A driving behavior analysis method based on a weighted cost function

Abstract

The application provides a driving behavior analysis method based on a weighted cost function, which comprises the following steps: collecting initial characteristic performance data of a plurality of driving trajectories; pre-processing the initial characteristic performance data to obtain characteristic performance data; dividing the driving trajectories into a plurality of decision cycle segments according to the duration of different driving events; constructing a cost function representing the importance of comfort, safety and speed of a driver during driving according to the characteristic performance data; processing the characteristic performance data corresponding to the plurality of decision cycle segments based on the cost function and an entropy weight method to obtain weight coefficients corresponding to the characteristic performance data; and clustering different driving trajectories based on a k-means method and the weight coefficients corresponding to the characteristic performance data of different driving trajectories, so as to realize clustering of driving personnel corresponding to different driving trajectories.

Description

Technical Field

[0001] This invention relates to a driving behavior analysis method, and more specifically, to a driving behavior analysis method based on a weighted cost function. Background Technology

[0002] During driving, drivers determine their vehicle's operation based on the traffic environment and their personal driving goals. When making decisions, drivers consider the costs involved, aiming to minimize those costs. This consideration of costs reflects their driving characteristics or style, categorizing them as cautious, aggressive, or average drivers. Classifying driving behavior not only serves as a warning to drivers, reminding them to pay attention to their actions and correct bad driving habits to avoid traffic accidents, but also provides a reference for vehicle insurance companies to adjust insurance premiums accordingly.

[0003] How to classify the driving trajectory of drivers, and thus classify the driving behavior of drivers, has become a technical challenge in this field. Summary of the Invention

[0004] To overcome the shortcomings of the above-mentioned technical problems, this invention provides a driving behavior analysis method based on a weighted cost function, which enables the clustering of drivers corresponding to different driving trajectories.

[0005] This invention provides a driving behavior analysis method based on a weighted cost function, characterized in that the method includes:

[0006] Collect initial characteristic performance data for several driving trajectories;

[0007] The initial characteristic performance data is processed to obtain characteristic performance data; the processing includes taking the derivative of the longitudinal acceleration in the initial characteristic performance data to obtain the longitudinal acceleration.

[0008] Each driving trajectory is divided into several decision cycle segments based on the duration of different driving events; preferably, the different driving events include constant speed behavior processes, acceleration behavior processes, and deceleration behavior processes;

[0009] A cost function is constructed based on characteristic performance data to characterize the driver's emphasis on comfort, safety, and speed during driving.

[0010] Based on the cost function and entropy weight method, the feature performance data corresponding to several decision period segments of each driving trajectory are processed to obtain the weight coefficients corresponding to the feature performance.

[0011] Clustering of different driving trajectories is performed based on the k-means method and the weight coefficients corresponding to the feature performance of different driving trajectories.

[0012] The characteristic performance data includes effective velocity v and effective longitudinal acceleration a. x Effective lateral acceleration a y Effective longitudinal acceleration jerk.

[0013] A further preferred embodiment of a driving behavior analysis method based on a weighted cost function includes determining the category to which the driver's driving trajectory belongs based on the weight coefficients corresponding to the characteristic performance of the driver's driving trajectory and the Euclidean distance between the cluster centers of different categories of driving trajectories after clustering; the category corresponding to the cluster center with the smallest Euclidean distance is the category to which the driver's driving trajectory belongs.

[0014] Compared with the prior art, the present invention has the following beneficial effects: the cost function can characterize the cost that the driver considers when making decisions on different performance indicators through the driver's driving trajectory, thereby characterizing the driver's driving behavior characteristics; by dividing the driving trajectory into several decision cycle segments according to the duration of different driving events, it is beneficial to avoid the problem of misjudgment of driving behavior caused by similar driving behavior characteristics under different circumstances when characterizing the characteristic performance data in the driving trajectory based on the cost function.

[0015] Furthermore, the driving behavior analysis method based on the weighted cost function further includes preprocessing the initial characteristic performance data before processing it; applying a moving average filter to the initial velocity v; and applying a moving average filter to the initial longitudinal acceleration a. x and lateral acceleration a y Perform wavelet filtering, selecting sym8 as the wavelet basis function, and decomposition level greater than or equal to 5.

[0016] The beneficial effect of adopting the above-described further technical solution is that, through the preprocessing, the initial a in the initial characteristic performance data is obtained. x and a y Wavelet filtering was performed to achieve a x and a y The curve became smoother, eliminating several small a values ​​caused by other interference. x or a y This leads to significant deviations in the calculation of the weight coefficients of each feature performance in the subsequent cost function.

[0017] Furthermore, the specific steps for dividing the driving trajectory into several decision-making cycle segments are as follows:

[0018] Calculate the radius of the driving trajectory; based on longitudinal acceleration α. x Construct the function y = a x Extract the function y = a x The complete waveform in the image is obtained by extracting the α portion of the complete waveform. x Peak or trough value;

[0019] Based on the driving trajectory radius and α x The peak or trough value divides the continuous driving trajectory into several decision cycle segments.

[0020] Furthermore, the method for calculating the radius of the driving trajectory is as follows: R is the radius of the driving trajectory, v is the speed, and a y It is lateral acceleration;

[0021] or,

[0022] The method for calculating the radius of the driving trajectory is as follows: R is the radius of the driving trajectory, v is the speed, and a y ε is the lateral acceleration; ε1 is any value from 0 to 0.0015.

[0023] The beneficial effect of adopting the above-mentioned further technical solution is that it enables the calculation of the driving trajectory radius; through While obtaining the radius of the driving trajectory, the problem of the radius not existing when driving at a constant speed is avoided, and the drawbacks caused by preprocessing the initial feature performance indicators are also resolved. This avoids the removal of a small amount of data with small accelerations during preprocessing, which would lead to a large deviation in the calculation of the weight coefficients of each feature performance in the subsequent cost function.

[0024] Furthermore, the complete waveform is α. x The waveform increases from 0 to the peak value and then decreases back to 0; or, the complete waveform is α. x It decreases from 0 to a trough value and then increases back to 0;

[0025] and / or

[0026] The specific process of dividing the decision-making cycle into several segments is as follows:

[0027] Specific segmentation criteria are established to divide the driving trajectory into several decision-making cycle segments.

[0028] The specific segmentation criteria are as follows:

[0029] When the driving radius R ≥ 1000m, it is a straight driving trajectory; when the driving radius R < 1000m, it is a turning driving trajectory.

[0030] When the driving radius R ≥ 1000m, the longitudinal acceleration αx Peak wave length > 1.0 m / s 2 The continuous driving trajectory at that time is the first decision cycle segment;

[0031] When the driving radius R ≥ 1000m, -1.0m / s 2 ≤Longitudinal acceleration α x Peak or trough value ≤ 1.0 m / s 2 The continuous driving trajectory at that time is the second decision cycle segment;

[0032] When the driving radius R ≥ 1000m, the longitudinal acceleration α x Valley value < -1.0 m / s 2 The continuous driving trajectory at that time is the third decision cycle segment;

[0033] When the driving radius R < 1000m, the longitudinal acceleration α x Peak wave length > 1.0 m / s 2 The continuous driving trajectory at that time is the fourth decision cycle segment;

[0034] When the driving radius R < 1000m, -1.0m / s 2 ≤Longitudinal acceleration α x Peak or trough value ≤ 1.0 m / s 2 The continuous driving trajectory at that time is the fifth decision cycle segment;

[0035] When the driving radius R < 1000m, the longitudinal acceleration α x Valley value < -1.0 m / s 2 The continuous driving trajectory at that time is the sixth decision cycle segment.

[0036] That is, the first decision cycle segment is an acceleration straight-line cycle segment, the second decision cycle segment is a constant speed straight-line cycle segment, and the third decision cycle segment is a deceleration straight-line cycle segment; the fourth decision cycle segment is an acceleration turning cycle segment, the fifth decision cycle segment is a constant speed turning cycle segment, and the sixth decision cycle segment is a deceleration turning cycle segment.

[0037] The beneficial effect of adopting the above-mentioned further technical solution is that it divides the driving data corresponding to each driver's driving trajectory into six basic events: accelerating straight, driving straight at a constant speed, driving straight at a deceleration, accelerating and turning, turning at a constant speed, and turning at a deceleration; at the same time, it can record the start and end times of each event to obtain the duration of the event.

[0038] Furthermore, the cost function is:

[0039]

[0040] In the formula, T represents the duration of the decision-making cycle; ω jω represents the weighting coefficients corresponding to the feature performance. j Including ω1, ω2, ω3, and ω4; ω1, ω2, ω3, and ω4 represent the longitudinal acceleration a. x Weighting coefficients, longitudinal acceleration (jerk), lateral acceleration (a) y The weighting coefficients of , velocity v, and ; v d v is the driver's desired driving speed. d The value is the maximum speed limit of the road being traveled.

[0041] Preferably, the higher the weight of a characteristic performance, the more important the driver is to that performance indicator; conversely, the lower the weight, the less important it is to the driver.

[0042] The cost function is:

[0043]

[0044] In the formula, T represents the duration of the decision-making cycle; ω j ω represents the weighting coefficients corresponding to the feature performance. j Including ω1, ω2, ω3, and ω4; ω1, ω2, ω3, and ω4 represent the longitudinal acceleration a. x Weighting coefficients, longitudinal acceleration (jerk), lateral acceleration (a) y The weighting coefficients of , velocity v, and ;

[0045] ε1 is any value from 0 to 0.0015; ε1 is preferably 0.001; Vt is the reciprocal of the sampling frequency of the original data; v d For the driver's desired driving speed, v is the speed during the first decision cycle. d For a x The speed decreased from the peak to 0.1 m / s 2 The average velocity thereafter; for the second decision period, v d v is the average velocity during the second decision-making period; for the third decision-making period, v d For a x Descending to -0.1 m / s 2 The average velocity before; for the fourth decision period, v d v is the average speed after exiting the curve; for the fifth decision period, v d The average velocity for the fifth decision cycle; for the sixth decision cycle, v d The average speed before entering the curve.

[0046] The beneficial effect of adopting the above-mentioned further technical solution is that the cost function can characterize the various feature performances of the driving trajectory, and at the same time obtain the weight coefficients corresponding to the feature performances.

[0047] Additionally

[0048] The introduction of ε1 avoids the problem of integrating to zero in the cost function when some accelerations are small during preprocessing, which is equivalent to running at a constant speed. It also avoids the problem of integrating to zero during the second and fifth decision periods when running at a constant speed in actual uniform speed operation. Furthermore, it avoids the drawback of removing some small acceleration data during preprocessing. However, because the value of ε1 is small, although it avoids the problem of integrating to zero and does not affect the calculation of larger acceleration values ​​in the cost function, the problem of integrating to zero with respect to ε1 when some accelerations are zero due to preprocessing still results in very small values. This leads to a large difference between ω1, ω2, and ω3 and ω4, which does not accurately reflect the actual situation. Through the aforementioned v... d Instead of using fixed values, different processing methods are used in the second and fifth decision cycles. This avoids the problem that the integral value of ε1 is very small when the acceleration is 0, which is caused by preprocessing, resulting in a large difference between ω1, ω2, and ω3 and ω4. Ultimately, it can truly reflect the real characteristics of the driving trajectory.

[0049] Furthermore, the specific process for obtaining the weight coefficients corresponding to the aforementioned feature performance is as follows:

[0050] Extract the feature performance data corresponding to several decision cycle segments of the driving trajectory;

[0051] The characteristic performance data corresponding to the decision cycle segment are converted into variable data according to the first transformation formula; the variable data includes the defined property velocity vw. T Definition of longitudinal acceleration a x w T 1. Define the property of lateral acceleration a y w T Definition of properties: longitudinal acceleration (jerkw) T ;

[0052] The variable data corresponding to several decision-making period segments of the driving trajectory are processed using the entropy weight method to obtain the weight coefficients corresponding to the variable data; the weight coefficients corresponding to the feature performance are obtained through the weight coefficients corresponding to the variable data.

[0053] Furthermore, the first conversion formula includes

[0054] or,

[0055] The first conversion formula includes ε1 can be any value from 0 to 0.0015; ε1 is preferably 0.001.

[0056] Furthermore, the process of processing the variable data corresponding to several decision-making period segments of the driving trajectory based on the entropy weight method is as follows: the variable data corresponding to several decision-making period segments are standardized using the second formula.

[0057] The second formula is Where, x ij x' represents the j-th variable data for the i-th decision period segment of the driving trajectory. ij For x ij Normalized data;

[0058] Using the third formula to evaluate x' ij Processing to obtain p ij p ij The third formula represents the proportion of the j-th variable data in the i-th decision period segment of the driving trajectory to the j-th variable data in all decision period segments of the driving trajectory; i = 1, ..., n, j = 1, ..., m;

[0059] Using the fourth formula to apply p ij Processing yields e j The e j Let be the entropy value of the j-th variable in the driving trajectory;

[0060] The fourth formula is: Where, k = 1 / ln(n);

[0061] Using the fifth formula to evaluate e j Processing yields d j The d j The information entropy redundancy of the j-th variable in the driving trajectory;

[0062] The fifth formula is: d j =1-e j j = 1, ..., m;

[0063] Using the sixth formula to analyze d j Processing to obtain ω j The ω j The weighting coefficients for the characteristic performance corresponding to the j-th variable of the driving trajectory; the sixth formula is:

[0064] The beneficial effect of adopting the above-mentioned further technical solution is that, by using the values ​​of the four performance indicators in all decision-making period segments of the driving trajectory corresponding to each driver, the entropy weight method is used to obtain the weight coefficients ω1, ω2, ω3, and ω4 of the four performance indicators of the driving trajectory corresponding to each driver.

[0065] Furthermore, the specific process of classifying different driving trajectories is as follows:

[0066] Based on the k-means method, different driving trajectories are divided into three categories according to the weight coefficients corresponding to the feature performance of the driving trajectory;

[0067] The longitudinal acceleration a of the first to third categories x The cluster centers of the corresponding weight coefficients gradually decrease;

[0068] The cluster centers of the weight coefficients corresponding to the longitudinal acceleration jerk in categories 1-3 gradually decrease;

[0069] Lateral accelerations of types 1-3 a y The cluster centers corresponding to the weight coefficients gradually increase;

[0070] The cluster centers of the weight coefficients corresponding to the first to third types of velocities v gradually increase;

[0071] K 11 The longitudinal acceleration a of the first type of driving trajectory x Cluster centers corresponding to the weight coefficients; K 12 K represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the first type of driving trajectory; 13 The lateral acceleration a for the first type of driving trajectory y Cluster centers corresponding to the weight coefficients; K 14 The cluster center is the weight coefficient corresponding to the speed v of the first type of driving trajectory;

[0072] K 21 The longitudinal acceleration a of the second type of driving trajectory x Cluster centers corresponding to the weight coefficients; K 22 K represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the first type of driving trajectory; 23 The lateral acceleration a for the second type of driving trajectory y Cluster centers corresponding to the weight coefficients; K 24 The cluster centers are the weight coefficients corresponding to the speed v of the second type of driving trajectory;

[0073] K 31 The longitudinal acceleration a of the third type of driving trajectory x Cluster centers corresponding to the weight coefficients; K 32 K represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the third type of driving trajectory; 33 The lateral acceleration a for the third type of driving trajectory y Cluster centers corresponding to the weight coefficients; K 34 The cluster center is the weight coefficient corresponding to the speed v of the third type of driving trajectory;

[0074] As shown in Table 1

[0075] Table 1

[0076] Category / Cluster Center <![CDATA[ω1]]> <![CDATA[ω2]]> <![CDATA[ω3]]> <![CDATA[ω4]]> Category 1 <![CDATA[K 11 ]]> <![CDATA[K 12 ]]> <![CDATA[K 13 ]]> <![CDATA[K 14 ]]> Category 2 <![CDATA[K 21 ]]> <![CDATA[K 22 ]]> <![CDATA[K 23 ]]> <![CDATA[K 24 <!-- 5 -->]]> Category 3 <![CDATA[k 31 ]]> <![CDATA[K 32 ]]> <![CDATA[K 33 ]]> <![CDATA[K 34 ]]>

[0077] Different categories of driving trajectories are defined and assigned physical meaning according to the marking method;

[0078] The specific definitions when different driving trajectories are divided into 3 categories are as follows:

[0079] Category 1 driving trajectories are defined as follows: the driver corresponding to the driving trajectory is a cautious driver.

[0080] The second type of driving trajectory is defined as: the driver corresponding to the driving trajectory is a normal type of driver;

[0081] The third type of driving trajectory is defined as follows: the driver corresponding to the driving trajectory is an aggressive driver.

[0082] Drivers exhibiting the first type of driving trajectory place a high value on comfort and are highly attentive to longitudinal acceleration a during driving. x The costs of longitudinal acceleration (jerk) and longitudinal acceleration (vw) are given great importance, while the cost of losing the expected travel speed is also considered. T Low requirements;

[0083] For drivers on the third type of driving trajectory, in order to achieve the desired speed, the loss of desired speed (vw) was increased. T At the cost of reducing longitudinal acceleration a x The trade-off for comfort is the jerk's acceleration and longitudinal thrust.

[0084] Preferably, or alternatively, based on the k-means method, different driving trajectories are divided into 5 categories according to the weight coefficients corresponding to the characteristic performance of the driving trajectory; the 5 categories of driving trajectories respectively represent cautious drivers, moderately cautious drivers, normal drivers, moderately aggressive drivers, and aggressive drivers;

[0085] The cluster centers of its performance metrics are shown in Table 1:

[0086] M 11 The longitudinal acceleration a of the first type of driving trajectory x The cluster centers corresponding to the weight coefficients; M 12 M represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the first type of driving trajectory; 13 The lateral acceleration a for the first type of driving trajectory y The cluster centers corresponding to the weight coefficients; M 14 The cluster center is the weight coefficient corresponding to the speed v of the first type of driving trajectory;

[0087] M 21 The longitudinal acceleration a of the second type of driving trajectory x The cluster centers corresponding to the weight coefficients; M 22 M represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the second type of driving trajectory; 23 The lateral acceleration a for the second type of driving trajectory y The cluster centers corresponding to the weight coefficients; M 24 The cluster centers are the weight coefficients corresponding to the speed v of the second type of driving trajectory;

[0088] M 31 The longitudinal acceleration a of the third type of driving trajectory x The cluster centers corresponding to the weight coefficients; M 32 M represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the third type of driving trajectory; 33 The lateral acceleration a for the third type of driving trajectory y The cluster centers corresponding to the weight coefficients; M 34 The cluster center is the weight coefficient corresponding to the speed v of the third type of driving trajectory;

[0089] M 41 The longitudinal acceleration a of the fourth type of driving trajectory x The cluster centers corresponding to the weight coefficients; M 42 M represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the fourth type of driving trajectory; 43 The lateral acceleration a for the fourth type of driving trajectory y The cluster centers corresponding to the weight coefficients; M 44 The cluster center is the weight coefficient corresponding to the speed v of the fourth type of driving trajectory;

[0090] M 51 The longitudinal acceleration a of the fifth type of driving trajectory x The cluster centers corresponding to the weight coefficients; M 52 M represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the fifth type of driving trajectory; 53 The lateral acceleration a for the fifth type of driving trajectory y The cluster centers corresponding to the weight coefficients; M 54 The cluster center is the weight coefficient corresponding to the speed v of the fifth type of driving trajectory;

[0091] As shown in Table 2

[0092] Table 2

[0093] Category / Cluster Center <![CDATA[ω1]]> <![CDATA[ω2]]> <![CDATA[ω3]]> <![CDATA[ω4]]> Category 1 <![CDATA[M 11 ]]> <![CDATA[M 12 ]]> <![CDATA[M 13 ]]> <![CDATA[M 14 ]]> Category 2 <![CDATA[M 21 ]]> <![CDATA[M 22 ]]> <![CDATA[M 23 ]]> <![CDATA[M 24 ]]> Category 3 <![CDATA[M 31 ]]> <![CDATA[M 32 ]]> <![CDATA[M 33 ]]> <![CDATA[M 34 ]]> Category 4 <![CDATA[M 41 ]]> <![CDATA[M 42 ]]> <![CDATA[M 43 ]]> <![CDATA[M 44 ]]> Category 5 <![CDATA[M 51 ]]> <![CDATA[M 52 ]]> <![CDATA[M 53 ]]> <![CDATA[M 54 ]]>

[0094] The beneficial effect of adopting the above-mentioned further technical solution is that it realizes the weight coefficients ω1, ω2, ω3, and ω4 of the four performance indicators in the driving trajectory corresponding to the driver, and uses the k-means method to cluster multiple drivers; based on the magnitude of the obtained performance indicator weights, the driver's driving style is determined; the larger the weight of the performance indicator, the more important the driver is to that performance indicator; conversely, the smaller the weight, the less important the driver is to that performance indicator.

[0095] By calculating the Euclidean distance between the driver and the cluster center of each driver type, the driver can be classified into the driver type with the smallest Euclidean distance.

[0096] As can be seen, the driving behavior analysis method based on the weighted cost function of the present invention enables rapid and accurate classification of drivers, can provide friendly reminders to drivers regarding their driving behavior to avoid traffic accidents, and can also provide reliable reference for vehicle insurance institutions to adjust drivers' insurance amounts. Attached Figure Description

[0097] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be described below.

[0098] Figure 1 The weighting coefficients are the four performance indicators of the driving trajectories of the 16 drivers in Example 1. Detailed Implementation

[0099] To make the objectives, technical solutions, and advantages of the present invention clearer, the various aspects of the present invention will be described in detail below with reference to specific embodiments. However, these specific embodiments are only used to illustrate the present invention and do not constitute any limitation on the scope of protection and the substantive content of the present invention.

[0100] Example 1

[0101] This embodiment provides a driving behavior analysis method based on a weighted cost function. The method includes: collecting initial characteristic performance data of several driving trajectories; processing the initial characteristic performance data to obtain characteristic performance data, wherein the processing involves extracting the longitudinal acceleration 'a' from the initial characteristic performance data. x By taking the derivative, we can obtain the longitudinal acceleration jerk.

[0102] The characteristic performance number

[0103] Data including velocity v and longitudinal acceleration a x Lateral acceleration a y Longitudinal acceleration (jerk).

[0104] Each driving trajectory is divided into several decision cycle segments based on the duration of different driving events; the different driving events include constant speed behavior processes, acceleration behavior processes, and deceleration behavior processes;

[0105] The specific steps for dividing the driving trajectory into several decision cycle segments are as follows:

[0106] Calculate the radius of the driving trajectory; the method for calculating the radius of the driving trajectory is as follows: R is the radius of the driving trajectory, v is the speed, and a y This is lateral acceleration.

[0107] Based on longitudinal acceleration α x Construct the function y = a x Extract the function y = a x The complete waveform in the image is obtained by extracting the α portion of the complete waveform. x Peak or trough value;

[0108] Based on the driving trajectory radius and α x The peak or trough value of the wave divides the continuous driving trajectory into several decision cycle segments; the complete waveform is α. x The waveform increases from 0 to the peak value and then decreases back to 0; or, the complete waveform is α. x It decreases from 0 to a trough value and then increases back to 0;

[0109] The specific process of dividing the decision-making cycle into several segments is as follows:

[0110] Specific segmentation criteria are established to divide the driving trajectory into several decision-making cycle segments.

[0111] The specific segmentation criteria are as follows:

[0112] When the driving radius R ≥ 1000m, it is a straight driving trajectory; when the driving radius R < 1000m, it is a turning driving trajectory.

[0113] When the driving radius R ≥ 1000m, the longitudinal acceleration α x Peak wave length > 1.0 m / s 2 The continuous driving trajectory at that time is the first decision cycle segment;

[0114] When the driving radius R ≥ 1000m, -1.0m / s 2 ≤Longitudinal acceleration α x Peak or trough value ≤ 1.0 m / s 2 The continuous driving trajectory at that time is the second decision cycle segment;

[0115] When the driving radius R ≥ 1000m, the longitudinal acceleration α x Valley value < -1.0 m / s 2The continuous driving trajectory at that time is the third decision cycle segment;

[0116] When the driving radius R < 1000m, the longitudinal acceleration α x Peak wave length > 1.0 m / s 2 The continuous driving trajectory at that time is the fourth decision cycle segment;

[0117] When the driving radius R < 1000m, -1.0m / s 2 ≤Longitudinal acceleration α x Peak or trough value ≤ 1.0 m / s 2 The continuous driving trajectory at that time is the fifth decision cycle segment;

[0118] When the driving radius R < 1000m, the longitudinal acceleration α x Valley value < -1.0 m / s 2 The continuous driving trajectory at any given time is the sixth decision cycle segment; that is, the first decision cycle segment is the acceleration straight-line cycle segment, the second decision cycle segment is the constant speed straight-line cycle segment, the third decision cycle segment is the deceleration straight-line cycle segment; the fourth decision cycle segment is the acceleration turning cycle segment, the fifth decision cycle segment is the constant speed turning cycle segment, and the sixth decision cycle segment is the deceleration turning cycle segment; the start and end times of each event are recorded to obtain the event duration;

[0119] The system divides the driving data corresponding to each driver's driving trajectory into six basic events: accelerating straight, driving straight at a constant speed, driving straight at a deceleration, accelerating and turning, driving at a constant speed, and driving straight at a deceleration.

[0120] A cost function is constructed based on characteristic performance data to characterize the driver's emphasis on comfort, safety, and speed during driving; the cost function is:

[0121]

[0122] In the formula, T represents the duration of the decision-making cycle; ω j ω represents the weighting coefficients corresponding to the feature performance. j Including ω1, ω2, ω3, and ω4; ω1, ω2, ω3, and ω4 represent the longitudinal acceleration a. x Weighting coefficients, longitudinal acceleration (jerk), lateral acceleration (a) y The weighting coefficients of , velocity v, and ; v d v represents the driver's desired driving speed. d The value is the maximum speed limit of the road being driven on; the greater the weight of the characteristic performance, the more important the driver is to that performance indicator; conversely, the less important it is to the driver.

[0123] Based on the cost function and entropy weight method, the feature performance data corresponding to several decision period segments of each driving trajectory are processed to obtain the weight coefficients corresponding to the feature performance.

[0124] The specific process for obtaining the weight coefficients corresponding to the aforementioned feature performance is as follows:

[0125] Extract the feature performance data corresponding to several decision cycle segments of the driving trajectory;

[0126] The characteristic performance data corresponding to the decision cycle segment are converted into variable data according to the first transformation formula; the variable data includes the defined property velocity vw. T Definition of longitudinal acceleration a x w T 1. Define the property of lateral acceleration a y w T Definition of properties: longitudinal acceleration (jerkw) T The first conversion formula includes:

[0127] The variable data corresponding to several decision-making period segments of the driving trajectory are processed based on the entropy weight method to obtain the weight coefficients corresponding to the variable data; the weight coefficients corresponding to the feature performance are obtained through the weight coefficients corresponding to the variable data.

[0128] The process of processing the variable data corresponding to several decision-making period segments of the driving trajectory based on the entropy weight method is as follows: the variable data corresponding to several decision-making period segments are standardized using the second formula.

[0129] The second formula is Where, x ij x' represents the j-th variable data for the i-th decision period segment of the driving trajectory. ij For x ij Normalized data;

[0130] Using the third formula to evaluate x' ij Processing to obtain p ij p ij The third formula represents the proportion of the j-th variable data in the i-th decision period segment of the driving trajectory to the j-th variable data in all decision period segments of the driving trajectory; i = 1, ..., n, j = 1, ..., m;

[0131] Using the fourth formula to apply p ij Processing yields e j The e j Let be the entropy value of the j-th variable in the driving trajectory;

[0132] The fourth formula is: Where, k = 1 / ln(n);

[0133] Using the fifth formula to evaluate e j Processing yields d j The d j The information entropy redundancy of the j-th variable in the driving trajectory;

[0134] The fifth formula is: d j =1-e j j = 1, ..., m;

[0135] Using the sixth formula to analyze d j Processing to obtain ω j The ω j The weighting coefficients for the characteristic performance corresponding to the j-th variable of the driving trajectory; the sixth formula is:

[0136] The weighting coefficients ω1, ω2, ω3, and ω4 of the four performance indicators for each driver's driving trajectory are obtained.

[0137] Clustering of different driving trajectories is performed based on the k-means method and the weight coefficients corresponding to the feature performance of different driving trajectories.

[0138] The specific process of classifying different driving trajectories is as follows:

[0139] Based on the k-means method, different driving trajectories are divided into three categories according to the weight coefficients corresponding to the feature performance of the driving trajectory;

[0140] The longitudinal acceleration a of the first to third categories x The cluster centers of the corresponding weight coefficients gradually decrease;

[0141] The cluster centers of the weight coefficients corresponding to the longitudinal acceleration jerk in categories 1-3 gradually decrease;

[0142] Lateral accelerations of types 1-3 a y The cluster centers corresponding to the weight coefficients gradually increase;

[0143] The cluster centers of the weight coefficients corresponding to the first to third types of velocities v gradually increase;

[0144] K 11 The longitudinal acceleration a of the first type of driving trajectory x Cluster centers corresponding to the weight coefficients; K 12 K represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the first type of driving trajectory; 13 The lateral acceleration a for the first type of driving trajectory yCluster centers corresponding to the weight coefficients; K 14 The cluster center is the weight coefficient corresponding to the speed v of the first type of driving trajectory;

[0145] K 21 The longitudinal acceleration a of the second type of driving trajectory x Cluster centers corresponding to the weight coefficients; K 22 K represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the first type of driving trajectory; 23 The lateral acceleration a for the second type of driving trajectory y Cluster centers corresponding to the weight coefficients; K 24 The cluster centers are the weight coefficients corresponding to the speed v of the second type of driving trajectory;

[0146] K 31 The longitudinal acceleration a of the third type of driving trajectory x Cluster centers corresponding to the weight coefficients; K 32 K represents the cluster center of the weight coefficients corresponding to the longitudinal acceleration jerk of the third type of driving trajectory; 33 The lateral acceleration a for the third type of driving trajectory y Cluster centers corresponding to the weight coefficients; K 34 The cluster center is the weight coefficient corresponding to the speed v of the third type of driving trajectory;

[0147] For example, driving data of the driving trajectories of 16 drivers can be collected, and the weighted coefficients of four performance characteristics corresponding to the driving trajectories of the 16 drivers can be obtained through the above method, such as... Figure 1 As shown in Table 1, the weight coefficients of the four performance characteristics were used to classify the 16 drivers into three driving styles using k-means.

[0148] Table 1

[0149] Category / Cluster Center <![CDATA[ω1]]> <![CDATA[ω2]]> <![CDATA[ω3]]> <![CDATA[ω4]]> Category 1 0.30 0.40 0.18 0.12 Category 2 0.29 0.34 0.20 0.16 Category 3 0.28 0.24 0.29 0.19

[0150] Different categories of driving trajectories are defined and assigned physical meaning according to the marking method;

[0151] The specific definitions when different driving trajectories are divided into 3 categories are as follows:

[0152] Category 1 driving trajectories are defined as follows: the driver corresponding to the driving trajectory is a cautious driver.

[0153] The second type of driving trajectory is defined as: the driver corresponding to the driving trajectory is a normal type of driver;

[0154] The third type of driving trajectory is defined as follows: the driver corresponding to the driving trajectory is an aggressive driver.

[0155] Drivers exhibiting the first type of driving trajectory place a high value on comfort and are highly attentive to longitudinal acceleration a during driving. x The costs of longitudinal acceleration (jerk) and longitudinal acceleration (vw) are given great importance, while the cost of losing the expected travel speed is also considered. T Low requirements;

[0156] For drivers on the third type of driving trajectory, in order to achieve the desired speed, the loss of desired speed (vw) was increased. T At the cost of reducing longitudinal acceleration a x The trade-off between comfort and longitudinal acceleration (jerk).

[0157] Based on the k-means method, different driving trajectories are divided into 5 categories according to the weight coefficients corresponding to the characteristic performance of the driving trajectory; the 5 categories of driving trajectories respectively represent cautious drivers, moderately cautious drivers, normal drivers, moderately aggressive drivers and aggressive drivers;

[0158] If, based on the k-means method, different driving trajectories are divided into 5 categories according to the weight coefficients corresponding to the characteristic performance of the driving trajectory; the 5 categories of driving trajectories respectively represent cautious drivers, moderately cautious drivers, normal drivers, moderately aggressive drivers, and aggressive drivers;

[0159] If we use k-means to divide the driving data of 16 drivers into 5 categories, the results are shown in Table 2:

[0160] Table 2

[0161] Category / Cluster Center <![CDATA[ω1]]> <![CDATA[ω2]]> <![CDATA[ω3]]> <![CDATA[ω4]]> Category 1 0.30 0.40 0.17 0.14 Category 2 0.31 0.36 0.21 0.13 Category 3 0.29 0.31 0.23 0.17 Category 4 0.30 0.23 0.25 0.23 Category 5 0.27 0.24 0.30 0.18

[0162] Based on the weight coefficients corresponding to the characteristic performance of other drivers' driving trajectories and the Euclidean distance between the cluster centers of different categories of driving trajectories after clustering, the category to which the driver's driving trajectory belongs is determined; the category corresponding to the cluster center with the smallest Euclidean distance is the category to which the driver's driving trajectory belongs.

[0163] Example 2

[0164] The similarities between this embodiment and Embodiment 1 will not be repeated here. This embodiment provides a driving behavior analysis method based on a weighted cost function, which further includes preprocessing the initial characteristic performance data before processing it; applying a moving average filter to the initial velocity v; and applying a moving average filter to the initial longitudinal acceleration a. x and lateral acceleration a y Wavelet filtering is performed, with the wavelet basis function chosen as sym8 and the decomposition level set to 7.

[0165] The method for calculating the radius of the driving trajectory is as follows: R is the radius of the driving trajectory, v is the speed, and a y The lateral acceleration is given by ε1 = 0.001.

[0166] The cost function is:

[0167]

[0168] In the formula, T represents the duration of the decision-making cycle; ω j ω represents the weighting coefficients corresponding to the feature performance. j Including ω1, ω2, ω3, and ω4; ω1, ω2, ω3, and ω4 represent the longitudinal acceleration a. x Weighting coefficients, longitudinal acceleration (jerk), lateral acceleration (a) y The weighting coefficients of , velocity v, and are given; ε1 is 0.001.

[0169] Vt is the reciprocal of the sampling frequency of the original data. d For the driver's desired driving speed, v is the speed during the first decision cycle. d For a x The speed decreased from the peak to 0.1 m / s 2 The average velocity thereafter; for the second decision period, v d v is the average velocity during the second decision-making period; for the third decision-making period, v d For a x Descending to -0.1 m / s 2 The average velocity before; for the fourth decision period, v d v is the average speed after exiting the curve; for the fifth decision period, v d The average velocity for the fifth decision cycle; for the sixth decision cycle, v d The average speed before entering the curve.

[0170] The first conversion formula includes The value of ε1 is 0.001.

[0171] The present invention has been described above with reference to specific embodiments. These specific embodiments are merely exemplary and should not be construed as limiting the scope of protection of the present invention. Those skilled in the art can make various modifications, changes, or substitutions without departing from the essence of the present invention. Therefore, various equivalent variations made according to the present invention still fall within the scope of the present invention.

Claims

1. A driving behavior analysis method based on a weighted cost function, characterized in that, The method includes: Collect initial characteristic performance data for several driving trajectories; The initial characteristic performance data is processed to obtain the characteristic performance data; Each driving trajectory is divided into several decision cycle segments based on the duration of different driving events; A cost function is constructed based on characteristic performance data to characterize the driver's emphasis on comfort, safety, and speed during driving. The cost function is: In the formula, T represents the duration of the decision-making cycle. ω j These are the weighting coefficients corresponding to the feature performance. ω j Including ω 1 , ω 2 , ω 3 , ω 4 ; ω 1 , ω 2 , ω 3 , ω 4 Longitudinal acceleration Weighting coefficients, longitudinal acceleration Weighting coefficients, lateral acceleration Weighting coefficients, speed Weighting coefficients; For the driver's desired driving speed, The value is the maximum speed limit of the road being traveled. or The cost function is: In the formula, T represents the duration of the decision-making cycle. ω j These are the weighting coefficients corresponding to the feature performance. ω j Including ω 1 , ω 2 , ω 3 , ω 4 ; ω 1 , ω 2 , ω 3 , ω 4 Longitudinal acceleration Weighting coefficients, longitudinal acceleration Weighting coefficients, lateral acceleration Weighting coefficients, speed Weighting coefficients; It can be any value between 0 and 0.0015; It is the reciprocal of the sampling frequency of the original data; For the driver's desired driving speed, in the first decision cycle segment, v d for a x The speed decreased from the peak to 0.1 m / s 2 The subsequent average speed; for the second decision period, v d This represents the average speed during the second decision-making cycle; for the third decision-making cycle, v d for a x Decreased to -0.1 m / s 2 The average speed before; for the fourth decision period, v d The average speed after exiting the curve; for the fifth decision cycle segment, v d This represents the average speed during the fifth decision-making cycle; for the sixth decision-making cycle, v d The average speed before entering the curve; Based on the cost function and entropy weight method, the feature performance data corresponding to several decision period segments of each driving trajectory are processed to obtain the weight coefficients corresponding to the feature performance. Clustering of different driving trajectories is performed based on the k-means method and the weight coefficients corresponding to the feature performance of different driving trajectories. The characteristic performance data includes speed. Longitudinal acceleration Lateral acceleration Longitudinal acceleration .

2. The driving behavior analysis method based on a weighted cost function according to claim 1, characterized in that, Also package This includes preprocessing the initial characteristic performance data before processing it; applying a moving average filter to the initial velocity v; and processing the initial longitudinal acceleration... and lateral acceleration Perform wavelet filtering, selecting sym8 as the wavelet basis function, and decomposition level greater than or equal to 5.

3. The driving behavior analysis method based on a weighted cost function according to claim 1 or 2, characterized in that, The specific steps for dividing the driving trajectory into several decision cycle segments are as follows: Calculate the radius of the driving trajectory; Based on longitudinal acceleration a x ， Construct function y= a x Extract the function y= a x The complete waveform in the image is obtained. a x Peak or trough value; Based on the driving trajectory radius and a x The peak or trough value divides the continuous driving trajectory into several decision cycle segments.

4. The driving behavior analysis method based on a weighted cost function according to claim 3, characterized in that, The method for calculating the radius of the driving trajectory is as follows: ; R The radius of the driving trajectory. For speed, It is lateral acceleration; or, The method for calculating the radius of the driving trajectory is as follows: ; R The radius of the driving trajectory. For speed, It is lateral acceleration. It can be any value from 0 to 0.0015.

5. The driving behavior analysis method based on a weighted cost function according to claim 4, characterized in that, The complete waveform is a x The waveform increases from 0 to the peak value and then decreases back to 0; or, the complete waveform is... a x It decreases from 0 to a trough value and then increases back to 0; and / or The specific process of dividing the decision-making cycle into several segments is as follows: Specific segmentation criteria are established to divide the driving trajectory into several decision-making cycle segments. The specific segmentation criteria are as follows: When the driving radius R ≥ 1000m, it is a straight driving trajectory; when the driving radius R < 1000m, it is a turning trajectory. Trajectory; When the driving radius R ≥ 1000m, the longitudinal acceleration a x Peak wave length > 1.0 m / s 2 The continuous driving trajectory at that time is the first decision Policy cycle period; When the driving radius R ≥ 1000m, -1.0m / s 2 ≤Longitudinal acceleration a x Peak or trough value ≤ 1.0 m / s 2 The continuous driving trajectory at that time is the second decision cycle segment; When the driving radius R ≥ 1000m, the longitudinal acceleration a x Valley value < -1.0 m / s 2 The continuous driving trajectory at that time is the third decision cycle segment; When the driving radius R < 1000m, the longitudinal acceleration a x Peak wave length > 1.0 m / s 2 The continuous driving trajectory at that time is the fourth decision Policy cycle period; When the driving radius R < 1000m, -1.0m / s 2 ≤Longitudinal acceleration a x Peak or trough value ≤ 1.0 m / s 2 The continuous driving trajectory at that time is the fifth decision cycle segment; When the driving radius R < 1000m, the longitudinal acceleration a x Valley value < -1.0 m / s 2 The continuous driving trajectory at that time is the sixth decision cycle segment.

6. The driving behavior analysis method based on a weighted cost function according to claim 5, characterized in that, get The specific process for the weighting coefficients corresponding to the feature performance is as follows: Extract the feature performance data corresponding to several decision cycle segments of the driving trajectory; The characteristic performance data corresponding to the decision cycle segment are converted into variable data according to the first transformation formula; the variable data includes the defined property velocity. Define the property of longitudinal acceleration Define the property of lateral acceleration Definition of longitudinal acceleration ; The variable data corresponding to several decision-making period segments of the driving trajectory are processed using the entropy weight method to obtain the weight coefficients corresponding to the variable data; the weight coefficients corresponding to the feature performance are obtained through the weight coefficients corresponding to the variable data.

7. The driving behavior analysis method based on a weighted cost function according to claim 6, characterized in that, The first conversion formula includes , , , ; or, The first conversion formula includes , , , ; It can be any value between 0 and 0.0015.

8. The driving behavior analysis method based on a weighted cost function according to claim 7, characterized in that, The process of processing the variable data corresponding to several decision-making period segments of the driving trajectory based on the entropy weight method is as follows: The second formula is used to standardize the variable data corresponding to several decision-making cycle segments. The second formula is in, For the j-th variable data of the i-th decision period segment of the driving trajectory; for Normalized data; Using the third formula Processed p ij , p ij The proportion of the j-th variable data in the i-th decision period segment of the driving trajectory to the j-th variable data in all decision period segments of the driving trajectory; The third formula is: , i=1, …, n, j=1, …, m; Using the fourth formula p ij Processing yields e j The e j Let be the entropy value of the j-th variable in the driving trajectory; The fourth formula is: , j=1, …, m ; where k=1 / ln(n); Using the fifth formula to evaluate e j Processed d j The d j The information entropy redundancy of the j-th variable in the driving trajectory; The fifth formula is: ,j=1, …, m; Through the sixth formula d j Processed ω j The ω j The weight coefficients of the characteristic performance corresponding to the j-th variable of the driving trajectory; The sixth formula is: ,j=1, …, m 。 9. The driving behavior analysis method based on a weighted cost function according to claim 8, characterized in that, right The specific process for classifying different driving trajectories is as follows: Based on the k-means method, different driving trajectories are divided into three categories according to the weight coefficients corresponding to the feature performance of the driving trajectory; Longitudinal acceleration of types 1-3 The cluster centers of the corresponding weight coefficients gradually decrease; Longitudinal acceleration of types 1-3 The cluster centers of the corresponding weight coefficients gradually decrease; Lateral accelerations of types 1-3 The cluster centers corresponding to the weight coefficients gradually increase; Types 1-3 speeds The cluster centers corresponding to the weight coefficients gradually increase; Different categories of driving trajectories are defined and assigned physical meaning according to the marking method; When different driving trajectories are divided into 3 categories, the specific definitions are as follows: Category 1 driving trajectories are defined as follows: the driver corresponding to the driving trajectory is a cautious driver. The second type of driving trajectory is defined as: the driver corresponding to the driving trajectory is a normal type of driver; The third type of driving trajectory is defined as: the driver corresponding to the driving trajectory is an aggressive driver; Drivers exhibiting the first type of driving trajectory place a high value on comfort and are highly attentive to longitudinal acceleration during driving. and longitudinal acceleration Both costs are given great importance, including the cost of losing the expected speed. Low requirements; In order to achieve the desired speed, drivers in the third type of driving trajectory increased the cost of losing that desired speed. However, the longitudinal acceleration was reduced. and longitudinal acceleration The cost of two kinds of comfort.

Images