A method for analyzing lane-changing behavior of a driver in an intelligent network environment

Abstract

The application discloses a driver lane-changing behavior analysis method in an intelligent network connection environment, and specifically comprises the following steps: step 1: based on a vehicle-road cooperation system architecture, a three-dimensional high-precision traffic simulation scene is constructed; step 2: based on the three-dimensional high-precision traffic simulation scene, simulated driving is performed to obtain related data of a lane-changing failure scene and related data of a forced lane-changing scene; step 3: based on the related data of the lane-changing failure scene in step 2, an acceleration failure model is constructed, the braking behavior in the lane-changing failure scene is analyzed, the influence rate of the braking behavior on the lane-changing failure is obtained, and the influence rate is displayed through an Internet of Things environment for reference of a driver; and step 4: based on the related data of the forced lane-changing scene in step 2, a threshold-exceeding model is constructed, the threshold-exceeding model is used to analyze the collision risk in the forced lane-changing scene, and the analysis result is transmitted to a user. The application can improve the driving efficiency and safety of an intelligent network connection vehicle in a driving process.

Description

Technical Field

[0001] The invention belongs to the technical field of traffic engineering. Background Art

[0002] Intelligent connected vehicles will be the future of transportation, and connected environments have become a key research area in traffic engineering. Lane-changing behavior is fundamental to vehicle driving behavior. Inappropriate lane-changing behavior can lead to traffic congestion and collisions, significantly impacting the traffic environment. The analysis of lane-changing behavior is complex and closely related to traffic congestion and safety. However, due to the scarcity of vehicle trajectory data in connected environments, existing research on lane-changing behavior modeling focuses primarily on lane changes in traditional driving environments, without analyzing the driver's driving behavior and decision-making process in connected environments, which can lead to accidents. Summary of the Invention

[0003] Purpose of the invention: In order to solve the problems existing in the above-mentioned prior art, the present application provides a method for analyzing the lane-changing behavior of a driver in an intelligent connected environment.

[0004] Technical solution: The present invention provides a method for analyzing a driver's lane-changing behavior in an intelligent connected environment, the method comprising:

[0005] Step 1: Build a 3D high-precision traffic simulation scenario based on the vehicle-road cooperative system architecture;

[0006] Step 2: Performing simulated driving based on a three-dimensional high-precision traffic simulation scenario to obtain simulated driving results; the simulated driving results include relevant data of a lane change failure scenario and relevant data of a forced lane change scenario;

[0007] Step 3: Based on the relevant data of the lane change failure scenario in Step 2, an accelerated failure model is constructed. The braking behavior in the lane change failure scenario is analyzed to obtain the impact rate of braking behavior on lane change failure. The impact rate is displayed in the IoT environment for the driver's reference.

[0008] Step 4: Build a super-threshold model based on the relevant data of the forced lane change scenario in Step 2, use this model to analyze the collision risk in the forced lane change scenario, and transmit the analysis results to the user.

[0009] Furthermore, the vehicle-road cooperative system architecture includes an intelligent road terminal, an intelligent vehicle terminal, and an intelligent cloud. Based on the vehicle-road cooperative system architecture, a three-dimensional high-precision traffic simulation scene is constructed, specifically: the intelligent road terminal collects road traffic status information, the intelligent vehicle terminal collects vehicle status information; and the intelligent cloud manages the collected information.

[0010] Build a two-dimensional high-precision traffic simulation model, set traffic flow operation information based on road traffic status information and vehicle status information, build a two-dimensional high-precision traffic simulation scene, and provide a data interface for retrieving real-time vehicle status information;

[0011] Importing autonomous driving scenario development software into a 2D high-precision traffic simulation model, editing and correcting road geometry, setting various road scene elements, and inserting elements of the surrounding real-world environment to generate a 3D high-precision traffic simulation model;

[0012] The three-dimensional high-precision traffic simulation model is imported into a three-dimensional development engine to generate a three-dimensional high-precision traffic simulation scene.

[0013] Furthermore, the expression of the accelerated failure model in step 3 is as follows:

[0014] S(t|X)=S0[tEXP(βX)]

[0015] Wherein, X represents a set of covariate vectors, β represents the estimated parameter vector corresponding to the covariate, βX represents the baseline hazard function and baseline survival function when all covariates are zero, EXP() represents the exponential function, t represents the occurrence time of the input parameter, and S0 represents the known basic survival function; the covariates include first driving characteristic information, characteristic information of the first driver, and braking behavior; the first driving characteristic information includes speed, acceleration, vehicle coordinates, driving lane, and vehicle body angle; the characteristic information of the first driver includes: age, gender, driver's license type, driving experience, and educational background; the braking behavior includes initial speed, minimum deceleration time, average deceleration, maximum deceleration, and deceleration change rate;

[0016] The survival function of the accelerated failure model is:

[0017]

[0018] Among them, β′ i The vector representing the regression coefficients of driving characteristics and driving environment, x iq represents the vector of covariates of vehicle driving characteristics and connected driving environment, γ′ represents the vector of regression coefficients of driver characteristic information coefficients, z i A vector representing driver feature information;

[0019] The duration data of survival time follows the Weibull distribution, and the survival function of the Weibull distribution is expressed as follows:

[0020] S(t)=EXP{-EXP[-P(β0+β1X1+…+ β nX n )]t P}

[0021] where P represents the scale parameter of Weibull distribution, X i is the i th covariate, n is the number of covariates, β i represents the coefficient of covariate, i.e., the random parameter of driver, β i The expression of is as follows:

[0022] β i =X+ψz i +Γδ

[0023] where ψ represents the covariance matrix, δ represents the random error term subject to independent standard normal distribution, and Γ is a lower triangular symmetric matrix.

[0024] Further, the parameters in the accelerated failure model are estimated by maximum likelihood estimation.

[0025] Further, the step 4 constructs the super threshold model of lane changing collision risk based on the generalized Pareto distribution, and the expression of the super threshold model F u (y) is as follows:

[0026]

[0027] where u is the amount of exceeding threshold, ξ is the shape parameter, σ' is the scale parameter, u represents the threshold, and G(.) represents the generalized Pareto distribution.

[0028] Further, the second driving feature information and the second driver feature information are taken as the covariates of the super threshold model, the second driving feature information includes speed, lane remaining distance, front vehicle distance and rear vehicle distance, and the second driver feature information includes gender and age.

[0029] The lane changing interval time GT is taken as the collision risk index of the super threshold model, and the expression of GT is as follows:

[0030] GT=t2-t1

[0031] where t1 represents the time when the host vehicle reaches point P1, t2 represents the time when the candidate vehicle reaches point P1, and P1 is the position of the host vehicle on the target lane after completing forced lane changing.

[0032] The negative value of the lane changing interval time is taken as the independent variable for establishing the super threshold model.

[0033] The first threshold interval is obtained by using the average exceeding amount function, the second threshold interval is obtained by using the threshold stability diagram, the intersection of the two threshold intervals is taken as the selection range of the threshold, and the lower bound of the selection range is determined as the final threshold.

[0034] Beneficial effects: This application can analyze lane-changing behavior to provide early warnings to drivers, thereby improving the driving efficiency and safety of smart connected vehicles during driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is an application environment diagram of the present invention.

[0036] Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0037] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0038] This application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with a data processing platform provided on the server 104 via a network. The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0039] like Figure 2 As shown, this embodiment includes the following steps:

[0040] S1: Build a three-dimensional high-precision traffic simulation scenario based on the vehicle-road cooperative system architecture, which includes an intelligent road side, an intelligent vehicle side, and an intelligent cloud side.

[0041] It should be noted that the intelligent road terminal uses intelligent and networked equipment to realize the communication function between roads and vehicles. It obtains road traffic status information through various sensors or GPS devices, processes and transmits the information, and can also receive status information and traffic control information from the intelligent vehicle-mounted system in the vehicle; the intelligent vehicle terminal is used to obtain the vehicle status information of the current vehicle (including vehicle speed, acceleration, location information, etc.), and can also obtain vehicle status information of other vehicles in the current driving environment. The on-board intelligent devices mainly communicate with the intelligent road terminal and the intelligent cloud through wireless network communication to communicate data, so that the vehicle can dynamically adjust the driving status of the vehicle according to changes in the current driving environment, thereby improving the safety and operation efficiency of the vehicle; the intelligent cloud is responsible for the basic data management of the entire vehicle-road cooperative system, and adopts cloud computing, data communication, big data and other technologies to provide real-time road supervision and driving for the vehicle-road cooperative system. Assisted decision-making, cloud data storage and other functions, cloud computing can set corresponding traffic information records for traffic operation modes based on the collected system-related information (such as intersection signal information, surveillance video, parking fee standards, etc.), so as to provide data support and reliable professional advice to relevant government departments for road repair management, daily maintenance, road traffic planning, etc., thereby improving the intelligence level of the entire vehicle-road cooperative system. The data communication system is used to support the data interaction between people, vehicles and roads in the entire system, and realize real-time communication between smart car terminals, smart car terminals, and smart car terminals, so as to ensure the safety and stability of system operation. Among them, the smart road terminal mainly includes a traffic road network based on a real environment, and the smart vehicle terminal mainly includes various vehicles operating in a virtual reality environment. The smart cloud is mainly realized through the background program pre-set by traffic simulation technology and three-dimensional development technology, collecting vehicle data and roadside data in all directions and at all times, and processing and analyzing the data. Specifically:

[0042] Collect road traffic status information, vehicle status information and manage the collected information through the intelligent road terminal, intelligent vehicle terminal and intelligent cloud terminal respectively;

[0043] On the intelligent cloud, a two-dimensional high-precision traffic simulation model is constructed based on open source traffic simulation software, traffic flow operation information is set based on the road traffic status information and vehicle status information, a two-dimensional high-precision traffic simulation scene is constructed, and a data interface is provided for retrieving real-time vehicle status information;

[0044] Based on the 2D high-precision traffic simulation model, the model is imported into the autonomous driving scenario development software through conversion of the OpenDrive high-precision map format. Based on the 2D road network, the road geometry is edited and modified, various road scene elements are set, and elements of the surrounding real environment are inserted to generate a 3D high-precision traffic simulation model.

[0045] Importing the three-dimensional high-precision traffic simulation model based on a three-dimensional development engine to generate a three-dimensional high-precision traffic simulation scene;

[0046] The two-dimensional high-precision traffic simulation scene and the three-dimensional high-precision traffic simulation scene are dynamically coupled and connected through traffic simulation technology and three-dimensional development technology, so that the traffic flow operation information in the two-dimensional high-precision traffic simulation scene is mapped to the three-dimensional high-precision traffic simulation model, and the final three-dimensional high-precision traffic simulation scene is generated.

[0047] S2: Perform simulated driving based on the three-dimensional high-precision traffic simulation scene to obtain simulated driving results.

[0048] The specific steps are:

[0049] A main vehicle is set in the three-dimensional high-precision traffic simulation scene, and a driving simulation device based on virtual reality is connected to the main vehicle;

[0050] Designing a plurality of lane-changing scenarios based on the three-dimensional high-precision traffic simulation scenario, and implementing a driving simulation function based on the plurality of lane-changing scenarios to obtain a first simulated driving result;

[0051] Based on the first simulated driving result, relevant data corresponding to the lane change failure scenario and the forced lane change scenario are extracted and classified and stored to obtain a second simulated driving result, which is the final simulated driving result. The relevant data includes driving status information, road condition information, surrounding driving environment information and front vehicle information, as well as characteristic data of the simulated driving participants.

[0052] S3: Select relevant data of the lane change failure scenario in the simulated driving results, analyze the braking behavior in the lane change failure scenario through a pre-built accelerated failure model, and obtain a first analysis result. The first analysis result is the impact rate of the braking behavior on the lane change failure. The first analysis result is displayed through the Internet of Things environment for the driver's reference.

[0053] It should be noted that the relevant data includes vehicle trajectory data and simulated driver characteristic data. The vehicle trajectory data includes at least one of the following: speed, acceleration, position, driving lane, and vehicle body angle. The characteristic data of the simulated driving participant includes at least one of the following: age, gender, driver's license type, driving experience, and educational background.

[0054] Preprocessing the vehicle driving trajectory data to extract vehicle speed change data during the braking process that can be used as braking behavior indicators to describe the vehicle's form characteristics, including initial speed, minimum deceleration time, average deceleration, maximum deceleration, and deceleration change rate;

[0055] According to the existing method, the accelerated failure model is constructed:

[0056] S(t|X)=S0[tEXP(βX)]

[0057] Where X represents a vector of a set of covariates, β represents an estimated parameter vector corresponding to the covariates, βX represents a baseline hazard function and a baseline survival function when all covariates are zero (X=0), EXP() represents an exponential function, t represents the occurrence time of an input parameter, and S0 represents a known basic survival function.

[0058] The time taken by the driver to reduce the host vehicle from the initial speed to the minimum speed is selected as the duration variable, and the speed survival time is defined as t iq , that is, the time interval between the start of the lane changing behavior of the host vehicle and the termination of the lane changing behavior, the speed survival time is modeled based on the accelerated failure model, and the survival function of the accelerated failure model is obtained:

[0059]

[0060] Where S0 is a known basic survival function, β' i represents a vector of regression coefficients of driving characteristics and driving environment, x iq represents a vector of covariates of vehicle driving characteristics and networked driving environment, γ' represents a vector of regression coefficients of driver characteristic information coefficients, and z i represents a vector of driver characteristic information.

[0061] The accelerated failure model assumes that the logarithmic conversion linear function of the survival time and the explanatory variable has a linear relationship:

[0062]

[0063] Where ln(t iq ) represents the survival time of the accelerated failure model, and the random error term ε iq is defined to follow an independent and identically normal distribution with a mean of zero and a standard deviation σ.

[0064] The survival time of the speed change is modeled based on the Weibull distribution, and the probability density function of the Weibull distribution is:

[0065] f(t)=λP×(λt) P-1 EXP[-(λt P )]

[0066] Where λ represents the location parameter of the Weibull distribution, and P represents the scale parameter of the Weibull distribution.

[0067] The location parameter is represented as the mean of the distribution through the covariate, as follows:

[0068] λ = EXP[-P(β0+β1X1+…+β n X n )]

[0069] where X i represents the covariates, β i represents the coefficients of the covariates.

[0070] Based on the above formula, the final survival function of the Weibull distribution is expressed as follows:

[0071] S(t = EXP{-EXP[-P(β0+β1X1+…+β n X n )]t P}

[0072] where: β0+β1X1+…+β n X n = βX represents the baseline risk function and baseline survival function when all covariates are zero (X = 0), t P represents the occurrence time of the parameter.

[0073] In the risk-based duration model, a random parameter model is introduced, and β i is defined as a random parameter variable of the driver as follows:

[0074] β i = X + ψz i + Γδ

[0075] where δ represents a random error term that follows an independent standard normal distribution, and Γ is a lower triangular symmetric matrix, i.e., a Cholesky matrix, the elements in Γ are used to calculate the standard deviation or variance of the random parameter, and the correlation existing in the coefficients is considered.

[0076] Based on Cholesky decomposition, the variance-covariance matrix of the random parameter is expressed as follows:

[0077] V = ΓΓ'

[0078] where the diagonal elements of the variance-covariance matrix represent the square values of the standard deviations of the correlated random parameters.

[0079] The parameters of the accelerated failure model are estimated by maximum likelihood estimation, i.e., the coefficients of each covariate are estimated to obtain the parameter value with the maximum likelihood function value.

[0080] Assuming that f(t|β) is the probability density function at time t, the maximum likelihood estimation of the parameter estimation of the accelerated failure model is based on the estimation of the coefficients of each covariate, so as to obtain the parameter value with the maximum likelihood function value, therefore, the input data is composed of multiple covariate coefficients, therefore, β=(β0,…,β n ) is the coefficient of each covariate, β0, β1 or other can be selected as input respectively, and the above β i is the random parameter of the driver of the operating variable, and for a sample of n, the likelihood function and the log-likelihood function are respectively:

[0081]

[0082]

[0083] The partial derivative of the unknown covariate coefficient is taken, and the coefficient value to be estimated is obtained by solving The estimation value satisfies:

[0084]

[0085] According to the vehicle speed survival time data obtained from the lane change failure scene experiment, the parameter estimation of the covariate coefficient of the accelerated failure model is carried out by using the above method, so as to obtain the parameter value with the maximum likelihood function value, and in the case of the parameter value, the probability of the sample data occurring is maximum.

[0086] Based on the Akaike information criterion, the Bayesian information criterion and the Mallows pseudo R square, the complexity and the fitting data goodness of the random parameter model are evaluated, the related data of the lane change failure scene in the simulation driving result is selected in the above step, the random parameter model is analyzed based on the related data, the randomness of the related data is judged based on the evaluation result, the data meeting the requirement is selected according to the randomness of the related data, the braking behavior index data of the accelerated failure model is extracted, the braking behavior index data is taken as the input data of the accelerated failure model, and the braking behavior characteristic analysis result is output, so as to realize the braking behavior characteristic analysis, and the braking behavior characteristic analysis result of the accelerated failure model is realized based on the trained model, and the working principle is the prior art. Specifically, the calculation formula of the Akaike information criterion is:

[0087] AIC=2k-2ln(L)

[0088] Wherein, k is the number of parameters, and L is the likelihood function.

[0089] The calculation formula of the Bayesian information criterion is:

[0090] BIC=kln(n)-2ln(L)

[0091] Based on the above-mentioned accelerated failure model, the braking behavior characteristics are analyzed to obtain a final analysis result, so as to predict the braking behavior of the lane change failure scenario, obtain a first analysis result, and display the first analysis result through the Internet of Things environment.

[0092] S4: Select relevant data of the forced lane change scenario in the simulated driving results, analyze the collision risk in the forced lane change scenario through a super-threshold model, and obtain a second analysis result, where the second analysis result is the collision risk. Send the second analysis result to the user end to prompt the driver to respond to the second analysis result.

[0093] Based on the vehicle trajectory data, driving characteristics and driver's personal characteristics of the main vehicle during the forced lane change process, a super-threshold model is constructed based on the generalized Pareto distribution of lane change collision risk to analyze the collision risk of the main vehicle in the forced lane change scenario. The lane change collision risk of the main vehicle is analyzed based on the driving environment and the driver's personal characteristics. The super-threshold model F u The expression for (y) is as follows:

[0094]

[0095] where y is the amount by which it exceeds the threshold, ξ is the shape parameter, σ' is the scale parameter, u represents the threshold, G(.) represents the generalized Pareto distribution, and ~ represents approximation.

[0096] Selection of parameters for the super-threshold model:

[0097] Select covariates: Extract the speed of the main vehicle, remaining lane distance, distance between the leading vehicle and the trailing vehicle during the time period of the forced lane change as different driving characteristics, and obtain the driver's characteristic information as a covariate. For a categorical variable with n categories of driver characteristic information, it is necessary to introduce n dummy variables, including: age (young), age (old), and gender (female). The covariates are introduced into the scale parameter σ' of the model:

[0098]

[0099] Among them, σ' i is the scale parameter of the generalized Pareto distribution in the connected driving environment, σ'0 is the intercept term, and β i and γ i are the covariate vectors of estimated parameters and driving characteristic information, and Ω i are the covariate vectors of estimated parameters and driver characteristic information, β1 represents the estimated parameter of speed, β2 represents the estimated parameter of lane remaining distance, β3 represents the estimated parameter of front vehicle distance, β4 represents the estimated parameter of rear vehicle distance, an estimated parameter indicating that the driver is young, an estimated parameter indicating that the driver is old, an estimated parameter indicating that the driver is female, γ1 represents speed, γ2 represents lane remaining distance, γ3 represents front vehicle distance, γ4 represents rear vehicle distance, Ω1 represents that the driver is young, Ω2 represents that the driver is old, and Ω3 represents that the driver is female.

[0100] Selection of the collision risk index: defining P as the position of the host vehicle on the target lane after completing the forced lane change, then the lane change interval time GT is equal to the time elapsed between the time t1 at which the host vehicle reaches P and the time t2 at which the rear vehicle reaches P:

[0101] GT = t2 - t1.

[0102] Selection of the threshold value of the over-threshold model includes an average excess function and a threshold stability diagram, and the average excess function is:

[0103] The average excess function e(u) based on the generalized Pareto distribution G(x; σ u , ξ) is calculated as follows:

[0104]

[0105] where X represents a random variable subject to independent identical distribution, u represents a certain sufficiently large threshold value set, σ u is a scale parameter, ξ represents a shape parameter, e(u) is a linear function with respect to u, and for a set of observation values X1, X2, …, X n , u > u0, u0 represents a random threshold value, and the average excess function can be calculated by the following formula:

[0106]

[0107] where n u is the number of observation values exceeding the threshold value u, and if the distribution of the average excess function is approximately the generalized Pareto distribution G(x; σ u , ξ) for the threshold value, then for the threshold value, the average excess function of the sample observation value sequence will be approximately a straight line, and the point set {(u, e n (u)): u < x max} is defined as the average residual life diagram, x max represents the maximum observation value in X1, X2, …, X n , and by selecting u0 > 0 as the threshold value, e n (u) is approximately linear with respect to u ≥ u0.

[0108] The threshold stability diagram method is as follows: The threshold stability diagram method is to determine the effect of changes in the value of u on the estimated value. Assuming that the generalized Pareto distribution is suitable for modeling observations exceeding the threshold, the super-threshold part of the threshold should also follow the same distribution. In this case, the shape parameter ξ of the two generalized Pareto distributions should be the same, and the scale parameter σ u =σ u0 +ξ(u-u0) can vary with u unless ξ=0. Reparameterize the scale parameter of the distribution to σ * =σ u +ξu, which will produce a modified scale parameter σ for u * The criterion for the threshold stability diagram is: if u0 is a valid sample of excess valid thresholds following a generalized Pareto distribution, then ξ and σ are expected to * The estimated value of should remain constant and above the threshold u0. Therefore, ξ and σ are * About the graph of u0, and choose so that ξ and σ * Maintain a constant minimum u0 as the threshold of the model.

[0109] After obtaining the threshold selection range obtained from the average excess function and the threshold stability diagram, the value ranges are intersected to obtain the final threshold selection range, and the final threshold is determined by selecting the lower bound of the value range.

[0110] After the threshold of the super-threshold model is determined, the parameters of the super-threshold model are usually estimated using maximum likelihood estimation, probability weight moment method or L-moment method, and the parameters include shape parameters and scale parameters.

[0111] The maximum likelihood estimation includes:

[0112] Define x=(x1,…,x n ) is a set of independent and identically distributed observation sequences, and its overall distribution function F obeys the generalized Pareto distribution with shape parameter ξ and scale parameter σ, then the probability density function of the observation population is:

[0113]

[0114] Get the log-likelihood function:

[0115]

[0116] Among them, x i Need to satisfy the restrictions of its domain:

[0117] x i ∈D(σ,ξ)

[0118]

[0119] When the value of ξ tends to 0, the generalized Pareto distribution becomes the exponential distribution, and the likelihood function is obtained as:

[0120]

[0121] The maximum likelihood estimate of σ is obtained as

[0122] When ξ > 0.5, the parameters are estimated by the maximum likelihood estimate, and when n→∞, the inverse of the observed information matrix in the maximum likelihood estimate is first calculated to obtain the maximum likelihood estimate of (σ, ξ) and the asymptotic covariance matrix of and the following formula is obtained:

[0123]

[0124]

[0125] where N2(μ,Σ) represents a two-dimensional Gaussian distribution with covariance matrix Σ and mean vector μ, M -1 is the covariance matrix of the maximum likelihood estimate and ;

[0126] The covariance matrix of is approximated as the following formula:

[0127]

[0128] where F is the distribution function, the horizontal bar is the average value, and the hat is the estimated value.

[0129] The quantile x p is calculated by the delta method, and its standard error and confidence interval are:

[0130]

[0131] where, is the value of at the point , which includes the uncertainty in the estimate ;

[0132] When the shape parameter ξ≠0, we have:

[0133]

[0134] When the shape parameter ξ=0, we have:

[0135]

[0136] x p Assuming it to be a constant, the likelihood function of the shape parameter ξ is obtained based on the above formula.

[0137] The probability weight moment method includes:

[0138] When μ = 0, that is, when G = G(x; P; ξ), based on the characteristics of the probability weight moment, we can get

[0139]

[0140] Where: r is the selected input parameter;

[0141] Special:

[0142]

[0143] but:

[0144]

[0145] Use the above sample probability weights to replace the overall probability weights ω0 and ω1 to obtain the probability weight estimates of the shape parameter ξ and scale parameter σ and When ξ>0, the efficiency of probability weight moment estimation is similar to maximum likelihood estimation. If the parameters are related to time or are affected by other explanatory variables, probability weight moment estimation is no longer applicable.

[0146] The L-moment method includes:

[0147] Define the probability distribution function of the random variable X to obey the generalized Pareto distribution G(x;σ,ξ)(ξ<1), and calculate the first two order L moments:

[0148] λ1=σ / (1-ξ)

[0149] λ2=σ / ((1-ξ)(2-ξ))

[0150] The first two-order sample L moments of the excess are obtained by the following formula:

[0151]

[0152]

[0153] in: and represents the 1st and 2nd order L moment estimates, x i,n and x j,n Represents different random parameters of input;

[0154] Then the L-moment estimates of ξ and σ can be obtained by the following formula:

[0155]

[0156]

[0157] The maximum likelihood estimation is used to estimate the parameters of the super-threshold model in the connected driving environment and the basic driving environment, and the parameter estimation results are obtained;

[0158] Based on the parameter estimation results, a fitting test is performed on the super-threshold model in each driving environment to obtain a test result;

[0159] Based on the test results, i.e., the Pr(.) value, the collision risk analysis model in the forced lane change scenario is obtained, including:

[0160]

[0161] Where Z represents the maximum negative value of the lane change interval, H(.) is the generalized Pareto distribution, and Pr(.) represents the probability.

[0162] The collision risk analysis model is evaluated based on the Akaike Information Criterion and the Bayesian Information Criterion, and a final collision risk analysis model is obtained based on the evaluation results, wherein the evaluation method is the same as step S3.

[0163] Based on the final collision risk analysis model, the collision risk in the forced lane change scenario is analyzed to predict the collision risk of the forced lane change scenario, and a second analysis result is obtained. The second analysis result is sent to the user end to prompt the driver to respond to the second analysis result, thereby avoiding the collision risk and improving the driving efficiency and safety of the intelligent connected vehicle during driving.

[0164] In the above-mentioned method for analyzing the driver's lane changing behavior in an intelligent connected environment, a three-dimensional high-precision traffic simulation scene is constructed based on the vehicle-road cooperative system architecture, and the vehicle-road cooperative system architecture includes an intelligent road end, an intelligent vehicle end and an intelligent cloud end; simulated driving is performed based on the three-dimensional high-precision traffic simulation scene to obtain a simulated driving result; relevant data of the lane changing failure scene in the simulated driving result is selected, and the braking behavior in the lane changing failure scene is analyzed through a pre-built acceleration failure model to obtain a first analysis result, and the first analysis result is displayed through the Internet of Things environment to prompt the driver to respond to the first analysis result; relevant data of the forced lane changing scene in the simulated driving result is selected, and the collision risk in the forced lane changing scene is analyzed through a super-threshold model to obtain a second analysis result, and the second analysis result is sent to the user end to prompt the driver to respond to the second analysis result. This application can analyze the lane changing behavior to provide early warning prompts to the driver to improve the driving efficiency and safety of the intelligent connected vehicle during driving.

[0165] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0166] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0167] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. To avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. A method for analyzing driver lane-changing behavior in an intelligent connected environment, characterized in that: The method comprises: Step 1: Build a 3D high-precision traffic simulation scenario based on the vehicle-road cooperative system architecture; Step 2: Performing simulated driving based on a three-dimensional high-precision traffic simulation scenario to obtain simulated driving results; the simulated driving results include relevant data of a lane change failure scenario and relevant data of a forced lane change scenario; Step 3: Based on the relevant data of the lane change failure scenario in Step 2, an accelerated failure model is constructed. The braking behavior in the lane change failure scenario is analyzed to obtain the impact rate of braking behavior on lane change failure. The impact rate is displayed in the IoT environment for the driver's reference. Step 4: Build a super-threshold model based on the data related to the forced lane change scenario in Step 2. Use this model to analyze the collision risk in the forced lane change scenario and transmit the analysis results to the user. The expression of the accelerated failure model in step 3 is as follows: S(t|X)=S0[tEXP(βX)] Wherein, X represents a set of covariate vectors, β represents the estimated parameter vector corresponding to the covariate, βX represents the baseline hazard function and baseline survival function when all covariates are zero, EXP() represents the exponential function, t represents the occurrence time of the input parameter, and S0 represents the known basic survival function; the covariates include first driving characteristic information, characteristic information of the first driver, and braking behavior; the first driving characteristic information includes speed, acceleration, vehicle coordinates, driving lane, and vehicle body angle; the characteristic information of the first driver includes: age, gender, driver's license type, driving experience, and educational background; the braking behavior includes initial speed, minimum deceleration time, average deceleration, maximum deceleration, and deceleration change rate; The survival function of the accelerated failure model is: Among them, β' i The vector representing the regression coefficients of driving characteristics and driving environment, x iq represents the vector of covariates of vehicle driving characteristics and connected driving environment, γ' represents the vector of regression coefficients of driver characteristic information coefficients, z i A vector representing driver feature information; The duration data of survival time follows the Weibull distribution, and the survival function of the Weibull distribution is expressed as follows: S(t)=EXP{-EXP[-P(β0+β1X1+…+β n X n )]t P } Where P represents the scale parameter of the Weibull distribution, X i is the i-th covariate, n is the number of covariates, β i The coefficient of the covariate is also the driver random parameter, β i The expression is as follows: b i =μ+ψz i +Gd where δ represents a random error term that follows an independent standard normal distribution, and Γ is a lower triangular symmetric matrix.

2. The method for analyzing driver lane-changing behavior in an intelligent connected environment according to claim 1, characterized in that: The VIS system architecture includes an intelligent roadside terminal, an intelligent vehicle terminal, and an intelligent cloud. The three-dimensional, high-precision traffic simulation scenario constructed based on the VIS system architecture specifically involves: the intelligent roadside terminal collecting road traffic status information, the intelligent vehicle terminal collecting vehicle status information; and the intelligent cloud terminal managing the collected information. Build a two-dimensional high-precision traffic simulation model, set traffic flow operation information based on road traffic status information and vehicle status information, build a two-dimensional high-precision traffic simulation scene, and provide a data interface for retrieving real-time vehicle status information; Importing autonomous driving scenario development software into the two-dimensional high-precision traffic simulation model, and editing and correcting the road geometry based on the two-dimensional road network, setting various road scene elements, and inserting elements of the surrounding real environment to generate a three-dimensional high-precision traffic simulation model; The three-dimensional high-precision traffic simulation model is imported into a three-dimensional development engine to generate a three-dimensional high-precision traffic simulation scene.

3. The method for analyzing driver lane-changing behavior in an intelligent connected environment according to claim 1, characterized in that: Maximum likelihood estimation is used to estimate the parameters in the accelerated failure model.

4. The method for analyzing driver lane-changing behavior in an intelligent connected environment according to claim 1, characterized in that: In step 4, a super-threshold model is constructed based on the lane-changing collision risk of the generalized Pareto distribution. The super-threshold model F u The expression for (y) is as follows: Where y exceeds the threshold, ξ is the shape parameter, and σ , is the scale parameter, u represents the threshold, and G(.) represents the generalized Pareto distribution.

5. The method for analyzing driver lane-changing behavior in an intelligent connected environment according to claim 1, characterized in that: The second driving characteristic information and the second driver characteristic information are used as covariates of the super-threshold model, wherein the second driving characteristic information includes speed, remaining lane distance, distance between the preceding vehicle and the following vehicle; and the second driver characteristic information includes gender and age; The lane change interval time GT is used as the collision risk indicator of the super-threshold model. The expression of GT is: GT = t2 - t1 Where t1 represents the time when the main vehicle arrives at point P1, t2 represents the time when the waiting vehicle arrives at point P1, and P1 is the position of the main vehicle in the target lane after completing the forced lane change; The negative value of the lane-changing interval is used as the independent variable to establish the super-threshold model; Selection of the threshold of the super-threshold model: the average excess function is used to obtain the first threshold interval, the threshold stability diagram is used to obtain the second threshold interval, the intersection of the two threshold intervals is used as the threshold selection range, and the lower limit of the selection range is determined as the final threshold.

Images