A virtual reality driving training system based on multimedia interaction

Abstract

The application discloses a virtual reality driving training system based on multimedia interaction and relates to the technical field of driving simulation, comprising: a scheduling and data management unit for constructing a personalized ability portrait; a virtual environment real-time generation unit for rendering a virtual driving environment according to the personalized ability portrait; a multi-modal interaction unit for collecting and fusing student driving operation data and vehicle state data to obtain student behavior data flow; and providing feedback to the student; an AI driving instructor analysis and decision unit for making decisions on the student behavior data flow based on a three-layer fusion architecture of driving skill development risk prediction and optimization algorithm to obtain driving teaching decisions; and a dynamic traffic and event simulation unit for constructing adaptive emergencies in the virtual driving environment according to the driving teaching decisions to train the student. The application constructs a system integrating perception, analysis, decision, feedback and evolution, and greatly improves the training quality.

Description

Technical Field

[0001] This invention relates to the field of driving simulation technology, and more specifically, to a virtual reality driving training system based on multimedia interaction. Background Technology

[0002] Traditional driver training presents several challenges for learners. First, limitations in facilities, vehicles, and instructors may prevent them from getting enough practice opportunities. Second, adverse weather conditions such as rain or snow can significantly impact training effectiveness, even leading to course cancellations. Third, traditional teaching methods lack personalized guidance and cannot be tailored to different learning paces and styles, greatly reducing learning efficiency. Furthermore, novice drivers face higher risks when practicing on real roads.

[0003] To overcome these limitations, driving simulation systems based on computer simulation technology have gradually emerged in recent years. Traditional driving training relies on physical vehicles and specific venues, a method that, while intuitive, is clearly limited. In recent years, computer-based driving simulators have begun to appear, but these systems often lack realistic operational feedback and a highly interactive experience, making it difficult to completely replace the feeling of actual driving. Some high-end products attempt to combine VR (virtual reality) technology to enhance immersion, but due to high costs, low comfort levels, and high technical complexity, their adoption is low.

[0004] Overall, the main drawbacks of existing technologies are their lack of interactivity and realism, resulting in poor learning outcomes; poor flexibility, making it impossible to adjust course content according to individual learning progress; and strong dependence on the physical environment, increasing operating costs.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] In view of the problems in related technologies, the present invention proposes a virtual reality driving training system based on multimedia interaction to overcome the above-mentioned technical problems existing in the existing related technologies.

[0007] Therefore, the specific technical solution adopted by the present invention is as follows: A multimedia interactive virtual reality driving training system, the system comprising: The scheduling and data management unit is used to perform feature fusion on trainee training requests and historical data to build personalized competency profiles. The virtual environment real-time generation unit is used to render a virtual driving environment based on personalized capability profiles and using maps and a 3D scene engine. The multimodal interaction unit is used to collect and integrate student driving operation data and vehicle status data in real time using interactive devices to obtain student behavior data stream; and to provide multi-dimensional interactive feedback to students based on the student behavior data stream and the driving teaching decisions of the AI ​​driving coach analysis and decision-making unit. The AI ​​driving coach analysis and decision-making unit is used to perform real-time analysis and decision-making on student behavior data streams based on a three-layer fusion architecture of driving skill development risk prediction and optimization algorithms, in order to obtain driving teaching decisions. The dynamic traffic and event simulation unit is used to construct adaptive emergencies in a virtual driving environment based on driving instruction decisions, and to train students until the students finish their training.

[0008] Furthermore, the scheduling and data management unit includes a user management and identity authentication module, a training request processing module, a student information management module, and a personalized ability profile generation module; The user management and identity authentication module is used for account registration, login authentication, and permission management for students and administrators; The training request processing module is used to parse training requests from logged-in and authenticated students and obtain training instructions. The student information management module is used to integrate training requests, training process data, and evaluation results to obtain historical data sources for students; The personalized competency profile generation module is used to extract the corresponding student's historical data source from the student information management module according to the training instructions, clean and extract features from the corresponding student's historical data source to obtain a standardized feature set, and construct a personalized competency profile based on the standardized feature set.

[0009] Furthermore, based on the training instructions, the corresponding historical data source of the student is extracted from the student information management module. This historical data source is then cleaned and its features are extracted to obtain a standardized feature set. Based on this standardized feature set, a personalized competency profile is constructed, including: The training instructions are parsed, and based on the parsing results, the corresponding historical data sources of students are extracted from the student information management module. The acquired historical data sources of the corresponding students are sequentially subjected to denoising, outlier filtering, and standardization to obtain multi-dimensional features; Multi-dimensional features are filtered to obtain a standardized feature set; a pre-set driving ability mapping model is used to map the standardized feature set to obtain a quantitative driving ability index. By integrating quantitative driving ability indicators, a personalized ability profile is constructed.

[0010] Furthermore, the real-time virtual environment generation unit includes a personalized scene configuration module, a scene data management module, and an environment rendering engine module; The personalized scene configuration module is used to analyze the personalized capability profile to obtain scene parameters and environment rendering parameters; The scene data management module is used to schedule and integrate map data, 3D road models, traffic facility models and environmental texture resources based on scene parameters to construct a scene framework; The environment rendering engine module is used to render the scene framework based on the environment rendering parameters and in conjunction with the 3D graphics engine to obtain the virtual driving environment.

[0011] Furthermore, the multimodal interaction unit includes an operation behavior acquisition module, a vehicle status perception module, an interaction data fusion and synchronization module, and a multi-channel feedback generation module; The operation behavior acquisition module is used to collect driving operation data of trainees in real time using interactive devices to obtain operation behavior data stream; The vehicle state perception module is used to acquire vehicle dynamic parameters in the virtual environment and preprocess the vehicle dynamic parameters to obtain vehicle state data stream. The interactive data fusion and synchronization module is used to sequentially perform timestamp alignment and filtering fusion processing on the operation behavior data and vehicle status data to obtain the student behavior data stream. The multi-dimensional feedback generation module is used to provide students with multi-dimensional feedback, including tactile, kinematic, and auditory feedback, based on the student behavior data stream and the driving teaching decisions of the AI ​​driving coach analysis and decision-making unit, combined with interactive devices.

[0012] Furthermore, the AI ​​driving coach analysis and decision-making unit includes a driving intention and risk prediction module, an event feedback linkage decision-making module, and a skill trajectory and training strategy module; The driving intention and risk prediction module is used to analyze the perception layer based on the student's behavior data stream and the driving skill development risk prediction and optimization algorithm to predict the student's driving intention, and evaluate the driving intention based on the preset risk threshold to obtain the risk assessment result. The event feedback linkage decision module is used to map teaching strategies based on risk assessment results and real-time training events, and to obtain event feedback decisions by utilizing the decision layer of the driving skill development risk prediction and optimization algorithm. The Skill Trajectory and Training Strategy module is used to optimize event feedback decisions based on students' historical data and path optimization algorithms at the path layer of the driving skill development risk prediction optimization algorithm, thereby obtaining driving teaching decisions.

[0013] Furthermore, based on the student behavior data stream, and combined with the perception layer of the driving skill development risk prediction and optimization algorithm, the driving intention of the student is predicted. This driving intention is then assessed using a preset risk threshold, yielding risk assessment results including: A driving intention recognition model constructed using a temporal neural network is used to identify the student's behavioral data stream and obtain the student's driving intention. Based on the student's driving intention, the system matches it with the scenario-intention-risk mapping table in the preset risk knowledge base, and calculates the risk probability value of the student's driving intention by combining it with real-time environmental status information. The risk probability value is compared with a preset risk threshold. Based on the comparison results, a risk assessment result is obtained, including risk level, risk type, risk probability, and core trigger.

[0014] Furthermore, based on the risk assessment results and real-time training events, the decision layer of the driving skill development risk prediction and optimization algorithm is used to map teaching strategies, resulting in event feedback decisions including: By combining natural language processing algorithms, the risk assessment results and real-time training events are analyzed, based on a standardized event information table containing events, types, and key parameters. Based on a pre-set teaching strategy rule base, a standardized event information table is matched using a multi-condition matching algorithm to obtain preliminary teaching intervention strategies. By using a teaching effectiveness prediction model, preliminary teaching intervention strategies are pre-evaluated and parameters are fine-tuned to obtain event feedback decisions.

[0015] Furthermore, based on historical student data and combined with path optimization algorithms, the event feedback decision-making is optimized at the path layer of the driving skill development risk prediction optimization algorithm, resulting in driving instruction decisions including: Based on students' historical data, key competency indicator sequences are extracted, and personalized skill development trajectories for students are constructed. Based on the preset capability development standard library in the path layer of the driving skill development risk prediction and optimization algorithm, the ideal skill development trajectory of the corresponding trainee is obtained. By using the dynamic time warping algorithm, the personalized skill development trajectory of trainees and the ideal skill development trajectory are aligned and analyzed to obtain the trainees' potential bottleneck nodes and lag dimensions. Based on path optimization algorithms, potential bottleneck nodes for trainees, and lag dimensions, the event feedback decision is optimized to obtain driving instruction decisions.

[0016] Furthermore, the dynamic traffic and event simulation unit includes a traffic flow simulation module, an event triggering module, a dynamic event injection module, and an adaptive difficulty adjustment module; The traffic flow simulation module is used to construct dynamic traffic flow in a virtual driving environment based on a microscopic traffic model, thereby obtaining the traffic environment. The event triggering module is used to obtain target training events by matching the driving teaching decisions with a preset event triggering rule base; The dynamic event injection module is used to integrate training events into the traffic environment in a multi-stage dynamic evolution manner to obtain sudden events; The adaptive difficulty adjustment module is used to adaptively adjust the parameters of sudden events based on the student's personalized ability profile and real-time operation performance, thus obtaining adaptive sudden events.

[0017] The beneficial effects of this invention are as follows: 1. This invention, through the organic integration of automation, unmanned operation, and artificial intelligence technologies, pioneers a completely new driver training model. Unmanned management enables 24 / 7 service, greatly enhancing service flexibility and convenience. In terms of product services, this invention utilizes virtual reality technology and multimodal interaction to provide trainees with an immersive driving experience, and combines this with an AI driving assistant to provide personalized teaching plans, ensuring that each trainee receives targeted guidance. In terms of management and operation, through cloud databases and automated management systems, the training process is standardized and data-driven optimized, improving overall operational efficiency. Regarding application scenarios, this invention is not only suitable for conventional driver training but can also conduct simulated training for special road conditions and extreme weather, meeting the needs of trainees at different levels. This series of innovative measures not only improves training efficiency and service quality but also brings about a transformation in the driver training industry, demonstrating broad market prospects and development potential.

[0018] 2. This invention breaks through the limitations of traditional driving simulators, which have fixed scenarios, single feedback, and static teaching. It constructs an intelligent driving training system that integrates perception, analysis, decision-making, feedback, and evolution by using a scheduling and data management unit, a real-time virtual environment generation unit, a multimodal interaction unit, an AI driving coach analysis and decision-making unit, and a dynamic traffic and event simulation unit. This significantly improves the quality and safety of training. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a virtual reality driving training system based on multimedia interaction according to an embodiment of the present invention. Figure 2 This is one of the application example diagrams of a virtual reality driving training system based on multimedia interaction according to an embodiment of the present invention; Figure 3This is a second example of an application of a multimedia interactive virtual reality driving training system according to an embodiment of the present invention; Figure 4 This is Figure 3, an application example of a multimedia interactive virtual reality driving training system according to an embodiment of the present invention.

[0021] In the picture: 1. Scheduling and data management unit; 2. Real-time virtual environment generation unit; 3. Multimodal interaction unit; 4. AI driving coach analysis and decision-making unit; 5. Dynamic traffic and event simulation unit. Detailed Implementation

[0022] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0023] According to an embodiment of the present invention, a virtual reality driving training system based on multimedia interaction is provided.

[0024] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1-4 As shown, according to an embodiment of the present invention, a virtual reality driving training system based on multimedia interaction includes: The scheduling and data management unit 1 is used to perform feature fusion on trainee training requests and historical data to build a personalized capability profile.

[0025] It should be explained that the scheduling and data management unit realizes a closed-loop support for secure user access, accurate demand analysis, full-domain data integration, and personalized profile generation: it ensures data security through identity authentication and access control, and achieves accurate matching of demands and data through training instruction parsing and historical data extraction, ultimately generating a personalized capability profile.

[0026] In this optional embodiment, the scheduling and data management unit 1 includes a user management and identity authentication module, a training request processing module, a student information management module, and a personalized ability profile generation module; The user management and identity authentication module is used for account registration, login authentication, and permission management for students and administrators; Specifically, when student Zhang San uses the system for the first time, he completes account registration via the web interface: filling in basic information such as name, driver's license type, and training stage, and uploading a photo of his ID card to complete real-name authentication. When administrator Li Si logs in through the backend account, the system automatically assigns him maintenance and teaching management permissions. Subsequently, when Zhang San initiates training, he logs in using a two-factor authentication method (account and password, SMS verification code). After the module verification is successful, a temporary access token is generated, allowing him to access system resources.

[0027] Identity authentication uses the OAuth 2.0 protocol to interface with identity verification interfaces, ensuring the accuracy of real-name authentication. Password storage uses the BCrypt algorithm with salt hashing to avoid plaintext leakage. Access management can be configured with an access matrix based on the RBAC model. Students can only view their own training data and initiate appointments; administrators can configure training scenario parameters and view student group profiles; and operations and maintenance personnel can monitor system resource usage but have no access to student data. Login logs for all accounts are recorded (time, IP, device model). When abnormal logins from other locations are detected, the account is automatically frozen and an alarm SMS is sent.

[0028] The training request processing module is used to parse training requests from logged-in and authenticated students and obtain training instructions. Specifically, JSON Schema is used to verify the completeness of request parameters, such as required training subjects and durations; if missing, an insufficient parameter message is returned. The request queue is sorted using a rule engine. At this time, there are 3 pending requests in the system. Zhang San's request to strengthen his weak points has higher priority than the other two students' regular familiarization training requests and is ranked first. The training requests are parsed and converted into a standardized instruction set, which is then pushed synchronously to the cross-unit scheduling module and the student information management module.

[0029] The student information management module is used to integrate training requests, training process data, and evaluation results to obtain historical data sources for students; Specifically, after receiving Zhang San's training instructions, the student information management module automatically associates them with his student ID and extracts historical data sources from the distributed database cluster. It integrates the training request data, training process data, and evaluation result data into Zhang San's specific historical data source, and archives it in the student archive database according to the student ID-skill type-timestamp format.

[0030] Structured data (training requests, evaluation results) is stored in a MySQL sharded database, partitioned by driver's license type and subject, supporting fast retrieval by skill type. Unstructured data (operation time-series curves, training video clips) is stored in MinIO object storage, with unique filenames generated using student IDs and timestamps, and linked to structured data via indexes. Real-time process data is temporarily cached in a Redis cluster and automatically archived to persistent storage after training. An ETL tool (Kettle) is used to batch process the previous day's training data every morning, removing duplicate records and adding data association labels.

[0031] The personalized competency profile generation module is used to extract the corresponding student's historical data source from the student information management module according to the training instructions, clean and extract features from the corresponding student's historical data source to obtain a standardized feature set, and construct a personalized competency profile based on the standardized feature set.

[0032] In this optional embodiment, according to the training instructions, the corresponding student's historical data source is extracted from the student information management module, the corresponding student's historical data source is cleaned and its features are extracted to obtain a standardized feature set; and based on the standardized feature set, a personalized competency profile is constructed, including: The training instructions are parsed, and based on the parsing results, the corresponding historical data sources of students are extracted from the student information management module. The acquired historical data sources of the corresponding students are sequentially subjected to denoising, outlier filtering, and standardization to obtain multi-dimensional features; Multi-dimensional features are filtered to obtain a standardized feature set; a pre-set driving ability mapping model is used to map the standardized feature set to obtain a quantitative driving ability index. By integrating quantitative driving ability indicators, a personalized ability profile is constructed.

[0033] It should be explained that, taking the generation of the profile of trainee Zhang San's hill start training as an example, the specific steps are as follows: Step 1: Parse the training instructions and extract historical data sources.

[0034] The training request processing module parses Zhang San's training instructions as target student = ZH001, target skill = hill start, and training type = specialized reinforcement, and synchronizes the parsing results to the personalized ability profile generation module; the personalized ability profile generation module initiates a data retrieval request to the student information management module based on the student ID and target skill.

[0035] The system retrieves the full data of Zhang San's last three hill start training sessions from the student information management module, including 2,800 operation timing data, 5 error event records, and 3 AI coach evaluation reports.

[0036] Step 2: Data cleaning and multi-dimensional feature extraction.

[0037] Kalman filtering is used to filter sensor noise in the operational data, preserving valid operational trajectories; extreme data are eliminated using the 3σ principle; parameters of different dimensions are converted into standardized values ​​in the [0,1] interval, such as the original clutch lift rate of 0.8 cm / s corresponding to a standardized value of 0.4, and the standardized value of 1.2 cm / s corresponding to 1.0. Operational accuracy features, temporal features, error features, and progress features are extracted to form a 12-dimensional multi-dimensional feature set.

[0038] Step 3: Feature selection and driving ability quantification.

[0039] The mutual information entropy algorithm is used to calculate the correlation between each feature and hill start capability. Irrelevant features with a correlation of <0.3 are removed, and core features are retained to form a standardized feature set. The standardized feature set is input into a preset driving capability mapping model (the driving capability mapping model is trained based on the GBDT algorithm) to output quantitative indicators.

[0040] Step 4: Integrate and build a personalized capability profile.

[0041] The system integrates indicators at different levels, including core skills, subdivided competency items, quantitative scores, and weakness markers. It also automatically matches suggested rule bases to the profile based on specific reinforcement needs in training instructions. Finally, it generates a structured profile, writes it into the trainee file database, and pushes it to the virtual environment generation unit 2.

[0042] The virtual environment real-time generation unit 2 is used to render a virtual driving environment based on a personalized capability profile and using a map and 3D scene engine.

[0043] It needs to be explained that through the entire process of personalized ability profile analysis, scene parameter generation, resource integration and construction, and realistic rendering, focusing on the learner's weaknesses, and through the technical collaboration of the personalized scene configuration module, scene data management module, and environment rendering engine module, abstract ability characteristics are transformed into personalized virtual driving environments that are adapted to skills, hardware, and experience, truly realizing the scene support value of virtual reality driving training that is tailored to individual needs.

[0044] In this optional embodiment, the virtual environment real-time generation unit 2 includes a personalized scene configuration module, a scene data management module, and an environment rendering engine module; The personalized scene configuration module is used to parse the personalized capability profile to obtain scene parameters and environment rendering parameters; Specifically, the system receives personalized capability profiles from the scheduling and data management unit 1 and performs semantic analysis on them. For example, if the profile shows that the trainee has weak lane change risk control ability or poor night driving adaptability, the system automatically identifies the training dimensions that need to be strengthened and generates corresponding scene parameters (such as the road type being an urban expressway and the traffic density being set to medium to high) and environmental rendering parameters (such as setting it to rainy night, low visibility, and insufficient lighting).

[0045] The scene data management module is used to schedule and integrate map data, 3D road models, traffic facility models and environmental texture resources based on scene parameters to construct a scene framework; Specifically, based on the parameter commands output by the personalized scene configuration module, high-precision map data of the corresponding area is retrieved from a local or cloud-based map database, and matching 3D road models (such as ramps and roundabouts), traffic facility models (such as traffic lights and signs), and environmental texture resources (such as road surface materials and architectural styles) are loaded. Through data integration and spatial stitching, a complete and geographically accurate virtual scene framework is constructed.

[0046] The environment rendering engine module is used to render the scene framework according to the environment rendering parameters and in combination with the 3D graphics engine to obtain a virtual driving environment.

[0047] Specifically, relying on mainstream 3D graphics engines (such as Unity or Unreal Engine), the system performs real-time light and shadow calculations, physical material representations, and weather effect simulations on the constructed scene framework based on environmental rendering parameters, ultimately generating a visually realistic and sensorially immersive virtual driving environment.

[0048] For example, lighting rendering can use directional light to simulate natural light, adjust the lighting angle, avoid backlighting, and enable global illumination (GI) to enhance the realism of the scene; auxiliary prompt rendering can set the green semi-linked prompt line as a self-illuminating material; sign rendering can render ramp signs with a 1.5x magnification parameter, and the text can be in bold; multi-sensory linkage rendering can push scene visual frame synchronization signals to multi-modal interaction unit 3 to synchronously trigger multi-dimensional interactive feedback.

[0049] The multimodal interaction unit 3 is used to collect and integrate student driving operation data and vehicle status data in real time using interactive devices to obtain student behavior data stream; and to provide multi-dimensional interactive feedback to students based on student behavior data stream and driving teaching decisions of AI driving coach analysis and decision-making unit.

[0050] It should be explained that the multimodal interaction unit 3 integrates functions such as operation behavior acquisition, vehicle status perception, data fusion and synchronization, and multi-dimensional feedback generation. Using the interactive device as a carrier, it accurately captures the student's operation and the virtual vehicle status; through data fusion, it achieves a deep correlation between operation behavior and vehicle response; and finally, it constructs an immersive interactive experience with multi-dimensional feedback. This not only meets the needs of novice learners for operation assistance but also provides real-time data and execution carriers for personalized teaching by AI driving instructors, fully demonstrating the core advantages of the system's data-driven and multi-sensory collaborative approach.

[0051] In this optional embodiment, the multimodal interaction unit includes an operation behavior acquisition module, a vehicle state perception module, an interaction data fusion and synchronization module, and a multi-channel feedback generation module; The operation behavior acquisition module is used to collect driving operation data of trainees in real time using interactive devices to obtain operation behavior data stream; Specifically, various interactive devices (such as steering wheels, pedals, gear shifters, etc.) are used to capture the student's driving operation data in real time, including steering angle, acceleration and braking force, gear shifting timing, etc. The collected data is encapsulated in the format of timestamp-device ID-parameter value to form a continuous operation behavior data stream.

[0052] The vehicle state perception module is used to acquire vehicle dynamic parameters in the virtual environment and preprocess the vehicle dynamic parameters to obtain vehicle state data stream. Specifically, the dynamic parameters of the vehicle in the virtual environment, such as speed, acceleration, tire friction, and vehicle tilt, are obtained and preprocessed (e.g., noise removal and data smoothing) to generate a vehicle status data stream that accurately reflects the current operating status of the vehicle.

[0053] The interactive data fusion and synchronization module is used to sequentially perform timestamp alignment and filtering fusion processing on the operation behavior data and vehicle status data to obtain the student behavior data stream. Specifically, after receiving the operation behavior data stream and the vehicle status data stream, the data is processed in three steps: time sequence alignment, filtering and fusion, and event correlation.

[0054] Timestamp alignment is based on 10ms, and the timestamps of the two types of data are uniformly calibrated to remove invalid data with a timestamp deviation >20ms; filtering and fusion use the Kalman filter algorithm to fuse operation behavior data and vehicle status data to generate associated data; fusion is performed according to event labels. For example, the operation event 10:02:31.300 clutch lifting and the vehicle event 10:02:31.310 speed dropping to 750rpm are associated to mark the semi-clutch control deviation event.

[0055] The final generated structured engineer behavior data stream contains four-dimensional information: timestamp, operation parameters, vehicle status, and event tags.

[0056] The multi-dimensional feedback generation module is used to provide students with multi-dimensional feedback, including tactile, kinematic, and auditory feedback, based on the student behavior data stream and the driving teaching decisions of the AI ​​driving coach analysis and decision-making unit, combined with interactive devices.

[0057] Specifically, the module receives student behavior data streams and real-time decisions from the AI ​​driving coach unit, and generates multi-dimensional feedback in conjunction with interactive devices. This feedback is not limited to traditional visual cues, but also includes tactile, kinematic, and auditory feedback. This multi-sensory stimulation enhances the student's learning experience. A specific example is as follows: The steering wheel provides tactile feedback through force feedback. When the clutch is over-engaged, it immediately outputs reverse resistance to slightly straighten the steering wheel, indicating that the clutch has been raised too quickly. When the clutch returns to its original position, the resistance disappears and a synchronous vibration indicates that the clutch position has been reached.

[0058] The motion feedback is provided by a six-degree-of-freedom motion platform. When the semi-clutch deviation causes the vehicle to roll backward slightly, the platform immediately tilts backward to simulate the feeling of rolling backward. After the correction operation, the platform returns to a horizontal attitude.

[0059] Auditory feedback is provided through spatial audio headphones, simultaneously outputting engine sound effects and AI voice guidance.

[0060] Visual feedback is provided through a VR headset, with a green indicator line overlaid on the semi-linkage position in the VR field of view. When the operation deviates, the indicator line flashes red, and returns to a steady light after correction, thus working in conjunction with tactile feedback.

[0061] AI driving coach analysis and decision-making unit 4 is used to perform real-time analysis and decision-making on student behavior data streams based on a three-layer fusion architecture of driving skill development risk prediction and optimization algorithms, and obtain driving teaching decisions.

[0062] It should be explained that the AI ​​driving coach analysis and decision-making unit 4, through a three-level linkage mechanism of perception layer, decision layer, and path layer, possesses comprehensive capabilities for real-time perception, intelligent judgment, and dynamic optimization. It can not only intervene immediately when a student exhibits risky behavior, but also continuously optimize the training path based on their long-term learning trajectory, truly realizing an intelligent and personalized driving teaching closed loop that tailors instruction to individual needs, significantly improving the teaching quality and training efficiency of virtual reality driving training.

[0063] In this optional embodiment, the AI ​​driving coach analysis and decision-making unit 4 includes a driving intention and risk prediction module, an event feedback linkage decision-making module, and a skill trajectory and training strategy module. The driving intention and risk prediction module is used to analyze the perception layer based on the student's behavior data stream and the driving skill development risk prediction and optimization algorithm to predict the student's driving intention, and evaluate the driving intention based on the preset risk threshold to obtain the risk assessment result. The event feedback linkage decision module is used to map teaching strategies based on risk assessment results and real-time training events, and to obtain event feedback decisions by utilizing the decision layer of the driving skill development risk prediction and optimization algorithm. The Skill Trajectory and Training Strategy module is used to optimize event feedback decisions based on students' historical data and path optimization algorithms at the path layer of the driving skill development risk prediction optimization algorithm, thereby obtaining driving teaching decisions.

[0064] It should be explained that the AI ​​driving coach analysis and decision-making unit 4, as the core intelligent hub of the entire virtual reality driving training system, aims to perform real-time analysis and hierarchical decision-making on the student behavior data stream from the multimodal interaction unit 3 based on the three-layer fusion architecture of the Driving Skills Development Risk Prediction and Optimization Algorithm (DSRPO). Ultimately, it generates scientific, accurate, and personalized driving teaching decisions, realizing a closed-loop teaching mechanism from risk perception to immediate feedback and long-term development.

[0065] The AI ​​driving coach analysis and decision-making unit 4 consists of three functional modules, which correspond to the three-layer structure of the DSRPO algorithm: perception layer, decision layer and path layer, forming a progressive decision-making process of risk identification, teaching response and development optimization.

[0066] In this optional embodiment, based on the student behavior data stream, the perception layer of the driving skill development risk prediction and optimization algorithm is analyzed to predict the student's driving intention. The driving intention is then assessed using a preset risk threshold to obtain the risk assessment results, including: A driving intention recognition model constructed using a temporal neural network is used to identify the student's behavioral data stream and obtain the student's driving intention. Based on the student's driving intention, the system matches it with the scenario-intention-risk mapping table in the preset risk knowledge base, and calculates the risk probability value of the student's driving intention by combining it with real-time environmental status information. The risk probability value is compared with a preset risk threshold. Based on the comparison results, a risk assessment result is obtained, including risk level, risk type, risk probability, and core trigger.

[0067] It should be explained that the driving intention and risk prediction module performs preliminary analysis and risk prediction on the trainee's real-time behavior. Its workflow is as follows: Step 1: Recognize driving intent.

[0068] The system receives student behavior data streams (such as steering wheel angle, accelerator / brake opening, head posture, etc.) from multimodal interaction units and inputs them into a driving intention recognition model built on temporal neural networks such as LSTM to identify the student's current driving intention, such as preparing to change lanes, emergency braking, or going straight through an intersection.

[0069] Step 2: Risk probability calculation.

[0070] The identified driving intentions are matched against a pre-set risk knowledge base. This knowledge base contains a scenario-intention-risk mapping table. For example, on an urban expressway, a lane change intention with a vehicle behind corresponds to a high risk; rainy weather and high-speed driving correspond to a risk of insufficient braking distance.

[0071] Calculate the risk probability value of the driving intention by combining the real-time status of the current virtual environment (such as vehicle distance, weather, lighting, and traffic density).

[0072] Step 3: Risk level determination.

[0073] The risk probability value is compared with a preset risk threshold (e.g., 70% is the high-risk threshold), and a structured risk assessment result is output, which includes risk level (high / medium / low), risk type (operational / judgmental / reactionary), risk probability, and core triggers (e.g., failure to observe blind spots, following too closely).

[0074] In this optional embodiment, based on the risk assessment results and real-time training events, the decision layer of the driving skill development risk prediction and optimization algorithm is used to map the teaching strategy, and the event feedback decision includes: By combining natural language processing algorithms, the risk assessment results and real-time training events are analyzed, based on a standardized event information table containing events, types, and key parameters. Based on a pre-set teaching strategy rule base, a standardized event information table is matched using a multi-condition matching algorithm to obtain preliminary teaching intervention strategies. By using a teaching effectiveness prediction model, preliminary teaching intervention strategies are pre-evaluated and parameters are fine-tuned to obtain event feedback decisions.

[0075] It should be explained that the event feedback-linked decision-making module, based on the risk assessment output from the perception layer, combines real-time training events injected by the dynamic traffic and event simulation unit to perform immediate mapping of teaching strategies. Its workflow is as follows: Step 1: Data analysis and standardization.

[0076] Natural Language Processing (NLP) algorithms are used to perform semantic parsing on risk assessment results and real-time events, extract key elements (such as event type, location, and severity), and generate a standardized event information table containing events, types, and parameters, thereby realizing the structured expression of unstructured information.

[0077] Step 2: Matching teaching strategies.

[0078] Based on a pre-defined teaching strategy rule base, strategies are matched using multi-condition matching algorithms (such as rule engines or decision trees) to generate preliminary teaching intervention strategies.

[0079] Step 3: Strategy pre-assessment and fine-tuning.

[0080] The teaching effectiveness prediction model, trained based on historical intervention data, is invoked to pre-evaluate the expected effects of the strategy (such as student response rate and probability of behavioral correction). If the predicted effects are not met, the intervention format, timing, or expression method is fine-tuned, ultimately generating an event feedback decision.

[0081] In this optional embodiment, based on student historical data and combined with a path optimization algorithm, the event feedback decision is optimized at the path layer of the driving skill development risk prediction optimization algorithm to obtain driving teaching decisions including: Based on students' historical data, key competency indicator sequences are extracted, and personalized skill development trajectories for students are constructed. Based on the preset capability development standard library in the path layer of the driving skill development risk prediction and optimization algorithm, the ideal skill development trajectory of the corresponding trainee is obtained. By using the dynamic time warping algorithm, the personalized skill development trajectory of trainees and the ideal skill development trajectory are aligned and analyzed to obtain the trainees' potential bottleneck nodes and lag dimensions. Based on path optimization algorithms, potential bottleneck nodes for trainees, and lag dimensions, the event feedback decision is optimized to obtain driving instruction decisions.

[0082] It should be noted that the Skills Trajectory and Training Strategy module goes beyond single-event interventions, focusing on the long-term development of learners' abilities and continuously optimizing teaching strategies. Its workflow is as follows: Step 1: Construct a personalized skills development trajectory.

[0083] Based on trainees' historical training data (skill scores, error frequency, intervention response, etc.), time series of key competency indicators are extracted to construct a personalized skill development trajectory that reflects their actual learning progress.

[0084] Step 2: Obtain your ideal skill development trajectory.

[0085] In the path layer of the DSRPO algorithm, a preset capability development standard library is invoked to generate an ideal skill development trajectory (i.e., the optimal learning path) that matches the learner's foundation.

[0086] Step 3: Trajectory alignment and bottleneck location.

[0087] The Dynamic Time Warping (DTW) algorithm is used to nonlinearly align the actual trajectory with the ideal trajectory, and to identify potential bottleneck nodes that are lagging behind or stagnating in development (such as slow improvement in nighttime following distance control capability) and lagging dimensions (such as success rate).

[0088] Step 4: Strategy optimization and instructional decision generation.

[0089] By combining path optimization algorithms (such as reinforcement learning strategy adjustment), the current event feedback decision is reconstructed based on bottleneck nodes and lag dimensions. For example, increasing the frequency of specialized training, designing complex challenge scenarios, and extending the intervention period can ultimately generate driving instruction decisions that take into account both short-term correction and long-term development.

[0090] The Dynamic Traffic and Event Simulation Unit 5 is used to construct adaptive emergencies in a virtual driving environment based on driving instruction decisions, and to train students until the students finish their training.

[0091] It should be explained that Dynamic Traffic and Event Simulation Unit 5, through a four-step closed-loop process of environment construction, event matching, dynamic injection, and difficulty adjustment, achieves a leap from static scenarios to dynamic, intelligent, and personalized teaching events. Dynamic Traffic and Event Simulation Unit 5 significantly improves the realism, relevance, and effectiveness of driver training, providing strong technical support for trainees to master the ability to cope with complex roads in a safe environment.

[0092] In this optional embodiment, the dynamic traffic and event simulation unit 5 includes a traffic flow simulation module, an event triggering module, a dynamic event injection module, and an adaptive difficulty adjustment module; The traffic flow simulation module is used to construct dynamic traffic flow in a virtual driving environment based on a microscopic traffic model, thereby obtaining the traffic environment. Specifically, the traffic flow simulation module is based on micro-traffic models such as Intelligent Driving Model (IDM) to simulate the autonomous behavior of a large number of NPC (non-player) vehicles in a virtual driving environment, including following, changing lanes, merging, and avoiding, forming a dynamic traffic environment with real traffic flow characteristics.

[0093] For example, on urban expressways, the system can be configured to handle medium to high density traffic flow, with NPC vehicles maintaining a reasonable distance and obeying traffic rules, providing a natural and credible background environment for the injection of subsequent emergencies.

[0094] The event triggering module is used to obtain target training events by matching the driving teaching decisions with a preset event triggering rule base; Specifically, the event triggering module receives driving instruction decisions from the AI ​​driving coach analysis and decision-making unit 4, and performs intelligent matching based on the system's preset event triggering rule base. The rule base stores templates of typical traffic events, such as sudden braking by the vehicle in front, pedestrians suddenly crossing the road, vehicles forcibly changing lanes from the side, and sudden severe weather, etc. Each type of event is associated with a specific teaching objective.

[0095] For example, when the teaching decision is aimed at improving braking reaction ability in rainy weather, the system automatically matches wet and slippery road surface and emergency braking event templates of the vehicle in front to generate target training events, ensuring that the event design is highly consistent with the teaching objectives.

[0096] The dynamic event injection module is used to integrate training events into the traffic environment in a multi-stage dynamic evolution manner to obtain sudden events; Specifically, the dynamic event injection module is responsible for accurately injecting target training events into the virtual environment and controlling their development process. Unlike simple one-time events, the dynamic event injection module supports multi-stage dynamic evolution, enhancing the realism of the scenario and the depth of teaching.

[0097] For example, in a tire blowout incident, the system first simulates a slight vehicle deviation, then gradually exacerbates the loss of steering control, and finally triggers the automatic stability system to intervene, guiding the trainee to complete the entire emergency procedure of stabilizing the steering, gently applying the brakes, and pulling over to the side of the road.

[0098] By releasing risk pressure in stages, we help trainees build a systematic emergency response capability, ultimately generating a sudden event with a time dimension and logical progression.

[0099] The adaptive difficulty adjustment module is used to adaptively adjust the parameters of sudden events based on the student's personalized ability profile and real-time operation performance, thus obtaining adaptive sudden events.

[0100] Specifically, the adaptive difficulty adjustment module dynamically adjusts the difficulty parameters of unexpected events based on the student's personalized ability profile and real-time operational performance.

[0101] For example, for beginners, delay the triggering time of the car in front's sudden braking, reduce the braking intensity, and give more reaction time; for advanced learners, add complex challenges, such as rainy nights, heavy fog, and sudden obstacles, to improve their comprehensive coping ability; for learners who repeatedly make mistakes, repeatedly inject similar events in the same scenario to strengthen memory and habit formation.

[0102] By adjusting in real time based on feedback, unexpected events are optimized into adaptive unexpected events, ensuring that the training difficulty is always within the trainee's zone of proximal development, neither too easy to lead to ineffective training, nor too difficult to cause frustration.

[0103] like Figures 2-4 The image shown is an application example diagram of a multimedia interactive virtual reality driver training system. Figure 2 China A and Figure 3 Connect B in the middle. Figure 3 C and Figure 4 (Connected to the C-axis) In practical applications, after logging into the system, trainees enter a highly realistic virtual driving environment. This environment is constructed using the latest 3D modeling technology and high-definition texture mapping. It not only has a lifelike appearance but also dynamic lighting effects, making trainees feel as if they are in a real driving scenario. Whether driving on city roads, country lanes, or in various weather conditions, everything is realistically reproduced in this system. To enhance immersion, the system integrates multiple input devices such as the steering wheel and pedals, combined with haptic feedback devices, allowing trainees to experience the driving process not only visually and audibly but also receive more realistic tactile feedback. For example, if the speed is too high when turning, the system will remind the trainee to slow down through haptic feedback. This multi-sensory interactive method makes the learning process more intuitive and effective.

[0104] Once the training begins, the AI ​​instructor guides the student step-by-step through each driving maneuver, breaking it down into detailed steps. For example, during vehicle startup, the AI ​​instructor guides the student through specific steps such as checking dashboard indicator lights, with detailed voice prompts and visual guidance for each step to ensure accurate execution. If the student deviates from the correct procedure, the AI ​​instructor immediately points out the error and demonstrates the correct approach.

[0105] As learners become more familiar with basic operations, the AI ​​instructor continues to provide detailed guidance during driving. For example, when turning, the AI ​​instructor will first explain the correct turning procedure: "Turn on the turn signal," "Check the rearview mirror to confirm safety," "Slowly turn the steering wheel," and "Gradually accelerate away from the curve." Throughout the process, the system monitors the learner's actions in real time and provides immediate feedback and adjustment suggestions based on the actual situation, helping learners correct mistakes promptly.

[0106] When learning to park, the AI ​​instructor guides you step-by-step, such as: "Turn on the turn signal," "Find a suitable parking space," "Slowly approach and stop," "Pull on the handbrake and turn off the engine." After each operation, the system automatically generates a feedback report, recording the student's performance in detail, pointing out strengths and weaknesses, and providing specific suggestions for improvement.

[0107] Throughout the learning process, the program records every detail of the learner's actions, and the AI ​​algorithm analyzes and provides feedback in real time. After each practice session, the system generates a detailed evaluation report, showcasing not only the learner's strengths but also listing areas for improvement. It also adjusts subsequent teaching content based on each learner's specific situation to ensure a personalized learning path.

[0108] In this way, the system not only solves the limitations of traditional driver training caused by factors such as venue and weather, but also provides more detailed and personalized guidance through the step-by-step teaching method of AI instructors. This allows each student to gradually master driving skills in a safe and efficient environment. This not only improves learning efficiency but also greatly enhances students' enthusiasm and participation, enabling them to gradually grow into qualified drivers without any external constraints.

[0109] In summary, by utilizing the aforementioned technical solutions of this invention, a completely new driver training model is created through the organic combination of automation, unmanned operation, and artificial intelligence technologies. Unmanned management enables 24 / 7 service, greatly enhancing service flexibility and convenience. In terms of product services, this invention utilizes virtual reality technology and multimodal interaction to provide trainees with an immersive driving experience, and combines this with an AI driving assistant to provide personalized teaching plans, ensuring that each trainee receives targeted guidance. In terms of management and operation, the standardization of the training process and data-driven optimization are achieved through cloud databases and automated management systems, improving overall operational efficiency. Regarding application scenarios, this invention is not only suitable for conventional driver training but can also conduct simulated training for special road conditions and extreme weather, meeting the needs of trainees at different levels. This series of innovative measures not only improves training efficiency and service quality but also brings about a transformation in the driver training industry, demonstrating broad market prospects and development potential. This invention breaks through the limitations of traditional driving simulators, such as fixed scenarios, single feedback, and static teaching. It constructs an intelligent driving training system that integrates perception, analysis, decision-making, feedback, and evolution by using a scheduling and data management unit 1, a virtual environment real-time generation unit 2, a multimodal interaction unit 3, an AI driving coach analysis and decision-making unit 4, and a dynamic traffic and event simulation unit 5. This significantly improves the quality and safety of training.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A virtual reality driving training system based on multimedia interaction, characterized in that, The system includes: The scheduling and data management unit is used to perform feature fusion on trainee training requests and historical data to build personalized competency profiles. The virtual environment real-time generation unit is used to render a virtual driving environment based on personalized capability profiles and using maps and a 3D scene engine. The multimodal interaction unit is used to collect and integrate student driving operation data and vehicle status data in real time using interactive devices to obtain student behavior data stream; and to provide multi-dimensional interactive feedback to students based on the student behavior data stream and the driving teaching decisions of the AI ​​driving coach analysis and decision-making unit. The AI ​​driving coach analysis and decision-making unit is used to perform real-time analysis and decision-making on student behavior data streams based on a three-layer fusion architecture of driving skill development risk prediction and optimization algorithms, in order to obtain driving teaching decisions. The dynamic traffic and event simulation unit is used to construct adaptive emergencies in a virtual driving environment based on driving instruction decisions, and to train students until the students finish their training.

2. The virtual reality driving training system based on multimedia interaction according to claim 1, characterized in that, The scheduling and data management unit includes a user management and identity authentication module, a training request processing module, a student information management module, and a personalized ability profile generation module. The user management and identity authentication module is used for account registration, login authentication and permission management for students and administrators; The training request processing module is used to parse the training requests of logged-in and authenticated students to obtain training instructions; The student information management module is used to integrate training requests, training process data, and evaluation results to obtain historical data sources for students. The personalized competency profile generation module is used to extract the corresponding student's historical data source from the student information management module according to the training instructions, clean and extract features from the corresponding student's historical data source to obtain a standardized feature set, and construct a personalized competency profile based on the standardized feature set.

3. The virtual reality driving training system based on multimedia interaction according to claim 2, characterized in that, According to the training instructions, the corresponding student's historical data source is extracted from the student information management module, and the corresponding student's historical data source is cleaned and features are extracted to obtain a standardized feature set. Based on standardized feature sets, personalized capability profiles are constructed, including: The training instructions are parsed, and based on the parsing results, the corresponding historical data sources of students are extracted from the student information management module. The acquired historical data sources of the corresponding students are sequentially subjected to denoising, outlier filtering, and standardization to obtain multi-dimensional features; Multi-dimensional features are filtered to obtain a standardized feature set; a pre-set driving ability mapping model is used to map the standardized feature set to obtain a quantitative driving ability index. By integrating quantitative driving ability indicators, a personalized ability profile is constructed.

4. The virtual reality driving training system based on multimedia interaction according to claim 1, characterized in that, The real-time virtual environment generation unit includes a personalized scene configuration module, a scene data management module, and an environment rendering engine module. The personalized scene configuration module is used to parse the personalized capability profile to obtain scene parameters and environment rendering parameters; The scene data management module is used to schedule and integrate map data, 3D road models, traffic facility models and environmental texture resources based on scene parameters to construct a scene framework; The environment rendering engine module is used to render the scene framework according to the environment rendering parameters and in combination with the 3D graphics engine to obtain a virtual driving environment.

5. The virtual reality driving training system based on multimedia interaction according to claim 1, characterized in that, The multimodal interaction unit includes an operation behavior acquisition module, a vehicle status perception module, an interaction data fusion and synchronization module, and a multi-channel feedback generation module. The operation behavior acquisition module is used to collect the driving operation data of the trainee in real time using the interactive device to obtain the operation behavior data stream. The vehicle state perception module is used to acquire vehicle dynamic parameters in a virtual environment and preprocess the vehicle dynamic parameters to obtain a vehicle state data stream. The interactive data fusion and synchronization module is used to perform timestamp alignment and filtering fusion processing on the operation behavior data and vehicle status data in sequence to obtain the student behavior data stream. The multi-dimensional feedback generation module is used to provide students with multi-dimensional feedback, including tactile, kinesthetic, and auditory feedback, based on the student behavior data stream and the driving teaching decisions of the AI ​​driving coach analysis and decision-making unit, combined with interactive devices.

6. The virtual reality driving training system based on multimedia interaction according to claim 1, characterized in that, The AI ​​driving coach analysis and decision-making unit includes a driving intention and risk prediction module, an event feedback linkage decision-making module, and a skill trajectory and training strategy module. The driving intention and risk prediction module is used to analyze the driving intention of the student based on the student behavior data stream and the perception layer of the driving skill development risk prediction and optimization algorithm, predict the student's driving intention, and evaluate the driving intention based on the preset risk threshold to obtain the risk assessment result. The event feedback linkage decision module is used to map teaching strategies based on risk assessment results and real-time training events, using the decision layer of the driving skill development risk prediction and optimization algorithm to obtain event feedback decisions. The skill trajectory and training strategy module is used to optimize event feedback decisions based on students' historical data and in conjunction with path optimization algorithms at the path layer of the driving skill development risk prediction optimization algorithm, thereby obtaining driving teaching decisions.

7. A virtual reality driving training system based on multimedia interaction according to claim 6, characterized in that, The process involves analyzing student behavior data streams and combining them with a driving skill development risk prediction and optimization algorithm at the perception layer to predict the student's driving intentions. These intentions are then assessed using a preset risk threshold, yielding risk assessment results including: A driving intention recognition model constructed using a temporal neural network is used to identify the student's behavioral data stream and obtain the student's driving intention. Based on the student's driving intention, the system matches it with the scenario-intention-risk mapping table in the preset risk knowledge base, and calculates the risk probability value of the student's driving intention by combining it with real-time environmental status information. The risk probability value is compared with the preset risk threshold. Based on the comparison results, a risk assessment result including risk level, risk type, risk probability and core trigger is obtained.

8. A virtual reality driving training system based on multimedia interaction according to claim 7, characterized in that, The process of mapping teaching strategies based on risk assessment results and real-time training events, using the decision layer of the driving skill development risk prediction and optimization algorithm, to obtain event feedback decisions includes: By combining natural language processing algorithms, the risk assessment results and real-time training events are analyzed, based on a standardized event information table containing events, types, and key parameters. Based on a pre-set teaching strategy rule base, a standardized event information table is matched using a multi-condition matching algorithm to obtain preliminary teaching intervention strategies. By using a teaching effectiveness prediction model, preliminary teaching intervention strategies are pre-evaluated and parameters are fine-tuned to obtain event feedback decisions.

9. A virtual reality driving training system based on multimedia interaction according to claim 8, characterized in that, Based on historical student data and combined with a path optimization algorithm, the event feedback decision is optimized at the path layer of the driving skill development risk prediction and optimization algorithm to obtain driving teaching decisions, including: Based on students' historical data, key competency indicator sequences are extracted, and personalized skill development trajectories for students are constructed. Based on the preset capability development standard library in the path layer of the driving skill development risk prediction and optimization algorithm, the ideal skill development trajectory of the corresponding trainee is obtained. By using the dynamic time warping algorithm, the personalized skill development trajectory of trainees and the ideal skill development trajectory are aligned and analyzed to obtain the trainees' potential bottleneck nodes and lag dimensions. Based on path optimization algorithms, potential bottleneck nodes for trainees, and lag dimensions, the event feedback decision is optimized to obtain driving instruction decisions.

10. A virtual reality driving training system based on multimedia interaction according to claim 1, characterized in that, The dynamic traffic and event simulation unit includes a traffic flow simulation module, an event triggering module, a dynamic event injection module, and an adaptive difficulty adjustment module; The traffic flow simulation module is used to construct dynamic traffic flow in a virtual driving environment based on a microscopic traffic model, thereby obtaining the traffic environment. The event triggering module is used to obtain the target training event by matching the driving teaching decision with a preset event triggering rule base; The dynamic event injection module is used to integrate training events into the traffic environment in a multi-stage dynamic evolution manner to obtain sudden events; The adaptive difficulty adjustment module is used to adaptively adjust the parameters of sudden events based on the student's personalized ability profile and real-time operation performance, so as to obtain adaptive sudden events.

Images