In-vehicle multi-variable environment intelligent simulation system for driving training

Abstract

The application discloses an intelligent simulation system for a multi-variable environment in a vehicle for driving training, and relates to the technical field of driving simulation.The system comprises the following steps: obtaining a plurality of driving sample scenes, identifying an external environment sample data set and an internal environment sample data set; constructing an internal and external mapping data set; training based on the internal and external mapping data set, and outputting an internal environment mapping model; obtaining a first external simulation scene for driving training of a current training user; inputting the internal environment mapping model, obtaining first internal simulation parameters; and connecting a simulation device, which simulates the internal environment according to the first internal simulation parameters.The application solves the technical problem that the driving training environment simulation in the prior art is single, lacks authenticity, and cannot provide a comprehensive, authentic and variable driving training environment, and achieves the technical effect of accurately simulating a multi-variable environment in a vehicle according to driving training requirements, and improves the authenticity and diversity of driving training.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of driving simulation, and particularly relates to an intelligent simulation system for a multi-variable environment in a vehicle for driving training. BACKGROUND

[0002] With the increasing number of vehicles, road traffic safety problems are increasingly prominent, and improving the driving skills and the ability to cope with complex environments of drivers becomes the key. Driving training is crucial for the growth of novice drivers and the skill improvement of experienced drivers, and the traditional driving training method mainly relies on actual road driving and simple field training, which has many limitations. In actual road driving training, although a real traffic environment can be provided, it is difficult for drivers to experience various road conditions and environmental changes in a short period of time due to the limitations of time, place, weather and other conditions. For example, in a specific season or region, a driver is difficult to frequently encounter rainstorm, snowstorm, heavy fog and other adverse weather scenes for training; at the same time, actual road training needs to consider traffic flow and safety factors, and cannot simulate special road conditions such as road construction and emergency accident scenes at will.

[0003] The existing driving training environment simulation is single and lacks real sense, and it is difficult to provide a comprehensive and real multi-variable driving training environment. SUMMARY

[0004] The present application provides an intelligent simulation system for a multi-variable environment in a vehicle for driving training, which solves the technical problem that the existing driving training environment simulation is single and lacks real sense, and it is difficult to provide a comprehensive and real multi-variable driving training environment.

[0005] The present application provides an intelligent simulation system for a multi-variable environment in a vehicle for driving training, which includes:

[0006] The sample data set identification module is used to obtain a plurality of driving sample scenes, and identify corresponding vehicle external environment sample data sets and vehicle internal environment sample data sets of the plurality of driving sample scenes; the vehicle internal and external mapping data set construction module is used to perform time sequence synchronization processing according to the vehicle external environment sample data sets and the vehicle internal environment sample data sets, and construct a vehicle internal and external mapping data set; the vehicle internal environment mapping model output module is trained based on the vehicle internal and external mapping data set, and outputs a vehicle internal environment mapping model; the first vehicle external simulation scene acquisition module is used to obtain a first vehicle external simulation scene used for driving training by a current training user; the first vehicle internal simulation parameter acquisition module is used to input the first vehicle external simulation scene into the vehicle internal environment mapping model, and obtain first vehicle internal simulation parameters corresponding to the first vehicle external simulation scene; and the vehicle internal environment simulation module is used to connect a simulation device, and the simulation device simulates a vehicle internal environment according to the first vehicle internal simulation parameters.

[0007] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0008] The sample data set identification module is used to obtain a plurality of driving sample scenes, and identify corresponding vehicle external environment sample data sets and vehicle internal environment sample data sets; the vehicle internal and external mapping data set construction module is used to perform time sequence synchronization processing, and construct a vehicle internal and external mapping data set; the vehicle internal environment mapping model output module is trained based on the vehicle internal and external mapping data set, and outputs a vehicle internal environment mapping model; the first vehicle external simulation scene acquisition module is used to obtain a first vehicle external simulation scene used for driving training by a current training user; the first vehicle internal simulation parameter acquisition module is used to input the first vehicle external simulation scene into the vehicle internal environment mapping model, and obtain first vehicle internal simulation parameters corresponding to the first vehicle external simulation scene; and the vehicle internal environment simulation module is used to connect a simulation device, and the simulation device simulates a vehicle internal environment according to the first vehicle internal simulation parameters. The technical effect of accurately simulating a variable vehicle internal environment according to driving training requirements is achieved, and the authenticity and diversity of driving training are improved. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0010] Figure 1A structural schematic diagram of an in-vehicle multi-variable environment intelligent simulation system for driving training is provided for the embodiments of the present application.

[0011] Figure 2 A flowchart of outputting an in-vehicle environment mapping model of the in-vehicle multi-variable environment intelligent simulation system for driving training is provided for the embodiments of the present application.

[0012] Label explanation: sample data set identification module 10, in-vehicle and out-of-vehicle mapping data set construction module 20, in-vehicle environment mapping model output module 30, first out-of-vehicle simulation scene acquisition module 40, first in-vehicle simulation parameter acquisition module 50, in-vehicle environment simulation module 60. DETAILED DESCRIPTION

[0013] The present application provides an in-vehicle multi-variable environment intelligent simulation system for driving training, which is used to solve the technical problem that the driving training environment simulation is single, lacks real sense, and it is difficult to provide a comprehensive real and variable driving training environment in the prior art.

[0014] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0015] As shown in the embodiments, the present application provides an in-vehicle multi-variable environment intelligent simulation system for driving training, which comprises: Figure 1

[0016] A sample data set identification module 10 is used to acquire a plurality of driving sample scenes, and identify corresponding out-of-vehicle environment sample data sets and in-vehicle environment sample data sets of the plurality of driving sample scenes.

[0017] ​Specifically, the sample dataset identification module 10 plays a role in data collection and preliminary processing in the intelligent simulation system of the driving training vehicle in a variable environment. First, a plurality of driving sample scenarios are obtained through various means, including but not limited to field driving data collection, scene library in driving simulation software, and typical scene cases provided by professional driving training institutions, etc. The obtained driving sample scenarios are diverse and cover various road conditions such as urban congested roads, highways, mountainous bends, etc.; different weather conditions such as sunny, rainy, snowy, foggy, etc.; and various traffic conditions such as busy traffic flow during peak hours, night low-visibility driving, etc. For each driving sample scenario, various sensors installed in different positions of the vehicle are used to identify the corresponding vehicle external environment sample dataset. The collection of the vehicle external environment sample dataset involves multiple aspects, such as installing high-precision meteorological sensors on the top of the vehicle to obtain real-time temperature, humidity, air pressure, and weather conditions (such as raindrop size, snowfall intensity, fog concentration, etc.); equipping the vehicle with multiple cameras and radars around the body to capture traffic flow, vehicle type, road signs, obstacle position, etc. information; at the same time, through the vehicle-mounted GPS and map data, the geographical position and road type (such as urban trunk road, rural road, etc.) of the vehicle are determined. At the same time, it is also responsible for identifying the vehicle internal environment sample dataset. Various sensors are arranged inside the vehicle, such as pressure sensors and position sensors installed on the seats to obtain the sitting posture and seat usage of the driver and passengers; light sensors are set near the instrument panel to detect the light intensity inside the vehicle; temperature and humidity sensors and noise sensors are installed in the vehicle space to monitor the temperature, humidity and noise level inside the vehicle. Through the cooperative work of these sensors, the sample dataset identification module 10 can comprehensively and accurately collect the vehicle external environment sample dataset and the vehicle internal environment sample dataset corresponding to each driving sample scenario, providing a solid data foundation for subsequent vehicle internal and external mapping dataset construction and vehicle internal environment mapping model training.

[0018] The vehicle internal and external mapping dataset construction module 20 is used for time sequence synchronization processing according to the vehicle external environment sample dataset and the vehicle internal environment sample dataset, and constructing a vehicle internal and external mapping dataset.

[0019] Specifically, the inside-outside mapping dataset construction module 20 undertakes the key data integration task in the entire driving training simulation system. The module first acquires the outside environment sample dataset and the inside environment sample dataset collected by the sample dataset identification module. These datasets contain rich environmental information, but due to the time difference in the collection process, time synchronization processing is needed. High-precision timestamp labeling technology is used to label the collection time for each piece of data in the outside environment sample dataset (such as outside temperature, humidity, traffic conditions, etc.) and the corresponding data in the inside environment sample dataset (such as inside temperature, noise, seat state, etc.). Then, through the data alignment algorithm, the outside environment data and the inside environment data at the same time are corresponded one by one according to the timestamp, forming a series of ordered data pairs. These data can clearly reflect the correlation between the outside environment factors and the inside environment state at a specific time point. For example, when the outside temperature suddenly rises at a certain time, the corresponding inside temperature change and other related environmental factors (such as the running state of the inside air conditioner, whether the driver opens the window, etc.) are reflected. After such time synchronization processing, the inside-outside mapping dataset is successfully constructed, which provides an accurate and time sequence logical data basis for the subsequent training of the inside environment mapping model, enabling the model to learn the dynamic relationship between the inside and outside environment changes, and thus more accurately simulate and predict the change trend of the inside environment under different outside environment conditions.

[0020] The inside environment mapping model output module 30 is trained based on the inside-outside mapping dataset and outputs the inside environment mapping model.

[0021] Specifically, the in-vehicle environment mapping model output module 30 undertakes the heavy task of constructing a high-precision in-vehicle environment mapping model in the intelligent simulation system for driving training in a variable in-vehicle environment. It optimizes the training process with the help of a generative adversarial network. First, the module obtains a trained generative adversarial network, which is composed of a generator and a discriminator that cooperate with each other and is connected to the input end of the in-vehicle environment mapping model, forming an organic whole. At the beginning of training, the in-vehicle and out-of-vehicle mapping dataset is input into the generator in the generative adversarial network. The generator uses its powerful neural network architecture and advanced learning algorithm to deeply analyze the complex relationships and features contained in the in-vehicle and out-of-vehicle mapping dataset. It continuously adjusts its parameters to learn how to generate realistic in-vehicle and out-of-vehicle environment data pairs, thereby outputting an in-vehicle and out-of-vehicle mapping generated dataset. These generated data can simulate the out-of-vehicle environment changes that may occur in real scenarios but are not directly presented in the original dataset and their corresponding in-vehicle environment states to some extent. Subsequently, the generated in-vehicle and out-of-vehicle mapping generated dataset is used to expand the original in-vehicle and out-of-vehicle mapping dataset, combining the virtually generated data with the actual collected real data to construct an expanded dataset with larger data volume, more extensive data distribution, and greater diversity. This expansion operation greatly enriches the sample space of the training data, enabling the in-vehicle environment mapping model to learn more diverse in-vehicle and out-of-vehicle environment mapping relationships. Based on the expanded dataset and the identification information that indicates the learning accuracy, the in-vehicle environment mapping model is trained. During the training process, the in-vehicle environment mapping model takes the expanded dataset as input and continuously adjusts its internal parameters to minimize the error between the model's predicted results and the actual expected output (indicated by the identification information). At the same time, the discriminator in the generative adversarial network continuously discriminates between the data generated by the generator and the data in the original dataset, and its discrimination accuracy is fed back to the generator as a key feedback signal. The generator continuously optimizes its generation ability based on the feedback of the discrimination accuracy, striving to generate data that can confuse the discriminator, i.e., data that is closer to the real data distribution. Through this repeated adversarial interaction training between the generator and the discriminator, as well as the in-depth learning of the in-vehicle environment mapping model on the expanded dataset, a high-performance and highly accurate in-vehicle environment mapping model is finally output. This model can accurately calculate and output highly matched in-vehicle simulation parameters based on any input out-of-vehicle simulation scenario information, providing solid core model support for the driving training simulation system and ensuring the authenticity and reliability of the simulated in-vehicle environment, significantly improving the effectiveness and quality of driving training.

[0022] The first out-of-vehicle simulation scenario acquisition module 40 is configured to acquire a first out-of-vehicle simulation scenario for the current training user to use for driving training.

[0023] Specifically, the first off-car simulation scene acquisition module 40 is responsible for accurately obtaining the first off-car simulation scene for driving training in the intelligent simulation system of the driving training car in a variable environment. First, interact with the training course setting module or user-defined training requirement interface of the system to obtain the training target, skill level, and individualized training requirements of the current training user. If the user is a novice driver, the training target may focus on mastering basic driving skills, such as vehicle control in normal road conditions and weather conditions; while for experienced drivers, it is more inclined to respond to training in complex road conditions and adverse weather scenarios. According to this information, appropriate scenes are selected from the rich scene library built into the system, which contains various types of off-car simulation scenes, such as busy traffic intersection scenes in urban centers with complex traffic light rules, dense pedestrian and vehicle flow; highway scenes involving different speed limits, lane changes, and vehicle spacing requirements; and various special weather scenes, such as reduced visibility in heavy rain, road flooding, slippery roads in heavy snow, and low visibility. The module will select the scene with the highest matching degree as the first off-car simulation scene according to the user's training needs. For example, if the user needs to enhance rain driving skills, the module will preferentially select a rain scene that includes different rainfall intensities, road slipperiness levels, and traffic conditions. In addition, the module also has the ability to dynamically adjust the simulation scene according to real-time feedback and training progress. If the user makes modifications to the current scene difficulty or type during training, the module will immediately re-evaluate the user's needs and select a more appropriate first off-car simulation scene from the scene library to ensure that the driving training always meets the user's expectations and training goals, providing the most targeted and effective driving training experience for the user.

[0024] The first in-car simulation parameter acquisition module 50 is used to input the first off-car simulation scene into the in-car environment mapping model to obtain the first in-car simulation parameters corresponding to the first off-car simulation scene.

[0025] In particular, the first in-vehicle simulation parameter acquisition module 50 plays a key role in parameter conversion and generation in the intelligent simulation system of the driving training vehicle in a variable environment. First, it receives the first off-vehicle simulation scene information determined by the first off-vehicle simulation scene acquisition module. This scene information covers a wealth of off-vehicle environment data, such as weather conditions (sunny, rainy, snowy, foggy, etc.), temperature, humidity, light intensity, traffic flow, road type (urban road, highway, rural road, etc.), and surrounding topographic features. Subsequently, these detailed first off-vehicle simulation scene data are input into the pre-trained in-vehicle environment mapping model. Based on the learning of a large number of in-vehicle and off-vehicle mapping data sets, the in-vehicle environment mapping model has established a complex mapping relationship model between off-vehicle environmental factors and in-vehicle environmental parameters. Upon receiving the off-vehicle simulation scene input, the model rapidly processes and analyzes these data through its internal neural network structure or other algorithmic mechanisms. According to the learned mapping rules, the model calculates the first in-vehicle simulation parameters corresponding to the current off-vehicle simulation scene. These parameters include, but are not limited to, the specific numerical value to which the in-vehicle temperature should be adjusted to simulate the impact of off-vehicle temperature on the in-vehicle environment; the humidity setting value reflecting the in-vehicle humidity trend under off-vehicle humidity changes; the noise level parameter simulating the effect of off-vehicle traffic noise, wind and rain sound, etc. entering the in-vehicle environment; the light intensity parameter creating an in-vehicle light atmosphere matching the off-vehicle weather and time; the seat position adjustment parameter providing appropriate driving posture suggestions according to driving scene requirements (such as high-speed driving or complex road conditions); and the simulated display state of the instrument panel display information (such as speedometer, tachometer, warning lights, etc.), providing realistic driving visual feedback for the driver. Through this process, the first in-vehicle simulation parameter acquisition module 50 successfully acquires the first in-vehicle simulation parameters highly matching the first off-vehicle simulation scene, providing accurate data basis for creating a realistic in-vehicle environment for the subsequent in-vehicle environment simulation module.

[0026] The in-vehicle environment simulation module 60 is used to connect simulation devices that simulate the in-vehicle environment according to the first in-vehicle simulation parameters.

[0027] Specifically, the in-vehicle environment simulation module 60 is a key execution part of the in-vehicle intelligent simulation system for realizing the realistic reproduction of the in-vehicle environment. The module is first connected with various professional simulation devices, including but not limited to a high-precision temperature regulation system, a humidity control system, an advanced noise simulation generator, an adjustable lighting device, a multifunctional seat adjustment device, and a highly simulated instrument panel display system, etc. When receiving the first in-vehicle simulation parameters from the first in-vehicle simulation parameter acquisition module, the module quickly distributes these parameters to the corresponding simulation devices. The temperature regulation system accurately controls the heating or cooling device in the vehicle according to the set temperature parameter, so that the in-vehicle temperature quickly and stably reaches the simulation requirement. For example, when simulating cold weather scenes, the in-vehicle temperature is lowered to the corresponding low temperature state, so that the driver can actually feel the impact of the cold environment on driving operations. The humidity control system adjusts the humidity level in the vehicle according to the humidity parameter to create a dry or humid in-vehicle environment atmosphere, just like driving in different climate conditions. The noise simulation generator generates realistic traffic noise, wind and rain sound, engine roar, etc. according to the noise level parameter, so that the driver is immersed in the real driving scene in terms of hearing. The adjustable lighting device simulates the in-vehicle lighting effect under different times (such as morning, noon, and evening) and weather (sunny, cloudy, rainy) according to the light intensity parameter, including direct sunlight, shadow blocking, and different degrees of light dimming, etc. affecting the driver's visual perception and operation judgment. The multifunctional seat adjustment device automatically adjusts the seat position, backrest angle, height, etc. according to the seat position adjustment parameter, providing the driver with a comfortable and reasonable sitting posture in line with the current driving scene. The highly simulated instrument panel display system accurately displays the vehicle speed, speed, fuel quantity, various warning lights, etc. according to the instrument panel display information parameter, which is matched with the simulated scene outside the vehicle, providing visual feedback for the driver, so that he can obtain the vehicle running state information like in real driving. Through the coordinated work of these simulation devices, the in-vehicle environment simulation module 60 successfully creates a highly realistic in-vehicle environment according to the first in-vehicle simulation parameters, greatly improving the authenticity and effectiveness of driving training, and enabling the driver to fully experience and cope with various variable driving scenes in a safe simulation environment.

[0028] In one possible implementation manner, as shown in Figure 2 The in-vehicle environment mapping model output module 30 comprises:

[0029] An adversarial network acquisition unit is configured to acquire a trained adversarial network, wherein the adversarial network is connected with an input end of the in-vehicle environment mapping model.

[0030] The data learning unit is configured to perform data learning on the indoor-outdoor mapping dataset according to the adversarial network, and output an indoor-outdoor mapping generated dataset.

[0031] The indoor environment mapping model output unit is configured to perform dataset expansion on the indoor-outdoor mapping dataset according to the indoor-outdoor mapping generated dataset, and perform training according to the expanded dataset and the identification information for identifying the learning accuracy, and output an indoor environment mapping model.

[0032] Specifically, in the driving training indoor variable environment intelligent simulation system, in order to improve the accuracy and generalization ability of the indoor environment mapping model, an adversarial network technology is used for optimization. First, a trained adversarial network is obtained, which is composed of a generator and a discriminator, and is connected to the input end of the indoor environment mapping model to form a collaborative working architecture.

[0033] In the data learning stage, the indoor-outdoor mapping dataset is input into the generator in the adversarial network. The generator uses its internal complex neural network structure and learning algorithm to deeply mine and learn the data features in the indoor-outdoor mapping dataset. By continuously adjusting its own parameters, the generator tries to generate new data similar to the original dataset but with certain changes, thereby outputting an indoor-outdoor mapping generated dataset. This generated dataset contains virtual indoor-outdoor environment data pairs created by the generator according to the learned patterns and rules. These data pairs can simulate scenarios that may occur in real situations but are not explicitly included in the original dataset to some extent.

[0034] The in-vehicle environment mapping model takes the expanded dataset as input, and gradually reduces the error between the output of the model and the expected output represented by the identification information by continuously adjusting the parameters inside the model. In this process, the discriminator in the adversarial network also plays an important role. The discriminator is responsible for distinguishing and judging the data generated by the generator and the data in the original dataset, and its discrimination accuracy is transmitted as a feedback signal to the generator, prompting the generator to continuously optimize its generation ability to generate high-quality data that is more difficult to distinguish by the discriminator. Through this adversarial training between the generator and the discriminator, and the learning of the in-vehicle environment mapping model on the expanded dataset, an in-vehicle environment mapping model with better performance and higher accuracy is finally output. This optimized model can more accurately predict the in-vehicle environment parameters according to the simulated outdoor scene, and provide more realistic and actual in-vehicle environment simulation for driving training.

[0035] In one possible implementation, the adversarial network acquisition unit further includes:

[0036] The generated dataset acquisition unit is configured to input the in-vehicle and out-of-vehicle mapping dataset into the generator, acquire a generated dataset after multiple learning by data learning of the generator, and output the generated dataset.

[0037] The discrimination accuracy acquisition unit is configured to input the generated dataset after multiple learning into the discriminator for discrimination, acquire discrimination accuracy, and output the discrimination accuracy.

[0038] The feedback optimization unit is configured to perform feedback optimization on the generator according to the discrimination accuracy if the discrimination accuracy is less than a preset threshold, until the discrimination accuracy of the discriminator is greater than or equal to the preset threshold, and acquire a trained adversarial network.

[0039] Specifically, first, the inside-out mapping dataset is input into the generator, which starts data learning based on its complex neural network structure and learning algorithm. It tries to extract key data features and potential mapping relationship patterns from the inside-out mapping dataset. During the learning process, the generator generates new simulated inside-out environment data pairs by continuously adjusting its internal parameters according to the learned patterns. These data pairs constitute the generated dataset after multiple learning. The goal of the generator is to generate data similar to the distribution of the real inside-out mapping dataset to deceive the discriminator.

[0040] In the adversarial network training process of the intelligent simulation system for the variable environment in the driving training vehicle, inputting the generated dataset after multiple learning into the discriminator to obtain the discrimination accuracy is a key step. After receiving the generated dataset, the discriminator starts its complex discrimination mechanism. It first performs detailed feature analysis on each inside-out environment data in the generated dataset. These features include, but are not limited to, temperature variation trend, humidity fluctuation range, light intensity distribution pattern, traffic flow variation law in the outside environment data, and temperature adjustment amplitude, humidity response characteristics, noise simulation level, and light simulation effect in the corresponding inside environment data. The discriminator uses pre-set discrimination rules and algorithms, which are formed based on the learning and summary of a large amount of real inside-out mapping data. By comparing the data features in the generated dataset with these rules, the discriminator calculates the similarity of each data pair to the real data distribution. For example, for the response change of the inside temperature when the outside temperature rises, there may be a certain variation range and law in the real data, and the discriminator checks whether the corresponding change in the generated data conforms to this law. After discriminating each data in the generated dataset, the discriminator calculates the discrimination accuracy according to the discrimination results of all data pairs. The discrimination accuracy is usually calculated by dividing the number of correctly discriminated data pairs by the total number of data pairs to obtain a proportion value. This proportion value directly reflects the discrimination ability of the discriminator for the generated dataset. If the discrimination accuracy is high, it means that the data generated by the generator is similar to the features and distribution of the real data, and the performance of the generator is good; otherwise, if the discrimination accuracy is low, it means that there are many differences between the data generated by the generator and the real data, and further optimization is needed. By obtaining the discrimination accuracy, the current generation effect of the generator can be accurately evaluated, providing key basis for subsequent feedback optimization, ensuring that the adversarial network can continuously improve its ability to generate high-quality simulated data, and thus providing more reliable data support for the training of the inside environment mapping model.

[0041] If the discrimination accuracy is less than the preset threshold, it indicates that the difference between the data generated by the generator and the real data is still large, and it is easy to be identified by the discriminator. At this time, according to the problem reflected by the discrimination accuracy, the relevant information is transmitted to the generator as feedback. The generator adjusts its internal parameters according to these feedbacks, and improves the way and quality of generating data. Then, the updated generated data set is input into the discriminator for discrimination again, and the above process is repeated to continuously optimize the generation ability of the generator. In this way, the generator continues to improve, and the discriminator continuously discriminates and feeds back until the discrimination accuracy of the discriminator is greater than or equal to the preset threshold. At this time, it indicates that the generator has been able to generate data that is realistic enough, and the distribution is highly similar to the real data, which is difficult to be distinguished by the discriminator. At this time, the trained adversarial network is successfully obtained, and the generator of the trained adversarial network can effectively generate high-quality simulation data, which provides a solid foundation for the subsequent expansion of the in-vehicle and out-of-vehicle mapping data set and the optimization training of the in-vehicle environment mapping model, thereby improving the accuracy and reliability of the entire driving training simulation system for simulating the in-vehicle environment.

[0042] In a possible implementation manner, the sample data set identification module 10 comprises:

[0043] A scene type identification unit is configured to identify scene types of the plurality of driving sample scenes.

[0044] A feature parameter label set construction unit is configured to identify scene feature parameters according to the scene types, and construct a feature parameter label set corresponding to each driving sample scene.

[0045] An out-of-vehicle environment sample data set acquisition unit is configured to acquire out-of-vehicle environment sample data sets corresponding to the plurality of driving sample scenes according to the feature parameter label set.

[0046] Specifically, in the intelligent simulation system of the variable environment in the driving training vehicle, identifying the scene type of multiple driving sample scenes is the primary step to build a high-quality simulation environment, and various data collection methods and analysis means are used to achieve this purpose. First, the visual information of the driving sample scene is captured by the vehicle-mounted camera, including road conditions, sky conditions, traffic signs, and the situation of surrounding vehicles and pedestrians, etc. For example, if the sky is blue and sunny and the road is dry in the picture taken by the camera, it is a preliminary indication of a sunny day scene; if there is rain, water on the road, and the wiper is in working condition in the picture, it is inclined to judge as a rainy day scene. At the same time, the meteorological sensor equipped on the vehicle also plays an important role. The environmental data collected by the temperature sensor, humidity sensor, and barometric pressure sensor provides a key basis for the judgment of the scene type. For example, in the case of high temperature, high humidity, and relatively low air pressure, combined with the signs of overcast sky in the visual information, it can assist in judging as a thunderstorm weather scene. In addition, the GPS positioning system combined with map data can determine the geographical location and road type of the vehicle, further assisting the identification of the scene type. If the vehicle is located in a mountain road and the surrounding environment shows many curves, large terrain undulations, combined with the vegetation and terrain features in the camera picture, it is helpful to judge as a mountain driving scene. Through the comprehensive analysis and processing of these multi-source data, the type of each driving sample scene is accurately identified, laying a foundation for the subsequent identification of feature parameters and sample data set collection for different scene types, and thus building a more realistic and comprehensive driving training simulation environment.

[0047] In the intelligent simulation system of the variable environment in the driving training vehicle, the identification of scene feature parameters according to the identified scene type and the construction of the corresponding feature parameter tag set are the key links to accurately simulate the vehicle environment. For sunny scenes, light is the most critical feature parameter. The intensity, angle, and distribution of sunlight are accurately measured by light sensors. For example, at noon, the sunlight is directly on the ground, and the light intensity may reach the maximum value. At this time, the sun angle is close to perpendicular to the ground, and the light distribution is relatively uniform. These parameters are crucial for simulating the light environment in the vehicle. At the same time, the color and transparency of the sky on a sunny day also affect the light atmosphere inside the vehicle. By capturing images with a camera and analyzing their color characteristics and clarity, relevant parameters are obtained. These parameters related to sunny light are integrated to construct feature parameter tags such as "sunny day-high light intensity-specific sun angle-uniform light distribution-blue sky", etc., forming the feature parameter tag set corresponding to the sunny day scene. In rainy scenes, humidity and wind speed become the key feature parameters to focus on. The humidity sensor monitors the outdoor environment humidity in real time, accurately recording the humidity range under different rainfall intensities, because high humidity not only affects the visibility of the vehicle window, but also affects the vehicle's humidity regulation system. The anemometer measures wind speed and direction, and different wind speeds will change the motion trajectory and impact force of raindrops, thereby affecting the vehicle's driving stability and the driver's vision. In addition, raindrop size and density are also important parameters, which can be obtained through special raindrop sensors or camera image analysis technology. These rain-related parameters are summarized to construct feature parameter tags such as "rainy day-high humidity range-different wind speed-specific raindrop size and density", etc., forming the feature parameter tag set of the rainy day scene. For night scenes, brightness and light distribution are the core feature parameters. The light sensor measures the brightness of the ambient light, including the combined brightness level of streetlights, other vehicle headlights, and moonlight. The camera is used to analyze the light distribution, such as the spacing and illumination range of streetlights, the angle and high-beam distribution of vehicle headlights, etc. At the same time, consider the light changes at different time periods at night, such as the gradual dimming of light in the evening, the dark environment and uniform light distribution at night, etc. These night light-related parameters are sorted out to construct feature parameter tags such as "night-low brightness range-specific light distribution-light changes at different time periods", etc., forming the feature parameter tag set of the night scene. Through such detailed feature parameter identification and tag set construction for different scene types, accurate guidance is provided for subsequent accurate collection of outdoor environment sample data sets, ensuring that the simulation system can accurately reproduce the environmental characteristics of various driving scenes.

[0048] According to the characteristic parameter tag set, a plurality of driving sample scene corresponding vehicle external environment sample data set is collected. For each characteristic parameter represented by the tag, the corresponding sensor is used for data collection. For example, for the light intensity, a light sensor is used; for the humidity, a humidity sensor is used; for the wind speed, a wind speed meter is used. By deploying these sensors in different scenes, the data of various environmental factors can be accurately collected, so as to construct a comprehensive and accurate vehicle external environment sample data set, provide a solid data foundation for subsequent construction of vehicle internal and external mapping data set and training of vehicle internal environment mapping model, and ensure that the simulation system can truly restore the vehicle external environment in different driving scenes, and improve the effectiveness and authenticity of driving training.

[0049] In a possible implementation manner, the vehicle internal environment mapping model output module 30 further includes:

[0050] A scene switching mode judgment unit is configured to judge whether the scene used by the current training user for driving training is in a scene switching mode.

[0051] A scene switching node acquisition unit is configured to acquire a scene switching node if the driving training scene of the current training user is in the scene switching mode, wherein the scene switching node is a node at which the first vehicle external simulation scene switches to the second vehicle external simulation scene.

[0052] A second vehicle internal simulation parameter acquisition unit is configured to acquire a second vehicle internal simulation parameter corresponding to the second vehicle external simulation scene according to the vehicle internal environment mapping model.

[0053] A switching simulation unit is configured to switch the simulation of the vehicle internal environment according to the second vehicle internal simulation parameter after the simulation device simulates the vehicle internal environment according to the first vehicle internal simulation parameter.

[0054] Specifically, in the driving training vehicle internal variable environment intelligent simulation system, the scene switching simulation function provides the driver with a richer and more challenging training experience. First, it is necessary to judge whether the scene used by the current training user for driving training is in a scene switching mode. This judgment is based on the training course logic preset by the system and the operation instruction of the user. For example, if the user selects a comprehensive training course containing scene switching, or manually triggers a scene switching instruction during training, it is determined that the current scene is in a scene switching mode.

[0055] When it is determined to be a scene switching mode, a scene switching node is immediately acquired, which refers to the accurate time or position point at which the first off-board simulation scene is switched to the second off-board simulation scene. The determination of this node is crucial for ensuring the continuity and authenticity of scene switching. The scene switching node is accurately located through monitoring of the time progress of the simulation scene and tracking of the position of the vehicle in the virtual environment. For example, when a city road driving scene is switched to a highway driving scene, the scene switching node is set to the point at which the vehicle travels to the junction of the city road and the highway entrance, or the simulation time reaches a predetermined switching time.

[0056] Once the scene switching node is acquired, the second in-vehicle simulation parameters corresponding to the second off-board simulation scene are quickly calculated and acquired according to the trained in-vehicle environment mapping model. The in-vehicle environment mapping model is trained based on a large number of in-vehicle and out-vehicle mapping data sets, and can accurately predict the adjustment requirements of in-vehicle environment parameters corresponding to different off-board environment changes. For the second off-board simulation scene, considering its unique environmental factors, such as the change in wind noise caused by higher speed on the highway, different lighting conditions, and more stable temperature environment, etc., the corresponding in-vehicle temperature, humidity, noise, and lighting parameters are calculated.

[0057] In terms of simulation equipment, first, the in-vehicle environment is simulated according to the first in-vehicle simulation parameters, so that the driver experiences a real driving environment in the initial scene. When the scene switching node is reached, the simulation equipment is quickly switched to simulate the in-vehicle environment with the second in-vehicle simulation parameters. For example, the in-vehicle temperature regulation system adjusts from the appropriate temperature in the city road scene to a slightly lower temperature in the highway scene to simulate the ventilation effect when driving at high speed; the noise simulation equipment increases the intensity of wind noise and tire noise to create a noisy environment on the highway; the lighting system adjusts the brightness and angle of the light according to the time and weather conditions of the new scene. Through such a scene switching simulation process, the driver's adaptability and response skills when switching between different driving scenes can be effectively improved, providing a training environment that is closer to actual driving conditions.

[0058] In one possible implementation manner, the switching simulation unit further includes:

[0059] A parameter difference index output unit is configured to perform parameter difference analysis on the first in-vehicle simulation parameters and the second in-vehicle simulation parameters, and output a parameter difference index.

[0060] A switching transition module activation unit is configured to activate a switching transition module if the parameter difference index is greater than a preset difference index.

[0061] A transition switching simulation unit is configured to simulate a transition of the in-vehicle environment according to the switching transition module.

[0062] Specifically, a detailed parameter difference analysis is performed on the first in-vehicle simulation parameters and the second in-vehicle simulation parameters. The first in-vehicle simulation parameters correspond to the in-vehicle environment settings in the initial driving scenario (e.g., sunny urban road driving), including temperature, humidity, noise level, and light intensity. The second in-vehicle simulation parameters are the ideal in-vehicle environment parameters corresponding to the post-switching scenario (e.g., rainy night highway driving). During the analysis, the differences between each parameter are compared one by one. For example, in terms of temperature, the in-vehicle temperature may be comfortable during sunny urban road driving, but it needs to be appropriately lowered during rainy night highway driving due to the high speed and low outdoor temperature. In terms of humidity, the humidity in a rainy night environment is significantly different from that in a sunny urban road environment. The noise level will significantly increase in a rainy night highway scenario due to wind and rain sounds and the noise generated by high-speed driving. The light intensity will also be significantly weakened due to the combined effects of night and rain. By accurately calculating the differences in these parameters in terms of numerical values, trends, and overall impact on the in-vehicle environment, a comprehensive parameter difference index is output.

[0063] The parameter difference index is compared with a preset difference index, which is a threshold set according to a large amount of actual driving scenario switching data and the sensitivity of human perception to environmental changes. If the calculated parameter difference index is greater than the preset difference index, it means that the in-vehicle environment changes significantly due to the scenario switch, and direct switching may cause a sense of strangeness to the driver, affecting the authenticity and effectiveness of the simulation training. At this time, the system activates the switching transition module.

[0064] The switching transition module simulates a transition of the in-vehicle environment according to the parameter difference index and the preset transition strategy. Taking the rainy night scenario as an example, assuming that it is switched from a sunny urban road, the in-vehicle temperature will not instantly drop to the low temperature required for rainy night highway driving, but will gradually decrease at a certain rate, simulating the gradual change process of the in-vehicle temperature affected by the external environment; the humidity will gradually increase from a low level, taking into account factors such as window fogging, and the in-vehicle humidity will be gradually adjusted by simulation equipment and the defogging effect will be simulated; the noise level will gradually increase from the relatively quiet environment of the urban road to the noisy environment of the rainy night highway, simulating the gradual increase of wind, rain, and tire friction sound on the wet road; the light intensity will also gradually dim from the bright state of the sunny day, simulating the gradual darkening effect of the light in the night and rainy day. Through such a transition switching simulation, the in-vehicle environment can naturally transition from the initial state to the post-switching scenario state, allowing the driver to more realistically experience the environmental changes brought about by the scenario switch during the simulation driving process, improving their ability to cope with complex and variable driving environments, and making the driving training more realistic.

[0065] In a possible implementation, the transition switching simulation unit further includes:

[0066] a preset transition step configuration unit, configured to configure a preset transition step.

[0067] an in-vehicle simulation parameter acquisition unit, configured to input the preset transition step into the switching transition module after the activation of the switching transition module, and acquire transition simulation parameters of the first in-vehicle simulation parameter and the second in-vehicle simulation parameter.

[0068] an in-vehicle environment transition switching unit, configured to perform transition switching simulation on the in-vehicle environment according to the transition simulation parameters.

[0069] Specifically, in the intelligent in-vehicle multi-variable environment simulation system for driving training, to realize smooth transition of the in-vehicle environment during scene switching, a preset transition step is configured to accurately control the gradual change process of the environment parameters. First, the preset transition step is configured according to the characteristics of the driving scene and the human perception law of environmental change. The preset transition step is a series of numerical values used to control the gradual change of the in-vehicle environment parameters. The setting of these numerical values is based on reasonable estimation of the change speed of different environmental factors and consideration of ensuring that the driver can naturally adapt to the environmental change. For example, for the temperature parameter, the transition step can be set to reduce or increase a certain number of degrees per minute to simulate the rate of temperature change in the real environment; for the humidity change, the transition step can be expressed as an increase or decrease of a certain percentage of humidity per second.

[0070] When the switching transition module is activated, the preset transition step is input into the module. The switching transition module calculates the transition simulation parameters according to these transition steps and the first in-vehicle simulation parameter and the second in-vehicle simulation parameter. Taking the example of switching from a sunny driving scene to a rainy night driving scene, if the initial in-vehicle temperature is 25℃ and the target temperature after switching is 18℃, the preset transition step is 1℃ per minute, then during the transition process, according to the time advancement, the module will calculate the corresponding transition simulation temperature values at different times, such as 24℃ after 1 minute, 23℃ after 2 minutes, and so on, to generate the temperature transition simulation parameters during the entire transition period. Similarly, for other environmental parameters such as humidity, noise, and illumination, the corresponding transition simulation parameters will also be calculated according to the respective preset transition steps and the initial and target parameter values.

[0071] Finally, the in-vehicle environment is switched and simulated by using the transition simulation parameters. The temperature adjusting system in the vehicle adjusts the temperature in the vehicle according to the temperature transition simulation parameters, so that the driver can gradually adapt to the change in temperature. The humidity control system simulates the gradual change in the humidity in the vehicle according to the humidity transition simulation parameters, while cooperating with the window fogging simulation effect. The noise simulation device gradually increases the noise level according to the noise transition simulation parameters, simulating the transition from a relatively quiet environment on a sunny day to a noisy environment on a rainy night. The lighting system gradually dims the light in the vehicle according to the light transition simulation parameters, simulating the transition from bright light to dim light. Through this transition simulation method based on the preset transition step, the in-vehicle environment can smoothly and naturally transition from the first in-vehicle simulation scene to the second in-vehicle simulation scene during the scene switching process, providing the driver with a more realistic, comfortable and actual driving experience, and effectively improving the quality and effect of the driving training.

[0072] In a possible implementation manner, the first in-vehicle simulation scene acquisition module 40 comprises:

[0073] a mixed scene judgment unit, configured to judge whether the first out-of-vehicle simulation scene is a mixed scene.

[0074] a plurality of mixed scene determination units, configured to determine a plurality of mixed scenes if the first out-of-vehicle simulation scene is a mixed scene.

[0075] a plurality of in-vehicle simulation parameter acquisition units, configured to acquire a plurality of in-vehicle simulation parameters corresponding to the plurality of mixed scenes respectively according to the in-vehicle environment mapping model.

[0076] a fusion in-vehicle simulation parameter output unit, configured to perform parameter fusion on the plurality of in-vehicle simulation parameters, and output a fusion in-vehicle simulation parameter.

[0077] a mixed simulation unit, configured to perform mixed simulation on the in-vehicle environment according to the fusion in-vehicle simulation parameter by using the simulation device.

[0078] Specifically, in the intelligent simulation system of the variable environment in the driving training vehicle, determining whether the first off-vehicle simulation scene is a mixed scene is an important beginning to realize complex environment simulation. First, a comprehensive feature extraction and analysis is performed on the first off-vehicle simulation scene, and various types of information in the scene are obtained through various data acquisition devices and technical means, such as meteorological sensors collecting temperature, humidity, air pressure, precipitation type and other meteorological data, cameras capturing visual information of road conditions, traffic signs, surrounding objects and weather phenomena, GPS positioning combined with map data to determine geographical location and topographic features, and vehicle sensors obtaining dynamic information such as vehicle speed and driving direction. For these collected data, a preset scene recognition algorithm and rules are used for comprehensive judgment. For example, if the meteorological data shows that there is both rain and low temperature, and the camera picture shows that the road is snow-covered or icy, and the geographical location information indicates that it is a mountain road, it is preliminarily judged that the scene is a mixed scene of rain and snow weather and mountain road conditions; if the camera captures road construction signs and traffic congestion flow conditions, and combines with map data to confirm that it is a city construction road section, it is inclined to determine that it is a mixed scene of construction and congestion. However, in order to ensure the accuracy of the judgment, the mutual relationship and influence degree between various environmental factors are further considered. For example, when determining whether it is a mixed scene of bad weather and special road conditions, not only the existence of bad weather elements (such as rain, snow, fog, etc.) and special road condition elements (such as curves, steep slopes, construction road sections, etc.) is confirmed, but also the influence of bad weather on the driving difficulty of special road conditions is analyzed whether it meets the characteristics of mixed scenes. For example, the water accumulation in heavy rain weather will increase the risk of driving on a curve, and if the road water accumulation data in the scene and the curve curvature data confirm this influence relationship, it will be more determined that the scene is a mixed scene. Through such detailed and multi-dimensional analysis and judgment, whether the first off-vehicle simulation scene is a mixed scene is accurately identified, which provides a reliable basis for subsequent processing of mixed scenes, and thus provides a real driving training environment for the driver.

[0079] When the first outdoor simulation scene is determined to be a mixed scene, it is necessary to further determine the multiple mixed scene elements, i.e., multiple specific sub-scenes, that constitute the mixed scene. First, the various types of data collected related to the scene are analyzed in depth, including meteorological data, road condition data, geographic information data, and vehicle state data, etc. For example, in a city driving scene determined to be a mixed scene, the meteorological data shows that it is raining and the temperature is low, the road condition data indicates that there are waterlogged sections and traffic congestion, the geographic information data determines that it is located in the downtown business district, and the vehicle state data shows that the vehicle speed is low and frequent start-stop. Based on these data, the multiple mixed scene elements are determined according to the preset scene classification rules and feature recognition algorithms. For the above example, the rainy weather is considered as a mixed scene element, characterized by increased humidity, obstructed vision, slippery road surface, etc.; traffic congestion is another mixed scene element, manifested as vehicle density, slow driving, and driver's high concentration on maintaining distance and frequent start-stop operations; waterlogged sections are also an important mixed scene element, which affects the vehicle's handling and increases the risk of vehicle skidding; and the low temperature environment is a separate scene element that affects the in-vehicle temperature regulation and the driver's thermal comfort. In determining these mixed scene elements, not only their independent characteristics are considered, but also their interactions and influence relationships are analyzed. For example, rainy weather and waterlogged sections are interrelated, as the rain leads to the formation of waterlogging, which further affects driving safety in rainy weather; traffic congestion and low temperature environment jointly affect the engine's slow warming during low-speed driving, which may affect the in-vehicle heating supply and thus the driver's comfort and attention. Through such comprehensive and detailed analysis, the multiple mixed scene elements can be accurately determined, laying a foundation for subsequent acquisition of corresponding in-vehicle simulation parameters and ultimately realizing realistic mixed scene simulation, enabling the driver to fully experience the complex and variable driving environment in the mixed scene during simulation training.

[0080] According to the in-vehicle environment mapping model, the corresponding in-vehicle simulation parameters are obtained for each determined mixed scene element. The in-vehicle environment mapping model is trained based on a large number of in-vehicle and out-of-vehicle mapping data sets, and can accurately predict the in-vehicle environment parameters corresponding to changes in different single environmental factors. For the heavy rain weather element, the model considers factors such as raindrop impact sound, increased humidity, and obstructed vision to affect in-vehicle noise, humidity, and lighting parameters, and calculates the corresponding in-vehicle simulation parameters; for the dense fog environment element, the model focuses on the impact of reduced visibility on in-vehicle lighting and driver's visual perception to determine the corresponding in-vehicle simulation parameters such as lighting intensity and contrast.

[0081] The plurality of in-vehicle simulation parameters are fused, in the fusion process, according to the weight distribution of each mixed scene element on the influence of the in-vehicle environment, the interaction relationship between different parameters is comprehensively considered, and each in-vehicle simulation parameter is weighted and summed or other reasonable fusion calculation. For example, in the mixed scene of heavy rain and thick fog, if the influence weight of heavy rain on noise is larger, and the influence weight of thick fog on light is larger, then in the fusion of in-vehicle simulation parameters, the parameter changes of these two aspects will be highlighted accordingly, and a fused in-vehicle simulation parameter considering the combined influence of heavy rain and thick fog is obtained.

[0082] The simulation device simulates the in-vehicle environment according to the fused in-vehicle simulation parameters, the in-vehicle noise simulation device emits a sound containing heavy rain sound and relatively quiet but special sound in thick fog environment according to the fused noise parameter; the humidity control system adjusts to the fused humidity level to simulate the high humidity brought by heavy rain and the possible humidity change in thick fog environment; the light system creates an in-vehicle light atmosphere that meets the dark light in heavy rain and the low visibility in thick fog according to the fused light parameter. Through such a mixed simulation process, the driver can experience the real in-vehicle environment change in the complex mixed scene in the simulation driving, and improve his ability to deal with complex and changeable environment in actual driving.

[0083] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0084] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0085] The present application is only an exemplary description of the present application, and should be considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application intends to include these modifications and changes.

Claims

1. An intelligent simulation system for variable in-vehicle environments used in driver training, characterized in that, The system includes: A sample dataset identification module is used to acquire multiple driving sample scenarios and identify the vehicle exterior environment sample dataset and vehicle interior environment sample dataset corresponding to the multiple driving sample scenarios. The vehicle interior and exterior mapping dataset construction module is used to construct the vehicle interior and exterior mapping dataset by performing time-series synchronization processing on the vehicle exterior environment sample dataset and the vehicle interior environment sample dataset. The in-vehicle environment mapping model output module is trained based on the in-vehicle and out-of-vehicle mapping dataset and outputs the in-vehicle environment mapping model. The first vehicle exterior simulation scene acquisition module is used to acquire the first vehicle exterior simulation scene used by the current training user for driving training. The first in-vehicle simulation parameter acquisition module is used to input the first external simulation scene into the in-vehicle environment mapping model and obtain the first in-vehicle simulation parameters corresponding to the first external simulation scene. An in-vehicle environment simulation module is provided, which is used to connect to a simulation device, and the simulation device simulates the in-vehicle environment according to the first in-vehicle simulation parameters. The in-vehicle environment mapping model output module further includes: A scene switching mode determination unit is used to determine whether the scene used by the current training user for driving training is a scene switching mode. A scene switching node acquisition unit is used to acquire a scene switching node if the current training user's driving training scenario is a scene switching mode, wherein the scene switching node is the node that switches from the first vehicle exterior simulation scenario to the second vehicle exterior simulation scenario. The second in-vehicle simulation parameter acquisition unit is used to acquire the second in-vehicle simulation parameters corresponding to the second external simulation scenario based on the in-vehicle environment mapping model. A switching simulation unit is used for the simulation device to simulate the in-vehicle environment using the second in-vehicle simulation parameters after simulating the in-vehicle environment according to the first in-vehicle simulation parameters. The switching simulation unit further includes: The parameter difference index output unit is used to perform parameter difference analysis on the first in-vehicle simulation parameters and the second in-vehicle simulation parameters, and output parameter difference index. A switching transition module activation unit is used to activate the switching transition module if the parameter difference index is greater than a preset difference index. A transition switching simulation unit is used to simulate a transition switching of the in-vehicle environment according to the switching transition module. The transition simulation unit includes: A preset transition step size configuration unit is used to configure a preset transition step size; The in-vehicle simulation parameter acquisition unit is used to activate the activation switching transition module, input the preset transition step size into the switching transition module, and acquire the transition simulation parameters of the first in-vehicle simulation parameters and the second in-vehicle simulation parameters. The in-vehicle environment transition switching unit is used to simulate the transition switching of the in-vehicle environment using the transition simulation parameters.

2. The intelligent simulation system for variable in-vehicle environment used for driver training as described in claim 1, characterized in that, The in-vehicle environment mapping model output module includes: An adversarial network acquisition unit is used to acquire a trained adversarial network, wherein the adversarial network is connected to the input end of the in-vehicle environment mapping model; A data learning unit is configured to learn data from the vehicle interior and exterior mapping dataset based on the adversarial network and output a vehicle interior and exterior mapping generated dataset. The in-vehicle environment mapping model output unit is used to expand the in-vehicle interior and exterior mapping dataset based on the in-vehicle interior and exterior mapping generated dataset, train the model based on the expanded dataset and the label information of the label learning accuracy, and output the in-vehicle environment mapping model.

3. The intelligent simulation system for variable in-vehicle environment used for driver training as described in claim 2, characterized in that, The adversarial network includes a generator and a discriminator, and the generator and the discriminator are trained alternately. The adversarial network acquisition unit includes: A dataset acquisition unit is provided, which is used to input the vehicle interior and exterior mapping dataset into the generator, perform data learning based on the generator, and acquire the generated dataset after multiple learning iterations. A discrimination accuracy acquisition unit is used to input the generated dataset after multiple learning sessions into the discriminator for discrimination and to obtain the discrimination accuracy. A feedback optimization unit is configured to optimize the generator based on the discrimination accuracy if the discrimination accuracy is less than a preset threshold, until the discrimination accuracy of the discriminator is greater than or equal to the preset threshold, thereby obtaining a trained adversarial network.

4. The intelligent simulation system for variable in-vehicle environment used for driver training as described in claim 1, characterized in that, The sample dataset identification module includes: A scene type identification unit, wherein the scene type identification unit is used to identify the scene type of multiple driving sample scenes; A feature parameter label set construction unit is used to identify scene feature parameters according to the scene type and construct a feature parameter label set corresponding to each driving sample scene. The vehicle exterior environment sample dataset collection unit is used to collect vehicle exterior environment sample datasets corresponding to the multiple driving sample scenarios according to the feature parameter label set.

5. The intelligent simulation system for variable in-vehicle environment used for driver training as described in claim 1, characterized in that, The first vehicle exterior simulation scene acquisition module also includes: A mixed scene determination unit is used to determine whether the first vehicle exterior simulation scene is a mixed scene; A multiple mixed scene determination unit is configured to determine multiple mixed scenes if the first vehicle exterior simulation scene is a mixed scene. Multiple in-vehicle simulation parameter acquisition units are used to acquire multiple in-vehicle simulation parameters corresponding to the multiple mixed scenarios according to the in-vehicle environment mapping model. A vehicle in-vehicle simulation parameter output unit is used to fuse the multiple vehicle in-vehicle simulation parameters and output fused vehicle in-vehicle simulation parameters. A hybrid simulation unit is used by the simulation device to perform a hybrid simulation of the in-vehicle environment based on the fused in-vehicle simulation parameters.

Images