CN114707573A - Unsupervised driving style analysis method based on basic driving operation event - Google Patents
Unsupervised driving style analysis method based on basic driving operation event Download PDFInfo
- Publication number
- CN114707573A CN114707573A CN202210176012.4A CN202210176012A CN114707573A CN 114707573 A CN114707573 A CN 114707573A CN 202210176012 A CN202210176012 A CN 202210176012A CN 114707573 A CN114707573 A CN 114707573A
- Authority
- CN
- China
- Prior art keywords
- event
- curve
- intensity
- time window
- driving
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides an unsupervised driving style analysis method based on basic driving operation events, which comprises the following steps: acquiring and preprocessing data; extracting a basic driving operation event; performing feature construction and extraction on each event in the basic driving operation events to obtain event intensity features; event intensity clustering is carried out through k-means, and event intensity category labels are marked; acquiring a dynamic time window, and constructing an event time variation curve which represents the driving style in the dynamic time window and is provided with an event intensity category label; and (3) clustering curves of all dynamic time windows based on a curve clustering algorithm fused with DTW to obtain various time window curves and marking driving style type labels. The invention takes the basic driving operation event as a basic unit, considers the event intensity and the event transfer characteristic, takes the change curve of the event along with the time as the characteristic for describing the driving style, embodies the dynamic decision information, the data continuity and the time characteristic of the driving behavior, retains the original information of the data and improves the analysis accuracy of the driving style.
Description
Technical Field
The invention relates to the technical field of traffic analysis, in particular to an unsupervised driving style analysis method based on basic driving operation events.
Background
With the development of social science and technology, driving style analysis has become a research hotspot. The driving style has important functions in the aspects of road safety, vehicle economy, vehicle insurance, intelligent vehicle design and the like: the driving style is detected and fed back to the driver in real time, so that the occurrence of road traffic accidents can be effectively reduced; the driving style greatly influences the fuel economy of the vehicle, and the more aggressive the driving behavior, the lower the fuel economy; the insurance cost of the vehicle also depends on the driving style of the vehicle user, and the interest of the insurance company can be maximized by setting different insurance costs for drivers with different driving styles; with the development of artificial intelligence, intelligent vehicles become a new trend for development of vehicle related industries, the decision making process of the intelligent vehicles depends on analysis of driving behaviors to a great extent, and effective analysis of driving styles is a key technology for ensuring decision making safety of the intelligent vehicles and improving acceptance of users. Therefore, accurate and efficient implementation of driving style recognition is a major concern at present.
In the background of big data, in the face of massive natural driving data, 800 working hours are needed for calibrating the natural driving data every hour, so that it is necessary to perform cluster analysis on the driving style by adopting an unsupervised method. The clustering analysis is an unsupervised machine learning algorithm used for data mining, can search the internal distribution structure of data, and can gather the data with similar characteristics into one class, and the data with different types have obvious difference, so that the similarity of the similar data and the difference between the data with different types are enlarged, and the interpretability is improved, and the follow-up research and analysis are facilitated.
However, the following disadvantages still exist in the existing driving style analysis: (1) a driving style analysis method which starts from a basic driving operation event and considers the intensity of the event and the transfer characteristic of the event at the same time is lacked; (2) in the prior art, the driving style is represented by mostly utilizing discrete characteristic points, the continuity of natural driving data cannot be reserved, and the time characteristic is omitted; (3) the existing driving style clustering algorithms are all clustering algorithms aiming at discrete points, and lack of clustering algorithms aiming at continuous curves with different lengths, so that efficient and accurate clustering analysis of driving styles of data samples with different lengths based on continuous curve characteristics cannot be realized.
Disclosure of Invention
The invention provides an unsupervised driving style analysis method based on a basic driving operation event, which aims to solve the problems that the conventional driving style analysis method does not start from the basic driving operation event and does not consider the event intensity and the event transfer characteristic; the driving style is represented by utilizing discrete characteristic points, the continuity of natural driving data cannot be reserved, and the time characteristic is omitted; the problem that the efficient and accurate cluster analysis of the driving style of the data samples with different lengths based on the characteristics of the continuous curve cannot be realized due to the lack of the clustering algorithm aiming at the continuous curves with different lengths is solved.
The invention provides an unsupervised driving style analysis method based on basic driving operation events, which comprises the following steps of: s1, natural driving data are obtained, and the natural driving data are preprocessed to obtain effective natural driving data; s2, extracting basic driving operation events based on the effective natural driving data; s3, based on the effective natural driving data, performing feature construction and feature extraction on each event in the basic driving operation events through driving behavior features to obtain event intensity features representing the intensity of each event; s4, based on the event intensity characteristics, performing event intensity clustering on each event in the basic driving operation events according to the straight events and the steering events through a k-means clustering algorithm, and marking event intensity category labels on various events obtained through clustering according to the statistic value of the event intensity characteristics; s5, acquiring a road line shape based on the effective natural driving data, and acquiring a dynamic time window according to a straight line section and a curve section based on the change of the road line shape; constructing a time-varying curve of the event which represents the driving style in the dynamic time window and is provided with the event intensity category label based on the event intensity category and the event transfer characteristic in the dynamic time window; clustering the curves of the dynamic time windows according to the straight-going event and the steering event based on a DTW-fused curve clustering algorithm to obtain various time window curves; and marking driving style type labels on various time window curves based on the event intensity characteristics and the event transfer characteristics.
In some embodiments of the invention, the natural driving data comprises speed, longitudinal acceleration and lateral acceleration, and the pre-processing comprises anomalous data culling and data smoothing.
In some embodiments of the present invention, the basic driving operation event includes six types of driving operation events, which are straight acceleration, straight constant velocity, straight deceleration, steering acceleration, steering constant velocity, and steering deceleration.
In some embodiments of the present invention, the method for extracting the basic driving operation event in step S2 is specifically as follows: s21, calculating the radius of the driving track based on the speed and the lateral acceleration:
wherein R is the radius of the driving track, v is the speed, ayIs the lateral acceleration; s22, constructing a longitudinal acceleration waveform based on the longitudinal acceleration; and S23, extracting basic driving operation events based on the radius of the running track and the wave peak value and/or the wave trough value of the longitudinal acceleration waveform.
In some embodiments of the invention, the basic driving maneuver event is extracted by: setting the threshold value of the radius of the running track to be 1000m when the radius of the running track R is>When the number is 1000m, extracting as a straight event; when the radius of the running track R<At 1000m, extracting as a steering event; setting the wave peak value threshold value of longitudinal acceleration waveform to be 1.0m/s2The wave trough value threshold of the longitudinal acceleration wave is-1.0 m/s2When the longitudinal acceleration waveform peak value>1.0m/s2Then, extracting as an acceleration event; when the ratio is-1.0 m/s2The wave peak value or the wave trough value of the longitudinal acceleration waveform is less than or equal to 1.0m/s2Extracting the event to be a uniform event; when the wave valley value of the longitudinal acceleration waveform is less than-1.0 m/s2And, extracted as a deceleration event.
In some embodiments of the present invention, the method for obtaining the event intensity feature characterizing each event intensity in step S3 is as follows: s31, making use of effective natural driving dataConstructing driving behavior characteristics based on the variables, wherein the driving behavior characteristics comprise an average value, a median, a tail-cutting average value, a standard deviation, a quartile, an absolute median, a percentile, a minimum value, a maximum value, a shannon entropy and an approximate entropy; s32, calculating the characteristic weight of each driving behavior of each variable through an entropy method; and S33, based on the characteristic weight of each driving behavior of each variable, calculating the strength characteristic of each variable by the following polynomial:
wherein, var _ idxiIntensity feature representing the i-th event var variable, var ═ effective natural driving data]Velocity, longitudinal acceleration, lateral acceleration]，w_svarkK-th driving behavior characteristic weight, f, representing var variableikThe driving behavior characteristic is the kth driving behavior characteristic of the ith event var variable, and K is the driving behavior characteristic number of each variable; s34, calculating variance contribution rate and accumulated principal component contribution rate of each principal component by a principal component analysis method based on intensity characteristics of each variable, and selecting a corresponding principal component with the accumulated principal component contribution rate of more than or equal to 90% as an effective principal component; s35, calculating the weight of each effective principal component by the following formula:
wherein σmRepresents the variance contribution rate, w _ p, corresponding to the selected mth effective principal componentmRepresents the weight of the mth effective principal component, and M represents the number of the effective principal components; s36, effective principal component weight w _ pmAnd the corresponding effective principal component load a calculated by the principal component analysis methodmvarCalculating an event intensity characteristic characterizing the intensity of each event by the following formula:
wherein, event _ idxiAn event intensity characteristic of the ith event.
In some embodiments of the invention, in step S4, based on the event intensity characteristics, after event intensity clustering is performed on each of the basic driving operation events according to the straight-going event and the steering event by a k-means clustering algorithm, a clustering result is determined based on the contour coefficient and the davison baudin index; the statistical value of the event intensity characteristic comprises the average value, the maximum value, the minimum value and the standard deviation of the event intensity characteristic of each type in the straight-going event and the steering event which are obtained after clustering, and event intensity category labels are marked on the events which are obtained through clustering based on the statistical value of the event intensity characteristic.
In some embodiments of the present invention, in step S5, based on a curve clustering algorithm that fuses DTWs, the curves of the dynamic time windows are clustered according to the straight-going event and the steering event, and the method for obtaining the curves of the various time windows is specifically as follows: 1) recording the curve of each dynamic time window as a curve set C; 2) calculating the distance between every two curves in the curve set C based on a DTW algorithm, selecting any one of the two curves with the maximum distance between every two curves as a curve a, and setting a curve similarity distance threshold T; 3) classifying the curve a into a curve set C1, and recording the curve set C as C-C1; 4) calculating the distance between each curve in the curve set C and the curve set C1 to obtain a curve b in the curve set C corresponding to the minimum distance, and recording the curve set C1' as C1+ b; 5) calculating an intra-similarity D (C1 ') of a curve set C1', if D (C1 ') > T, turning to the step 2), otherwise, classifying a curve b into a curve set C1, respectively recording a curve set C1 ═ Cl + b, and a curve set C ═ C-C1, and turning the algorithm to the step 4); 6) and when the algorithm is operated until the curve set C is an empty set, the algorithm is terminated.
In some embodiments of the present invention, the curve similarity distance threshold T is a percentile of a distance set between every two curves in the curve set C calculated based on the DTW algorithm, the percentile of the dynamic time window of the straight line road segment is 87 percentile, and the percentile of the dynamic time window of the curve road segment is 50 percentile.
In some embodiments of the present invention, in step S5, the method for labeling the driving style type of each type of time window curve based on the event intensity category and the event transition feature is as follows: step one, calculating the class z time by the following formulaEvent transition probability of window curve:
wherein the class z time window curve represents a certain class of time window curve in the class of time window curves obtained by clustering the curves of the dynamic time windows based on a DTW-fused curve clustering algorithm, and thetatRepresenting the current event, thetat+1Indicating the event at the next moment, eventoEvent intensity category, event, representing the current time corresponding to the cluster of step S4lAn event intensity class, n (θ) representing the cluster corresponding to the step S4 at the next timet＝evento，θt+1＝eventl) Representing slave events within a certain class of time windowoTransfer to eventlThe number of such transition forms, n (θ)t+1＝eventl) Indicates that the t +1 time is eventlThe number of (2); step two, calculating the Shannon entropy of the event transition probability of the z-th time window curve by the following formula:
wherein, eventrThe number of event intensity categories involved in the z-th category of time window curve; step three, calculating the transfer tendency of the class z time window curve by the following formula:
wherein, o is the current event intensity category,
in the form of a weight of a transition,
the intensity class span difference of events before and after transition of each transition form of the z-th class time window curve, wherein O is N, and O, N respectively represent the number of the intensity classes of the events clustered in the step S4, and the eventsz.For the event intensity category after the transfer,
event intensity category before transfer; step four, calculating the average value of the event intensities of all events contained in the class z time window curve by the following formula to serve as the event intensity characteristic of the class z time window curve:
wherein s is the number of curves in the z-th class time window curve,
the event intensity category corresponding to the tth moment of the s-th curve is defined, and T is the duration time of the s-th curve; and fifthly, marking cautious, normal and aggressive driving style labels on various types of time window curves from small to large according to numerical values based on the event transition probability Shannon entropy of the z-th type time window curve, the transition tendency of the z-th type time window curve and the numerical value of the event intensity characteristic of the z-th type time window curve.
Compared with the prior art, the invention has the following beneficial effects:
(1) the natural driving data acquired by the invention is used as continuous time sequence data to describe complex driving behaviors of a driver when the driver completes a driving task, and direct analysis is not beneficial to understanding differences of driving styles from a microscopic level, so that basic driving operation events are extracted according to effective natural driving data and are used as basic units for describing the complex driving behaviors of the driver, the driving behavior characteristics can be embodied from a microscopic level, decision information of the driving behaviors can be embodied, and the driving style analysis is facilitated; in addition, in addition to taking the basic driving operation event as a basic unit, the analysis method also considers the event intensity category and the event transfer characteristic, takes the change curve of the event along with the time as the characteristic for describing the driving style, embodies the dynamic decision information of the driving behavior and the continuity and the time characteristic of the natural driving data, better retains the original information of the natural driving data, and improves the interpretability and the accuracy of the driving style analysis result.
(2) The method carries out feature construction based on effective natural driving data, can comprehensively represent data information, and improves the accuracy of a subsequent algorithm; the number of features is reduced through feature extraction, more representative features are obtained, and the calculation complexity is reduced.
(3) The driving style analysis method disclosed by the invention adopts two clustering algorithms based on k-means and DTW, and the k-means and DTW are fused into the analysis method disclosed by the invention, so that the driving style is analyzed under an unsupervised condition, and the limit of manual calibration is broken through.
(4) The dynamic time window is determined according to the road alignment, so that a sufficient number of events are contained in the time window, the dynamic decision of the driving behavior is reflected, the real-time online cluster analysis of the driving style is realized, the constantly changing real-time driving style of a driver can be effectively identified, and the composition of the long-term driving style of the driver is explained based on the real-time driving style; the invention provides a large-scale curve clustering method considering the global fusion DTW with dynamic time distortion similarity measurement aiming at curve clustering, which can be used for clustering curves with different lengths based on a large-scale curve clustering algorithm of the fusion DTW and avoiding local optimization caused by adopting a hierarchical clustering algorithm in the prior art.
Drawings
In order to more clearly illustrate the technical solution in the embodiment of the present invention, the drawings required to be used in the embodiment of the present invention will be described below.
FIG. 1 is a graph illustrating a basic driving maneuver event in an unsupervised driving style analysis method based on the basic driving maneuver event according to an embodiment of the present invention;
FIG. 2 is a histogram of the distribution of straight-ahead events and steering events during a basic driving maneuver in accordance with one embodiment of the present invention;
FIG. 3 is a driving behavior feature weight graph of the construction of variables in a straight-ahead event and a steering event in accordance with one embodiment of the present invention;
FIG. 4 is a graph of an intensity profile of each variable during a straight-ahead event and a steering event in accordance with one embodiment of the present invention;
FIG. 5 is a boxplot of the intensity profile of each variable in a straight-ahead event and a steering event in accordance with one embodiment of the present invention;
FIG. 6 is a graph of an intensity profile of each event in a straight-ahead event and a steering event in accordance with one embodiment of the present invention;
FIG. 7 is a graph illustrating event intensity cluster indicator changes for a straight-ahead event and a turn event, in accordance with an embodiment of the present invention;
FIG. 8 is a diagram illustrating event intensity clustering results for a straight-ahead event and a turn event according to an embodiment of the present invention;
FIG. 9 is a histogram of dynamic time window length distribution according to an embodiment of the present invention;
FIG. 10 is a diagram illustrating the result of dividing the time window into a straight-ahead event and a turning event according to an embodiment of the present invention;
FIG. 11 is a graph of event intensity categories for straight-ahead events and steering events over time for one embodiment of the present invention;
FIG. 12 is a diagram illustrating a clustering result of a time window curve including a straight event and a turning event according to an embodiment of the present invention;
FIG. 13 is a time window driving style category plot including a straight-ahead event and a steering event in accordance with one embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the following detailed description of various aspects of the present invention is provided with specific examples, which are only used for illustrating the present invention and do not limit the scope and spirit of the present invention.
The embodiment provides an unsupervised driving style analysis method based on basic driving operation events. Fig. 1 is a graph illustrating a basic driving operation event in the unsupervised driving style analysis method based on the basic driving operation event according to the present embodiment; fig. 2 shows a histogram of distribution of straight-going events and steering events in the basic driving operation event of the present embodiment; FIG. 3 is a driving behavior feature weight diagram showing the construction of variables in the straight-ahead event and the steering event according to the present embodiment; FIG. 4 is a graph showing the intensity profile of each variable in the straight-ahead event and the steering event of the present embodiment; FIG. 5 is a box plot showing the intensity profile of each variable in the straight-ahead event and the steering event of the present embodiment; FIG. 6 illustrates a respective event intensity profile for a straight-ahead event and a steering event for the present embodiment; FIG. 7 is a graph showing the event intensity cluster indicator change for the straight-ahead event and the steering event of the present embodiment; FIG. 8 is a diagram illustrating the event intensity clustering results of the straight-going event and the turning event according to the present embodiment; FIG. 9 shows a dynamic time window length distribution histogram of the present embodiment; FIG. 10 is a diagram illustrating the result of dividing the time window straight-going event and the steering event according to the embodiment; FIG. 11 is a graph showing the event intensity categories of the straight travel event and the steering event of the present embodiment as a function of time; FIG. 12 is a diagram illustrating a clustering result of a time window curve including a straight event and a turning event according to the embodiment; fig. 13 shows a time window driving style category plot for the present embodiment including a straight-ahead event and a steering event.
The unsupervised driving style analysis method based on the basic driving operation event comprises the following steps of:
s1, natural driving data are obtained, and the natural driving data are preprocessed to obtain effective natural driving data;
s2, extracting basic driving operation events based on the effective natural driving data;
s3, based on effective natural driving data, performing feature construction and feature extraction on each event in the basic driving operation events through driving behavior features to obtain event intensity features representing the intensity of each event;
s4, based on the event intensity characteristics, performing event intensity clustering on each event in the basic driving operation events according to the straight events and the steering events through a k-means clustering algorithm, and marking event intensity category labels on various events obtained by clustering according to the statistic value of the event intensity characteristics;
s5, acquiring a road alignment based on the effective natural driving data, and acquiring a dynamic time window according to a straight line section and a curve section based on the change of the road alignment; constructing a change curve of an event with an event intensity category label and representing the driving style in the dynamic time window along with time based on the event intensity category and the event transfer characteristic in the dynamic time window; clustering curves of all dynamic time windows according to a straight-going event and a turning event based on a DTW-fused curve clustering algorithm to obtain various time window curves; and marking driving style type labels on various time window curves based on the event intensity category and the event transfer characteristics.
In this embodiment, the manner of acquiring the natural driving data in step S1 is not limited, and those skilled in the art can select the acquisition manner according to actual situations. In this embodiment, it is preferable that the natural driving data at step S1 is collected by using a RADS type 8-degree-of-freedom panoramic driving simulation system developed by the department of transportation, highway science Research Institute (RIOS), which performs traffic simulation, in which a driver under test performs free driving and autonomous lane changing according to individual driving habits and simulated traffic (a preset road route), and collects natural driving data of the driver under test. In the present embodiment, the natural driving data includes speed, longitudinal acceleration, and lateral acceleration. In the driving simulation experiment process, due to the operation error of a tested driver or the reason of equipment, the acquired natural driving data has the problem of abnormity or defect, the natural driving data needs to be preprocessed, and the preprocessing of the embodiment comprises abnormal data elimination and data smoothing. In this embodiment, the specific method of the abnormal data elimination and the data smoothing processing is not limited, and those skilled in the art can reasonably select the method according to actual needs, for example, a moving average method can be used to perform the data smoothing processing; the method for eliminating the abnormal data can be as follows: removing abnormal value according to data normal threshold in the field, wherein the abnormal value is larger than 10m/s in longitudinal acceleration and lateral acceleration2And less than-10 m/s2The data point of (2) is replaced by the average of its first 3 and last 3 adjacent data points, and the data point with a velocity greater than 36m/s is replaced by the average of its first 3 and last 3 adjacent data points.
In the present embodiment, the basic driving operation events extracted in step S2 include six types of driving operation events, which are straight acceleration, straight constant speed, straight deceleration, steering acceleration, steering constant speed, and steering deceleration, as shown in fig. 1-2.
In the present embodiment, the method for extracting the basic driving operation event in step S2 is specifically as follows:
s21, calculating the radius of the running track based on the speed and the lateral acceleration:
wherein R is the radius of the driving track, v is the speed, ayIs the lateral acceleration;
s22, constructing a longitudinal acceleration waveform based on the longitudinal acceleration (in this embodiment, the complete waveform of the longitudinal acceleration is a process in which the longitudinal acceleration increases (decreases) from 0 to a certain value and then decreases (increases) to 0);
and S23, extracting basic driving operation events based on the radius of the running track and the wave peak value and/or the wave trough value of the longitudinal acceleration waveform.
In the present embodiment, the step S23 extracts a basic driving operation event by the following conditions: setting the threshold value of the radius of the running track to be 1000m when the radius of the running track R is>When the number is 1000m, extracting as a straight event; when the radius of the running track R<At 1000m, extracting as a steering event; setting the wave peak value threshold value of longitudinal acceleration waveform to be 1.0m/s2The wave trough value threshold of the longitudinal acceleration wave is-1.0 m/s2When the longitudinal acceleration waveform peak value>1.0m/s2Then, extracting as an acceleration event; when the ratio is-1.0 m/s2The wave peak value or the wave trough value of the longitudinal acceleration waveform is less than or equal to 1.0m/s2Extracting the event to be a uniform event; when longitudinal acceleration waveform trough value<-1.0m/s2And, extracted as a deceleration event. It should be noted that, during the actual driving process, the driver is subjected to the driving behaviorThe direct control to the driver, whether the driver is accelerating, driving at a constant speed or decelerating, is more or less accompanied by the change of the acceleration, for example, the acceleration behavior is not generally accelerated at a constant acceleration, the deceleration behavior is not generally decelerated at a constant acceleration, the constant speed behavior is not generally guaranteed to be always zero (the acceleration change is more or less slight, but the acceleration is smaller and can be ignored under the condition of the body feeling or external observation of the vehicle), in order to more accurately judge the basic driving operation event and prevent the misjudgment of the constant speed behavior accompanied by the slight acceleration change as the acceleration behavior or the deceleration behavior, the embodiment constructs a longitudinal acceleration waveform by the longitudinal acceleration in the collected and preprocessed effective natural driving data, and based on a preset longitudinal acceleration waveform peak value threshold value, The preset longitudinal acceleration waveform trough value threshold value and the wave peak value and trough value of the constructed longitudinal acceleration waveform are used for judging acceleration, uniform speed and deceleration events, the influence of acceleration change caused by more or less acceleration, uniform speed and deceleration driving when a driver artificially controls driving behaviors is considered, the acceleration, uniform speed and deceleration events can be more accurately judged, and event misjudgment is prevented. In the embodiment, six driving operation events including straight-going acceleration, straight-going constant speed, straight-going deceleration, steering acceleration, steering constant speed and steering deceleration can be automatically extracted by comprehensively considering the threshold of the radius of the running track, the threshold of the wave peak value of the longitudinal acceleration waveform and the threshold of the wave trough value of the longitudinal acceleration waveform.
In this embodiment, the method for acquiring the event intensity feature characterizing each event intensity in step S3 specifically includes the following steps:
s31, constructing the driving behavior characteristics of the following table 1 based on the variables by taking effective natural driving data as variable parameters, and expressing the operating characteristics of the driving behaviors on the variables; the driving behavior characteristics comprise an average value, a median, a cropping average value, a standard deviation, a quartering potential difference, an absolute median potential difference, a percentile, a minimum value, a maximum value, a Shannon entropy and an approximate entropy;
TABLE 1 driving behavior characteristics of variable parameters and configurations
S32, calculating a characteristic weight of each driving behavior of each variable by using an entropy method (in this embodiment, the entropy method is a commonly used method for calculating a bottom-layer index weight, and measures a dispersion degree of variable data according to an entropy value, and the smaller the entropy value of the variable, the larger the dispersion degree of the variable, the larger the weight of the variable, the entropy method is known in the art and is not described herein again), as shown in fig. 3, the characteristic weight of the driving behavior constructed by each variable in the straight-going event and the steering event obtained in this embodiment is shown;
s33, based on the characteristic weight of each driving behavior of each variable, calculating the strength characteristic of each variable through the following polynomial (fig. 4 shows the strength characteristic distribution diagram of each variable in the straight-going event and the steering event obtained in the embodiment; fig. 5 shows the strength characteristic box diagram of each variable in the straight-going event and the steering event obtained in the embodiment):
wherein, var _ idxiIntensity feature representing the i-th event var variable, var ═ effective natural driving data]Velocity, longitudinal acceleration, lateral acceleration]，w_svarkK-th driving behavior characteristic weight, f, representing var variableikThe driving behavior characteristic is the kth driving behavior characteristic of the ith event var variable, and K is the driving behavior characteristic number of each variable;
s34, calculating variance contribution rate and accumulated principal component contribution rate of each principal component based on intensity characteristics of each variable by a principal component analysis method, and selecting a corresponding principal component with the accumulated principal component contribution rate of more than or equal to 90% as an effective principal component (in the embodiment, the principal component analysis method is a top-level index weight calculation method, and the principal component analysis method is a method known in the art and is not described herein again);
s35, calculating the weight of each effective principal component by the following formula:
wherein σmRepresents the variance contribution rate, w _ p, corresponding to the selected mth effective principal componentmRepresents the weight of the mth effective principal component, and M represents the number of the effective principal components;
s36, effective principal component weight w _ pmAnd the corresponding effective principal component load a calculated by the principal component analysis methodmvarAn event intensity characteristic representing the intensity of each event is calculated by the following formula (fig. 6 shows the distribution diagram of the intensity characteristics of each event in the straight-going event and the steering event obtained by the embodiment):
wherein, event _ idxiAn event intensity characteristic of the ith event.
In this embodiment, in step S4, based on the event intensity characteristics, event intensity clustering is performed on each event in the basic driving operation events according to the straight-going event and the turning event by using a K-means clustering algorithm, and the events with similar intensities are clustered into a class, so as to obtain an event intensity category (in this embodiment, the K-means algorithm is a classical clustering algorithm for iterative solution, and is not described herein again). The clustering number is selected to be 1-10 in sequence, the clustering result is analyzed based on the profile coefficient SC and the Davison baudian index DBI, and the optimal clustering number is determined (in the embodiment, the profile coefficient SC and the Davison baudian index DBI are common internal indexes of clustering performance measurement, and the clustering result is evaluated by calculating the intra-class compactness and the inter-class dispersion, wherein the threshold value of the profile coefficient SC is [ -1,1], the closer the value is to 1, the better the clustering effect is, and the smaller the value of the Davison Baoding index DBI is, the better the clustering effect is). As shown in fig. 7, which shows a graph of the event intensity cluster index change of the straight event and the turning event, it can be seen that, for the straight event, the clustering effect is the best when the cluster number is 3; for turn events, the clustering effect is best when the number of clusters is 2. Fig. 8 shows a schematic diagram of event intensity clustering results of straight-going events and turning events, corresponding to the optimal clustering numbers of fig. 7. In this embodiment, the statistical values of the event intensity features include an average value, a maximum value, a minimum value, and a standard deviation of the intensity features of each type of event in the straight-going event and the steering event, which are obtained after clustering, and event intensity category labels are marked on the various types of events obtained by clustering based on the statistical values of the event intensity features. Specifically, the variation trends of the average value, the maximum value and the minimum value of the intensity characteristics of each type of event in the straight-going event and the steering event obtained after clustering are consistent, the standard deviation difference is small, the larger the average value, the maximum value and the minimum value of the intensity characteristics of each type of event in the straight-going event and the steering event obtained after clustering are, the stronger the event intensity is, and a label representing high intensity is marked when the label is marked; the smaller the average value, the maximum value and the minimum value of the intensity characteristics of each type of events in the straight-going events and the steering events obtained after clustering, the lower the intensity of the events, and the label representing the low intensity when the label is marked. The standard deviation for each class is small, indicating that the intensity profile for each class of events is relatively concentrated. In this embodiment, it is preferable that the category labels of the straight-going event and the turning event are mapped to category values having physical meanings, and the higher the intensity is, the larger the mapping value is (for example, the values mapped in fig. 8: the event intensity category of the straight-going event includes category 1, category 2 and category 3; and the event intensity category of the turning event includes category 1 and category 2).
In this embodiment, the step S5 is to obtain the road alignment based on the effective natural driving data, and to obtain the dynamic time window according to the straight line section and the curved line section based on the change of the road alignment, specifically, by: and determining the road alignment according to the running track radius R, and taking the road alignment as a basis for determining the length of the dynamic time window, wherein the dynamic time window comprises a straight-going event when the running track radius R is greater than 1000m and the duration t is greater than 4s, and the dynamic time window comprises a steering event when the running track radius R is less than 1000m and the duration t is greater than 4 s. Fig. 9 shows a dynamic time window length distribution histogram, and fig. 10 shows a time window straight-going event and steering event partitioning result diagram. In this embodiment, the straight line segment is a segment including a straight event (straight constant speed, straight deceleration, straight acceleration), and the curved line segment is a segment including a turning event (turning constant speed, turning deceleration, turning acceleration).
In the present embodiment, the step S5 is to construct a time-varying curve of the event with the event intensity category label, which represents the driving style in the dynamic time window, based on the event intensity category and the event transition feature in the dynamic time window, as shown in fig. 11, which shows a time-varying graph of the event intensity categories of the straight-going event and the turning event, where the horizontal axis of the curve is the time axis, and the vertical axis is the event intensity category value corresponding to the time point. In the present embodiment, the event transition feature represents a feature in which the current event intensity category at the current time is transitioned to the next event intensity category at the next time.
In this embodiment, in step S5, based on the curve clustering algorithm fused with DTW, the curves of the dynamic time windows are clustered according to the straight events and the turning events to obtain various time window curves (for example, fig. 12 shows a schematic diagram of a time window curve clustering result including the straight events and the turning events, in fig. 12, a category A, B, C indicates a type of various time window curves obtained by clustering the curves of the dynamic time windows based on the curve clustering algorithm fused with DTW, which reflects a driving style corresponding to each time window obtained after clustering, and indicates that the clustering result is 3 categories, i.e., 3 driving styles, and can ensure interpretability and acceptability of the clustering result), which is specifically as follows:
1) recording the curves of the dynamic time windows as a curve set C;
2) calculating the distance between every two curves in the curve set C based on a DTW algorithm, selecting any one of the two curves with the maximum distance between every two curves as a curve a, and setting a curve similarity distance threshold T;
3) the curve a is classified into a curve set C1, and the curve set C is recorded as C-C1;
4) calculating the distance between each curve in the curve set C and the curve set C1 to obtain a curve b in the curve set C corresponding to the minimum distance, and recording the curve set C1' as C1+ b;
5) calculating the similarity D (C1 ') in the curve set C1', if D (C1 ') > T, going to step 2), otherwise, putting the curve b into the curve set C1, respectively recording the curve set C1 as C1+ b and the curve set C as C-C1, and going to step 4);
6) and when the algorithm is operated until the curve set C is an empty set, the algorithm is terminated.
In the present embodiment, the calculation method of step 4) in step S5 is as follows:
definition 1: curve Lp(any one of the curves in the curve set C) and the curve set C1 (L)1,L2,...,Lm) Is defined as curve LpThe average of the distances from all the curves in the curve set C1, i.e.,
wherein m represents the number of curves in the curve set C1, LzRepresents the z-th curve, L, in the curve set C1pRepresents any curve in the curve set C.
In the present embodiment, the calculation method of step 5) in step S5 is as follows:
definition 2: curve set C1' (L)1,L2,...,Ln) Is defined as the maximum distance between two curves in the curve set C1', i.e.,
wherein n represents the number of curves in the curve set C1', LiRepresents the ith curve, L, in the curve set C1jRepresents the jth curve in the curve set C1'.
In this embodiment, in step S5, based on the curve clustering algorithm fused with DTW, the curves of the dynamic time windows are clustered according to the straight-going event and the turning event, and the method for obtaining the curves of the various time windows is performed step by step according to steps 1) -6) above, it is noted that the term appearing in the next step is the term after the previous step, if the term does not appear in the previous step, the term is the term after the previous step, and so on, until the term appears recently (for example, the term "curve set C": and 5) when the step 5) is consistent with D (C1 ') > T, turning to the step 2), wherein the curve set C appears in the step 2), the last step of the step 2) is the step 5) which is consistent with D (C1 ') > T, the curve set C does not appear, and then the last step 4) of the step 5) is checked, wherein the term curve set C appears, the curve set C is not limited in the step 4) and then the last step 3) of the step 4) is checked, the step 3) processes the curve set C, and the curve set C used when the step 5) is consistent with D (C1 ') > T is turned to the step 2) is the curve set C processed in the step 3). In this embodiment, the meaning of the curve set C ═ C-C1 in step 3) is: the curve set C processed in the step 3) represents a curve set obtained by removing the curve set C1 from the curve set C in the previous step, namely the step 2); the meanings of the curve set C1 ═ C1+ b in step 4), the curve set C1 ═ C1+ b in step 5) and the curve set C — C1 in step 5) can be understood similarly and correspondingly.
In this embodiment, the curve similarity distance threshold T set in step S5 in step 2) is a percentile of a distance set between every two curves in the curve set C calculated based on the DTW algorithm, the percentile of the dynamic time window of the straight line segment is 87 percentile, the percentile of the dynamic time window of the curve segment is 50 percentile, that is, the curve similarity distance threshold T of the dynamic time window of the straight line segment is 87 percentile, and the curve similarity distance threshold T of the dynamic time window of the curve segment is 50 percentile.
In this embodiment, in step S5, based on the event intensity category and the event transition feature, the method for labeling the driving style type of each type of time window curve specifically includes:
step one, calculating the event transition probability of a class z time window curve by the following formula:
wherein the class z time window curve represents a certain class of time window curve in the class of time window curves obtained by clustering the curves of the dynamic time windows based on a DTW-fused curve clustering algorithm, and thetatRepresenting the current event, thetat+1Event, representing the next moment in timeoEvent intensity category, event, representing the current time corresponding to the cluster of step S4lAn event intensity class, n (θ) representing the cluster corresponding to the step S4 at the next timet＝evento，θt+1＝eventl) Representing slave events within a certain class of time windowoTransfer to eventlThe number of such transition forms, n (θ)t+1＝eventl) Indicates that the t +1 time is eventlThe number of (2);
step two, calculating the Shannon entropy of the event transition probability of the z-th time window curve by the following formula:
wherein, eventrThe number of event intensity categories involved in the z-th category of time window curve;
step three, calculating the transfer tendency of the class z time window curve by the following formula:
wherein, o is the current event intensity category,
in the form of a weight of a transition,
the difference of the event intensity class span before and after transition (i.e. before and after transition) for each transition form of the class z time window curveThe difference between the class values having physical meanings mapped by the intensity class label), O ═ N (O ═ N indicates that O and N have the same meaning), and O, N each indicate the number of event intensity classes of the cluster of step S4, eventz.For the event intensity category after the transfer,
event intensity category before transfer;
step four, calculating the average value of the event intensities of all events contained in the class z time window curve by the following formula to serve as the event intensity characteristic of the class z time window curve:
wherein s is the number of curves in the z-th class time window curve,
the event intensity category corresponding to the tth moment of the sth curve is T, and the T is the duration time of the sth curve;
and fifthly, marking cautious, normal and aggressive driving style labels on various types of time window curves from small to large according to numerical values based on the event transition probability Shannon entropy of the z-th type time window curve, the transition tendency of the z-th type time window curve and the numerical value of the event intensity characteristic of the z-th type time window curve. In the embodiment, the three indexes of event transition probability shannon entropy, transition tendency and event intensity characteristic are all positive indexes, and the smaller the numerical values of the three indexes are, the more cautious the driving style is, and a cautious type label is marked; the larger the numerical values of the three indexes are, the more aggressive the driving style is represented, and aggressive labels are marked; the numerical values of the three indexes are centered, and the more normal the driving style is, the normal type label is marked. As fig. 13 shows a time window driving style category graph containing a straight going event and a turning event, it can be seen that: the lateral acceleration of the time window containing the straight-going event has no obvious difference; while cautious and normal types are relatively slow compared to aggressive types; the prudent type longitudinal acceleration is not obviously changed and is basically maintained near zero, so that the prudent type longitudinal acceleration is maintained at a relatively low speed for uniform running; compared with the prudent type, the normal type has obvious acceleration and deceleration behaviors in an example time range, and has more diversity and more frequent operation; for the aggressive type, the speed is relatively large, and the longitudinal acceleration change is obvious, so the aggressive type is more prone to high-intensity and difficult operation. For the time window containing the steering event, under the condition that the lateral acceleration is approximately the same, the cautious turning speed is relatively low, the cautious turning speed enters the curve at a relatively low speed in the early stage of the turning process and slightly carries out deceleration action, the turning process is continuously accelerated until the turning is finished along with the propulsion of the turning process, the driving behavior characteristic that a cautious driver firstly enters the curve at a low speed to reduce the driving risk and slowly accelerates until the turning is finished along with the propulsion of the turning process in the common knowledge recognition is met, the driving risk irritation is relatively low, and the comfort is relatively high; the aggressive driver has higher speed when turning, obvious acceleration and deceleration behaviors show that the aggressive driving behaviors still keep stronger operating characteristics when turning, and the aggressive driver continuously accelerates and decelerates and keeps higher speed in the turning process, so that the aggressive driver pursues driving excitement and tends to continuously adjust own driving strategy along with the change of the surrounding environment, therefore, the driving behaviors have poorer comfortable feelings, and the excitement caused by the aggressive driver is stronger due to the continuous change of the acceleration and is consistent with the aggressive driving style known by the common sense of people; the normal type passes through the curve at a constant speed, the operation is not too cautious and not too frequent stimulation as the aggressive type, the curve is passed through by a driving behavior operation mode which is relatively gentle and has traffic efficiency, and the artificial common sense cognition is met. As can be seen from fig. 13 and the above analysis, the result obtained by the method for analyzing the driving style of the present embodiment matches the characteristics of actual cautious, normal and aggressive driving, which indicates that the method for analyzing the driving style of the present embodiment is effective and highly accurate.
The present invention has been described in conjunction with specific embodiments which are intended to be exemplary only and are not intended to limit the scope of the invention, which is to be given the full breadth of the appended claims and any and all modifications, variations or alterations that may occur to those skilled in the art without departing from the spirit of the invention. Therefore, various equivalent changes made according to the present invention still fall within the scope covered by the present invention.
Claims (10)
1. An unsupervised driving style analysis method based on basic driving operation events is characterized by comprising the following steps of:
s1, natural driving data are obtained, and the natural driving data are preprocessed to obtain effective natural driving data;
s2, extracting basic driving operation events based on the effective natural driving data;
s3, based on the effective natural driving data, performing feature construction and feature extraction on each event in the basic driving operation events through driving behavior features to obtain event intensity features representing the intensity of each event;
s4, based on the event intensity characteristics, performing event intensity clustering on each event in the basic driving operation events according to the straight events and the steering events through a k-means clustering algorithm, and marking event intensity category labels on various events obtained through clustering according to the statistic value of the event intensity characteristics;
s5, acquiring a road line shape based on the effective natural driving data, and acquiring a dynamic time window according to a straight line section and a curve section based on the change of the road line shape; constructing a time-varying curve of the event which represents the driving style in the dynamic time window and is provided with the event intensity category label based on the event intensity category and the event transfer characteristic in the dynamic time window; clustering the curves of the dynamic time windows according to the straight-going event and the steering event based on a DTW-fused curve clustering algorithm to obtain various time window curves; and marking driving style type labels on various time window curves based on the event intensity category and the event transfer characteristics.
2. The method of claim 1, wherein the natural driving data includes speed, longitudinal acceleration, and lateral acceleration, and the pre-processing includes anomalous data culling and data smoothing.
3. The method of claim 1, wherein the basic driving maneuver events include six types of driving maneuver events, the six types of driving maneuver events being straight acceleration, straight cruise, straight deceleration, steering acceleration, steering cruise, and steering deceleration.
4. The method of claim 2, wherein the method of extracting the basic driving operation event in step S2 is specifically as follows:
s21, calculating the radius of the running track based on the speed and the lateral acceleration:
wherein R is the radius of the driving track, v is the speed, ayIs the lateral acceleration;
s22, constructing a longitudinal acceleration waveform based on the longitudinal acceleration;
and S23, extracting basic driving operation events based on the radius of the running track and the wave peak value and/or the wave trough value of the longitudinal acceleration waveform.
5. The method of claim 4, wherein the basic driving maneuver event is extracted by:
setting a running track radius threshold value as 1000m, and extracting a straight-going event when the running track radius R > is 1000 m; when the radius R of the running track is less than 1000m, extracting as a steering event;
setting the wave peak value threshold value of longitudinal acceleration waveform to be 1.0m/s2The wave trough value threshold of the longitudinal acceleration wave is-1.0 m/s2When the longitudinal acceleration waveform peak value>1.0m/s2Then, extracting as an acceleration event; when the ratio is-1.0 m/s2The wave peak value or the wave trough value of the longitudinal acceleration waveform is less than or equal to 1.0m/s2Extracting the event to be a uniform event; when longitudinal acceleration waveform trough value<-1.0m/s2And, extracted as a deceleration event.
6. The method according to claim 1, wherein the step S3 of obtaining the event intensity characteristics characterizing each event intensity is as follows:
s31, constructing driving behavior characteristics based on the variables by taking effective natural driving data as variables, wherein the driving behavior characteristics comprise an average value, a median, a tailcut average value, a standard deviation, a quartile difference, an absolute median difference, a percentile, a minimum value, a maximum value, a Shannon entropy and an approximate entropy;
s32, calculating the characteristic weight of each driving behavior of each variable through an entropy method;
and S33, based on the characteristic weight of each driving behavior of each variable, calculating the strength characteristic of each variable by the following polynomial:
wherein, var _ idxiIntensity feature representing the i-th event var variable, var ═ effective natural driving data]Velocity, longitudinal acceleration, lateral acceleration]，w_svarkK-th driving behavior characteristic weight, f, representing var variableikThe driving behavior characteristic is the kth driving behavior characteristic of the ith event var variable, and K is the driving behavior characteristic number of each variable;
s34, calculating variance contribution rate and accumulated principal component contribution rate of each principal component by a principal component analysis method based on intensity characteristics of each variable, and selecting a corresponding principal component with the accumulated principal component contribution rate of more than or equal to 90% as an effective principal component;
s35, calculating the weight of each effective principal component by the following formula:
wherein σmRepresents the variance contribution rate, w _ p, corresponding to the selected mth effective principal componentmRepresents the weight of the mth effective principal component,m represents the number of effective main components;
s36, effective principal component weight w _ pmAnd the corresponding effective principal component load a calculated by the principal component analysis methodmvarCalculating an event intensity characteristic characterizing the intensity of each event by the following formula:
wherein, event _ idxiAn event intensity characteristic of the ith event.
7. The method as claimed in claim 1, wherein in step S4, after event intensity clustering is performed on each of the basic driving operation events according to the straight-going event and the turning event by a k-means clustering algorithm based on the event intensity characteristics, a clustering result is determined based on the contour coefficient and the davison baudin index;
the event intensity characteristic statistic value comprises the average value, the maximum value, the minimum value and the standard deviation of the event intensity characteristics of each type of events in the straight-going events and the steering events obtained after clustering, and event intensity category labels are marked on the events obtained by clustering on the basis of the event intensity characteristic statistic value.
8. The method according to claim 1, wherein in step S5, based on a curve clustering algorithm with DTW fusion, the curves of the dynamic time windows are clustered according to the straight-going event and the turning event, and the method for obtaining the curves of the various time windows is specifically as follows:
1) recording the curve of each dynamic time window as a curve set C;
2) calculating the distance between every two curves in the curve set C based on a DTW algorithm, selecting any one of the two curves with the maximum distance between every two curves as a curve a, and setting a curve similarity distance threshold T;
3) classifying the curve a into a curve set C1, and recording the curve set C as C-C1;
4) calculating the distance between each curve in the curve set C and the curve set C1 to obtain a curve b in the curve set C corresponding to the minimum distance, and recording the curve set C1' as C1+ b;
5) calculating an internal similarity D (C1 ') of a curve set C1', if D (C1 ') > T, turning to the step 2), otherwise, classifying a curve b into a curve set C1, and respectively recording a curve set C1 as C1+ b and a curve set C as C-C1, and turning the algorithm to the step 4);
6) and when the algorithm is operated until the curve set C is an empty set, the algorithm is terminated.
9. The method according to claim 8, wherein the curve similarity distance threshold T is a percentile of a set of distances between two curves in the curve set C calculated based on the DTW algorithm, the percentile of the dynamic time window of the straight line segment is 87 percentile, and the percentile of the dynamic time window of the curve segment is 50 percentile.
10. The method according to claim 8, wherein in step S5, based on the event intensity category and the event transition feature, the driving style type labels are marked on various types of time window curves as follows:
step one, calculating the event transition probability of a class z time window curve by the following formula:
wherein the class z time window curve represents a certain class of time window curve in the class of time window curves obtained by clustering the curves of the dynamic time windows based on a DTW-fused curve clustering algorithm, and thetatRepresenting the current event, thetat+1Event, representing the next moment in timeoEvent intensity category, event, representing the current time corresponding to the cluster of step S4lIndicates the event intensity class, n (θ) corresponding to the cluster of step S4, at the next timet＝evento，θt+1＝eventl) Representing slave events within a certain class of time windowoTransfer to eventlNumber of such transition forms, n (θ)t+1＝eventl) Indicating that the t +1 moment is eventlThe number of (2);
step two, calculating the Shannon entropy of the event transition probability of the z-th time window curve by the following formula:
wherein, eventrThe number of event intensity categories involved in the z-th category of time window curve;
step three, calculating the transfer tendency of the class z time window curve through the following formula:
wherein, o is the current event intensity category,
in the form of a weight of a transition,
the intensity class span difference of events before and after transition of each transition form of the z-th class time window curve, wherein O is N, and O, N respectively represent the number of the intensity classes of the events clustered in the step S4, and the eventsz.For the event intensity category after the transfer,
event intensity category before transfer;
step four, calculating the average value of the event intensities of all events contained in the class z time window curve by the following formula to be used as the event intensity characteristic of the class z time window curve:
wherein s is the number of curves in the z-th class time window curve,
the event intensity category corresponding to the tth moment of the sth curve is T, and the T is the duration time of the sth curve;
and fifthly, marking cautious, normal and aggressive driving style labels on various types of time window curves from small to large according to numerical values based on the event transition probability Shannon entropy of the z-th type time window curve, the transition tendency of the z-th type time window curve and the numerical value of the event intensity characteristic of the z-th type time window curve.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210176012.4A CN114707573A (en) | 2022-02-25 | 2022-02-25 | Unsupervised driving style analysis method based on basic driving operation event |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210176012.4A CN114707573A (en) | 2022-02-25 | 2022-02-25 | Unsupervised driving style analysis method based on basic driving operation event |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114707573A true CN114707573A (en) | 2022-07-05 |
Family
ID=82166558
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210176012.4A Pending CN114707573A (en) | 2022-02-25 | 2022-02-25 | Unsupervised driving style analysis method based on basic driving operation event |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114707573A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117238131A (en) * | 2023-09-14 | 2023-12-15 | 中国民航大学 | Traffic flow characteristic analysis method in Internet of vehicles environment |
-
2022
- 2022-02-25 CN CN202210176012.4A patent/CN114707573A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117238131A (en) * | 2023-09-14 | 2023-12-15 | 中国民航大学 | Traffic flow characteristic analysis method in Internet of vehicles environment |
| CN117238131B (en) * | 2023-09-14 | 2024-05-07 | 中国民航大学 | Traffic flow characteristic analysis method in Internet of vehicles environment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106971194B (en) | Driving intention recognition method based on improved HMM and SVM double-layer algorithm | |
| CN110569783B (en) | Method and system for identifying lane changing intention of driver | |
| CN113486822B (en) | Surrounding vehicle track prediction method and system based on driving intention | |
| CN111547064B (en) | Driving style recognition and classification method for automobile adaptive cruise system | |
| CN104951764B (en) | Hot-short Activity recognition method based on secondary spectral clustering and HMM-RF mixed models | |
| CN110120218A (en) | Expressway oversize vehicle recognition methods based on GMM-HMM | |
| CN111738337B (en) | Driver distraction state detection and identification method in mixed traffic environment | |
| CN113112796B (en) | Construction method and system of driving behavior characteristics and driving behavior analysis method | |
| CN110386144B (en) | GHMM/GGAP-RBF hybrid model for identifying driver braking intention and identification method | |
| CN113232669B (en) | Driving style identification method based on machine learning | |
| CN114493191B (en) | Driving behavior modeling analysis method based on network about vehicle data | |
| CN114169444B (en) | Driving style classification method considering risk potential field distribution under vehicle following working condition | |
| CN112699793A (en) | Fatigue driving detection optimization identification method based on random forest | |
| CN116363712B (en) | Palmprint palm vein recognition method based on modal informativity evaluation strategy | |
| CN111627132B (en) | Driving inclination-based automobile driving characteristic evaluation and early warning method | |
| CN116028884A (en) | Prototype network-based vehicle lane change risk assessment method under small sample | |
| CN114707573A (en) | Unsupervised driving style analysis method based on basic driving operation event | |
| CN115092165A (en) | Clustering model-based driver style identification method under different cycle conditions | |
| CN113581188A (en) | Commercial vehicle driver driving style identification method based on Internet of vehicles data | |
| CN112308136B (en) | Driving distraction detection method based on SVM-Adaboost | |
| CN113591780A (en) | Method and system for identifying driving risk of driver | |
| CN116432108B (en) | Driving behavior evaluation and driving style online identification method under following scene | |
| CN117077042B (en) | Rural level crossing safety early warning method and system | |
| CN112937592B (en) | Method and system for identifying driving style based on headway | |
| CN116091254B (en) | Commercial vehicle risk analysis method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |