CN110858312A - Driver driving style classification method based on fuzzy C-means clustering algorithm - Google Patents

Abstract

The invention discloses a driver's driving style classification method based on a fuzzy C-means clustering algorithm. First, the real-time driving data of the driver is collected through a driving simulator, and the experimental data is preliminarily screened to eliminate invalid data; and then based on the fuzzy C-means Clustering algorithm to perform preliminary clustering: take the reaction time and the following time distance as input variables, and output the preliminary driving style; finally, on the basis of the preliminary clustering, the standard deviation of the transient index speed fluctuation is introduced for reclustering, and the response Time, following speed, and standard deviation of speed fluctuations are used as input variables to output the re-clustered driving style. The feature indicators used in this method have significant differences, the classification algorithm is simple and the classification effect is good, and the driving styles of different drivers are considered comprehensively, which can cover the driving styles of most drivers.

Description

Classification method of driver's driving style based on fuzzy C-means clustering algorithm

technical field

The invention relates to the technical field of driving style classification, in particular to a driver's driving style classification method based on a fuzzy C-means clustering algorithm.

Background technique

There are obvious differences between drivers of different styles in overtaking and lane changing frequency, rapid acceleration/deceleration, speed fluctuation, steering wheel angular velocity fluctuation, close-range car-following and other behaviors, which affect the frequency of dangerous driving behaviors. Aggressive, the greater the probability of a traffic accident, the frequency of rear-end collisions or conflicts among aggressive drivers is more than 4 times that of normal drivers. More research results have verified that differences in drivers’ driving styles are significantly correlated with driving risk. .

In recent years, with the rapid development of intelligent driving vehicles, the research on driver characteristics has broad application prospects in the personalized customization of human-machine collaborative driving. On the basis of realizing the basic functions of the assisted driving system, drivers of different styles have great differences due to the timing and necessity of the early warning of driving assistance and the connection of the controller. In order to improve the intelligence of human-machine co-driving and adapt to the individual needs of different drivers, the driving style of the driver can be studied through driving data, and the system can be adjusted according to the type of driver to better adapt to the driver's driving style. Driving style, improve the intelligence level of smart cars and the driving comfort of human-machine driving.

Cluster analysis is a kind of multivariate statistical analysis. The basic idea is that there are different degrees of affinity between variables, find some statistics that can measure this affinity, and use it as a basis for classification to classify variables. Fuzzy set and fuzzy reasoning theory have been regarded as effective methods to deal with complex problems such as uncertain rules since their birth. Fuzzy clustering establishes the uncertainty description of samples to categories, which can better reflect the real world, and has become the mainstream of cluster analysis research.

The classification of driving style in the prior art also has the following disadvantages:

First, the driving style is evaluated by means of self-filled questionnaires and expert scores, which are too subjective and subject to great interference from external factors. The validity of the questionnaires and expert experience are high, and the results are not accurate enough.

Second, the amount of data extracted based on the real vehicle driving experiment is huge and complex, the post-processing is difficult and dangerous, and the external environment has a great influence on the experimental results.

Third, the classification of driving styles is relatively simple and cannot cover most drivers.

Fourth, the evaluation index for driving style classification is single, and the classification results are not convincing enough.

SUMMARY OF THE INVENTION

In view of the deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a driving style classification method based on a fuzzy C-means clustering algorithm. The clustering algorithm classifies the driver's driving style.

In order to achieve the above object, the present invention is realized through the following technical solutions: a driver's driving style classification method based on a fuzzy C-means clustering algorithm, comprising the following steps:

S1. Collect experimental data through a driving simulator, and conduct preliminary screening of the experimental data to eliminate invalid data;

S2. Based on the fuzzy C-mean clustering algorithm, perform preliminary clustering: take the reaction time and the following time distance as input variables, and output the preliminary driving style;

S3. On the basis of the initial clustering, the transient index speed fluctuation standard deviation is introduced for re-clustering, and the reaction time, the following speed and the speed fluctuation standard deviation are used as input variables, and the re-clustered driving style is output.

Further, before collecting experimental data through the driving simulator, basic information is collected on the tested drivers, including: age, gender, driving age, driving type, whether there is a traffic accident, and self-assessment of driving style.

Further, in S2, the reaction time is the time elapsed from the moment when the dangerous stimulus begins to occur until the driver just depresses the brake pedal.

Further, in S2, the following time distance is: following time distance=average inter-vehicle distance/average vehicle speed.

Further, in S2, the preliminary driving styles are divided into proficient driving styles, conservative driving styles, and aggressive driving styles; in S3, the re-clustered driving styles include: conservative driving style, dull driving style, aggressive driving style, and steady driving style , risky driving style and skilled driving style.

Further, the conservative driving style is expressed as: the reaction time RT is long, the following time distance is large, the state of long-distance weak following is maintained, the speed fluctuates greatly, the speed of the preceding vehicle is sensitive, and the dangerous state is relatively sensitive. in a safer driving environment;

Slow driving style is expressed as: long reaction time and insensitivity to changes in the speed of the preceding vehicle;

Aggressive driving style is expressed as follows: the following time distance is small, the vehicle tends to follow closely, the reaction time is short, the vehicle speed fluctuates greatly, it is prone to sudden acceleration/deceleration, and the vehicle speed is fast;

Steady driving style is expressed as: the following distance is moderate, the reaction time is moderate, and the speed remains stable;

The risky driving style is expressed as: the following time distance is small, the reaction time is long and the speed fluctuation is large;

Skilled driving style is expressed as: short reaction times, steady speed, and a tendency to follow closely.

Beneficial effects of the present invention:

The present invention is a method for classifying the driving style of drivers based on the fuzzy C-means clustering algorithm. First, the experimental data is collected through a driving simulator, and the experimental data is preliminarily screened to obtain the experimental data; then, based on the fuzzy C-means clustering algorithm, the Preliminary clustering: take the reaction time and the following time distance as input variables, and output the preliminary driving style; finally, on the basis of the preliminary clustering, the standard deviation of the transient index speed fluctuation is introduced for re-clustering, and the reaction time and the following speed are re-clustered. and speed fluctuation standard deviation as input variables, the output is again clustered driving style. The feature indicators adopted by this method have significant differences, the classification algorithm is simple and the classification effect is good. The invention provides a new idea and a new method for analyzing the driving style of the driver, and has a good application prospect in future smart cars.

Description of drawings

In order to illustrate the specific embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that are required to be used in the description of the specific embodiments or the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. In the drawings, each element or section is not necessarily drawn to actual scale.

FIG. 1 is a schematic flowchart of a method for classifying the driving style of a driver based on a fuzzy C-means clustering algorithm according to the present invention.

Detailed ways

Embodiments of the technical solutions of the present invention will be described in detail below with reference to the accompanying drawings. The following examples are only used to more clearly illustrate the technical solutions of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

It should be noted that, unless otherwise specified, the technical or scientific terms used in this application should have the usual meanings understood by those skilled in the art to which the present invention belongs.

As shown in Figure 1, the present invention provides a technical solution: a driver's driving style classification method based on a fuzzy C-means clustering algorithm, comprising the following steps:

The first step is to collect experimental data through the driving simulator, and conduct preliminary screening of the experimental data to eliminate invalid data and obtain valid experimental data;

Before collecting experimental data through the driving simulator, collect basic information on the test drivers, including: age, gender, driving experience, driving type, whether there is a traffic accident, and self-assessment of driving style;

The driving simulator is a driving simulator designed and developed based on the Unreal 4 (UE4) engine, combined with the CarSim vehicle dynamics simulation model and the Logitech G29 force feedback steering wheel pedal set;

Collecting experimental data through the driving simulator is as follows: complete the driving scenario simulation test of car start-stop, stable following, approaching, fading away, and emergency brake light in the driving simulator, and store the experimental data in the UE4 software, and finally 34 One test driver completed the experiment;

Removing invalid data is specifically: importing the experimental data of 34 test drivers into IBM SPSS software, performing integrity analysis, filtering processing, and normality inspection on the original experimental data, and then obtaining valid experimental data, which is obtained in this embodiment. 154 samples available.

The second step is to use Matlab software to write the program code of the fuzzy C-means clustering algorithm. Based on the fuzzy C-means clustering algorithm, preliminary clustering is carried out: the reaction time and the following time distance are used as input variables to output the preliminary driving style and preliminary driving. Styles include: conservative driving style, skilled driving style and aggressive driving style;

Among them, the reaction time RT is defined as: the time from the moment when the dangerous stimulus (such as the front car brake light is on) starts to the time the driver just depresses the brake pedal; the reaction time RT can characterize the driver's risk perception and Response capability;

The following time gap TH is defined as: following time gap=average inter-vehicle distance/average vehicle speed. It can effectively describe the tightness of car following.

The fuzzy C-means clustering algorithm is a classification algorithm, and the specific steps are as follows:

First, initialize the membership matrix U with a random number between 0 and 1, so that it satisfies the constraints. The constraints are:

where u _ij is between 0 and 1.

Then, use the cluster center formula to calculate c cluster centers, and the cluster center formula is;

c _i is the cluster center of fuzzy group I, and m∈[1,∞) is a weighting index.

Again, the objective function is calculated according to the objective function formula. The algorithm stops if it is less than a certain threshold, or if it is less than a certain threshold for the variable which is equivalent to the last value of the value function. The objective function formula is:

d _ij =||c _i -x _j || is the Euclidean distance between the ith cluster center and the jth data point.

Finally, use the optimal solution formula with the new U matrix to return to the cluster center to obtain the optimal solution. The optimal solution formula is:

The third step: On the basis of the initial clustering, the transient index speed fluctuation standard deviation is introduced for re-clustering, the reaction time, the following speed and the speed fluctuation standard deviation are used as input variables, and the driving style that is clustered again is output. The clustered driving styles include: conservative driving style, dull driving style, aggressive driving style, steady driving style, risky driving style, and skilled driving style.

Among them, the standard deviation SDS of the transient index speed fluctuation reflects the driver's ability to control the speed of the vehicle, and can intuitively reflect the driver's driving level. Smaller speed fluctuation standard deviation SDS: less than 20km/h, indicating that the driver has strong ability to control the speed of the vehicle and skilled driving skills; large speed fluctuation standard deviation SDS: 20km/h or more, indicating that the driver has abrupt acceleration/deceleration behavior It is more sensitive to changes in the speed of the preceding vehicle, and is prone to aggressive driving behavior. The specific algorithm of re-clustering adopts the above-mentioned fuzzy C-means clustering algorithm.

Conservative driving style is expressed as: long reaction time RT, large following time distance, maintain a long-distance weak following state, large speed fluctuation, sensitive to the speed change of the preceding vehicle, more sensitive to dangerous state, moderate and relatively safe The survey found that the average driving experience of such drivers is relatively short, and the level of driving skills is relatively low;

Obtuse driving style is expressed as: long reaction time, insensitivity to changes in the speed of the preceding vehicle, and poor perception of driving risks;

Aggressive driving style is expressed as follows: the following time distance is small, the vehicle tends to follow closely, the reaction time is short, the vehicle speed fluctuates greatly, it is prone to sudden acceleration/deceleration, and the vehicle speed is fast;

Steady driving style is expressed as: moderate following distance, moderate reaction time, stable speed, solid driving skills, and a tendency to drive safely;

The risky driving style is expressed as: the following time is small, the reaction time is long, and the speed fluctuates greatly, which has a high driving risk;

Skilled driving style is expressed as: short reaction time, stable speed, and tendency to closely follow the car. Combined with the driver information collection table, it is found that such drivers have the longest average driving experience and outstanding driving skills.

In the above driving styles, the short reaction time means that the reaction time is 3s or less; the long reaction time is that the reaction time is greater than 3s; the following time distance is small, the following time distance is within 3m, and the moderate following time distance is 3m- 5m, and the following distance is greater than 5m.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions described in the foregoing embodiments can still be modified, or some or all of the technical features thereof can be equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present invention. The scope of the invention should be included in the scope of the claims and description of the present invention.

Claims (6)

1. A driver driving style classification method based on a fuzzy C-means clustering algorithm is characterized by comprising the following steps:

s1, acquiring experimental data through a driving simulator, primarily screening the experimental data, and removing invalid data;

s2, performing preliminary clustering based on the fuzzy C-means clustering algorithm: taking the reaction time and the following distance as input variables, and outputting a preliminary driving style;

and S3, introducing the speed fluctuation standard deviation of the transient index for reclustering on the basis of the primary clustering, taking the reaction time, the following speed and the speed fluctuation standard deviation as input variables, and outputting the driving style of the secondary clustering.

2. The method for classifying the driving style of the driver based on the fuzzy C-means clustering algorithm as claimed in claim 1, wherein before experimental data is collected by a driving simulator, basic information collection is performed on the tested driver, and the method comprises the following steps: age, sex, driving age, driving type, whether traffic accidents occur or not and driving style are self-rated.

3. The method for classifying the driving style of the driver based on the fuzzy C-means clustering algorithm as claimed in claim 1, wherein in S2, the reaction time is: the time elapsed since the driver just stepped on the brake pedal is counted from the time when the dangerous stimulus starts to occur.

4. The method for classifying the driving style of the driver based on the fuzzy C-means clustering algorithm as claimed in claim 3, wherein in S2, the following vehicle distance is: the following distance is the average car-to-car distance/average car speed.

5. The method for classifying the driving styles of drivers based on the fuzzy C-means clustering algorithm as claimed in claim 4, wherein in S2, the preliminary driving styles are classified into a skilled driving style, a conservative driving style and an aggressive driving style; the driving styles re-clustered in S3 include: conservative driving style, sluggish driving style, aggressive driving style, robust driving style, risky driving style, and proficient driving style.

6. The method for classifying the driving style of the driver based on the fuzzy C-means clustering algorithm as claimed in claim 5, wherein the conservative driving style is expressed as: the method has the advantages that the reaction time RT is long, the following vehicle distance is large, the long-distance weak following state is kept, the speed fluctuation is large, the method is sensitive to the speed change of a front vehicle and the dangerous state, and the method is moderately kept in a safer driving environment;

the dull driving style is expressed as: the reaction time is long and is insensitive to the speed change of the front vehicle;

the aggressive driving style is represented as: the following distance is small, the vehicle tends to follow the vehicle tightly, the reaction time is short, the vehicle speed fluctuation is large, the rapid acceleration/deceleration condition is easy to occur, and the vehicle speed is fast;

the robust driving style is represented as: the following vehicle distance is moderate, the reaction time is moderate, and the vehicle speed is kept stable;

the risky driving style is expressed as: the following vehicle distance is small, the reaction time is long and the vehicle speed fluctuation is large;

the skilled driving style is expressed as: the reaction time is short, the vehicle speed is stable, and the vehicle tends to closely follow the vehicle.

Images