CN111882923A - Intelligent networking automobile behavior identification method - Google Patents

Intelligent networking automobile behavior identification method Download PDF

Info

Publication number
CN111882923A
CN111882923A CN202010684027.2A CN202010684027A CN111882923A CN 111882923 A CN111882923 A CN 111882923A CN 202010684027 A CN202010684027 A CN 202010684027A CN 111882923 A CN111882923 A CN 111882923A
Authority
CN
China
Prior art keywords
driving behavior
data
sequence
vehicle
lane
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
Application number
CN202010684027.2A
Other languages
Chinese (zh)
Inventor
罗映
王金祥
王淑超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Netlink Intelligent Vehicle Industry Technology Research Institute Co ltd
Original Assignee
Shandong Netlink Intelligent Vehicle Industry Technology Research Institute Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Netlink Intelligent Vehicle Industry Technology Research Institute Co ltd filed Critical Shandong Netlink Intelligent Vehicle Industry Technology Research Institute Co ltd
Priority to CN202010684027.2A priority Critical patent/CN111882923A/en
Publication of CN111882923A publication Critical patent/CN111882923A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/167—Driving aids for lane monitoring, lane changing, e.g. blind spot detection

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Traffic Control Systems (AREA)

Abstract

An intelligent networking automobile behavior identification method is used for improving the accuracy of identification results of continuous driving behavior samples. Using Viterbi algorithm to decode GMM-HMM driving behavior identification model, finding out output probability P (I | Lambda) of model parameter Lambda to output driving behavior sequence I, selecting the state with maximum probability as current driving behavior state value, so as to obtain identification driving behavior sequence value I; (2) extracting a driving behavior sequence with a specified length by adopting a sliding time window method; taking the number of the driving behavior states in each time window as the basis for calculating the probability change of the driving behavior; (3) processing subsequent observation data of the traffic vehicle subjected to the lane changing operation; (4) the data of an I-80 road section and a US-101 road section in an NGSIM data set are used for training and testing; after data preprocessing is completed, GMM data clustering and HMM training work is carried out. The invention can improve the accuracy of the identification result of the continuous driving behavior.

Description

Intelligent networking automobile behavior identification method
Technical Field
The invention relates to the technical field of intelligent networked automobiles, in particular to an intelligent networked automobile behavior identification method.
Background
With the rapid development of the road transportation industry and the rapid increase of the motor vehicle holding amount in China, the production and life of people are seriously affected by property loss and casualties caused by frequent traffic safety accidents. The driving behavior of the driver plays an important role in road traffic safety, so that the driver behavior needs to be correctly identified. But the driving behavior is an consciousness form and is difficult to be directly obtained through instrument measurement. Therefore, it is important to accurately and early recognize the behavior of the driver for the development of a safe driving system and the development of intelligent traffic.
At present, it is still a common practice to identify driving behaviors by using HMM (hidden markov model), and most of them identify driving behaviors of samples by training a plurality of driving behavior classifiers and using a maximum likelihood method. However, the maximum likelihood method can only provide the behavior with the maximum likelihood, but cannot express the possibility of the behavior through probability, and only can realize the problem of identifying a single driving behavior sample, and the accuracy of the identification result of a continuous driving behavior sample is low.
Disclosure of Invention
The invention aims to provide an intelligent networking automobile behavior identification method which is used for improving the accuracy of identification results of continuous driving behavior samples.
The technical scheme adopted by the invention for solving the technical problems is as follows: an intelligent networking automobile behavior identification method is characterized by comprising the following steps:
(1) decoding the GMM-HMM driving behavior identification model by using a Viterbi algorithm, solving the output probability P (I | lambda) of the model parameter lambda to the output driving behavior sequence I, and selecting the state with the maximum probability as the current driving behavior state value so as to obtain the identification driving behavior sequence value I;
(2) extracting a driving behavior sequence with a specified length by adopting a sliding time window method; taking the number of the driving behavior states in each time window as the basis for calculating the probability change of the driving behavior;
Figure BDA0002584942470000021
in the formula Pi(i-1, 2,3) respectively represents a left lane change,Probability of straight line driving and right lane changing; xiiThe number of hidden states in a sliding time window is i; n is the specified length of the sliding time window sequence;
(3) processing subsequent observation data of the traffic vehicle which carries out lane changing operation, outputting a sequence according to driving behaviors, monitoring the observation data from a lane changing initial point if the lane changing behavior appears in a track sequence, and when the lateral displacement of the lane changing vehicle is more than one lane width, considering that the lane changing is finished, and processing the subsequent observation data of the lane changing vehicle at the moment;
(4) the data of an I-80 road section and a US-101 road section in an NGSIM data set are used for training and testing; extracting the lateral displacement speed and the lateral displacement variation deviation of the vehicle from the data set as original data for training, and performing data smoothing filtering and abnormal value processing on the continuous variation parameters; after data preprocessing is completed, GMM data clustering and HMM training work is carried out.
Further, when a driving behavior sequence with a specified length is extracted, the width of a sliding time window is set to be 1s, the time step length is set to be 0.1s, 10 hidden states are collected by the instant window every time, and 1 sampling point is moved forward every time when the instant window is updated; if the length of the truncated sequence is N sampling points, the information of (N-1) nodes in the adjacent extracted 2 sequences is the same, the sampling frequency is 10Hz, and if the historical time domain of the input sequence is Tp, the sequence length is N-10 Tp.
Further, for the track changing track in the NGSIM data set, a track changing starting point and a characterization parameter corresponding to the starting point moment need to be extracted; in order to avoid the misjudgment and interference of the small-amplitude transverse displacement of the vehicle or the continuous lane change to the lane change starting point, the transverse displacement and the track curvature of the vehicle are used as the judgment standard for judging whether the lane change is carried out by the vehicle, and for the single complete lane change process, the transverse displacement and the track curvature at the lane change starting point and the lane change ending point meet the following requirements:
Figure BDA0002584942470000031
wherein y (n) is the lateral position of the vehicle at time n; t is the lane change time; l is the lane width; w is a compensation value; and theta (n) is the curvature of the vehicle running track at the moment n.
Further, the driving behavior data is fitted using a complete covariance matrix.
The invention has the beneficial effects that: the invention provides an intelligent networking automobile behavior identification method which comprises the following steps that (1) GMM is used for carrying out cluster analysis on observation characteristic values under all driving behaviors, the driving behaviors are divided according to cluster results, and statistical description under all driving behaviors is realized; (2) a multi-driving behavior identification model is established, and compared with the currently common multi-classification driving behavior identification model, the model introduces probability as result output. (3) In the process of identifying the driving behaviors, the positions of the lane-changing vehicles are subjected to data processing, so that the identification of the continuous driving behaviors is realized.
Drawings
FIG. 1 is a schematic diagram of a sliding window driving behavior sequence extraction;
FIG. 2 is a diagram of observation data processing criteria;
FIG. 3 is a diagram of observed data processing logic decision;
FIG. 4a is study section I-80;
FIG. 4b is the study section US-101;
FIG. 5a is a right switch track cluster;
FIG. 5b is a cluster of straight travel trajectories;
FIG. 5c is a left lane change track cluster;
FIG. 6 is a graph of Bayesian value information criterion values as a function of Gaussian mixture numbers;
FIG. 7a is a graph of the effect of a complete covariance matrix fit;
FIG. 7b is a graph of the effect of diagonal covariance matrix fitting;
FIG. 8a is a graph of lateral displacement velocity versus lateral displacement;
FIG. 8b is a probability density distribution diagram of GMM clustering driving behaviors according to observed data;
FIG. 9 is a schematic view of a continuous driving behavior trajectory;
FIG. 10a is a schematic diagram of probability output of continuous driving behavior under recognition of a random forest prediction algorithm and a GMM-HMM model;
FIG. 10b is a result of continuous driving behavior recognition under the recognition of the random forest prediction algorithm and the GMM-HMM model;
FIG. 11 shows the recognition result of the GMM-HMM model for continuous driving behavior.
Detailed Description
The main contents of the present invention will be described in detail below with reference to fig. 1 to 11.
(1) And decoding the GMM-HMM driving behavior identification model by using a Viterbi algorithm, solving the output probability P (I | lambda) of the model parameter lambda to the output driving behavior sequence I, and selecting the state with the maximum probability as the current driving behavior state value so as to obtain the identification driving behavior sequence value I.
(2) And (3) extracting a driving behavior sequence with a specified length by adopting a sliding time window method, as shown in figure 1. The width of the sliding time window is set to be 1s, the time step length is set to be 0.1s, the instant window collects 10 hidden states each time, and the sampling points are 1 forward sampling point each time. And if the length of the interception sequence is n sampling points, the information of (n-1) nodes in the adjacent extracted 2 sequences is the same. The sampling frequency is 10Hz, and if the historical time domain of the input sequence is Tp, the sequence length is N-10 Tp.
And taking the number of the driving behavior states in each time window as the basis for calculating the probability change of the driving behavior.
Figure BDA0002584942470000041
In the formula, Pi(i ═ 1,2,3) respectively represent the probability of a left lane change, a straight-line drive, and a right lane change; xiiThe number of hidden states in a sliding time window is i; n is the specified length of the sliding time window sequence.
(3) Lane change vehicle observation data processing
In the model training process, all observation data are concentrated to the current lane, so that after the lane change of the traffic vehicle at the current side is finished, the lateral displacement observation sequence can continuously approach to one lane width, the hidden Markov model generates wrong identification results, and meanwhile, continuous multi-behavior identification cannot be realized, so that the subsequent observation data of the traffic vehicle for performing the lane change operation are processed, as shown in FIG. 2.
And according to the driving behavior output sequence, if the track sequence has a lane changing behavior, monitoring the observation data from the initial lane changing point, and when the lateral displacement of the lane changing vehicle is greater than one lane width, considering that the lane changing is finished, and starting to process the subsequent observation data of the lane changing vehicle.
Therefore, the continuous driving behavior identification of the traffic vehicle can be realized by adjusting the lane change observation data.
(4) Data pre-processing
The data of the I-80 and US-101 segments in the NGSIM dataset are used for training and testing. Extracting the lateral displacement speed and the lateral displacement variation deviation of the vehicle from the data set as original data for training, and performing data smoothing filtering and abnormal value processing on the continuous variation parameters; after data preprocessing is completed, GMM data clustering and HMM training work is carried out.
And for the track changing track in the NGSIM data set, extracting a track changing starting point and a characterization parameter corresponding to the starting point moment. In order to avoid the misjudgment and interference of the small-amplitude transverse displacement of the vehicle or the continuous lane change to the lane change starting point, the transverse displacement and the track curvature of the vehicle are used as the judgment standard for judging whether the lane change is carried out by the vehicle, and for the single complete lane change process, the transverse displacement and the track curvature at the lane change starting point and the lane change ending point meet the following requirements:
Figure BDA0002584942470000051
wherein y (n) is the lateral position of the vehicle at time n; t is the lane change time; l is the lane width; w is a compensation value; and theta (n) is the curvature of the vehicle running track at the moment n. The transverse displacement and the track curvature at the starting point and the ending point in the lane changing process can be respectively limited through the formula, and the track data can be classified according to the limitation. The discrimination of the left and right lane changing sequence is determined according to the transverse coordinates of the start and end positions of the sequence.
137 pieces of left lane changing, 306 pieces of right lane changing and 520 pieces of straight line are extracted, and the training sample tracks are shown in fig. 5a, 5b and 5 c. In order to extract the characteristic value of the vehicle side upward direction, the intersection point of the lane changing track and the lane line is used as a reference point, so that all the sample tracks intersect at the point.
Based on the extracted driving behavior samples, a training sample with 300 sets of observation sequences was created, wherein,
Figure BDA0002584942470000061
i.e. the length of each set of observation sequence samples is 60. And setting initial parameters of the hidden Markov model, and setting an initial value of the initial probability distribution pi by a mean value method. In the actual driving process, the driver generally does not switch between the left lane change and the right lane change in a very short time, so the switching probability between the left lane change and the right lane change in the state transition probability matrix a is set to be 0, and the rest values are uniformly distributed.
The GMM is used to fit the probability distribution of the output variable in each hidden state. Fig. 6 shows that the bayesian information criterion value varies with the number of gaussian mixtures, and the number of gaussian distributions of the GMM is set to be M-3 according to the characteristics of the training data and the actual recognition effect.
The currently commonly used covariance matrix forms include a diagonal covariance matrix and a complete covariance matrix, in order to determine the optimal covariance matrix form suitable for the GMM model, two covariance matrix forms are respectively adopted as the covariance matrix of the GMM, and the fitting results are shown in fig. 7a and 7 b.
As can be seen from fig. 7a and 7b, the complete covariance matrix can better fit the driving behavior data to the diagonal covariance matrix, and therefore the GMM covariance matrix form employs the complete covariance matrix.
Fig. 8a and 8b are the result of the GMM clustering the driving behaviors according to the observation data and the corresponding probability density distribution diagram.
As can be seen from the analysis of FIG. 8, the lateral displacement speed data of the right lane-changing behavior is within the range of [ -2, 0], and the lateral displacement data is within the range of [0, -4 ]; the lateral displacement speed data of the left lane changing behavior is in the interval of [0, 2], and the lateral displacement data is in the interval of [0, 4 ]; the characteristic data of the straight line driving is in the coordinate origin interval. In the case of behavior recognition, the change in the probability of the respective behavior can be understood in practice as the change in the probability of the individual ellipse when a point moves in the lateral offset-lateral offset velocity plane.
In order to verify the accuracy of the driving behavior recognition model of the preceding vehicle for recognizing the continuous driving behaviors, the real vehicle acquires a continuous section of 120s of driving behavior data for verification, and the driving track of the real vehicle is shown in fig. 9. The driving behavior is recognized by using a GMM-HMM recognition model, as shown in fig. 10a and 10b, and compared with a Random Forest (RF) classification algorithm.
From an analysis of FIGS. 10a and 10b, it can be seen that: the random forest prediction algorithm has high error rate and long prediction delay in the identification process. Compared with a random forest prediction algorithm, the GMM-HMM front vehicle driver behavior recognition model provided by the invention correctly recognizes various driving behaviors, and the accuracy is obviously improved. Meanwhile, in the aspect of prediction delay, the prediction delay of the GMM-HMM model is shorter than that of a random forest prediction algorithm, and the change of driving behaviors can be recognized more quickly.
In order to verify the universality of the recognition model, 300 groups of verification samples corresponding to the left lane changing, the right lane changing and the straight-line driving behaviors are imported into the trained GMM-HMM model for behavior recognition, and the recognition result is shown in FIG. 11.
From the results in fig. 11, it can be seen that the average correct recognition rate of the GMM-HMM front-vehicle driver behavior recognition model proposed herein reaches 94%, while the average prediction delay is 176ms in terms of the prediction delay, which enables faster recognition of the change in driving behavior, indicating that the model is feasible for driving behavior recognition.
The invention has the following advantages:
(1) the GMM is used for carrying out cluster analysis on the observation characteristic values under each driving behavior, the driving behaviors are divided according to the cluster result, and statistical description under each driving behavior is realized;
(1) a multi-driving behavior identification model is established, and compared with the currently common multi-classification driving behavior identification model, the model introduces probability as result output.
(3) In the process of identifying the driving behaviors, the positions of the lane-changing vehicles are subjected to data processing, so that the identification of the continuous driving behaviors is realized.

Claims (4)

1. An intelligent networking automobile behavior identification method is characterized by comprising the following steps:
(1) decoding the GMM-HMM driving behavior identification model by using a Viterbi algorithm, solving the output probability P (I | lambda) of the model parameter lambda to the output driving behavior sequence I, and selecting the state with the maximum generation probability as the current driving behavior state value so as to obtain the identification driving behavior sequence value I;
(2) extracting a driving behavior sequence with a specified length by adopting a sliding time window method; taking the number of the driving behavior states in each time window as the basis for calculating the probability change of the driving behavior;
Figure FDA0002584942460000011
in the formula Pi(i ═ 1,2,3) respectively represent the probability of a left lane change, a straight-line drive, and a right lane change; xiiThe number of hidden states in a sliding time window is i; n is the specified length of the sliding time window sequence;
(3) processing subsequent observation data of the traffic vehicle which carries out lane changing operation, outputting a sequence according to driving behaviors, monitoring the observation data from a lane changing initial point if the lane changing behavior appears in a track sequence, and when the lateral displacement of the lane changing vehicle is more than one lane width, considering that the lane changing is finished, and processing the subsequent observation data of the lane changing vehicle at the moment;
(4) the data of an I-80 road section and a US-101 road section in an NGSIM data set are used for training and testing; extracting the lateral displacement speed and the lateral displacement variation deviation of the vehicle from the data set as original data for training, and performing data smoothing filtering and abnormal value processing on the continuous variation parameters; after data preprocessing is completed, GMM data clustering and HMM training work is carried out.
2. The method for identifying the behaviors of the intelligent networked automobile according to claim 1, wherein when a driving behavior sequence with a specified length is extracted, the width of a sliding time window is set to be 1s, the time step length is set to be 0.1s, 10 hidden states are collected by the instant window every time, and 1 sampling point is moved forward every time when the instant window is updated; if the length of the truncated sequence is N sampling points, the information of (N-1) nodes in the adjacent extracted 2 sequences is the same, the sampling frequency is 10Hz, and if the historical time domain of the input sequence is Tp, the sequence length is N-10 Tp.
3. The method for identifying the behaviors of the intelligent networked automobile according to claim 1, wherein a lane change starting point and a characterization parameter corresponding to the starting point moment are extracted from a lane change track in an NGSIM data set; in order to avoid the misjudgment and interference of the small-amplitude transverse displacement of the vehicle or the continuous lane change to the lane change starting point, the transverse displacement and the track curvature of the vehicle are used as the judgment standard for judging whether the lane change is carried out by the vehicle, and for the single complete lane change process, the transverse displacement and the track curvature at the lane change starting point and the lane change ending point meet the following requirements:
Figure FDA0002584942460000021
wherein y (n) is the lateral position of the vehicle at time n; t is the lane change time; l is the lane width; w is a compensation value; and theta (n) is the curvature of the vehicle running track at the moment n.
4. The method as claimed in claim 1, wherein the driving behavior data is fitted using a complete covariance matrix.
CN202010684027.2A 2020-07-15 2020-07-15 Intelligent networking automobile behavior identification method Pending CN111882923A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010684027.2A CN111882923A (en) 2020-07-15 2020-07-15 Intelligent networking automobile behavior identification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010684027.2A CN111882923A (en) 2020-07-15 2020-07-15 Intelligent networking automobile behavior identification method

Publications (1)

Publication Number Publication Date
CN111882923A true CN111882923A (en) 2020-11-03

Family

ID=73154860

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010684027.2A Pending CN111882923A (en) 2020-07-15 2020-07-15 Intelligent networking automobile behavior identification method

Country Status (1)

Country Link
CN (1) CN111882923A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112232525A (en) * 2020-12-15 2021-01-15 鹏城实验室 Driving mode characteristic construction and screening method and device and storage medium
CN112560782A (en) * 2020-12-26 2021-03-26 浙江天行健智能科技有限公司 Vehicle lane changing behavior identification method based on random forest algorithm
CN113704682A (en) * 2021-09-01 2021-11-26 清华大学 Traffic information credible identification method based on driving control data analysis
CN114124472A (en) * 2021-11-02 2022-03-01 华东师范大学 Vehicle-mounted network CAN bus intrusion detection method and system based on GMM-HMM

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109460023A (en) * 2018-11-09 2019-03-12 上海理工大学 Driver's lane-changing intention recognition methods based on Hidden Markov Model
CN109471436A (en) * 2018-11-09 2019-03-15 上海理工大学 Based on mixed Gaussian-Hidden Markov Model lane-change Model Parameter Optimization method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109460023A (en) * 2018-11-09 2019-03-12 上海理工大学 Driver's lane-changing intention recognition methods based on Hidden Markov Model
CN109471436A (en) * 2018-11-09 2019-03-15 上海理工大学 Based on mixed Gaussian-Hidden Markov Model lane-change Model Parameter Optimization method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
武馨宇: "智能车可行驶区域建模及驾驶辅助信息显示技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112232525A (en) * 2020-12-15 2021-01-15 鹏城实验室 Driving mode characteristic construction and screening method and device and storage medium
CN112560782A (en) * 2020-12-26 2021-03-26 浙江天行健智能科技有限公司 Vehicle lane changing behavior identification method based on random forest algorithm
CN113704682A (en) * 2021-09-01 2021-11-26 清华大学 Traffic information credible identification method based on driving control data analysis
CN113704682B (en) * 2021-09-01 2022-11-01 清华大学 Traffic information credible identification method based on driving control data analysis
CN114124472A (en) * 2021-11-02 2022-03-01 华东师范大学 Vehicle-mounted network CAN bus intrusion detection method and system based on GMM-HMM
CN114124472B (en) * 2021-11-02 2023-07-25 华东师范大学 Vehicle-mounted network CAN bus intrusion detection method and system based on GMM-HMM

Similar Documents

Publication Publication Date Title
CN111882923A (en) Intelligent networking automobile behavior identification method
CN103996287B (en) A kind of vehicle compulsory based on decision-tree model changes decision-making technique
CN113486822A (en) Surrounding vehicle track prediction method and system based on driving intention
CN104951764A (en) Identification method for behaviors of high-speed vehicle based on secondary spectrum clustering and HMM (Hidden Markov Model)-RF (Random Forest) hybrid model
CN108280484B (en) Driver acceleration characteristic online classification and identification method
CN112863245B (en) Vehicle track change real-time prediction method based on deep neural network
CN110569783A (en) Method and system for identifying lane changing intention of driver
CN111627132B (en) Driving inclination-based automobile driving characteristic evaluation and early warning method
CN113859266B (en) Method and system for predicting lane change track of unstructured road target vehicle
CN107274668A (en) A kind of congestion in road modeling method based on vehicle detection
CN112053589A (en) Target vehicle lane changing behavior adaptive identification model construction method
CN111599182B (en) Automatic lane changing behavior extraction method based on natural driving data
Bouhoute et al. On the application of machine learning for cut-in maneuver recognition in platooning scenarios
CN112528774A (en) Intelligent sorting system and method for unknown radar signals in complex electromagnetic environment
CN111340074B (en) Driver braking intention identification method based on improved RBF neural network
CN116279484A (en) Multi-wind-grid driver forced lane change prediction method integrating evolutionary game and machine learning
Zhang et al. Car-following behavior model learning using timed automata
Quek et al. Driver identification by driving style
CN117809458A (en) Real-time assessment method and system for traffic accident risk
CN109649399A (en) A kind of driver distraction's method of discrimination based on lateral running speed
CN112163521A (en) Vehicle driving behavior identification method, device and equipment
CN113761715B (en) Method for establishing personalized vehicle following model based on Gaussian mixture and hidden Markov
CN114707573A (en) Unsupervised driving style analysis method based on basic driving operation event
CN113306558B (en) Lane changing decision method and system based on lane changing interaction intention
CN113722982A (en) Automobile sensor attack detection and defense method based on multi-model fusion

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20201103