A microscopic simulation method for non-motor vehicle traffic flow

Abstract

The present invention belongs to the technical field of traffic engineering and provides a microscopic simulation method for non-motor vehicle traffic flow, which includes the following steps: clustering the driving trajectories of non-motor vehicles to divide non-motor vehicle drivers; constructing a dynamic perception space model to obtain the position and speed information of non-motor vehicles; constructing a behavior decision-making model, loading the position and speed information of non-motor vehicles into the behavior decision-making model to obtain a behavior decision-making result; according to the obtained behavior decision-making result, using an improved social force model to calculate the force condition and real-time update the position and speed information of non-motor vehicles. The method provided by the present invention makes up for the deficiency that the existing model enables non-motor vehicle drivers to complete the behavior decision of crossing the line only through simple rules, is closer to the actual behavior decision situation, can provide a prerequisite for accurately matching the kinematic formula subsequently, and more realistically reproduces the driving trajectory of non-motor vehicles.

Description

Technical Field

[0001] The present invention belongs to the technical field of traffic engineering, and particularly relates to a microscopic simulation method for non-motor vehicle traffic flow. Background Art

[0002] On urban low-grade roads or the auxiliary roads of arterial roads, the motor vehicle lane and the non-motor vehicle lane are often separated by painting lines. In recent years, with the rapid popularization of electric bicycles, the internal non-motor vehicle flow is affected by complex interaction methods, and there are a large number of non-motor vehicles crossing the line and driving into the motor vehicle lane, which exacerbates the conflict between motor vehicles and non-motor vehicles. At the same time, affected by factors such as experience, psychology, and physiology, non-motor vehicle drivers have differences in traffic characteristics such as perceiving the surrounding environment, making behavioral decisions, and driving trajectories, resulting in complex interaction characteristics of the non-motor vehicle flow. The existing non-motor vehicle lane design can no longer meet the new non-motor vehicle mixed traffic mode.

[0003] The traffic system consists of people, vehicles, roads, and the environment. Diverse travel modes make the urban traffic system a complex large system. Since there are many individual elements in the system and they interact with each other, it is impossible to describe the traffic operation process in detail through a simple mathematical model during the explanation. Therefore, advanced computer microscopic simulation technology should be relied on for explanation and reproduction. This can not only truly reproduce complex interaction situations, but also verify the renovation plan in advance, saving costs and better describing the complex traffic system. Currently, the microscopic simulation models for non-motor vehicle traffic flow mainly include the cellular automaton model, the vector field model, the lattice gas model, and the social force model. Among them, as a continuous flow model, the social force model does not need to delimit lanes and grids during modeling, which enables the model to more realistically reflect the traffic state. In addition, the boundary force in the social force model can express the inhibitory effect of the painted boundary on non-motor vehicle drivers, which cannot be shown by other models.

[0004] However, the existing social force model is difficult to reproduce the trajectory characteristics of non-motor vehicles crossing the line; at the same time, the existing research has greatly simplified the interaction rules and insufficiently considered factors in decision-making behavior, which makes it difficult for the existing model to accurately reproduce non-motor vehicle interaction and decision-making behavior from the microscopic level and cannot provide an analysis tool for the design of the motor vehicle and non-motor vehicle lane sections. Summary of the Invention

[0005] In order to solve the above problems existing in the prior art, the present invention proposes a microscopic simulation method for non-motor vehicle traffic flow, which is constructed based on an improved social force model.

[0006] The present invention provides a microscopic simulation method for non-motor vehicle traffic flow, and the method includes the following steps:

[0007] Step 1: Cluster the driving trajectories of non-motor vehicles to divide non-motor vehicle drivers accordingly;

[0008] Step 2: Construct a dynamic perception space model to obtain the position and speed information of non-motor vehicles;

[0009] Step 3: Construct a behavior decision-making model, load the position and speed information of the non-motor vehicles obtained in the above Step 2 into the behavior decision-making model, so as to obtain a behavior decision-making result;

[0010] Step 4: According to the behavior decision-making result obtained in Step 3, calculate the force condition by using an improved social force model, and update the position and speed information of the non-motor vehicles in real time.

[0011] The micro-simulation method for non-motor vehicle traffic flow provided by the present invention has the following beneficial effects:

[0012] (1) The present invention uses the K-means clustering algorithm to cluster the driving trajectories of non-motor vehicles, thereby classifying the types of non-motor vehicle drivers, and applying the classification results to the non-motor vehicle micro-simulation model, which can provide a simulation tool for the refined design and management of non-motor vehicle lanes.

[0013] (2) The present invention first proposes a dynamic perception space model, combines the compression characteristics of non-motor vehicle flows with the calculation of the perception space size, dynamically determines the interaction state between moving individuals, and ensures that a reasonable distance can be maintained between non-motor vehicles.

[0014] (3) The present invention respectively establishes a lane-changing behavior decision-making model for non-motor vehicle drivers crossing the line into the motor vehicle lane and crossing the line back to the original non-motor vehicle lane, and at the same time considers the influencing factors of the lane-changing behaviors of different types of non-motor vehicle drivers, making up for the deficiency that the existing model only enables non-motor vehicle drivers to complete the lane-changing behavior decision-making through simple rules. The behavior decision-making of non-motor vehicles in the model is closer to the actual behavior decision-making situation, which can provide a prerequisite for accurately matching the kinematic formula subsequently.

[0015] (4) The present invention improves the boundary force in the existing social force model. By proposing a virtual boundary force, different types of non-motor vehicle drivers can complete the lane-changing behavior, more realistically reproduce the driving trajectories of non-motor vehicles, and improve the theoretical research of the existing social force model. Description of the Drawings

[0016] Figure 1 It is a schematic diagram of the K-means clustering situation.

[0017] Figure 2 It is a schematic diagram of the non-motor vehicle dynamic perception space.

[0018] Figure 3 It is a schematic diagram of the change situation of the virtual boundary when a non-motor vehicle crosses the line into the motor vehicle lane.

[0019] Figure 4 This is a schematic diagram of the changes in the virtual boundary when a non-motor vehicle crosses the line and returns to the original non-motor vehicle lane.

[0020] Figure 5 It is a schematic diagram of modular design for simulation testing of the model.

[0021] Figure 6 This is a comparison chart of the actual and simulated crossing trajectories of different types of non-motor vehicle drivers; Figure 6-1 It is the adventurer's actual crossing track. Figure 6-2 It is the measured crossing trajectory of the opportunist. Figure 6-3 It is the measured crossing trajectory of the cautious; Figure 6-4 It is the simulated crossing track of the adventurer. Figure 6-5 It is the simulated crossing trajectory of the opportunist. Figure 6-6 It is the simulated crossing trajectory of the cautious person.

[0022] Figure 7 It is the velocity-density basic diagram of the simulated data and the measured data. DETAILED DESCRIPTION

[0023] To make the technical solution, purpose and advantages of the present invention clearer, the present invention is further described in detail below through specific implementation examples. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0024] The present invention provides a non-motor vehicle traffic flow micro-simulation method, which is based on an improved social force model and comprises the following steps:

[0025] Step 1: Cluster the driving trajectories of non-motor vehicles to divide the non-motor vehicle drivers;

[0026] Step 2: Construct a dynamic perception space model to obtain the position and speed information of non-motor vehicles;

[0027] Step 3: constructing a behavior decision model, loading the position and speed information of the non-motor vehicle obtained in step 2 into the behavior decision model, thereby obtaining a behavior decision result;

[0028] Step 4: Based on the behavioral decision results obtained in step 3, the improved social force model is used to calculate the force situation and update the position and speed information of the non-motor vehicle in real time.

[0029] According to the present invention, before performing the step 1, the kinematic parameters of the non-motor vehicle crossing the line and / or the parameters of the motor vehicle interacting with the non-motor vehicle crossing the line are first collected, and this is used as the data basis for the heterogeneity of the non-motor vehicle driver behavior division. Among them, the kinematic parameters of the non-motor vehicle crossing the line include but are not limited to: the speed of the overtaking vehicle, the speed difference between the front and rear vehicles when overtaking, the time of occupying the motor vehicle lane when crossing the line, the density of non-motor vehicle flow, the lateral position of the non-motor vehicle, the lateral gap of overtaking, the overtaking distance, the lateral spacing, and the longitudinal spacing. The parameters of the motor vehicle interacting with the non-motor vehicle crossing the line include but are not limited to: whether the motor vehicle honks and the headway of the motor vehicle.

[0030] In step 1, the K-means clustering algorithm may be used to cluster the driving trajectories of non-motor vehicles. The operation of clustering the driving trajectories of non-motor vehicles using the K-means clustering algorithm to divide the non-motor vehicle drivers includes the following steps:

[0031] Step 1: Randomly select k objects from n sample data as the initial object centers;

[0032] The “sample data” refers to the non-motor vehicle driving trajectory data, and the “object” refers to the kinematic parameters of the non-motor vehicle during crossing the line;

[0033] Step 2: Calculate the distance from each sample data to each cluster center, and assign each object to the cluster with the closest distance;

[0034] Step 3: After all objects are completed, recalculate the k cluster centers;

[0035] Step 4: Compare with the k cluster centers calculated last time. If the cluster center of any cluster changes, go to step 2. If not, clustering is completed.

[0036] When the K-means clustering algorithm is used to cluster the driving trajectories of non-motor vehicles, the clustering effect is best when the speed of overtaking vehicles, the time of crossing the line and occupying the motor vehicle lane, and the lateral clearance of overtaking are selected. Figure 1 The clustering result scatter plot is shown in Figure 2. Figure 1 The clustering results are divided into three point groups, namely square point group, circular point group and five-pointed star point group. The three cluster point groups represent different non-motor vehicle drivers, among which the square point group represents adventurers, the circular point group represents opportunists, and the five-pointed star point group represents cautious people, thus completing the division of non-motor vehicle drivers.

[0037] According to the present invention, in the second step, the dynamic perception space is the psychological space that a non-motor vehicle driver maintains a certain distance from other moving individuals (such as non-motor vehicles and motor vehicles) during cycling. When other moving individuals invade this space, the non-motor vehicle driver will have an unsafe and uncomfortable psychology and will make behavior decisions such as avoidance or crossing the line.

[0038] As Figure 2 shown, since a non-motor vehicle driver pays more attention to the front situation than the side during cycling, the dynamic perception space is set as a semi-ellipse, which is enclosed by two identical short semi-axes and a long semi-axis.

[0039] The size of the dynamic perception space changes dynamically with the traffic flow speed and density. For example, as the non-motor vehicle speed increases, in order to ensure cycling safety and comfort, the non-motor vehicle driver will maintain a greater distance from other moving individuals, and the size of the dynamic perception space will increase; due to the compression characteristics of non-motor vehicle traffic flow, as the non-motor vehicle flow density increases, the distance between moving individuals will shrink accordingly, and the size of the dynamic perception space will decrease.

[0040] The calculation expression of the dynamic perception space model is:

[0041]

[0042] In formula (1), a and b are the major axis radius and minor axis radius of the non-motor vehicle elliptical contour; α i (t) and α′ i (t) are the longitudinal perception coefficient and transverse perception coefficient between non-motor vehicles at time t; is the speed of non-motor vehicle i at time t; θ is the angle between the non-motor vehicle and the road forward direction; β is the non-motor vehicle flow compression coefficient, indicating that when the non-motor vehicle flow speed decreases and the density increases, the non-motor vehicle flow spacing will be compressed, and the expression is:

[0043] β = k j / ku f (2)

[0044] In formula (2): k j is the jam density; k is the non-motor vehicle flow density; u f is the free flow speed.

[0045] By using the dynamic perception space model constructed in the above manner, the perception space size of the non-motor vehicle driver can be calculated, and the position and speed information of non-motor vehicles within a certain range can be obtained therefrom.

[0046] According to the present invention, in step three, by loading the position and speed information of the non-motor vehicle obtained in step two into the behavior decision-making model, a behavior decision-making result is obtained. The behavior decision-making result includes a free behavior, an avoidance behavior, and a crossing behavior; wherein, the crossing behavior includes a behavior of crossing into the motor vehicle lane and a behavior of crossing back to the original non-motor vehicle lane.

[0047] The behavior decision-making model includes:

[0048] When there are no vehicles or obstacles in the dynamic perception space of the non-motor vehicle driver, the non-motor vehicle driver will take a free behavior.

[0049] When there are other moving individuals in the dynamic perception space of the non-motor vehicle driver, potential conflicts may occur. Only when the two moving individuals completely overlap in time and space, will the potential conflict evolve into a real conflict. At this time, the non-motor vehicle will take an avoidance behavior, and a conflict judgment mechanism is used to discriminate the conflict point. The expression is:

[0050]

[0051]

[0052]

[0053] In formulas (3), (4), and (5): are the positions of moving individual i and moving individual j; is the potential collision position; are the speeds of moving individual i and moving individual j; TC i , TC j is the time taken by moving individual i and moving individual j from the current position to the potential collision position; TC ij is the probability of collision between moving individual i and moving individual j. The smaller this value is, the greater the probability of collision between the two. When this value approaches infinity, the two will not collide.

[0054] In a preferred specific embodiment, the behavior decision-making model further includes an operation for calculating the decision-making probability of the non-motor vehicle driver's crossing behavior. A binary Logistic regression model can be used to calculate the decision-making probability of the non-motor vehicle driver's crossing behavior. Since the influencing factors for the non-motor vehicle driver to cross into the motor vehicle lane and cross back to the original non-motor vehicle lane are different, and there are differences in the crossing behavior decision-making of non-motor vehicle drivers with different driving behaviors, it is necessary to establish binary Logistic regression models for the two crossing behaviors (including the crossing decision-making behavior of the non-motor vehicle driver crossing into the motor vehicle lane and the crossing decision-making behavior of the non-motor vehicle driver crossing back to the original non-motor vehicle lane) respectively.

[0055] Among them, the expression of the behavior decision model for non-motor vehicle drivers to cross the line and enter the motor vehicle lane is as follows:

[0056]

[0057] In formula (6): P ad , P op , P ca are the probabilities of risk-takers, opportunists, and cautious people among non-motor vehicle drivers crossing the line and entering the motor vehicle lane; x 1 , x 2 , …, x n are the nth factors affecting the behavior of non-motor vehicle drivers crossing the line and entering the motor vehicle lane; β 0 , β 1 …, β n , ε are regression coefficients.

[0058] The expression of the behavior decision model for non-motor vehicle drivers to cross the line and return to the original non-motor vehicle lane is as follows:

[0059]

[0060] In formula (7): P′ ad , P′ op , P′ ca are the probabilities of risk-takers, opportunists, and cautious people among non-motor vehicle drivers crossing the line and returning to the original non-motor vehicle lane; x′ 1 , x′ 2 , …, x′ n are the nth factors affecting the behavior of non-motor vehicle drivers crossing the line and returning to the original non-motor vehicle lane; β′ 0 , β′ 1 …, β′ n , ε′ are regression coefficients.

[0061] According to the present invention, the fourth step may include the following situations:

[0062] 1. When the behavior decision result is judged as a free behavior, the non-motor vehicle is under the action of a driving force, and the force expression is:

[0063]

[0064] In formula (8): m i is the mass of non-motor vehicle i; τ i is the duration of non-motor vehicle i; is the speed of non-motor vehicle i at time t; is the expected direction, obtained by pointing from the current position to the target point position; is the expected speed of different types of non-motor vehicle drivers. When λ=1, it represents the risk-taker; when λ=2, it represents the opportunist; and when λ=3, it represents the cautious driver.

[0065] 2. When the behavior decision result is determined to be the behavior of avoiding non-motor vehicles, the non-motor vehicle is subjected to the repulsive force of non-motor vehicles and the avoidance force of non-motor vehicles. The force expression is:

[0066]

[0067] In formula (9): is the resultant force on non-motor vehicle i when it takes the action of giving way to non-motor vehicle b at time t; is the repulsive force exerted on non-motor vehicle i by non-motor vehicle b at time t; is the avoidance force exerted on non-motor vehicle i at time t.

[0068] Non-motor vehicle repulsion The expression is:

[0069]

[0070] In formula (10): A ib B is the repulsive force strength of non-motor vehicle b on non-motor vehicle i; ib b is the influence range of the repulsive force of non-motor vehicle b on non-motor vehicle i; ib is the distance between the outer contours of non-motor vehicle b and non-motor vehicle i; is the vector pointing from non-motor vehicle b to non-motor vehicle i.

[0071] Non-motor vehicle avoidance force The expression is:

[0072]

[0073] In formula (11): is the speed of non-motor vehicle i at time t; Δt is the time increment; is the adjusted speed of non-motor vehicle i, and the expression is:

[0074]

[0075]

[0076] In formula (12) and (13): ω(ψ) is the individual counterclockwise rotation angle.

[0077] 3. When the behavior decision result is determined to be the behavior of avoiding motor vehicles, the non-motor vehicle is subjected to the repulsive force of the motor vehicle and the avoidance force of the non-motor vehicle, and the force expression is:

[0078]

[0079] In formula (14): is the resultant force on non-motor vehicle i when it takes an evasive action against motor vehicle c at time t; is the repulsive force from motor vehicle c on non-motor vehicle i at time t; is the evasive force from non-motor vehicles on non-motor vehicle i at time t.

[0080] Motor vehicle repulsive force The expression is:

[0081]

[0082] In formula (15): A ic is the intensity of the repulsive force from motor vehicle c on non-motor vehicle i; A ho is the intensity of the horn force of the motor vehicle; B ic is the influence range of the repulsive force from motor vehicle c on non-motor vehicle i; B ho is the influence range of the motor vehicle horn; b ic is the distance between the outer contours of motor vehicle c and non-motor vehicle i; is the vector from motor vehicle c to non-motor vehicle i.

[0083] Non-motor vehicle evasive force The expression is the same as formula (11) and will not be elaborated here.

[0084] When the behavior decision result is judged as a crossing line behavior, the non-motor vehicle will be affected by the marked boundary. However, when the non-motor vehicle is affected by the boundary force in the existing social force model, the non-motor vehicle cannot complete the crossing line behavior. To solve this problem, the present invention sets the marked boundary between the motor vehicle lane and the non-motor vehicle lane as a virtual boundary that can be crossed. The virtual boundary can cooperate with the crossing vehicle to complete lateral offset, and the virtual boundary is discretized into a point set, which is used as the attraction point when the non-motor vehicle crosses the line. The positions of the attraction points are different for different non-motor vehicle driver types. The specific description is as follows:

[0085] IV. When the non-motor vehicle driver judges through the behavior decision result that it crosses the line into the motor vehicle lane, as Figure 3 shown, the virtual boundary will shift towards the motor vehicle lane side. The offset amount ω of the virtual boundary is equal to the difference between the lateral clearance d k for non-motor vehicle overtaking and the lateral distance d v from the marked boundary before crossing the line. The force direction of the non-motor vehicle under the virtual boundary force points to the virtual boundary attraction point. The position of the virtual boundary attraction point is parallel to the non-motor vehicle overtaking point. The force expression is:

[0086]

[0087] In formula (16): The virtual boundary force strength of different non-motor vehicle driver types is: λ = 1 represents adventurers, λ = 2 represents opportunists, and λ = 3 represents cautious drivers; is the virtual boundary influence range; d sp is the distance between non-motor vehicle i and the virtual boundary; It is the direction in which the non-motor vehicle is attracted by the virtual boundary point.

[0088] 5. When the non-motor vehicle driver is judged to have crossed the line and returned to the original non-motor vehicle lane through the behavioral decision result, if Figure 4 As shown in the figure, the virtual boundary will be offset to the non-motorized vehicle lane side, and the virtual boundary offset ω is equal to the lateral distance d between the non-motorized vehicle and the marked boundary before crossing the line v , the direction of the virtual boundary force on the non-motor vehicle points to the virtual boundary attraction point. The position of the virtual boundary attraction point is obtained by summing the vector perpendicular to the virtual boundary direction and the vector pointing to the desired target point. The force expression is the same as formula (16) and will not be repeated here.

[0089] In a specific implementation of the present invention, the step 2 further includes calibrating relevant parameters of the dynamic perception space, and the relevant parameters of the dynamic perception space include a longitudinal perception coefficient and a lateral perception coefficient.

[0090] Since the non-motor vehicle driver is not disturbed by other non-motor vehicles in a free driving state, that is, other non-motor vehicles are outside the dynamic perception space of the non-motor vehicle, the length and width of the dynamic perception space are closely related to the longitudinal spacing and lateral spacing maintained by the non-motor vehicle in a free driving state. The longitudinal perception coefficient and lateral perception coefficient under different vehicle models and driver types are calibrated. The expressions of the longitudinal perception coefficient and lateral perception coefficient are as follows:

[0091]

[0092] In formula (17): α i (t) is the longitudinal perception coefficient of non-motor vehicle i at time t; α′ i (t) is the lateral perception coefficient of non-motor vehicle i at time t; a and b are the major axis radius and minor axis radius of the non-motor vehicle elliptical contour, respectively; is the speed of non-motor vehicle i at time t; θ is the angle between the non-motor vehicle and the direction of the road; l i The longitudinal distance maintained by non-motor vehicles in a free riding state; b i It is the lateral distance maintained by non-motor vehicles when they are riding freely.

[0093] Table 1 shows a summary of the longitudinal perception coefficients and lateral perception coefficients of different specific vehicle models and non-motor vehicle driver types.

[0094] Table 1 Summary of longitudinal perception coefficients and lateral perception coefficients for different vehicle models and non-motor vehicle driver types

[0095]

[0096] In a specific embodiment of the present invention, in the third step, an operation of calibrating relevant parameters of the binary Logistic regression model in the decision-making of non-motor vehicle drivers' crossing behavior is further included. The relevant parameters of the binary Logistic regression model in the decision-making of non-motor vehicle drivers' crossing behavior include the influencing factors affecting non-motor vehicle drivers' crossing into the motor vehicle lane and crossing back to the original non-motor vehicle lane, as well as the corresponding regression coefficients, which are specifically described as follows:

[0097] 1. Calibration of relevant parameters of the binary Logistic regression model for non-motor vehicle drivers' crossing into the motor vehicle lane

[0098] Through actual investigation and crossing behavior analysis, it is obtained that the influencing factors affecting the decision-making of non-motor vehicle drivers' crossing into the motor vehicle lane include: the speed of the overtaking vehicle, the speed difference between the front and rear vehicles during overtaking, the density of the non-motor vehicle flow, the lateral position of the non-motor vehicle, and the headway of the motor vehicle. Table 2 shows the value variable factor situation corresponding to a specific model affecting the decision-making of non-motor vehicle drivers' crossing into the motor vehicle lane.

[0099] Table 2 Factors affecting the decision-making of non-motor vehicle drivers' crossing into the motor vehicle lane

[0100]

[0101] The stepwise backward optimization method is used to select the significance index, and a 95% confidence level is set. When the significance level sig. value is not greater than 0.05, it is considered that the parameter is significant for the dependent variable. According to the survey results and the classification of non-motor vehicle driver behavior heterogeneity, a binary Logistic regression model is established for different non-motor vehicle driver types, and the significance of the influencing factor indicators for each type of non-motor vehicle driver is shown in Table 3.

[0102] Table 3 Significance test

[0103]

[0104] Eliminating the parameters with sig. values ​​greater than 0.05 and selecting the independent variables that have a significant impact on the dependent variable, it can be concluded that for risk-takers, the speed of overtaking vehicles, the lateral position of non-motor vehicles, and the time headway parameters of motor vehicles have a significant impact on the decision-making of non-motor vehicle drivers crossing the line into the motor vehicle lane; for opportunists, the density of non-motor vehicle flow, the lateral position of non-motor vehicles, and the time headway of motor vehicles have a significant impact on the decision-making of non-motor vehicle drivers crossing the line into the motor vehicle lane; for cautious people, the speed of overtaking vehicles, the lateral position of non-motor vehicles, and the time headway of motor vehicles have a significant impact on the decision-making of non-motor vehicle drivers crossing the line into the motor vehicle lane. Table 4 shows a specific parameter estimation of each type of non-motor vehicle driver.

[0105] Table 4 Parameter estimates

[0106]

[0107] From this, we can get the probability P of the risk-taker, the opportunist and the cautious person crossing the line and entering the motor vehicle lane. ad , P op and P ca The calculation formula is as follows:

[0108]

[0109]

[0110]

[0111] 2. Calibration of parameters related to the binary classification logistic regression model for non-motor vehicle drivers crossing the line and returning to the original non-motor vehicle lane

[0112] Through actual investigation and analysis of crossing-line behavior, it is concluded that the factors affecting the decision-making of non-motor vehicle drivers to return to the original non-motor vehicle lane include: the speed of the overtaking vehicle, the density of non-motor vehicle flow, the lateral gap of overtaking, whether the motor vehicle honks, and the time of crossing the line and occupying the motor vehicle lane. Table 5 shows a specific factor affecting the decision-making of non-motor vehicle drivers to return to the original non-motor vehicle lane by crossing the line.

[0113] Table 5 Factors affecting the decision-making of non-motor vehicles returning to the original non-motor vehicle lane

[0114]

[0115] The stepwise backward optimization method was used to select the significant index, and the 95% confidence level was set. When the significant level sig. value was not greater than 0.05, the parameter was considered to be significant for the dependent variable. According to the survey results and the classification of the heterogeneity of non-motor vehicle drivers' behavior, a binary logistic regression model was established for different non-motor vehicle drivers. The significance of the influencing factor indicators of each type of non-motor vehicle drivers is shown in Table 6.

[0116] Table 6 Significance test

[0117]

[0118] Eliminating the parameters with sig. values ​​greater than 0.05 and selecting the independent variables that have a significant impact on the dependent variable, it can be concluded that for risk-takers, the speed of the overtaking vehicle, whether the motor vehicle honks, and the time of crossing the line to occupy the motor vehicle lane have a significant impact on the non-motor vehicle driver's decision to return to the original non-motor vehicle lane; for opportunists, the non-motor vehicle flow density, whether the motor vehicle honks, and the time of crossing the line to occupy the motor vehicle lane have a significant impact on the non-motor vehicle driver's decision to return to the original non-motor vehicle lane; for cautious people, the speed of the overtaking vehicle, the lateral gap of the overtaking vehicle, and whether the motor vehicle honks have a significant impact on the non-motor vehicle driver's decision to return to the original non-motor vehicle lane. The estimated parameters of each type of non-motor vehicle driver are shown in Table 7

[0119] Table 7 Parameter estimation table

[0120]

[0121] Thus, we can get the probability P of the adventurer, the opportunist and the cautious person crossing the line and returning to the original non-motor vehicle lane. a′ , P op ′ and P ca The calculation formula is as follows:

[0122]

[0123]

[0124]

[0125] In a specific embodiment of the present invention, in the step 4, it also includes an operation of calibrating the parameters related to the improved social force model of the non-motor vehicle, and the parameters related to the improved social force model of the non-motor vehicle include driving force, repulsive force, and virtual boundary force parameters, which are specifically described as follows:

[0126] Driving force parameters include desired speed and duration:

[0127] 1. Calibration of the expected speed includes:

[0128] The expected speed refers to the speed that the non-motor vehicle driver expects to reach without any interference. The speed when there is no interference from other vehicles in the dynamic perception space of the non-motor vehicle driver is set as the expected speed. According to the classification of non-motor vehicle models and non-motor vehicle driver categories, the measured data are counted and divided, as shown in Table 8.

[0129] Table 8 Summary of expected speeds for different vehicle models and non-motor vehicle driver types

[0130]

[0131] (2) The calibration of duration specifically includes:

[0132] Duration refers to the time it takes for a non-motor vehicle driver to change from the current speed to the desired speed, mainly including the reaction time of the non-motor vehicle driver when being stimulated by the outside world and the control time of taking evasive actions. According to the heterogeneity of non-motor vehicle driver behavior, the reaction time of the risk-taker is 1.98s, the reaction time of the opportunist is 2.17s, and the reaction time of the cautious is 2.78s. The calculation of the control time is based on the formula: (desired speed-average speed) / acceleration, and the duration of each type of non-motor vehicle driver is calculated, as shown in Table 9.

[0133] Table 9 Summary of duration of different types of vehicles and non-motor vehicle drivers

[0134]

[0135] The repulsive force parameters include the repulsive force intensity A of non-motor vehicles ib , Non-motor vehicle repulsive force influence range B ib , the repulsive force intensity of motor vehicle A ic , the influence range of motor vehicle repulsion force B ic , the force intensity of motor vehicle horn A ho , the impact range of motor vehicle horn B ho .

[0136] The maximum likelihood estimation method is used to calibrate the above parameters. The force conditions of the non-motor vehicle are calculated by selecting the parameter values. The trajectory of the non-motor vehicle is obtained through simulation and the error analysis is performed with the measured data to obtain the optimal parameters.

[0137] Assume that the position of non-motor vehicle i at time t+1 is P (i,t+1) Through the parameter θ prediction, P (i,t) Move to P (i,t+1) Moving distance Δd′ θ , and Δd′ θ It follows a normal distribution with a mean of μ and a standard deviation of σ. The measured moving distance Δd θ It also obeys the normal distribution with mean μ and standard deviation σ, and the likelihood function of parameter θ can be obtained:

[0138]

[0139] For the convenience of calculation, take the logarithm of both sides of formula (24) to find the value of the model parameter θ that maximizes this logarithmic function. The model after taking the logarithm is as follows:

[0140]

[0141] By debugging and calibrating the values of each parameter, the estimated results of the repulsive force parameters for different types of non-motor vehicle drivers are obtained, as shown in Table 10.

[0142] Table 10 Summary of Estimated Results of Repulsive Force Parameters for Different Types of Non-Motor Vehicle Drivers

[0143]

[0144] The parameters related to the virtual boundary force include the intensity of the virtual boundary force and the influence range of the virtual boundary

[0145] The establishment of the virtual boundary force draws on the modeling idea of attraction in the social force model. The effect of the virtual boundary force is actually the attraction of discrete points to non-motor vehicles crossing the line. Therefore, the parameter calibration of the virtual boundary force draws on the calibration method of the attraction model. The formula for the intensity of the virtual boundary force in the model is mv ov / τ, and the values of for different types of non-motor vehicle drivers are calculated respectively; the influence range of the virtual boundary represents the lateral distance between the non-motor vehicle and the virtual boundary when the non-motor vehicle begins to be affected by the virtual boundary force. According to the survey results and the above analysis, the influence range is closely related to the lateral clearance of non-motor vehicle overtaking. The calibration results are shown in Table 11.

[0146] Table 11 Summary of Estimated Results of Virtual Boundaries for Different Vehicle Types and Non-Motor Vehicle Driver Types

[0147]

[0148] As Figure 5 shown, the present invention uses the Python programming tool to conduct simulation tests on the model, and modularizes the simulation test process, which is divided into: creating an individual list module, a perception module, a decision-making module, a mechanical calculation module, an updating individual position module, an output data module, and a visualization module. The main contents of each module include:

[0149] (1) Create individual list module: This module is used to initialize individual information by creating a list. This list includes model parameters related to non-motor vehicles and motor vehicles. The motor vehicle list includes information such as individual number, coordinates, mass, reaction time, speed, acceleration, size, expected target point, and related mechanical parameters. The non-motor vehicle list divides non-motor vehicle drivers into three categories, namely, adventurers, opportunists, and cautious people. Individual behavior information is established for different types of non-motor vehicle drivers, including individual number, mass, size, expected target point, speed, acceleration, position, and related mechanical parameter information.

[0150] (2) Perception module: This module is the part where non-motor vehicle drivers perceive the information of individuals around them. First, the shape and size of the individual dynamic perception space are calculated through the model. Secondly, the dynamic perception space is connected with the moving individuals. Then, the individual attributes falling into the dynamic perception space are monitored, including motor vehicles, non-motor vehicles, and virtual boundaries. Finally, the motion information of other individuals in the perception space is obtained, including speed, acceleration, turning angle, and spacing. The perceived information is transmitted to the decision module as the basic data of the decision module.

[0151] (3) Decision-making module: This module is responsible for calculating the decision-making probability of individual motion behavior. By sensing the motion information of surrounding individuals, it calculates the probability of individual avoidance, the probability of crossing the line into the motor vehicle lane, and the probability of crossing the line back to the original non-motor vehicle lane. By setting (0, 1) to judge, it serves as a prerequisite for whether to change the individual mechanical formula next.

[0152] (4) Mechanical calculation module: This module calculates the combined force that motor vehicles and non-motor vehicles will be subject to at the next moment after making a decision, including driving force, repulsive force and virtual boundary force. After the three forces are summed in the vector direction, the acceleration of the moving individual at the next moment is calculated and the acceleration data at the next moment is passed to the module for updating the individual position.

[0153] (5) Individual position update module: This module is responsible for calculating the position of the moving individual at the next moment, and obtains the coordinates of the moving individual at the next moment through calculation.

[0154] (6) Output data module: This module is the data output part. By calling the text file, the speed, acceleration and trajectory information of the moving individual at each moment are stored and these data are plotted in the form of a chart.

[0155] (7) Visualization module: This module is mainly used to display the individual motion information obtained by the above calculation in the form of animation, including the size of the simulation area, the appearance of the moving individual, the simulation time, etc.

[0156] The specific simulation test process includes the following steps:

[0157] Step 1: Input the kinematic parameters of the moving individual and the simulation scenario parameters to initialize the simulation scenario.

[0158] Step 2: The moving individual scans the individual attributes (motor vehicles, non-motor vehicles, virtual boundaries) in the dynamic perception space and judges the distance and speed difference from other individuals.

[0159] Step 3: Judge whether the non-motor vehicle makes a decision to cross the line and enter the motor vehicle lane. If so, generate a virtual boundary, calculate the virtual boundary force, and jump to Step 5; otherwise, jump to Step 4.

[0160] Step 4: Judge whether the non-motor vehicle makes a decision to cross the line and return to the original non-motor vehicle lane. If so, generate a virtual boundary and calculate the virtual boundary force; otherwise, do not calculate the virtual boundary force.

[0161] Step 5: Judge whether the non-motor vehicle makes a decision to take an avoidance behavior. If so, generate a deflection angle and calculate the non-motor vehicle avoidance force; otherwise, do not calculate the non-motor vehicle avoidance force.

[0162] Step 6: Judge whether there is a motor vehicle in the dynamic perception space. If so, calculate the motor vehicle repulsion force; otherwise, do not calculate the motor vehicle repulsion force.

[0163] Step 7: Judge whether there is a non-motor vehicle in the dynamic perception space. If so, calculate the non-motor vehicle repulsion force; otherwise, do not calculate the non-motor vehicle repulsion force.

[0164] Step 8: Calculate the driving force according to the direction of the expected target point.

[0165] Step 9: Update the position of the moving individual according to the resultant force situation.

[0166] Step 10: Judge whether all moving individuals have reached the expected position points. If all have reached the expected position points, the simulation ends; otherwise, return to Step 2.

[0167] To ensure the authenticity and reliability of the simulation test results, multiple repeated experiments need to be carried out in the same simulation environment.

[0168] To verify that the model can reproduce the behavior of non-motor vehicle drivers crossing the line of different types, it is necessary to build an environment consistent with the reality in the simulation to verify the authenticity of the simulation. In terms of road attributes, a motor-vehicle and non-motor-vehicle lane section with a length of 50 m is set, the width of the motor vehicle lane is 3.5 m, and the width of the non-motor vehicle lane is 2.5 m. In terms of traffic flow input, motor vehicles arrive regularly in a uniform pattern, with a flow rate of 956 pcu / h, and non-motor vehicles arrive according to a negative binomial distribution, with a flow rate of 2,033 vehicles / h. Among them, electric bicycles account for 57%, adventurers account for 25.6%, opportunists account for 56.8%, and cautious people account for 17.6%. The simulation time step is 0.1 s, and the output data includes indicators such as non-motor vehicle trajectories, the number of line crossings, speed, density, etc. Extract the non-motor vehicle line-crossing trajectories in the simulation results and compare them with the measured trajectories. As Figure 6 shown, the method of trajectory coverage rate is used to evaluate the simulation trajectory results. The method is to divide the trajectory coordinate system into squares of 0.5 m × 0.5 m. If the simulation trajectory coincides with the true estimate, it is counted as 1, otherwise it is counted as 0. The results show that the non-motor vehicle trajectory coverage rate obtained from the simulation results is 85%, indicating that the model can reproduce the non-motor vehicle line-crossing behavior well.

[0169] To verify that the microscopic simulation model proposed by the present invention can reflect the macroscopic traffic flow state, the speed-density fundamental diagram is used for verification. The measured data is input into the simulation model. To ensure the authenticity and reliability of the data, three repeated experiments will be carried out. Extract the non-motor vehicle speed and density data in the middle section of the simulation and compare them with the measured data. As Figure 7 shown, from the results of the simulation data, as the non-motor vehicle density increases, the non-motor vehicle speed shows a downward trend. When the non-motor vehicle density is less than 0.1 vehicle / m 2 , the change in non-motor vehicle speed is not significant, indicating that at this time, non-motor vehicles are not greatly affected by other vehicles and show a free flow state. When the non-motor vehicle density is in the range of 0.1 - 0.3 vehicles / m 2 , the change in non-motor vehicle speed is the largest. When the non-motor vehicle density is greater than 0.3 vehicles / m 2 , the change in non-motor vehicle speed gradually slows down. When the non-motor vehicle density is greater than 0.45 vehicles / m 2 , the non-motor vehicle speed gradually becomes 0. From the overall trend, the simulation data and the measured data maintain a strong linear relationship, proving that the model can reproduce a non-motor vehicle flow consistent with the actual scenario.

[0170] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A microscopic simulation method for non-motor vehicle traffic flow, characterized in that, the method comprises the following steps: Step 1: Cluster the driving trajectories of non-motor vehicles to divide non-motor vehicle drivers accordingly; Step 2: Construct a dynamic perception space model to obtain the position and speed information of non-motor vehicles; Step 3: Construct a behavior decision-making model, and load the position and speed information of non-motor vehicles obtained in the above Step 2 into the behavior decision-making model to obtain a behavior decision result; Step 4: According to the behavior decision result obtained in Step 3, calculate the force condition using an improved social force model, and update the position and speed information of non-motor vehicles in real time; wherein, the calculation expression of the dynamic perception space model is: In formula (1), a and b are the major axis radius and minor axis radius of the elliptical contour of the non-motor vehicle; α i (t) and α i ′(t) are the longitudinal perception coefficient and lateral perception coefficient between non-motor vehicles at time t; is the speed of non-motor vehicle i at time t; θ is the angle between the non-motor vehicle and the road forward direction; β is the non-motor vehicle flow compression coefficient, indicating that when the non-motor vehicle flow speed decreases and the density increases, the non-motor vehicle flow spacing will be compressed, and the expression is: β = k j / ku f (2) In formula (2): k j is the blocking density; k is the density of non-motor vehicle flow; u f is the free flow speed; wherein, the parameters related to the improved social force model include a driving force parameter, a repulsive force parameter, and a virtual boundary force parameter; wherein, the driving force parameter includes an expected speed and a duration; The repulsive force parameters include the intensity A of the repulsive force of non-motor vehicles ib , the influence range B of the repulsive force of non-motor vehicles ib , the intensity A of the repulsive force of motor vehicles ic , the influence range B of the repulsive force of motor vehicles ic , the intensity A of the horn force of motor vehicles ho , the influence range B of the horn of motor vehicles ho ; Calibrate the repulsive force parameter using the maximum likelihood estimation method. After selecting parameter values, calculate the force condition of non-motor vehicles, obtain the non-motor vehicle trajectory through simulation and conduct error analysis with the measured data to obtain the optimal parameters; Assume that the position of non-motor vehicle i at time t+1 is P (i,t+1) Predicted by parameter θ, P (i,t) Move to P (i,t+1) Moving distance Δd′ θ , and Δd′ θ obeys a normal distribution with mean μ and standard deviation σ. The measured moving distance Δd θ also obeys a normal distribution with mean μ and standard deviation σ. The likelihood function of parameter θ can be obtained as follows: For the convenience of calculation, take the logarithm of both sides of formula (24), and find the value of the model parameter θ that maximizes this logarithmic function. The model after taking the logarithm is as follows: Through debugging and calibration of the values of each parameter, obtain the estimation results of the repulsive force parameters for non-motor vehicle drivers of each type; The virtual boundary force parameters include the acting intensity of the virtual boundary force and the influence range of the virtual boundary The acting intensity of the virtual boundary force is calculated by mv ov / τ, and the values for different types of non-motor vehicle drivers are calculated respectively; the virtual boundary influence range represents the lateral spacing between the non-motor vehicle and the virtual boundary when the non-motor vehicle begins to be affected by the virtual boundary force.

2. The method according to claim 1, characterized in that: Before performing the above Step 1, collect the kinematic parameters during the process of non-motor vehicles crossing the line, and / or the parameters during the interaction between motor vehicles and non-motor vehicles crossing the line.

3. The method according to claim 2, characterized in that: The kinematic parameters during the process of non-motor vehicles crossing the line include: the speed of the overtaking vehicle, the time occupying the motor vehicle lane when crossing the line, the speed difference between the front and rear vehicles during overtaking, the headway of the motor vehicle, the density of the non-motor vehicle flow, the lateral position of the non-motor vehicle, the lateral clearance during overtaking, the overtaking distance, the lateral spacing, and the longitudinal spacing; The parameters during the interaction between motor vehicles and non-motor vehicles crossing the line include: whether the motor vehicle honks the horn, and the headway of the motor vehicle.

4. The method according to any one of claims 1-3, characterized in that: In the above Step 1, use the K-means clustering algorithm to cluster the driving trajectories of non-motor vehicles to divide non-motor vehicle drivers accordingly.

5. The method according to claim 4, characterized in that: The operation of using the K-means clustering algorithm to cluster the driving trajectories of non-motor vehicles to divide non-motor vehicle drivers accordingly comprises the following steps: Step 1: Randomly select k objects from n sample data as the initial object centers; Step 2: Calculate the distance from each sample data to each clustering center respectively, and assign each object to the cluster with the closest distance; Step 3: After all objects are completed, recalculate the k cluster centers; Step 4: Compare with the k cluster centers obtained in the previous calculation. If the clustering center of any cluster changes, go to Step 2. If no change occurs, the clustering is completed; According to the clustering results, non-motor vehicle drivers are divided into three categories: adventurers, opportunists, and cautious people.

6. The method according to claim 5, Features: The behavioral decision model includes: When there are no vehicles or obstacles in the dynamic perception space of the non-motor vehicle driver, the non-motor vehicle driver will take free actions; When there are other moving individuals in the dynamic perception space of the non-motor vehicle driver, potential conflicts may occur. Only when the two moving individuals completely intersect in time and space, the potential conflicts will evolve into real conflicts. At this time, the non-motor vehicle will take evasive actions and use the conflict judgment mechanism to judge the conflict points. The expression is: In formulas (3), (4), and (5): are the positions of moving individuals i and j; is the potential collision position; are the velocities of moving individuals i and j; TC i , TC j is the time taken for moving individuals i and j to reach the potential collision position from the current position; TC ij is the probability of collision between moving individuals i and j.

7. The method according to claim 6, Features: The behavior decision model also includes an operation of calculating the probability of the non-motor vehicle driver's crossing the line behavior decision.

8. The method according to claim 7, Features: A binary logistic regression model was used to calculate the decision probability of non-motor vehicle drivers crossing the line. Among them, the behavioral decision model expression of non-motor vehicle drivers crossing the line into the motor vehicle lane is: In formula (6): P ad , P op , P ca are the probabilities of risk-takers, opportunists, and cautious non-motor vehicle drivers crossing the line and entering the motor vehicle lane; x 1 , x 2 , …, x n are the nth factors affecting the behavior of non-motor vehicle drivers crossing the line and entering the motor vehicle lane; β 0 , β 1 …, β n , ε are regression coefficients; The behavioral decision model expression of non-motor vehicle drivers crossing the line and returning to the original non-motor vehicle lane is: In formula (7): P a ' d , P o ' p , P c ' a is the probability that the adventurous, opportunistic, and cautious non-motor vehicle drivers cross the line and return to the original non-motor vehicle lane; x 1 ′ 、x 2 ′ , …, x n ′ is the nth factor that affects the behavior of non-motor vehicle drivers crossing the line and returning to the original non-motor vehicle lane; β 0 ′ , β 1 ′ …、β n ′ , ε′ is the regression coefficient.

9. The method according to claim 8, Features: The step 4 includes the following situations:

1. When the behavior decision result is judged as free behavior, the non-motor vehicle is acted upon by a driving force, and the force expression is: In formula (8): m i is the mass of non-motor vehicle i; τ i is the duration of non-motor vehicle i; is the speed of non-motor vehicle i at time t; The desired direction is obtained by pointing from the current position to the target point; is the expected speed of different types of non-motor vehicle drivers, when λ=1, it represents the risk-taker, when λ=2, it represents the opportunist, and when λ=3, it represents the cautious driver; 2. When the behavior decision result is determined to be the behavior of avoiding non-motor vehicles, the non-motor vehicle is subjected to the repulsive force of non-motor vehicles and the avoidance force of non-motor vehicles. The force expression is: In formula (9): is the resultant force on non-motor vehicle i when it takes an avoidance action against non-motor vehicle b at time t; is the repulsive force exerted on non-motor vehicle i by non-motor vehicle b at time t; is the avoidance force on non-motor vehicle i at time t; Non-motor vehicle repulsive force The expression is: In formula (10): A ib is the repulsive force intensity of non-motor vehicle b on non-motor vehicle i; B ib is the influence range of the repulsive force of non-motor vehicle b on non-motor vehicle i; b ib is the distance between the outer contours of non-motor vehicle b and non-motor vehicle i; is the vector from non-motor vehicle b to non-motor vehicle i; Non-motor vehicle avoidance force The expression is: In formula (11): is the speed of non-motor vehicle i at time t; Δt is the time increment; is the adjusted speed of non-motor vehicle i, and the expression is: In formula (12) and (13): ω(ψ) is the individual counterclockwise rotation angle; 3. When the behavior decision result is determined to be the behavior of avoiding motor vehicles, the non-motor vehicle is subjected to the repulsive force of the motor vehicle and the avoidance force of the non-motor vehicle, and the force expression is: In formula (14): is the resultant force on non-motor vehicle i when it takes an evasive action against motor vehicle c at time t; is the repulsive force on non-motor vehicle i from the motor vehicle at time t; is the evasive force on non-motor vehicle i from other non-motor vehicles at time t; Motor vehicle repulsive force The expression is: In formula (15): A ic is the intensity of the repulsive force exerted by motor vehicle c on non-motor vehicle i; A ho is the acting intensity of the horn sound force on the motor vehicle; B ic is the influence range of the repulsive force of motor vehicle c on non-motor vehicle i; B ho is the influence range of the motor vehicle horn; b ic is the distance between the outer contours of motor vehicle c and non-motor vehicle i; is the vector of motor vehicle c pointing to non-motor vehicle i; Non-motor vehicle avoidance force The expression is as shown in formula (11); IV. When it is judged by the behavior decision result that the non-motor vehicle driver crosses the line and enters the motor vehicle lane, the virtual boundary will shift towards the motor vehicle lane side, and the virtual boundary offset ω is equal to the non-motor vehicle overtaking lateral clearance d k and the lateral distance d v from the marked boundary before crossing the line. The force direction of the non-motor vehicle under the virtual boundary force points to the virtual boundary attraction point, and the position of the virtual boundary attraction point is parallel to the non-motor vehicle overtaking point. The force expression is as follows: In formula (16): is the virtual boundary force strength of different non-motor vehicle driver types, λ=1 represents risk-takers, λ=2 represents chance-takers; λ = 3 represents a cautious person; is the influence range of the virtual boundary; d sp is the distance between non-motor vehicle i and the virtual boundary; is the direction of the non-motor vehicle being attracted to the virtual boundary point; 5. When the non-motor vehicle driver judges through the behavior decision result that they need to return to the original non-motor vehicle lane by crossing the line, the virtual boundary will shift towards the non-motor vehicle lane side, and the virtual boundary offset ω is equal to the lateral distance d between the non-motor vehicle and the marked boundary before crossing the line. v , the force direction of the non-motor vehicle under the virtual boundary force points to the virtual boundary attraction point. The position of the virtual boundary attraction point is obtained by summing the vector perpendicular to the virtual boundary direction and the vector pointing to the desired target point. The force expression is shown in Equation (16).

10. The method according to claim 9, Features: In the step 2, it also includes an operation of calibrating the relevant parameters of the dynamic perception space, wherein the relevant parameters of the dynamic perception space include a longitudinal perception coefficient and a lateral perception coefficient; The expressions of the longitudinal perception coefficient and the lateral perception coefficient are: In formula (17): α i (t) is the longitudinal perception coefficient of non-motor vehicle i at time t; α i ′(t) is the lateral perception coefficient of non-motor vehicle i at time t; a and b are the major axis radius and minor axis radius of the elliptical contour of the non-motor vehicle respectively; is the speed of non-motor vehicle i at time t; θ is the angle between the non-motor vehicle and the forward direction of the road; l i is the longitudinal spacing maintained by the non-motor vehicle in the free-riding state; b i is the lateral spacing maintained by the non-motor vehicle in the free-riding state.

11. The method according to claim 10, Features: In the step three, it also includes the operation of calibrating the relevant parameters of the binary classification logistic regression model in the non-motor vehicle driver's crossing line behavior decision.

12. The method according to claim 11, Features: The parameters of the binary logistic regression model in the non-motor vehicle driver's crossing lane behavior decision-making include the factors that affect the non-motor vehicle driver's crossing the lane into the motor vehicle lane and crossing the lane back to the original non-motor vehicle lane; among them: The parameters of the binary logistic regression model for the non-motor vehicle driver crossing the line into the motor vehicle lane include: the speed of the overtaking vehicle, the speed difference between the front and rear vehicles when overtaking, the non-motor vehicle flow density, the lateral position of the non-motor vehicle, and the headway of the motor vehicle; The relevant parameters of the binary classification Logistic regression model for non-motor vehicle drivers to cross the line and return to the original non-motor vehicle lane include: the speed of the overtaking vehicle, the density of the non-motor vehicle flow, the lateral clearance during overtaking, whether the motor vehicle honks, and the time of occupying the motor vehicle lane by crossing the line.

Images