An automatic driving standard processing method

Abstract

The application provides an automatic driving standard processing method, which comprises the following steps: obtaining at least one automatic driving function and at least one market to be put on line of a to-be-uploaded item input by a user; sending each automatic driving function and each market to be put on line to a large language model, so that the large language model matches at least one target automatic driving standard suitable for each automatic driving function applied in each market to be put on line in a database, wherein the database comprises automatic driving standards suitable for all candidate markets, and the market to be put on line is one of all candidate markets; and receiving a focus table corresponding to each target automatic driving standard output by the large language model according to a preset structured table, wherein the focus table is used to indicate information that needs to be focused on by a technical personnel.

Description

Technical Field

[0001] This application relates to the field of information processing technology, and in particular to a standard processing method for autonomous driving. Background Technology

[0002] Before a car can be launched on the market, automakers need to review it according to the market's regulatory standards. Only cars that meet these standards can be released to the market. Currently, automakers typically obtain regulatory standards manually by managing spreadsheets. They search for relevant regulations on various official websites, compile these standards into tables, and manually identify each record in each regulatory document, as well as the functional labels constrained by that record. Different technicians may categorize the functional labels of the same regulation inconsistently, affecting the efficiency of vehicle testing. Furthermore, manually downloading and reading each regulation is inefficient, and humans cannot keep track of updates, leading to the use of outdated standards, impacting launch times. Human operation is also prone to errors, resulting in omissions of regulations.

[0003] Alternatively, one could subscribe to a commercial regulatory database. The advantage of subscribing to such databases is that the data source is authoritative and updated more promptly; large contract manufacturers and multinational Tier 1 suppliers typically use this approach. However, in practice, regulations are only differentiated by system (chassis, powertrain, safety, environmental, etc.) and cannot be further subdivided into functional tags for intelligent driving. Manual judgment is still required to determine which function of the car is affected by the regulations. Furthermore, while providing regulatory search capabilities, manual downloading of regulatory standard documents is still necessary in actual application, requiring engineers to manually determine which clauses the car must meet based on project parameters. Engineers need to manually download PDFs, read them page by page to extract technical requirements, and the subscription fee is higher, placing a heavy burden on small and medium-sized suppliers.

[0004] Alternatively, companies can build their own regulatory databases. However, database construction still relies on manual data entry, has a long initial construction period, and regulatory standard documents are generally stored in portable document format (PDF). Searching can only find documents, not specific regulatory entries. System maintenance requires dedicated technical personnel to continuously invest in the project, which is costly. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide at least one method for processing autonomous driving standards. By combining a large language model, this method determines the information related to autonomous driving functions in the target autonomous driving standards that technicians need to focus on, based on at least one autonomous driving function possessed by the project to be launched and at least one market to be launched. This solves the technical problem in the prior art that relies on manual processing, which makes it impossible to accurately and comprehensively determine the regulatory standards that the vehicle to be launched needs to comply with in the market to be launched. This achieves the technical effect of improving the accuracy of querying regulatory standards and reducing the cost of regulatory review.

[0006] This application mainly includes the following aspects: In a first aspect, embodiments of this application provide an autonomous driving standard processing method, the method comprising: acquiring at least one autonomous driving function and at least one market for a project to be launched, input by a user; sending each autonomous driving function and each market to be launched to a large language model, so that the large language model matches in a database at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched, wherein the database includes autonomous driving standards applicable to all candidate markets, and the market to be launched is one of all candidate markets; receiving a attention table corresponding to each target autonomous driving standard output by the large language model according to a preset structured table, the attention table being used to indicate information of the target autonomous driving standard that needs to be paid attention to by technical personnel.

[0007] Optionally, the database is constructed as follows: From the preset data sources corresponding to all regulatory systems involved in all candidate markets, an autonomous driving standard set is determined. The autonomous driving standard set includes standard texts corresponding to multiple autonomous driving standards covered by all regulatory systems. For each autonomous driving standard in the autonomous driving standard set, the standard text of the autonomous driving standard is sent to the large language model. The large language model parses the standard text according to multiple preset structured fields to obtain the standard structured data corresponding to the autonomous driving standard, and decomposes the technical structure of the standard text. The technical structure is used to describe the technical requirements of each level of the standard text. For each autonomous driving standard in the autonomous driving standard set, multiple preset classification labels are sent to the large language model. The large language model determines at least one target classification label that matches the autonomous driving standard from the multiple preset classification labels. The standard text, the standard structured data, the technical structure, and the at least one target classification label corresponding to each autonomous driving standard are detected to form the constructed database.

[0008] Optionally, the set of autonomous driving standards is determined as follows: All regulatory systems involved in all candidate markets are identified, and each regulatory system is sent to the large language model to receive multiple sets of search keywords generated by the large language model for each regulatory system. These multiple sets of search keywords include various technical fields involved in autonomous driving technology. For each regulatory system, a search is performed in a preset data source corresponding to that regulatory system according to the multiple sets of search keywords to find at least one candidate standard from the preset data source corresponding to that regulatory system. Candidate standards that meet the similarity criteria are identified from all candidate standards found under all regulatory systems and merged. The merged candidate similarity and each candidate standard that does not meet the similarity criteria are used as multiple autonomous driving standards, so that the standard texts corresponding to the multiple autonomous driving standards are combined to form the set of autonomous driving standards.

[0009] Optionally, the technical structure of the standard text can be decomposed in the following way: identifying the chapter name of each level of the standard text; for each minimum level of the standard text, sending the text content of the minimum level to the large language model to obtain the semantic description parsed by the large language model for the text content of the minimum level, the semantic description including functional description or technical verification indicators; and determining the chapter name of each level of the standard text and the semantic description corresponding to each minimum level as the technical structure of the standard text.

[0010] Optionally, the method further includes: for each minimum level of the standard text, sending at least one preset supplementary field to the large language model, so that the large language model parses the text content of the minimum level according to at least one preset supplementary field to obtain the text content corresponding to each preset supplementary field; and adding the text content corresponding to each minimum level of the standard text in each preset supplementary field to the technical structure of the standard text.

[0011] Optionally, the method further includes: detecting each autonomous driving standard in the database according to a preset update cycle, and determining the update information of the database, so as to update the database through the update information, wherein the update information includes at least one of the following: newly added autonomous driving standards not involved in the database, revisions to existing autonomous driving standards in the database, and error content corresponding to existing autonomous driving standards in the database; and pushing the update information according to a preset push cycle to indicate the information changes in the database.

[0012] Optionally, at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched is matched by the following method: identifying at least one first autonomous driving standard matching each market to be launched from all autonomous driving standards in the database; determining at least one second autonomous driving standard applicable to the at least one autonomous driving function from the at least one first autonomous driving standard according to at least one target classification label corresponding to each first autonomous driving standard; and determining at least one target autonomous driving standard from each second autonomous driving standard according to the preset priority corresponding to each second autonomous driving standard.

[0013] Optionally, the attention table corresponding to each target autonomous driving standard is determined by the following method: for each target autonomous driving standard, the preset structured table is filled with the information corresponding to the target autonomous driving standard involved in the database to construct the attention table, wherein the attention table is presented according to the minimum level of each target autonomous driving standard.

[0014] Secondly, embodiments of this application also provide an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory through the bus. The machine-readable instructions are executed by the processor to perform the steps of the autonomous driving standard processing method described in the first aspect or any possible implementation of the first aspect.

[0015] Thirdly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the autonomous driving standard processing method described in the first aspect or any possible implementation of the first aspect.

[0016] This application provides an autonomous driving standard processing method, which includes: obtaining at least one autonomous driving function and at least one market for a project to be launched, input by a user; sending each autonomous driving function and each market to be launched to a large language model, so that the large language model matches at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched in a database, wherein the database includes autonomous driving standards applicable to all candidate markets, and the market to be launched is one of all candidate markets; receiving a attention table corresponding to each target autonomous driving standard output by the large language model according to a preset structured table, wherein the attention table is used to indicate information of the target autonomous driving standard that needs to be paid attention to by technical personnel. This application obtains at least one autonomous driving function and at least one market for a project to be launched, and uses a large language model to determine at least one target autonomous driving standard based on the at least one autonomous driving function and the at least one market for which the project will be launched. Furthermore, by analyzing at least one target autonomous driving standard, it outputs information related to the autonomous driving function that technical personnel need to pay attention to for each target autonomous driving standard. This solves the technical problem in the prior art that relies on manual processing, which makes it impossible to accurately and comprehensively determine the regulatory standards that a vehicle to be launched needs to comply with in the market. This achieves the technical effect of improving the accuracy of querying regulatory standards and reducing the cost of regulatory review.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of a standard processing method for autonomous driving provided in an embodiment of this application is shown.

[0020] Figure 2 This paper illustrates a functional block diagram of an autonomous driving standard processing device provided in an embodiment of this application.

[0021] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0023] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] In existing technologies, automotive companies face the following shortcomings in managing regulations and standards: Firstly, manual data collection is inefficient. Since different regions use different regulations and standards, manually logging into websites and compiling all standards is time-consuming and may not even yield the latest versions, impacting processing efficiency. Secondly, manual classification of regulations and standards relies on experience; for example, manual assessment of whether a clause in a regulation applies to automatic emergency braking or driver assistance systems leads to inconsistent human judgment and calibration, hindering unified management and consuming significant time and labor costs, resulting in low efficiency. Thirdly, updates are delayed and lack warnings. There is no automatic notification when any regulation or standard is updated, requiring manual periodic website logins to check for updates, which can easily lead to missed effective dates and failure to inspect vehicles according to the latest standards, causing compliance risks. Fourthly, the itemization of regulations relies on manual processing, resulting in low efficiency in project compliance assessment. Manual processing is required for each market, increasing workload and impacting vehicle launch times if multiple markets are involved.

[0025] Based on this, this application provides an autonomous driving standard processing method. By acquiring at least one autonomous driving function and at least one market for a project to be launched, and using a large language model to determine at least one target autonomous driving standard based on the at least one autonomous driving function and the at least one market, the method analyzes the at least one target autonomous driving standard to output information related to the autonomous driving function that technical personnel need to focus on for each target autonomous driving standard. This solves the technical problem in the prior art where reliance on manual processing leads to the inability to accurately and comprehensively determine the regulatory standards that a vehicle to be launched needs to comply with in the market, thus improving the accuracy of regulatory standard queries and reducing regulatory review costs. Specifically: Please see Figure 1 , Figure 1 This is a flowchart illustrating a standard processing method for autonomous driving provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the standard processing method for autonomous driving includes the following steps: S101: Obtain user input regarding at least one autonomous driving function and at least one market to be launched for the project to be launched.

[0026] Among them, "projects to be launched" refers to a certain version of the autonomous driving system or a vehicle using a certain version of the autonomous driving system, "autonomous driving function" refers to the functions that the autonomous driving system has when performing autonomous driving operations, and "markets to be launched" refers to the actual markets in which the autonomous driving system will be launched.

[0027] For example, at least one autonomous driving function includes: AEB (Automatic Emergency Braking), LDW (Lane Departure Warning), LKA (Lane Keeping Assist), ACC (Adaptive Cruise Control), DMS (Driver Monitoring System), NOA (Navigation Assist), OTA (Over-The-Air Update), SOTIF (Safety in the Expected Function), ALKS (Automatic Lane Keeping System), ELK (Emergency Lane Keeping), BSM (Blind Spot Monitoring), APA (Automatic Parking Assist), TJA (Traffic Jam Assist), FCW (Forward Collision Warning), AES (Automatic Emergency Steering), FCTB (Forward Crossing Braking), etc. At least one market to be launched includes: Europe, Asia, Southeast Asia, South America, etc.

[0028] For example, a lookup table can be pre-set for projects to be launched and at least one autonomous driving function they involve. Then, when a user only needs to provide the project to be launched and its corresponding market (at least one market), they can determine the at least one autonomous driving function possessed by the project by consulting the lookup table. Taking a real-world business scenario as an example, if a car model equipped with version A of an autonomous driving system is being exported to the Middle East and Southeast Asia, the user only needs to input that the project to be launched is version A of the autonomous driving system and that at least one market is the Middle East and Southeast Asia to perform subsequent evaluation operations. Furthermore, the lookup table can also be set up with the project to be launched and the names of the chips it uses, with each chip name relating to its corresponding autonomous driving function. This allows the user to determine the at least one autonomous driving function corresponding to the project simply by inputting the project name.

[0029] S102: Send each autonomous driving function and each market to be launched to the large language model, so that the large language model can match the database with at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched.

[0030] The database includes autonomous driving standards applicable to all candidate markets, and the market to be launched is one of all candidate markets.

[0031] Specifically, at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched is matched in the following way: at least one first autonomous driving standard matching each market to be launched is identified from all autonomous driving standards in the database; at least one second autonomous driving standard matching the at least one autonomous driving function is determined from at least one first autonomous driving standard according to at least one target classification label corresponding to each first autonomous driving standard; at least one target autonomous driving standard is determined from each second autonomous driving standard according to the preset priority corresponding to each second autonomous driving standard.

[0032] In other words, after identifying at least one autonomous driving function required by the project to be launched and at least one target market, a large language model is used to extract all first autonomous driving standards adapted to each target market from the database. Then, from these first autonomous driving standards, at least one second autonomous driving standard adapted to each autonomous driving function of the project is selected. Thus, for each autonomous driving function of the project, a pre-defined priority autonomous driving standard is selected from all second autonomous driving standards applicable to that function in a target market as the target autonomous driving standard to regulate that function. The pre-defined priority can be the priority of the mandatory autonomous driving standard, meaning the selected target autonomous driving standard refers to the mandatory autonomous driving standard selected from all second autonomous driving standards applicable to that function in a target market; alternatively, the pre-defined priority can also be all priorities involved in the autonomous driving standard, effectively using all second autonomous driving standards applicable to a function in a target market as the target autonomous driving standard for that function. The setting of the pre-defined priority can be configured according to the actual application and is not mandatory.

[0033] For example, if the target market is South Korea, and South Korea has announced specific autonomous driving standards and also implements EU standards, then the at least one target autonomous driving standard for South Korea can include its own announced autonomous driving standards and the vehicle safety standards issued by the European New Car Assessment Programme (Euro NCAP). In other words, the autonomous driving standards of which regulatory systems are actually followed in each target market should be pre-configured, and a comparison relationship between each candidate market and its corresponding at least one regulatory system should be pre-set. This way, when the target market is known, the at least one regulatory system corresponding to the target market can be determined by querying the comparison relationship. Then, from the autonomous driving standards involved in each of the retrieved regulatory systems, the autonomous driving standards suitable for the autonomous driving functions of the project to be launched can be further selected as the at least one target autonomous driving standard that the project should comply with.

[0034] For example, if a project to be launched involves five autonomous driving functions: automatic emergency braking, lane departure warning, lane keeping assist, adaptive cruise control, and pilot assist, and since the project needs to be launched in Southeast Asia, the database is consulted to find all the autonomous driving standards that must be followed in Southeast Asia. From these standards, a target autonomous driving standard is identified that regulates at least one of the automatic emergency braking, lane departure warning, lane keeping assist, adaptive cruise control, and pilot assist systems. This determines at least one target autonomous driving standard that the project must comply with for its launch in Southeast Asia. A single target autonomous driving standard may regulate any one of the automatic emergency braking, lane departure warning, lane keeping assist, adaptive cruise control, and pilot assist systems, or it may regulate multiple of these systems.

[0035] Furthermore, the database should include all autonomous driving standards used to regulate autonomous driving systems in each of the candidate markets. An autonomous driving standard may be applied in multiple candidate markets or only in one candidate market. Here, "all candidate markets" can be understood as all markets globally.

[0036] Specifically, the database is constructed as follows: From the preset data sources corresponding to all regulatory systems involved in all candidate markets, an autonomous driving standard set is determined. This set includes standard texts corresponding to multiple autonomous driving standards covered by all regulatory systems. For each autonomous driving standard in the set, the standard text is sent to the large language model. The large language model parses the standard text according to multiple preset structured fields to obtain the standard structured data corresponding to that autonomous driving standard. The technical structure of the standard text is then extracted, describing the technical requirements at each level. For each autonomous driving standard in the set, multiple preset classification labels are sent to the large language model. The large language model determines at least one target classification label matching the autonomous driving standard from among the preset classification labels. The standard text, standard structured data, technical structure, and at least one target classification label corresponding to each autonomous driving standard are then detected to form the constructed database.

[0037] Specifically, the set of autonomous driving standards is determined as follows: All regulatory systems involved in all candidate markets are identified, and each regulatory system is sent to the large language model to receive multiple sets of search keywords generated by the large language model for each regulatory system. These multiple sets of search keywords include various technical fields involved in autonomous driving technology. For each regulatory system, a search is performed in a preset data source corresponding to that regulatory system according to the multiple sets of search keywords to find at least one candidate standard from the preset data source corresponding to that regulatory system. Candidate standards that meet the similarity criteria found under all regulatory systems are identified and merged. The merged candidate similarity and each candidate standard that does not meet the similarity criteria are used as multiple autonomous driving standards. The standard texts corresponding to these multiple autonomous driving standards are then combined to form the set of autonomous driving standards.

[0038] In other words, the set of autonomous driving standards should contain the standard texts of all autonomous driving standards that each candidate market in the world needs to comply with. Therefore, it is necessary to conduct an autonomous driving standard search across all regulatory systems involved in the global market to collect all autonomous driving standards related to autonomous driving within each regulatory system, in order to construct the set of autonomous driving standards according to all autonomous driving standards involved in each candidate market.

[0039] The various regulatory frameworks encompass a range of autonomous driving standards, including the UN Global Technical Regulations for Automated Driving Systems announced by the World Forum for Harmonization of Vehicle Regulations (UNECE WP.29) within the United Nations Economic Commission for Europe; autonomous driving standards announced in the Official Journal of the European Union; autonomous driving standards announced by the International Organization for Standardization (ISO); autonomous driving standards announced by the Society of Automotive Engineers (SAE); autonomous driving standards in Chinese national standards (such as Chinese mandatory standards GB and Chinese recommended standards GB / T); Chinese industry standards; the China New Car Assessment Programme (C-NCAP); regulations from the National Highway Traffic Safety Administration (NHTSA) and the Federal Motor Vehicle Safety Standards (FMVSS) series in the United States; automotive standards in Japanese Industrial Standards (JIS); automotive safety standards issued by the Korea Automobile Manufacturers Association (KMVSS); and automotive safety standards issued by the European New Car Assessment Programme (Euro NCAP). In other words, autonomous driving standards generally refer to the implementation standards for autonomous driving within the new car assessment procedures stipulated by the regulatory framework. An autonomous driving system can only be launched in a market if it meets the conditions stipulated by at least one target autonomous driving standard corresponding to the market it is intended to enter.

[0040] Each candidate market involves a regulatory framework it must comply with. Furthermore, multiple candidate markets may all need to comply with the autonomous driving standards under a single regulatory framework, while a candidate market may only need to comply with the autonomous driving standards under the regulatory framework specific to its location. Therefore, by taking the union of the regulatory frameworks for each candidate market, we can determine all the regulatory frameworks for autonomous driving systems in the global market.

[0041] Since the regulatory framework for automobiles includes not only standards for autonomous driving but also standards related to vehicle body design, lighting design, and other aspects, it is necessary to search the regulatory framework for autonomous driving standards to retrieve those relevant to autonomous driving systems from all automotive-related regulations. Therefore, for each regulatory framework, the framework is sent to a large language surface model. This model generates multiple sets of search keywords corresponding to that framework. These keywords are then input into a pre-defined data source for that framework to retrieve at least one candidate standard related to autonomous driving within that framework, and the standard text of the candidate standard is downloaded.

[0042] The default data source for each regulatory system refers to the official website and database of that regulatory system. Access to this default data source is achieved by setting up a data collection interface. For example, an independent parsing adapter can be configured for each default data source to handle different page structures and data formats. The parsing adapter should at least parse the name and latest version number of each regulatory system within that system to facilitate searching using multiple sets of search keywords.

[0043] For example, targeting the World Forum for Harmonization of Vehicle Regulations within the United Nations Economic Commission for Europe, the system accesses its official regulatory database page via an interface. It then uses multiple sets of corresponding search keywords to determine the latest version of autonomous driving rules and parses a list of standard markup language (HTML) documents to obtain the number, name, latest revision number, and publication date of each regulatory standard. Furthermore, the interface should be compatible with the access interfaces of the pre-defined data sources corresponding to each regulatory system.

[0044] For each regulatory framework, the corresponding sets of search keywords are designed differently to ensure comprehensive coverage across sub-domains and minimize omissions. Furthermore, each search result is automatically parsed and returned to the page by a large language model to extract key metadata information such as the regulatory number, name, publication date, and effective date of the retrieved candidate standards. Additionally, the language of the multiple sets of search keywords should be set according to the language used by the preset data source for that regulatory framework; that is, the language of the multiple sets of search keywords used in the retrieval should be the same as the preset data source for the regulatory framework.

[0045] For example, the various technical fields involved in autonomous driving technology include the field of autonomous driving, the field of cybersecurity and data privacy protection involved in autonomous driving applications, and the field of testing involved in testing autonomous driving systems. Therefore, multiple sets of search keywords include search keywords corresponding to the field of autonomous driving, search keywords corresponding to cybersecurity and data privacy protection, and search keywords corresponding to the testing field. For example, search keywords corresponding to the field of autonomous driving include: automatic lane keeping, automatic emergency braking, driver monitoring according to regulations, combined driving assistance, automatic emergency braking, driver attention monitoring, etc.; search keywords corresponding to cybersecurity and data privacy protection include: cybersecurity management system, software updates, OTA (Over-The-Air) technology, etc.; search keywords corresponding to the testing field include: autonomous driving testing requirements, safety performance testing, New Car Assessment Program, industry group standards, etc.

[0046] Since the same autonomous driving standard may appear in the preset data sources of multiple regulatory systems, a deduplication strategy is needed to avoid duplicate autonomous driving standards in the standard set. Generally, if one regulatory system decides to directly use an autonomous driving standard already drafted under another regulatory system, the regulatory number and name of the autonomous driving standard in one regulatory system will generally directly reference the regulatory number and name of the autonomous driving standard in the other regulatory system. Furthermore, if it is determined that the regulatory numbers and names of candidate standards retrieved under at least two regulatory systems according to multiple sets of search keywords are the same, then it is considered that the content of candidate standards with the same regulatory numbers and names under at least two regulatory systems actually refers to the content of the same autonomous driving standard.

[0047] For example, a standardized numbering format is used to unify the numbers of all candidate standards from different preset data sources into a standardized form, and candidate standards with the same number under the standardized form are selected. Natural language processing technology is used to calculate the semantic similarity of the regulatory names of all candidate standards from different preset data sources, so as to select candidate standards with the same regulatory name according to semantic similarity. The selected candidate standards with different regulatory names but the same number are merged, and the selected candidate standards with the same number and the same regulatory name are also merged, so that the merged candidate standards and the unmerged candidate standards are used as the autonomous driving standards in the autonomous driving standard set.

[0048] In other words, different preset data sources may result in different names for the same candidate regulations due to variations in translation or expression conventions. For example, Regulation No. 157 retrieved from the World Forum for Harmonization of Vehicle Regulations (WFR) preset data source and UNR157 retrieved from the Official Journal of the European Union (OEU) preset data source are both the same regulatory standard. The system directly identifies both regulatory standards as UNR157 and merges them into a single record, retaining the information of this regulatory standard from both the WFR and OEU preset data sources, as well as the most complete standard text downloaded from their respective preset databases. Thus, when the target market is the European market, the information and standard text obtained from the OEU preset data source are used. When the target market is a market other than Europe, and that market also requires compliance with the autonomous driving standards stipulated by the WFR, the information and standard text obtained from the WFR preset data source are used. Subsequently, when performing technical structure decomposition on the standard text, only one decomposition needs to be performed, while the standard structured data can be extracted separately for Europe and other markets.

[0049] Furthermore, when the names of regulations are semantically similar but the regulation numbers are different, the two regulations will not be merged. The two regulations may be related to the same function of autonomous driving, but because the numbers are different, they are not considered to be merged.

[0050] The pre-defined structured fields include the regulation number, regulation name, regulatory system, applicable region, mandatory nature, effective date, repeal date, publication date, issuing agency, original link, abstract, and usage status. Each pre-defined structured field can be adjusted according to actual needs, and this application does not impose any restrictions on this.

[0051] The large language model parses each autonomous driving standard in each preset data source and extracts information from multiple preset structured fields for each autonomous driving standard to obtain the standard structured data corresponding to each autonomous driving standard. When extracting information from multiple preset structured fields, some information can be directly extracted from the webpage of the preset data source. If it cannot be extracted from the webpage, it is extracted from the standard text of the autonomous driving standard. The standard text refers to the portable file format of the autonomous driving standard.

[0052] For example, taking the autonomous driving standard UNR157 retrieved from the preset data source of the Official Journal of the European Union as an example, its regulation number is UNR157; the regulation title is "Uniform provisions concerning the approval of vehicles with regard to Automated Lane Keeping Systems"; the regulatory system is the World Forum for Harmonization of Vehicle Regulations (UNECE) WP.29 within the United Nations Economic Commission for Europe; the applicable regions are the European Union, Japan, and South Korea; it is mandatory; the effective date is January 22, 2021; the repeal date is blank; the publication date is January 22, 2021; the issuing agency is the United Nations Economic Commission for Europe (UNECE); the original link is the link to the official document page of the United Nations Economic Commission for Europe; the abstract is the world's first conditional autonomous driving (Level 3 autonomous driving standard), specifying the system requirements for Automated Lane Keeping Systems (ALKS) under specific conditions (roads with a speed of 60 km / h or less, highways, and roads without pedestrians or cyclists); the status is currently in effect.

[0053] Among them, multiple preset category labels are used to represent autonomous driving functions. These preset category labels include, but are not limited to: AEB (Automatic Emergency Braking), LDW (Lane Departure Warning), LKA (Lane Keeping Assist), ACC (Adaptive Cruise Control), DMS (Driver Monitoring System), NOA (Navigation Assist), OTA (Over-The-Air Update), SOTIF (Safety in the Expected Function), ALKS (Automatic Lane Keeping System), ELK (Emergency Lane Keeping), BSM (Blind Spot Monitoring), APA (Automatic Parking Assist), TJA (Traffic Jam Assist), FCW (Forward Collision Warning), AES (Automatic Emergency Steering), and FCTB (Forward Crossing Braking).

[0054] Furthermore, after parsing the actual content of each autonomous driving standard, the large language model selects at least one functional label related to that autonomous driving standard from multiple preset classification labels as at least one target classification label for matching that autonomous driving standard. For example, the two target classification labels obtained by declassifying the General Technical Requirements for Automotive Software Upgrades GB 44496-2024 are Remote Online Upgrade (OTA) and Cybersecurity. The seventh major upgrade of the China New Car Assessment Programme (C-NCAP) added a test item for preventing accidental accelerator pedal use; therefore, the two target classification labels obtained by declassifying the seventh major upgrade of the C-NCAP are Automatic Emergency Braking and Forward Crossing Braking. Thus, at least one target classification label corresponding to each autonomous driving standard can directly determine whether the system to be launched needs to comply with that autonomous driving standard, or whether that autonomous driving standard is the target autonomous driving standard that the system to be launched needs to comply with.

[0055] For example, error correction is performed on each piece of information parsed by the large language model to avoid errors in the parsed standard structured data, at least one target classification label, etc. Typical error correction scenarios include: the large language model misidentifies the issuing authority of the Korean Autonomous Vehicle Management Act Enforcement Rules as belonging to the European regulatory system; therefore, when identifying autonomous driving rules from various countries, the issuing authority should be directly set to the respective country's issuing authority. Another example is when the large language model matches the number GB / T 39902-2021 as an automotive standard, but the rule engine checks the number and finds that the standard is actually for urban rail transit; this should be automatically marked as requiring manual confirmation and prevented from being added to the database. The original text should then be reread according to the link to check whether the parsed regulation number and name match the original content, and whether the regulation content is relevant to autonomous driving. Furthermore, the numbering format of each regulation is rigorously verified. For example, regulations with numbers starting with UNR are automatically changed to UNECE WP.29, eliminating the abbreviation. Chinese national mandatory standards (numbers starting with GB) and recommended standards (numbers starting with GB / T) are compared to check if the number prefixes match the mandatory nature of preset structured fields, thus determining the accuracy of the identified mandatory nature of the preset structured fields. Industry standards (such as QC / T automotive industry standards and YD / T communication industry standards) are also compared based on their number prefixes to check the accuracy of the identified mandatory nature of the preset structured fields. In other words, detection rules can be set to address recognition errors in the large language model, correcting errors generated during regulation parsing and improving recognition accuracy.

[0056] Specifically, the technical structure of the standard text is deconstructed in the following way: the chapter name of each level of the standard text is identified; for each minimum level of the standard text, the text content of the minimum level is sent to the large language model to obtain the semantic description parsed by the large language model for the text content of the minimum level, the semantic description including functional description or technical verification indicators; the chapter name of each level of the standard text and the semantic description corresponding to each minimum level are determined as the technical structure of the standard text.

[0057] Each autonomous driving standard presents its content in a hierarchical manner, with the lowest level generally specifying the conditions that the autonomous driving system must meet according to regulations. Therefore, for each autonomous driving standard, the standard text is converted into text content, retaining the original paragraph structure and chapter numbers. This involves extracting all text content and using pre-set regular expression patterns (matching multi-level numbering structures such as "X", "XX", "XXX", "XXXX") to extract the chapter names or titles at each level. When identifying each chapter name, only the actual regulatory requirements can be retained; the introductory section of the autonomous driving standard can be selectively deleted. Furthermore, the large language model only sends the lowest level text content to itself for semantic analysis. This analysis determines the semantic description of the lowest level as a functional description or technical verification indicator. Thus, the chapter names at each level and the semantic description at each lowest level constitute the technical structure of the standard text.

[0058] For example, if the hierarchy of autonomous driving is designed to be four layers, the pre-set regular expressions include: “X”, “XX”, “XXX”, and “XXXX”. The system is structured as follows: The first level is the chapter level, for example, the level name of chapter 4 is "Navigation Combination Driving Assistance System", which summarizes the overall safety and performance requirements of this functional domain; the second level is a section level under a chapter level, for example, the level name of section 6 of chapter 4 is subdomains such as "Basic Control Requirements", "Scene Recognition Requirements", and "Human-Machine Interaction Requirements"; the third level is a subsection level under a section level, for example, the level name of subsection 2 of section 6 of chapter 4 is "Lane Change Control", which describes the specific technical requirements scenario when the system performs automatic lane changes; the fourth level is an entry level under a subsection level, for example, the second entry of subsection 2 of section 6 of chapter 4, which determines the technical verification indicator as a rearward safe distance greater than or equal to the distance corresponding to the time of collision (TTC), and stipulates that during lane changes, the safe distance between the nearest vehicle behind the vehicle in the current lane and the vehicle should not be less than the safe distance corresponding to the collision time determined at the speed of the following vehicle.

[0059] In this context, the functional description refers to a summary description of the lowest-level text content, while the technical verification metrics refer to the technical indicators that the lowest-level text content specifies the actual requirements for the autonomous driving system. In other words, because regular expressions cannot understand the semantic boundaries and engineering implications defined by the technical metrics, a large language model is needed to parse the lowest-level text content. For example, the parsed technical verification metric might be a rearward safe distance greater than or equal to the time to rear collision (TTC).

[0060] Specifically, the method further includes: for each minimum level of the standard text, sending at least one preset supplementary field to the large language model, so that the large language model can parse the text content of the minimum level according to at least one preset supplementary field to obtain the text content corresponding to each preset supplementary field; and adding the text content corresponding to each minimum level of the standard text in each preset supplementary field to the technical structure of the standard text.

[0061] Among them, at least one of the following pre-defined supplementary fields includes: requirement domain, verification method, and affected domain controller and algorithm module. The requirement domain refers to the minimum level of text content used to measure the performance of the autonomous driving system under specific requirements. The verification method refers to the verification method used to verify whether the autonomous driving system conforms to the minimum level of text content. The affected domain controller and algorithm module refers to the module used by the autonomous driving system to verify the minimum level of text content.

[0062] Furthermore, the pre-defined supplementary fields refer to the fields other than the semantic description that are parsed when parsing the text content at the lowest level. The text content at the lowest level is summarized by parsing the pre-defined supplementary fields and the semantic description.

[0063] For example, the second item of the second subsection of the sixth section of Chapter 4 is parsed out, and the technical verification index is determined to be the distance corresponding to the rearward safe distance being greater than or equal to the rear vehicle collision time (TTC). The large language model generates the following supplementary fields: the requirement domain is lane change safety in lane control; the verification method is field test, that is, a closed field to simulate the scenario of the rear vehicle approaching; the affected domain controller and algorithm module are the affected perception module (which needs to detect the speed and distance of the rear vehicle by rearward millimeter-wave radar) and the decision planning module (which needs to calculate the rear vehicle collision time in real time and determine the lane change safety window).

[0064] For example, the second item in subsection 1 of section 6 of chapter 4 is parsed out. The chapter name of this item is "Maximum Operating Speed ​​Limit," the technical verification indicator is that the system operating speed does not exceed 120 km / h, the requirement domain is speed management under longitudinal control, and the verification method is bench testing and real vehicle testing. The above automatically filled fields directly transform the regulatory provisions into an executable requirement description for engineering.

[0065] Furthermore, by acquiring the standard text corresponding to each autonomous driving standard and parsing out the standard structured data, technical structure, and at least one target classification label for each autonomous driving standard, a database is constructed.

[0066] Specifically, the method further includes: detecting each autonomous driving standard in the database according to a preset update cycle, and determining the update information of the database, so as to update the database through the update information, wherein the update information includes at least one of the following: newly added autonomous driving standards not involved in the database, revisions to existing autonomous driving standards in the database, and error content corresponding to existing autonomous driving standards in the database; and pushing the update information according to a preset push cycle to notify the database of information changes.

[0067] In other words, the autonomous driving standards in the database are regularly checked. This check allows us to see if there are any new additions, revisions, or errors. For example, we can search for autonomous driving standards from preset data sources in each regulatory system every week and determine if there are any new additions or revisions. If three new regulatory updates are detected that week, such as the new draft of the China New Car Assessment Procedure, the updated AES (Automatic Emergency Steering) test plan in the European New Car Assessment Programme, and the revised automotive safety standards issued by the Korea Automobile Manufacturers Association that add human-machine interaction requirements for combined driver assistance (L2+), the new additions and revisions in the autonomous driving standards are categorized and organized into update information according to the regulatory system. This update information is then automatically pushed to the compliance team leader so that they are aware of the updates to the autonomous driving standards in a timely manner.

[0068] For example, it is also possible to check monthly whether there are autonomous driving standards that will take effect, be implemented, or be enforced in one or three months. For instance, if the vehicle information security technical requirements have taken effect, it is marked as implemented. If an autonomous driving standard issued by a certain regulatory system adds a mandatory clause for vehicles, it should be marked as a red warning. If the transition period for emission regulations in a Southeast Asian country is about to end, it should be marked as a yellow concern.

[0069] For example, daily quality checks can be performed to monitor database errors. For instance, if it detects that two original page links for autonomous driving standards have become invalid due to website redesigns of their corresponding preset data sources (i.e., the original page links are inaccessible and return a 404 Not Found error), the system can automatically attempt to relocate the correct page links through a search engine and update them. The corrected information can then be sent to technical personnel to inform them that the information has been modified.

[0070] For example, a monthly review can be conducted each month. This review will show that 12 new autonomous driving standards have been added to the database this month, mainly focusing on the United Nations Economic Commission for Europe and Chinese national standards. It will also identify that the coverage of local standards in Southeast Asia is still relatively weak, and it is recommended that the focus be on adding autonomous driving standards in Southeast Asia that are not currently included in the database next month.

[0071] For example, an emergency notification can also be issued. If the effective date of a mandatory regulation is advanced, such as a sudden notice that moves the deadline for cybersecurity certification forward by 6 months, there is no need to wait for a regular push notification to send this information. Instead, an emergency notification should be generated and sent to the relevant personnel immediately upon detection of the information.

[0072] Furthermore, an anti-duplicate message mechanism was implemented, which performs duplicate detection on regularly pushed messages (daily, weekly, monthly, etc.) to prevent the same messages from being repeatedly sent to technical personnel. Specifically, a two-layer anti-duplicate mechanism was set up. The first layer is a memory status marker: each push task is recorded in memory, ensuring that the same push message is not triggered repeatedly within the same execution cycle based on the executed push message and timestamp. The second layer is a persistent file lock: a lock file named after the date is created in the file system. Even if the system restarts unexpectedly, the status of the lock file can be checked to prevent duplicate push messages from being sent. In actual operation, tasks are generally triggered to be pushed again after a system restart. The file lock mechanism successfully intercepted duplicate pushes, verifying the effectiveness of the two-layer mechanism.

[0073] After the database is built, it can be continuously monitored to verify the accuracy of information and correct errors. First, when an autonomous driving standard is acquired, the corresponding standard text, structured data, technical structure, and at least one target classification label need to be added to the database. For example, when the intelligent standard GB / T 39066-2020 is collected and added to the database, the rule engine detects that the actual standard corresponding to this number is related to building economics terminology and not to autonomous driving. Therefore, it should be directly blocked from being added to the database, and the standard should be marked for manual verification to prevent erroneous standards from entering the official database. The integrity of the standard's structured data can also be checked, i.e., whether each preset structured field of the autonomous driving standard has been identified to avoid omissions. If omissions are found, the preset structured fields are marked as incomplete. It can also check whether at least one target classification label belongs to multiple preset classification labels and verify whether each date is logically consistent. For example, if the effective date of the autonomous driving standard is earlier than the publication date, a warning is triggered. Furthermore, a full database scan was performed monthly, and 11 problematic records were detected among all autonomous driving standards in the database. Among these, 7 standards had serious mismatches between their national standard numbers and names. For example, the standard numbered GB / T 39900-2021 was recorded in the database as "Performance Requirements for Lane Departure Warning Systems for Passenger Vehicles," but the actual standard corresponding to this number was from another field. The correct standard number for lane departure warning systems should be GB / T26773-2011. 3 standards were duplicates, meaning the same autonomous driving rule was entered into the database in two different data collection cycles. 1 standard had an incorrect original link. Therefore, duplicate autonomous driving standards should be deleted, the 7 incorrect numbers corrected, and standards not belonging to the autonomous driving field removed. If the correct number cannot be automatically confirmed, it should be marked for manual processing. For incorrect original links, the correct page link address should be re-determined through a search engine. Furthermore, a monthly sampling accuracy assessment should be conducted by randomly selecting a predetermined number of autonomous driving standards from the database and comparing the corresponding structured data and at least one target classification label with the standard text for each standard, focusing on confirming the accuracy of the regulation number, regulation name, system source, and target classification label. For example, if there are errors in the sampled number of autonomous driving standards, such as the omission of a target classification label for an autonomous driving standard, which involves both automatic emergency braking and forward collision warning but only labels automatic emergency braking, this omission should be fed back to the classification rule engine so that a new rule can be added when the classification label is determined. When dealing with regulations and standards related to collision warning, it is important to determine whether the autonomous driving standard involves automatic emergency braking and forward collision warning.

[0074] Furthermore, during the quality inspection process, duplicate records in the database are deleted and confirmed, erroneous records are corrected, and missing function tags are added. All information change operations in the database are recorded in the repair log, and data before and after the changes is retained for comparison. The basis for the data changes is also marked to ensure the repair process is traceable and auditable.

[0075] S103: Receive the attention table corresponding to each target autonomous driving standard, which is output by the large language model according to the preset structured table.

[0076] The attention table is used to indicate information that technical personnel need to pay attention to regarding the target autonomous driving standard.

[0077] In other words, the large language model fills the pre-defined structured table in the constructed database with standard structured data, technical structure and at least one target classification label. This means that the information that technical personnel need to pay attention to is filled into the pre-defined structured table to construct the attention table.

[0078] Specifically, the attention table corresponding to each target autonomous driving standard is determined in the following way: for each target autonomous driving standard, the preset structured table is filled with the information corresponding to the target autonomous driving standard involved in the database to construct the attention table, wherein the attention table is presented according to the minimum level of each target autonomous driving standard.

[0079] In other words, at least the minimum level of the technical structure of each target autonomous driving standard in the database should be extracted so that technicians can use the tables to determine whether the project to be launched meets the technical verification indicators specified by each target autonomous driving standard. This allows for the review of the project before its actual launch, timely identification of any parts that do not meet the technical verification indicators, and their modification.

[0080] For example, if the target market is the Middle East or Southeast Asia, the Middle East generally adopts the UN Global Technical Regulations for Automated Driving Systems announced by the World Forum for Harmonization of Vehicle Regulations (UNECE WP.29) within the United Nations Economic Commission for Europe. Southeast Asian countries, however, have localized differences, typically having their own automated driving standards. The first step is to filter applicable automated driving standards from the database according to the target market. If the project involves five automated driving functions—AEB (Automatic Emergency Braking), LDW (Lane Departure Warning), LKA (Lane Keeping Assist), ACC (Adaptive Cruise Control), and NOA (Navigate on Autopilot)—then the scope is further narrowed down by category tags to find multiple target automated driving standards.

[0081] This process involves several steps. First, after identifying all autonomous driving standards for a specific autonomous driving function in a target market, these standards can be filtered by their mandatory nature. This allows for the differentiation between mandatory and recommended standards. Furthermore, each standard is prioritized based on its mandatory nature and implementation time, with mandatory and already effective standards having higher priority. For example, the highest priority standard is one that must be confirmed as compliant with autonomous driving requirements before the standard operating procedure (SOP) for the project and must be enforced. The second highest priority standard is one involving cybersecurity and is mandatory. The second lowest priority standard is a recommended standard, and the lowest priority standard is one used for industry trend reference. Finally, when selecting target autonomous driving standards, a preset priority can be defined as the highest and second highest priority standards. This allows for the selection of mandatory standards from all standards for a specific autonomous driving function in a target market.

[0082] The pre-defined structured table requires the following information: the regulatory number of the target autonomous driving standard, the chapter name of each target's minimum level of technical verification indicators within the target autonomous driving standard, the technical verification indicators for each target's minimum level, the priority of the target autonomous driving standard, and its effective date. Target autonomous driving standards can be prioritized, with higher-priority standards listed first. This allows technical personnel to verify whether projects to be deployed meet each target's minimum level in order of priority.

[0083] For example, one row in the attention table corresponds to a minimum target level. Thus, part of one row in the output attention table would be: A target autonomous driving standard has the regulatory number UNR152; the chapter title of a minimum target level of this standard is "Triggering Conditions for Automatic Emergency Braking Systems in Urban Scenarios"; the technical verification indicator for this minimum target level is: a warning is issued when the collision time for a pedestrian or cyclist is less than or equal to 2.6 seconds, and automatic braking is initiated when the collision time is less than or equal to 1.2 seconds; the priority is the highest; and the effective date is July 2024, already in effect. Another row in the output attention table would be: Another target autonomous driving standard has the regulatory number UN R79 Rev.4; the chapter title of a minimum target level is "Safety Function Requirements for Electronic Control Systems"; the technical verification indicator for this minimum target level is: lateral control-type assisted steering functions (lane keeping assist or ELK emergency lane keeping) should issue a warning within 15 seconds of the driver's hands leaving the steering wheel; the priority is the highest; and the effective date is mandatory application by the Gulf Organization (GSO). Furthermore, the concern form can be directly exported and provided to the project engineering team as a compliance checklist.

[0084] Furthermore, when inputting multiple markets for market launch, the system can automatically compare these markets to check if each market involves duplicate minimum target levels that need to be implemented. The table can display duplicate minimum target levels only once and mark the applicable markets. It can also highlight differences between multiple markets. For example, regarding the Middle East and Southeast Asia markets, the Middle East directly uses the regulatory framework announced by the World Forum for Harmonization of Vehicle Regulations (WFORD), while Thailand and Indonesia each have their own localized regulatory frameworks, with Thailand requiring Thai labels and local test reports. While the technical specifications for automatic emergency braking are consistent in both markets, both originating from the WFORD regulatory framework, the certification processes and testing institutions differ, requiring inclusion in the attention table. In conclusion, comparing multiple markets helps project teams determine which technical specifications can be reused, avoiding multiple verifications and ensuring differences only exist at the certification process level.

[0085] Therefore, this application focuses on building a structured database. It collects raw regulatory information from multiple regulatory systems, classifies, tags, and structures the collected information, and then writes it into the database according to standardized regulatory records. This allows for the reading of complete regulatory standard document information from the database, as well as the identification of specific content segments within those documents. Based on user-specified projects and markets to be launched, the application queries regulatory records and entries in the database, and generates a project-specific attention table after filtering by region, function, mandatory nature, and priority. Furthermore, it continuously monitors changes in the regulatory data in the database, triggering periodic or real-time notifications according to a push strategy. It also performs multiple rounds of periodic checks on the database content, automatically fixing or marking detected issues for manual verification, and pushing the fixes to technical personnel.

[0086] Furthermore, this application achieves bidirectional data flow through structured database processing, namely, collecting regulations and standards and extracting information, thereby enabling on-demand querying and output of attention tables, regular monitoring to continuously maintain database quality and provide timely warnings, ensuring fully automated management of regulatory data throughout the entire lifecycle from collection, storage, disassembly, application to quality closure, without the need for manual operation by technical personnel, thus improving processing efficiency.

[0087] Based on the same application concept, this application also provides an autonomous driving standard processing device corresponding to the autonomous driving standard processing method provided in the above embodiments. Since the principle of the device in this application to solve the problem is similar to the autonomous driving standard processing method in the above embodiments of this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0088] like Figure 2 As shown, Figure 2 This application provides a functional block diagram of an autonomous driving standard processing device 10. The autonomous driving standard processing device 10 includes: an acquisition module 101, used to acquire at least one autonomous driving function and at least one market for a project to be launched, input by a user; a matching module 102, used to send each autonomous driving function and each market to be launched to a large language model, so that the large language model matches at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched in a database, wherein the database includes autonomous driving standards applicable to all candidate markets, and the market to be launched is one of all candidate markets; and a table generation module 103, used to receive a attention table corresponding to each target autonomous driving standard output by the large language model according to a preset structured table, wherein the attention table indicates information that needs to be paid attention to by technicians for the target autonomous driving standard.

[0089] Based on the same application concept, see [link / reference] Figure 3 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this application. The electronic device 20 includes a processor 201, a memory 202, and a bus 203. The memory 202 stores machine-readable instructions that can be executed by the processor 201. When the electronic device 20 is running, the processor 201 and the memory 202 communicate through the bus 203. When the machine-readable instructions are executed by the processor 201, the steps of the autonomous driving standard processing method described in any of the above embodiments are executed.

[0090] Specifically, when the machine-readable instructions are executed by the processor 201, they can perform the following processing: obtaining at least one autonomous driving function and at least one market for the project to be launched, input by the user; sending each autonomous driving function and each market to be launched to a large language model, so that the large language model matches at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched in the database, wherein the database includes autonomous driving standards applicable to all candidate markets, and the market to be launched is one of all candidate markets; receiving a attention table corresponding to each target autonomous driving standard output by the large language model according to a preset structured table, wherein the attention table is used to indicate information that the target autonomous driving standard needs to be paid attention to by technical personnel.

[0091] Based on the same concept, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the autonomous driving standard processing method provided in the above embodiments.

[0092] Specifically, the storage medium can be a general-purpose storage medium, such as a portable disk or hard disk. When the computer program on the storage medium is run, it can execute the above-mentioned autonomous driving standard processing method. By acquiring at least one autonomous driving function and at least one market to be launched for the project to be launched, and using a large language model, it determines at least one target autonomous driving standard based on the at least one autonomous driving function and the at least one market to be launched for the project to be launched. Furthermore, by analyzing at least one target autonomous driving standard, it outputs information related to the autonomous driving function that technical personnel need to pay attention to for each target autonomous driving standard. This solves the technical problem in the prior art that relies on manual processing, which makes it impossible to accurately and comprehensively determine the regulatory standards that the vehicle to be launched needs to comply with in the market to be launched. This achieves the technical effect of improving the accuracy of querying regulatory standards and reducing the cost of regulatory review.

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0094] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0095] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0096] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A standard processing method for autonomous driving, characterized in that, The method includes: Obtain user input regarding at least one autonomous driving function and at least one market to be launched for the project to be launched; Each autonomous driving function and each market to be launched are sent to a large language model so that the large language model can match at least one target autonomous driving standard applicable to each autonomous driving function in each market to be launched in the database. The database includes autonomous driving standards applicable to all candidate markets, and the market to be launched is one of all candidate markets. The system receives a focus table for each target autonomous driving standard, output by the large language model according to a preset structured table. The focus table is used to indicate each piece of standard information that needs to be focused on by technicians for the target autonomous driving standard.

2. The method according to claim 1, characterized in that, The database is constructed in the following manner: From the preset data sources corresponding to all regulatory systems involved in all candidate markets, an autonomous driving standard set is determined. The autonomous driving standard set includes the standard texts corresponding to multiple autonomous driving standards covered by all regulatory systems. For each autonomous driving standard in the set of autonomous driving standards, the standard text of the autonomous driving standard is sent to the large language model. The large language model parses the standard text according to multiple preset structured fields to obtain the standard structured data corresponding to the autonomous driving standard. The technical structure of the standard text is decomposed, and the technical structure is used to describe the technical requirements of each level of the standard text. For each autonomous driving standard in the set of autonomous driving standards, multiple preset classification labels are sent to the large language model, and the large language model determines at least one target classification label that matches the autonomous driving standard from among the multiple preset classification labels. The database is constructed by detecting the standard text, standard structured data, technical structure, and at least one target classification label corresponding to each autonomous driving standard.

3. The method according to claim 2, characterized in that, The set of autonomous driving standards was determined in the following way: All regulatory systems involved in all candidate markets are identified, and each regulatory system is sent to the large language model to receive multiple sets of search keywords generated by the large language model for each regulatory system. The multiple sets of search keywords include various technical fields involved in autonomous driving technology. For each regulatory system, a search is performed in the preset data source corresponding to that regulatory system according to the multiple sets of search keywords corresponding to that regulatory system, so as to find at least one candidate standard from the preset data source corresponding to that regulatory system; Candidate standards that meet the similarity criteria are identified from all candidate standards found under all regulatory systems and then merged. The merged candidate similarities and each candidate standard that does not meet the similarity criteria are used as multiple autonomous driving standards. The standard texts corresponding to the multiple autonomous driving standards are then combined to form the autonomous driving standard set.

4. The method according to claim 2, characterized in that, The technical structure of the standard text is deconstructed in the following way: Identify the chapter name at each level of the standard text; For each minimum level of the standard text, the text content of that minimum level is sent to the large language model to obtain the semantic description parsed by the large language model for the text content of that minimum level. The semantic description includes functional description or technical verification indicators. The technical structure of the standard text is determined by the chapter names of each level and the semantic descriptions corresponding to each smallest level.

5. The method according to claim 4, characterized in that, The method further includes: For each minimum level of the standard text, at least one preset supplementary field is sent to the large language model so that the large language model can parse the text content of the minimum level according to at least one preset supplementary field to obtain the text content corresponding to each preset supplementary field. The technical structure of the standard text is to add the text content corresponding to each minimum level of the standard text to each preset supplementary field.

6. The method according to claim 2, characterized in that, The method further includes: Each autonomous driving standard in the database is detected according to a preset update cycle, and the update information of the database is determined so as to update the database. The update information includes at least one of the following: newly added autonomous driving standards not involved in the database, revisions to existing autonomous driving standards in the database, and error content corresponding to existing autonomous driving standards in the database. The updated information is pushed out according to a preset push cycle to notify the database of information changes.

7. The method according to claim 1, characterized in that, The following methods are used to identify at least one target autonomous driving standard applicable to each market to be launched, for each autonomous driving function: Identify at least one first autonomous driving standard that matches each market to be launched from all the autonomous driving standards in the database; Based on at least one target classification label corresponding to each first autonomous driving standard, at least one second autonomous driving standard that matches the at least one autonomous driving function is determined from at least one first autonomous driving standard. Based on the preset priority corresponding to each second autonomous driving standard, at least one target autonomous driving standard is determined among the various second autonomous driving standards.

8. The method according to claim 1, characterized in that, The attention table for each target autonomous driving standard is determined using the following method: For each target autonomous driving standard, the preset structured table is populated with the information corresponding to that target autonomous driving standard in the database to construct the attention table, wherein the attention table is presented according to the minimum level of each target autonomous driving standard.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the standard autonomous driving processing method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the autonomous driving standard processing method as described in any one of claims 1 to 8.

Images