Test method and device of intelligent cabin software, intelligent cabin and vehicle

Abstract

The invention relates to an intelligent cabin software testing method and device, an intelligent cabin and a vehicle. The method comprises the following steps: acquiring a test requirement document of the intelligent cabin software; according to the test demand document, generating a decision tree and a dynamic demand graph corresponding to the test demand document; generating a function knowledge graph corresponding to the intelligent cabin software according to the decision tree and the dynamic demand graph corresponding to the test demand document; according to the function knowledge graph corresponding to the intelligent cabin software, generating a test case of the intelligent cabin software; and according to the test case of the intelligent cabin software, generating and executing a test script of the intelligent cabin software to obtain a test result of the intelligent cabin software. According to the method, the automatic process of compiling the test case and generating the test script can be realized, and the test efficiency of the intelligent cabin software is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicles, and in particular to a test method and device for intelligent cockpit software, an intelligent cockpit, and a vehicle. Background Art

[0002] With the rapid development of new energy vehicles, the complexity and functional diversity of intelligent cockpit software are increasing continuously. Among them, in order to ensure the quality of intelligent cockpit software, it is necessary to conduct Hardware-in-the-Loop (HIL) testing on the intelligent cockpit software.

[0003] In the prior art, the test steps corresponding to multiple test cases of the intelligent cockpit software are executed in parallel through a test bench, and the test data of the test bench are received and analyzed to obtain the test results of the intelligent cockpit software.

[0004] However, in the above method, it is necessary to manually write test cases, which takes a long time and results in low test efficiency of the intelligent cockpit software. Summary of the Invention

[0005] One of the purposes of the present invention is to provide a test method for intelligent cockpit software, which can realize the automated process of test case writing and test script generation, and improve the test efficiency of intelligent cockpit software; the second purpose is to provide a test device for intelligent cockpit software; the third purpose is to provide an intelligent cockpit; the fourth purpose is to provide a vehicle; the fifth purpose is to provide a computer-readable storage medium; the sixth purpose is to provide a computer program product.

[0006] In order to achieve the above purposes, the technical solutions adopted by the present invention are as follows:

[0007] A test method for intelligent cockpit software, the method includes:

[0008] Obtain a test requirement document of the intelligent cockpit software; and generate a decision tree and a dynamic requirement graph corresponding to the test requirement document according to the test requirement document; wherein, the decision tree includes at least one scenario requirement slice; the dynamic requirement graph includes at least one functional requirement slice, and the dynamic requirement graph represents the change track of functional test requirements;

[0009] Generate a functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and the dynamic requirement graph corresponding to the test requirement document; wherein, the functional knowledge graph represents the functional structure and relationship of the intelligent cockpit software;

[0010] Generate test cases for the intelligent cockpit software according to the functional knowledge graph corresponding to the intelligent cockpit software;

[0011] Generate and execute a test script for the intelligent cockpit software according to the test cases of the intelligent cockpit software to obtain the test results of the intelligent cockpit software.

[0012] Further, generating a decision tree corresponding to the test requirement document according to the test requirement document includes:

[0013] Perform semantic recognition processing on the test requirement document to construct a domain knowledge graph; wherein, the domain knowledge graph includes at least one initial scenario requirement shard; the domain knowledge graph represents the association relationship between each initial scenario requirement shard;

[0014] Generate a decision tree corresponding to the test requirement document according to the domain knowledge graph; wherein, the decision tree includes a root node and at least one leaf node; the root node represents the requirement type; the leaf node represents the final scenario requirement shard.

[0015] Further, performing semantic recognition processing on the test requirement document to construct a domain knowledge graph includes:

[0016] Perform semantic recognition processing on the test requirement document to obtain at least one initial scenario requirement shard; wherein, the initial scenario requirement shard has attribute information;

[0017] Construct an initial domain knowledge graph according to the attribute information of the at least one initial scenario requirement shard; wherein, the initial domain knowledge graph includes the attribute information of each initial scenario requirement shard and the association relationship between each initial scenario requirement shard;

[0018] Perform semantic chain annotation on each initial scenario requirement shard in the initial domain knowledge graph to obtain the domain knowledge graph.

[0019] Further, generating a decision tree corresponding to the test requirement document according to the domain knowledge graph includes:

[0020] Calculate the semantic entropy value of the semantic vector of each initial scenario requirement shard in the domain knowledge graph to obtain an entropy value matrix corresponding to the domain knowledge graph; wherein, the entropy value matrix includes at least one semantic association entropy value; the semantic association entropy value represents the semantic similarity between every two adjacent initial scenario requirement shards;

[0021] Determine a boundary threshold according to the document complexity of the test requirement document;

[0022] If it is determined that the semantic association entropy value is less than the boundary threshold, the two initial scenario requirement slices corresponding to the semantic association entropy value are merged to obtain a new initial scenario requirement slice corresponding to the semantic association entropy value;

[0023] According to the association relationships between the initial scenario requirement slices, a decision tree corresponding to the test requirement document is constructed.

[0024] Furthermore, generating a dynamic requirement graph corresponding to the test requirement document according to the test requirement document includes:

[0025] Preprocessing the test requirement document to obtain at least one triple; wherein, the triple represents the association relationship between functional text, table parameters, and chart elements;

[0026] Constructing a global requirement graph according to the requirement change history data of the intelligent cockpit software and the at least one triple; wherein, the global requirement graph includes at least one initial functional requirement slice; the global requirement graph represents the initial change trajectory of functional test requirements;

[0027] Performing graph pruning processing on the global requirement graph to obtain the dynamic requirement graph.

[0028] Furthermore, the preprocessing the test requirement document to obtain at least one triple includes:

[0029] Performing word segmentation and semantic encoding processing on each functional text in the test requirement document to obtain a representation set corresponding to each functional text; wherein, the representation set includes at least one word segmentation representation of a functional word segment;

[0030] Parsing and transforming each table in the test requirement document to obtain the table parameters in each table;

[0031] Performing recognition processing on each chart in the test requirement document to obtain the chart elements in each chart;

[0032] Performing semantic alignment processing on the word segmentation representations, the table parameters, and the chart elements to obtain the at least one triple.

[0033] Furthermore, the constructing a global requirement graph according to the requirement change history data of the intelligent cockpit software and the at least one triple includes:

[0034] Construct an initial subgraph corresponding to each document section in the test requirement document according to the requirement change history data of the intelligent cockpit software and the at least one triple; wherein, the initial subgraph includes at least one subgraph node and at least one subgraph edge; the subgraph node represents functional text, or table, or flowchart data in each document section; the subgraph edge represents the association relationship between each subgraph node.

[0035] Based on the attention mechanism, determine the semantic association degree between every two of the initial subgraphs.

[0036] Construct the global requirement graph according to each of the initial subgraphs and each of the semantic association degrees.

[0037] Further, the pruning process of the global requirement graph to obtain the dynamic requirement graph includes:

[0038] Determine the similarity between the preset scenario template and each initial functional requirement slice in the global requirement graph.

[0039] If it is determined that the similarity is greater than the preset threshold, it is determined that there is an association relationship between the preset scenario template corresponding to the similarity and the initial functional requirement slice.

[0040] According to the test priority corresponding to the initial functional requirement slice and the association relationship between the preset scenario template and the initial functional requirement slice, perform hierarchical pruning on the global requirement graph to obtain the dynamic requirement graph.

[0041] Further, the generation of the functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and the dynamic requirement graph corresponding to the test requirement document includes:

[0042] Construct an initial functional knowledge graph corresponding to the test requirement document according to the decision tree and the dynamic requirement graph corresponding to the test requirement document, and the test record data of the intelligent cockpit software; the initial functional knowledge graph includes at least one functional node and at least one functional edge; the functional node is used to represent the functional module of the intelligent cockpit software, and the functional edge represents the interaction relationship between two functional nodes.

[0043] Based on the graph neural network, process the initial functional knowledge graph corresponding to the test requirement document to generate at least one predicted test path and at least one predicted test scenario; wherein; the key functions of the target intelligent cockpit software are included in the predicted test path.

[0044] Optimize the initial functional knowledge graph corresponding to the test requirement document according to the at least one predicted test path and the at least one predicted test scenario, to obtain the functional knowledge graph corresponding to the intelligent cockpit software.

[0045] Further, generating the test script of the intelligent cockpit software according to the test cases of the intelligent cockpit software includes:

[0046] Generate the initial test script of the intelligent cockpit software according to the test cases, interface information and test environment information of the intelligent cockpit software;

[0047] Verify the script information of the initial test script of the intelligent cockpit software to obtain a verification result; wherein, the script information includes script syntax, interface information, integrity of the test path, trigger coverage rate, and environmental adaptability;

[0048] Optimize the initial test script of the intelligent cockpit software according to the verification result to obtain the test script of the intelligent cockpit software.

[0049] A test device for intelligent cockpit software, the device includes:

[0050] The first generation module is used to obtain the test requirement document of the intelligent cockpit software; and generate the decision tree and dynamic requirement graph corresponding to the test requirement document according to the test requirement document; wherein, the decision tree includes at least one scenario requirement slice; the dynamic requirement graph includes at least one functional requirement slice, and the dynamic requirement graph represents the change track of functional test requirements;

[0051] The second generation module is used to generate the functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and dynamic requirement graph corresponding to the test requirement document; wherein, the functional knowledge graph represents the functional structure and relationship of the intelligent cockpit software;

[0052] The third generation module is used to generate the test cases of the intelligent cockpit software according to the functional knowledge graph corresponding to the intelligent cockpit software;

[0053] The test module is used to generate and execute the test script of the intelligent cockpit software according to the test cases of the intelligent cockpit software to obtain the test result of the intelligent cockpit software.

[0054] Further, the first generation module is specifically configured to: perform semantic recognition processing on the test requirement document to construct a domain knowledge graph; wherein, the domain knowledge graph includes at least one initial scenario requirement slice; the domain knowledge graph represents the association relationship between each initial scenario requirement slice; generate a decision tree corresponding to the test requirement document according to the domain knowledge graph; wherein, the decision tree includes a root node and at least one leaf node; the root node represents the requirement type; the leaf node represents the final scenario requirement slice.

[0055] Further, the first generation module is specifically configured to: perform semantic recognition processing on the test requirement document to obtain at least one initial scenario requirement slice; wherein, the initial scenario requirement slice has attribute information; construct an initial domain knowledge graph according to the attribute information of the at least one initial scenario requirement slice; wherein, the initial domain knowledge graph includes the attribute information of each initial scenario requirement slice and the association relationship between each initial scenario requirement slice; perform semantic chain annotation on each initial scenario requirement slice in the initial domain knowledge graph to obtain the domain knowledge graph.

[0056] Further, the first generation module is also specifically configured to: calculate the semantic entropy value of the semantic vector of each initial scenario requirement slice in the domain knowledge graph to obtain an entropy value matrix corresponding to the domain knowledge graph; wherein, the entropy value matrix includes at least one semantic association entropy value; the semantic association entropy value represents the semantic similarity between every two adjacent initial scenario requirement slices; determine a boundary threshold according to the document complexity of the test requirement document; if it is determined that the semantic association entropy value is less than the boundary threshold, merge the two initial scenario requirement slices corresponding to the semantic association entropy value to obtain a new initial scenario requirement slice corresponding to the semantic association entropy value; construct a decision tree corresponding to the test requirement document according to the association relationship between each initial scenario requirement slice.

[0057] Further, the first generation module is also specifically configured to: preprocess the test requirement document to obtain at least one triple; wherein, the triple represents the association relationship between the function text, the table parameter and the chart element; construct a global requirement graph according to the requirement change history data of the intelligent cockpit software and the at least one triple; wherein, the global requirement graph includes at least one initial function requirement slice; the global requirement graph represents the initial change trajectory of the function test requirement; perform graph pruning processing on the global requirement graph to obtain the dynamic requirement graph.

[0058] Further, the first generation module is further specifically configured to: perform word segmentation and semantic encoding processing on each function text in the test requirement document to obtain a representation set corresponding to each function text; wherein, the representation set includes at least one word segmentation representation of a function word segment; parse and transform each table in the test requirement document to obtain table parameters in each table; perform recognition processing on each chart in the test requirement document to obtain chart elements in each chart; perform semantic alignment processing on each of the word segmentation representations, each of the table parameters, and each of the chart elements to obtain the at least one triple.

[0059] Further, the first generation module is further specifically configured to: construct an initial subgraph corresponding to each document chapter in the test requirement document according to the requirement change history data of the intelligent cockpit software and the at least one triple; wherein, the initial subgraph includes at least one subgraph node and at least one subgraph edge; the subgraph node represents a function text, or a table, or flowchart data in each document chapter; the subgraph edge represents the association relationship between each subgraph node; based on the attention mechanism, determine the semantic association degree between every two of the initial subgraphs; construct the global requirement graph according to each of the initial subgraphs and each of the semantic association degrees.

[0060] Further, the first generation module is further specifically configured to: determine the similarity between the preset scenario template and each initial functional requirement slice in the global requirement graph; if it is determined that the similarity is greater than a preset threshold, determine that there is an association relationship between the preset scenario template corresponding to the similarity and the initial functional requirement slice; perform hierarchical pruning processing on the global requirement graph according to the test priority corresponding to the initial functional requirement slice and the association relationship existing between the preset scenario template and the initial functional requirement slice to obtain the dynamic requirement graph.

[0061] Further, the second generation module is specifically configured to: construct an initial functional knowledge graph corresponding to the test requirement document according to the decision tree and the dynamic requirement graph corresponding to the test requirement document, and the test record data of the intelligent cockpit software; the initial functional knowledge graph includes at least one functional node and at least one functional edge; the functional node is used to represent the functional module of the intelligent cockpit software, and the functional edge represents the interaction relationship between two functional nodes; based on the graph neural network, process the initial functional knowledge graph corresponding to the test requirement document to generate at least one predicted test path and at least one predicted test scenario; wherein, the key functions of the target intelligent cockpit software are included in the predicted test path; optimize the initial functional knowledge graph corresponding to the test requirement document according to the at least one predicted test path and the at least one predicted test scenario to obtain the functional knowledge graph corresponding to the intelligent cockpit software.

[0062] Further, the test module is specifically configured to: generate an initial test script for the intelligent cockpit software according to the test cases, interface information, and test environment information of the intelligent cockpit software; verify the script information of the initial test script for the intelligent cockpit software to obtain a verification result; wherein, the script information includes script syntax, interface information, integrity of the test path, trigger coverage, and environmental adaptability; optimize the initial test script for the intelligent cockpit software according to the verification result to obtain the test script for the intelligent cockpit software.

[0063] An intelligent cockpit includes: a memory, a processor;

[0064] The memory stores computer execution instructions;

[0065] The processor executes the computer execution instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementation manners of the first aspect.

[0066] A vehicle, the vehicle includes an intelligent cockpit, and the intelligent cockpit is used to execute the above first aspect and / or various possible implementation manners of the first aspect.

[0067] A computer-readable storage medium stores computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the above first aspect and / or various possible implementation manners of the first aspect.

[0068] A computer program product includes a computer program, and when the computer program is executed by a processor, it implements the above first aspect and / or various possible implementation manners of the first aspect.

[0069] The beneficial effects of the present invention:

[0070] The present invention identifies the requirements document of the intelligent cockpit software test requirements document, constructs a decision tree and a dynamic requirements graph, combines the scenario requirements slices in the decision tree and the functional requirements slices in the dynamic requirements graph, constructs a functional knowledge graph representing the interaction relationship between the functional modules of the intelligent cockpit software, and then can generate test cases and test scripts to obtain the test results of the intelligent cockpit software, realizing the automated process of test case writing and test script generation, and improving the test efficiency of the intelligent cockpit software. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is an application scenario diagram provided by an embodiment of the present invention;

[0072] Figure 2 It is the flowchart of the test method for the intelligent cockpit software provided by an embodiment of the present invention Figure 1 ;

[0073] Figure 3 It is a schematic flowchart of the AI-driven method for the HIL test of the intelligent cockpit software provided by an embodiment of the present invention;

[0074] Figure 4 It is the flowchart of the test method for the intelligent cockpit software provided by an embodiment of the present invention Figure 2 ;

[0075] Figure 5 It is a schematic structural diagram of the test device for the intelligent cockpit software provided by an embodiment of the present invention;

[0076] Figure 6 It is a schematic structural diagram of the intelligent cockpit provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand the other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for explaining the present invention, rather than for limiting the protection scope of the present invention.

[0078] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Therefore, only the components related to the present invention are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0079] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present invention are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards, and corresponding operation entrances are provided for users to choose to authorize or reject.

[0080] Figure 1 This is an application scenario diagram provided by an embodiment of the present invention. As Figure 1 shown, with the rapid development of new energy vehicles, the complexity and functional diversity of intelligent cockpit software are increasing continuously. Among them, in order to ensure the quality of intelligent cockpit software, it is necessary to perform Hardware-in-the-Loop (HIL for short) testing on the intelligent cockpit software through a test platform.

[0081] In one example, the test steps corresponding to multiple test cases of the intelligent cockpit software are executed in parallel through a test bench, and the test data of the test bench is received for analysis to obtain the test results of the intelligent cockpit software. However, in the above method, it is necessary to manually write test cases, which takes a long time and results in low test efficiency of the intelligent cockpit software.

[0082] In view of this, the embodiments of the present invention propose a testing method for intelligent cockpit software. By performing requirement document recognition on the test requirement document of the intelligent cockpit software, constructing a decision tree and a dynamic requirement graph, and combining the scenario requirement slices in the decision tree and the function requirement slices in the dynamic requirement graph, a function knowledge graph representing the interaction relationship between the function modules of the intelligent cockpit software is constructed. Furthermore, test cases and test scripts can be generated to obtain the test results of the intelligent cockpit software, which can solve the problem of low test efficiency of the intelligent cockpit software.

[0083] Next, the technical solutions of the present invention will be described in detail through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0084] Figure 2 This is the flow of the testing method for intelligent cockpit software provided by an embodiment of the present invention Figure 1 As Figure 2 shown, the method includes:

[0085] 201. Obtain the test requirement document of the intelligent cockpit software; and generate a decision tree and a dynamic requirement graph corresponding to the test requirement document according to the test requirement document; wherein, the decision tree includes at least one scenario requirement shard; the dynamic requirement graph includes at least one functional requirement shard, and the dynamic requirement graph represents the change track of the functional test requirements.

[0086] Exemplarily, the execution entity of this embodiment may be an electronic device in the intelligent cockpit, hereinafter referred to as the device. The device obtains the test requirement document of the intelligent cockpit software, and slices the test requirement document through a requirement document slicing algorithm, that is, uses a Bidirectional Encoder Representations from Transformers (abbreviated as BERT) to identify the semantic structure of the test requirement document, and performs intelligent semantic segmentation on the test requirement document according to the semantic structure of the test requirement document to obtain multiple shards, and each shard corresponds to a semantic label with a shard, and these labels can be test scenario requirements or test functional requirements; according to the semantic labels of each shard, each shard is classified to obtain at least one scenario requirement shard representing test scenario requirements and at least one functional requirement shard representing test functional requirements. Based on the decision tree construction algorithm, all the scenario requirement shards are processed. The Classification and Regression Trees (abbreviated as CART) algorithm can be used to randomly select an initial scenario requirement shard from all the scenario requirement shards, and through the semantic label of the initial scenario requirement shard, all the remaining scenario requirement shards are repeatedly recursively classified until the stop condition is met. Each scenario requirement shard corresponds to a node to form a tree structure, and a decision tree is constructed. The decision tree includes all the scenario requirement shards, and the decision tree represents the association relationship between each scenario requirement shard. Based on the graph construction algorithm, all the functional requirement shards and the historical test requirement document are processed. Specifically, through a deep learning model constructed based on the Transformer architecture, the relationship extraction is performed on each functional requirement shard and each historical functional requirement shard in the historical test requirement document to identify the association relationship between each functional requirement shard and each historical functional requirement shard in pairs, and according to these association relationships, the two shards with an association relationship are linked to construct a dynamic requirement graph. The dynamic requirement graph includes all the functional requirement shards, and the dynamic requirement graph represents the change track of the functional test requirements.

[0087] For example, Figure 3 is a schematic flowchart of an AI-driven method for HIL testing of intelligent cockpit software provided by an embodiment of the present invention, as Figure 3As shown, the controller of the issued version test bench can compile the version through the automatic upgrade program, and then send the test requirement document of the intelligent cockpit software to the device. A system is deployed in the device. Based on the requirement recognition module in the system, an artificial intelligence (AI) model, namely a bidirectional encoder model, is used to identify the semantics and logic of the test requirement document. According to the identified semantics and logic, the test requirement document is further semantically segmented to identify the functional requirement slices corresponding to the software functional requirements, the performance requirements, compatibility requirements, etc. in the scenario requirements corresponding to the scenario requirement slices; for example, "voice interaction" and "gesture control" belong to the functional requirement slices, and "response time" and "resource occupancy rate" belong to the scenario requirement slices. Based on a general language model such as a bidirectional encoder model, a knowledge graph specific to the intelligent cockpit is incorporated. The knowledge graph specific to the intelligent cockpit includes a preset intelligent cockpit standard document, a preset historical requirement database, and a preset test case library, which are used to train the bidirectional encoder model so that the obtained model has the semantic understanding ability, decision tree construction ability, and graph construction ability unique to in-vehicle systems. Through the obtained model, all scenario requirement slices are processed to obtain a decision tree; and through the obtained model, all functional requirement slices and historical test requirement documents are processed to obtain a dynamic requirement graph.

[0088] 202. Generate a functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and the dynamic requirement graph corresponding to the test requirement document; wherein, the functional knowledge graph represents the functional structure and relationship of the intelligent cockpit software.

[0089] Exemplarily, the device can call a large language model (LLM) such as a bidirectional encoder model to perform semantic recognition processing on the scenario requirement slices in the constructed decision tree and the functional requirement slices in the dynamic requirement graph, and obtain an identification result including multiple key information such as function points, performance indicators, and constraint conditions. And through a graph neural network constructed based on the Transformer architecture, graph convolution processing is performed on the decision tree, the dynamic requirement graph, and the identification result to obtain the functional structure features representing each functional structure of the intelligent cockpit software and the relationship features representing the association relationships between the functional structure features, such as association relationships such as "belong to" and "affect". And according to these relationship features, link association processing is performed on the functional structure features with association relationships to obtain the functional knowledge graph corresponding to the intelligent cockpit software to represent the functional structure and relationship of the intelligent cockpit software.

[0090] For example, in combination with Figure 3, the decision tree and the dynamic requirement graph are used as input features for subsequent AI model recognition; technologies such as regular expressions and named entity recognition are used to extract features, and feature screening and optimization are carried out in combination with the domain knowledge base; the extracted features are input into the trained AI model for requirement understanding and recognition; then the pre-trained AI model is loaded using the deep learning framework to encode, decode, and classify the input features, and the recognition result is output; the recognition result of the AI model is verified and optimized to ensure the accuracy and integrity of the recognition result; in combination with domain knowledge and business rules, the recognition result is verified and optimized to obtain key information including function points, performance indicators, constraint conditions, etc.; the optimized recognition result is output to the test case writing module for processing through an interface or file transfer method to obtain a functional knowledge graph as input data for test case generation.

[0091] 203. Generate test cases for the intelligent cockpit software according to the functional knowledge graph corresponding to the intelligent cockpit software.

[0092] Exemplarily, based on the structured data output by the requirement recognition module, including the scenario requirement slices in the decision tree and the function requirement slices in the dynamic requirement graph, based on machine learning algorithms, such as the graph neural network model, the graph traversal algorithm (such as depth-first search) is used to process the functional knowledge graph corresponding to the intelligent cockpit software to obtain test cases for the intelligent cockpit software.

[0093] For example, in combination with Figure 3, based on machine learning algorithms, such as large language models, set the parameters and constraints of the model, that is, based on methods such as random search or Bayesian optimization, set parameters such as the hyperparameters and optimization algorithms of the bi-encoder model, as well as constraint conditions such as the loss function, to obtain a trained model. Through the trained model, identify the key information of the functional knowledge graph, such as keywords or tags, and process the key information of the functional knowledge graph, such as keywords or tags, to select a suitable test case template, where the suitable test case template is the test case template associated with the keyword or tag mapping among multiple preset test case templates. The test case template may include functional test cases corresponding to keywords related to functional testing, performance test cases corresponding to keywords related to performance testing, exception test cases corresponding to keywords related to exception testing, etc. The functional test cases are used to test whether the functions of the cockpit software work as expected, the performance test cases are used to test the metric data of the cockpit software under different load conditions, such as response time, throughput, and resource usage, and the exception test cases are used to test the behavior and robustness of the cockpit software under abnormal or error conditions. Fill the key information of the functional knowledge graph into the selected test case template, automatically generate test cases, and build a test case library. Among them, domain knowledge and business rules can be combined to evaluate, process, and optimize the test cases. For test cases that do not meet the requirements or have problems, they can be modified or regenerated.

[0094] 204. Generate and execute the test script of the intelligent cockpit software according to the test case of the intelligent cockpit software to obtain the test result of the intelligent cockpit software.

[0095] Exemplarily, based on the script generation module of the device, receive the test cases from the use case writing module. These test cases include key information such as test steps and expected results. According to the key information in the test cases, select a suitable test script template, where the suitable test script template is the test script template that matches the key information among multiple preset test script templates. The test script template may include automated test scripts, manual test scripts, etc.; fill the key information in the test cases into the selected test script template, perform script compilation, and automatically generate the test script. Based on the test execution and result analysis module, it is responsible for executing the HIL test script and collecting the test results. The test results include whether the functions of the intelligent cockpit software work as expected, whether the performance metric data under different load conditions is qualified, and the automated execution situation of the cockpit software under abnormal or error conditions. Through data analysis and visualization techniques, evaluate and display the test results to help testers quickly locate problems and optimize the test process.

[0096] For example, in combination with Figure 3, the device-based test execution and result analysis module receives the test tasks sent by the cloud platform through the HIL test bench, and uses a file parsing library or an Application Programming Interface (API) to parse the task content; the HIL test bench executes the test script according to the test task, monitors the script execution status, uses a logging framework to record the execution log, and uses a database or a file to store the test results; the HIL test bench detects problems using an assertion or an exception capture mechanism during the test, uses a report generation library or a custom template to generate a test report, and feeds back the test report to the cloud platform for display through an API or a file transfer method. Among them, the HIL test bench outputs the test results to a database or a file, and notifies relevant personnel by means of email, message queue, etc.; uses a database to store the test results, uses the Simple Mail Transfer Protocol (SMTP) to send emails, and uses a message queue to achieve asynchronous notification.

[0097] In this embodiment, a test method for intelligent cockpit software is provided. By identifying the requirements document of the intelligent cockpit software, constructing a decision tree and a dynamic requirements graph, and combining the scenario requirements slices in the decision tree and the function requirements slices in the dynamic requirements graph, a function knowledge graph representing the interaction relationship between the function modules of the intelligent cockpit software is constructed. Furthermore, test cases and test scripts can be generated to obtain the test results of the intelligent cockpit software, and an automated process for test case writing and test script generation can be realized, improving the test efficiency of the intelligent cockpit software and reducing the test cost.

[0098] Figure 4 The flowchart of the test method for intelligent cockpit software provided by an embodiment of the present invention Figure 2 , as Figure 4 shown, the method includes:

[0099] 301. Obtain the test requirements document of the intelligent cockpit software.

[0100] Exemplarily, this step can refer to step 201, which will not be elaborated here. After step 301, steps 302 and 304 can be executed synchronously.

[0101] 302. Perform semantic recognition processing on the test requirements document to construct a domain knowledge graph; where the domain knowledge graph includes at least one initial scenario requirement slice; the domain knowledge graph represents the association relationship between each initial scenario requirement slice.

[0102] Exemplarily, after step 301, based on the general language model, integrate the special knowledge graph of the intelligent cockpit to enable the model to have the semantic understanding ability unique to in-vehicle systems. The resulting model performs semantic recognition processing on the test requirement document, the historical requirement library, and the test case library to obtain at least one initial scenario requirement slice. Further process each initial scenario requirement slice to identify the association relationships between the initial scenario requirement slices, and link the associated initial scenario requirement slices to obtain the domain knowledge graph.

[0103] Among them, the requirement recognition module first receives the intelligent cockpit software test requirement document from the requirement input module. These documents may include functional requirements, performance requirements, user stories, etc. Receive the requirement document through the file upload interface or API interface, support multiple file formats, preprocess the received requirement document, including denoising, word segmentation, part-of-speech tagging, etc., for subsequent natural language processing, and use the natural language toolkit to perform text preprocessing to ensure the cleanliness and standardization of text data; extract key information from the preprocessed requirement document, such as requirement slices of function points, performance indicators, constraint conditions, etc., for constructing the domain knowledge graph.

[0104] In one example, step 302 includes the following steps:

[0105] The first step of step 302 is to perform semantic recognition processing on the test requirement document to obtain at least one initial scenario requirement slice; among them, the initial scenario requirement slice has attribute information.

[0106] The second step of step 302 is to construct an initial domain knowledge graph according to the attribute information of at least one initial scenario requirement slice; among them, the initial domain knowledge graph includes the attribute information of each initial scenario requirement slice and the association relationships between the initial scenario requirement slices.

[0107] The third step of step 302 is to perform semantic chain annotation on each initial scenario requirement slice in the initial domain knowledge graph to obtain the domain knowledge graph.

[0108] Exemplarily, through a dynamic semantic boundary recognition algorithm, the test requirement document is semantically recognized and processed to obtain at least one initial scenario requirement shard; wherein each initial scenario requirement shard has attribute information, such as belonging to a functional module, a performance metric, or a constraint condition; through the attribute information of each initial scenario requirement shard, the association relationship between each initial scenario requirement shard can be identified, and based on the knowledge graph construction algorithm, the attribute information of all initial scenario requirement shards and the association relationship between each initial scenario requirement shard are processed to generate an initial domain knowledge graph; a Bidirectional Long Short Term Memory-Conditional Random Field (BiLSTM-CRF) is used to perform entity chain annotation on the initial domain knowledge graph to obtain the domain knowledge graph.

[0109] For example, collect intelligent cockpit standard documents, historical requirement libraries, and test case libraries. First, perform entity definition. Entities belonging to functional modules include: voice interaction, gesture control, Human Machine Interface (HMI) interface, Advanced Driving Assistance System (ADAS) linkage, Over-the-Air Technology (OTA) upgrade, etc. Entities belonging to performance metrics include: response time, resource occupancy rate, compatibility range (such as -30°C to 85°C). Entities belonging to constraint conditions include: safety level, regulatory requirements. Then, perform entity relationship extraction. For example, define relationship types such as "belong to", "affect", "depend on", etc. For example, "voice wake-up success rate ≥ 95%" is associated with the "voice interaction" module and the "performance requirement" classification. During the semantic chain annotation process, nested entities can be identified, such as "navigation positioning accuracy < 1m under charging state" contains two entities, "charging state" and "navigation positioning accuracy"); and composite metrics are parsed, such as "system latency ≤ 300ms when multiple devices are connected simultaneously" is decomposed into two dimensions, "concurrent connection number" and "latency". Then, through the aforementioned knowledge graph construction algorithm, the final domain knowledge graph is obtained.

[0110] 303. Generate a decision tree corresponding to the test requirement document according to the domain knowledge graph; wherein the decision tree includes a root node and at least one leaf node; the root node represents the requirement type; the leaf node represents the final scenario requirement shard.

[0111] Exemplarily, based on the Transformer architecture combined with a domain adaptation pre-training model in the intelligent cockpit field, a trained model is obtained to process the initial scenario requirements slices in the domain knowledge graph, resulting in a root node and at least one leaf node. The root node represents the requirement type, and each leaf node represents the final scenario requirement slice. Through the association relationships between the leaf nodes, a decision tree can be obtained.

[0112] In one example, step 303 includes the following steps:

[0113] The first step of step 303 is to calculate the semantic entropy value for the semantic vector of each initial scenario requirement slice in the domain knowledge graph, obtaining an entropy value matrix corresponding to the domain knowledge graph. Among them, the entropy value matrix includes at least one semantic association entropy value. The semantic association entropy value represents the semantic similarity between every two adjacent initial scenario requirement slices.

[0114] The second step of step 303 is to determine a boundary threshold according to the document complexity of the test requirement document.

[0115] The third step of step 303 is to merge the two initial scenario requirement slices corresponding to the semantic association entropy value if it is determined that the semantic association entropy value is less than the boundary threshold, so as to obtain a new initial scenario requirement slice corresponding to the semantic association entropy value.

[0116] The fourth step of step 303 is to construct a decision tree corresponding to the test requirement document according to the association relationships between the initial scenario requirement slices.

[0117] Exemplarily, through the BERT-flow model, the sentence-level semantic vectors corresponding to each initial scenario requirement slice in the domain knowledge graph are mapped to the manifold space, and the semantic association entropy value between two adjacent semantic vectors is calculated to quantify the boundary fuzziness. For example, the semantic similarity between every two adjacent semantic vectors is calculated through BERT-flow as the semantic association entropy value, and then an entropy value matrix is sorted out. According to the document complexity of the test requirement document, the boundary threshold corresponding to this document complexity is determined through a preset mapping relationship. For example, the simple document threshold is 0.85 and the complex document threshold is 0.65. Among them, the document complexity of the test requirement document can be obtained based on the syntactic parser analyzing the syntactic complexity of the sentences in the test requirement document. If the adjacent sentence entropy value, that is, the semantic association entropy value between two adjacent semantic vectors < the boundary threshold, the corresponding two initial scenario requirement slices are merged into the same slice (semantically continuous) to obtain new initial scenario requirement slices. The domain knowledge graph is processed through a graph attention network to propagate constraint information such as preconditions and performance indicators, and propagate the association relationships or dependencies between each initial scenario requirement slice. For example, "Gesture control needs to respond within 1s after the screen is awakened" depends on the "Screen wake-up time" indicator, and then the final slice result is obtained, including at least one final scenario requirement slice and the association relationships between each scenario requirement slice. Through a decision tree construction algorithm, the at least one final scenario requirement slice and the association relationships between each scenario requirement slice are processed to obtain a final decision tree. Among them, the decision tree includes a root node: requirement type (function / performance / compatibility); it also includes branch conditions: constraint strength (such as security-related requirements being sliced first); it also includes leaf nodes: final slice result (such as aggregating "voice interaction"-related requirements into one slice).

[0118] 304. Preprocess the test requirement document to obtain at least one triple; where the triple represents the association relationship between functional text, table parameters, and chart elements.

[0119] Exemplarily, after step 301, through cross-modal semantic alignment and correlation modeling, the test requirement document is preprocessed to obtain mixed-modal data such as text, tables, charts, and schematic diagrams in the requirement document, and through a pre-trained model, joint embedding of text and images, tables, flowcharts, and user interface (UI) design diagrams is achieved to capture the implicit correlation between "text description - visual elements". Specifically, through the Contrastive Language–Image Pretraining (CLIP) model, format conversion processing and embedding vector generation processing are respectively performed on the input text and images, tables, flowcharts, and user interface (UI) design diagrams to obtain the embedding vectors of the text, images, tables, flowcharts, and UI design diagrams, and the similarity between the embedding vector of the text and other embedding vectors is calculated pairwise as the implicit correlation relationship between each text and each other diagram or table. The higher the similarity, the closer the implicit correlation. For example: Semantically align the text description of "voice interaction response time < 500ms" in the requirement document with the corresponding parameter row in the performance test table to convert the unstructured data corresponding to the corresponding parameter row in the performance test table into corresponding structured data. Store the correlation relationship between functional text, table parameters, and chart elements in the form of triples (such as "voice interaction, performance indicator, wake-up rate ≥ 95%").

[0120] In one example, step 304 includes the following steps:

[0121] The first step of step 304: Perform word segmentation and semantic encoding processing on each functional text in the test requirement document to obtain a representation set corresponding to each functional text; where the representation set includes at least one word segmentation representation of a functional word segment.

[0122] The second step of step 304: Parse and transform each table in the test requirement document to obtain the table parameters in each table.

[0123] The third step of step 304: Identify each chart in the test requirement document to obtain the chart elements in each chart.

[0124] The fourth step of step 304: Perform semantic alignment processing on each word segmentation representation, each table parameter, and each chart element to obtain at least one triple.

[0125] Exemplarily, use the Bidirectional Encoder Representations from Transformers (BERT)-Whole Word Masking (WWM) model to perform word segmentation and semantic encoding on each functional text in the test requirement document, obtaining a set of representations corresponding to each functional text; where the set of representations includes at least one word segmentation representation of a functional word segment; based on the document intelligence technology that combines Optical Character Recognition (OCR) and a pre-trained model, parse and transform each table in the test requirement document to obtain the table parameters in each table, and based on the object detection model and the graph neural network, perform recognition processing on each chart in the test requirement document to obtain the chart elements in each chart, and use a multi-task learning framework to perform cross-modal semantic alignment processing on the respective word segmentation representations, the respective table parameters, and the respective chart elements to obtain at least one triple.

[0126] For example, use the BERT-Whole Word Masking model to perform word segmentation and semantic encoding, retain professional terms (such as "CAN bus"), and identify the constraints in the requirements (such as "when the vehicle speed > 120 km / h, the intervention priority is the highest"). Use OCR + pre-trained model to identify the compatibility test matrix (such as the operating system version, hardware platform support list), and convert the table data into structured knowledge (such as "in-vehicle navigation supports versions 12.0 and above"). Use the object detection model to identify the controls (buttons, menu levels) in the User Interface Design (UI) design diagram, and model the user operation process through the graph neural network (such as "click the voice button → wake-up word recognition → execute the instruction"). Construct a text-image joint loss function to force the model to learn the shared semantic space of "text description" and "visual elements", for example: align the text of "gesture control sensitivity" with the screenshot of the gesture recognition area in the UI design diagram. Use a multi-task learning framework to synchronously extract text entities (such as "voice wake-up rate"), table parameters (such as "≥95%"), and chart elements (such as "microphone array position"). Finally, store the knowledge in the form of triples (such as <voice interaction, performance indicator, wake-up rate ≥ 95%>).

[0127] 305. Construct a global requirement graph based on the requirement change history data of the intelligent cockpit software and at least one triple; where the global requirement graph includes at least one initial functional requirement shard; the global requirement graph represents the initial change trajectory of the functional test requirements.

[0128] Exemplarily, based on a preset time-series graph construction algorithm, the requirement change historical data (such as new functions, modified metrics) and all the obtained triples are processed. Each new function text or each new chart element associated with each triple is identified from the requirement change historical data, and each triple is associated and connected with each new function text or each new chart element associated with the triple to construct a global requirement graph. The global requirement graph includes all the initial functional requirement fragments in the triples and the requirement change historical data. At the same time, the global requirement graph represents the initial change trajectory of the functional test requirements to record the evolution trajectory of the requirement elements (such as the parameter change of the "gesture control" function from V1.0 to V2.0).

[0129] In one example, step 305 includes the following steps:

[0130] The first step of step 305: construct an initial subgraph corresponding to each document chapter in the test requirement document according to the requirement change historical data of the intelligent cockpit software and at least one triple; wherein, the initial subgraph includes at least one subgraph node and at least one subgraph edge; the subgraph node represents the functional text, or table, or flowchart data in each document chapter; the subgraph edge represents the association relationship between the subgraph nodes.

[0131] The second step of step 305: determine the semantic association degree between every two initial subgraphs based on the attention mechanism.

[0132] The third step of step 305: construct a global requirement graph according to each initial subgraph and each semantic association degree.

[0133] Exemplarily, taking each document chapter in the test requirement document as a unit, process the requirement change history data of the in-vehicle infotainment software and all triples, and combine the association relationships between various functional texts, table parameters, and chart elements in each triple to construct an initial sub-graph corresponding to each document chapter (such as "functional requirement sub-graph", "performance requirement sub-graph"), where the initial sub-graph includes at least one sub-graph node and at least one sub-graph edge; each sub-graph node represents a functional text, a table, or a flowchart data in each document chapter; each sub-graph edge represents the association relationship between the sub-graph nodes. For example, map the text, table, and flowchart data in the "voice interaction" chapter into sub-graph nodes and sub-graph edges. Call a preset attention mechanism, and through this attention mechanism, perform semantic association calculation on every two initial sub-graphs to calculate the semantic association degree between the initial sub-graphs. For example, "response time" in the "performance requirement sub-graph" is associated with "hardware platform" in the "compatibility requirement sub-graph". According to each initial sub-graph and the semantic association degree between each initial sub-graph, associate the initial sub-graphs with association relationships to construct a global requirement graph to support cross-shard requirement tracing. For example, trace back from "navigation positioning accuracy" to "sensor calibration process".

[0134] 306. Perform graph pruning processing on the global requirement graph to obtain a dynamic requirement graph.

[0135] Exemplarily, call a predefined typical test scenario template for the in-vehicle infotainment (such as "extreme temperature test", "multi-device concurrent connection"), and through a graph matching algorithm, screen out the requirement fragments strongly related to the scenario, that is, perform graph pruning processing on the global requirement graph to obtain a dynamic requirement graph. For example, in the "charging scenario test", automatically associate cross-modal requirements such as "power consumption limit of the entertainment system during charging" and "communication protocol of the battery management system".

[0136] In one example, step 306 includes the following steps:

[0137] The first step of step 306: Determine the similarity between each initial functional requirement shard in the preset scenario template and the global requirement graph.

[0138] The second step of step 306: If it is determined that the similarity is greater than the preset threshold, determine that there is an association relationship between the preset scenario template corresponding to the similarity and the initial functional requirement shard.

[0139] The third step of step 306: According to the test priority corresponding to the initial functional requirement shard and the association relationship between the preset scenario template and the initial functional requirement shard, perform hierarchical pruning processing on the global requirement graph to obtain a dynamic requirement graph.

[0140] Exemplarily, the device obtains a preset scenario template. For example, "distracted driving monitoring" includes requirement segments: "driver fatigue detection", "steering wheel grip monitoring", "voice alarm trigger". Through the Siamese network, calculate the similarity between each initial functional requirement segment in the global requirement graph and the preset scenario template. When the similarity is greater than the preset threshold, it is determined that there is an association relationship between the preset scenario template corresponding to the similarity and the initial functional requirement segment. For example, cosine similarity > 0.75, trigger the association, and calculate the test priority corresponding to the initial functional requirement segment. For example, safety-related > functional integrity > performance optimization. According to the test priority corresponding to each initial functional requirement segment and the association relationship between each preset scenario template and the initial functional requirement segment, perform hierarchical pruning on the global requirement graph to obtain a dynamic requirement graph. For example, in the "safety test special project", retain the initial functional requirement segments related to "emergency braking" and prune the initial functional requirement segments of non-critical functions to obtain the final functional requirement segments.

[0141] 307. Construct an initial functional knowledge graph corresponding to the test requirement document according to the decision tree and dynamic requirement graph corresponding to the test requirement document, and the test record data of the intelligent cockpit software; the initial functional knowledge graph includes at least one functional node and at least one functional edge; the functional node is used to represent the functional module of the intelligent cockpit software, and the functional edge represents the interaction relationship between two functional nodes.

[0142] Exemplarily, obtain and integrate multi-source data of the intelligent cockpit, including the decision tree and dynamic requirement graph obtained through the test requirement document, and the test record data of the intelligent cockpit software. The test record data includes historical test data, user operation logs, sensor data, and vehicle communication protocols such as Controller Area Network (CAN) bus data. Use a joint extraction model such as an end-to-end model based on Bidirectional Encoder Representations from Transformers (BERT) to process the multi-source data to obtain the corresponding initial functional knowledge graph; the initial functional knowledge graph includes at least one functional node and at least one functional edge; each functional node is used to represent the functional module of the intelligent cockpit software, and each functional edge represents the interaction relationship between two functional nodes.

[0143] For example, entities (such as "voice control module", "navigation system") and relationships (such as "dependency", "call") are extracted from unstructured text in multi-source data, and a schema layer is constructed by combining design rules; for example, the dependency relationship between "voice wake-up function" and "voice recognition module" is identified from software requirement documents. Through entity alignment and semantic similarity calculation, knowledge from different sources (such as user behavior data and protocol documents) is fused, that is, through the identified dependency relationships, entities identified in unstructured text are aligned with structured text in multi-source data, and semantic similarity calculation is performed on the embedding vectors corresponding to each aligned entity. Multiple entities from different sources with semantic similarity within a preset range are fused to obtain functional knowledge. Each entity in the functional knowledge corresponds to a functional node, and the inter-entity relationship between every two entities corresponds to each functional edge. Then, a graph is constructed, updated functional knowledge is obtained, and the graph is updated in real time according to the updated functional knowledge to reflect software version iteration or user behavior changes, so as to obtain the initial functional knowledge graph corresponding to the intelligent cockpit software to represent the functional structure and relationships of the intelligent cockpit software. For example, it can structurally represent the functional modules, interfaces, interaction logics and their dependencies of the intelligent cockpit software.

[0144] Among them, the multi-source data is converted into a format supported by the graph database. Through data cleaning and normalization, redundant fields (such as debug information in logs) are removed, terms are standardized (such as unifying "GPS module" into "navigation and positioning system"), and data consistency is verified through a rule engine.

[0145] 308. Based on the graph neural network, the initial functional knowledge graph corresponding to the test requirement document is processed to generate at least one predicted test path and at least one predicted test scenario; among them, the predicted test path includes the key functions of the target intelligent cockpit software.

[0146] Exemplarily, using the graph neural network, the functional nodes (functional modules) and functional edges (dependency relationships) in the initial functional knowledge graph are embedded and represented. Through a graph traversal algorithm (such as depth-first search), at least one predicted test path covering the key functions is generated. For example, an interaction chain of "navigation system startup → voice interaction → map update" is predicted. Reinforcement learning is used to simulate abnormal user operations (such as continuously triggering voice commands multiple times), and the model is trained in combination with historical failure data to process each predicted test path and identify the corresponding predicted test scenario, such as a high-risk scenario (such as resource competition during concurrent multi-module calls).

[0147] Among them, during the use of the model, for the cockpit software variants of different vehicle models, the model parameters can be adjusted through transfer learning to reuse the core test logic and reduce repeated modeling.

[0148] Specifically, based on a probability model, a probability set corresponding to the intelligent cockpit software is determined, and the probability set includes the triggering probabilities of different test function combinations in the intelligent cockpit software; based on a graph neural network, the initial function knowledge graph, historical test cases, and user behavior pattern data of the intelligent cockpit software are calculated and processed to obtain at least one initial predicted test path; each initial predicted test path includes at least one key test function of the target intelligent cockpit software; calculate the coverage rate and the corresponding triggering probability of each initial predicted test path, and determine the priority of each initial predicted test path according to the coverage rate and the corresponding triggering probability of each initial predicted test path; determine the initial predicted test path with the highest priority as the predicted test path; and generate a predicted test scenario corresponding to the predicted test path according to the predicted test path.

[0149] For example, the input data includes a functional dependency graph (initial function knowledge graph), historical test cases, and user behavior patterns (such as high-frequency operation paths); based on a Markov chain or a Bayesian network, the triggering probabilities of different function combinations are calculated. For example, after the user starts the navigation, there is an 80% probability of triggering voice search. Based on the coverage priority strategy, a greedy algorithm is used to select the initial predicted test path that covers the most key nodes, and the test order of the test functions in the initial predicted test path is optimized by a genetic algorithm to reduce the execution time, resulting in the final predicted test path. Extreme conditions (such as low battery state, network latency) are inserted into the predicted test path to simulate abnormal scenarios in a real environment, obtaining the predicted test scenario.

[0150] 309. Optimize the initial function knowledge graph corresponding to the test requirement document according to at least one predicted test path and at least one predicted test scenario to obtain the function knowledge graph corresponding to the intelligent cockpit software.

[0151] Exemplarily, based on the real-time feedback closed-loop process, all the obtained predicted test paths and all the predicted test scenarios can be pre-executed to obtain the test execution results, and the test execution results (such as coverage rate, defect distribution) are fed back to the initial function knowledge graph. For example, the weights of the function nodes in the initial function knowledge graph are adjusted (such as high-defect modules are tested first), and combined with image recognition (UI interface screenshots) and voice interaction logs, the association relationship of "user interface → voice command" in the graph is supplemented to obtain the final function knowledge graph.

[0152] 310. Generate test cases for the intelligent cockpit software according to the function knowledge graph corresponding to the intelligent cockpit software.

[0153] Exemplarily, this step can refer to step 202 and will not be elaborated here.

[0154] 311. Generate an initial test script for the intelligent cockpit software based on the test cases, interface information, and test environment information of the intelligent cockpit software.

[0155] Exemplarily, in combination with Figure 3 , use an AI script generator to generate an efficient and executable test script according to the test cases in the use case library, in combination with the interface information and test environment of the intelligent cockpit software. And generate a script library to support the classification management, version control, and intelligent retrieval of scripts.

[0156] Among them, based on the verification mechanism of test cases, the verification of test cases needs to ensure their logical correctness, executability, and adaptability to the target environment.

[0157] 312. Verify the script information of the initial test script of the intelligent cockpit software to obtain a verification result; among them, the script information includes script syntax, interface information, integrity of the test path, trigger coverage, and environmental adaptability.

[0158] Exemplarily, according to the preset verification mechanism of test scripts, verify the script information of the initial test script of the intelligent cockpit software to obtain a verification result; among them, the script information includes script syntax, interface information, integrity of the test path, trigger coverage, and environmental adaptability.

[0159] For example, based on the static verification mechanism, check for syntax and interface compliance. For example, verify whether the script syntax conforms to the target language specification through a static analysis tool (such as a parser based on the Abstract Syntax Tree (AST)), and check whether the interface calls match the API definitions of the intelligent cockpit (such as in-vehicle entertainment system interfaces, CAN bus communication protocols); verify the logical integrity. Based on the dependency relationship of the functional knowledge graph, verify whether the predicted test path in the script covers the key functional links (such as the complete interaction chain of "voice wake-up → navigation settings → screen feedback"), and detect whether there are unhandled exception branches (such as the recovery logic after a network interruption). Based on the dynamic verification mechanism, for simulation execution and coverage analysis, the test script can be pre-executed in a virtual test environment, and the code coverage (such as branch coverage) can be statistically analyzed through instrumentation technology to ensure that all preset conditions are triggered by the test cases; for environmental adaptability verification, verify whether the test script is compatible with different hardware configurations (such as screen resolution, voice chip model, etc.) and software versions to avoid execution failures due to environmental differences.

[0160] 313. Optimize the initial test script of the intelligent cockpit software according to the verification result to obtain the test script of the intelligent cockpit software.

[0161] Exemplarily, based on the verification result, the optimization strategy optimizes the script information in the initial test script of the intelligent cockpit software according to the verification result to obtain the test script of the intelligent cockpit software.

[0162] For example, based on the defect repair and logic complement mechanism, for error location and correction, according to the error reporting logs of dynamic verification (such as interface call timeout, assertion failure), combined with the functional knowledge graph to trace the defect nodes in the functional link (such as missing dependency module initialization steps), automatically generate patches or prompt manual repair; for coverage complement, if the coverage rate does not meet the standard, the reinforcement learning model can be used to analyze the uncovered paths in the test script (such as the low battery mode in the cold start scenario), automatically generate supplementary test cases and update the test script. Based on the efficiency and stability improvement mechanism, for redundant step elimination, data flow analysis is used to identify duplicate operations in the test script (such as initializing the same module multiple times), and redundant steps are merged to shorten the execution time. For concurrent optimization, for multi-module interaction scenarios (such as executing voice commands and touch screen operations simultaneously), the thread scheduling strategy of the test script is adjusted to avoid deadlocks or timeouts caused by resource competition. Based on the environment adaptability enhancement mechanism, for the parameterized configuration template, according to vehicle model differences (such as screen size, operating system version), the environment variables in the test script (such as resolution parameters, API addresses) are abstracted into configuration files to support dynamic adaptation; for version compatibility extension, the logic of the test script can be adjusted through transfer learning to make it compatible with new and old version interfaces (such as the mapping of the old CAN protocol identifier to the new version). At the same time, the key information dependencies in the test script can also be optimized, including execution logs and performance data, such as script execution time, peak resource occupancy, interface response latency, etc., for locating performance bottlenecks; it also includes coverage reports, such as identifying uncovered code branches or functional paths to guide the extension of test cases or test scripts; it also includes defect databases, such as historical defect records (such as interface timeout frequency, common exception types) for training anomaly prediction models; it also includes environment configuration libraries, such as hardware parameters, software versions, and communication protocols of different vehicle models, to support the dynamic adaptation and optimization of test scripts.

[0163] Taking another example, the verification and optimization instance is the voice interaction test script. When dynamically executed, it is found that "the voice wake-up success rate is lower than the threshold (90%)", and the log shows that "VoiceAPI.init() is not correctly called in the low battery mode". The initialization logic for the low battery scenario is supplemented according to the functional knowledge graph, and the battery status detection condition is inserted into the voice interaction test script. The wake-up word length limit of different vehicle models is adapted through parameterized configuration to avoid API errors caused by overly long instructions.

[0164] In this embodiment, based on the AI-driven method for intelligent cockpit software HIL testing, on the basis of the above embodiment, by introducing artificial intelligence technology, an automated process from requirement identification to test case writing and script generation is realized, improving test efficiency and quality and reducing test costs.

[0165] Figure 5 The following is a schematic structural diagram of a test device for intelligent cockpit software provided by an embodiment of the present invention. As Figure 5 shown, the device includes:

[0166] A first generation module 401, configured to obtain a test requirement document of the intelligent cockpit software; and generate a decision tree and a dynamic requirement graph corresponding to the test requirement document according to the test requirement document; wherein, the decision tree includes at least one scenario requirement slice; the dynamic requirement graph includes at least one functional requirement slice, and the dynamic requirement graph represents the change track of the functional test requirements.

[0167] A second generation module 402, configured to generate a functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and the dynamic requirement graph corresponding to the test requirement document; wherein, the functional knowledge graph represents the functional structure and relationship of the intelligent cockpit software.

[0168] A third generation module 403, configured to generate test cases for the intelligent cockpit software according to the functional knowledge graph corresponding to the intelligent cockpit software.

[0169] A test module 404, configured to generate and execute a test script for the intelligent cockpit software according to the test cases of the intelligent cockpit software to obtain a test result of the intelligent cockpit software.

[0170] Further, the first generation module 401 is specifically configured to: perform semantic recognition processing on the test requirement document to construct a domain knowledge graph; wherein, the domain knowledge graph includes at least one initial scenario requirement slice; the domain knowledge graph represents the association relationship between each initial scenario requirement slice; generate a decision tree corresponding to the test requirement document according to the domain knowledge graph; wherein, the decision tree includes a root node and at least one leaf node; the root node represents the requirement type; the leaf node represents the final scenario requirement slice.

[0171] Further, the first generation module 401 is specifically configured to: perform semantic recognition processing on the test requirement document to obtain at least one initial scenario requirement slice; wherein, the initial scenario requirement slice has attribute information; construct an initial domain knowledge graph according to the attribute information of at least one initial scenario requirement slice; wherein, the initial domain knowledge graph includes the attribute information of each initial scenario requirement slice and the association relationship between each initial scenario requirement slice; perform semantic chain annotation on each initial scenario requirement slice in the initial domain knowledge graph to obtain a domain knowledge graph.

[0172] Further, the first generation module 401 is further specifically configured to: calculate the semantic entropy value for the semantic vector of each initial scenario requirement slice in the domain knowledge graph to obtain an entropy value matrix corresponding to the domain knowledge graph; wherein, the entropy value matrix includes at least one semantic association entropy value; the semantic association entropy value represents the semantic similarity between every two adjacent initial scenario requirement slices; determine a boundary threshold according to the document complexity of the test requirement document; if it is determined that the semantic association entropy value is less than the boundary threshold, merge the two initial scenario requirement slices corresponding to the semantic association entropy value to obtain a new initial scenario requirement slice corresponding to the semantic association entropy value; construct a decision tree corresponding to the test requirement document according to the association relationship between the initial scenario requirement slices.

[0173] Further, the first generation module 401 is further specifically configured to: preprocess the test requirement document to obtain at least one triple; wherein, the triple represents the association relationship between the function text, the table parameter and the chart element; construct a global requirement graph according to the requirement change history data of the intelligent cockpit software and at least one triple; wherein, the global requirement graph includes at least one initial function requirement slice; the global requirement graph represents the initial change track of the function test requirement; perform graph pruning processing on the global requirement graph to obtain a dynamic requirement graph.

[0174] Further, the first generation module 401 is further specifically configured to: perform word segmentation and semantic encoding processing on each function text in the test requirement document to obtain a representation set corresponding to each function text; wherein, the representation set includes at least one word segmentation representation of the function word segmentation; parse and transform each table in the test requirement document to obtain the table parameters in each table; identify and process each chart in the test requirement document to obtain the chart elements in each chart; perform semantic alignment processing on the word segmentation representations, the table parameters and the chart elements to obtain at least one triple.

[0175] Further, the first generation module 401 is further specifically configured to: construct an initial subgraph corresponding to each document chapter in the test requirement document according to the requirement change history data of the intelligent cockpit software and at least one triple; wherein, the initial subgraph includes at least one subgraph node and at least one subgraph edge; the subgraph node represents the function text, or the table, or the flowchart data in each document chapter; the subgraph edge represents the association relationship between the subgraph nodes; determine the semantic association degree between every two initial subgraphs based on the attention mechanism; construct a global requirement graph according to the initial subgraphs and the semantic association degrees.

[0176] Further, the first generation module 401 is further specifically configured to: determine the similarity between each initial functional requirement slice in the preset scenario template and the global requirement graph; if it is determined that the similarity is greater than the preset threshold, determine that there is an association relationship between the preset scenario template corresponding to the similarity and the initial functional requirement slice; perform hierarchical pruning processing on the global requirement graph according to the test priority corresponding to the initial functional requirement slice and the association relationship existing between the preset scenario template and the initial functional requirement slice, so as to obtain a dynamic requirement graph.

[0177] Further, the second generation module 402 is specifically configured to: construct an initial functional knowledge graph corresponding to the test requirement document according to the decision tree corresponding to the test requirement document, the dynamic requirement graph, and the test record data of the intelligent cockpit software; the initial functional knowledge graph includes at least one functional node and at least one functional edge; the functional node is used to represent the functional module of the intelligent cockpit software, and the functional edge represents the interaction relationship between two functional nodes; based on the graph neural network, process the initial functional knowledge graph corresponding to the test requirement document to generate at least one predicted test path and at least one predicted test scenario; wherein; the key functions of the target intelligent cockpit software are included in the predicted test path; optimize the initial functional knowledge graph corresponding to the test requirement document according to at least one predicted test path and at least one predicted test scenario, so as to obtain the functional knowledge graph corresponding to the intelligent cockpit software.

[0178] Further, the test module 404 is specifically configured to: generate an initial test script for the intelligent cockpit software according to the test cases, interface information, and test environment information of the intelligent cockpit software; verify the script information of the initial test script for the intelligent cockpit software to obtain a verification result; wherein, the script information includes script syntax, interface information, integrity of the test path, trigger coverage, and environment adaptability; optimize the initial test script for the intelligent cockpit software according to the verification result to obtain the test script for the intelligent cockpit software.

[0179] The device in this embodiment can execute the technical solutions in the above method, and the specific implementation process and technical principle are the same, which will not be elaborated here.

[0180] Figure 6 The structural schematic diagram of the intelligent cockpit provided by an embodiment of the present invention is as Figure 6 shown, and the intelligent cockpit may include: at least one processor 501 and a memory 502.

[0181] The memory 502 is used to store a program. Specifically, the program may include program code, and the program code includes computer execution instructions.

[0182] The memory 502 may include high-speed random access memory (RAM), and may also include non-volatile memory, such as at least one disk memory.

[0183] The processor 501 is configured to execute the computer-executable instructions stored in the memory 502 to implement the test method of the intelligent cockpit software described in the foregoing method embodiments. Among them, the processor 501 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0184] Optionally, the intelligent cockpit may further include a receiver 503 and a transmitter 504. In a specific implementation, if the receiver 503, the transmitter 504, the memory 502, and the processor 501 are implemented independently, the receiver 503, the transmitter 504, the memory 502, and the processor 501 may be interconnected through a bus and communicate with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc., but it does not mean that there is only one bus or one type of bus.

[0185] Optionally, in a specific implementation, if the receiver 503, the transmitter 504, the memory 502, and the processor 501 are integrated on a chip, the receiver 503, the transmitter 504, the memory 502, and the processor 501 may communicate through an internal interface.

[0186] The present invention also provides a vehicle, including an intelligent cockpit, which is configured to execute the method in the foregoing embodiments.

[0187] The present invention also provides a computer-readable storage medium, in which computer program instructions are stored. When the processor executes the computer program instructions, the solutions in the foregoing embodiments are implemented.

[0188] The present invention also provides a computer program product, including a computer program, which implements the solutions in the foregoing embodiments when executed by a processor.

[0189] The above-mentioned computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium accessible by a general-purpose or special-purpose computer.

[0190] An exemplary readable storage medium is coupled to the processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application-specific integrated circuit. Of course, the processor and the readable storage medium can also exist as discrete components in an intelligent driving application running device.

[0191] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including the above method embodiments; and the foregoing storage medium includes various media such as ROM, RAM, magnetic disk or optical disk that can store program codes.

[0192] Finally, it should be noted that the above embodiments are only preferred embodiments given to fully illustrate the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or transformations made by those skilled in the art on the basis of the present invention are all within the protection scope of the present invention.

Claims

1. A testing method for intelligent cockpit software, characterized in that, The method includes: Obtaining a test requirement document for the intelligent cockpit software; and generating a decision tree and a dynamic requirement graph corresponding to the test requirement document according to the test requirement document; wherein, the decision tree includes at least one scenario requirement slice; the dynamic requirement graph includes at least one functional requirement slice, and the dynamic requirement graph represents the change track of functional test requirements; Generating a functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and the dynamic requirement graph corresponding to the test requirement document; wherein, the functional knowledge graph represents the functional structure and relationships of the intelligent cockpit software; Generating test cases for the intelligent cockpit software according to the functional knowledge graph corresponding to the intelligent cockpit software; Generating and executing a test script for the intelligent cockpit software according to the test cases for the intelligent cockpit software to obtain a test result for the intelligent cockpit software.

2. The method according to claim 1, characterized in that, The generating the decision tree corresponding to the test requirement document according to the test requirement document includes: Performing semantic recognition processing on the test requirement document to construct a domain knowledge graph; wherein, the domain knowledge graph includes at least one initial scenario requirement slice; the domain knowledge graph represents the association relationships between the respective initial scenario requirement slices; Generating the decision tree corresponding to the test requirement document according to the domain knowledge graph; wherein, the decision tree includes a root node and at least one leaf node; the root node represents the requirement type; the leaf node represents the final scenario requirement slice.

3. The method according to claim 2, wherein The performing semantic recognition processing on the test requirement document to construct a domain knowledge graph includes: Performing semantic recognition processing on the test requirement document to obtain at least one initial scenario requirement slice; wherein, the initial scenario requirement slice has attribute information; Constructing an initial domain knowledge graph according to the attribute information of the at least one initial scenario requirement slice; wherein, the initial domain knowledge graph includes the attribute information of each initial scenario requirement slice and the association relationships between the respective initial scenario requirement slices; Performing semantic chain annotation on each initial scenario requirement slice in the initial domain knowledge graph to obtain the domain knowledge graph.

4. The method according to claim 2, wherein The generating the decision tree corresponding to the test requirement document according to the domain knowledge graph includes: Calculating the semantic entropy value of the semantic vector of each initial scenario requirement slice in the domain knowledge graph to obtain an entropy value matrix corresponding to the domain knowledge graph; wherein, the entropy value matrix includes at least one semantic association entropy value; the semantic association entropy value represents the semantic similarity between every two adjacent initial scenario requirement slices; Determining a boundary threshold according to the document complexity of the test requirement document; If it is determined that the semantic association entropy value is less than the boundary threshold, then merging the two initial scenario requirement slices corresponding to the semantic association entropy value to obtain a new initial scenario requirement slice corresponding to the semantic association entropy value; Constructing the decision tree corresponding to the test requirement document according to the association relationships between the respective initial scenario requirement slices.

5. The method according to claim 1, wherein Generating a dynamic requirement graph corresponding to the test requirement document according to the test requirement document, including: Preprocessing the test requirement document to obtain at least one triple; wherein, the triple represents the association relationship between functional text, table parameters, and chart elements; Constructing a global requirement graph according to the requirement change history data of the intelligent cockpit software and the at least one triple; wherein, the global requirement graph includes at least one initial functional requirement shard; the global requirement graph represents the initial change track of functional test requirements; Performing graph pruning processing on the global requirement graph to obtain the dynamic requirement graph.

6. The method according to claim 5, wherein The preprocessing the test requirement document to obtain at least one triple includes: Performing word segmentation and semantic encoding processing on each functional text in the test requirement document to obtain a representation set corresponding to each functional text; wherein, the representation set includes at least one word segmentation representation of a functional word segment; Parsing and transforming each table in the test requirement document to obtain table parameters in each table; Performing recognition processing on each chart in the test requirement document to obtain chart elements in each chart; Performing semantic alignment processing on each word segmentation representation, each table parameter, and each chart element to obtain the at least one triple.

7. The method according to claim 5, characterized in that The constructing a global requirement graph according to the requirement change history data of the intelligent cockpit software and the at least one triple includes: Constructing an initial subgraph corresponding to each document chapter in the test requirement document according to the requirement change history data of the intelligent cockpit software and the at least one triple; wherein, the initial subgraph includes at least one subgraph node and at least one subgraph edge; the subgraph node represents functional text, or a table, or flowchart data in each document chapter; the subgraph edge represents the association relationship between each subgraph node; Determining the semantic association degree between every two of the initial subgraphs based on the attention mechanism; Constructing the global requirement graph according to each initial subgraph and each semantic association degree.

8. The method according to claim 5, wherein The performing graph pruning processing on the global requirement graph to obtain the dynamic requirement graph includes: Determining the similarity between the preset scenario template and each initial functional requirement shard in the global requirement graph; If it is determined that the similarity is greater than the preset threshold, determining that there is an association relationship between the preset scenario template corresponding to the similarity and the initial functional requirement shard; Performing hierarchical pruning processing on the global requirement graph according to the test priority corresponding to the initial functional requirement shard and the association relationship existing between the preset scenario template and the initial functional requirement shard to obtain the dynamic requirement graph.

9. The method according to claim 1, wherein The generating a functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and the dynamic requirement graph corresponding to the test requirement document includes: Construct an initial functional knowledge graph corresponding to the test requirement document according to the decision tree and dynamic requirement graph corresponding to the test requirement document, and the test record data of the intelligent cockpit software; at least one functional node and at least one functional edge are included in the initial functional knowledge graph; the functional node is used to represent the functional module of the intelligent cockpit software, and the functional edge represents the interaction relationship between two functional nodes. Based on the graph neural network, process the initial functional knowledge graph corresponding to the test requirement document to generate at least one predicted test path and at least one predicted test scenario; wherein, the key functions of the target intelligent cockpit software are included in the predicted test path. Optimize the initial functional knowledge graph corresponding to the test requirement document according to the at least one predicted test path and the at least one predicted test scenario to obtain the functional knowledge graph corresponding to the intelligent cockpit software.

10. The method according to any one of claims 1-9, characterized in that, The generation of the test script of the intelligent cockpit software according to the test case of the intelligent cockpit software includes: Generate an initial test script of the intelligent cockpit software according to the test case, interface information and test environment information of the intelligent cockpit software. Verify the script information of the initial test script of the intelligent cockpit software to obtain a verification result; wherein, the script information includes script syntax, interface information, integrity of the test path, trigger coverage rate and environmental adaptability. Optimize the initial test script of the intelligent cockpit software according to the verification result to obtain the test script of the intelligent cockpit software.

11. A test device for intelligent cockpit software, characterized in that, The device includes: A first generation module, configured to obtain a test requirement document of an intelligent cockpit software; and generate a decision tree and a dynamic requirement graph corresponding to the test requirement document according to the test requirement document; wherein, the decision tree includes at least one scenario requirement slice; the dynamic requirement graph includes at least one functional requirement slice, and the dynamic requirement graph represents the change track of functional test requirements. A second generation module, configured to generate a functional knowledge graph corresponding to the intelligent cockpit software according to the decision tree and the dynamic requirement graph corresponding to the test requirement document; wherein, the functional knowledge graph represents the functional structure and relationship of the intelligent cockpit software. A third generation module, configured to generate a test case of the intelligent cockpit software according to the functional knowledge graph corresponding to the intelligent cockpit software. A test module, configured to generate and execute a test script of the intelligent cockpit software according to the test case of the intelligent cockpit software to obtain a test result of the intelligent cockpit software.

12. An intelligent cockpit, characterized in that, Including: A memory, a processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory, so that the processor executes the method according to any one of claims 1-10.

13. A vehicle, characterized in that, The vehicle includes an intelligent cockpit; the intelligent cockpit is used to execute the method according to any one of claims 1-10.

14. A computer-readable storage medium, characterized in that, Computer execution instructions are stored in the computer-readable storage medium, and when the computer execution instructions are executed by a processor, they are used to implement the method according to any one of claims 1-10.

Images