Search

CN-121989975-A - Vehicle control method based on multi-mode information

CN121989975ACN 121989975 ACN121989975 ACN 121989975ACN-121989975-A

Abstract

The invention relates to the technical field of vehicle control and discloses a vehicle control method based on multi-mode information, which comprises the steps of respectively constructing multi-mode characteristics with different characteristics based on different characteristic parameters of the multi-mode information; the multi-mode feature is input into a first model, a driving scene, a driving intention, a driving risk and a driving preference are obtained through output, a power output range is determined according to the driving scene, the driving intention and the driving risk, and a torque distribution result is obtained through mapping in the power output range according to the driving preference. The invention improves the characteristics of the multi-mode characteristics, and simultaneously gives consideration to the model quantization capability, thereby improving the accuracy of subsequent torque distribution.

Inventors

  • Zou Zhuolun
  • CHEN BO
  • JIN TINGXIN
  • GU SHAOWEI

Assignees

  • 重庆长安汽车股份有限公司

Dates

Publication Date
20260508
Application Date
20260407

Claims (14)

  1. 1. A method of vehicle control based on multimodal information, the method comprising: respectively constructing multi-mode features with different characteristics based on different characteristic parameters of the multi-mode information; inputting the multi-mode features into a first model, and outputting to obtain driving scenes, driving intentions, driving risks and driving preferences; And determining a power output range according to the driving scene, the driving intention and the driving risk, and mapping a torque distribution result in the power output range according to the driving preference.
  2. 2. The method according to claim 1, wherein the constructing the multi-modal feature of different characteristics based on different characteristic parameters of the multi-modal information includes: Acquiring multi-mode information, wherein the multi-mode information comprises driver operation information, vehicle body state information, environmental road condition information, other vehicle information, in-vehicle NVH information and sensor health monitoring information; Respectively constructing a plurality of short-time features with different characteristics based on the driver operation information, the vehicle body state information, the environmental road condition information, the other vehicle information, the in-vehicle NVH information and the sensor health monitoring information; and carrying out time alignment and normalization processing on each short-time feature, and sequentially intercepting the short-time sequence features from the aligned data through a fixed time window.
  3. 3. The method of claim 2, wherein the constructing short-term features of a plurality of different characteristics based on the driver operation information, the body state information, the environmental road condition information, the other vehicle information, the in-vehicle NVH information, and the sensor health monitoring information, respectively, comprises: constructing at least one of an accelerator opening feature, a brake pedal feature, a steering wheel operation feature, a gear mode operation feature, and a driver tolerance feature based on the driver operation information; Constructing a vehicle body dynamic characteristic and a vehicle body dynamic energy characteristic based on the vehicle body state information; constructing road condition features based on the environmental road condition information; Constructing traffic interaction characteristics of other vehicles and the vehicle based on the other vehicle information; Constructing occupant comfort status features based on the in-vehicle NVH information; sensor health features are constructed based on the sensor health monitoring information.
  4. 4. A method according to claim 3, wherein said constructing a gear mode operating feature based on said driver operating information comprises: Constructing a switching operation event sequence according to the gear shifting operation of a driver in a preset time period, the driving mode switching operation and the time point of each switching operation; Calculating the driving mode switching times in a unit mileage, the motion mode accumulated duration in the preset time period, the number of gear events in the preset time period and the gear event proportion in the preset time period based on the switching operation event sequence; And combining the driving mode switching times in the unit mileage, the motion mode accumulated duration in the preset time period, the number of gear events in the preset time period and the gear event proportion in the preset time period into a sixth mode vector to obtain the gear mode operation characteristic.
  5. 5. A method according to claim 3, wherein said constructing a driver tolerance level feature based on said driver operation information comprises: Acquiring actual longitudinal acceleration of a plurality of sampling points of the vehicle in a preset time period; Calculating the corresponding longitudinal acceleration change rate after each sampling point according to the actual longitudinal acceleration of the adjacent sampling points; Collecting discomfort events including subjective feedback of an occupant and abnormal gestures detected by a seat sensor; time alignment is carried out on the longitudinal acceleration change rate and the uncomfortable event, and a target longitudinal acceleration change rate set corresponding to the uncomfortable event is screened out; And counting the upper and lower limit intervals of the longitudinal acceleration change rate by using the target longitudinal acceleration change rate set to obtain the tolerance degree characteristic of the driver.
  6. 6. The method of claim 3, wherein the constructing a body dynamics feature and a body dynamics energy feature based on the body state information comprises: Collecting longitudinal speed, longitudinal acceleration change rate, yaw rate, lateral acceleration and wheel speed of the vehicle as the vehicle body state information in a preset time period; Processing the output torque of the driving machine, the vehicle mass and the longitudinal acceleration by adopting a simplified longitudinal slope estimation model to obtain a corresponding road slope; obtaining tire attachment utilization rate and stability margin indexes through the road gradient and the vehicle body state information mapping; Constructing a seventh mode vector corresponding to the vehicle body dynamic feature, wherein the seventh mode vector comprises the longitudinal speed, the longitudinal acceleration change rate, the yaw rate, the lateral acceleration, the wheel speed, the tire attachment utilization rate and the stability margin index; constructing a driving machine rotating speed, a driving machine output torque, driving machine efficiency, transmission system efficiency, a battery charge state, a battery health state, a battery temperature, a maximum discharge power and instantaneous equivalent energy consumption cost in a preset time period into an eighth mode vector for representing the power energy characteristics of the vehicle body; the instantaneous equivalent energy consumption cost is calculated by the following method: and carrying out weighted summation on the fuel flow, the battery discharge power and a thermal penalty term in the preset time period to obtain the instantaneous equivalent energy consumption cost, wherein the thermal penalty term is obtained through calculation of the output torque of the driving machine and the battery temperature.
  7. 7. A method according to claim 3, wherein said constructing road condition features based on said environmental road condition information comprises: acquiring road related information in a preset distance in front of a current position of a vehicle, wherein the road related information comprises along-road coordinates, road speed limit, road gradient, road curvature and road type; Constructing a forward-looking road section sequence comprising a plurality of road section information with preset time period length based on the road related information; Extracting target road section information of a plurality of road sections closest to the vehicle from the forward-looking road section sequence to serve as basic road characteristics; predicting a forward looking torque demand within a future preset time period using the base road characteristics, the current vehicle speed and the driver intent; The base road characteristic and the forward torque demand are combined into a ninth mode vector as the road condition characteristic.
  8. 8. The method of claim 3, wherein the constructing traffic interaction features of other vehicles with the host vehicle based on the other vehicle information comprises: Acquiring external vehicle information in a preset time period, wherein the external vehicle information comprises relative motion parameters of the vehicle and an external vehicle and external vehicle behavior intention; calculating high-rise characteristics for evaluating the collision risk degree of the vehicle and the external vehicle based on the external vehicle information; And combining the external vehicle information and the high-level features into a tenth mode vector as the traffic interaction feature.
  9. 9. The method of claim 3, wherein the constructing occupant comfort state features based on the in-vehicle NVH information comprises: Collecting NVH original data and uncomfortable events in a vehicle within a preset time period, wherein the uncomfortable events comprise subjective feedback of passengers and abnormal postures detected by seat sensors; An eleventh mode vector is constructed as the occupant comfort state feature based on the in-vehicle NVH raw data and the discomfort event.
  10. 10. The method of claim 3, wherein constructing the multi-modal feature based on the characteristic parameters of the multi-modal information further comprises: Extracting driver operating parameters from the accelerator opening feature, the brake pedal feature, the steering wheel operating feature and the gear mode operating feature respectively; And integrating and obtaining the driver style portrait features based on the driver operation parameters.
  11. 11. The method of claim 10, wherein determining a power output range from the driving scenario, the driving intent, and the driving risk, and mapping torque distribution results within the power output range from the driving preference, comprises: Calculating a first-stage torque range based on the driving scenario, the driving risk, the vehicle body dynamic feature and the vehicle body dynamic energy feature; According to the driving intention and the driver style portrait characteristic, the first-stage torque range is contracted to obtain a second-stage torque range; And constructing a multi-objective optimization function according to the weight corresponding to the driving preference, and solving an optimal solution of the multi-objective optimization function in the second-stage torque range to obtain the torque distribution result.
  12. 12. The method of claim 11, wherein the calculating a first level torque range based on the driving scenario, the driving risk, the body dynamics feature, and the body dynamics energy feature comprises: calculating allowable torque of a power system according to the power energy characteristics of the vehicle body; calculating the tire attachment limit torque according to the dynamic characteristics of the vehicle body; determining an upper limit of the first-stage torque range based on a minimum of the powertrain allowable torque and the tire adhesion limit torque; Determining a lower limit of the first-stage torque range based on a maximum value of a negative value of a limit braking torque of the motor and the tire adhesion limit torque; and correcting the upper limit and the lower limit of the first-stage torque range according to the complexity of the driving scene and the height of the driving risk.
  13. 13. The method of claim 11, wherein the contracting the first-stage torque range to obtain a second-stage torque range based on the driving intent and the driver style representation feature comprises: Inquiring an allowable floating bandwidth upper limit and an allowable floating bandwidth lower limit of a torque range according to the driving intention; calculating a driving coefficient representing the driving degree through the normalization of the driver style image feature vector; Respectively superposing the upper limit of the allowed floating bandwidth and the lower limit of the allowed floating bandwidth which are processed by the aggressive coefficient based on the forward torque demand to obtain a soft constraint contraction range; the second stage torque range is determined from an intersection of the soft constraint shrink range and the first stage torque range.
  14. 14. The method of claim 11, wherein the objective optimization function comprises: Wherein, the The target optimization function is represented as a function of the target optimization, Representing the solution of the objective optimization function, As a function of the performance of the object, As a function of the energy consumption of the power plant, As a function of the comfort level of the objective, As a function of the stability of the object, Corresponding to the performance weight, Corresponding to energy consumption weight, Corresponding to the comfort weight, Corresponding to the stability weight, through the target weight vector The target weight vector is obtained and used for representing driving preference output by the first model.

Description

Vehicle control method based on multi-mode information Technical Field The invention relates to the technical field of vehicle control, in particular to a vehicle control method based on multi-mode information. Background Along with the rapid development of new energy automobiles and intelligent network-connected automobiles, a vehicle power system is driven by a traditional single engine, and evolves into a complex structure of multiple motors, multi-shaft driving and motor-engine cooperation, so that the whole vehicle has higher-degree-of-freedom torque distribution capability. Meanwhile, technologies such as high-precision maps, environment perception, V2X communication and in-vehicle human factor perception are increasingly mature, and vehicles can acquire rich external environment and driver state information in real time. However, although some technologies have begun to comprehensively consider multi-modal information in terms of longitudinal control and torque distribution strategies, the related technologies are almost directly used for feature extraction of multi-modal information no matter what form of data the multi-modal information is, some text data are all used for coding feature extraction by adopting a bert model, and the characteristics of different modal data are often difficult to consider by adopting a one-cut feature extraction mode. In addition, in the related art, input multi-mode data is directly analyzed through a large model, then a torque distribution result is directly output, the quality of the model is difficult to quantify due to over-abstract consideration of influence of multi-mode information in different aspects, and further negative output of torque distribution is difficult to adjust. Disclosure of Invention The invention provides a vehicle control method and device based on multi-mode information and a vehicle, and aims to solve the problem that conventional vehicle control torque distribution based on multi-mode information is not accurate enough. According to the first aspect, the vehicle control method based on the multi-modal information comprises the steps of respectively constructing multi-modal features of different characteristics based on different characteristic parameters of the multi-modal information, inputting the multi-modal features into a first model, outputting to obtain driving scenes, driving intentions, driving risks and driving preferences, determining a power output range according to the driving scenes, the driving intentions and the driving risks, and mapping in the power output range according to the driving preferences to obtain a torque distribution result. According to the technical means, the method integrates the multi-source information and synchronously outputs four key information of driving scene, intention, risk and preference. Compared with a control strategy of single parameter response of directly outputting a torque result by a traditional model, the method breaks through the limitation of sensing dimension and avoids the problem of multi-objective weighing stiffness. And the torque distribution result is dynamically matched through multidimensional information, the quality of the torque distribution can be directly measured through four key information of driving scene, intention, risk and preference, the problem is convenient to trace, the negative output of the torque distribution is easy to adjust, the synergistic compromise of safety, dynamic property, economy and comfort is realized, the intellectualization and suitability of vehicle control are obviously improved, and the method is suitable for individual requirements under complex driving conditions. In some alternative embodiments, the multi-modal characteristics of different characteristics are respectively constructed based on different characteristic parameters of the multi-modal information, wherein the multi-modal information comprises driver operation information, vehicle body state information, environment road condition information, other vehicle information, in-vehicle NVH information and sensor health monitoring information, the short-time characteristics of different characteristics are respectively constructed based on the driver operation information, the vehicle body state information, the environment road condition information, the other vehicle information, the in-vehicle NVH information and the sensor health monitoring information, the short-time characteristics are aligned and normalized according to time, and short-time sequence characteristics are sequentially obtained by intercepting aligned data through a fixed time window. According to the above technical means, the acquisition range of the multi-mode information is comprehensively covered from the driver operation information, the vehicle body state information, the environmental road condition information, other vehicle information, the in-vehicle NVH information and the sensor health monitoring