Intelligent simulation method, system and equipment for test scene of real-time control system and medium

Abstract

The invention discloses a test scene intelligent simulation method, system and device for a real-time control system and a medium, and the method comprises the steps: receiving a test demand, carrying out the analysis of the test demand based on a large language model and a test domain knowledge graph, and generating a structured test demand; calling a corresponding generative artificial intelligence model, and generating a cross-modal test data set associated in time sequence and semantics; based on a structured test requirement, a cross-modal test data set and a test domain knowledge graph rule, automatically generating a test script and arranging a time sequence test scene process so as to drive the real-time control system and collect response of the real-time control system; and analyzing a test execution result, optimizing decision logic of data generation or protocol simulation through reinforcement learning based on the analysis result, and updating the decision logic to the test domain knowledge graph. Compared with the prior art, full-link automation of real-time control system testing from scene construction and execution to autonomous evolution is achieved.

Description

Technical Field

[0001] This invention relates to the field of real-time control system testing and simulation technology, and in particular to an intelligent simulation method, system, device and medium for testing scenarios of real-time control systems. Background Technology

[0002] The functional reliability and safety of real-time control systems are directly related to equipment operational safety, and their testing process needs to cover normal scenarios, failure scenarios, and extreme scenarios. Existing testing methods mainly rely on manually generated test data and testing based on real equipment or on-site packet capture and playback data, which presents the following technical challenges: Low efficiency in test scenario construction: Manually creating test data and writing test scripts is time-consuming and laborious, with the data preparation cycle for a single scenario taking several days, which is difficult to meet the needs of large-scale testing. Insufficient coverage of fault scenarios: Extreme fault and combined fault scenarios need to be implemented through complex means such as code instrumentation, which is costly and difficult to fully cover. Poor scenario reusability: When changing test projects or adding new scenarios, test data and scripts need to be modified again, resulting in poor adaptability; Low level of automation: The processes of test data generation, interface adaptation, and script execution lack coordination and require manual intervention, which can easily lead to errors due to human operation; Multimodal data processing is challenging: Real-time control system testing involves multiple types of data, such as text protocol data, image data, and radar point cloud data, which are difficult to process efficiently using traditional methods.

[0003] A search revealed Chinese Patent Publication No. CN120560999A, which discloses a software automated testing method and system based on generative artificial intelligence. This method uses a multimodal large model to identify software UI components to construct a knowledge graph and retrieves historical test cases to assist in generating functional test scripts for the software interface, thus automating the testing of general software user interfaces. However, this solution primarily targets the functional and interface testing of general software. Its knowledge graph is built around the UI component structure and does not involve the generation and simulation of multimodal sensor data in the physical environment. It fails to form a complete closed loop from requirements to execution to model self-optimization, and cannot solve problems such as high-fidelity physical scene simulation and cross-modal data collaborative testing required for real-time control systems.

[0004] Therefore, how to automatically generate and execute real-time control system tests covering multimodal data and complex scenarios is a technical problem that needs to be solved. Summary of the Invention

[0005] The purpose of this invention is to overcome the defects of the prior art by providing an intelligent simulation method, system, device and medium for testing scenarios of real-time control systems.

[0006] The objective of this invention can be achieved through the following technical solutions: According to a first aspect of the present invention, a method for intelligent simulation of test scenarios for a real-time control system is provided, comprising: Upon receiving test requirements, the system parses the test requirements based on a large language model and a test domain knowledge graph to generate structured test requirements that include physical scene type, data type, and interface protocol information. Based on the structured testing requirements, a cross-modal test dataset that is temporally and semantically correlated is generated. The cross-modal test dataset includes sensor simulation data for simulating the physical environment and control protocol simulation data for simulating interaction with the real-time control system under test. Based on the structured test requirements, cross-modal test dataset, and preset rules in the test domain knowledge graph, test scripts are automatically generated and time-sequential test scenario processes are arranged to drive the real-time control system under test and collect its response. The test execution results are analyzed, and the decision logic for data generation or protocol simulation is optimized through reinforcement learning. The optimization results are then updated to the test domain knowledge graph.

[0007] As a preferred technical solution, the generation of a cross-modal test dataset that is temporally and semantically correlated specifically includes: Based on the physical scenario type in the structured test requirements, determine the spatiotemporal correlation constraints between the simulation data of each sensor and the protocol simulation data; The generative artificial intelligence model is invoked to generate sensor simulation data and protocol simulation data that are consistent in timestamp, spatial coordinate system and event logic, based on the spatiotemporal correlation constraints.

[0008] As a preferred technical solution, the sensor simulation data includes scene image data generated by a diffusion model and three-dimensional point cloud data generated by a generative adversarial network; the control protocol simulation data includes text protocol data generated by a large language model.

[0009] As a preferred technical solution, the time-sequential test scenario orchestration process specifically includes: The data in the cross-modal test dataset, the instructions in the test script, and the interface calls determined by the preset rules are time-series aligned and logically dependently bound to generate a unified scene execution timeline.

[0010] As a preferred technical solution, the time-sequential test scenario process is executed as follows: according to the scenario execution timeline, the test is triggered by device simulation, the protocol simulation data in the cross-modal test dataset is sent to the corresponding interface of the real-time control system under test, and the sensor simulation data is synchronously injected into the corresponding simulation sensor channel, while the response data of the system under test is collected.

[0011] As a preferred technical solution, the decision logic is optimized through reinforcement learning, specifically including: When the test execution results indicate that the expected goals have not been achieved, root cause analysis should be performed on the execution results to locate the problematic link; Based on the results of the root cause analysis, reinforcement learning is used to adjust the prompting strategies, model parameters, or decision priorities of the data generation and interface adaptation.

[0012] As a preferred technical solution, the method further includes: storing the structured test requirements, cross-modal test datasets, test scripts, scenario execution timelines and test execution results generated during each test simulation in a versioned manner to form a test asset library.

[0013] As a preferred technical solution, the test requirements are described in natural language.

[0014] According to a second aspect of the present invention, a test scenario intelligent simulation system for implementing the method is provided, the system comprising: The requirement parsing module, based on a large language model and a knowledge graph of the testing domain, parses the test requirements input by users into structured test requirements; The multimodal data generation module, based on the structured testing requirements, calls the corresponding generative artificial intelligence model to generate a cross-modal test dataset that is temporally and semantically related. The scenario construction and execution module automatically generates test scripts and orchestrates and executes a time-sequential test scenario process based on the structured test requirements, cross-modal test dataset, and test domain knowledge graph. The knowledge optimization module is used to analyze test execution results, optimize the decision logic of data generation or protocol simulation through reinforcement learning, and update the optimization results to the test domain knowledge graph.

[0015] As a preferred technical solution, the scene construction and execution module includes: The timing logic orchestration module is used to perform timing alignment and logic binding on cross-modal test data, test scripts, and interface calls; The equipment simulation module is used to interact with the real-time control system under test through a preset interface protocol to perform tests and collect response data.

[0016] According to a third aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described thereon.

[0017] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.

[0018] Compared with the prior art, the present invention has the following advantages: 1. This invention achieves end-to-end automation of the entire link of real-time control system testing through intelligent requirement analysis, automatic generation of multimodal data, scenario orchestration and execution, and closed-loop self-optimization, eliminating the need for manual intervention and improving testing efficiency, scenario coverage depth, and overall reliability.

[0019] 2. By coordinating multiple generative artificial intelligence models and applying temporal and semantic constraints, this invention can generate a physically consistent and spatiotemporally correlated multimodal test dataset, ensuring the consistency of sensor data, control protocols, and test commands in time, space, and logic. This enables a more realistic simulation of the operating environment of the physical system and effectively improves the confidence level of test results and the defect detection rate.

[0020] 3. Based on reinforcement learning and knowledge graphs, this invention achieves adaptive optimization of testing strategies and data generation logic, enabling the testing system to continuously learn and evolve, thereby improving test coverage and long-term performance.

[0021] 4. This invention can automatically construct extreme and combined failure scenarios without the need for manual code instrumentation or building complex physical environments, thereby improving the test coverage integrity and security of extreme and combined failure scenarios.

[0022] 5. By replacing or expanding the domain knowledge graph, this invention can quickly adapt to different real-time control systems such as train control and automatic driving, thereby improving the system's versatility and maintainability. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a simulation data flow diagram of the test scenario for this invention; Figure 3 This is a system architecture diagram of the present invention; Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] Example 1: like Figure 1 As shown, this invention provides an intelligent simulation method for test scenarios of a real-time control system, whose data flow and interaction are as follows: Figure 2 As shown, the method includes: Step S1: Receive and parse the test requirements to generate structured test requirements. Users input their testing requirements through a natural language interface, such as: "Design a test scenario for the train autonomous perception system where an obstacle appears in front of it while it is traveling in a tunnel, triggering emergency braking. The scenario should include camera images, LiDAR point clouds, and Train Control (ATP) protocol data." After receiving the request, the system invokes a Large Language Model (LLM) loaded with knowledge from the rail transit domain. This LLM, combined with a pre-built test domain knowledge graph, performs collaborative parsing. The knowledge graph stores information such as common sensor types in train systems, their corresponding data formats, communication protocols, typical fault modes, and physical constraint rules. The LLM utilizes this knowledge to decompose the natural language request and transform it into a structured test requirement, the content of which can be represented as: Physical scene type: tunnel environment, static obstacle in front; Data types: 3D LiDAR point cloud data, RGB image data, vehicle control network protocol data; Interface protocol information: ATP protocol, specific destination and source addresses, braking command message identifier; Physical and logical constraints: obstacles should be approximately 50-100 meters from the front of the vehicle; images and point clouds should be spatially aligned; ATP alarm messages should be issued within a specific delay after the obstacle is detected.

[0026] Integrate the above information to generate a machine-readable JSON or XML structured test requirements document.

[0027] Step S2: Generate a cross-modal test dataset with temporal and semantic associations. The cross-modal data generation module receives the above-mentioned structured test requirements, schedules multiple generative AI models in parallel, and collaboratively generates data based on the spatiotemporal constraints implicit in the requirements.

[0028] Determine the associated constraints: The system derives from the structured requirements that the scene image data, 3D point cloud data and text protocol data must meet the following: (a) Spatial alignment: The pixel blocks in the image and the 3D point clusters in the point cloud should represent the same obstacle, and their positions and sizes should match; (b) Temporal causality: The obstacle appears at time T0, the system perceives it at time T0+Δt1, and the ATP alarm is triggered at time T0+Δt2, where Δt1 and Δt2 must meet the system specifications.

[0029] Collaborative data generation: Scene image data: Using diffusion models such as Stable Diffusion, with the input prompt: "Inside the tunnel, there is a square metal obstacle on the track ahead, about 80 meters away, from the train's front view, in dim lighting." Generate high-resolution, labeled fault images; 3D point cloud data: Call the point cloud generation model trained on Generative Adversarial Network (GAN), input the same scene semantic description as the image, and additionally specify the approximate size and reflectivity of the obstacles to generate the corresponding .pcd format point cloud file, ensuring that its 3D bounding box basically coincides with the 2D projection box of the obstacle in the image after coordinate transformation.

[0030] Text protocol data: The large language model is invoked to fill in the binary protocol message data stream containing fields such as timestamp, obstacle recognition ahead, suggested braking deceleration, and emergency braking request according to the ATP protocol format template in the knowledge graph, and a time series synchronized with sensor data is generated.

[0031] All generated data is packaged into a cross-modal test dataset, which is internally linked by a unified scene ID and timestamp to ensure that it describes the same physical event in terms of timing and semantics.

[0032] Step S3: Orchestrate and execute a time-series test scenario process Based on structured testing requirements and test logic templates in the knowledge graph, an executable test script is automatically generated, which includes initialization instructions, data loading loops, interface call functions, etc. Read the cross-modal test dataset and test scripts, and based on the scene rules in the knowledge graph, perform time series alignment and logical dependency binding to create a scene execution timeline: Time T0: Inject the initial frame point cloud and image; Time T1: Begin injecting the point cloud and image sequence containing obstacles; At time T1+150ms: The test script calls the interface check function to check whether the system under test has issued an identification signal; At time T1+200ms: a simulation protocol message is injected into the interface of the system under test to simulate the braking command issued by the vehicle controller; Time T2: Collect the actual braking response curve, status code, and other data of the system under test.

[0033] The equipment simulation is executed according to this precise timeline, connecting to the real-time control system under test via a physical interface adapter. The system, following the timeline, precisely controls the image injection card, point cloud simulator, and ATP protocol board, synchronously injecting simulation data into the controller under test. Simultaneously, the controller's output signals are acquired in real time via a data acquisition card.

[0034] Step S4: Analyze the results and implement self-optimizing learning After the test is completed, the system automatically analyzes the collected data and compares the expected behavior with the actual behavior. For example, it finds that the system issues the braking command later than expected. The 5 Whys and other root cause analysis algorithms are used in conjunction with knowledge graphs to analyze possible causes, such as insufficient obstacle point cloud density leading to recognition delays and slight deviations in the ATP protocol message format. The reinforcement learning unit treats the timeout as a negative reward and may decide to adjust the data generation decision logic when generating similar scenarios next time, for example, by increasing the point density parameter of the point cloud generation model by 20% or adding a highest priority field to the protocol generation prompt template. The optimized decision logic is transformed into a structured rule, stored in the test domain knowledge graph, and associated with the obstacle intrusion scenario node in the tunnel. All input, process data, and outputs from this test are packaged, versioned, and stored to form a reusable test asset. When similar needs arise regarding obstacles within the station in the future, the system can quickly retrieve and adapt this asset.

[0035] The method of this invention analyzes natural language requirements through large language models and domain knowledge graphs, coordinates multiple generative AI models to automatically generate cross-modal test scenarios with temporal and physical associations, and drives hardware-in-the-loop testing. Finally, it achieves closed-loop self-optimization of test strategies and knowledge graphs based on reinforcement learning, realizing full-process automation of test scenarios from intelligent construction and automatic execution to continuous evolution.

[0036] Example 2: like Figure 3 As shown, this invention provides an intelligent simulation system for test scenarios to implement the above-described method. This system adopts a layered architecture and is deployed in a hardware-in-the-loop test environment, mainly comprising: 1. User Access Layer: This layer serves as the unified entry point for the system's interaction with the external environment, responsible for receiving diverse input commands and executing control measures, including: Natural Language Requirements Parsing Module: Provides a web interface or API to receive test requirements described by users in natural language and pass them to the core capability layer for in-depth parsing; Document-based requirement analysis module: Supports uploading structured requirement documents, automatically extracts key information, and supplements or replaces natural language input; Hardware system control module: Provides remote control functions for hardware-in-the-loop (HIL) test bench, such as power-on / power-off, reset, and analog signal on / off of the controller under test; Software testing system scheduling module: responsible for scheduling and managing external testing software or simulation tools to ensure the testing environment is ready; Manual scenario triggering module: Allows testers to insert manual intervention points into automated processes, such as injecting preset fault signals at specific times to achieve mixed test scenarios.

[0037] 2. Core Competency Layer: This layer integrates the core functional units for realizing intelligent simulation, including: Protocol parsing module: Performs in-depth parsing of the requirements passed in from the access layer, calls the large language model of the basic support layer, and combines knowledge graphs to transform vague requirements into structured test requirements that include clear physical scenarios, data types, and interface protocols; Text data generation module: Based on structured requirements, it calls the LLM inference service of the basic support layer to generate text protocol data or test script drafts that conform to specific industrial communication protocol formats. Image / Point Cloud Generation Module: Based on the physical scene described in the structured requirements, it calls the generative adversarial network inference service and diffusion model of the basic support layer to generate high-fidelity scene image data and 3D point cloud data, and ensures that the two are aligned in spatial position. Control parameter generation module: assists in generating control parameters, threshold configurations, and other data required for testing; Script generation module: Based on the integrated structured requirements, generated test data, and interface rules, it automatically assembles and generates executable test scripts. The timing logic orchestration module is responsible for aligning and sorting the generated cross-modal test data, test script instructions, and hardware interface calls according to event logic and causal relationships, and orchestrating them into a precise scene execution timeline. Equipment simulation module: As the driving unit for interaction with real hardware, it loads the scene execution timeline, precisely controls the hardware interface board during test execution, injects protocol data, image / point cloud data, etc. into the real-time control system under test in sequence and synchronously, and collects its response output in real time.

[0038] 3. Basic support layer: This layer provides unified algorithm, computing power, and data resource support for the core capabilities of the upper layer, including: LLM Inference Module: Deploys a large language model finely tuned with domain knowledge, providing semantic understanding and content generation capabilities for requirements parsing, text protocol generation, etc. GAN inference module: Deploys a generative adversarial network model dedicated to generating high-quality images and 3D point clouds, which is the basis for synthesizing sensor data; Reinforcement learning module: Runs reinforcement learning algorithms, receives feedback from test execution results, optimizes data generation strategies or parameters by calculating reward values, and drives continuous system improvement; Vector Database Module: Stores and manages the embedded vectors of test cases and generated data samples, and supports rapid retrieval and reuse of semantically similar cases; Structured database module: Stores structured data such as system configuration, user information, test task logs, and version information; The knowledge graph module stores entities (devices, protocols, fault types), attributes, and their complex relationships (dependencies, triggers, constraints) in the testing domain using a graph structure. It guides requirement analysis and scenario construction, and also serves as a storage carrier for reinforcement learning optimization results, recording experiential rules and enabling the accumulation and evolution of knowledge.

[0039] System workflow: When a user inputs a natural language testing requirement through the access layer, the protocol parsing submodule of the core capability layer immediately collaborates with the LLM inference submodule and knowledge graph submodule of the basic support layer to structure the requirement. Subsequently, the text data generation submodule, image / point cloud generation submodule, etc., call the corresponding AI models at the underlying level to generate relevant multimodal test datasets. The script generation submodule and the timing logic orchestration submodule integrate these data with the hardware interface logic to form an executable test plan. Finally, the device simulation submodule, with the cooperation of the hardware system control submodule of the access layer, drives the HIL bench to perform the test and feeds the results back to the system. The reinforcement learning submodule of the basic support layer optimizes the strategy based on the results and updates the knowledge graph submodule, completing an intelligent closed loop.

[0040] The system of this invention receives diverse inputs through the user access layer, transforms the requirements into executable time-series tests through the core capability layer, and provides algorithm and knowledge support through the basic support layer. The collaborative work of each layer enables the intelligent construction, automatic execution, and continuous evolution of test scenarios.

[0041] Example 3: The electronic device of this invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0042] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0043] The processing unit executes the various methods and processes described above, such as methods S1 to S4. For example, in some embodiments, methods S1 to S4 may be implemented as computer software programs tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of methods S1 to S4 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S4 by any other suitable means (e.g., by means of firmware).

[0044] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0045] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0046] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0047] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent simulation of test scenarios in a real-time control system, characterized in that, include: Upon receiving test requirements, the system parses the test requirements based on a large language model and a test domain knowledge graph to generate structured test requirements that include physical scene type, data type, and interface protocol information. Based on the structured testing requirements, a cross-modal test dataset that is temporally and semantically correlated is generated. The cross-modal test dataset includes sensor simulation data for simulating the physical environment and control protocol simulation data for simulating interaction with the real-time control system under test. Based on the structured test requirements, cross-modal test dataset, and preset rules in the test domain knowledge graph, test scripts are automatically generated and time-sequential test scenario processes are arranged to drive the real-time control system under test and collect its response. The test execution results are analyzed, and the decision logic for data generation or protocol simulation is optimized through reinforcement learning. The optimization results are then updated to the test domain knowledge graph.

2. The intelligent simulation method for test scenarios of a real-time control system according to claim 1, characterized in that, The generation of the cross-modal test dataset, which is temporally and semantically correlated, specifically includes: Based on the physical scenario type in the structured test requirements, determine the spatiotemporal correlation constraints between the simulation data of each sensor and the protocol simulation data; The generative artificial intelligence model is invoked to generate sensor simulation data and protocol simulation data that are consistent in timestamp, spatial coordinate system and event logic, based on the spatiotemporal correlation constraints.

3. The intelligent simulation method for test scenarios of a real-time control system according to claim 1, characterized in that, The sensor simulation data includes scene image data generated by a diffusion model and 3D point cloud data generated by a generative adversarial network; the control protocol simulation data includes text protocol data generated by a large language model.

4. The intelligent simulation method for test scenarios of a real-time control system according to claim 1, characterized in that, The orchestration of time-sequential test scenarios specifically includes: The data in the cross-modal test dataset, the instructions in the test script, and the interface calls determined by the preset rules are time-series aligned and logically dependently bound to generate a unified scene execution timeline.

5. The intelligent simulation method for test scenarios of a real-time control system according to claim 4, characterized in that, The execution of the time-series test scenario process is as follows: according to the scenario execution timeline, the test is triggered by device simulation, the protocol simulation data in the cross-modal test dataset is sent to the corresponding interface of the real-time control system under test, and the sensor simulation data is synchronously injected into the corresponding simulation sensor channel, while the response data of the system under test is collected.

6. The intelligent simulation method for test scenarios of a real-time control system according to claim 1, characterized in that, The decision-making logic is optimized through reinforcement learning, specifically including: When the test execution results indicate that the expected goals have not been achieved, root cause analysis should be performed on the execution results to locate the problematic link; Based on the results of the root cause analysis, reinforcement learning is used to adjust the prompting strategies, model parameters, or decision priorities of the data generation and interface adaptation.

7. The intelligent simulation method for test scenarios of a real-time control system according to claim 1, characterized in that, The method further includes: versioning and storing the structured test requirements, cross-modal test datasets, test scripts, scenario execution timelines and test execution results generated during each test simulation to form a test asset library.

8. The intelligent simulation method for test scenarios of a real-time control system according to claim 1, characterized in that, The test requirements are described in natural language.

9. A test scenario intelligent simulation system for implementing the method as described in any one of claims 1-8, characterized in that, The system includes: The requirement parsing module, based on a large language model and a knowledge graph of the testing domain, parses the test requirements input by users into structured test requirements; The multimodal data generation module, based on the structured testing requirements, calls the corresponding generative artificial intelligence model to generate a cross-modal test dataset that is temporally and semantically related. The scenario construction and execution module automatically generates test scripts and orchestrates and executes a time-sequential test scenario process based on the structured test requirements, cross-modal test dataset, and test domain knowledge graph. The knowledge optimization module is used to analyze test execution results, optimize the decision logic of data generation or protocol simulation through reinforcement learning, and update the optimization results to the test domain knowledge graph.

10. The intelligent simulation system for test scenarios according to claim 9, characterized in that, The scenario construction and execution module includes: The timing logic orchestration module is used to perform timing alignment and logic binding on cross-modal test data, test scripts, and interface calls; The equipment simulation module is used to interact with the real-time control system under test through a preset interface protocol to perform tests and collect response data.

11. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 8.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 8.

Images