Information processing method, information processing device, and program

Abstract

Provided is an information processing method for setting a parameter related to a setting target, wherein an information processing device: uses a first model of the setting target to execute a first search targeting a first search range; sets a second search range on the basis of the result of the first search; uses a second model of the setting target to execute a second search targeting the second search range, wherein the first model is a surrogate model having a higher speed and a lower accuracy than the second model; and sets the parameter on the basis of the result of the second search.

Description

Information Processing Method, Information Processing Apparatus, and Program

[0001] The present disclosure relates to an information processing method, an information processing apparatus, and a program.

[0002] In a DC-DC converter used in an electric vehicle or the like, a plurality of feedback controls are combined to control charging and discharging according to the driving situation, and stable and highly responsive current and voltage controls are required. In production equipment such as a component mounter or an assembly robot used in a factory or the like, a plurality of servo control motors are combined, and complex and sophisticated operation control is required. In such an apparatus, the operation of the control target is controlled according to a large number of control parameters. In order to obtain desired performance, the user empirically adjusts the control parameters.

[0003] Such a process of adjusting control parameters becomes more difficult as the operating conditions become diverse and complex and the number of parameters increases. The number of trials in the adjustment process also increases, imposing a great deal of labor and time. Therefore, there is an increasing demand to automate these adjustment processes. In addition, in cases such as mass production of a variety of products in small quantities or in the development of derivative models where the operating conditions or required specifications differ slightly, the adjustment process for coping with new requirements or performance improvements while making use of past design assets is becoming more complex.

[0004] On the other hand, a technique of surrogate modeling the relationship between the parameters and responses of the equipment to be controlled by regression and optimizing the parameters using the surrogate model is disclosed in, for example, Patent Document 1.

[0005] In addition, a technique of determining a search range of parameters based on similar parameters and predicted values of responses using a knowledge database storing the results adjusted in the past is disclosed in, for example, Patent Document 2.

[0006] However, when the number of control parameters increases from tens to hundreds, and the number of operating conditions that must satisfy the performance also increases to tens, the number of possible combinations becomes enormous. Therefore, it can become difficult to efficiently set appropriate search priorities or search ranges. As a result, it can become difficult to efficiently set appropriate parameter values, i.e., to efficiently adjust the parameters. In particular, when there are multiple control loops, and the number of controllers increases due to a multi-loop structure or multi-mode configuration, the combinations of control parameters become even more complex, increasing the possibility that multiple parameter combinations exist that yield the desired performance. As a result, it can become even more difficult to efficiently set appropriate parameters.

[0007] Patent No. 7439289 Patent No. 6650786

[0008] This disclosure aims to provide an information processing method, an information processing device, and a program that enable the efficient setting of appropriate parameters.

[0009] An information processing method according to one aspect of the present disclosure is an information processing method for setting parameters relating to a target to be set, wherein the information processing device performs a first search using a first model of the target to be set, with respect to a first search range, sets a second search range based on the result of the first search, performs a second search using a second model of the target to be set, the first model being a surrogate model that is faster and less accurate than the second model, and sets the parameters based on the result of the second search.

[0010] An information processing device according to one aspect of the present disclosure is an information processing device comprising a circuit configuration, the circuit configuration performs a first search using a first model of the target to be set, targets a first search range, sets a second search range based on the result of the first search, performs a second search using a second model of the target to be set, the first model is a surrogate model that is faster and less accurate than the second model, and sets the parameters based on the result of the second search.

[0011] A program according to one aspect of the present disclosure is a program for causing an information processing device to perform processing, wherein the processing includes: performing a first search using a first model of the target to be set, targeting a first search range; setting a second search range based on the result of the first search; performing a second search using a second model of the target to be set, the first model being a surrogate model that is faster and less accurate than the second model; and setting the parameters based on the result of the second search.

[0012] These comprehensive or specific embodiments may be implemented by systems, devices, methods, integrated circuits, computer programs, or recording media, or by any combination of systems, devices, methods, integrated circuits, computer programs, or recording media.

[0013] Further advantages and effects of one aspect of this disclosure will be made apparent from the specification and drawings. Such advantages and / or effects are provided by several embodiments and features described in the specification and drawings, but not all of them are necessarily provided in order to obtain one or more identical features.

[0014] This is a block diagram showing an example of a parameter adjustment system according to the embodiment of this disclosure. This is a simplified diagram showing an example of the hardware configuration of the parameter adjustment system. This is a flowchart showing an example of the operation of the parameter adjustment system according to the embodiment of this disclosure. This is a diagram showing an example of control parameters. This is a diagram showing an example of the search result of the first search. This is a diagram showing an example of the operation screen of the UI unit when a parameter search is performed. This is a diagram showing an example of the operation screen of the UI unit when setting the parameter search range. This is a schematic diagram showing an example of the display area of ​​one parameter in the parameter display and operation area.

[0015] (Summary of the Disclosure) An information processing method according to a first aspect of the Disclosure is an information processing method for setting parameters relating to a target to be set, wherein the information processing device performs a first search using a first model of the target to be set, with respect to a first search range, sets a second search range based on the result of the first search, performs a second search using a second model of the target to be set, the first model being a surrogate model that is faster and less accurate than the second model, and sets the parameters based on the result of the second search.

[0016] According to the first embodiment, appropriate parameters can be efficiently set by the first and second searches.

[0017] In the information processing method according to a second aspect of this disclosure, in the first aspect, the second search range is preferably narrower than the first search range.

[0018] According to the second embodiment, appropriate parameters can be efficiently set by a second search targeting a second search range that is narrower than the first search range.

[0019] In the third aspect of the information processing method of this disclosure, in the first aspect, it is preferable to further prune the second search range using the first model before performing the second search.

[0020] According to the third embodiment, pruning allows for further optimization of parameters and further efficiency in setting parameters.

[0021] In the fourth aspect of the information processing method of this disclosure, in the third aspect, the evaluation conditions in the pruning are preferably stricter than the evaluation conditions in the first search.

[0022] According to the fourth embodiment, parameters can be further optimized by setting evaluation conditions, and parameter setting can be made even more efficient.

[0023] In the information processing method according to the fifth aspect of this disclosure, in any one of the first to fourth aspects, when setting the parameters, it is preferable to perform an applicability evaluation to the actual machine to be set based on the result of the second search, and set the parameters based on the result of the applicability evaluation.

[0024] According to the fifth aspect, the parameters can be further optimized and the parameter setting can be made more efficient through applicability evaluation.

[0025] In the sixth aspect of the information processing method of this disclosure, in the fifth aspect, it is preferable to further reset the execution conditions of the second search, which include at least one of the second search range and the initial search value of the second search, based on the result of the applicability evaluation, and to re-execute the second search based on the reset execution conditions.

[0026] According to the sixth embodiment, the parameters can be further optimized and the parameter setting can be made more efficient by re-executing the second search.

[0027] In the information processing method according to the seventh aspect of this disclosure, in the sixth aspect, when resetting the execution conditions, the reset execution conditions are obtained from an input device having an operation screen, the operation screen displays a bar-shaped first figure indicating the second search range and a point-shaped second figure indicating the initial search value, the second search range is reset by pinch-in, pinch-out, slide, or drag operation on the first figure, and the initial search value is reset by drag operation on the second figure.

[0028] According to the seventh aspect, a highly user-friendly interface allows for further optimization of parameters and more efficient parameter setting.

[0029] In the information processing method according to the eighth aspect of this disclosure, in any one of the first to seventh aspects, in the execution of the first search, it is preferable to perform a multi-stage search that is slower and more accurate in the later stages, and in the execution of the second search, it is preferable to perform a multi-stage search that is slower and more accurate in the later stages.

[0030] According to the eighth aspect, parameters can be further optimized and parameter setting can be made more efficient through multi-stage exploration.

[0031] An information processing method according to a ninth aspect of the present disclosure is an information processing method for setting parameters relating to a target to be set, wherein an information processing device sets execution conditions for a search including at least one of a search range and an initial search value, executes the search using a model of the target to be set based on the set execution conditions, sets the parameters based on the results of the search, and in setting the execution conditions, the execution conditions are obtained from an input device having an operation screen, the operation screen displays a bar-shaped first figure indicating the search range and a dot-shaped second figure indicating the initial search value, the search range is set by pinch-in, pinch-out, slide, or drag operation on the first figure, and the initial search value is set by drag operation on the second figure.

[0032] According to the ninth aspect, a highly user-friendly interface allows for parameter optimization and streamlined parameter setting.

[0033] In the ninth embodiment, the information processing method according to the tenth aspect of the present disclosure is such that the first figure has a first endpoint corresponding to the lower limit of the search range and a second endpoint corresponding to the upper limit of the search range, the search range is reduced by a pinch-in operation on the first and second endpoints, the search range is expanded by a pinch-out operation on the first and second endpoints, the search range is moved by a slide operation on the first and second endpoints, the lower limit is changed by a drag operation on the first endpoint, the upper limit is changed by a drag operation on the second endpoint, and the initial search value is changed by a drag operation on the second figure.

[0034] According to the tenth embodiment, a highly user-friendly interface allows for parameter optimization and streamlining of parameter setting.

[0035] An information processing device according to an eleventh aspect of the present disclosure is an information processing device comprising a circuit configuration for setting parameters relating to a target to be set, wherein the circuit configuration performs a first search using a first model of the target to be set, with respect to a first search range, sets a second search range based on the result of the first search, performs a second search using a second model of the target to be set, the first model being a surrogate model that is faster and less accurate than the second model, and sets the parameters based on the result of the second search.

[0036] According to the eleventh embodiment, appropriate parameters can be efficiently set by the first and second searches.

[0037] An information processing device according to a twelfth aspect of the present disclosure is an information processing device comprising a circuit configuration for setting parameters relating to a target to be set, wherein the circuit configuration sets execution conditions for a search including at least one of a search range and an initial search value, executes the search using a model of the target to be set based on the set execution conditions, sets the parameters based on the results of the search, and in setting the execution conditions, obtains the execution conditions from an input device having an operation screen, the operation screen displays a bar-shaped first figure indicating the search range and a dot-shaped second figure indicating the initial search value, the search range is set by pinch-in, pinch-out, slide, or drag operations on the first figure, and the initial search value is set by drag operations on the second figure.

[0038] According to the twelfth embodiment, a highly user-friendly interface allows for parameter optimization and streamlining of parameter setting.

[0039] A program according to a thirteenth aspect of this disclosure is a program for causing an information processing device for setting parameters relating to a target to be set to execute a process, wherein the process includes: executing a first search using a first model of the target to be set, targeting a first search range; setting a second search range based on the result of the first search; executing a second search using a second model of the target to be set, targeting the second search range, wherein the first model is a surrogate model that is faster and less accurate than the second model; and setting the parameters based on the result of the second search.

[0040] According to the 13th embodiment, appropriate parameters can be efficiently set by the first and second searches.

[0041] A program according to a fourteenth aspect of this disclosure is a program for causing an information processing device for setting parameters relating to a target to be set to execute a process, wherein the process sets execution conditions for a search including at least one of a search range and an initial search value, executes the search using a model of the target to be set based on the set execution conditions, sets the parameters based on the results of the search, and in setting the execution conditions, the execution conditions are obtained from an input device having an operation screen, the operation screen displays a bar-shaped first figure indicating the search range and a dot-shaped second figure indicating the initial search value, the search range is set by pinch-in, pinch-out, slide, or drag operation on the first figure, and the initial search value is set by drag operation on the second figure.

[0042] According to the 14th embodiment, a highly operable user interface allows for parameter optimization and streamlining of parameter setting.

[0043] This disclosure can also be implemented as a program that causes a computer to execute each characteristic configuration included in such a method or apparatus, or as a system that operates using such a program. It goes without saying that such a computer program can be distributed via a computer-readable, non-temporary recording medium such as a CD-ROM, or via a communication network such as the Internet.

[0044] (Embodiments of the Present Disclosure) Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Elements denoted by the same reference numerals in different drawings indicate the same or corresponding elements. Each embodiment can be arbitrarily combined and applied. In addition, the components, the arrangement positions of the components, the connection forms, the order of operations, etc. shown in the following embodiments are examples and are not intended to limit the present disclosure. The present disclosure is limited only by the claims. Therefore, among the components in the following embodiments, the components not described in the independent claims indicating the highest concept of the present disclosure are not necessarily required to achieve the problems of the present disclosure, but are described as constituting a more preferable form.

[0045] Note that the accompanying drawings and the following description are provided for those skilled in the art to fully understand the present disclosure, and it is not intended to limit the subject matter described in the claims thereby. For example, each figure is a schematic diagram and is not necessarily drawn precisely. Therefore, for example, the scales etc. in each figure do not necessarily match.

[0046] <Configuration of the Parameter Adjustment System> Figure 1 is a block diagram showing an example of a parameter adjustment system 100 according to an embodiment of the present disclosure. As shown in Figure 1, the parameter adjustment system 100 is a system for setting parameters related to the actual machine of the adjustment target 109, and comprises a parameter adjustment unit 103, a user interface unit (hereinafter referred to as the "UI unit") 101, a database unit (hereinafter referred to as the "DB unit") 102, a simulation model 110, a proxy model generation unit 111, and a proxy model 112. The parameter adjustment unit 103, simulation model 110, proxy model generation unit 111, and proxy model 112 are examples of parameter adjustment devices according to the present disclosure, and are also examples of computers. The UI unit 101 is an example of a display unit according to the present disclosure. The adjustment target 109 is the parameter of the present disclosure This is an example of a target to be set, and is also called the target to be optimized. The simulation model 110 is an example of a simulation unit that simulates the operation of the adjustment target 109, and is also called a simulator. The proxy model generation unit 111 is a processing unit for generating a proxy model 112. The proxy model 112 is a model in which the input / output relationship of the adjustment target 109 and / or the simulation model 110 is reduced in dimension and / or transformed in dimension using a mathematical model or the like. The proxy model generation unit 111 may be included in the parameter adjustment unit 103. The proxy model 112 generated by the proxy model generation unit 111 calculates and outputs the simulated response and / or partial response of the adjustment target 109 and / or the simulation model 110.

[0047] The parameter adjustment unit 103 is connected to the UI unit 101, DB unit 102, adjustment target 109, simulation model 110, proxy model generation unit 111, and proxy model 112 in a communicative manner, and acquires and / or provides data and / or information to and from each of them. Communication is performed by wired, wireless, or a combination of wired and wireless. The acquisition of data and / or information can also be described as the reception or input of data and / or information. The provision of data and / or information can also be described as the transmission or output of data and / or information.

[0048] The parameter adjustment system 100 searches for and adjusts m control parameters of the adjustment target 109 so that a predetermined operation (operation condition) to be performed on the adjustment target 109 satisfies an objective value representing a predetermined performance. Here, m is a natural number.

[0049] In this specification, a parameter for defining an operation condition is described as an operation parameter or an operation condition parameter. Examples of operation parameters include input voltage, output voltage, output current, moving speed, moving distance, or stop position. Also, an internal parameter of the adjustment target 109 that realizes an operation condition with desired performance is described as a control parameter or a control condition parameter. Examples of control parameters include control gain, filter time constant, or filter coefficient. An m-dimensional space composed of the range from the minimum value to the maximum value of each control parameter that can be set for the adjustment target 109 may be referred to as a search space. Also, a vector P = (P1, P2,..., Pm) in the search space having m control parameters as elements may be referred to as a parameter vector. Furthermore, a set of values of each control parameter Pi (i = 1 to m) and / or a set of parameter vectors Pj (j = 1 to m) may be referred to as a parameter set. The target of parameter search by the parameter adjustment system 100 is a parameter vector P including m control parameters that satisfy an objective value representing a predetermined performance within the search space.

[0050] Here, search means automatically searching for parameter values that satisfy an objective value using various known algorithms. In the present embodiment, the parameter adjustment system 100 performs a search for a parameter vector that is a set of m parameter values.

[0051] Adjustment means varying parameter values by search or manually. Also, a parameter related to search is described as a hyperparameter. Examples of hyperparameters include search priority (priority order), search range, number of searches, or search conditions (objective value setting).

[0052] The UI unit 101 is implemented by a UI device such as a display, keyboard, mouse, or touch panel. The UI unit 101 receives input of adjustment items from the user by the user operating the UI unit 101. The adjustment items are, for example, operating conditions, desired performance, or preferred parameter search range. The UI unit 101 provides the adjustment items received from the user to the parameter adjustment unit 103 (specifically, the adjustment item setting unit 104, which will be described later). The fact that the UI unit 101 receives adjustment items from the user and provides them to the parameter adjustment unit 103 may mean that the user sets the adjustment items.

[0053] The UI unit 101 displays information and / or trial history set in the parameter adjustment unit 103 in text and / or graphically. Here, "trial" refers to a series of operations in which the parameter adjustment unit 103 repeatedly performs the control parameter adjustment process to optimize the parameter value set for the adjustment target 109. The parameter adjustment process includes evaluating candidate parameter values ​​and determining the next candidate parameter value to be set for the adjustment target 109 based on the evaluation results.

[0054] In the example shown in Figure 1, the UI unit 101 is shown separately from the parameter adjustment unit 103, but it may be incorporated into the parameter adjustment unit 103. Also, the UI unit 101 may be assumed to have separate input and output devices, or it may be an integrated device.

[0055] The DB unit 102 is implemented, for example, by a storage device such as a hard disk drive in a data management server or data center. As will be explained in more detail below, the DB unit 102 stores the profile of the adjustment target 109 and a history of previously set parameters. The DB unit 102 may also store information about simulation models 110 or substitute models 112 that have been used in the past.

[0056] This information stored in the DB unit 102 is read and / or written by the parameter adjustment unit 103 (specifically, the adjustment item setting unit 104, which will be described later) and / or the UI unit 101.

[0057] In the example shown in Figure 1, the DB unit 102 is shown separately from the parameter adjustment unit 103, but it may also be incorporated into the parameter adjustment unit 103.

[0058] The parameter adjustment unit 103 searches for a parameter vector that achieves the target value by repeatedly determining the search space using m control parameters related to the adjustment target 109, shrinking and expanding the sub-search space, determining a trial vector from the initial search value vector, and obtaining its evaluation value.

[0059] Specifically, the parameter adjustment unit 103 searches for and adjusts the control parameters of the adjustment target 109 by determining the search range for the control parameters of the adjustment target 109 based on the adjustment items received from the user by the UI unit 101 and the information stored in the DB unit 102. For example, the parameter adjustment unit 103 searches for and adjusts the control parameters as follows.

[0060] The adjustment item setting unit 104 sets information related to the adjustment target 109, such as the search range for the control parameters of the adjustment target 109 (settable minimum and maximum values), default initial values, operating conditions, or desired target values, and sets the hyperparameters related to the search. The adjustment item setting unit 104 may also select a simulation model or surrogate model and / or a method for generating the surrogate model that is similar to the adjustment target 109. This information may be set by the user from the UI unit 101, or it may be set based on history information such as past adjustment history obtained from the DB unit 102.

[0061] The search range / initial value setting unit 105 sets a partial search range, which is the range of parameters that can be implemented, within the search range of the control parameters of the adjustment target 109, and sets an initial vector, which is an implementable parameter vector that satisfies predetermined conditions.

[0062] The parameter search unit 106 generates a combination of control parameter values ​​(i.e., a control parameter vector) that improves the target value based on past trial results.

[0063] The trial vector setting unit 107 generates a trial vector consisting of the generated control parameter vector and one or more operating parameter values ​​that constitute the operating conditions used for adjustment, and sets the trial vector for the adjustment target 109, the simulation model 110, and the surrogate model 112. As a result, the parameter adjustment unit 103 can control the adjustment target 109, the simulation model 110, and the surrogate model 112 to operate according to the trial vector. These controls may be performed simultaneously or sequentially. Alternatively, only one of these controls may be performed. In this specification, a trial vector means a combination of the control parameter vector and the values ​​of one or more operating parameters that constitute the operating conditions (i.e., the operating parameter vector).

[0064] Furthermore, when a device or functional unit (processing unit) sets information such as parameters, it may mean one or more of the following: the device or functional unit (processing unit) stores such information in a data table; the device or functional unit (processing unit) operates according to such information; and the device or functional unit (processing unit) controls other devices or functional units (processing units) to operate according to such information.

[0065] The adjustment target 109 is an example of a control system that is the target of setting or adjusting control parameters, such as an actual DC-DC converter or an actual production equipment such as a component mounting machine. A trial vector is set for these actual machines, they are made to operate under the set operating conditions, and the operating results are acquired by sensors (not shown).

[0066] The simulation model 110 is a model (or simulator) that simulates the operation of the actual device 109 to be adjusted. This simulator may be implemented in hardware, in software, or as a combination of hardware and software. Furthermore, the simulation model 110 can use multiple models implemented at various resolutions or granularities. A trial vector is set for the simulation model 110, it is made to operate under the set operating conditions, and the operation results of the simulation model 110 are obtained based on calculation results from a sensor (not shown) or a sensor model that simulates the actual sensor. The simulation model 110 that simulates the device to be adjusted 109 may also be called a virtual device, a virtual setting target, or a virtual adjustment target.

[0067] The operation evaluation unit 108 detects and evaluates the operation results of the adjustment target 109, the simulation model 110, or the surrogate model 112 (described later) that operated according to the set trial vector. As performance values ​​related to operation, for example, physical quantities such as voltage, current, displacement, velocity, vibration, noise, and / or heat generation at the adjustment target can be used. Furthermore, these multiple performance values ​​or evaluation values ​​in accordance with evaluation criteria may be used as a single target value for the search, for example, by normalizing or weighting them. In this specification, performance values, evaluation values, and target values ​​may be used interchangeably without particular distinction.

[0068] The parameter search unit 106 determines whether the target value has improved based on the operation results evaluated by the operation evaluation unit 108. The parameter search unit 106 uses the results of the series of trials as a prior distribution to generate the next trial vector.

[0069] Furthermore, the search range / initial value setting unit 105 determines whether the target value has stopped improving based on the operation results evaluated by the operation evaluation unit 108, and generates the next search range and / or search initial value based on this series of trial results.

[0070] The parameter adjustment unit 103 stores the distribution of trial values ​​(parameter values) for each parameter, the distribution of vectors for each trial (or setting), and the distribution of evaluation values ​​for each trial (or setting) in the DB unit 102. In other words, the parameter adjustment unit 103 writes the history of a series of trials to the DB unit 102.

[0071] As described above, the parameter adjustment unit 103 searches for a parameter vector that can obtain the desired performance within the search space, but the search space becomes very vast when there are many parameters. For example, Figure 4A shows an example of control parameters. In Figure 4A, each row represents one parameter, and it is assumed that 10 parameters are to be set. For example, for the parameter param1 in the first row, the default value is defined as 10000, the minimum settable value is 10, the maximum settable value is 100000, and the settable step width is 1000. Similarly, each parameter from param2 to param10 is defined, and the parameter adjustment unit 103 searches for a parameter vector within this 10-dimensional search space.

[0072] While the vectors within this search space are basically configurable for the adjustment target 109, in actual operation, to prevent damage or failure and ensure safe operation, the operating results must fall within predetermined protection criteria, such as maximum output voltage, maximum output current, and the presence or absence of oscillation, vibration, and heat generation. Although these protection criteria are more relaxed than the desired performance criteria, they represent the operating range (operable range) that must be observed to prevent damage or accidents to the actual machine. Therefore, the range in which parameters can actually be set exists in a subspace, which is a part of the search space. Until now, the tendency has been to adjust based on the experience of skilled users, using the area around which setting parameters that yield operating results within the criteria exist as a subspace. This often resulted in adjustments that sided on the safety side, making it difficult to extract performance improvements through new combinations of parameter values.

[0073] On the other hand, using the simulation model 110, it is possible to safely try all vectors in the search space without destroying the actual device. However, since running the simulation model 110, which simulates the operation of the actual device with sufficient accuracy, takes a very long time, it is not practical to try all vectors in the search space.

[0074] Therefore, in the parameter adjustment system 100 according to this embodiment, partial evaluation is performed using a surrogate model 112, which simulates partial features of the adjustment target 109 and / or the simulation model 110, and is a model obtained by reducing the dimensionality of its input / output relationship and / or transforming the dimensionality using a mathematical model or the like. The surrogate model 112 is a faster and less accurate model than the simulation model 110.

[0075] The surrogate model generation unit 111 generates a surrogate model 112. As the surrogate model 112, for example, a regression model that expresses the relationship between a parameter vector and an output response by regression can be used. Alternatively, a prediction model that predicts the output response from a parameter vector using a known machine learning model, such as a neural network based on deep learning, or PINNs (Physics Informed Neural Networks) that consider the governing equations and differential equations of physical laws can be used. Alternatively, the time-axis response can be converted to a frequency-axis response by dimensional transformation, and the amplitude and phase of its characteristic frequencies can be modeled and used. For example, Bode plots or Nyquist plots used in control systems are one such example, and the frequency response of the tuning target 109 and / or the simulation model 110 can be measured to determine the stability or responsiveness of the system. Alternatively, the tuning target 109 and / or the simulation model 110 can be linearly transformed to obtain a transfer function model or a state-space model, and the stability and responsiveness of the system can be determined based on these mathematical models. Alternatively, a portion of the adjustment target 109 and / or the simulation model 110, or a portion of the response result, may be represented by a surrogate model, and the different parts may be composed of multiple different surrogate models.

[0076] Multiple surrogate models 112 may be generated in the manner described above, and each may model the adjustment target 109 and / or the simulation model 110 using a different representation method. By setting trial vectors in the surrogate models 112, the surrogate models 112 calculate outputs corresponding to the operation results of the adjustment target 109 and / or the simulation model 110. These surrogate models 112 can reduce the computational load by relaxing or limiting the accuracy or operating range, and can be executed faster than the simulation model 110.

[0077] The parameter adjustment unit 103 first performs a search using the proxy model 112 to find a subspace that is actually executable from the configurable search space. The configurable search space is an example of the first search range, the subspace is an example of the second search range, the search to find the subspace is an example of the first search, and the proxy model 112 is an example of the first model.

[0078] The search range / initial value setting unit 105 sets the range from the minimum configurable value (min value) to the maximum configurable value (max value) of each parameter shown in Figure 4A as the first search range, and sets the default value (default value) as the initial search value. If no default value is specified, a random vector may be used as the initial search value, a vector specified by the user from the UI unit 101 may be used, or a similar vector stored in the DB unit 102 may be used.

[0079] The parameter search unit 106 starts a first search with the set first search range and initial vector, and searches for a vector whose evaluation in the surrogate model 112 satisfies predetermined conditions. For example, if the frequency response model of the simulation model 110 is set in the surrogate model 112, the operation evaluation unit 108 calculates evaluation values ​​indicating the stability and responsiveness of the simulation model 110 from the output of the surrogate model 112. Multiple vectors are obtained in which these evaluation values ​​fall within a predetermined range (preferably a range that is feasible and slightly relaxed than the protection standard). At this time, various known algorithms can be used for the search algorithm, for example, Bayesian optimization or a genetic algorithm can be used. Alternatively, random search or grid search may be used to perform a wider search than these algorithms which focus on searching the space around the local optimum. In the first search, rather than searching for a vector with better performance, the search hyperparameters are set to search a wide range in which feasible vectors exist, and the search is performed by performing a predetermined number of trials.

[0080] After performing the first search as described above, the parameter adjustment unit 103 aggregates the trial vectors whose evaluation values ​​fall within a predetermined range. For example, as shown in the search results in Figure 4B, suppose that out of the 1000 vectors tried in the first search, 390 vectors had evaluation values ​​within a predetermined range. The parameter adjustment unit 103 calculates the distribution or statistics of the values ​​for each parameter of the 390 vectors. In the example shown in Figure 4B, the parameter adjustment unit 103 calculates various statistical values ​​for each parameter, such as the mean, standard deviation, minimum value (min), percentiles (10%, 50%, 90%), and maximum value (max). Based on these statistical values, the parameter adjustment unit 103 sets a search range narrower than the first search range and performs a search using the simulation model 110 within that search range. A search range narrower than the first search range is an example of a second search range, the simulation model 110 is an example of a second model, and the search using the simulation model 110 is an example of a second search.

[0081] The parameter adjustment unit 103 sets, for example, the range from the minimum value to the maximum value in Figure 4B as the second search range for each parameter. Alternatively, the range from 10% to 90% of the percentile may be set as the second search range, or the range of mean ± standard deviation may be set as the second search range. By limiting the second search range more than the first search range in this way, the number of searches in the simulation model 110, which takes a long time to execute, can be reduced.

[0082] Furthermore, the parameter adjustment unit 103 sets the vector with the best performance from among the vectors whose evaluation values ​​fell within a predetermined range in the first search (390 vectors in Figure 4B) as the initial search value (initial vector) for the second search. The parameter adjustment unit 103 may use a vector randomly selected from among the vectors whose evaluation values ​​fell within a predetermined range in the first search as the initial search value for the second search, or it may use a vector specified by the user from the UI unit 101, or it may use a similar vector stored in the DB unit 102. Alternatively, the parameter adjustment unit 103 may select multiple vectors that are as far apart as possible from among the vectors whose evaluation values ​​fell within a predetermined range in the first search, so that a search can be performed over as wide a range as possible even within the subspace.

[0083] Next, the parameter adjustment unit 103 searches for a vector that can obtain the desired target value by performing a second search using the simulation model 110, targeting the set second search range and initial search values.

[0084] The parameter search unit 106 starts a second search with the set second search range and initial vector, and searches for a vector in which the target value evaluated in the simulation model 110 satisfies predetermined conditions. Based on past trial results, the parameter search unit 106 generates a control parameter vector that improves the target value and inputs it to the trial vector setting unit 107.

[0085] The trial vector setting unit 107 performs a preliminary evaluation using the surrogate model 112 before performing the evaluation using the simulation model 110. In other words, the parameter adjustment unit 103 prunes the second search range using the first model before performing the second search. The operation evaluation unit 108 evaluates the execution result of the surrogate model 112 and calculates a provisional target value. The parameter search unit 106 determines that if this provisional target value does not fall within a predetermined range, the evaluation using the simulation model 110 is unlikely to improve the target value. In this case, the parameter search unit 106 performs a second search pruning on this trial vector and instructs the trial vector setting unit 107 to cancel the execution and evaluation of the simulation model 110. This pruning operation reduces the number of times the simulation model 110 is executed for trial vectors that are unlikely to improve the target value, thereby shortening the search time. Preferably, the range for determining the provisional target value is narrower (limited) than the range of target values ​​used in the first search. In other words, it is desirable that the evaluation conditions in pruning be stricter than those in the first search, which allows for further optimization of the parameters and makes parameter setting more efficient.

[0086] If the provisional target value in the proxy model 112 falls within a predetermined range, the parameter search unit 106 instructs the trial vector setting unit 107 to execute the simulation model 110 and perform evaluation by the operation evaluation unit 108. The operation evaluation unit 108 evaluates the operation results of the simulation model 110 and calculates the target value. The parameter search unit 106 determines whether the target value has improved and generates the next trial vector using the series of trial results as a prior distribution. This operation is repeated to search for a vector that satisfies the target value.

[0087] Furthermore, the parameter search unit 106 and / or the search range / initial value setting unit 105 determine that a local optimum may have been reached if the desired performance cannot be obtained for a predetermined number of times or a predetermined time and the target value during the search does not improve. In this case, the parameter search unit 106 and / or the search range / initial value setting unit 105 change the initial conditions and retry the second search. Here, it is important to search a different subspace within the search space, so the search range / initial value setting unit 105 first re-selects the initial vector. The initial vector to be re-selected may be randomly selected from a plurality of initial vector candidates obtained in the first search, or a vector that is farther away from the current best vector may be selected. Alternatively, a vector specified by the user from the UI unit 101 may be selected, or a vector that has been slightly modified by the user from the current best vector may be used, or a similar vector stored in the DB unit 102 may be used.

[0088] The search range / initial value setting unit 105 may reset the second search range, which is a search subspace based on the first search, in addition to resetting the initial vector of the second search. The parameter adjustment unit 103 may reset the second search range, for example, by expanding the currently set second search range by a predetermined percentage. Alternatively, the parameter adjustment unit 103 may reset the second search range by re-executing the first search after changing the proxy model 112 or its evaluation conditions.

[0089] The parameter adjustment unit 103 performs the first and second searches a predetermined number of times and / or for a predetermined period of time as described above, selects the trial vector that comes closest to the desired performance as the best vector, and adopts the trial vectors that fall within the predetermined performance range and the protection standard range as a set of candidate vectors.

[0090] The parameter adjustment unit 103, similar to the first and second searches, determines the search range and initial search value from multiple vectors that yielded the desired performance in the second search, and performs adjustments to the actual machine. In other words, the parameter adjustment unit 103 performs an applicability evaluation of the adjustment target 109 to the actual machine based on the results of the second search, and sets the parameters of the adjustment target 109 based on the results of the applicability evaluation. Specifically, the parameter adjustment unit 103 performs adjustments to the adjustment target 109, which is the actual machine, using the best vector and the set of candidate vectors obtained in the second search. By using the best vector as the initial search value for the applicability evaluation and searching for vectors within a search range that falls within the protection standard range, it becomes possible to perform adjustments according to the individual differences of the actual machine.

[0091] Through this exploration process, the parameter adjustment system 100 displays information or trial history set in the parameter adjustment unit 103 on the UI unit 101, making it available to the user. As a result, the user can manipulate the exploration conditions while referring to (for example, comparing) previously set parameters, enabling them to efficiently set appropriate control parameters. Furthermore, it becomes possible to efficiently explore and adjust control parameters. In addition, the user can easily grasp the progress during exploration or adjustment from the displayed information or trial history. This makes it possible for even inexperienced users, such as newcomers, to efficiently set appropriate control parameters.

[0092] In Figure 1, the illustrated processing blocks may be implemented in one device or in multiple devices. For example, the UI unit 101, DB unit 102, parameter adjustment unit 103, simulation model 110, proxy model generation unit 111, and proxy model 112 may be implemented in one device, while the adjustment target 109 may be implemented in another device. Alternatively, the parameter adjustment unit 103 may be implemented in a server device on the cloud, and the UI unit 101 may be implemented in a local device such as a PC (Personal Computer), smartphone, or tablet, with the server device and the local device connected via a communication network (not shown). Furthermore, there may be multiple adjustment targets 109, simulation models 110, or proxy models 112.

[0093] Furthermore, the UI unit 101 and / or parameter adjustment unit 103 may not only manually set adjustment items entered by the user, but may also automatically set control parameters and acquire operation results by utilizing existing automatic control technologies. As an automatic control technology, for example, an RPA (Robotic Process Automation) may be used to automatically control the GUI (Graphical User Interface) of adjustment software that manually adjusts the parameters of the adjustment target using image recognition. This makes it possible to automatically adjust parameters even for adjustment targets 109 of existing technologies that do not have a dedicated communication interface.

[0094] <Configuration of Parameter Adjustment Device> Figure 2 is a simplified diagram showing an example of the hardware configuration of the parameter adjustment system 100. The parameter adjustment system 100 is implemented by a server device, a computer device such as a PC, or an information processing device. For example, as shown in Figure 2, the parameter adjustment system 100 may include a storage device 121, a processing device 122, a UI device 123, and a communication device 124, which are interconnected via a bus 125 or a communication network. The processing device 122 has a circuit configuration such as a processor or a controller. The storage device 121 stores a program 129 or instruction for causing the processing device 122 to perform a desired process or for causing the processing device 122 to realize a desired function. The program 129 or instruction may be downloaded from an external server device via a communication network or the like, or read from a removable storage medium such as flash memory.

[0095] The storage device 121 is implemented by RAM (Random Access Memory), flash memory, or a hard disk drive, etc. The storage device 121 stores the installed program 129 or instructions, along with files and data used to execute the program 129 or instructions. The storage device 121 may also include a non-transitory storage medium such as non-volatile memory.

[0096] Furthermore, if the DB unit 102 is incorporated into the parameter adjustment system 100, the data accumulated in the DB unit 102 may be stored in the storage device 121.

[0097] The processing unit 122 may be implemented by a general-purpose processor or controller circuit configuration, or by a dedicated processor or controller circuit configuration. When implemented by a general-purpose processor, the processing unit 122 may be implemented by one or more CPUs (Central Processing Units), GPUs (Graphics Processing Units), Processing Circuits, or FPGAs (Field Programmable Gate Arrays), which may be configured with one or more processor cores. The processing unit 122 executes the functions and processes of the parameter adjustment system 100 based on the program 129 or instructions stored in the storage device 121 and data such as parameters used to execute the program 129 or instructions.

[0098] The UI device 123 may be configured to include input devices such as a keyboard, mouse, camera, and microphone, output devices such as a display, speaker, headset, and printer, or input / output devices such as a touch panel. The UI device 123 provides an interface between the user and the parameter adjustment system 100. For example, the user operates the parameter adjustment system 100 by operating the GUI displayed on the display or touch panel using a keyboard, mouse, or finger.

[0099] If the UI unit 101 is incorporated into the parameter adjustment system 100, the UI unit 101 may be implemented by part or all of the UI device 123.

[0100] The communication device 124 may be implemented by various communication circuits that perform communication processing with external devices, the Internet, a LAN (Local Area Network), or a VPN (Virtual Private Network).

[0101] The hardware configuration of the parameter adjustment system 100 described above is merely an example. The parameter adjustment system 100 according to this disclosure may be implemented by other suitable hardware configurations.

[0102] <Operation of the parameter adjustment system> Figure 3 is a flowchart showing an example of the operation of the parameter adjustment system 100 according to the embodiment of this disclosure.

[0103] First, in step S301, the parameter adjustment unit 103 sets the adjustment items.

[0104] Next, in step S302, the parameter adjustment unit 103 generates a substitute model 112. The substitute model generation unit 111 generates the substitute model 112 based on the response of the actual machine and / or simulation model 110 of the adjustment target 109 using the initial parameters.

[0105] Next, in step S303, the parameter adjustment unit 103 performs a first search using the surrogate model 112, targeting the first search range. By performing a parameter search using the surrogate model 112, the parameter adjustment unit 103 narrows down the search range and obtains provisional parameter candidates.

[0106] Next, in step S304, the parameter adjustment unit 103 sets a second search range that is narrower than the first search range based on the results of the first search.

[0107] Next, in step S305, the parameter adjustment unit 103 uses the proxy model 112 to prune the second search range.

[0108] Next, in step S306, the parameter adjustment unit 103 performs a second search using the simulation model 110, targeting the second search range after pruning. The parameter adjustment unit 103 obtains parameter candidates by performing a parameter search using the simulation model 110 while simultaneously pruning the second search range using the proxy model 112.

[0109] Next, in step S307, the parameter adjustment unit 103 performs an evaluation of the applicability of the adjustment target 109 to the actual machine based on the results of the second search.

[0110] Next, in step S308, the parameter adjustment unit 103 sets the parameters of the adjustment target 109 based on the results of the applicability evaluation. The parameter adjustment unit 103 obtains the final parameters by performing fine-tuning of the parameters of the adjustment target 109 using the actual machine.

[0111] Next, in step S309, the parameter adjustment unit 103 determines whether predetermined termination conditions are met. These termination conditions include, for example, that the desired performance has been achieved with the final parameters, or that a predetermined time has elapsed since the start of the adjustment.

[0112] If the predetermined termination conditions are not met (step S309: NO), the parameter adjustment unit 103 then prompts the user to reset the execution conditions for the second search (search range, initial value, or target value, etc.) in step S310. After resetting the execution conditions for the second search according to the data input from the user, the parameter adjustment unit 103 returns to step S304 and re-executes the processes from step S304 onward.

[0113] The reset in step S310 may also be the regeneration of the proxy model 112, in which case the parameter adjustment unit 103 will re-execute the processes from step S302 onwards. The reset in step S310 may also be the reset of the execution conditions for the first search (search range, initial value, or target value, etc.), in which case the parameter adjustment unit 103 will re-execute the processes from step S303 onwards. The reset in step S310 may also be the reset of the pruning conditions, in which case the parameter adjustment unit 103 will re-execute the processes from step S305 onwards.

[0114] If the predetermined termination conditions are met (step S309: YES), the parameter adjustment unit 103 terminates the process.

[0115] If the target value for each step of generating the proxy model 112 (step S302), the first search (step S303), or the second search (step S306) is not reached, or if a predetermined time has elapsed, the parameter adjustment unit 103 may proceed to step S309 without moving to the next step, and in step S310, prompt the user to reset the execution conditions. After resetting the execution conditions according to the data input from the user, the parameter adjustment unit 103 returns to the original step or an earlier step to perform a parameter re-search.

[0116] For example, if the proxy model 112 could not be generated in step S302 due to reasons such as not being able to obtain a sufficient amount of response data from the actual device, or not being able to obtain response data under predetermined conditions, the user may be prompted to change the generation conditions for the proxy model 112. For example, the user may be asked to generate the proxy model 112 under conditions that limit the range of operation, or to use a different type of proxy model 112. If the generation conditions are changed in step S310 after moving to step S309 according to the data input from the user, the proxy model generation unit 111 will re-execute the processing from step S301 or step S302 onward according to the changed generation conditions.

[0117] Furthermore, if, for example, the criteria for the search objective value in the first search were too strict or too lenient, or the accuracy of the surrogate model 112 was insufficient, and provisional parameters that satisfy the search objective value could not be obtained in the first search in step S303, the user may be prompted to change the generation conditions of the surrogate model 112 or to change the execution conditions of the first search. For example, the user may be suggested to loosen or tighten the criteria for the search objective value in the first search, or to use a different type of surrogate model 112. In step S310, after moving to step S309, if the generation conditions or execution conditions are changed according to the data input from the user, the parameter adjustment unit 103 re-executes any of the processes from steps S301 to S303 onward according to the changed generation conditions or execution conditions.

[0118] Furthermore, if, for example, the pruning criteria were too strict or too lenient, the criteria for the search objective value in the second search were too strict or too lenient, or the second search range was too wide or too narrow, and parameters that satisfy the search objective value could not be obtained in the second search in step S306, the user may be prompted to change the pruning criteria or change the execution conditions of the second search. For example, if the number of executions of the simulation model 110 is too few due to an excessive number of prunings, the user may be prompted to relax the pruning criteria, change the second search range, or change the initial vector of the second search. If, in step S310 after moving to step S309, the pruning criteria or the execution conditions of the second search are changed according to the data input from the user, the parameter adjustment unit 103 will re-execute any of the processes from steps S301 to S306 onward according to the changed criteria or execution conditions.

[0119] Furthermore, if, for example, the range of variation due to individual differences in the actual machine 109 to be adjusted is large, and parameters that satisfy the desired performance cannot be obtained in the applicability evaluation in step S307, the user may be prompted to change the execution conditions of the applicability evaluation. For example, the user may be asked to relax the variation range condition or change the initial vector of the applicability evaluation. If, in step S310 after moving to step S309, the execution conditions of the applicability evaluation are changed according to the data input from the user, the parameter adjustment unit 103 will re-execute any of the processes from steps S301 to S307 onward according to the changed execution conditions.

[0120] The parameter adjustment unit 103 gradually narrows the search range by repeating these operations, sequentially changing the evaluation means from the surrogate model 112 to the simulation model 110 and then to the actual machine being adjusted 109. This shortens the time required for parameter adjustment. Note that the narrowing down process is not limited to three stages: surrogate model 112, simulation model 110, and the machine being adjusted 109; it may also involve four or more stages of narrowing down. In other words, in the execution of the first search using the surrogate model 112, the parameter adjustment unit 103 may use multiple surrogate models 112 with different speeds and accuracy to perform a multi-stage search that becomes slower and more accurate in later stages. Similarly, in the execution of the second search using the simulation model 110, the parameter adjustment unit 103 may use multiple surrogate models 112 with different speeds and accuracy to perform a multi-stage search that becomes slower and more accurate in later stages.

[0121] <Display Processing> The display processing performed in the parameter adjustment system 100 will be described below. In the display processing, the user can control parameter adjustment operations, set the parameter search range, or display set values ​​and response results by operating the UI unit 101.

[0122] Figure 5 shows an example of the operation screen 500 of the UI unit 101 when parameter search is performed. The operation screen 500 includes, for example, a maximum search time setting area 501, a maximum search range setting area 502, a search interruption button 503, a search resume button 504, an elapsed time display area 505, a valid trial count display area 506, a parameter display / operation area 507, a proxy model response display area 508, and a simulation model response display area 509. The parameter display / operation area 507 displays a search range display bar 510, a parameter vector display bar 511, and a parameter distribution display bar 512.

[0123] In the maximum search time setting area 501, the maximum execution time for parameter search is set. In the example shown in Figure 5, the maximum search time is set to 1.5 hours, in which case the search will end 1.5 hours after the start of the search.

[0124] In the maximum search range setting area 502, the search range for each parameter is set collectively. The search range for each parameter displayed in the parameter display / operation area 507 (corresponding to the search range display bar 510) can be changed collectively to the percentage set in the setting area 502. In the example shown in Figure 5, the maximum search range is set to 80%, in which case the search range for each parameter can be reduced to 80%. Alternatively, by setting the maximum search range to 110%, the search range for each parameter can be expanded to 110%.

[0125] When the search interruption button 503 is pressed or touched, the ongoing search is interrupted. When the search resume button 504 is pressed or touched, the interrupted search is resumed. These actions may be toggled with a single button. The user can interrupt and resume the search at any time.

[0126] The elapsed time display area 505 displays the elapsed time of the current search. The elapsed time is displayed as a progress bar relative to the maximum execution time set in the maximum search time setting area 501, and the remaining time (40 minutes in the example in Figure 5) is displayed.

[0127] The effective trial count display area 506 displays the number of effective trials in the current search. The effective trial count refers to the number of times the trial vector was evaluated by the simulation model 110 without being pruned. The effective trial count display area 506 displays the total number of trials (1234 in the example in Figure 5) and the number of effective trials (445 in the example in Figure 5) in a bar graph. Below the bar graph, the total number of trials, the number of prunings (789 in the example in Figure 5), and the number of simulation runs (the same as the effective trial count) are displayed. In the second search using the simulation model 110, pruning is performed by the surrogate model 112, so the user can know whether the trial vector was pruned or evaluated using the simulation model 110. If the number or percentage of pruning is too high or too low, the efficiency of the search will decrease, so the user can use this indicator to verify the validity of the search conditions such as pruning.

[0128] The parameter display and operation area 507 displays the parameter currently being searched for and its search range, a reference parameter for comparison, and the initial values ​​of each parameter being searched. To avoid complicating the diagram, it is omitted in Figure 5, but the values ​​of each parameter are displayed as markers and text on a number line. The parameter display and operation area 507 uses a number line to show a numerical axis (horizontal axis) normalized by the minimum and maximum values ​​that can be set, and displays multiple parameters in parallel coordinates. The parameter vector is displayed as a parameter vector display bar 511, which is a polyline connecting each parameter. The search range of each parameter is displayed as a search range display bar 510, which is a bar-shaped first figure shown as a dashed line. The left end of the search range display bar 510 indicates the lower limit of the search, and the right end indicates the upper limit of the search. A solid partial line is displayed on the dashed line of the search range display bar 510, and this partial line is the parameter distribution display bar 512, which shows the distribution of parameters tried in the current search. During the search, the polyline representing the currently being tested vector changes, allowing the user to see where within the search range the search is taking place and how it differs from the reference parameters. When the search is paused, the parameter display / operation area 507 switches to an operable mode. The user can change the settings by clicking or dragging the markers for each parameter or the bars or handles that indicate the search range. Alternatively, double-clicking the parameter display / operation area 507 may transition to the search range setting screen described later. Furthermore, the polyline connecting the initial search values ​​of each parameter shows handles (point-shaped second figures) corresponding to the initial values ​​of each parameter, and the user can change the initial search value of the corresponding parameter by dragging these second figures.

[0129] The proxy model response display area 508 displays an indication that the proxy model frequency response is being searched, and also shows the result of calculating the response by the proxy model 112 using the current trial vector. The proxy model response display area 508 also displays a comparison with the response using the reference vector.

[0130] The simulation model response display area 509 displays an indication that the simulation time response is being explored, along with the result of calculating the response by the simulation model 110 using the current trial vector. The simulation model response display area 509 also shows a comparison with the response using the reference vector.

[0131] Figure 6 shows an example of the operation screen 600 of the UI unit 101 when setting the parameter search range. The operation screen 600 has a parameter display / operation area 507 and a parameter range table 601. The parameter range table 601 corresponds to the search results of the first search shown in Figure 4B.

[0132] As described above, the parameter display and operation area 507 is in an operable state. The search range display bar 510 for each parameter is linked to the value of a predetermined cell in the parameter range table 601. The parameter range table 601 displays, for example, the statistical values ​​of each element of a vector that satisfies predetermined criteria obtained from the search results of the proxy model 112, and each cell is editable. For example, if the user wants to set the minimum value (min) column of the parameter range table 601 as the lower search limit and the maximum value (max) column as the upper search limit, the user can do so by double-clicking, for example, the top cell in the minimum value (min) column and for example, the top cell in the maximum value (max) column. At this time, the search range is set in conjunction with the values ​​in these columns and is reflected in the search range display bar 510 of the parameter display and operation area 507. Also, if the user wants to use the value of a different cell for some parameters, the user can double-click that cell or set it as the search range from the context menu. Alternatively, you can directly change the value in the cell, or you can change it by manipulating the search range display bar 510, as described later.

[0133] Figure 7 is a schematic diagram showing an example of a display area 701 for one parameter in the parameter display and operation area 507. The display area 701 shows a search range display bar 510, with the area around its left end being the search lower limit handle 703, the area around its right end being the search upper limit handle 704, and the area near the center being the translation handle 705. The user can change the lower limit of the search range by dragging the search lower limit handle 703. The user can also change the upper limit of the search range by dragging the search upper limit handle 704. Furthermore, the user can translate the median value without changing the width from the search lower limit to the search upper limit by dragging the translation handle 705. When using a UI device such as a touch panel capable of multipoint or multitouch, operations such as pinch-in and pinch-out can be performed. While pinch-in and pinch-out gestures are usually assigned to zooming in or out of the entire screen, for the parameter display area 701, pinch-in may be assigned to zooming in, pinch-out to zooming in, and slide to move the search range. For example, as shown in the leftmost diagram of the hand, the user can zoom in on the lower search limit handle 703 (first endpoint) and the upper search limit handle 704 (second endpoint). Also, as shown in the center diagram of the hand, the user can zoom in on the lower search limit handle 703 (first endpoint) and the upper search limit handle 704 (second endpoint). Furthermore, as shown in the rightmost diagram of the hand, the user can move the search range by sliding on the lower search limit handle 703 (first endpoint) and the upper search limit handle 704 (second endpoint).

[0134] <Effects, etc.> According to the embodiment of this disclosure, the processing unit 122, as an information processing device, performs a first search targeting a first search range using a proxy model 112 (first model) of the adjustment target 109 (setting target). The processing unit 122 also sets a second search range that is narrower than the first search range based on the results of the first search. The processing unit 122 also performs a second search targeting the second search range using a simulation model 110 (second model) of the adjustment target 109. The first model is a proxy model that is faster and less accurate than the second model. The processing unit 122 sets the parameters of the adjustment target 109 based on the results of the second search. As a result, appropriate parameters can be efficiently set by the first and second searches.

[0135] Furthermore, according to the embodiments of this disclosure, the display screen shown on the UI unit 101 displays the range, vector, trial value, or response of each parameter, making it operable. This allows the user to easily manipulate the search range for a large number of parameter combinations. As a result, the search range can be narrowed down efficiently. In addition, the user can easily grasp the progress during the search or adjustment by visually confirming the displayed information or trial history. This makes it possible for even inexperienced users, such as newcomers, to efficiently set appropriate control parameters.

[0136] <Other> Although embodiments have been described above with reference to the drawings, this disclosure is not limited to the examples described above. Those skilled in the art will be able to conceive of various modifications or alterations within the scope of the claims. Such modifications or alterations will also be understood to fall within the technical scope of this disclosure. Furthermore, the components in the embodiments may be combined in any way without departing from the spirit of this disclosure.

[0137] One or more of the functional units (processing units) described in the above embodiment may be integrated as needed, or one functional unit (processing unit) may be divided into multiple sub-functional units (sub-processing units).

[0138] The order of steps in the flowchart described in the above embodiment is merely an example and is not limited to the order shown.

[0139] The details described in the above embodiments may be combined as appropriate, provided they are not contradictory or unless explicitly stated to be incompatible.

[0140] Furthermore, the notation "...part" in the above-described embodiment may be replaced with other notations such as "...circuitry," "...assembly," "...device," "...unit," or "...module."

[0141] This disclosure can be implemented using software, hardware, or software integrated with hardware.

[0142] Each functional block used in the description of the above embodiments may be implemented partially or entirely as an integrated circuit called an LSI (Large Scale Integration), and each process described in the above embodiments may be controlled partially or entirely by one LSI or a combination of LSIs. An LSI may consist of individual chips, or it may consist of one chip that includes some or all of the functional blocks. An LSI may include data input and output sections. Depending on the degree of integration, LSIs may also be referred to as ICs, system LSIs, super LSIs, or ultra LSIs.

[0143] The integrated circuit implementation method is not limited to LSIs; it may also be implemented using dedicated circuits, general-purpose processors, or dedicated processors. Furthermore, a Field Programmable Gate Array (FPGA) that can be programmed after LSI manufacturing, or a reconfigurable processor capable of reconfiguring the connections and / or settings of circuit cells within the LSI, may be used. This disclosure may also be implemented as digital or analog processing.

[0144] Furthermore, if advances in semiconductor technology or other derived technologies lead to the emergence of integrated circuit technologies that can replace LSIs, then functional blocks may be integrated using those technologies. The application of biotechnology, for example, is a possible possibility.

[0145] This disclosure is applicable to all types of devices, systems, and equipment having communication capabilities (collectively referred to as communication equipment). Communication equipment may include a radio transceiver and a processing / control circuit. The radio transceiver may include a receiver and a transmitter, or both as functions. The radio transceiver (transmitter, receiver) may include an RF (Radio Frequency) module and one or more antennas. The RF module may include an amplifier, an RF modulator / demodulator, or similar. Non-exclusive examples of communication devices include telephones (mobile phones, smartphones, etc.), tablets, personal computers (PCs) (laptops, desktops, notebooks, etc.), cameras (digital still / video cameras, etc.), digital players (digital audio / video players, etc.), wearable devices (wearable cameras, smartwatches, tracking devices, etc.), game consoles, digital book readers, telehealth / telemedicine devices, vehicles or mobile transport with communication capabilities (cars, airplanes, ships, etc.), and combinations of the above-mentioned devices.

[0146] Communication devices are not limited to portable or movable devices, but also include all kinds of devices, systems, and equipment that are difficult to carry or are fixed and cannot be carried, such as smart home devices (appliances, lighting equipment, smart meters or measuring instruments, control panels, etc.), vending machines, and any other "things" that may exist on an IoT (Internet of Things) network.

[0147] Communication includes data communication via cellular systems, wireless LAN systems, and communication satellite systems, as well as data communication using combinations of these.

[0148] Furthermore, the communication device also includes devices such as controllers and sensors that are connected to or linked to a communication device that performs the communication functions described in this disclosure. For example, this includes controllers and sensors that generate control signals and data signals used by the communication device that performs the communication functions of the communication device.

[0149] Furthermore, communication equipment includes infrastructure facilities such as base stations, access points, and any other devices, devices, and systems that communicate with or control the aforementioned non-limited types of equipment.

[0150] Although specific examples of this disclosure have been described in detail above, these are merely illustrative and do not limit the scope of the claims. The technologies described in the claims include various modifications and changes to the specific examples illustrated above.

[0151] This disclosure can be used as a parameter setting method for setting parameters of a control system, and is useful, for example, for operation and management methods of a control system.

Claims

1. An information processing method for setting parameters relating to a target to be set, wherein the information processing device performs a first search using a first model of the target to be set, targeting a first search range, sets a second search range based on the results of the first search, performs a second search using a second model of the target to be set, the first model being a surrogate model that is faster and less accurate than the second model, and sets the parameters based on the results of the second search.

2. The information processing method according to claim 1, wherein the second search range is narrower than the first search range.

3. The information processing method according to claim 1, further comprising pruning the second search range using the first model before performing the second search.

4. The information processing method according to claim 3, wherein the evaluation conditions in the pruning are stricter than the evaluation conditions in the first search.

5. The information processing method according to claim 1, wherein, in setting the parameters, an applicability evaluation to the target machine is performed based on the result of the second search, and the parameters are set based on the result of the applicability evaluation.

6. The information processing method according to claim 5, further comprising resetting the execution conditions for the second search, which include at least one of the second search range and the initial search value of the second search, based on the results of the applicability evaluation, and re-executing the second search based on the reset execution conditions.

7. The information processing method according to claim 6, wherein, in resetting the execution conditions, the reset execution conditions are obtained from an input device having an operation screen, the operation screen displays a bar-shaped first figure indicating the second search range and a dot-shaped second figure indicating the initial search value, the second search range is reset by pinch-in, pinch-out, slide, or drag operation on the first figure, and the initial search value is reset by drag operation on the second figure.

8. The information processing method according to claim 1, wherein in the execution of the first search, multiple stages of searching are performed, with the speed decreasing and accuracy decreasing in the later stages, and in the execution of the second search, multiple stages of searching are performed, with the speed decreasing and accuracy decreasing in the later stages.

9. An information processing method for setting parameters relating to a target to be set, wherein an information processing device sets execution conditions for a search that include at least one of a search range and an initial search value; executes the search using a model of the target to be set based on the set execution conditions; sets the parameters based on the results of the search; in setting the execution conditions, the execution conditions are obtained from an input device having an operation screen; the operation screen displays a bar-shaped first figure indicating the search range and a dot-shaped second figure indicating the initial search value; the search range is set by a pinch-in, pinch-out, slide, or drag operation on the first figure; and the initial search value is set by a drag operation on the second figure.

10. The information processing method according to claim 9, wherein the first figure has a first endpoint corresponding to the lower limit of the search range and a second endpoint corresponding to the upper limit of the search range, the search range is reduced by a pinch-in operation on the first and second endpoints, the search range is expanded by a pinch-out operation on the first and second endpoints, the search range is moved by a slide operation on the first and second endpoints, the lower limit is changed by a drag operation on the first endpoint, the upper limit is changed by a drag operation on the second endpoint, and the initial search value is changed by a drag operation on the second figure.

11. An information processing device comprising a circuit configuration for setting parameters relating to a target to be set, wherein the circuit configuration includes: performing a first search using a first model of the target to be set, targeting a first search range; setting a second search range based on the results of the first search; performing a second search using a second model of the target to be set, targeting the second search range; the first model being a surrogate model that is faster and less accurate than the second model; and setting the parameters based on the results of the second search.

12. An information processing device comprising a circuit configuration for setting parameters relating to a target to be set, wherein the circuit configuration sets execution conditions for a search including at least one of a search range and an initial search value, executes the search using a model of the target to be set based on the set execution conditions, sets the parameters based on the results of the search, and in setting the execution conditions, obtains the execution conditions from an input device having an operation screen, the operation screen displays a bar-shaped first figure indicating the search range and a dot-shaped second figure indicating the initial search value, the search range is set by pinch-in, pinch-out, slide, or drag operation on the first figure, and the initial search value is set by drag operation on the second figure.

13. A program for causing an information processing device to perform a process for setting parameters relating to a target to be set, wherein the process includes: performing a first search using a first model of the target to be set, targeting a first search range; setting a second search range based on the results of the first search; performing a second search using a second model of the target to be set, targeting the second search range, wherein the first model is a surrogate model that is faster and less accurate than the second model; and setting the parameters based on the results of the second search.

14. A program for causing an information processing device for setting parameters relating to a target to be set to execute a process, wherein the process includes: setting execution conditions for a search including at least one of a search range and an initial search value; executing the search using a model of the target to be set based on the set execution conditions; setting the parameters based on the results of the search; in setting the execution conditions, the execution conditions are obtained from an input device having an operation screen; the operation screen displays a bar-shaped first figure indicating the search range and a dot-shaped second figure indicating the initial search value; the search range is set by pinch-in, pinch-out, slide, or drag operation on the first figure; and the initial search value is set by drag operation on the second figure.

Images