CN-122009212-A - Vehicle, control method and device thereof, and storage medium

CN122009212ACN 122009212 ACN122009212 ACN 122009212ACN-122009212-A

Abstract

The embodiment of the application provides a vehicle, a control method, a device and a storage medium thereof, wherein driving scene data are acquired, feature extraction is carried out on the driving scene data to determine driving scene features, comprehensive exposure is calculated based on the driving scene features and weight coefficients corresponding to the driving scene features, the comprehensive exposure and the driving scene features are input into a preset vehicle control parameter prediction model to output vehicle control parameters, and the vehicle is controlled based on the vehicle control parameters. According to the method, the driving scene data are obtained, the multidimensional features are extracted, the comprehensive exposure capable of quantifying the dynamic composite risk is obtained through calculation by combining with the preset weight, and then the exposure and the scene features are input into the control parameter prediction model together, and the optimal control instruction is generated on line and executed. Therefore, the problems of response lag, performance attenuation and safety risk improvement caused by control parameter solidification of the vehicle in a complex and changeable composite scene are effectively solved.

Inventors

LIANG XINFENG
LIU GUANGHAO
YANG GUOKE
HAN SHOUNING

Assignees

广州汽车集团股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260211

Claims (10)

1. A control method of a vehicle, characterized by comprising: Acquiring driving scene data, and carrying out feature extraction on the driving scene data to determine driving scene features; Based on the driving scene characteristics and weight coefficients corresponding to the driving scene characteristics, calculating to obtain comprehensive exposure; And inputting the comprehensive exposure and the driving scene characteristics into a preset vehicle control parameter prediction model to output vehicle control parameters, and controlling the vehicle based on the vehicle control parameters.
2. The method of controlling a vehicle according to claim 1, wherein the driving scene features include a travel data feature, an environment data feature, a traffic data feature, and an occupant data feature, wherein the feature extracting the driving scene data to determine the driving scene features includes: Preprocessing the driving scene data to obtain target driving scene data, wherein the target driving scene data comprises driving data, environment data, traffic data and passenger data; Extracting dynamic characteristics and/or driving style characteristics of the driving data to obtain the driving data characteristics; extracting road characteristics and/or weather influence characteristics from the environmental data to obtain the environmental data characteristics; extracting traffic density characteristics and/or congestion scene characteristics from the traffic data to obtain the traffic data characteristics; And extracting load influence characteristics from the passenger data to obtain the passenger data characteristics.
3. The method according to claim 2, wherein the calculating the comprehensive exposure based on the driving scene feature and the weight coefficient corresponding to the driving scene feature includes: calculating to obtain a driving exposure degree, an environment exposure degree, a traffic exposure degree and a passenger exposure degree based on the driving data characteristic, the environment data characteristic, the traffic data characteristic, the passenger data characteristic and a first weight coefficient respectively; Calculating a main exposure degree based on the driving exposure degree, the environment exposure degree, the traffic exposure degree, the passenger exposure degree and a second weight coefficient; determining interaction items among the driving data features, the environment data features, the traffic data features and the passenger data features, and calculating to obtain additional exposure degree based on the features corresponding to the interaction items and a third weight coefficient; The integrated exposure is calculated based on the primary exposure and the additional exposure.
4. A control method of a vehicle according to claim 3, characterized by further comprising: In response to the driving scene data being changed, determining relevant driving scene features corresponding to the changed driving scene data; Determining a first sensitivity coefficient of the first weight coefficient and/or a second sensitivity coefficient of a third weight coefficient according to the related driving scene characteristics; Updating the first weight coefficient and/or the third weight coefficient based on the first sensitivity coefficient and/or the second sensitivity coefficient.
5. The control method of a vehicle according to claim 3 or 4, characterized by further comprising: acquiring historical driving scene data and historical comprehensive exposure and real event occurrence probability corresponding to the historical driving scene data; determining the occurrence probability of the predicted event corresponding to the historical driving scene data according to the comprehensive exposure interval in which the historical comprehensive exposure is located; Calculating a loss value based on the predicted event occurrence probability and the actual event occurrence probability; Calculating a first gradient of the first weight coefficient and/or a second gradient corresponding to the third weight coefficient based on the loss value; Updating the first weight coefficient and/or the third weight coefficient based on the first gradient and/or the second gradient.
6. The control method of a vehicle according to claim 1, characterized by further comprising: Acquiring historical driving scene data and corresponding historical comprehensive exposure and historical vehicle control parameters; extracting features of the historical driving scene data to determine historical driving scene features; Constructing a model training set of the preset vehicle control parameter prediction model according to the historical driving scene characteristics, the historical comprehensive exposure corresponding to the historical driving scene characteristics and the historical vehicle control parameters; And training the preset vehicle control parameter prediction model by using the model training set.
7. The method of claim 6, wherein training the predictive model of the preset vehicle control parameters using the training set of models comprises: Inputting any input data in the model training set into an initial preset vehicle control parameter prediction model to output predicted vehicle control parameters, wherein the any input data is the historical driving scene characteristics in the model training set and the historical comprehensive exposure corresponding to the historical driving scene characteristics; calculating discrimination loss based on the predicted vehicle control parameters and the historical vehicle control parameters corresponding to the model training set to obtain a calculation result; and updating the model parameters of the initial preset vehicle control parameter prediction model by using the calculation result until the updated initial preset vehicle control parameter prediction model meets preset convergence conditions, so as to obtain the preset vehicle control parameter prediction model.
8. A control device for a vehicle, comprising: The acquisition module is used for acquiring driving scene data; The feature extraction module is used for carrying out feature extraction on the driving scene data to determine driving scene features; The first determining module is used for calculating and obtaining comprehensive exposure degree based on the driving scene characteristics and the weight coefficients corresponding to the driving scene characteristics; The second determining module is used for inputting the comprehensive exposure and the driving scene characteristics into a preset vehicle control parameter prediction model so as to output vehicle control parameters; And the control module is used for controlling the vehicle based on the vehicle control parameters.
9. A computer-readable storage medium, characterized in that a control program of a vehicle is stored thereon, which when executed by a processor implements the control method of a vehicle according to any one of claims 1-7.
10. A vehicle characterized by comprising a memory, a processor and a control program of the vehicle stored on the memory and capable of running on the processor, the processor implementing the control method of the vehicle according to any one of claims 1-7 when executing the control program of the vehicle.

Description

Vehicle, control method and device thereof, and storage medium Technical Field The embodiment of the application relates to the technical field of vehicles, in particular to a vehicle, a control method and device thereof and a storage medium. Background Vehicle electrical control systems, such as ABS (Anti-lock Braking System, anti-lock brake system) and ESP (Electronic Stability Program ), are central to the active safety of modern automobiles. These systems achieve adjustments to the vehicle state through pre-calibrated control parameters (e.g., trigger thresholds, pressure gradients, intervention timing, etc.), which are typically optimally calibrated under limited standardized test scenarios (e.g., linear braking of specific adhesion coefficients, steady state gyration, etc.), and cured in a controller. However, the actual driving environment is complex and changeable, and often presents dynamic and compound characteristics, such as a scene of interweaving a plurality of adverse factors such as 'rainy night + curve + congestion'. Under such non-standard, multi-dimensional composite scenes, static control parameters based on single or limited scene calibration tend to be difficult to achieve optimal control effects, and even may cause potential safety hazards. Disclosure of Invention The embodiment of the application provides a vehicle, a control method, a control device and a storage medium thereof, and aims to solve the technical problems that the control performance of the vehicle is reduced and the safety is insufficient in a dynamic interweaving scene because the vehicle cannot adapt to a complex and changeable composite driving scene in real time due to adoption of a pre-calibrated static control parameter. In order to achieve the above objective, an embodiment of the present application provides a vehicle control method, which obtains driving scene data, performs feature extraction on the driving scene data to determine driving scene features, calculates a comprehensive exposure based on the driving scene features and weight coefficients corresponding to the driving scene features, inputs the comprehensive exposure and the driving scene features into a preset vehicle control parameter prediction model to output vehicle control parameters, and controls a vehicle based on the vehicle control parameters. According to the method, the driving scene data are obtained, the multidimensional features are extracted, the comprehensive exposure capable of quantifying the dynamic composite risk is obtained through calculation by combining with the preset weight, and then the exposure and the scene features are input into the control parameter prediction model together, and the optimal control instruction is generated on line and executed. Therefore, the limitation of the traditional pre-calibrated static parameters is broken through by introducing the real-time risk assessment index of the comprehensive exposure degree, and the active matching of the control performance and the complex scene is realized, so that the safety and stability of the vehicle in the dynamic complex environment are improved, and the problems of response lag, performance attenuation and safety risk improvement caused by the solidification of the control parameters of the vehicle in the complex and variable complex scene are effectively solved. According to one embodiment of the application, the driving scene features comprise driving data features, environment data features, traffic data features and passenger data features, wherein the driving scene data is subjected to feature extraction to determine driving scene features, and the driving scene features comprise preprocessing driving scene data to obtain target driving scene data, wherein the target driving scene data comprises driving data, environment data, traffic data and passenger data, extracting dynamic features and/or driving style features of the driving data to obtain driving data features, extracting road features and/or weather influence features of the environment data to obtain environment data features, extracting traffic density features and/or congestion scene features of the traffic data to obtain traffic data features, and extracting load influence features of the passenger data to obtain passenger data features. According to one embodiment of the application, the comprehensive exposure is calculated based on driving scene characteristics and weight coefficients corresponding to the driving scene characteristics, and comprises the steps of calculating the driving exposure, the environment exposure, the traffic exposure and the passenger exposure based on driving data characteristics, environment data characteristics, traffic data characteristics, passenger data characteristics and first weight coefficients respectively, calculating the main exposure based on the driving exposure, the environment exposure, the traffic exposure, the passenger exposure and second weig