User assistance device

Abstract

This user assistance device according to an embodiment acquires abnormal state information relating to an abnormal state caused by a control device, and acquires processing context information in software processing of the control device that caused the abnormal state. Subsequently, the user assistance device, on the basis of the abnormal state information and the processing context information, when creating a prompt instructing a generative AI system to generate a method for coping with the abnormal state, transmits the prompt to the generative AI system, and when receiving a response from the generative AI, presents the content of the response.

Description

Assistive devices

[0001] An embodiment of the present invention relates to a technique for presenting information for resolving an abnormal state that occurs in a control device that operates a machine tool or the like.

[0002] In the past, there has been a technique for notifying a user of specific information associated with each abnormal state in order to assist in resolving an abnormal state that has occurred in a control device. Examples of means for notifying the specific information include displaying an alarm number or a warning, or making an audio notification.

[0003] Japanese Patent Application Laid-Open No. 2018-097616

[0004] One possible method is to input specific information about the notified abnormal condition into a generating AI device to obtain auxiliary information to resolve the abnormal condition. However, when only specific information about the abnormal condition is provided to the generating AI device, it is difficult to obtain an appropriate response as auxiliary information because the context of the processing flow that led to the abnormal condition is not taken into account. Therefore, a system that can present more appropriate auxiliary information to the user to resolve the abnormal condition that has occurred in the control device is desired.

[0005] In an embodiment, the user assistance device acquires abnormal condition information regarding an abnormal condition that has occurred in a control device, and acquires processing context information for the software processing of the control device that caused the abnormal condition. Then, based on the abnormal condition information and the processing context information, the user assistance device creates a prompt to instruct a generation AI system to generate a method of dealing with the abnormal condition, sends the prompt to the generation AI system, and upon receiving a response from the generation AI, presents the content of the response.

[0006] FIG. 1 is a block diagram showing the functional configuration of a user assistance device in a first embodiment. FIG. 2 is a diagram illustrating an example of information provided by a control device in the first embodiment. FIG. 3 is a diagram explaining a function call history obtained from a call stack in the first embodiment. FIG. 4 is a diagram illustrating an example of processing context information acquired by a user assistance device in the first embodiment. FIG. 5 is a diagram illustrating a first example of a prompt created in the first embodiment. FIG. 6 is a diagram illustrating a second example of a prompt created in the first embodiment. FIG. 7 is a block diagram showing the functional configuration of a user assistance device in a second embodiment. FIG. 8 is a diagram illustrating an example of a response correction prompt created in the second embodiment.

[0007] The following describes an exemplary embodiment of the present invention. In order to obtain a solution to resolve an abnormal state that has occurred in a control device, the user assistance device of this embodiment inputs processing context information related to the internal processing of the control device as well as specific information about the abnormal state to a generation AI system as a prompt, and presents the response to the prompt as auxiliary information to the user.

[0008] Here, the generative AI system is an interactive response device that uses a machine learning model that has undergone additional learning (e.g., fine tuning or instruction tuning) to generate countermeasures for questions about abnormal conditions based on data such as control device manuals, specifications, past inquiries and their responses, or a model that can access the above data and generate answers.

[0009] 1 is a block diagram showing the functional configuration of a user assistance device 1 in embodiment 1. The user assistance device 1 is an information processing device (computer) that includes a control unit (processor), a storage unit (memory), and various input / output interfaces, and presents a method for dealing with an abnormal state that occurs in a control device.

[0010] The control unit includes an abnormal state information acquisition unit 11, a processing context information acquisition unit 12, a prompt creation unit 13, a transmission unit 14, a reception unit 15, and a presentation unit 16, and these functional units operate by the control unit reading and executing various software stored in the memory unit.

[0011] The abnormal state information acquiring unit 11 acquires specific information such as an alarm number output by software processing, and its description, as abnormal state information relating to an abnormal state.

[0012] The processing context information acquisition unit 12 acquires processing context information in the software processing of the control device that caused the abnormal condition. The acquired processing context information includes, for example, any of the following information: - Information specific to the software executing the processing that caused the abnormal condition; - Program counter in the processing that caused the abnormal condition; - Internal information of the processor in the processing that caused the abnormal condition: values ​​of registers including the stack pointer and data being processed; - Call stack in the processing that caused the abnormal condition: including the order in which functions were called up to the processing that caused the abnormal condition, the return addresses of each call location, and the values ​​of registers temporarily saved when the function was called; - Assembly code related to the call stack in the processing that caused the abnormal condition: assembly code near the return addresses of each call location up to the processing that caused the abnormal condition; - Source code related to the call stack in the processing that caused the abnormal condition: source code that generated the assembly code; - Related documents linked to the source code.

[0013] The related documents can be obtained, for example, by RAG (Retrieval Augmented Generation). Specifically, for example, there are specifications that explain the internal specifications created during software development of a control device. By creating a database based on these specifications, the processing context information acquisition unit 12 can search the database based on information such as alarm names and function names in the calling order to obtain the related documents.

[0014] The prompt creation unit 13 creates a prompt that instructs the generation AI system to create a method of dealing with the abnormal state based on the abnormal state information and the processing context information. For example, the processing context information may be described in the prompt as related information. In this case, the prompt creation unit 13 may include in the prompt a statement requesting the generation AI system not to include the processing context information described in the prompt in the response.

[0015] The sending unit 14 sends the prompt generated by the prompt creating unit 13 to the generating AI system. The receiving unit 15 receives a response from the generating AI. The presentation unit 16 presents the content of the response from the generating AI to the user by screen display, audio, or the like.

[0016] 2 is a diagram illustrating an example of information provided by the control device in the first embodiment. During notification processing of an abnormal state, the control device generates notification data in a predetermined format, such as the JSON format. Here, the alarm number, which serves as specific information about the abnormal state, the call stack (top address and data), the program counter, data from various registers, and specific information about the software (version number, series information, etc.) are described. The processing context information acquisition unit 12 appropriately acquires the data and related information provided by the control device in this manner as processing context information.

[0017] 3 is a diagram illustrating a function call history (stack trace) obtained from a call stack in the first embodiment. Here, the stack trace is illustrated as being interpreted based on a predetermined stack layout used by a compiler in the control device.

[0018] For example, the data "80001040" at the top address "80001020" of the call stack indicates the start address of the caller's stack frame, and the area from "80001020" to just before "80001040" is the stack frame of the function for which the alarm occurred. Address "80001024" also stores the value of the link register, i.e., the program counter to which processing returns when the function finishes executing. The stack frame also includes areas to store registers to be used by the function, padding to make the stack size a multiple of 16, and an area to store arguments when the function is called.

[0019] By tracing the caller of the function in this way, the processing context information acquisition unit 12 can acquire the function call history from the entry point to the alarm generation process.

[0020] 4 is a diagram illustrating an example of processing context information acquired by the user assistance device 1 in the first embodiment. After acquiring the data of FIG. 2 generated by the control device, the processing context information acquisition unit 12 adds information about each stack frame, as well as the corresponding source code, assembly code, line number of the function call location in the code, and related documents based on the stack trace of FIG. 3, and provides the resulting information to the prompt creation unit 13 as processing context information.

[0021] Specifically, for example, once specific information about the software is known, the assembly code and source code, as well as the line number of the call location in these, can be identified from the value of the link register saved in each stack frame, i.e., the address of the program counter of the caller.

[0022] 5 is a diagram showing a first example of a prompt created in the first embodiment. The prompt creation unit 13 assigns the role of a software developer of the control device to the generating AI system, and then writes in the prompt an instruction to present an appropriate method for resolving the abnormal condition (alarm) that has occurred based on the specific information (alarm number) of the abnormal condition and related information.

[0023] The prompt includes, for example, a list of functions in the stack trace showing the flow of the process that generated the alarm in order of execution, the source code of each process, assembly code, related documents, etc. as related information.

[0024] 6 is a diagram showing a second example of a prompt created in the first embodiment. In the first example, no conditions are set for the content of the response from the generating AI system. However, there is a possibility that information that should be kept secret for security reasons, such as the address of the call history or the software structure provided to the AI ​​system as related information, may be presented as is.

[0025] In the second example, the prompt generator 13 includes a statement in the prompt (enclosed by a dashed line) requesting that processing context information not be included in the response, thereby preventing confidential information from being presented to the user as a response.

[0026] According to the first embodiment, the user assistance device 1 acquires and inputs not only information (such as an alarm number) identifying the abnormal state of the control device but also processing context information into the generation AI system, and then asks how to resolve the abnormal state. Furthermore, by requesting the generation AI system not to include this context information in the response, the user assistance device 1 can present the user with more appropriate assistance information for resolving the abnormal state of the control device while taking security into consideration.

[0027] 7 is a block diagram showing the functional configuration of the user assistance device 1 according to the second embodiment. In the second embodiment, compared to the first embodiment, the control unit of the user assistance device 1 further includes a response correction unit 17, and accordingly, the functions of the transmission unit 14, the reception unit 15, and the presentation unit 16 are expanded.

[0028] The response correction unit 17 creates a response correction prompt that requests that processing context information be deleted from the response of the generation AI system. Accordingly, the sending unit 14 sends the response correction prompt to the generation AI system, and the receiving unit 15 receives a new response from the generation AI system. The presentation unit 16 then presents the content of the new response to the response correction prompt, from which the processing context information from the generation AI has been deleted, instead of the response to the prompt created by the prompt creation unit 13.

[0029] 8 is a diagram illustrating an example of a response correction prompt created in the second embodiment. The response correction prompt has a role of correcting the resolution method for the created alarm (dashed box A) and an instruction to delete the processing context information provided from the description of the resolution method (dashed box B).

[0030] As a result, it is expected that a new response will be obtained in which the function name, which is a description that matches the processing context information, has been deleted from the transcribed response from the generation AI (dashed frame C).

[0031] In the first embodiment, the prompt creation unit 13 made a request that did not include processing context information in the response, but the processing of the response correction unit 17 may be performed instead of or in addition to this.

[0032] According to the second embodiment, the user assistance device 1 enters processing context information in addition to specific information about the abnormal state in the prompt to obtain an appropriate method of dealing with the abnormal state of the control device from the generating AI system, and further instructs the generating AI system to correct the obtained response by deleting the processing context information. Therefore, the user assistance device 1 can provide the user with appropriate assistance information that takes security, etc. into consideration.

[0033] Here, the machine learning model in the generative AI system used in the above-described embodiment may be fine-tuned using non-public data, etc. For example, in a testing process during software development of a control device, a test may be performed to check whether an intended alarm occurs when a certain situation arises. When an alarm occurs during this test, processing context information is automatically collected by the processing context information acquisition unit 12, and a prompt can be created by the prompt creation unit 13 using this processing context information. Since the reason for the alarm is known to the software designer and test case designer, information combining the prompt and the reason for the alarm can be obtained. This information can be used as data for fine-tuning.

[0034] According to any of the above-described embodiments, the user assistance device 1 acquires not only information identifying the abnormal state of the control device (such as an alarm number) but also processing context information, inputs the acquired information into the generation AI system, and then asks the user about a countermeasure for resolving the abnormal state. Therefore, the user assistance device 1 can obtain a countermeasure based on the processing flow leading up to the abnormal state of the control device as a response, and can present the user with more appropriate assistance information for resolving the abnormal state.

[0035] Furthermore, even if the same alarm is implemented for different abnormal conditions in the software of a control device, the generation AI system can distinguish between them by providing processing context information, allowing the user to obtain appropriate auxiliary information. Furthermore, as a result, it is not necessary to subdivide alarms, which saves resources for alarm-specific information and explanations in software development and reduces the amount of work required to prepare alarms with new names and modify processing to trigger them.

[0036] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to the individual embodiments described above. Various additions, substitutions, modifications, partial deletions, etc. are possible to these embodiments without departing from the gist of the invention or the idea and intent of the present invention derived from the content of the claims and their equivalents. For example, in the above-described embodiments, the order of each operation and the order of each process are shown as examples and are not limited to these. The same applies when numerical values ​​or mathematical expressions are used in the description of the above-described embodiments.

[0037] The following supplementary note is further disclosed regarding the above-described embodiment and modified examples. (Supplementary note 1) A user assistance device (1) that presents a method for dealing with an abnormal state that has occurred in a control device, comprising: an abnormal state information acquisition unit (11) that acquires abnormal state information related to the abnormal state; a processing context information acquisition unit (12) that acquires processing context information in software processing of the control device that has caused the abnormal state; a prompt creation unit (13) that creates a prompt that instructs a generation AI system to generate a method for dealing with the abnormal state based on the abnormal state information and the processing context information; a transmission unit (14) that transmits the prompt to the generation AI system; a reception unit (15) that receives a response from the generation AI; and a presentation unit (16) that presents the content of the response.

[0038] (Supplementary Note 2) The user assistance device (1) described in Supplementary Note 1, wherein the processing context information includes at least one of: specific information of the software that executes the software processing; a program counter, internal information of the processor, a call stack, assembly code, and source code in the software processing; and documentation related to the source code.

[0039] (Supplementary Note 3) The user assistance device (1) according to (Supplementary Note 1) or (Supplementary Note 2), wherein the prompt creation unit (13) includes a statement in the prompt requesting that the processing context information not be included in the response.

[0040] (Appendix 4) A user assistance device (1) described in any of (Appendix 1) to (Appendix 3) comprising a response correction unit (17) that creates a response correction prompt requesting that the processing context information be deleted from the response, wherein the sending unit (14) sends the response correction prompt to the generation AI system, and the presentation unit (16) presents the content of a new response to the response correction prompt from the generation AI in which the processing context information has been deleted, instead of the response to the prompt.

[0041] REFERENCE SIGNS LIST 1 User assistance device 11 Abnormal state information acquisition unit 12 Processing context information acquisition unit 13 Prompt creation unit 14 Transmission unit 15 Reception unit 16 Presentation unit 17 Response correction unit

Claims

1. A user assistance device that presents a method for dealing with an abnormal condition that has occurred in a control device, comprising: an abnormal condition information acquisition unit that acquires abnormal condition information regarding the abnormal condition; a processing context information acquisition unit that acquires processing context information in the software processing of the control device that caused the abnormal condition; a prompt creation unit that creates a prompt that instructs a generation AI system to generate a method for dealing with the abnormal condition based on the abnormal condition information and the processing context information; a transmission unit that transmits the prompt to the generation AI system; a receiving unit that receives a response from the generation AI; and a presentation unit that presents the content of the response.

2. The user assistance device of claim 1, wherein the processing context information includes at least one of the following: specific information about the software executing the software processing; a program counter, internal processor information, a call stack, assembly code, and source code in the software processing; and documentation related to the source code.

3. The user assistance device according to claim 1 or 2, wherein the prompt creation unit includes in the prompt a statement requesting that the processing context information not be included in the response.

4. A user assistance device as described in any one of claims 1 to 3, comprising a response correction unit that creates a response correction prompt requesting that the processing context information be deleted from the response, wherein the sending unit sends the response correction prompt to the generation AI system, and the presentation unit presents, instead of the response to the prompt, the content of a new response to the response correction prompt from the generation AI in which the processing context information has been deleted.

Images