A method for identifying a driver's driving style based on the characteristics of acceleration distribution

Abstract

The present invention is applicable to the technical field of driver driving style recognition, and provides a driver driving style recognition method based on acceleration distribution characteristics. By statistical analysis, the recognition benchmark boundaries of the driver's acceleration style and the steering style benchmark boundary are obtained. Only by collecting the vehicle speed, longitudinal acceleration, and lateral acceleration data can the decoupled acceleration style and steering style be obtained. This method can simultaneously obtain the driver's acceleration and steering styles, thereby providing a basis for driver behavior analysis. In addition, the comprehensive driving style after comprehensive weighting of the acceleration style and the steering style can also be obtained.

Description

Technical Field

[0001] The present invention belongs to the technical field of driver driving style recognition, and particularly relates to a method for recognizing a driver's driving style based on acceleration distribution characteristics. Background Art

[0002] With the advancement of the new trend of "intelligentization, electrification, networking, and sharing", the automotive industry is undergoing an unprecedented technological and product revolution. At the same time, users' demands for driving experience are gradually shifting from single ride comfort to personalized performance requirements. Drivers are no longer satisfied with traditional standard driving modes but expect the vehicle to be adaptively adjusted according to their individual driving styles. Therefore, the research focus in academia and industry has gradually shifted to the recognition and application of driver driving styles, and various driving mode designs have been proposed to meet the increasingly diverse user needs. A driver's driving style is the stable driving behavior shown by the driver when operating the vehicle. Generally, a driver's driving style is stable in the short term and is related to factors such as the driver's personality and driving experience. Currently, the methods for recognizing a driver's driving style have the following limitations:

[0003] First, methods based on machine learning or deep learning require a large amount of driving style labeled data from drivers, and the algorithms are complex. In addition, it is necessary to select recognition feature parameters for driver data, and then cluster and train the feature parameters to obtain a driving style recognition model. Second, methods based on learning have a common problem, that is, poor interpretability, and their recognition process is opaque. Finally, due to the different operating styles of each vehicle, using the driver's acceleration operation amounts, including accelerator pedal opening, accelerator pedal change rate, brake pedal opening, brake pedal opening change rate, and gear position, as recognition features for the driver's driving style is not universal. Drivers with the same driving style may have different operation amounts when operating different vehicles. At the same time, the current driving style recognition algorithms can only estimate the comprehensive driving style of the driver, but cannot separately estimate the acceleration style and steering style of the driver. The two recognized operation styles are not decoupled, and there are certain limitations in applying and matching driving modes. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a method for recognizing a driver's driving style based on acceleration distribution characteristics, aiming to solve the problems proposed in the above background art.

[0005] The embodiments of the present invention are implemented as follows. A method for recognizing a driver's driving style based on acceleration distribution characteristics includes the following steps:

[0006] Step 1: Based on the publicly available existing driver driving behavior dataset, extract the existing driver's three-dimensional natural driving data in the dataset, including the longitudinal speed, longitudinal acceleration, and lateral acceleration of the existing driver;

[0007] Step 2: Divide the obtained existing driver natural driving data into a two-dimensional dataset of speed-longitudinal acceleration and a two-dimensional dataset of speed-lateral acceleration;

[0008] Based on the existing driver natural driving data, obtain the acceleration style recognition area of speed-longitudinal acceleration and the steering style recognition area of speed-lateral acceleration to analyze the acceleration style and steering style of the existing driver respectively;

[0009] For the acceleration style analysis, use a triangular closed curve to fit the acceleration style recognition reference boundary, where is the probability distribution of small-probability data of the existing driver's acceleration, is the probability distribution of common data of the existing driver's acceleration, and is solved by the following formula i :

[0010] ;

[0011] where, i represents the i th triangular closed curve; N represents the total amount of existing driver data; represents the i number of data points included in the th closed-loop curve; i represents the probability that the data points included in the

[0012] th triangular closed curve account for the total data; Through the above formula, solve the envelope line in the shape of a triangle with the data volume being the total data volume, and use it as the acceleration style recognition reference boundary of the existing driver;

[0013] Similarly, for the steering style analysis, use a triangular closed curve to fit the steering style recognition boundary, where is the probability distribution of small-probability data of the existing driver's acceleration, is the probability distribution of common data of the existing driver's acceleration, and is solved by the following formula j :

[0014] ;

[0015] where, j represents the j th triangular closed curve; N represents the total amount of existing driver data; represents thej The number of data points included in a closed-loop curve; Representing the j Probability that the data points included in the

[0016] By solving the above formula, an envelope line in the shape of a triangle with a data volume equal to the total data volume is obtained, and it is used as the recognition benchmark boundary for the steering style of existing drivers;

[0017] Step 3: Determine the driving style recognition time window T w , within a time window, the longitudinal speed, longitudinal acceleration, and lateral acceleration data of the driver to be tested are collected through vehicle sensors, and the collected data is used as the data input for the recognition of the driving style of the driver to be tested;

[0018] Step 4: Based on the existing driver acceleration style recognition benchmark boundary and steering style recognition benchmark boundary obtained in Step 2, recognize the driving style of the driver to be tested, and realize the decoupling of the acceleration style and steering style recognition by separately recognizing the acceleration style and steering style of the driver to be tested.

[0019] As a preferred embodiment of the present invention, in the said Step 1, the longitudinal acceleration is obtained according to the ratio of two speed sampling points of the existing driver to the sampling time, that is:

[0020] ;

[0021] In the formula, is the longitudinal speed at the k + 1 moment, is the longitudinal speed at the k moment, is the timestamp at the k + 1 moment, is the timestamp at the k moment, is the acceleration at the k moment.

[0022] In a further technical solution, in the said Step 3, the longitudinal speed, longitudinal acceleration, and lateral acceleration data of the vehicle are obtained through an IMU inertial measurement unit or GPS.

[0023] In a further technical solution, the said Step 4 includes the following specific steps:

[0024] Step 4.1: Acceleration style recognition, based on the acceleration style recognition benchmark boundary, define the probability interval for ordinary acceleration style drivers, where the upper bound of the interval is , and the lower bound of the interval is ; Define the lower bound of the probability interval for aggressive acceleration style drivers as , and the upper bound as ; Define the upper bound of the probability interval for conservative acceleration style drivers as​ , the lower bound is , and the specific interval is expressed as follows:

[0025] ;

[0026] The probability that the driving data of the driver to be tested exceeds the boundary of the acceleration style recognition benchmark in the total data is , if falls within the conservative acceleration style interval, it is determined that the acceleration style of the driver to be tested is conservative; if falls within the normal acceleration style interval, it is determined that the acceleration style of the driver to be tested is normal; if falls within the aggressive acceleration style interval, it is determined that the acceleration style of the driver to be tested is aggressive;

[0027] Step 4.2: Quantify the acceleration style and normalize the acceleration style, that is, by combining with the acceleration style interval to convert into an acceleration style coefficient between 0 and 1 ;

[0028] Step 4.3: Steering style recognition. Similarly, based on the steering style recognition benchmark boundary, define the probability interval of drivers with normal steering styles, where the upper bound of the interval is , and the lower bound of the interval is ; Define the lower bound of the probability interval of drivers with aggressive steering styles as , and the upper bound is ; Define the upper bound of the probability interval of drivers with conservative steering styles as , and the lower bound is ; The specific interval is expressed as follows:

[0029] ;

[0030] The probability that the driving data of the driver to be tested exceeds the boundary of the steering style recognition benchmark in the total data is , if falls within the conservative steering style interval, it is determined that the steering style of the driver to be tested is conservative; if falls within the normal steering style interval, it is determined that the steering style of the driver to be tested is normal; if falls within the aggressive steering style interval, it is determined that the steering style of the driver to be tested is aggressive;

[0031] And synchronously quantify the steering style to determine the steering style coefficient ;

[0032] Step 4.4: Determine the comprehensive driving style coefficient, based on the obtained acceleration style coefficient and the steering style coefficient , combined with the acceleration style weight coefficient and the steering style coefficient , the comprehensive driving style coefficient can be obtained. The calculation formula of the comprehensive driving style coefficient is as follows:

[0033] ;

[0034] In the formula, represents the comprehensive driving style coefficient; and represent the acceleration style weight coefficient and the steering style coefficient respectively, and both take 0.5.

[0035] As a preferred embodiment of the present invention, in the step 4.2, the calculation formula of the acceleration style coefficient is as follows:

[0036] ;

[0037] Among them, the ordinary acceleration style coefficient is 0.5, the aggressive type is greater than 0.5, and the conservative type is less than 0.5.

[0038] As a preferred embodiment of the present invention, in the step 4.3, the calculation formula of the steering style coefficient is as follows:

[0039] ;

[0040] Among them, the ordinary steering style coefficient is 0.5, the aggressive steering style coefficient is greater than 0.5 and less than 1, and the conservative steering style coefficient is less than 0.5 and greater than 0.

[0041] A driver driving style recognition method based on the acceleration distribution characteristics provided by the embodiment of the present invention obtains the decoupled acceleration style and steering style by statistical analysis of the driver acceleration style recognition reference boundary and the steering style reference boundary, and only needs to collect the vehicle speed, longitudinal acceleration and lateral acceleration data. This method can obtain the driver's acceleration and steering styles at the same time, so as to provide a basis for driver behavior analysis. In addition, the comprehensive driving style after comprehensive weighting of the acceleration style and the steering style can also be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flowchart of a driver driving style recognition method based on the acceleration distribution characteristics provided by the embodiment of the present invention;

[0043] Figure 2It is the acceleration and steering style recognition reference boundary in a driver driving style recognition method based on acceleration distribution characteristics provided by an embodiment of the present invention;

[0044] Figure 3 They are the distribution characteristics of data points of aggressive and conservative acceleration styles. Detailed implementation manners

[0045] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0046] The following describes the specific implementation of the present invention in detail with reference to specific embodiments.

[0047] As Figure 1 and 2 shown, a driver driving style recognition method based on acceleration distribution characteristics provided by an embodiment of the present invention includes the following steps:

[0048] Step 1: Based on the publicly available existing driver driving behavior data set, extract the existing driver's three-dimensional natural driving data in the data set, including the existing driver's longitudinal speed, longitudinal acceleration and lateral acceleration. Since this data is required as the recognition reference condition, a large amount and variety of existing driver's driving data is required, including various driving conditions to characterize the general existing driver's driving habits.

[0049] Step 2: Divide the obtained existing driver's natural driving data into a speed-longitudinal acceleration two-dimensional data set and a speed-lateral acceleration two-dimensional data set. Based on the existing driver's natural driving data, obtain the speed-longitudinal acceleration acceleration style recognition area and the speed-lateral acceleration steering style recognition area to analyze the acceleration style and steering style of the existing driver respectively. The finally obtained driving style recognition reference boundary is as Figure 2 shown (where a is the speed-longitudinal acceleration distribution characteristic; b is the speed-lateral acceleration distribution characteristic). For the acceleration style analysis, use a triangular closed curve to fit the acceleration style recognition reference boundary, where is the distribution probability of small-probability data of the existing driver's acceleration, is the distribution probability of common data of the existing driver's acceleration, and is solved by the following formula i :

[0050] ;

[0051] where i represents the i th triangular closed curve; NRepresents the total amount of existing driver data; Represents the i Number of data points included in the th closed-loop curve; i Represents the probability that the data points included in the

[0052] th triangular closed curve account for the total data. By the above formula, the envelope line in the shape of a triangle with the data volume being the total data volume can be solved, and it is used as the recognition benchmark boundary for the acceleration style of existing drivers.

[0053] Similarly, for the steering style analysis, a triangular closed curve is used to fit the recognition boundary of the steering style, where Is the probability distribution probability of the low-probability data of the acceleration of existing drivers, Is the probability distribution probability of the common data of the acceleration of existing drivers, and is solved by the following formula j :

[0054] ;

[0055] Among them, j Represents the j th triangular closed curve; N Represents the total amount of existing driver data; Represents the j Number of data points included in the th closed-loop curve; j Represents the probability that the data points included in the

[0056] th triangular closed curve account for the total data. By the above formula, the envelope line in the shape of a triangle with the data volume being the total data volume can be solved, and it is used as the recognition benchmark boundary for the steering style of existing drivers. The recognition benchmark boundary for the driving style of existing drivers is a preset condition for driving style recognition.

[0057] Step 3: Determine the driving style recognition time window T w , within a time window, collect the longitudinal speed, longitudinal acceleration, and lateral acceleration data of the driver to be measured through vehicle sensors, and use the collected data as the data input for the driving style recognition of the driver to be measured.

[0058] Step 4: Based on the existing driver acceleration style recognition benchmark boundary and steering style recognition benchmark boundary obtained in Step 2, identify the driving style of the driver to be tested. The existing driver driving style recognition benchmark boundary represents the average acceleration distribution characteristics of diverse drivers, which can characterize the driving habits of ordinary existing drivers, i.e., the commonly used acceleration distribution, and has universality. In addition, the percentage of the driving data of the driver to be tested outside the benchmark boundary is positively correlated with the aggressiveness of the driver to be tested. That is, the larger the percentage of data outside the benchmark boundary, the more aggressive the driver to be tested; conversely, the smaller the data outside the benchmark boundary, the more conservative the driver to be tested. By separately identifying the acceleration style and steering style of the driver to be tested, the decoupling of acceleration style and steering style recognition is achieved.

[0059] As a preferred embodiment of the present invention, in Step 1, the longitudinal acceleration can be obtained according to the ratio of two speed sampling points of the driver to the sampling time, i.e.:

[0060] ;

[0061] In the formula, is the longitudinal speed at the (k + 1)-th moment, is the longitudinal speed at the k-th moment, is the timestamp at the (k + 1)-th moment, is the timestamp at the k-th moment, is the acceleration at the k-th moment.

[0062] As a preferred embodiment of the present invention, in Step 3, data such as the longitudinal speed, longitudinal acceleration, and lateral acceleration of the vehicle can be obtained through an IMU inertial measurement unit or GPS.

[0063] As a preferred embodiment of the present invention, Step 4 includes the following specific steps:

[0064] Step 4.1: Acceleration style recognition. Based on the acceleration style recognition benchmark boundary, define the probability interval for ordinary acceleration style drivers, where the upper bound of the interval is , and the lower bound of the interval is ; define the lower bound of the probability interval for aggressive acceleration style drivers as , and the upper bound as ; define the upper bound of the probability interval for conservative acceleration style drivers as , and the lower bound as , and the specific interval expression is shown as the following formula:

[0065] ;

[0066] The distribution characteristics of data points for aggressive and conservative acceleration styles are as shown in Figure 3As shown, the probability that the driving data of the driver to be tested exceeds the acceleration style recognition benchmark boundary accounts for the total data is . If falls within the conservative acceleration style interval, it is determined that the acceleration style of the driver to be tested is conservative; if falls within the normal acceleration style interval, it is determined that the acceleration style of the driver to be tested is normal; if falls within the aggressive acceleration style interval, it is determined that the acceleration style of the driver to be tested is aggressive;

[0067] Step 4.2: Quantify the acceleration style to accurately identify the needs of the acceleration style of the driver to be tested; normalize the acceleration style, that is, by combining with the acceleration style interval, is converted into an acceleration style coefficient between 0 and 1;

[0068] Step 4.3: Steering style recognition. Similarly, based on the steering style recognition benchmark boundary, define the probability interval of drivers with normal steering styles, where the upper bound of the interval is , and the lower bound of the interval is ; define the lower bound of the probability interval of drivers with aggressive steering styles as , and the upper bound as ; define the upper bound of the probability interval of drivers with conservative steering styles as , and the lower bound as ; the specific interval expression is shown as the following formula:

[0069] ;

[0070] The probability that the driving data of the driver to be tested exceeds the steering style recognition benchmark boundary accounts for the total data is . If falls within the conservative steering style interval, it is determined that the steering style of the driver to be tested is conservative; if falls within the normal steering style interval, it is determined that the steering style of the driver to be tested is normal; if falls within the aggressive steering style interval, it is determined that the steering style of the driver to be tested is aggressive. Similarly, quantify the steering style to determine the steering style coefficient .

[0071] Step 4.4: Determine the comprehensive driving style coefficient. Based on the obtained acceleration style coefficient and the steering style coefficient , and then combined with the acceleration style weight coefficient and the steering style coefficient , the comprehensive driving style coefficient can be obtained.

[0072] As a preferred embodiment of the present invention, in the step 4.2, the acceleration style coefficient is calculated as follows:

[0073] ;

[0074] Among them, the acceleration style coefficient of the normal type is 0.5, the aggressive type is greater than 0.5, and the conservative type is less than 0.5.

[0075] As a preferred embodiment of the present invention, in the step 4.3, the steering style coefficient is calculated as follows:

[0076] ;

[0077] Among them, the steering style coefficient of the normal type is 0.5, the aggressive type steering style coefficient is greater than 0.5 and less than 1, and the conservative type steering style coefficient is less than 0.5 and greater than 0.

[0078] As a preferred embodiment of the present invention, in the step 4.4, the comprehensive driving style coefficient is calculated as follows:

[0079] ;

[0080] In the formula, represents the comprehensive driving style coefficient; and represent the acceleration style weight coefficient and the steering style coefficient respectively, and both take 0.5.

[0081] The above is only a preferred embodiment of the present invention, and it is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for identifying a driver's driving style based on acceleration distribution characteristics, characterized in that: The following steps are involved: Step 1: Based on a publicly available existing driver driving behavior dataset, extract the existing driver's three-dimensional natural driving data in the dataset, including the existing driver's longitudinal speed, longitudinal acceleration, and lateral acceleration; Step 2: dividing the obtained existing driver's three-dimensional natural driving data into a speed-longitudinal acceleration two-dimensional data set and a speed-lateral acceleration two-dimensional data set; Based on the natural driving data of existing drivers, the speed-longitudinal acceleration acceleration style recognition area and the speed-lateral acceleration steering style recognition area are obtained to analyze the acceleration style and steering style of the driver to be tested respectively; Step 3: Determine the driving style recognition time window T w , within a time window, the longitudinal speed, longitudinal acceleration and lateral acceleration data of the driver to be tested are collected through the vehicle sensor, and the collected data are used as the data input for the driving style recognition of the driver to be tested; Step 4: Based on the existing driver acceleration style recognition benchmark boundary and steering style recognition benchmark boundary obtained in step 2, the driving style of the driver to be tested is recognized, and the acceleration style and steering style recognition are decoupled by respectively recognizing the acceleration style and steering style of the driver to be tested; In step 2, for the accelerated style analysis, a triangular closed curve is used to fit the accelerated style recognition reference boundary, where is the probability distribution of the existing driver's acceleration small probability data, is the probability distribution of common data of existing driver acceleration, which can be solved by the following formula: i : ； in, i Representative i A triangular closed curve; N Represents the total amount of existing driver data; Representative i The number of data points contained in a closed-loop curve; Representative i The probability that the data points contained in the triangular closed curve account for the total data; By solving the above formula, the total amount of data is The triangular envelope is used as the benchmark boundary for identifying the acceleration style of existing drivers; Similarly, for the steering style analysis, a triangular closed curve is used to fit the steering style recognition boundary, where is the probability distribution of the existing driver's acceleration small probability data, is the probability distribution of common data of existing driver acceleration, which can be solved by the following formula: j : ； in, j Representative j A triangular closed curve; N Represents the total amount of existing driver data; Representative j The number of data points contained in a closed-loop curve; Representative j The probability that the data points contained in the triangular closed curve account for the total data; By solving the above formula, the total amount of data is The triangular envelope is used as the reference boundary for identifying the steering style of existing drivers.

2. The method for identifying a driver's driving style based on acceleration distribution characteristics according to claim 1, characterized in that: In step 1, the longitudinal acceleration is obtained according to the ratio of the two speed sampling points of the existing driver to the sampling time, that is: ； In the formula, is the longitudinal velocity at the k+1th moment, is the longitudinal velocity at the kth moment, is the timestamp of the k+1th moment, is the timestamp of the kth moment, is the acceleration at the kth moment.

3. The method for identifying a driver's driving style based on acceleration distribution characteristics according to claim 2, characterized in that: In step 3, the longitudinal velocity, longitudinal acceleration and lateral acceleration data of the vehicle are obtained through an IMU inertial measurement unit or a GPS.

4. The method for identifying a driver's driving style based on acceleration distribution characteristics according to claim 2, characterized in that: The step 4 comprises the following specific steps: Step 4.1: Acceleration style recognition: Based on the acceleration style recognition benchmark boundary, define the probability interval of drivers with normal acceleration style, where the upper limit of the interval is , the lower bound of the interval is ; Define the lower bound of the probability interval of aggressive acceleration style drivers as , the upper bound is ; The upper bound of the probability interval of conservative acceleration style drivers is defined as , the lower bound is , the specific interval is expressed as follows: ； The probability that the driving data of the driver to be tested exceeds the benchmark boundary of acceleration style recognition accounts for the total data is ,like If the acceleration style falls within the conservative acceleration style range, the acceleration style of the driver to be tested is determined to be conservative; If the acceleration style falls within the normal acceleration style range, the acceleration style of the driver to be tested is judged to be normal; If the acceleration style falls within the aggressive acceleration style range, the acceleration style of the driver to be tested is determined to be aggressive; Step 4.2: Quantify the accelerated style and normalize the accelerated style, that is, by The combination with the accelerated style interval will Converted to an acceleration style coefficient between 0 and 1 ; Step 4.3: Steering style recognition. Similarly, based on the steering style recognition benchmark boundary, define the probability interval of drivers with normal steering style, where the upper limit of the interval is , the lower bound of the interval is ; The lower bound of the probability interval of aggressive steering style drivers is defined as , the upper bound is ; Define the upper bound of the probability interval of conservative steering style drivers as , the lower bound is ; The specific interval expression is shown in the following formula: ； The probability that the driving data of the driver to be tested exceeds the reference boundary of steering style recognition accounts for the total data is ,like If the steering style falls within the conservative steering style range, the steering style of the driver to be tested is determined to be conservative; If the steering style falls within the normal range, the driver's steering style is judged to be normal. If the steering style falls within the radical steering style range, the steering style of the driver to be tested is determined to be radical; Simultaneously quantify the steering style and determine the steering style coefficient ; Step 4.4: Determine the comprehensive driving style coefficient based on the obtained acceleration style coefficient and the steering style coefficient , combined with the acceleration style weight coefficient and the steering style coefficient The comprehensive driving style coefficient can be obtained. The calculation formula of the comprehensive driving style coefficient is as follows: ； In the formula, represents the comprehensive driving style coefficient; and Represent the acceleration style weight coefficient and the steering style coefficient respectively.

5. The method for identifying a driver's driving style based on acceleration distribution characteristics according to claim 4, characterized in that: In step 4.2, the acceleration style coefficient The calculation formula is as follows: ； Among them, the ordinary acceleration style coefficient is 0.5, the radical type is greater than 0.5, and the conservative type is less than 0.

5.

6. The method for identifying a driver's driving style based on acceleration distribution characteristics according to claim 5, characterized in that: In step 4.3, the style coefficient The calculation formula is as follows: ； Among them, the ordinary steering style coefficient is 0.5, the aggressive steering style coefficient is greater than 0.5 and less than 1, and the conservative steering style coefficient is less than 0.5 and greater than 0.

7. The method for identifying a driver's driving style based on acceleration distribution characteristics according to claim 6, characterized in that: In step 4.4, and Take 0.5 for both.

Images