
22. An information optimization method, characterized in that: Let the information processing device, performing a first optimization process of the information using the first evaluation criterion, setting a search range for the information based on the plurality of evaluation values calculated in the first optimization process, A second optimization process of the information is performed using a second evaluation criterion for the set search range.
23. An information processing device for optimizing control parameters in a device for performing multiple actions based on multiple control parameters, characterized in that: including a selection unit, a control unit, an acquisition unit, a calculation unit and an update unit, In the control parameter optimization process, The selection unit selects a representative action including at least one action among all actions executable by the device, The control unit instructs the device to execute the representative action selected by the selection unit, The acquisition unit acquires measurement data related to the operation of the device measured during the execution of the representative operation. The calculation unit calculates an evaluation value of a predetermined evaluation index based on the measurement data acquired by the acquisition unit, The updating unit updates the control parameter based on the evaluation value calculated by the calculating unit, The control parameter optimization process is executed multiple times while changing an evaluation criterion of the control parameter optimization process.
24. A program for causing an information processing device to function as a selection unit, a control unit, an acquisition unit, a calculation unit, and an update unit, the information processing device being used to optimize the control parameters in a device that performs a plurality of actions based on a plurality of control parameters, the program being characterized in that: In the control parameter optimization process, The selection unit selects a representative action including at least one action among all actions that can be performed by the device, The control unit causes the device to execute the representative action selected by the selection unit, The acquisition unit acquires measurement data related to the operation of the device measured during the execution of the representative operation. The calculation unit calculates an evaluation value of a predetermined evaluation index based on the measurement data acquired by the acquisition unit, The updating unit updates the control parameter based on the evaluation value calculated by the calculating unit, The control parameter optimization process is executed multiple times while changing an evaluation criterion of the control parameter optimization process.
