Two-wheeled electric vehicle street crossing behavior prediction method based on multilayer decision model

Abstract

The invention discloses a two-wheeled electric vehicle street-crossing behavior prediction method based on a multilayer decision model, and the method comprises the steps: constructing a signal intersection two-wheeled electric vehicle parking behavior data set based on video survey data; constructing a two-wheeled electric vehicle multi-layer decision model; and inputting the constructed data set into a multilayer decision model of the two-wheeled electric vehicle, estimating model parameters by using a maximum likelihood estimation method, and predicting the street crossing behavior of the two-wheeled electric vehicle. According to the method, microscopic operation characteristics of the two-wheeled electric vehicle are considered, and individual characteristics are associated with complex decisions such as parking position decisions and red light behavior decisions at the signalized intersection; meanwhile, multiple factors including heterogeneity are brought into independent variables of behavior decision making in modeling, and a new view angle is provided for describing the behavior mechanism of the two-wheeled electric vehicle driver.

Description

Technical Field

[0001] This invention relates to the technical field of two-wheeled electric vehicle parking behavior research and two-wheeled electric vehicle management, and in particular to a method for predicting the street crossing behavior of two-wheeled electric vehicles based on a multi-level decision model. Background Technology

[0002] In recent years, the demand for two-wheeled electric vehicles among urban residents has surged, surpassing cars to become the most widely used mode of transportation. With the significant increase in the number of two-wheeled electric vehicles on urban roads, traffic safety issues have become increasingly prominent. Compared to motor vehicles, the traffic behavior of two-wheeled electric vehicles is more complex and variable, especially exhibiting strong unstructured characteristics at signalized intersections, such as random stopping, irregular starting, and running red lights, posing a serious challenge to urban traffic management and safety.

[0003] Existing research primarily focuses on signal control and path optimization for motor vehicles, as well as pedestrian crossing behavior. However, it lacks systematic research on the detailed behavioral characteristics of two-wheeled electric vehicles at signalized intersections, particularly in-depth modeling and analysis of their multi-level decision-making processes. Actual video recordings show that two-wheeled electric vehicles at signalized intersections make multi-level decisions based on factors such as travel purpose, available space ahead, signal status, and intersection clearance. These decisions include choosing different parking locations, whether to stop, and starting strategies during red lights, exhibiting significant heterogeneity and uncertainty. Therefore, a modeling method that comprehensively considers the behavioral characteristics of two-wheeled electric vehicles and the dynamic characteristics of intersection space is urgently needed to quantify their multi-level behavioral decision-making mechanisms, providing data support and theoretical basis for urban traffic organization optimization and intelligent signal control. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings and disadvantages of the existing technology and propose a method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model. This method considers the micro-operational characteristics of two-wheeled electric vehicles and links individual characteristics with complex decisions such as parking position decisions at signalized intersections and red light behavior decisions. At the same time, this method attempts to incorporate multiple factors, including heterogeneity, into the independent variables of behavioral decisions in the modeling process, providing a new perspective for describing the behavioral mechanism of two-wheeled electric vehicle drivers.

[0005] To achieve the above objectives, the technical solution provided by this invention is: a method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model, comprising the following steps:

[0006] 1) A dataset of parking behavior of two-wheeled electric vehicles at signalized intersections was constructed through video surveillance. The dataset includes parking location data and red light compliance data.

[0007] 2) Construct a multi-layer decision-making model for two-wheeled electric vehicles. The model consists of three layers. The first layer is the parking position decision model for two-wheeled electric vehicles, which is used to predict the parking position of two-wheeled electric vehicles. The second layer is used to determine whether there is enough space in front of two-wheeled electric vehicles. The third layer contains two models: a red light behavior decision model for two-wheeled electric vehicles and an early start time distribution model for two-wheeled electric vehicles. These models are used to predict the behavior of two-wheeled electric vehicles in front of red lights and to calculate the probability density distribution of the early start time if two-wheeled electric vehicles choose to start early.

[0008] 3) Input the dataset constructed in step 1) into the multi-level decision model of two-wheeled electric vehicles established in step 2), and use the maximum likelihood estimation method to estimate the model parameters. Finally, predict the crossing behavior of two-wheeled electric vehicles based on the fitted multi-level decision model of two-wheeled electric vehicles.

[0009] Furthermore, in step 1), the types of two-wheeled electric vehicle drivers are first classified according to their travel purpose, and a mapping relationship is established between the travel purpose and the appearance of the two-wheeled electric vehicle in the video survey. The distribution of drivers classified by travel purpose as pick-up and drop-off type, commuter type, and delivery type corresponds to the distribution of drivers classified by appearance as multi-person type, single-person type, and cargo type.

[0010] Then, through video surveillance, the parking behavior of two-wheeled electric vehicles was statistically analyzed to construct a dataset of two-wheeled electric vehicle parking behavior at signalized intersections. The dataset contains two parts: parking location data and red light compliance data. The parking location data reflects the obvious clustering characteristics of two-wheeled electric vehicles with different travel purposes, which are concentrated near two stop lines. These two stop lines are the rightmost lane of the straight-ahead motor vehicle lane in the direction conflicting with the travel direction of two-wheeled electric vehicles and the stop line of motor vehicles in the same direction as the travel direction of two-wheeled electric vehicles, referred to as stop line 1 and stop line 2. The red light compliance data includes the red light compliance of drivers with different travel purposes and different intersection clearance conditions, divided into three categories: normal start after stopping, early start after stopping, and non-stop passage. At the same time, the dataset also includes the red light waiting time of two-wheeled electric vehicle drivers under these three red light compliance conditions.

[0011] Furthermore, in step 2), the multi-layer decision-making model for the two-wheeled electric vehicle is described in detail below:

[0012] The first layer involves an individual entering an intersection and determining their parking position, i.e., selecting their parking line using a two-wheeled electric vehicle parking position decision model. The second layer involves an individual deciding whether to proceed based on whether there is enough space ahead. If there is insufficient space, they need to slow down and stop until there is enough space ahead. When the vehicle reaches its chosen parking position, it enters the third layer and responds to the traffic light. If the traffic light is green, it proceeds directly. If the traffic light is red, it enters the two-wheeled electric vehicle red light behavior decision model and decides to take one of the following three actions: proceed directly without stopping, stop until the green light before proceeding, or stop but start moving in advance before the red light turns green. For the latter two actions, the individual needs to stop until the signal condition meets their own passage conditions before they can pass through the signalized intersection. If the individual starts moving in advance, the advance start time distribution model of the two-wheeled electric vehicle needs to determine the advance time.

[0013] Furthermore, a parking location decision model for two-wheeled electric vehicles is constructed based on multivariate logistic regression, and its utility function is as follows:

[0014]

[0015] In the formula, Let n represent the probability of a two-wheeled electric vehicle stopping at each parking line, where n takes the values ​​0, 1, and 2, representing the cases of no parking line, parking line 1, and parking line 2, respectively; X1, X2, and X3 are the driver types in the two-wheeled electric vehicle parking location decision model, namely multi-person, single-person, and cargo-carrying types, respectively; β1, β2, and β3 are the regression coefficients corresponding to their respective variables; and β is a constant term.

[0016] Furthermore, a decision-making model for red-light behavior of two-wheeled electric vehicles is constructed based on multivariate logistic regression, and its utility function is as follows:

[0017]

[0018] In the formula, Let m represent the probabilities of each behavior, taking values ​​of 1, 2, and 3, representing the three behaviors: starting the green light normally after stopping, starting early after stopping, and proceeding without stopping, respectively; Y1, Y2, and Y3 represent the driver type, namely multi-person, single-person, and cargo-carrying; Y4, Y5, and Y6 represent the waiting time, respectively low, medium, and high; Y7 and Y8 represent the intersection clearance status, respectively not cleared and cleared; Y9 and Y... 10 Y 11 The number of vehicles starting in the surrounding area is categorized as few, medium, and many; α1, α2, ..., α 11 α represents the regression coefficients for each independent variable; α is the constant term.

[0019] Furthermore, a model for the early start time distribution of two-wheeled electric vehicles in a parked but early start state is established as follows:

[0020] Since the distribution of the early start time of two-wheeled electric vehicles shows a decreasing proportion of participants as the start time increases, an exponential distribution is used to describe the probability density f(t) of the early start time:

[0021] f(t) = λe -λt

[0022] In the formula, λ is the parameter of the exponential distribution, which is determined based on historical datasets; t is the advance start time of the two-wheeled electric vehicle.

[0023] Furthermore, in step 3), the maximum likelihood estimation method is used to estimate the model parameters of the two-wheeled electric vehicle parking position decision model and the two-wheeled electric vehicle red light behavior decision model. The estimation principle is to find a set of parameter values ​​that maximize the probability of observing the current sample data under this set of parameter values.

[0024] Then, predict the crossing behavior of two-wheeled electric vehicles. The specific process is as follows: based on the fitted multi-layer decision model of two-wheeled electric vehicles, input the data required for each layer, and predict the parking position and parking behavior of two-wheeled electric vehicles in front of traffic lights in the first and third layers; if the two-wheeled electric vehicle is in a stopped but early start state, the early start time distribution of the two-wheeled electric vehicle will be calculated in the third layer.

[0025] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0026] 1. This invention proposes the concept of "dual parking lines" (i.e., parking line 1 and parking line 2) for the first time, which characterizes the real parking behavior of two-wheeled electric vehicles, reflects the spatial usage behavior of two-wheeled electric vehicles in current urban traffic, and makes up for the lack of modeling of parking space selection in existing studies.

[0027] 2. Compared with traditional single-behavior modeling methods, the two-wheeled electric vehicle street crossing decision-making behavior description method proposed in this invention covers three levels of behavioral decision-making: "parking location selection - whether to stop - red light behavior and prediction of early start time". It can systematically and in stages reflect the behavior path of the two-wheeled electric vehicle driver from approaching the intersection to passing through the intersection, thereby improving the explanatory power and predictive ability of the model.

[0028] 3. This invention maps travel purpose to the visual appearance features of electric vehicles, and establishes a driver type identification mechanism. This enables the multi-layer decision-making model for two-wheeled electric vehicles to determine individual behavioral characteristics through video analysis without relying on questionnaire data, thereby improving the applicability of the model and the efficiency of data collection.

[0029] 4. This invention supports model parameter correction based on the behavioral characteristics of two-wheeled electric vehicles in different cities or environments, and has good scalability and engineering application value. Attached Figure Description

[0030] Figure 1 This is a framework diagram of the method of the present invention.

[0031] Figure 2 This is a schematic diagram of the "double parking line" concept proposed in this invention on an actual road.

[0032] Figure 3 This is a flowchart of the multi-level decision-making model for two-wheeled electric vehicles according to the present invention. Detailed Implementation

[0033] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0034] like Figure 1 As shown, this embodiment discloses a method for predicting the pedestrian crossing behavior of two-wheeled electric vehicles based on a multi-level decision model. By collecting and analyzing actual video recording data, a multi-level parking behavior decision model for two-wheeled electric vehicles is established to predict their pedestrian crossing behavior. The method includes the following steps:

[0035] 1) Select several typical signal intersections in a city and use drones to take aerial video recordings of each collection point. The recording time should be 10 to 15 minutes, and the collection time should be during the morning rush hour on weekdays.

[0036] First, the types of drivers were classified, and a direct mapping relationship was established between the purpose of travel and the appearance of the two-wheeled electric vehicles in the video survey. The distribution of drivers classified by travel purpose as pick-up and drop-off, commuter, and delivery corresponds to the classification by appearance as multi-person, single-person, and cargo-carrying drivers. The classification criteria are as follows: it can be seen from the collected videos that the two-wheeled electric vehicles driven by delivery drivers are equipped with square delivery boxes, and this is used as the appearance feature to identify cargo-carrying individuals. Pick-up and drop-off drivers, because they carry passengers, have a significantly larger total body coverage area and a significantly wider overall outline than other vehicle types, and this is used as the appearance feature to identify multi-person two-wheeled electric vehicle drivers. Individuals who do not have these two characteristics are considered to be general commuter individuals, i.e., single-person drivers.

[0037] Using the physical analysis software Tracker, coordinate data of two-wheeled electric vehicles and traffic facilities at signalized intersections were collected at different times using manual marking methods.

[0038] Then, parking behavior data for two-wheeled electric vehicles was collected and statistically analyzed to construct a dataset of parking behavior for two-wheeled electric vehicles at signalized intersections. This dataset contains two parts: parking location data and red light compliance data. The parking location data reflects a clear clustering characteristic of two-wheeled electric vehicles with different travel purposes, concentrated near two parking lines. These two parking lines are the rightmost lane of the straight-ahead motor vehicle lane (opposing the direction of travel for two-wheeled electric vehicles) and the parking line for motor vehicles traveling in the same direction as the two-wheeled electric vehicles; these are referred to as parking line 1 and parking line 2. This phenomenon is called the "double parking line phenomenon." Figure 2 As shown; the red light compliance data includes drivers' red light compliance under different travel purposes and different intersection clearance conditions, divided into three categories: normal start after stopping, early start after stopping, and non-stop passage. At the same time, the dataset also includes the red light waiting time of two-wheeled electric vehicle riders under these three categories of red light compliance conditions;

[0039] Overall, in this dataset, most vehicles choose parking line 1 as their parking line. The parking process is generally as follows: vehicles that arrive first and decide to park usually do not change their lateral position. After driving straight to the chosen parking line, they generally park within 1 meter of the parking line. Vehicles that arrive later prioritize adjusting their direction to park alongside the vehicles parked in front, increasing the width of the two-wheeled electric vehicle parking cluster until the left boundary approaches the inner lane of the motor vehicle lane. Subsequent vehicles then park behind the already parked two-wheeled electric vehicles, forming a queue. Subsequent vehicles also prioritize parking alongside the vehicles in front.

[0040] Finally, data on red light compliance of two-wheeled electric vehicles was collected and statistically analyzed. The main analysis focused on individual red light compliance under different travel purposes and different intersection clearance conditions. The percentage of drivers who chose to stop and start normally, stop and start early, and pass through without stopping under different conditions was statistically analyzed. In addition, the time required for an individual to wait for the traffic light to turn green was calculated by subtracting the individual's arrival time from the time the green light turned on in the current cycle, and the red light compliance behavior of drivers under different waiting times was statistically analyzed.

[0041] 2) Construct a multi-layer decision-making model for two-wheeled electric vehicles. The model consists of three layers. The first layer is the parking position decision model for two-wheeled electric vehicles (referred to as the "parking position decision model" in the figure), which is used to predict the parking position of two-wheeled electric vehicles. The second layer is used to determine whether there is enough space in front of two-wheeled electric vehicles. The third layer contains two models: a red light behavior decision model for two-wheeled electric vehicles (referred to as the "red light behavior decision model" in the figure) and an early start time distribution model for two-wheeled electric vehicles (referred to as the "early start time distribution model" in the figure), which are used to predict the behavior of two-wheeled electric vehicles in front of red lights and calculate the probability density distribution of the early start time if two-wheeled electric vehicles choose to start early.

[0042] like Figure 3 As shown, the decision-making process of this multi-layer decision-making model is as follows: The first layer is when an individual enters the intersection and determines its own parking position, that is, selects its own parking line through the two-wheeled electric vehicle parking position decision model; the second layer is when the individual decides whether to pass based on whether there is enough space ahead. If there is not enough space, it needs to slow down and stop to wait until there is enough space ahead; when the vehicle moves to its chosen parking position, it enters the third layer and responds to the traffic light. If the traffic light is green, it passes directly. If the traffic light is red, it enters the two-wheeled electric vehicle red light behavior decision model and decides to take one of the following three actions: pass directly without stopping, stop until the green light before passing, or stop but start in advance before the red light turns green; for the latter two actions, it is necessary to stop until the signal state meets its own passage conditions before it can pass through the signalized intersection. If it starts in advance, it is necessary to determine the advance time through the two-wheeled electric vehicle advance start time distribution model;

[0043] The method for constructing a parking location decision model for two-wheeled electric vehicles is as follows:

[0044] Based on multivariate logistic regression, a decision-making model for parking locations of two-wheeled electric vehicles is constructed, and its utility function is:

[0045]

[0046] In the formula, Let n represent the probability that a two-wheeled electric vehicle stops at each parking line. n can take the values ​​0, 1, and 2, representing the cases of no parking line, parking line 1, and parking line 2, respectively. X1, X2, and X3 are the driver types in this decision model, namely multi-person, single-person, and cargo-carrying types, respectively. β1, β2, and β3 are the regression coefficients corresponding to their respective variables. β is a constant term.

[0047] The third layer contains two models. First, a decision-making model for red-light behavior of two-wheeled electric vehicles is established based on multiple logistic regression, and its utility function is:

[0048]

[0049] In the formula, Let m represent the probabilities of each behavior, where m can take values ​​of 1, 2, and 3, representing the three behaviors: starting the green light normally after stopping, starting early after stopping, and proceeding without stopping, respectively; Y1, Y2, and Y3 represent the driver type, namely multi-person, single-person, and cargo-carrying; Y4, Y5, and Y6 represent the waiting time, namely low, medium, and high; Y7 and Y8 represent the intersection clearance status, namely not cleared and cleared; Y9 and Y... 10 Y 11 The number of vehicles starting in the surrounding area is categorized as few, medium, and many; α1, α2, ..., α 11 These are the regression coefficients for each independent variable; α is the constant term.

[0050] Then, a model for the early start time distribution of two-wheeled electric vehicles is established to describe the "parked but early start" state:

[0051] Since the distribution of the early start time of two-wheeled electric vehicles shows a decreasing proportion of participants as the start time increases, an exponential distribution is used to describe the probability density f(t) of the early start time:

[0052] f(t) = λe -λt

[0053] In the formula, λ is the parameter of the exponential distribution, which can be selected based on historical datasets; t is the advance start time of the two-wheeled electric vehicle.

[0054] 3) Input the dataset constructed in step 1) into the multi-level decision model of the two-wheeled electric vehicle established in step 2), and use the maximum likelihood estimation method to estimate the model parameters of the parking position decision model and the red light behavior decision model of the two-wheeled electric vehicle. The basic principle of estimation is to find a set of parameter values ​​that maximize the probability of observing the current sample data under this set of parameter values.

[0055] Then, predict the crossing behavior of two-wheeled electric vehicles. The specific process is as follows: based on the fitted multi-layer decision model of two-wheeled electric vehicles, input the data required for each layer, and predict the parking position and parking behavior of two-wheeled electric vehicles in front of traffic lights in the first and third layers; if the two-wheeled electric vehicle is in the "parked but started early" state, the early start time distribution of the two-wheeled electric vehicle will be calculated in the third layer.

[0056] In this example, the fitted two-wheeled electric vehicle parking location decision model is as follows:

[0057]

[0058] The fitted red light behavior decision model for two-wheeled electric vehicles is as follows:

[0059]

[0060] By inputting the dataset into a multi-level decision model for two-wheeled electric vehicles, it is possible to predict the parking location and parking decisions when crossing the street.

[0061] For the early start time distribution model of two-wheeled electric vehicles in the multi-level decision-making model for two-wheeled electric vehicles, in this example, we take λ = 0.2394, that is:

[0062] f(t) = 0.2394e -0.2394t

[0063] Calculations show that if the two-wheeled electric vehicle in this case makes the behavioral decision of "stopping but starting early", the average early start time of the two-wheeled electric vehicle is 4.18s.

[0064] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model, characterized in that, Includes the following steps: 1) A dataset of parking behavior of two-wheeled electric vehicles at signalized intersections was constructed through video surveillance. The dataset includes parking location data and red light compliance data. 2) Construct a multi-layer decision-making model for two-wheeled electric vehicles. The model is divided into three layers. The first layer is the parking location decision model for two-wheeled electric vehicles, which is used to predict the parking location of two-wheeled electric vehicles. The second layer determines whether there is enough space in front of the two-wheeled electric vehicle; the third layer contains two models, namely the two-wheeled electric vehicle red light behavior decision model and the two-wheeled electric vehicle early start time distribution model, which are used to predict the behavior of the two-wheeled electric vehicle in front of the red light and calculate the probability density distribution of the early start time if the two-wheeled electric vehicle chooses to start early. 3) Input the dataset constructed in step 1) into the multi-level decision model of two-wheeled electric vehicles established in step 2), and use the maximum likelihood estimation method to estimate the model parameters. Finally, predict the crossing behavior of two-wheeled electric vehicles based on the fitted multi-level decision model of two-wheeled electric vehicles.

2. The method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model according to claim 1, characterized in that, In step 1), the types of two-wheeled electric vehicle drivers are first classified according to their travel purpose, and a mapping relationship is established between the travel purpose and the appearance of the two-wheeled electric vehicle in the video survey. The distribution of drivers classified by travel purpose as pick-up and drop-off type, commuter type, and delivery type corresponds to the distribution of drivers classified by appearance as multi-person type, single-person type, and cargo type. Then, through video surveillance, the parking behavior of two-wheeled electric vehicles was statistically analyzed to construct a dataset of two-wheeled electric vehicle parking behavior at signalized intersections. The dataset contains two parts: parking location data and red light compliance data. The parking location data reflects the obvious clustering characteristics of two-wheeled electric vehicles with different travel purposes, which are concentrated near two stop lines. These two stop lines are the rightmost lane of the straight-ahead motor vehicle lane in the direction conflicting with the travel direction of two-wheeled electric vehicles and the stop line of motor vehicles in the same direction as the travel direction of two-wheeled electric vehicles, referred to as stop line 1 and stop line 2. The red light compliance data includes the red light compliance of drivers with different travel purposes and different intersection clearance conditions, divided into three categories: normal start after stopping, early start after stopping, and non-stop passage. At the same time, the dataset also includes the red light waiting time of two-wheeled electric vehicle drivers under these three red light compliance conditions.

3. The method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model according to claim 2, characterized in that, In step 2), the multi-layer decision-making model for the two-wheeled electric vehicle is described in detail below: The first layer involves an individual entering an intersection and determining their parking position, i.e., selecting their parking line using a two-wheeled electric vehicle parking position decision model. The second layer involves an individual deciding whether to proceed based on whether there is enough space ahead. If there is insufficient space, they need to slow down and stop until there is enough space ahead. When the vehicle reaches its chosen parking position, it enters the third layer and responds to the traffic light. If the traffic light is green, it proceeds directly. If the traffic light is red, it enters the two-wheeled electric vehicle red light behavior decision model and decides to take one of the following three actions: proceed directly without stopping, stop until the green light before proceeding, or stop but start moving in advance before the red light turns green. For the latter two actions, the individual needs to stop until the signal condition meets their own passage conditions before they can pass through the signalized intersection. If the individual starts moving in advance, the advance start time distribution model of the two-wheeled electric vehicle needs to determine the advance time.

4. The method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model according to claim 3, characterized in that, A parking location decision model for two-wheeled electric vehicles is constructed based on multivariate logistic regression, and its utility function is as follows: In the formula, Let n represent the probability of a two-wheeled electric vehicle stopping at each parking line, where n takes the values ​​0, 1, and 2, representing the cases of no parking line, parking line 1, and parking line 2, respectively; X1, X2, and X3 are the driver types in the two-wheeled electric vehicle parking location decision model, namely multi-person, single-person, and cargo-carrying types, respectively; β1, β2, and β3 are the regression coefficients corresponding to their respective variables; and β is a constant term.

5. The method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model according to claim 4, characterized in that, A red light behavior decision-making model for two-wheeled electric vehicles is constructed based on multivariate logistic regression, and its utility function is as follows: In the formula, Let m represent the probabilities of each behavior, taking values ​​of 1, 2, and 3, representing the three behaviors: starting the green light normally after stopping, starting early after stopping, and proceeding without stopping, respectively; Y1, Y2, and Y3 represent the driver type, namely multi-person, single-person, and cargo-carrying; Y4, Y5, and Y6 represent the waiting time, respectively low, medium, and high; Y7 and Y8 represent the intersection clearance status, respectively not cleared and cleared; Y9 and Y... 10 Y 11 The number of vehicles starting in the surrounding area is categorized as few, medium, and many; α1, α2, ..., α 11 α represents the regression coefficients for each independent variable; α is the constant term.

6. The method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model according to claim 5, characterized in that, A model for the early start time distribution of a two-wheeled electric vehicle in a parked but early start state is established as follows: Since the distribution of the early start time of two-wheeled electric vehicles shows a decreasing proportion of participants as the start time increases, an exponential distribution is used to describe the probability density f(t) of the early start time: f(t)=λe -λt In the formula, λ is the parameter of the exponential distribution, which is determined based on historical datasets; t is the advance start time of the two-wheeled electric vehicle.

7. The method for predicting the pedestrian behavior of two-wheeled electric vehicles based on a multi-level decision model according to claim 6, characterized in that, In step 3), the maximum likelihood estimation method is used to estimate the model parameters of the two-wheeled electric vehicle parking position decision model and the two-wheeled electric vehicle red light behavior decision model. The estimation principle is to find a set of parameter values ​​that maximize the probability of observing the current sample data under this set of parameter values. Then, predict the crossing behavior of two-wheeled electric vehicles. The specific process is as follows: based on the fitted multi-layer decision model of two-wheeled electric vehicles, input the data required for each layer, and predict the parking position and parking behavior of two-wheeled electric vehicles in front of traffic lights in the first and third layers; if the two-wheeled electric vehicle is in a stopped but early start state, the early start time distribution of the two-wheeled electric vehicle will be calculated in the third layer.

Images