Vehicle automatic braking system parameter adjusting method and system based on artificial intelligence

Abstract

The invention discloses a vehicle automatic braking system parameter adjusting method and system based on artificial intelligence. The method comprises the following steps: S1, collecting real-time data of an AEB system in a vehicle driving process by using various sensors assembled on a vehicle, and transmitting the real-time data to a cloud platform; s2, the cloud platform firstly carries out time sequence and space alignment processing and processing on the collected data; s3, an appropriate algorithm is automatically selected through artificial intelligence, and various parameters of the AEB system are calculated and adjusted according to the processed data; and S4, after AEB parameter adjustment is completed, performing a multi-scene simulation test in a virtual environment, and evaluating the performance of the AEB system in different driving scenes. By introducing an AEB parameter adjustment method based on artificial intelligence, real-time dynamic adjustment of the AEB system is realized, manual intervention is reduced, continuous optimization is performed by using real-time data feedback, and the intelligence and adaptive ability of the system are improved, so that the performance and safety of the AEB system in different driving scenes are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of automotive autonomous driving, and particularly relates to a parameter adjustment method and system for an automatic emergency braking system. By utilizing artificial intelligence technology, real-time dynamic adjustment of AEB system parameters is achieved, improving the safety and efficiency of the autonomous driving system. Background Art

[0002] In the field of automotive autonomous driving, the AEB system is one of the key technologies to ensure vehicle safety. Currently, there are many defects in the parameter adjustment method of the AEB system: 1. Lack of real-time dynamic adjustment ability: Traditional AEB systems mostly rely on fixed parameter settings and are difficult to adjust parameters in real time according to dynamic factors such as road conditions, traffic conditions, and weather during actual driving. For example, when the road surface is wet and slippery in rainy weather, the braking distance of the vehicle will increase significantly, but the traditional system cannot adjust the braking parameters in time to adapt to this change, resulting in poor braking effect in case of emergency and increasing the collision risk.

[0003] 2. Excessive manual intervention: The parameter adjustment of existing AEB systems often relies on manual settings or a large number of static test results. This not only consumes a lot of manpower, material resources, and time, but also it is difficult for manual setting to comprehensively consider various complex driving scenarios, resulting in the parameter settings not being optimal and affecting the system performance.

[0004] 3. Unable to make full use of real-time data feedback: Current AEB systems are difficult to use real-time data for continuous optimization. When the road or environment changes, the system cannot quickly adjust parameters according to the new data, making the AEB system unable to perform optimally in different driving situations.

[0005] 4. Slow algorithm optimization process: Some AEB systems based on machine learning can optimize parameters, but they require a lot of time and have high requirements for the quality of training data. Therefore, rapid adjustment and optimization cannot be achieved.

[0006] The main reasons for these defects include: lack of adaptive algorithms, existing systems are mostly based on fixed control rules and cannot automatically optimize to adapt to complex and changing environments; strong coupling between hardware and software, cumbersome testing and verification are required when the system is upgraded or adjusted, restricting the ability of real-time adjustment and response to environmental changes; long testing and verification cycles, parameter optimization depends on long-term testing and it is difficult to quickly feedback new driving scenarios; lag in data and models, models trained based on historical data cannot reflect changes in the real situation in time. Summary of the Invention

[0007] The present invention aims to provide a method and system for parameter adjustment of an automatic emergency braking system, so as to solve the problems existing in the parameter adjustment of the existing AEB system, such as lack of adaptability, poor real-time performance, excessive manual intervention, and slow optimization process.

[0008] To solve the above technical problems, the present invention is implemented by the following solutions: A method for adjusting parameters of a vehicle automatic braking system based on artificial intelligence, comprising the following steps: Step S1: Data collection: Use a variety of sensors installed on the vehicle to collect real-time data and historical test data of the AEB system during vehicle driving. The real-time data includes vehicle state information, information about obstacles ahead, and environmental information, and transmit the collected data to the cloud platform; Step S2: Data processing: The cloud platform first performs time-series and spatial alignment processing on the collected data, then performs data cleaning, and then extracts features related to the AEB system parameters from the cleaned data. For example, extract features such as the distance, speed, and acceleration of obstacles from radar data; extract features such as the shape, size, and position of obstacles from image data.

[0009] Step S3: AEB parameter adjustment: Use artificial intelligence to automatically select a suitable algorithm, and calculate and adjust various parameters of the AEB system according to the processed data; Step S4: Multi-scenario simulation test, feedback, and optimization: After completing the AEB parameter adjustment, automatically simulate multiple scenarios in a virtual environment for simulation testing, evaluate the performance of the AEB system under different driving scenarios, and if the evaluation results do not meet expectations, continue to optimize the parameters according to the feedback.

[0010] For further optimization, the data collection in step S1 specifically includes: Step S1.1: Install a variety of types of sensors on the test vehicle to form an all-round data acquisition network; the sensors include but are not limited to millimeter-wave radar, cameras, wheel speed sensors, acceleration sensors, gyroscopes, humidity sensors, temperature sensors, and light sensors.

[0011] For example, a high-precision millimeter-wave radar is installed at the front of the vehicle. Its detection range can reach more than 200 meters, and the angular resolution is between 1° and 3°. It is used to accurately measure the distance, speed, and angle information of the obstacles in front. Multiple cameras are arranged around the vehicle body, such as front-view, rear-view, and side-view cameras. The resolution of the front-view camera reaches 1080P and above, and it has a wide-angle field of view. It is used to identify the shape, type (such as vehicles, pedestrians, non-motor vehicles, etc.) of the obstacles and their moving directions. At the same time, wheel speed sensors are installed at the wheels to accurately monitor the wheel rotation speed, and the vehicle speed is calculated based on this. The accuracy can reach 0.1 km / h. An acceleration sensor and a gyroscope are installed at the chassis position to obtain the acceleration and attitude change information of the vehicle. In addition, the road surface conditions and weather information are monitored through humidity sensors, temperature sensors, and light sensors. The accuracy of the humidity sensor can reach ±2%RH, and the accuracy of the temperature sensor is within ±0.5°C.

[0012] Step S1.2: To ensure the real-time and accuracy of the data, set each sensor to a different acquisition frequency; the millimeter-wave radar acquires data at a frequency of 50 - 100 times per second to quickly capture the dynamic information of the obstacles in front; the camera captures images at a rate of 25 - 30 frames per second to ensure continuous monitoring of the surrounding environment; the acquisition frequencies of the wheel speed sensor, acceleration sensor, and gyroscope are higher, reaching 100 - 200 times per second, to obtain the real-time changes in the vehicle's motion state.

[0013] Step S1.3: The data collected by the sensors is transmitted to the vehicle-mounted data processing unit through the Controller Area Network (CAN) bus in the vehicle. In the vehicle-mounted data processing unit, the data is initially packaged and format-converted, and then the data is transmitted to the cloud platform in real time through the 4G / 5G communication module. To ensure the stability and security of data transmission, an encryption transmission protocol, such as the SSL / TLS protocol, is used to encrypt the data to prevent the data from being stolen or tampered with during transmission.

[0014] For further optimization, in step S2, the cloud platform preprocesses the collected data, specifically including: Step 2.1: Time series alignment: After the cloud platform receives the data, it first performs time series alignment. Due to the differences in the sampling moments and data transmission delays of different sensors, it is necessary to unify the time reference. Using high-precision time synchronization technology, such as the Network Time Protocol (NTP), and taking the clock of the cloud platform as the reference, calibrate the timestamps of the data of each sensor. For the data with a large timestamp deviation, use methods such as linear interpolation or polynomial fitting for time compensation to ensure that the data of different sensors at the same moment can accurately reflect the actual state of the vehicle and the surrounding environment. For example, if the timestamp difference between the data of the millimeter-wave radar and the camera is 50 milliseconds, through the analysis and calculation of the front and back data, correct the timestamp to make the two data accurately aligned in time.

[0015] Step 2.2: Spatial alignment: Since the installation positions and measurement coordinate systems of different sensors are different, spatial alignment is required. Based on the vehicle's mechanical structure and the installation parameters of the sensors, a spatial transformation model of the sensors is established. For camera data, the internal and external parameters of the camera are obtained through camera calibration technology, and the pixel coordinates in the image are converted into world coordinates; for millimeter-wave radar data, according to its installation angle and position, the distance and angle information measured by the radar are converted into the unified vehicle coordinate system using the coordinate transformation formula. Through spatial alignment, it is ensured that the information such as the position and speed of the front obstacle obtained from different sensors is consistent in space, providing an accurate data basis for subsequent data analysis and processing.

[0016] Step 2.3: Data cleaning and feature extraction: After temporal and spatial alignment, the data is cleaned. Outliers are removed, such as values of the distance measured by the millimeter-wave radar that far exceed the reasonable range, which are judged and removed by setting thresholds; for missing values, data filling algorithms are used for processing, such as filling based on the mean, median or linear interpolation method of adjacent data. Then feature extraction is carried out. Features such as acceleration and deceleration are calculated from the vehicle speed data; features such as the size and shape of the obstacle (analyzed through camera images), relative speed and acceleration are extracted from the front obstacle information; features such as the road surface friction coefficient (calculated according to humidity, temperature and other data combined with empirical models) and slope are extracted from the road surface condition data. These features will provide key bases for the subsequent adjustment of AEB parameters.

[0017] For further optimization, in step S3, an artificial intelligence is used to automatically select appropriate algorithms, specifically including: Step S3.1: Construct an algorithm candidate pool. According to different driving scenarios and the requirements of the AEB system, determine the algorithms applicable to each scenario, including but not limited to PID control algorithms, fuzzy control algorithms and machine learning algorithms; organize these applicable algorithms into an algorithm candidate pool, and label the applicable scenario ranges and performance indicators for each algorithm.

[0018] Among them, the PID control algorithm is applicable to some relatively stable and linear scenarios, and can adjust the AEB parameters according to the current error, error integral, and error derivative to make the system reach a stable state. For example, when the vehicle is driving at a constant speed on the highway, the PID control algorithm can better adjust the braking parameters. The fuzzy control algorithm is applicable to scenarios with uncertainty and ambiguity, such as driving in bad weather. The fuzzy control algorithm can reason according to fuzzy rules to adjust the AEB parameters. The fuzzy control algorithm does not require an accurate mathematical model and can better handle complex environments. Machine learning algorithms include supervised learning, unsupervised learning, and reinforcement learning, etc. Supervised learning algorithms, such as linear regression, logistic regression, decision trees, etc., can learn the relationship between parameters and braking effects based on historical data to predict the optimal parameters. The reinforcement learning algorithm interacts with the environment and continuously tries different parameter settings to maximize a certain reward function, which is applicable to dynamically changing scenarios.

[0019] Step S3.2: Algorithm evaluation and selection: Evaluate the algorithms according to four indicators: accuracy, response speed, stability, and adaptability, and select a suitable algorithm according to the algorithm selection strategy.

[0020] The algorithm selection strategy specifically includes: Rule-based selection: Select a suitable algorithm from the algorithm candidate pool according to the result of scene recognition and preset rules. For example, if the scene recognition is driving on the highway and the vehicle is in a constant speed state, the PID control algorithm is preferentially selected. Performance evaluation-based selection: Conduct real-time performance evaluation on each candidate algorithm and select the algorithm with the best performance according to the evaluation indicators. Historical data can be used to conduct offline evaluation of the algorithms, or online evaluation can be carried out during the actual operation process. For example, within a certain period of time, count the accuracy, response speed and other indicators of each algorithm in different scenarios and select the algorithm with the best comprehensive performance. Hybrid strategy: Combine rule-based selection and performance evaluation-based selection; First, screen out a part of applicable algorithms according to the scene rules, and then conduct performance evaluation on these algorithms, and finally select the optimal algorithm.

[0021] The scene recognition is to classify different driving scenes using scene classification algorithms. Common scenes include driving on urban roads, driving on highways, driving in parking lots, driving in bad weather, etc. Each scene can be further subdivided. For example, driving on urban roads can be divided into congested sections, non-congested sections, etc.

[0022] Scene classification includes rule-based classification and machine learning classification. The rule-based classification is to judge the current driving scene according to some preset rules. For example, if the vehicle speed is lower than 20 km / h, and the distance to the obstacle in front is relatively close and the vehicle starts and stops frequently, then it can be judged as a scene of an urban congested section.

[0023] Machine learning is classified as using classification algorithms, such as decision trees, support vector machines, neural networks, etc., to classify driving scenarios. By learning from a large amount of historical data, the model can automatically identify the characteristic patterns of different scenarios. For example, training a neural network model with vehicle state data and environmental data as inputs and scenario categories as outputs.

[0024] The four evaluation metrics include: Accuracy: The degree to which the AEB parameters predicted by the algorithm can enable the vehicle to make a correct braking response in the actual scenario. For example, when approaching an obstacle ahead, the braking parameters selected by the algorithm can enable the vehicle to stop just within the safe distance. Response speed: The ability of the algorithm to quickly output appropriate parameters after receiving the data. In an emergency, the response speed is crucial and can reduce the braking delay time. Stability: The consistency and reliability of the algorithm in outputting parameters under different conditions. For example, when encountering similar scenarios continuously for multiple times, the algorithm can select stable parameters. Adaptability: The ability of the algorithm to adapt to different driving scenarios and environmental changes. For example, when the weather changes from sunny to rainy, the algorithm can timely adjust the parameters to adapt to the new road conditions.

[0025] Furthermore, in step S3.2, for dynamically changing scenarios, a machine learning algorithm is selected to adjust the AEB parameters, specifically including: Step S3.2.1: Algorithm selection and model training: Using the cloud platform to build a model combining a convolutional neural network (CNN) and a long short-term memory network (LSTM), and training the model to enable the model to learn the relationship between data features and AEB parameters.

[0026] Adopting a deep learning algorithm, such as a model combining a convolutional neural network (CNN) and a long short-term memory network (LSTM), to analyze the processed data. CNN is used to extract features in the image data, such as the shape and texture information of obstacles; LSTM is used to process time series data and capture the changing trends of the motion states of the vehicle and obstacles. In the early stage, a large amount of vehicle data under different driving scenarios is collected, including scenarios such as normal driving, following a vehicle, overtaking, and encountering an obstacle, as well as the ideal parameter settings of the corresponding AEB system. These data are labeled and divided into a training set and a test set. The training set is used to train the model, and the weights and biases of the model are continuously adjusted through the backpropagation algorithm to enable the model to learn the relationship between data features and AEB parameters. During the training process, the early stopping method is adopted to prevent the model from overfitting, and the training is stopped when the performance of the model on the test set no longer improves.

[0027] Step S3.2.2: Generation of Parameter Adjustment Rules: The trained model automatically generates parameter adjustment rules based on the input data. For different driving scenarios and vehicle states, the model outputs corresponding AEB system parameter adjustment schemes; such as braking start time, braking force, braking deceleration, etc. For example, when the model detects an obstacle approaching rapidly ahead and the vehicle speed is high, it will generate an adjustment rule of braking in advance and increasing the braking force; if the vehicle is driving at a low speed and the obstacle is at a far distance, the model will generate a rule of applying a smaller braking force and delaying braking appropriately. At the same time, in order to deal with possible incorrect decisions of the model, a parameter penalty rule is formulated. If the parameters generated by the model cause the vehicle to be unstable in the simulation test (such as the vehicle losing control due to excessive braking) or fail to effectively avoid collisions, the output of the model is penalized and the weights of the model are adjusted to make it more cautious and accurate in subsequent decisions.

[0028] Step S3.2.3: Automatic Parameter Adjustment: The cloud platform sends parameter adjustment instructions to the AEB system of the vehicle according to the generated parameter adjustment rules. After receiving the instructions, the AEB system adjusts its own parameter settings in real time. During the adjustment process, the dynamic characteristics and safety limitations of the vehicle are considered to ensure that the adjusted parameters will not have a negative impact on the stability and safety of the vehicle. For example, the increase in braking force should not exceed the maximum friction force between the tire and the road surface to prevent wheel lock-up; the advance of the braking start time should consider the driver's reaction time to avoid unnecessary emergency braking that causes discomfort to the driver.

[0029] For further optimization, in step S3.2, when the parameters generated by the model cause the vehicle to be unstable in the simulation test or fail to effectively avoid collisions, the output of the model is penalized, specifically including: Step S3.2.2.1: Penalty Based on Performance Metrics: Define metrics for evaluating the performance of the AEB system, including but not limited to braking distance, collision speed, and vehicle stability, and set a range or threshold for each performance metric. For example, the braking distance should be within the safe distance, and the collision speed should be as low as possible.

[0030] When the parameter settings of the AEB system cause a certain performance metric to exceed the ideal range or threshold, these parameters are penalized. For example, if the actual braking distance exceeds the safe distance, the score of the parameter setting is reduced, and a corresponding penalty coefficient is given according to the degree of exceeding.

[0031] Step S3.2.2: Penalty Based on Scenario Adaptability First, classify the driving scenarios, categorizing different driving scenarios such as highway scenarios, urban street scenarios, rainy day scenarios, etc. Set specific parameter adjustment targets and limitations according to the characteristics and safety requirements of each scenario. For example, in highway scenarios, the AEB system is required to respond quickly and provide sufficient braking force; in rainy day scenarios, the impact of slippery roads on braking performance needs to be considered. When the parameter settings of the AEB system do not meet the requirements of a specific scenario, penalties are imposed on these parameters. For example, if in a rainy day scenario, the parameter settings cause the vehicle to skid during braking, penalties are imposed on this parameter setting. The degree of penalty is determined according to the importance of the scenario and the degree of deviation of the parameter settings from the requirements. For example, in highway scenarios, more severe penalties are given to parameter settings that do not meet the requirements.

[0032] For further optimization, step S4 specifically includes: Step S4.1: Multi-Scenario Simulation Test Verify the adjusted parameters in a multi-scenario simulation test environment. The simulation environment simulates various real driving scenarios, including different weather conditions (sunny, rainy, snowy, foggy), road conditions (highway, urban street, rural road, curve, slope), obstacle types (vehicles, pedestrians, animals of different sizes and shapes), and different driving states of the vehicle (accelerating, decelerating, cruising, turning), etc. In each simulation scenario, the simulated vehicle travels according to the adjusted AEB parameters, and indicators such as the vehicle's motion state, the change in distance from obstacles, and whether a collision occurs are monitored. Through a large number of simulation tests, comprehensively evaluate the performance of the adjusted AEB parameters in different scenarios.

[0033] Step S4.2: Result Feedback and Evaluation Feed back the results of the simulation test to the developers and the optimization module of the cloud platform. The developers evaluate the results from a professional perspective, checking whether the braking effect of the AEB system in different scenarios meets the expectations, such as whether a collision is successfully avoided, whether the braking process is smooth, and the impact on vehicle stability. At the same time, the optimization module of the cloud platform conducts a quantitative analysis of the results through preset evaluation indicators, such as calculating the average braking distance, collision avoidance rate, vehicle stability index, etc. If the test results meet the expectations, that is, all evaluation indicators reach the preset standards, the system will solidify these parameters and apply them to the actual AEB system; if the results do not meet the expectations, the optimization module analyzes the problem based on the feedback data, optimizes the algorithm, adjusts the parameters again, and conducts simulation tests until the test results meet the expectations. Through continuous feedback and optimization, make the parameter adjustment of the AEB system more accurate and efficient, and improve the performance and safety of the AEB system in various actual driving scenarios.

[0034] An AEB parameter adjustment system based on artificial intelligence, comprising: Data collection module: used to collect real-time data of the AEB system by using various sensors installed on the vehicle and transmit the data to the cloud platform; Data processing and adjustment module: set on the cloud platform, used to perform time alignment and space alignment processing on the collected data, automatically select appropriate algorithms by using artificial intelligence, and calculate and adjust various parameters of the AEB system according to the processed data; Multi-scenario simulation test module: set on the cloud platform, used to verify the performance of the AEB system in different driving scenarios through multi-scenario tests in a virtual environment after completing the AEB parameter adjustment; Feedback and optimization module: used to feedback the parameters adjusted by the AEB parameter adjustment module to the developers for evaluation. If the evaluation result does not meet the expectation, the parameters will be continuously optimized according to the feedback.

[0035] Compared with the prior art, the beneficial effects of the present invention are: 1. The method of the invention can intelligently adjust the parameters of the AEB system according to the real-time collected vehicle state information, front obstacle information and environmental information, adapt to the complex and changeable actual driving environment, and improve the braking effect and safety of the AEB system.

[0036] 2. The automated adjustment process in the method of the present invention reduces manual participation, improves the adjustment efficiency and accuracy, and reduces the adjustment cost and time.

[0037] 3. During the parameter adjustment process of the method of the invention, the changes of the vehicle state and the front obstacle are continuously monitored, and the parameter adjustment strategy is dynamically adjusted in real time to ensure that the AEB system is always in the best working state. Brief description of the drawings

[0038] Figure 1 It is a flowchart of the method for adjusting the parameters of the vehicle automatic braking system based on artificial intelligence according to the present invention. Detailed implementation manners

[0039] The present invention will be further described below in conjunction with embodiments, but it should not be understood that the above-mentioned subject scope of the present invention is limited to the following embodiments. Embodiment 1:

[0040] As Figure 1 shown, the method for adjusting the parameters of the vehicle automatic braking system based on artificial intelligence includes the following steps: Step S1: Data Collection: Assume that on an autonomous driving test vehicle, sensors such as high-precision radars, cameras, and lidars are installed. During the vehicle's driving process, these sensors collect data in real time. For example, when the vehicle is driving on urban roads, the sensors capture information such as the distance, speed of the vehicle ahead, and the wetness of the road surface. And these data are transmitted to the cloud platform in the form of real-time data streams through wireless communication technology, and the cloud platform stores them in a dedicated database for subsequent processing.

[0041] Step S2: Data Processing: After receiving the data, the cloud platform first performs data alignment. For time alignment, the cloud platform uses a time synchronization algorithm based on the timestamps of each sensor's data to ensure that all data is consistent in time. For spatial alignment, using the position information and calibration parameters of the sensors, the data from different sensors are matched in space to make the data accurately reflect the actual scenario.

[0042] Step S3: AEB Parameter Adjustment: The preprocessed data enters the AEB parameter adjustment module, which automatically selects a suitable algorithm using artificial intelligence and calculates and adjusts the various parameters of the AEB system according to the processed data.

[0043] For dynamically changing scenarios, taking the machine learning algorithm as an example for AEB parameter adjustment, specifically including: Step S3.2.1: Algorithm Selection and Model Training: Use the cloud platform to build a model that combines a convolutional neural network (CNN) and a long short-term memory network (LSTM). The model learns from a large amount of historical data and real-time data to establish the mapping relationship between the vehicle state, driving environment, and AEB parameters.

[0044] Step S3.2.2: Parameter Adjustment Rule Generation: The trained model automatically generates parameter adjustment rules according to the input data. For different driving scenarios and vehicle states, the model outputs corresponding AEB system parameter adjustment schemes. When the vehicle approaches an obstacle that is decelerating ahead, the model automatically calculates the optimal braking parameters, such as braking force and braking timing, based on data such as the current vehicle speed, obstacle distance, and relative speed, and sends these parameters to the vehicle's AEB system.

[0045] Step S3.2.3: Automatic Parameter Adjustment: The cloud platform sends parameter adjustment instructions to the vehicle's AEB system according to the generated parameter adjustment rules. After receiving the instructions, the AEB system adjusts its own parameter settings in real time.

[0046] Step S4: Multi-scenario simulation testing, feedback, and optimization: The AEB system performs braking operations according to the adjusted parameters, and the braking effect data of the vehicle is uploaded to the cloud platform again. The cloud platform uses the multi-scenario simulation testing module to simulate different driving scenarios and evaluate the performance of the adjusted AEB system. For example, it simulates emergency braking situations at different speeds and on different road conditions to verify whether indicators such as the braking distance and braking stability of the vehicle meet the expectations.

[0047] If the results of the simulation testing meet the expectations, the developers will solidify these parameters and apply them to the actual AEB system. If the results do not meet the expectations, the cloud platform will further optimize the parameters of the machine learning model or adjust the algorithm based on the feedback data, recalculate the parameters of the AEB system, and conduct multi-scenario simulation testing again until satisfactory results are obtained. Through such a closed-loop optimization process, the performance and safety of the AEB system are continuously improved. Experiment and verification:

[0048] To verify the effectiveness of the present invention, a series of experiments were conducted. The experimental vehicle was equipped with a complete AEB system and the parameter adjustment system of the present invention, and was tested in a variety of simulated driving scenarios, including different vehicle speeds, obstacle types, road surface conditions, and weather conditions.

[0049] I. Set scenarios and parameters for experiments: 1). Set the following three scenarios respectively: Scenario 1: On a dry road surface, the vehicle is traveling at a speed of 50 km / h, and a stationary simulated vehicle is placed in front.

[0050] Scenario 2: On a slippery road surface, the vehicle is traveling at a speed of 60 km / h, and the simulated vehicle in front is slowly traveling at a speed of 20 km / h.

[0051] Scenario 3: In a rainy environment, the vehicle is traveling at a speed of 70 km / h, and a pedestrian suddenly crosses the road in front.

[0052] 2). Experimental parameter settings: Vehicle parameters: The vehicle mass is 1500 kg, the friction coefficient between the tire and the dry road surface is set to 0.8, and the friction coefficient between the tire and the slippery road surface is set to 0.4.

[0053] Initial parameters of the AEB system: The braking start time delay is set to 0.5 s, and the maximum braking force is set to 8000 N.

[0054] II. Collection of experimental data: Under each experimental scenario, the running data of the vehicle, including speed, acceleration, distance to the obstacle, relative speed, etc., are collected in real time by in-vehicle sensors, and at the same time, the actual braking parameters of the AEB system, such as braking start time, braking force, braking deceleration, etc., are recorded.

[0055] III. Analysis of Experimental Results: 1) Results of Experiment for Scenario 1: Traditional AEB system: The braking start time is 0.5 s, the braking force reaches 8000 N, the vehicle stops at a distance of 3.5 m from the obstacle, and the braking distance is 12 m.

[0056] The AEB system adjusted by the present invention: According to the collected data, the deep learning model evaluates the collision risk as high risk. The parameter adjustment strategy advances the braking start time to 0.3 s, and the braking force is adjusted to 9000 N according to the vehicle dynamics model and the optimization algorithm. The vehicle stops at a distance of 1.5 m from the obstacle, and the braking distance is shortened to 8 m, effectively avoiding the collision, and the braking process is stable without obvious tire locking phenomenon.

[0057] 2) Results of Experiment for Scenario 2: Traditional AEB system: Due to the slippery road surface, the braking force cannot be effectively transmitted. The braking start time is 0.5 s, the braking force is 8000 N, but the braking deceleration of the vehicle is only 3 m / s². The vehicle finally collides with the simulated vehicle in front, and the collision speed is 15 km / h.

[0058] The AEB system adjusted by the present invention: The model identifies the slippery road surface and the slow-moving vehicle in front, evaluates the collision risk as medium risk, advances the braking start time to 0.4 s, adjusts the braking force to 6000 N according to the road surface friction coefficient and the vehicle dynamics model, and adopts a segmented braking strategy. First, the vehicle is smoothly decelerated with a lower braking force, and then the braking force is gradually increased. Finally, the vehicle stops at a distance of 0.5 m from the simulated vehicle in front, successfully avoiding the collision, and the vehicle maintains good stability during the braking process.

[0059] 3) Results of Experiment for Scenario 3: Traditional AEB system: In rainy weather, there is a certain delay in the sensor's recognition of pedestrians. The braking start time is 0.6 s, the braking force is 8000 N, and the vehicle stops at a distance of 2 m from the pedestrian. Although a direct collision is avoided, the vehicle shows a slight side-slip phenomenon during the braking process.

[0060] The adjusted AEB system of the present invention: The deep learning model combines camera and radar data to quickly and accurately identify pedestrians, assess the collision risk as high risk, immediately activate braking, advance the braking start time to 0.35 s, adjust the braking force to 10,000 N according to weather conditions and vehicle speed, and maintain vehicle stability during braking by adjusting the distribution of braking force. Finally, the vehicle stops at a distance of 0.5 m from the pedestrian, successfully avoiding the collision, and the driving trajectory of the vehicle remains stable without skidding.

[0061] From the comparison of the above experimental results, it can be seen that the method and system for adjusting the parameters of the AEB system based on artificial intelligence of the present invention can effectively improve the performance of the AEB system in different driving scenarios, significantly shorten the braking distance, avoid or reduce the occurrence of collision accidents, and the braking process is smoother, ensuring the safety of the vehicle and the occupants.

[0062] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A method for adjusting parameters of a vehicle automatic braking system based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Data collection: using various sensors installed on the vehicle to collect real-time data of the AEB system during vehicle driving, and transmitting the collected data to the cloud platform; the real-time data includes vehicle status information, front obstacle information and environmental information; Step S2: Data processing: The cloud platform first performs time series and spatial alignment processing on the collected data, then performs data cleaning, and then extracts features related to AEB system parameters from the cleaned data; Step S3: AEB parameter adjustment: using artificial intelligence to automatically select a suitable algorithm, calculate and adjust various parameters of the AEB system based on the processed data; Step S4: Multi-scenario simulation test and feedback, optimization: After completing the AEB parameter adjustment, conduct a multi-scenario simulation test in a virtual environment to evaluate the performance of the AEB system in different driving scenarios. If the evaluation results do not meet expectations, continue to optimize the parameters based on the feedback.

2. The method for adjusting parameters of a vehicle automatic braking system based on artificial intelligence according to claim 1, characterized in that: The step S1 data collection specifically includes: Step S1.1: Assemble various types of sensors on the test vehicle to form a comprehensive data collection network; the sensors include but are not limited to millimeter wave radar, camera, wheel speed sensor, acceleration sensor, gyroscope, humidity sensor, temperature sensor and light sensor, Step S1.2: Setting different types of sensors to different acquisition frequencies; Step S1.3: The data collected by the sensor is transmitted to the on-board data processing unit through the controller area network CAN bus in the vehicle. In the on-board data processing unit, the data is initially packaged and format converted, and then the data is transmitted to the cloud platform in real time through the 4G / 5G communication module.

3. The method for adjusting parameters of a vehicle automatic braking system based on artificial intelligence according to claim 2, characterized in that: In step S3, artificial intelligence is used to automatically select a suitable algorithm, which specifically includes: Step S3.1: construct an algorithm candidate pool, and determine the algorithms applicable to each scenario according to different driving scenarios and AEB system requirements, including but not limited to PID control algorithms, fuzzy control algorithms, and machine learning algorithms; organize these applicable algorithms into an algorithm candidate pool, and annotate the applicable scenario range and performance indicators for each algorithm; Step S3.2: Evaluate the algorithm using four indicators: accuracy, response speed, stability, and adaptability, and select the appropriate algorithm based on the algorithm selection strategy; the algorithm selection strategy specifically includes: Rule-based selection: Select the appropriate algorithm from the algorithm candidate pool based on the results of scene recognition and preset rules; Selection based on performance evaluation: Perform real-time performance evaluation on each candidate algorithm and select the algorithm with the best performance based on the evaluation indicators; Hybrid strategy: combines rule-based selection and performance evaluation-based selection; first, select a part of applicable algorithms according to scenario rules, then perform performance evaluation on these algorithms, and finally select the optimal algorithm.

4. The method for adjusting parameters of a vehicle automatic braking system based on artificial intelligence according to claim 3 is characterized in that: In step S3.2, for dynamically changing scenarios, a machine learning algorithm is selected to adjust AEB parameters, specifically including: Step S3.2.1: Algorithm selection and model training: Use the cloud platform to build a model that combines a convolutional neural network (CNN) and a long short-term memory (LSTM) network, and train the model so that the model learns the relationship between data features and AEB parameters; Step S3.2.2: Parameter adjustment rule generation: The trained model automatically generates parameter adjustment rules based on the input data. For different driving scenarios and vehicle states, the model outputs corresponding AEB system parameter adjustment solutions. When the parameters generated by the model cause the vehicle to become unstable or fail to effectively avoid collisions in the simulation test, the model output is penalized and the model weight is adjusted. Step S3.2.3: Automatic parameter adjustment: The cloud platform sends parameter adjustment instructions to the vehicle's AEB system based on the generated parameter adjustment rules. After receiving the instructions, the AEB system adjusts its own parameter settings in real time.

5. The method for adjusting parameters of a vehicle automatic braking system based on artificial intelligence according to claim 4 is characterized in that: In step S3.2.2, when the parameters generated by the model cause the vehicle to become unstable or fail to effectively avoid collision in the simulation test, the output of the model is penalized, specifically including: Step S3.2.2.1: Penalty based on performance indicators: Define the indicators used to evaluate the performance of the AEB system, including but not limited to braking distance, collision speed, and vehicle stability, and set a range or threshold for each performance indicator; Step S3.2.2.2: Penalty based on scenario adaptability: First, classify the driving scenarios, classify different driving scenarios, and set specific parameter adjustment targets and restrictions according to the different characteristics and safety requirements of each scenario; when the parameter settings of the AEB system cannot meet the requirements of a specific scenario, these parameters will be penalized, and the degree of penalty is determined according to the importance of the scenario and the degree to which the parameter settings deviate from the requirements.

6. The method for adjusting parameters of a vehicle automatic braking system based on artificial intelligence according to claim 5, characterized in that: The step S4 specifically includes: Step S4.1: Multi-scenario simulation test: The adjusted parameters are verified in a multi-scenario simulation test environment, which simulates various real driving scenarios, including different weather conditions, road conditions, obstacle types, and different driving states of the vehicle. In each simulation scenario, the simulated vehicle drives according to the adjusted AEB parameters, and monitors three indicators: the vehicle's motion state, the change in distance to the obstacle, and whether a collision occurs. Step S4.2: Result feedback and evaluation: The results of the simulation test are fed back to the developer and the optimization module of the cloud platform. The developer evaluates the results and checks whether the braking effect of the AEB system in different scenarios meets expectations. If the test results meet expectations, that is, the various evaluation indicators meet the preset standards, the system will solidify these parameters and apply them to the actual AEB system; if the results do not meet expectations, the optimization module will analyze the cause of the problem based on the feedback data, optimize the algorithm, and perform parameter adjustment and simulation testing again until the test results meet expectations.

7. An AEB parameter adjustment system based on artificial intelligence, characterized in that: include: Data collection module: used to collect real-time data of the AEB system using various sensors installed on the vehicle and transmit the data to the cloud platform; Data processing and adjustment module: It is set up on the cloud platform and is used to perform time and space alignment processing on the collected data, automatically select the appropriate algorithm using artificial intelligence, and calculate and adjust various parameters of the AEB system based on the processed data; Multi-scenario simulation test module: This module is set up on the cloud platform and is used to verify the performance of the AEB system in different driving scenarios through multi-scenario tests in a virtual environment after the AEB parameters are adjusted. Feedback and optimization module: used to feed back the parameters adjusted by the AEB parameter adjustment module to the developer for evaluation. If the evaluation results do not meet expectations, the parameters will continue to be optimized based on the feedback.

Images