A method for identifying driving style of an interactive vehicle

Abstract

The application relates to a driving style recognition method of an interactive vehicle, which comprises the following steps: obtaining first trip information of a driving vehicle and second trip information of the interactive vehicle based on a preset acquisition period; obtaining average longitudinal acceleration, average longitudinal speed and the ratio of the scene speed limit of the interactive vehicle and the arrival gap difference of the driving vehicle and the interactive vehicle to the corresponding conflict position according to the first trip information and the second trip information, so as to obtain the driving characteristics of the interactive vehicle; performing clustering analysis on the driving characteristics corresponding to a plurality of continuous preset acquisition periods to obtain the driving style true value corresponding to each preset acquisition period; and recognizing the driving style of a target vehicle interacting with the driving vehicle based on the driving style true value and the driving characteristics. The driving style of the interactive vehicle is recognized by calculating three driving characteristics existing in any driving condition, so that the driving personnel of the driving vehicle can select to interact with a vehicle with a specific style according to personal preference.

Description

Technical Field

[0001] This invention relates to the field of driving style recognition technology, and more particularly to a driving style recognition method for interactive vehicles. Background Technology

[0002] Driving style recognition algorithms are generally implemented based on selected driving features and classification methods for those features. Classification methods for driving features are typically implemented through a set of rules, a model, or by using machine learning.

[0003] Rule-based methods, also known as threshold-based methods, determine a driver's driving style using a set of rules. Examples include calculating the number of aggressive maneuvers or using fuzzy logic-based methods. Model-based methods describe driving styles using a set of predefined feature equations and model each style by adjusting the model's parameters. Machine learning-based methods train a model using data and then use the model for inference. They are adaptive and allow for individual algorithm design for each driver; machine learning algorithms represent a future trend in driving style recognition.

[0004] Existing driving style recognition methods can only identify driving styles under four conditions: starting acceleration, braking deceleration, lane changing, and turning. This results in different driving features being needed to determine the driver's style under different conditions. For example, the maximum steering wheel angle is used for turning, and braking deceleration is used for braking acceleration. However, the specific conditions involved in driving are numerous and cannot be exhaustively represented. Therefore, driving style recognition algorithms that use different driving features for different conditions are difficult to satisfy all driving styles.

[0005] Therefore, there is a need to provide a driving style recognition method that is applicable to all driving conditions, effectively reduces the complexity of the recognition algorithm, and can accurately determine the driving style of the interactive vehicle to solve the above technical problems. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a driving style recognition method for interactive vehicles. It solves the problem in existing driving style recognition methods that require different driving features for different operating conditions, leading to an increase in algorithm complexity as the number of operating conditions increases.

[0007] The technical effects of this invention are achieved through the following:

[0008] A method for recognizing the driving style of an interactive vehicle includes:

[0009] The first journey information of the driving vehicle is obtained based on a preset collection period;

[0010] The second travel information of the interactive vehicle that interacts with the driving vehicle is obtained based on the preset collection period;

[0011] The average longitudinal acceleration of the interactive vehicle and the ratio of the average longitudinal velocity to the scene speed limit are obtained based on the second travel information.

[0012] Based on the currently collected first travel information, second travel information and the previously collected first travel information, second travel information, the arrival gap difference between the driving vehicle and the interactive vehicle at the corresponding conflict position is obtained. Combined with the average longitudinal acceleration of the interactive vehicle and the ratio of the average longitudinal speed to the scene speed limit, the driving characteristics of the interactive vehicle corresponding to the current collection period are obtained.

[0013] Cluster analysis is performed on the driving features corresponding to several consecutive preset collection periods to obtain the true value of the driving style corresponding to each preset collection period.

[0014] A Naive Bayes classifier is trained based on the ground truth of the driving style and the driving features to identify the driving style of the target vehicle interacting with the driving vehicle online. Three driving features are proposed, applicable to all driving conditions: the ratio of the average longitudinal acceleration of the interacting vehicle to the scene speed limit, and the difference in arrival gap between the driving vehicle and the interacting vehicle at the conflict position. This allows the driving style of the interacting vehicle to be identified by calculating these three driving features, enabling drivers to choose to interact with vehicles of specific styles according to their preferences. This eliminates the need for numerous branches within the algorithm that correspond one-to-one with different driving conditions, significantly reducing the algorithm's complexity.

[0015] Furthermore, cluster analysis is performed on the driving features corresponding to several consecutive preset collection periods to obtain the true values ​​of driving styles corresponding to each preset collection period, including:

[0016] The driving features corresponding to several consecutive preset acquisition cycles are normalized.

[0017] Three cluster centers are set, and the normalized driving features are subjected to cluster analysis to obtain the true value of driving style corresponding to each preset collection period and the three cluster centers corresponding to all the true values ​​of driving style.

[0018] The cluster centers are used to classify the true values ​​of the driving style into three categories: conservative, normal, and aggressive. Since the ratio of the average longitudinal acceleration of the interactive vehicle, the average longitudinal speed of the interactive vehicle to the scene speed limit, and the difference in arrival gap between the driving vehicle and the interactive vehicle at the conflict location are all positively correlated with the aggressiveness of the interactive vehicle, setting three cluster centers to perform cluster analysis on these three driving characteristics can distinguish the three driving styles corresponding to the three cluster centers: conservative, normal, and aggressive.

[0019] Further, training a Naive Bayes classifier based on the true values ​​of the driving style and the driving features includes:

[0020] Select a portion of consecutive acquisition cycles from a number of consecutive preset acquisition cycles, and obtain the corresponding driving features and driving style ground values ​​as training data.

[0021] A Naive Bayes classifier is trained based on the training data to obtain Gaussian distributions corresponding to the three driving styles, wherein the three dimensions of the Gaussian distribution are the three driving features corresponding to the current driving style.

[0022] Further, a Naive Bayes classifier is trained based on the ground truth of the driving style and the driving features, followed by:

[0023] Select several consecutive collection cycles from the remaining preset collection cycles, and obtain the corresponding driving features and driving style true values ​​as verification data.

[0024] The joint probability density is calculated based on the driving features and the Gaussian distributions corresponding to the three driving styles to obtain the driving style recognition result;

[0025] The identification results are compared with the ground truth values ​​of driving styles to ensure the accuracy of driving style identification. Based on a trained Naive Bayes classifier, online driving style identification is performed on selected validation data from consecutive collection periods to obtain the joint probability density of the three driving styles. The driving style with the highest joint probability density among the three types is then selected as the identification result for the interactive vehicle, and compared with the ground truth values ​​to improve the accuracy of driving style identification.

[0026] Further, the recognition result is compared with the true value of the driving style to determine the accuracy of the driving style recognition, followed by:

[0027] Acquire the driving characteristics of the target vehicle that interacts with the driving vehicle within a preset collection period;

[0028] The joint probability density is calculated based on the driving characteristics and the Gaussian distributions corresponding to the three driving styles;

[0029] The aggressiveness of the target vehicle is determined by weighting the corresponding driving style based on the joint probability density. By weighting the probability densities of three driving style types using the online identified driving styles, the aggressiveness of the interactive vehicle can be obtained. Compared to determining the aggressiveness based on the three driving style types, this method provides the driver with a more accurate style recognition result for the interactive vehicle.

[0030] Furthermore, both the first travel information and the second travel information include the position, longitudinal speed, longitudinal acceleration, and heading angle of the corresponding vehicle collected at each collection time point.

[0031] Further, based on the second travel information, the ratio of the average longitudinal acceleration of the interactive vehicle and the average longitudinal speed of the interactive vehicle to the scene speed limit is obtained, including:

[0032] The average longitudinal acceleration of the interactive vehicle is obtained based on the preset acquisition period and the longitudinal acceleration of the interactive vehicle.

[0033] Obtain scene speed limits, which are determined by traffic speed limits or by a combination of traffic speed limits and objective environmental factors;

[0034] The average longitudinal speed of the interactive vehicle is obtained based on the preset collection time and the longitudinal speed of the interactive vehicle, and then the ratio of the average longitudinal speed of the interactive vehicle to the scene speed limit is obtained by combining the scene speed limit.

[0035] Further, based on the currently collected first trip information, second trip information, and the previously collected first trip information, second trip information, the arrival gap difference between the driving vehicle and the interacting vehicle at the corresponding conflict location is obtained, including:

[0036] The conflict locations of the driving vehicle and the interacting vehicle are determined based on the current collection time point.

[0037] The first arrival gap and the second arrival gap between the driving vehicle and the interacting vehicle are obtained based on the conflict location, the first travel information and the second travel information collected at the current collection time point, and the first travel information and the second travel information collected at the previous collection time point.

[0038] The arrival gap difference is obtained based on the first arrival gap and the second arrival gap.

[0039] Furthermore, based on the positions of the driving vehicle and the interacting vehicle collected at the current collection time, the conflicting positions between them are determined, including:

[0040] Obtain the heading angle of the driving vehicle and the heading angle of the interactive vehicle;

[0041] The difference in heading angle is obtained based on the heading angle of the driving vehicle and the heading angle of the interacting vehicle;

[0042] When the difference in heading angle is less than a first preset angle, it is determined that the driving vehicle and the interacting vehicle are traveling in the same direction, and the conflict position is the front position of the vehicle in front of the two vehicles.

[0043] When the difference in heading angle is greater than or equal to a first preset angle and less than or equal to a second preset angle, it is determined that the driving vehicle and the interacting vehicle are driving at an intersection, and the conflict position is the longitudinal intersection position of the two vehicles.

[0044] When the difference in heading angle is greater than the second preset angle, it is determined that the driving vehicle and the interactive vehicle are traveling in opposite directions. The conflict position is the meeting position of the two vehicles after they have traveled at the longitudinal speed corresponding to the current collection time point.

[0045] Further, based on the conflict location, the first travel information and the second travel information collected at the current collection time point, and the first travel information and the second travel information collected at the previous collection time point, the first arrival gap and the second arrival gap between the driving vehicle and the interacting vehicle are obtained, including:

[0046] The first travel information is used to obtain the longitudinal speed of the driving vehicle at the current collection time point, and the second travel information is used to obtain the longitudinal speed of the interactive vehicle at the current collection time point, so as to determine the first arrival gap in combination with the conflict position.

[0047] Based on the first travel information, the longitudinal speed of the driving vehicle at the start time of the previous collection time point is obtained, and based on the second travel information, the longitudinal speed of the interactive vehicle at the start time of the previous collection time point is obtained, so as to determine the second arrival gap in combination with the conflict position.

[0048] As described above, the present invention has the following beneficial effects:

[0049] 1) Three driving features are proposed that are applicable to all driving conditions: the average longitudinal acceleration of the interactive vehicle, the ratio of the average longitudinal velocity of the interactive vehicle to the scene speed limit, and the difference in arrival gap between the driving vehicle and the interactive vehicle when they reach the conflict position. By calculating these three driving features, the driving style of the interactive vehicle can be identified, so that the driver of the driving vehicle can choose to interact with the vehicle with a specific style according to his or her own preferences. This eliminates the need to set up a large number of branches corresponding to different driving conditions within the algorithm, which greatly reduces the complexity of the algorithm.

[0050] 2) Since the average longitudinal acceleration of the interactive vehicle, the ratio of the average longitudinal speed of the interactive vehicle to the scene speed limit, and the difference in arrival gap between the driving vehicle and the interactive vehicle at the conflict position are all positively correlated with the aggressiveness of the interactive vehicle, setting three cluster centers to perform cluster analysis on the three driving characteristics can distinguish the three types of driving styles corresponding to the three cluster centers: conservative, normal, and aggressive.

[0051] 3) Based on the trained Naive Bayes classifier, online driving style recognition is performed on the selected continuous collection period of validation data to obtain the joint probability density of the three driving styles. The driving style with the largest joint probability density among the three driving styles is selected as the recognition result of the interactive vehicle and compared with the true value of the driving style to improve the accuracy of driving style recognition.

[0052] 4) The driving style identified online is weighted by the probability density of the three driving style types to obtain the aggressiveness of the interactive vehicle. Compared with the determination of the three driving style types, it can provide the driver with a more accurate style recognition result of the interactive vehicle. Attached Figure Description

[0053] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0054] Figure 1 A flowchart illustrating a driving style recognition method for an interactive vehicle provided in an embodiment of this specification;

[0055] Figure 2 This is an interactive diagram provided for an embodiment of the present specification when the driving vehicle and the interacting vehicle are traveling in the same direction.

[0056] Figure 3 This is an interactive diagram provided by an embodiment of the present specification when the driving vehicle and the interacting vehicle are traveling in intersections;

[0057] Figure 4 This is an interactive diagram provided for an embodiment of the present specification when the driving vehicle and the interacting vehicle are traveling in opposite directions. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0060] Example 1:

[0061] like Figure 1 As shown in the embodiments of this specification, a driving style recognition method for interactive vehicles is provided, including:

[0062] S100: Acquires the first journey information of the driving vehicle during the driving process based on a preset collection period;

[0063] S200: Based on the preset collection period, acquire the second travel information of the interactive vehicle that interacts with the driving vehicle;

[0064] In this embodiment, the driving vehicle is the vehicle currently in a driving state. This vehicle is equipped with a driving style recognition system for interactive vehicles, used to identify the driving style of the interactive vehicles interacting with it in real time. The driving style recognition system has a preset collection period, which is a constant. Every time the preset collection period is repeated, it collects the first trip information of the driving vehicle during its driving process, and at the same time, it collects the second trip information of the interactive vehicle.

[0065] The first and second travel information both include the vehicle's position, longitudinal speed, longitudinal acceleration, and heading angle collected at each data collection time point. The time interval between adjacent data collection time points is a preset data collection period; in this embodiment, the preset data collection period is 2 seconds.

[0066] S300: Based on the second travel information, obtain the average longitudinal acceleration of the interactive vehicle and the ratio of the average longitudinal speed to the scene speed limit;

[0067] In one specific implementation, step S300, based on the second travel information, obtains the ratio of the average longitudinal acceleration of the interactive vehicle and the average longitudinal speed of the interactive vehicle to the scene speed limit, including:

[0068] The average longitudinal acceleration of the interactive vehicle is obtained based on the preset acquisition period and the longitudinal acceleration of the interactive vehicle.

[0069] Specifically, the ratio of the longitudinal acceleration of the interactive vehicle to the preset collection period is the average longitudinal acceleration within the current preset collection period.

[0070] Obtain scene speed limits, which are determined by traffic speed limits or by a combination of traffic speed limits and objective environmental factors;

[0071] Specifically, when the scene speed limit is determined solely by the traffic speed limit, the scene speed limit is the maximum driving speed on the current road where the vehicle is located.

[0072] When the speed limit for a given scenario is determined jointly by traffic speed limits and objective environmental factors, the scenario speed limit is the maximum speed required of the vehicle on the current road, taking into account both the traffic speed limit and objective environmental factors. These objective environmental factors can include road conditions, weather conditions, or traffic congestion.

[0073] The average longitudinal speed of the interactive vehicle is obtained based on the preset collection time and the longitudinal speed of the interactive vehicle, and then the ratio of the average longitudinal speed of the interactive vehicle to the scene speed limit is obtained by combining the scene speed limit.

[0074] S400: Based on the currently collected first travel information, second travel information and the previously collected first travel information, second travel information, the arrival gap difference between the driving vehicle and the interactive vehicle at the corresponding conflict position is obtained, and combined with the average longitudinal acceleration of the interactive vehicle and the ratio of the average longitudinal speed to the scene speed limit, the driving characteristics of the interactive vehicle corresponding to the current collection period are obtained.

[0075] In one specific implementation, the arrival gap difference between the driving vehicle and the interacting vehicle at the corresponding conflict location is obtained based on the currently collected first travel information, second travel information, and the previously collected first travel information and second travel information, including:

[0076] The conflict locations of the driving vehicle and the interacting vehicle are determined based on the current collection time point.

[0077] The first arrival gap and the second arrival gap between the driving vehicle and the interacting vehicle are obtained based on the conflict location, the first travel information and the second travel information collected at the current collection time point, and the first travel information and the second travel information collected at the previous collection time point.

[0078] The arrival gap difference is obtained based on the first arrival gap and the second arrival gap.

[0079] In one specific implementation, determining the conflict location between the driving vehicle and the interacting vehicle based on their positions collected at the current collection time includes:

[0080] Obtain the heading angle of the driving vehicle and the heading angle of the interactive vehicle;

[0081] The difference in heading angle is obtained based on the heading angle of the driving vehicle and the heading angle of the interacting vehicle;

[0082] When the difference in heading angle is less than a first preset angle, it is determined that the driving vehicle and the interacting vehicle are traveling in the same direction, and the conflict position is the front position of the vehicle in front of the two vehicles.

[0083] When the difference in heading angle is greater than or equal to a first preset angle and less than or equal to a second preset angle, it is determined that the driving vehicle and the interacting vehicle are driving at an intersection, and the conflict position is the longitudinal intersection position of the two vehicles.

[0084] When the difference in heading angle is greater than the second preset angle, it is determined that the driving vehicle and the interactive vehicle are traveling in opposite directions. The conflict position is the meeting position of the two vehicles after they have traveled at the longitudinal speed corresponding to the current collection time point.

[0085] In this application, the longitudinal direction of the vehicle is located in the extension direction of the road where the vehicle is located. The longitudinal direction of the vehicle refers to the projection direction of the vehicle's front direction onto the extension direction of the road where the vehicle is located. When the driving vehicle and the interacting vehicle are determined to be traveling in the same direction, the lanes of the two vehicles are parallel, and the longitudinal directions of the two vehicles are respectively in the same direction.

[0086] In this embodiment, the first preset angle is 30 degrees. When the angle between the frontal direction of the driving vehicle and the frontal direction of the interacting vehicle is less than 30 degrees, it is determined that the driving vehicle and the interacting vehicle are traveling in the same direction. The conflict position between the driving vehicle and the interacting vehicle is the front end of the vehicle that is located further forward in the longitudinal direction. Here, the frontal direction is the direction from the rear of the vehicle to the front of the vehicle.

[0087] Specifically, the conflict location is an area within a preset distance in front of the vehicle at a relatively forward position. The length of the area is the preset distance, and the width of the area is the coverage width between the lane where the driving vehicle is located and the lane where the interacting vehicle is located. The preset distance can be set by those skilled in the art.

[0088] For example, when the angle between the direction of the driving vehicle's front and the direction of the interactive vehicle's front is 0 degrees, the corresponding interactive diagram is as follows: Figure 2 As shown.

[0089] In this embodiment, the second preset angle is 150 degrees. When the difference in heading angle is greater than or equal to 30 degrees and less than or equal to 150 degrees, it is determined that the driving vehicle and the interacting vehicle are traveling in intersection, and the conflict position is the longitudinal intersection position of the two vehicles.

[0090] The center of the conflict location is the intersection point between the projection directions of the two vehicles, and the conflict location is the overlapping area of ​​the ground projections of the two vehicles when they are traveling in their respective projection directions.

[0091] For example, when the angle between the direction of the driving vehicle's front and the direction of the interactive vehicle's front is 90 degrees, the corresponding interactive diagram is as follows: Figure 3 As shown.

[0092] In this embodiment, when the angle between the front direction of the driving vehicle and the front direction of the interactive vehicle is greater than 150 degrees, it is determined that the driving vehicle and the interactive vehicle are traveling in opposite directions. The conflict position is the meeting position of the two vehicles after they travel at the longitudinal speed corresponding to the current collection time point.

[0093] Specifically, the longitudinal speed refers to the component of the vehicle's current travel direction speed in the aforementioned projection direction, and the conflict position is the meeting position when both vehicles are traveling at their corresponding component speeds.

[0094] The centerline of the meeting point is the same horizontal line where the front ends of both vehicles arrive simultaneously, and this centerline is perpendicular to the projection direction. The conflict area is a region of a predetermined length based on the centerline, and its width is the coverage width between the lane of the driving vehicle and the lane of the interacting vehicle. The predetermined length can be set by those skilled in the art.

[0095] For example, when the angle between the direction of the driving vehicle's front and the direction of the interactive vehicle's front is 180 degrees, the corresponding interactive diagram is as follows: Figure 4 As shown. Figure 2 , Figure 3 and Figure 4 The word "in" refers to the vehicle being driven in this application.

[0096] It should be noted that the above-mentioned determinations of same-direction driving, merging driving, and oncoming driving are based on the premise that there is interaction between the driving vehicle and the interacting vehicle. For example, when a vehicle is traveling in the opposite direction to the driving vehicle, this vehicle cannot be considered an interacting vehicle of the driving vehicle. Therefore, even if the angle between the frontal directions of the driving vehicle and the frontal directions of the other vehicle is 180 degrees, it cannot be determined that the two vehicles are traveling in opposite directions.

[0097] In one specific implementation, the first arrival gap and the second arrival gap between the driving vehicle and the interacting vehicle are obtained based on the conflict location, the first travel information collected at the current collection time point, the second travel information, and the first travel information and the second travel information collected at the previous collection time point, including:

[0098] The first travel information is used to obtain the longitudinal speed of the driving vehicle at the current collection time point, and the second travel information is used to obtain the longitudinal speed of the interactive vehicle at the current collection time point, so as to determine the first arrival gap in combination with the conflict position.

[0099] Based on the first travel information, the longitudinal speed of the driving vehicle at the start time of the previous collection time point is obtained, and based on the second travel information, the longitudinal speed of the interactive vehicle at the start time of the previous collection time point is obtained, so as to determine the second arrival gap in combination with the conflict position.

[0100] Specifically, the difference between the time required for a driving vehicle to reach the corresponding conflict area at the current data collection time and the time required for an interactive vehicle to reach the conflict area is the first arrival gap.

[0101] The difference between the time required for the driving vehicle to reach the conflict area at the previous data collection time and the time required for the interactive vehicle to reach the conflict area is the second arrival gap. It should be noted that the conflict area mentioned is still the conflict area corresponding to the current data collection time.

[0102] S500: Perform cluster analysis on the driving features corresponding to several consecutive preset collection periods to obtain the true value of the driving style corresponding to each preset collection period.

[0103] In one specific implementation, step S500 involves performing cluster analysis on the driving features corresponding to several consecutive preset collection periods to obtain the true values ​​of driving styles corresponding to each preset collection period, including:

[0104] The driving features corresponding to several consecutive preset acquisition cycles are normalized.

[0105] Three cluster centers are set, and the normalized driving features are subjected to cluster analysis to obtain the true value of driving style corresponding to each preset collection period and the three cluster centers corresponding to all the true values ​​of driving style.

[0106] The cluster centers are used to classify the true values ​​of the driving style into three types of driving styles: conservative, normal, and aggressive.

[0107] First, 100 consecutive sets of driving features are acquired at a 2-second acquisition cycle. Each set of driving features includes: the average longitudinal acceleration of the interactive vehicle, the ratio of the average longitudinal speed of the interactive vehicle to the scene speed limit, and the arrival gap difference. The 100 sets of driving features are then normalized to prevent imbalances in the weights of each driving feature during clustering, ensuring that the three normalized driving features all satisfy the conditions of a mean of 0 and a variance of 1.

[0108] Three cluster centers are set, and the normalized driving features are used to perform driving style clustering analysis using K-Means clustering. After the clustering calculation is completed, the ground truth value of driving style corresponding to each group of driving features is obtained, as well as the three cluster centers corresponding to all driving style ground truth values. The K-Means clustering method is prior art and will not be described in detail in this application.

[0109] Since all three driving characteristics mentioned above are positively correlated with the aggressiveness of the interactive vehicle, the three cluster centers are divided into three driving styles: conservative, normal, and aggressive.

[0110] S600: Train a Naive Bayes classifier based on the true values ​​of the driving style and the driving features to identify the driving style of the target vehicle interacting with the driving vehicle online.

[0111] In one specific implementation, step S600 trains a Naive Bayes classifier based on the true value of the driving style and the driving features, including:

[0112] Select a portion of consecutive acquisition cycles from a number of consecutive preset acquisition cycles, and obtain the corresponding driving features and driving style ground values ​​as training data.

[0113] A Naive Bayes classifier is trained based on the training data to obtain Gaussian distributions corresponding to the three driving styles, wherein the three dimensions of the Gaussian distribution are the three driving features corresponding to the current driving style.

[0114] First, the first 70 sets of driving features are obtained from the aforementioned 100 sets of driving features as training data. Driving features during the driving process are a typical type of continuous data. When applying the Naive Bayes algorithm, it is often assumed that the driving features conform to a Gaussian distribution, and the mean and variance of the training data are used to estimate the distribution parameters of each driving feature in each driving style, i.e., the mean and variance, calculated as follows. It should be noted that the aforementioned driving features during the driving process refer to the real-time driving features of the driving vehicle and the interactive vehicle during the driving process, which is continuous data. The three driving features in this application's technical solution are equivalent to sampled data obtained by collecting the continuous data according to a preset collection period. The Naive Bayes algorithm corresponding to the Naive Bayes classifier is existing technology and will not be described further in this application.

[0115] The formula for calculating the mean is:

[0116] Variance calculation formula:

[0117] Where i represents the i-th driving style and j represents the j-th driving characteristic. ij This refers to the number of samples with driving style i and driving characteristic j; This refers to the sample mean where driving style is i and driving characteristic is j; This refers to the sample variance of driving style i and driving characteristic j; x k It iterates through all sample values ​​with driving style i and driving feature j.

[0118] This results in three Gaussian distributions, each representing a driving style. Each Gaussian distribution is a three-dimensional Gaussian distribution, where each dimension represents the distribution of the corresponding driving characteristic under that driving style.

[0119] In one specific implementation, a Naive Bayes classifier is trained based on the ground truth of the driving style and the driving features, followed by:

[0120] Select several consecutive collection cycles from the remaining preset collection cycles, and obtain the corresponding driving features and driving style true values ​​as verification data.

[0121] The joint probability density is calculated based on the driving features and the Gaussian distributions corresponding to the three driving styles to obtain the driving style recognition result;

[0122] The recognition results are compared with the true values ​​of the driving style to ensure the accuracy of driving style recognition.

[0123] Specifically, the last 30 driving features out of the aforementioned 100 sets of driving features and their corresponding ground truth values ​​for driving styles were selected as validation data. Based on the trained Naive Bayes classifier, online driving style recognition was performed on the last 30 sets of driving features. The driving style with the highest joint probability density among the three categories was taken as the recognition result and compared with the ground truth values ​​for driving styles. If they were the same, the driving style recognition was considered accurate. The ratio of the number of accurately recognized sets to the number of sets in the validation data was the recognition accuracy rate.

[0124] When the recognition accuracy of 30 sets of verification data reaches the preset standard, the driving style obtained by the currently trained Naive Bayes classifier is considered to be highly reliable. Therefore, the trained Naive Bayes classifier can be used for online recognition of the driving style of the interacting vehicle to determine which of the three driving styles—conservative, normal, or aggressive—the currently driving vehicle belongs to. The preset standard can be set by those skilled in the art.

[0125] By proposing three driving features applicable to all driving conditions—the average longitudinal acceleration of the interactive vehicle, the ratio of the average longitudinal velocity of the interactive vehicle to the scene speed limit, and the difference in arrival gap between the driving vehicle and the interactive vehicle at the conflict position—the driving style of the interactive vehicle can be identified by calculating these three driving features. This allows drivers to choose to interact with vehicles of a specific style according to their preferences, eliminating the need to set up a large number of branches within the algorithm that correspond one-to-one with different driving conditions, thus greatly reducing the complexity of the algorithm.

[0126] In one specific implementation, the recognition result is compared with the true value of the driving style to determine the accuracy of the driving style recognition, followed by:

[0127] Acquire the driving characteristics of the target vehicle that interacts with the driving vehicle within a preset collection period;

[0128] The joint probability density is calculated based on the driving characteristics and the Gaussian distributions corresponding to the three driving styles;

[0129] The aggressiveness of the target vehicle is determined by weighting the corresponding driving style based on the joint probability density.

[0130] Specifically, an evaluation criterion is set, with three driving styles—conservative, normal, and aggressive—corresponding to different scores. Using the joint probability density of each of the three driving styles as weights, and combining this with the parameters corresponding to different driving styles, a specific score reflecting the aggressiveness of the target vehicle is obtained, thus determining the aggressiveness of the interactive vehicle. Compared to simply defining three driving style types, this approach provides drivers with more accurate style recognition results for the interactive vehicle.

[0131] For example, the evaluation criteria set by those skilled in the art are: a score of 90 for conservative, 70 for normal, and 50 for aggressive. The joint probability densities of the three driving styles (conservative, normal, and aggressive) corresponding to a set of driving characteristics are 0.2, 0.3, and 0.5, respectively. Then, the weighted value of the joint probability density obtained based on the evaluation criteria, that is, the specific score of the aggressiveness of the target vehicle, is 90*(0.2 / 0.2+0.3+0.5)+70*(0.3 / 0.2+0.3+0.5)+50*(0.5 / 0.2+0.3+0.5)=64. This score represents the aggressiveness of the target vehicle. That is, the lower the score, the more aggressive the driver's driving state of the target vehicle is, and the higher the score, the more conservative the driver's driving state of the target vehicle is.

[0132] It should be noted that the technical solution of this application can be applied to the field of autonomous driving decision-making, and combined with autonomous driving decision-making algorithms to provide them with style recognition results of surrounding vehicles, so that autonomous driving decision-making algorithms can avoid a lot of interaction with extremely aggressive drivers and improve the safety of autonomous driving algorithms; it can also be applied to advanced driver assistance systems to provide drivers with style recognition results of surrounding vehicles, so that drivers can choose to interact with vehicles of a specific style according to their own preferences, thereby improving drivers' satisfaction with the system.

[0133] While the present invention has been described through preferred embodiments, it is not limited to the embodiments described herein, and various changes and modifications are made without departing from the scope of the invention.

[0134] Where there is no conflict, the above embodiments and features described herein can be combined with each other.

[0135] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for recognizing the driving style of an interactive vehicle, characterized in that, include: The first journey information of the driving vehicle is obtained based on a preset collection period; The second travel information of the interactive vehicle that interacts with the driving vehicle is obtained based on the preset collection period; Both the first travel information and the second travel information include the position, longitudinal speed, longitudinal acceleration, and heading angle of the corresponding vehicle collected at each collection time point; the preset collection period is the time interval between adjacent collection time points; The average longitudinal acceleration of the interactive vehicle and the ratio of the average longitudinal velocity to the scene speed limit are obtained based on the second travel information. The speed limit in the scenario is determined by traffic speed limits, or by a combination of traffic speed limits and objective environmental factors; Based on the currently collected first travel information, second travel information and the previously collected first travel information, second travel information, the arrival gap difference between the driving vehicle and the interactive vehicle at the corresponding conflict position is obtained. Combined with the average longitudinal acceleration of the interactive vehicle and the ratio of the average longitudinal speed to the scene speed limit, the driving characteristics of the interactive vehicle corresponding to the current collection period are obtained. Cluster analysis is performed on the driving features corresponding to several consecutive preset collection periods to obtain the true value of the driving style corresponding to each preset collection period. A Naive Bayes classifier is trained based on the true values ​​of the driving style and the driving features to identify the driving style of the target vehicle interacting with the driving vehicle online. The step of obtaining the arrival gap difference between the driving vehicle and the interacting vehicle at the corresponding conflict location based on the currently collected first travel information, second travel information, and the previously collected first travel information and second travel information includes: The conflict locations of the driving vehicle and the interacting vehicle are determined based on the current collection time point. The first arrival gap and the second arrival gap between the driving vehicle and the interacting vehicle are obtained based on the conflict location, the first travel information and the second travel information collected at the current collection time point, and the first travel information and the second travel information collected at the previous collection time point. The arrival gap difference is obtained based on the first arrival gap and the second arrival gap.

2. The driving style recognition method for interactive vehicles according to claim 1, characterized in that, Cluster analysis is performed on the driving features corresponding to several consecutive preset collection periods to obtain the true values ​​of driving styles corresponding to each preset collection period, including: The driving features corresponding to several consecutive preset collection periods are normalized. Three cluster centers are set, and the normalized driving features are subjected to cluster analysis to obtain the true value of driving style corresponding to each preset collection period and the three cluster centers corresponding to all the true values ​​of driving style. The cluster centers are used to classify the true values ​​of the driving style into three types of driving styles: conservative, normal, and aggressive.

3. The driving style recognition method for interactive vehicles according to claim 2, characterized in that, Training a Naive Bayes classifier based on the true values ​​of the driving style and the driving features includes: Select a portion of consecutive acquisition cycles from a number of consecutive preset acquisition cycles, and obtain the corresponding driving features and driving style ground values ​​as training data. A Naive Bayes classifier is trained based on the training data to obtain Gaussian distributions corresponding to the three driving styles, wherein the three dimensions of the Gaussian distribution are the three driving features corresponding to the current driving style.

4. The driving style recognition method for interactive vehicles according to claim 3, characterized in that, Based on the ground truth of the driving style and the driving features, a Naive Bayes classifier is trained, followed by: Select several consecutive collection cycles from the remaining preset collection cycles, and obtain the corresponding driving features and driving style true values ​​as verification data. The joint probability density is calculated based on the driving features and the Gaussian distributions corresponding to the three driving styles to obtain the driving style recognition result; The recognition results are compared with the true values ​​of the driving style to ensure the accuracy of driving style recognition.

5. The driving style recognition method for interactive vehicles according to claim 4, characterized in that, The recognition result is compared with the true value of the driving style to determine the accuracy of the driving style recognition, followed by: Acquire the driving characteristics of the target vehicle that interacts with the driving vehicle within a preset collection period; The joint probability density is calculated based on the driving characteristics and the Gaussian distributions corresponding to the three driving styles; The aggressiveness of the target vehicle is determined by weighting the corresponding driving style based on the joint probability density.

6. The driving style recognition method for interactive vehicles according to claim 1, characterized in that, The average longitudinal acceleration of the interactive vehicle and the ratio of the average longitudinal speed of the interactive vehicle to the scene speed limit are obtained based on the second travel information, including: The average longitudinal acceleration of the interactive vehicle is obtained based on the preset acquisition period and the longitudinal acceleration of the interactive vehicle. Obtain scene speed limits, which are determined by traffic speed limits or by a combination of traffic speed limits and objective environmental factors; The average longitudinal speed of the interactive vehicle is obtained based on the preset collection period and the longitudinal speed of the interactive vehicle, and then the ratio of the average longitudinal speed of the interactive vehicle to the scene speed limit is obtained by combining the scene speed limit.

7. The driving style recognition method for interactive vehicles according to claim 1, characterized in that, Based on the positions of the driving vehicle and the interacting vehicle collected at the current collection time, the conflict positions corresponding to the two are determined, including: Obtain the heading angle of the driving vehicle and the heading angle of the interactive vehicle; The difference in heading angle is obtained based on the heading angle of the driving vehicle and the heading angle of the interacting vehicle; When the difference in heading angle is less than a first preset angle, it is determined that the driving vehicle and the interacting vehicle are traveling in the same direction, and the conflict position is the front position of the vehicle in front of the two vehicles. When the difference in heading angle is greater than or equal to a first preset angle and less than or equal to a second preset angle, it is determined that the driving vehicle and the interacting vehicle are driving at an intersection, and the conflict position is the longitudinal intersection position of the two vehicles. When the difference in heading angle is greater than the second preset angle, it is determined that the driving vehicle and the interactive vehicle are traveling in opposite directions. The conflict position is the meeting position of the two vehicles after they have traveled at the longitudinal speed corresponding to the current collection time point.

8. The driving style recognition method for interactive vehicles according to claim 1, characterized in that, Based on the conflict location, the first travel information and the second travel information collected at the current collection time point, and the first travel information and the second travel information collected at the previous collection time point, the first arrival gap and the second arrival gap between the driving vehicle and the interacting vehicle are obtained, including: The first travel information is used to obtain the longitudinal speed of the driving vehicle at the current collection time point, and the second travel information is used to obtain the longitudinal speed of the interactive vehicle at the current collection time point, so as to determine the first arrival gap in combination with the conflict position. Based on the first travel information, the longitudinal speed of the driving vehicle at the start time of the previous collection time point is obtained, and based on the second travel information, the longitudinal speed of the interactive vehicle at the start time of the previous collection time point is obtained, so as to determine the second arrival gap in combination with the conflict position.

Images