A User Style Classification Strategy for Intelligent Driving Method and System

By classifying the driving styles of surrounding vehicles and drivers and combining this with a deep learning module to form a customized driving model, the problem of homogeneous driving styles is solved, and a personalized autonomous driving experience is achieved.

CN117163039BActive Publication Date: 2026-11-17NINGBO LOTUS ROBOTICS CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202210593480.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-27
Publication Date
2026-11-17
Estimated Expiration
2042-05-27

AI Technical Summary

Technical Problem

Existing autonomous driving methods struggle to meet the personalized driving needs of different users in the same scenario, resulting in severe homogenization of driving styles and failing to provide users with a personalized experience.

Method used

By automatically classifying the driving styles of surrounding vehicles and drivers, merging the results from both, and combining them with a deep learning module to form a customized driving model, personalized driving strategies can be achieved.

Benefits of technology

Under the premise of safety and comfort, the vehicle is more in line with the user's psychological expectations during autonomous driving, providing a unique and personalized driving experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117163039B_ABST
    Figure CN117163039B_ABST
Patent Text Reader

Abstract

This invention provides a user style classification strategy-based intelligent driving method and system. The method includes the following steps: automatically classifying the driving styles of surrounding vehicles to obtain a first classification result; receiving the driver's classification of the driving styles of surrounding vehicles to obtain a second classification result; merging the first and second classification results, and the autonomous driving system executing the corresponding driving strategy. In this invention, the driver's driving style is learned by acquiring the driver's classification of the driving styles of surrounding vehicles, thereby achieving a customized intelligent driving mode. During autonomous driving, the vehicle, under the premise of safety and comfort, adopts a driving strategy that better matches the user's psychological expectations, providing the user with a unique and personalized experience.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of autonomous driving, and in particular to an intelligent driving method and system based on a user style classification strategy. Background Technology

[0002] Current autonomous driving methods primarily prioritize safety, comfort, and efficiency, failing to personalize the system to meet diverse user needs. While autonomous driving technology enhances vehicle intelligence, the driving styles of these systems often suffer from severe homogenization, hindering the provision of personalized experiences. Although current research explores different autonomous driving technologies for various cities and road conditions, achieving some degree of diversification, it cannot satisfy the personalized needs of different users or even the same user seeking different autonomous vehicle operations within the same scenario. Summary of the Invention

[0003] One objective of this invention is to provide an intelligent driving method and system that can implement user style classification strategies for different autonomous vehicle operations in the same scenario.

[0004] Specifically, this invention provides a user style classification strategy-based intelligent driving method, comprising the following steps:

[0005] The driving styles of surrounding vehicles are automatically classified, and the first classification result is obtained.

[0006] Receive the driver's classification of the driving style of surrounding vehicles and obtain the second classification result;

[0007] The first and second classification results are merged, and the autonomous driving system executes the corresponding driving strategy.

[0008] Furthermore, after receiving the driver's classification of the driving styles of surrounding vehicles, the credibility of the driver's modified behavior is judged. If the credibility judgment is passed, a second classification result is obtained.

[0009] Furthermore, when judging the credibility of the driver's modification behavior, a confidence level is given. If the confidence level is greater than or equal to a set threshold, the driver's modification behavior is approved; if the confidence level is less than the set threshold, the driver's modification behavior is rejected.

[0010] Furthermore, when the driver categorizes the driving styles of surrounding vehicles, the driver categorizes the surrounding vehicles a set number of times.

[0011] Furthermore, the set number of times is once.

[0012] Furthermore, it also includes: assigning identity information to each driver, and after identifying the identity information, sending both the first classification result and the second classification result corresponding to the identity information into the deep learning module to form a driving model.

[0013] Furthermore, after identifying the identity information, the driving parameters corresponding to the identity information are also sent to the deep learning module.

[0014] Furthermore, the driving model is readable.

[0015] In particular, the present invention also discloses a classification strategy intelligent driving system, including a memory and a processor, wherein the memory stores a program for executing the classification strategy intelligent driving method.

[0016] In this invention, the driver's driving style is learned by acquiring the driver's classification of the driving styles of surrounding vehicles, thereby realizing a customized intelligent driving mode. During the autonomous driving process, the vehicle adopts a driving strategy that is more in line with the user's psychological expectations under the premise of safety and comfort, bringing the user a unique personalized experience.

[0017] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0018] The following sections will describe some specific embodiments of the invention in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0019] Figure 1 This is a flowchart of a user style classification strategy intelligent driving method according to an embodiment of the present invention. Detailed Implementation

[0020] Figure 1 This is a flowchart of a user style classification strategy intelligent driving method according to an embodiment of the present invention.

[0021] Specifically, this invention discloses a user style classification strategy-based intelligent driving method, comprising the following steps:

[0022] S1. Automatically classify the driving styles of surrounding vehicles to obtain the first classification result.

[0023] S2. Receive the driver's classification of the driving styles of surrounding vehicles and obtain the second classification result.

[0024] S3. Merge the results of the first classification and the second classification, and the autonomous driving system executes the corresponding driving strategy.

[0025] In this invention, the driver's driving style is learned by acquiring the driver's classification of the driving styles of surrounding vehicles, thereby realizing a customized intelligent driving mode. During the autonomous driving process, the vehicle adopts a driving strategy that is more in line with the user's psychological expectations under the premise of safety and comfort, bringing the user a unique personalized experience.

[0026] According to an embodiment of the present invention, in step S1, the driving styles of surrounding vehicles are automatically classified. The classification method used can be a supervised BP neural network or other methods mentioned in the prior art, or an unsupervised K-means clustering or other methods. One type of unsupervised K-means clustering specifically includes the following steps:

[0027] First, on the offline side, the following steps are included:

[0028] S101. Collect trajectory data and establish a trajectory dataset. It can be understood that trajectory data refers to the trajectory data of a vehicle during its driving process.

[0029] S102. Perform kinematic analysis on the trajectory dataset collected in step S101. It can be understood that in this step, kinematic analysis refers to studying the vehicle's motion from a geometric perspective, including the changes in the vehicle's spatial pose and velocity over time. A model is then built based on this data, and the kinematic analysis results are output from the model.

[0030] S103. Perform EKF filtering on the kinematic analysis results from step S102. It is understood that EKF stands for Extended Kalman Filter, a highly efficient recursive filter (autoregressive filter).

[0031] S104. Perform K-means clustering on the filtering results in step S103. K-means clustering is an iterative clustering analysis algorithm, which will not be described in detail in this embodiment.

[0032] Secondly, on the online side, the following steps are included:

[0033] S111. Determine the distribution of kinematic parameters for different styles.

[0034] S112, Obtain the historical trajectory of the subject (Agent).

[0035] S113. Based on the distribution of different style kinematic parameters in step S111 and the main historical trajectory in S112, the style similarity of the main historical trajectory in S112 is compared with the distribution of different style kinematic parameters in step S111.

[0036] S114. Determine the style of the main historical trajectory in S112 based on the comparison results in S113, and output the style.

[0037] Understandably, the concepts of "offline" and "online" are common in modeling applications. In the field of model learning, models are used to generate features, which may require feature consumption support. Feature consumption support can be divided into two main categories: offline and online. Offline support refers to feature extraction support during the training phase, while online support refers to the support needed to acquire features during model deployment (model serving).

[0038] Furthermore, since the driving style classification involves other vehicles, the implementation process typically does not directly collect driving data from other vehicles. Instead, it utilizes the vehicle's own sensors, such as images and radar, to acquire information about other vehicles. After the system processes the data, it generates the information needed for driving style classification, such as trajectories or vehicle bounding boxes. This information is then further processed (e.g., kinematic analysis and EKF filtering), combined with machine statistical learning methods such as K-means clustering, to ultimately achieve the classification of driving styles for other vehicles.

[0039] According to an embodiment of the present invention, the specific implementation process of this embodiment is as follows:

[0040] S101. The vehicle's display terminal shows the classification of surrounding vehicles' driving styles by the autonomous driving system. Vehicles with overly conservative styles are displayed in green, overly aggressive styles in red, and normal styles in blue. It is understood that the conservative and aggressive styles in this embodiment are relative classifications, determined based on characteristic parameters of the driver's driving process. Aggressive driving styles generally have higher values ​​for characteristic parameters such as average speed, maximum speed, and maximum throttle opening; conservative driving styles generally have lower values ​​for characteristic parameters. Currently, domestic research on driving style discrimination often relies on a few characteristic parameters to determine the driver's driving style, using methods such as BP neural networks as the identification method. This is existing technology and will not be elaborated upon here. It is understood that in this step, the vehicle's automatic classification result is the first classification result.

[0041] S102. The autonomous driving system adopts different driving strategies for vehicles with different driving styles. Specifically, it adopts a strategy of avoiding and maintaining distance from aggressive vehicles, and a strategy of prioritizing lane changing or overtaking for conservative vehicles.

[0042] S103. During vehicle operation, the driver can modify their classification of the driving styles of surrounding vehicles. This modification is limited to once per trip. In this step, after receiving the driver's classification of the driving styles of surrounding vehicles, the system assesses the credibility of the modification. If the credibility assessment passes, a second classification result is obtained; the driver's classification of the driving styles of surrounding vehicles becomes the second classification result. It is understandable that after obtaining the first and second classification results, the driving style system merges them. During this merging process, the autonomous driving system's attention behavior will better align with the user's psychological expectations, thereby guiding the vehicle to make driving behaviors more consistent with the user's driving habits under autonomous driving conditions.

[0043] S104. In addition to the autonomous driving system, various passive data information such as engine speed, accelerator and brake pedal opening, vehicle speed, and safe distance from other vehicles are recorded by the autonomous driving system during the user's driving process. This data is then sent to online and cloud learning modules to form a unique driving style model under that user ID. Furthermore, the refined driving style can be exported, allowing users to customize their autonomous driving experience by choosing the style of other users, such as family members, friends, or celebrities.

[0044] According to one embodiment of the present invention, in step S103, the confidence level judgment is specifically carried out in the following manner: The driving style system will judge the user's modification behavior and give a confidence level. If the confidence level exceeds a threshold, the modification is considered reasonable, and the system will change the classification of the vehicle's driving style and learn the change, adopting different driving strategies for vehicles with different driving styles. If the confidence level is too low, the modification is considered unreasonable, and the system will warn the user not to make any changes. It is understood that confidence level is a statistical concept, and its specific concept and application will not be elaborated here.

[0045] According to one embodiment of the present invention, in step S103, it is also necessary to determine the identity of each driver and assign identity information to each driver, which is represented by a user ID. After identifying the identity information, the first classification result and the second classification result corresponding to the identity information are both sent to the deep learning module to form a driving model belonging to that driver. In addition, the driving parameters corresponding to the identity information are also sent to the deep learning module. The driving model is recorded in a readable storage medium, so that the driving model can be read for implementation on different vehicles or exported for modification.

[0046] In particular, the present invention also discloses a classification strategy intelligent driving system, including a memory and a processor, wherein the memory stores a program for executing the classification strategy intelligent driving method.

[0047] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A user style classification strategy-based intelligent driving method, characterized in that, Includes the following steps: The driving styles of surrounding vehicles are automatically classified, and the first classification result is obtained. The system receives the driver's classification of the driving styles of surrounding vehicles, assesses the credibility of the driver's modified behavior, and obtains a second classification result if the credibility assessment is passed. The first and second classification results are merged, and the autonomous driving system executes the corresponding driving strategy.

2. The intelligent driving method based on the classification strategy according to claim 1, characterized in that, When judging the credibility of a driver's modification behavior, a confidence level is given. If the confidence level is greater than or equal to a set threshold, the driver's modification behavior is approved; if the confidence level is less than the set threshold, the driver's modification behavior is rejected.

3. The intelligent driving method based on the classification strategy according to claim 1, characterized in that, When a driver categorizes the driving styles of surrounding vehicles, the driver categorizes the surrounding vehicles a set number of times.

4. The intelligent driving method based on the classification strategy according to claim 3, characterized in that, The set number of times is once.

5. The intelligent driving method based on the classification strategy according to claim 1, characterized in that, Also includes: Each driver is assigned identity information. After the identity information is identified, the first classification result and the second classification result corresponding to the identity information are both sent to the deep learning module to form a driving model.

6. The intelligent driving method based on the classification strategy according to claim 5, characterized in that, After identifying the identity information, the driving parameters corresponding to the identity information are also sent to the deep learning module.

7. The intelligent driving method based on the classification strategy according to claim 5, characterized in that, The driving model is readable.

8. A classification strategy intelligent driving system, characterized in that, It includes a memory and a processor, the memory storing a program for executing the classification strategy intelligent driving method as described in any one of claims 1-7.

Citation Information

Patent Citations

  • Automatic driving control method and device, and vehicle

    CN112477884A

  • Personal driving style learning for autonomous driving

    CN112805198A