A method and system for developing civil aircraft airborne software based on artificial intelligence

Abstract

The present application belongs to the technical field of civil aircraft airborne software development. The present application provides a method and system for civil aircraft airborne software development based on artificial intelligence. When processing the privacy and security requirements of sensitive data, the disclosed embodiment adds a sensitive data determiner, which can not only protect the security of local sensitive data, but also obtain high-quality services from external large models, and has obvious advantages in terms of security, service level and quality; it adopts a model fine-tuning method, combined with a local knowledge base, to reduce the cost of use; at the same time, it can also obtain high-quality services in the field of localized professional knowledge, with the advantages of low development and deployment costs and high service quality. The provided intelligent agent can cover the entire life cycle of civil aircraft airborne software development, and has the advantages of significantly improving the efficiency, quality and security of civil aircraft software development.

Description

Technical Field

[0001] The disclosed embodiments relate to the technical field of civil aircraft onboard software development, and in particular, to a civil aircraft onboard software development method and system based on artificial intelligence. Background Art

[0002] DO178B / C is an important technical standard for the development of civil aircraft airborne software. It plays an extremely important role in the aviation field. According to the impact of software failures on aircraft safety, it classifies software into five levels: A (most serious) to E (no impact). It also provides airborne software development and verification activities, including requirements, design, coding, integration, testing, configuration management and quality assurance. The large language model technology based on artificial intelligence is an important technology in the field of natural language processing. It uses deep learning technology to process large-scale language data. The goal is to improve the accuracy and efficiency of high-level natural language processing tasks. In recent years, with the rapid development of deep learning technology and the increasing abundance of data resources, large language models have made significant progress in various fields, especially in natural language generation, understanding and reasoning.

[0003] In the traditional airborne software development process, it is necessary to manually write software requirements, design, code, write test cases, conduct software testing, and manually review the data at each stage of software development. The development cycle is long and the verification cost is high. With the development of large language model technology, artificial intelligence has significantly improved software development efficiency, development quality and security in automatic code generation and completion, software defect detection and code quality analysis, software automated testing, requirements and document analysis, etc. Domestic and foreign artificial intelligence large language models have made certain progress in assisting software development. For example, Github's Copilot and Baidu's Comate support programmers in the process of writing code, assisting code prompts, but do not support full life cycle software development. At present, the full life cycle development method combined with artificial intelligence technology in the field of civil aircraft high-security airborne software development is a technical blank.

[0004] Therefore, it is necessary to improve one or more problems existing in the above-mentioned related technical solutions.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] The purpose of the embodiments of the present disclosure is to provide a method and system for developing civil aircraft onboard software based on artificial intelligence, thereby overcoming one or more problems caused by the limitations and defects of related technologies, at least to a certain extent.

[0007] According to a first aspect of an embodiment of the present disclosure, a method for developing civil aircraft airborne software based on artificial intelligence is provided, the method comprising: Providing artificial intelligence services for access requests using an artificial intelligence service platform; wherein the access requests include a first request, a second request, a third request, and a fourth request; Converting software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, and obtaining corresponding first feedback information, completing relevant software requirement auxiliary writing tasks according to the first feedback information, so as to obtain software requirement analysis results and analysis reports thereof, and storing the software requirement analysis results in the software development library; Writing a software design document based on the software requirements analysis result, converting the software design document into the second request to the artificial intelligence service platform, and obtaining corresponding second feedback information, completing the relevant software coding auxiliary writing task according to the second feedback information to obtain the software code, and storing the software code in the software development library; Converting the integration task in the software integration process into the third request to the artificial intelligence service platform, combining the software code to obtain the first version of the software and a software integration report, and storing the software integration report in the software development library; The test tasks in the software testing process are converted into the fourth request to the artificial intelligence service platform, and a test case is generated in combination with the test case template. The civil aircraft onboard software and the test report are obtained according to the test case and the initial version of the software, and the test report is stored in the software development library.

[0008] Furthermore, the method further comprises: Marking high-quality information in the software requirement analysis results, the software code, the software integration report, and the test report to obtain high-quality data, and using the high-quality data to fine-tune the local large model; Generate corresponding software airworthiness compliance documents according to the software development objectives relative to the regulations and standards of civil aircraft software, and use the software airworthiness compliance documents to conduct quality review of the civil aircraft airborne software; A local knowledge base is generated based on the regulations and standards of the civil aircraft software, the software requirement analysis results, the software code, the software integration report and the test report.

[0009] Furthermore, the artificial intelligence service platform includes: A local large model, an external large model and a sensitive data determiner; wherein the sensitive data determiner is used to determine whether the access request contains sensitive information; If included, access the artificial intelligence interface of the local large model to protect relevant data information; If not included, access the artificial intelligence interface of the external large model to obtain higher quality data information.

[0010] Furthermore, the steps of converting the software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, obtaining corresponding first feedback information, and completing the relevant software requirement auxiliary writing task according to the first feedback information to obtain the software requirement analysis results and the analysis report thereof include: Writing the software requirements and the software design data and converting them into the first request; Sending the first request to the artificial intelligence service platform, the artificial intelligence service platform processes the first request to generate the first feedback information, and provides feedback; The software requirement analysis result and the analysis report thereof are obtained according to the first feedback information.

[0011] Furthermore, the software requirements and software ICD are input into the artificial intelligence service platform for instant checking and integrity checking; wherein the instant checking includes writing information according to the software requirements prompts, checking the correctness of the software ICD reference and the correctness of the traceability; the integrity checking includes checking the consistency, accuracy and traceability integrity of the software requirements.

[0012] Furthermore, the step of converting the integration task in the software integration process into the third request to the artificial intelligence service platform, and combining the software code to obtain the first version of the software and the software integration report includes: generating the third request according to the integration task; The artificial intelligence service platform obtains the corresponding software code in the software development library according to the third request, constructs a target code according to the software code, and generates a log file; If the construction is successful, the artificial intelligence service platform integrates the target code into the target machine to obtain the first version of the software, checks the completeness and correctness of the integration process, and generates the software integration report; If the build fails, the log file is read, the cause of the error is analyzed, and the corresponding software code is downloaded from the software development library again for compilation and integration; if the failure exceeds the specified number of times, the software integration report containing the error log is generated.

[0013] Further, the step of converting the test task in the software testing process into the fourth request to the artificial intelligence service platform, generating a test case in combination with the test case template, and obtaining the civil aircraft airborne software and the test report according to the test case and the first version of the software includes: Converting the test task into the fourth request; The artificial intelligence service platform generates the test case according to the fourth request and the software requirement; The test case is run on the initial version of the software to generate the test report and determine whether the test result is passed.

[0014] According to a second aspect of an embodiment of the present disclosure, there is provided a civil aircraft airborne software development system based on artificial intelligence, the system comprising: An artificial intelligence service platform, configured to provide artificial intelligence services for access requests; wherein the access requests include a first request, a second request, a third request, and a fourth request; A software requirement agent, used to convert the software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, and obtain the corresponding first feedback information, complete the relevant software requirement auxiliary writing task according to the first feedback information, so as to obtain the software requirement analysis results and analysis report thereof, and store the software requirement analysis results in the software development library; A software coding agent, configured to write a software design document based on the software requirement analysis result, convert the software design document into the second request to the artificial intelligence service platform, obtain corresponding second feedback information, complete related software coding auxiliary writing tasks according to the second feedback information to obtain software code, and store the software code in a software development library; A software integration agent, used to convert the integration task in the software integration process into the third request to the artificial intelligence service platform, combine the software code to obtain the first version of the software and the software integration report, and store the software integration report in the software development library; The software testing agent is used to convert the test tasks in the software testing process into the fourth request to the artificial intelligence service platform, generate test cases in combination with the test case template, obtain the civil aircraft onboard software and test reports based on the test cases and the initial version of the software, and store the test reports in the software development library.

[0015] Furthermore, the system also includes: A fine-tuning training module, used for marking high-quality information in the software requirement analysis results, the software code, the software integration report and the test report to obtain high-quality data, and using the high-quality data to perform fine-tuning training on the local large model; A software airworthiness agent is used to generate corresponding software airworthiness compliance documents according to the software development goals relative to the regulations and standards of civil aircraft software, and to conduct quality review of the civil aircraft airborne software using the software airworthiness compliance documents; The local knowledge base is used to store the regulations and standards of the civil aircraft software, the software requirement analysis results, the software code, the software integration report and the test report locally.

[0016] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects: In the embodiments of the present disclosure, through the above-mentioned civil aircraft airborne software development method and system based on artificial intelligence, on the one hand, when processing the privacy and security requirements of sensitive data, a sensitive data determiner is added, which can not only protect the security of local sensitive data, but also obtain high-quality services of external large models, and has obvious advantages in terms of security, service level and quality; on the other hand, the model fine-tuning method is adopted, combined with the local knowledge base, to reduce the cost of use; at the same time, high-quality services in the field of localized professional knowledge can also be obtained, with the advantages of low development and deployment costs and high service quality. The provided intelligent agent can cover the entire life cycle of civil aircraft airborne software development, and has the advantages of significantly improving the efficiency, development quality and security of civil aircraft software development.

[0017] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0019] Figure 1 A step diagram showing a method for developing civil aircraft airborne software based on artificial intelligence in an exemplary embodiment of the present disclosure; Figure 2 A schematic diagram showing an artificial intelligence-based civil aircraft airborne software development system in an exemplary embodiment of the present disclosure; Figure 3 A specific flow chart of a civil aircraft onboard software development system based on artificial intelligence in an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0020] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0021] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0022] This example implementation first provides a method for developing civil aircraft airborne software based on artificial intelligence. The method can be applied to a terminal device, such as a mobile phone, a personal digital assistant, a laptop, a tablet computer, a smart watch, or other mobile terminal [flexibly adjusted according to specific circumstances, such as the terminal device is a server, etc.]. Figure 1 As shown in , the method may include the following steps: Step S101: using an artificial intelligence service platform to provide artificial intelligence services for access requests; wherein the access requests include a first request, a second request, a third request, and a fourth request; Step S102: converting the software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, and obtaining corresponding first feedback information, completing the relevant software requirement auxiliary writing task according to the first feedback information, so as to obtain the software requirement analysis results and analysis report thereof, and storing the software requirement analysis results in the software development library; Step S103: writing a software design document based on the software requirement analysis result, converting the software design document into the second request to the artificial intelligence service platform, and obtaining corresponding second feedback information, completing the relevant software coding auxiliary writing task according to the second feedback information to obtain the software code, and storing the software code in the software development library; Step S104: converting the integration task in the software integration process into the third request to the artificial intelligence service platform, combining the software code to obtain the first version of the software and a software integration report, and storing the software integration report in the software development library; Step S105: convert the test task in the software testing process into the fourth request to the artificial intelligence service platform, generate a test case in combination with the test case template, obtain the civil aircraft onboard software and the test report according to the test case and the first version of the software, and store the test report in the software development library.

[0023] Through the above-mentioned civil aircraft airborne software development method and system based on artificial intelligence, on the one hand, when dealing with the privacy and security requirements of sensitive data, a sensitive data determiner is added, which can not only protect the security of local sensitive data, but also obtain high-quality services of external large models, and has obvious advantages in terms of security, service level and quality; on the other hand, the model fine-tuning method is adopted, combined with the local knowledge base, to reduce the cost of use; at the same time, high-quality services in the field of localized professional knowledge can also be obtained, with the advantages of low development and deployment costs and high service quality. The provided intelligent agent can cover the entire life cycle of civil aircraft airborne software development, and has the advantages of significantly improving the efficiency, development quality and security of civil aircraft software development.

[0024] Next, we will refer to Figure 1 Each step of the above method in this example implementation is described in more detail.

[0025] In step S101, select available high-quality large models to deploy locally and configure API access interfaces, where the number of deployed large models can be one or more; select high-quality accessible external large models and configure API access interfaces; select locally deployable knowledge base training tools, deploy them locally, and configure API access interfaces. Based on the definition of sensitive data, all artificial intelligence requests are classified and identified as sensitive data requests or non-sensitive data requests using rule-based combined with machine learning and other methods.

[0026] In one embodiment, a sensitive data determiner is developed to filter all access requests and identify access requests as knowledge-sensitive information and intellectual property-non-sensitive information. For intellectual property-sensitive information, the artificial intelligence interface of the locally deployed large model is accessed to protect the relevant data information. For information that is not sensitive to intellectual property, the artificial intelligence interface of the external large model is accessed to obtain higher quality data information. The locally deployed large model is optimized through fine-tuning or prompt word engineering to obtain more accurate or higher quality output. The locally deployed or externally accessed artificial intelligence interface provides a service access interface in the form of an API. The tool chain in the development process of civil aircraft onboard software accesses the services provided by the artificial intelligence by calling the API.

[0027] In step S102, based on the software requirement compilation tool, a software requirement agent is developed by developing a plug-in or by calling the plug-in. The main function of the agent is to convert the tasks in the software requirement compilation process into requests to the artificial intelligence service platform, obtain corresponding feedback information, and complete the relevant software requirement auxiliary writing tasks according to the feedback information. The requirement development includes high-level requirement development and low-level requirement development (software design).

[0028] In one embodiment, software requirements and software ICD are input as knowledge into the artificial intelligence large language model, which can support instant inspection and integrity inspection. In instant inspection, artificial intelligence will automatically prompt to write information according to software requirements, check the correctness of ICD references, the correctness of traceability, etc.; integrity inspection includes the consistency of requirements, the integrity of traceability, the accuracy of requirements, etc.

[0029] In step S103, a software coding agent is developed based on a software code compilation tool by developing a plug-in or by calling a method. Its main function is to convert the tasks in the software coding process into requests to the artificial intelligence service platform, obtain corresponding feedback information, and complete related software coding auxiliary writing tasks based on the feedback information.

[0030] In one embodiment, software design documents are manually compiled according to requirements. The design description is in the form of functions to clearly define the input, output, and implemented logical functions. Artificial intelligence is used to assist in generating code corresponding to the design description, and the generated code is manually confirmed or revised.

[0031] It supports checking the correctness of the code through artificial intelligence, and also supports artificial intelligence calling third-party coding rule checking tools to check the generated code and generate code analysis reports.

[0032] Supports the generation of a bidirectional traceability matrix between requirements, design, and code through artificial intelligence.

[0033] Artificial intelligence automatically modifies the generated code based on the problems reported by the coding rule checking tool until the coding rules are met, or confirms it manually and adds explanations.

[0034] In step S104, a software integration agent is developed based on a software integration tool by developing a plug-in or by calling a software integration agent. The main function of the agent is to convert the tasks in the software integration process into requests to the artificial intelligence service platform, obtain corresponding feedback information, and complete the relevant software integration tasks according to the feedback information.

[0035] In one embodiment, artificial intelligence automatically downloads the code or configuration files required for integration from the development library or controlled library according to instructions, automatically calls the compilation command, builds the generated target code, and checks the generated log.

[0036] When the build is successful, the artificial intelligence will call the loading instruction to automatically integrate the target code into the target machine, run the integration process test cases at the same time, check the completeness and correctness of the integration process, and generate an integration report. After the integration report is manually confirmed, the integration process ends.

[0037] When a build fails, AI reads the log file, analyzes the possible causes of the error, and tries to download the files required for compilation from the library again for compilation integration. If the number of attempts exceeds the specified number, an integration report containing the error log is generated for manual inspection.

[0038] In step S105, based on the software integration tool, a software integration agent is developed by developing a plug-in or by calling the agent. The main function of the agent is to convert the tasks in the software testing process into requests to the artificial intelligence service platform, obtain corresponding feedback information, and complete the relevant software testing tasks according to the feedback information.

[0039] In one embodiment, according to the test case template, artificial intelligence generates corresponding test cases based on software requirements or design, and is manually confirmed or revised.

[0040] Supports the generation of test case analysis reports and test case tracking reports through artificial intelligence, which can be confirmed manually.

[0041] Supports artificial intelligence to automatically call the integration process according to instructions, generate target code, and load it into the target machine.

[0042] Supports automatic downloading of test cases, loading them into the test platform, automatically running the test cases, and automatically judging whether the test results are passed.

[0043] Supports the generation of test reports and manual confirmation of test reports.

[0044] The method also includes: developing a software airworthiness intelligent entity based on application or by calling according to DO178B / C software airworthiness standards, the main function of which is to generate a software airworthiness compliance document based on the development goals of airborne software in DO178B / C and the services provided by calling the local large model according to the software development library / controlled library.

[0045] Based on the software development library / controlled library, mark high-quality software requirements, designs, codes, test cases, test results and other data, fine-tune and train the local large model, and optimize the local large model.

[0046] Based on the regulations and standards related to civil aircraft software and the problem reports and design revision data accumulated by the projects, a local knowledge base is generated, so that the system can acquire knowledge related to the development of airborne software.

[0047] In one embodiment, artificial intelligence is used to support knowledge extraction from the content of regulatory standards and problem reports and design data revisions of software projects, thereby identifying a knowledge base of software problems.

[0048] Based on the identified knowledge base, complete the software life cycle data analysis, identify potential problems, generate an analysis report, and confirm it manually.

[0049] Supports analysis of software life cycle data according to different software level requirements of DO178B / C.

[0050] If the life cycle data is complete, the corresponding airworthiness compliance document can be generated and confirmed manually.

[0051] If the life cycle data is incomplete or has problems, a corresponding analysis report will be generated and confirmed manually.

[0052] In a specific embodiment, by deploying and optimizing the local big model, combining the external big model interface, developing the artificial intelligence service access interface, the tool chain in the software development process, such as requirements and design compilation, code writing, continuous integration, test verification and other tools, through customized development, integrated artificial intelligence access interface, users can obtain the services provided by artificial intelligence by using the functions integrated by the tools in the software development process. The steps are as follows: Build a local artificial intelligence service platform, local large models, knowledge base, external large model access interface, develop sensitive data determiners, develop software requirements, coding, integration, testing, airworthiness and other intelligent agents, where the software agents are integrated into the software tool chain, or can be called by the software tool chain, and provide artificial intelligence services.

[0053] Based on the high-quality data marked by the software development library / controlled library, the local large model is fine-tuned to optimize the service quality of the local large model in the field of airborne software development.

[0054] Extract and generate the knowledge base based on software development library / controlled library (i.e. software controlled library), regulations and standards, and problem reports.

[0055] Software requirements and software design data are compiled through software requirement agents, and artificial intelligence agent-related services are automatically executed based on user requests.

[0056] Software coding agents compile software codes to automatically execute artificial intelligence agent-related services based on user requests.

[0057] Carry out software integration tasks through software integration agents, based on user requests or automatically execute artificial intelligence agent-related services.

[0058] Carry out software testing tasks through software testing agents, based on user requests or automatically execute artificial intelligence agent-related services.

[0059] The software airworthiness documentation is compiled through the software airworthiness agent, and the artificial intelligence agent-related services are automatically executed based on user requests.

[0060] In a specific embodiment, assuming that a software that complies with DO178B / CA level needs to be developed, the implementation process of the present invention is as follows: 1) Build a local artificial intelligence service platform In this example, the external big model adopts Baidu's Qianfan big model, the locally deployed model adopts the ollama model running framework, the local big model adopts llama3, and the big model framework application integration development framework adopts langchain. The Qianfan big model needs to be accessed through the Internet by configuring the API key, and the llama3 big model can be accessed locally based on the open source ollama framework. The artificial intelligence service platform provides web access to platform services; the web service platform mainly supports knowledge base creation, agent configuration, intelligent question and answer, project configuration and other functions; knowledge base creation is mainly to use the langchain framework to combine DO178B / C and other regulatory standard documents with embedded models, vector databases, etc. to build a DO178B / C knowledge retrieval system; agent configuration is mainly to configure the agents used in the current project for requirements, coding, integration, and testing tasks. Intelligent Q&A provides questions and answers combined with knowledge bases, such as questions about the DO178B / C standards. In this example, a "cockpit display software" project was created, and the software requirements, coding, integration, testing, and airworthiness agents used were configured. Different software projects may need to configure different agents due to different development requirements. For example, different software uses different compilers, so the required integrated agents are also different.

[0061] 2) Develop a sensitive data determiner The sensitive data determiner is mainly used to determine service requests for demand and code data. It adopts a two-level classification method of rules plus machine learning. The first level classifies the request data through document identification, model, performance indicators, key algorithms and other tags. In the second level, a sensitive data classifier is trained through machine learning, and the sensitive data is finally classified through this classifier. In this example, it is assumed that the rule is that if the document classification is confidential, the service provided by the locally deployed large model is requested, and if the document classification is public, the next step of classification is carried out through the sensitive data classifier. By marking some sensitive texts, such as "the operating performance of XX model is XX", training is carried out, and the trained model is used as a determiner. According to the final judgment result of the determiner (sensitive / non-sensitive), the service request method is selected. Sensitive data requests the local large model, and non-sensitive data requests the external large model.

[0062] 3) Developing software requirements agents In this example, the software requirement agent of the "cockpit display software" project is developed and configured on the artificial intelligence service platform through the web. The software requirements use the IBM Doors tool. Based on Doors, a requirement compilation agent call plug-in is developed. The plug-in can obtain the services provided by the software requirement agent of this project in the form of web API, including standard compliance check, traceability check, analysis report generation and traceability matrix generation. In this example, if a requirement is written in Doors, "HUD should display the airspeed digital symbol in the center of the left side according to ipAirspeed", the document classification is public, and the standard compliance check service of the requirement agent is selected through the Doors plug-in, and the requirement agent is called through the web API. Assuming that one of the standards in the requirement checklist is "the requirement should contain the keyword 'should'", the requirement agent will generate the following prompt words according to this requirement of the requirement checklist: "You are a software requirement review expert and need to conduct a review based on the requirement standards. One of the requirement standards is that the requirement should contain the keyword "should". Please judge the requirement described below to determine whether it meets the standard." Then send the requirement "The HUD should display the airspeed digital symbol in the center of the left side according to ipAirspeed." After calling the large model, the large model generates "The requirement description is "The HUD should display the airspeed digital symbol in the center of the left side according to ipAirspeed", which contains the keyword "should", so it meets the requirement standards." Based on the returned information, the intelligent agent determines that the requirement meets the requirement standard and returns the result to the user in the form of an analysis report.

[0063] 4) Develop software-coded agents In this example, the software coding agent of the "cockpit display software" project is developed and configured on the artificial intelligence service platform through the web. A coding agent calling plug-in is developed based on vscode. The plug-in can import the software design into the code in the form of comments, generate code in an auxiliary way, and support calling the testBed code rule checking tool to check whether the code complies with the coding rules. It supports the generation of code analysis reports for user confirmation. In this example, the low-level software requirement is "the software should call DrawNumber (ipAirspeed) to generate drawing instructions through the DrawAirspeed (position, ipAirspeed) function"; this requirement has been imported into the code as a comment. Click the Generate Code button in vscode, and the plug-in will generate the drawing instructions through the web The API method is to call the coding agent, which will generate the following prompt words "The low-level software requirement is "The software should generate drawing instructions by calling DrawNumber (ipAirspeed) through the DrawAirspeed (position, ipAirspeed) function", please generate the corresponding C code for me", and after calling the large model, the large model generates "The following is a sample C code that meets this requirement: ...", and the agent returns the result to VScode to call the plug-in, and VScode calls the plug-in to insert the code into the file.

[0064] 5) Develop software integrated agents In this example, the software integration agent of the "cockpit display software" project is developed and configured on the artificial intelligence service platform through the web. The software integration is carried out through the Bamboo tool. The software integration agent task is developed based on Bamboo. When the user's component task is started, the agent will automatically call the integration task, download the design data from the software development library / controlled library, call the software tool chain, compile and integrate to generate the target code, and check the compilation log according to the integration checklist, and generate an integration report for the user to confirm. In this example, the call for software integration is integrated in the vscode plug-in. Select the integration menu in vscode, and the plug-in will call the software integration agent in the form of webAPI. The agent downloads the source code by calling the bamboo tool, compiles and integrates, generates a compilation log, and sends the prompt word along with the log to the large model. The prompt word is as follows "The following is the error message encountered during the compilation process, please provide the corresponding solution". The agent summarizes the information generated by the large model, generates a software integration report, and returns it to the user.

[0065] 6) Developing software testing agents In this example, a software test agent for the "cockpit display software" project is developed and configured on the artificial intelligence service platform through the web. The test cases are developed based on Excel. A test agent calling plug-in is developed based on vscode to support the generation of test cases based on requirements or design assistance, and to generate test case analysis reports. By calling the software integration agent, the software target code is generated, the software loader is called, and it is automatically loaded into the target machine. At the same time, the test platform loaded with the test case automatically executes the test case and obtains the result. After comparing with the expected result, a test report is generated for user confirmation. In this example, the requirement "HUD should be based on ipAirsp eed displays the airspeed digital symbol in the center on the left side", click Generate Test Case, the test agent will be called through webAPI, the agent uploads the test case template, and generates the following prompt words "This is a test case template, please write a test case according to the template, please write a test case according to the requirement "HUD should display the airspeed digital symbol in the center on the left side according to ipAirspeed", the large model returns "According to the provided requirement "HUD should display the airspeed digital symbol in the center on the left side according to ipAirspeed", the following are test cases that conform to the template:......", the test agent generates the corresponding Excel test case based on the returned content.

[0066] 7) Developing software airworthiness agents In this example, the software airworthiness agent of the "cockpit display software" project is developed and configured on the artificial intelligence service platform through the web; the software airworthiness compliance document is compiled based on Word, and an agent plug-in is developed based on vscode. According to the software development goals of DO178B / C, the data information in the software controlled library is automatically captured, and the software airworthiness compliance document is generated for user confirmation; in this example, the software airworthiness complies with the standard DO178B, and an airworthiness agent call plug-in is developed in vscode. The plug-in will call the software airworthiness agent plug-in through webAPI. The agent will select the DO178B file and the software life cycle data list and send them to the big model, and generate the following prompt words "Please give me a corresponding declaration of compliance document based on the requirements of the DO178B document and the software life cycle data list". The big model generates "The following is an example of a declaration of compliance document corresponding to your standard requirements:..." The agent returns the text, generates the corresponding word file according to the software compliance document template, and returns it to the user.

[0067] 8) Fine-tune training to optimize the local large model Screen the data in software development libraries and controlled libraries, mark high-quality software requirements, designs, codes, test cases and other data, and fine-tune, train and optimize local large models.

[0068] In this example, the life cycle data of the software project that has passed the airworthiness review is marked with high-quality data in JSON or CSV format, such as the correspondence between requirements and test cases. Then, the pre-trained model llama3 is loaded through frameworks such as torchtransformers datasets to preprocess and batch the data, and fine-tune the parameters, such as the learning rate of 1e5. The model is fine-tuned using Trainer, and finally the fine-tuned model is saved and released after optimization.

[0069] 9) Generate local knowledge base Based on tools such as LangChain, the knowledge of airworthiness regulations and software problem report libraries is sliced, vectorized and stored, and then combined with RAG (retrieval enhanced generation technology) to maintain the timeliness of the knowledge base; in this example, the knowledge data of regulations and standards such as DO178B / C are pre-processed (such as text splitting) through the langchain framework, and then converted into embedded vectors and stored in the vector database created by FAISS to provide query functions; in this example, if you ask "What are the requirements for coding standards in DO178C?" through the artificial intelligence service platform, the local big model will answer the following based on the knowledge base: "The coding standard requirements of DO-178C mainly cover the following aspects to ensure that the software complies with aviation safety standards and maintains high quality and low error rate during the development process:..."

[0070] It should be noted that, although the steps of the method in the present disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step for execution, and / or one step may be decomposed into multiple steps for execution, etc. In addition, it is also easy to understand that these steps may be, for example, executed synchronously or asynchronously in multiple modules / processes / threads.

[0071] Furthermore, in this example implementation, a civil aircraft airborne software development system based on artificial intelligence is also provided. Figure 2 As shown in , the system may include an artificial intelligence service platform, a software requirement agent, a software coding agent, a software integration agent, and a software testing agent. Among them: An artificial intelligence service platform, configured to provide artificial intelligence services for access requests; wherein the access requests include a first request, a second request, a third request, and a fourth request; A software requirement agent, used to convert the software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, and obtain the corresponding first feedback information, complete the relevant software requirement auxiliary writing task according to the first feedback information, so as to obtain the software requirement analysis results and analysis report thereof, and store the software requirement analysis results in the software development library; A software coding agent, configured to write a software design document based on the software requirement analysis result, convert the software design document into the second request to the artificial intelligence service platform, obtain corresponding second feedback information, complete related software coding auxiliary writing tasks according to the second feedback information to obtain software code, and store the software code in a software development library; A software integration agent, used to convert the integration task in the software integration process into the third request to the artificial intelligence service platform, combine the software code to obtain the first version of the software and the software integration report, and store the software integration report in the software development library; The software testing agent is used to convert the test tasks in the software testing process into the fourth request to the artificial intelligence service platform, generate test cases in combination with the test case template, obtain the civil aircraft onboard software and test reports based on the test cases and the initial version of the software, and store the test reports in the software development library.

[0072] The system also includes: A fine-tuning training module, used for marking high-quality information in the software requirement analysis results, the software code, the software integration report and the test report to obtain high-quality data, and using the high-quality data to perform fine-tuning training on the local large model; A software airworthiness agent is used to generate corresponding software airworthiness compliance documents according to the software development goals relative to the regulations and standards of civil aircraft software, and to conduct quality review of the civil aircraft airborne software using the software airworthiness compliance documents; The local knowledge base is used to store the regulations and standards of the civil aircraft software, the software requirement analysis results, the software code, the software integration report and the test report locally.

[0073] like Figure 3 As shown, this is a schematic diagram of the specific process of the civil aircraft airborne software development system based on artificial intelligence.

[0074] Through the above-mentioned civil aircraft airborne software development method and system based on artificial intelligence, on the one hand, when dealing with the privacy and security requirements of sensitive data, a sensitive data determiner is added, which can not only protect the security of local sensitive data, but also obtain high-quality services of external large models, and has obvious advantages in terms of security, service level and quality; on the other hand, the model fine-tuning method is adopted, combined with the local knowledge base, to reduce the cost of use; at the same time, high-quality services in the field of localized professional knowledge can also be obtained, with the advantages of low development and deployment costs and high service quality. The provided intelligent agent can cover the entire life cycle of civil aircraft airborne software development, and has the advantages of significantly improving the efficiency, development quality and security of civil aircraft software development.

[0075] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0076] It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of a module or unit described above can be further divided into multiple modules or units for concretization. The components displayed as modules or units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. Those of ordinary skill in the art can understand and implement it without paying creative work.

[0077] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims.

Claims

1. A method for developing civil aircraft airborne software based on artificial intelligence, characterized in that: The method includes: Providing artificial intelligence services for access requests using an artificial intelligence service platform; wherein the access requests include a first request, a second request, a third request, and a fourth request; Converting software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, and obtaining corresponding first feedback information, completing relevant software requirement auxiliary writing tasks according to the first feedback information, so as to obtain software requirement analysis results and analysis reports thereof, and storing the software requirement analysis results in the software development library; Writing a software design document based on the software requirements analysis result, converting the software design document into the second request to the artificial intelligence service platform, and obtaining corresponding second feedback information, completing the relevant software coding auxiliary writing task according to the second feedback information to obtain the software code, and storing the software code in the software development library; Converting the integration task in the software integration process into the third request to the artificial intelligence service platform, combining the software code to obtain the first version of the software and a software integration report, and storing the software integration report in the software development library; The test tasks in the software testing process are converted into the fourth request to the artificial intelligence service platform, and a test case is generated in combination with the test case template. The civil aircraft onboard software and the test report are obtained according to the test case and the initial version of the software, and the test report is stored in the software development library.

2. The method for developing civil aircraft airborne software based on artificial intelligence according to claim 1, characterized in that: The method further includes: Marking high-quality information in the software requirement analysis results, the software code, the software integration report, and the test report to obtain high-quality data, and using the high-quality data to fine-tune the local large model; Generate corresponding software airworthiness compliance documents according to the software development objectives relative to the regulations and standards of civil aircraft software, and use the software airworthiness compliance documents to conduct quality review of the civil aircraft airborne software; A local knowledge base is generated based on the regulations and standards of the civil aircraft software, the software requirement analysis results, the software code, the software integration report and the test report.

3. The method for developing civil aircraft airborne software based on artificial intelligence according to claim 2, characterized in that: The artificial intelligence service platform includes: A local large model, an external large model and a sensitive data determiner; wherein the sensitive data determiner is used to determine whether the access request contains sensitive information; If included, access the artificial intelligence interface of the local large model to protect relevant data information; If not included, access the artificial intelligence interface of the external large model to obtain higher quality data information.

4. The method for developing civil aircraft airborne software based on artificial intelligence according to claim 3, characterized in that: The steps of converting the software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, obtaining corresponding first feedback information, and completing the relevant software requirement auxiliary writing task according to the first feedback information to obtain the software requirement analysis results and the analysis report thereof include: Writing the software requirements and the software design data and converting them into the first request; Sending the first request to the artificial intelligence service platform, the artificial intelligence service platform processes the first request to generate the first feedback information, and provides feedback; The software requirement analysis result and the analysis report thereof are obtained according to the first feedback information.

5. The method for developing civil aircraft airborne software based on artificial intelligence according to claim 4, characterized in that: The software requirements and software ICD are input into the artificial intelligence service platform for instant check and integrity check; wherein the instant check includes writing information according to the software requirement prompts, checking the correctness of the software ICD reference and the correctness of the traceability; the integrity check includes checking the consistency, accuracy and traceability integrity of the software requirements.

6. The method for developing civil aircraft airborne software based on artificial intelligence according to claim 4, characterized in that: The step of converting the integration task in the software integration process into the third request to the artificial intelligence service platform, combining the software code to obtain the first version of the software and the software integration report includes: generating the third request according to the integration task; The artificial intelligence service platform obtains the corresponding software code in the software development library according to the third request, constructs a target code according to the software code, and generates a log file; If the construction is successful, the artificial intelligence service platform integrates the target code into the target machine to obtain the first version of the software, checks the completeness and correctness of the integration process, and generates the software integration report; If the build fails, the log file is read, the cause of the error is analyzed, and the corresponding software code is downloaded from the software development library again for compilation and integration; if the failure exceeds the specified number of times, the software integration report containing the error log is generated.

7. The method for developing civil aircraft airborne software based on artificial intelligence according to claim 4, characterized in that: The step of converting the test task in the software testing process into the fourth request to the artificial intelligence service platform, generating a test case in combination with the test case template, and obtaining the civil aircraft airborne software and the test report according to the test case and the first version of the software includes: Converting the test task into the fourth request; The artificial intelligence service platform generates the test case according to the fourth request and the software requirement; The test case is run on the initial version of the software to generate the test report and determine whether the test result is passed.

8. An artificial intelligence-based civil aircraft airborne software development system, characterized in that: The system includes: An artificial intelligence service platform, configured to provide artificial intelligence services for access requests; wherein the access requests include a first request, a second request, a third request, and a fourth request; A software requirement agent, used to convert the software requirements and software design data in the software compilation process into the first request to the artificial intelligence service platform, and obtain the corresponding first feedback information, complete the relevant software requirement auxiliary writing task according to the first feedback information, so as to obtain the software requirement analysis results and analysis report thereof, and store the software requirement analysis results in the software development library; A software coding agent, configured to write a software design document based on the software requirement analysis result, convert the software design document into the second request to the artificial intelligence service platform, obtain corresponding second feedback information, complete related software coding auxiliary writing tasks according to the second feedback information to obtain software code, and store the software code in a software development library; A software integration agent, used to convert the integration task in the software integration process into the third request to the artificial intelligence service platform, combine the software code to obtain the first version of the software and the software integration report, and store the software integration report in the software development library; The software testing agent is used to convert the test tasks in the software testing process into the fourth request to the artificial intelligence service platform, generate test cases in combination with the test case template, obtain the civil aircraft onboard software and test reports based on the test cases and the initial version of the software, and store the test reports in the software development library.

9. The civil aircraft airborne software development system based on artificial intelligence according to claim 8 is characterized in that: The system also includes: A fine-tuning training module, used for marking high-quality information in the software requirement analysis results, the software code, the software integration report and the test report to obtain high-quality data, and using the high-quality data to perform fine-tuning training on the local large model; A software airworthiness agent is used to generate corresponding software airworthiness compliance documents according to the software development goals relative to the regulations and standards of civil aircraft software, and to conduct quality review of the civil aircraft airborne software using the software airworthiness compliance documents; The local knowledge base is used to store the regulations and standards of the civil aircraft software, the software requirement analysis results, the software code, the software integration report and the test report locally.

Images