Method for representing mixed traffic interaction behavior based on safety field theory

Abstract

The application discloses a kind of mixed traffic interaction behavior characterization method based on safety field theory;Its core steps include: obtaining the high-density mixed traffic trajectory data of target section;Build comprehensive safety field model, the model is by introducing the equivalent mass correction formula based on speed-mass coupling effect and the differential potential field function distinguishing between networked automatic driving vehicles and artificial driving vehicles, to quantify the superimposed influence of road boundary and surrounding vehicles on target vehicle;Based on the model, calculate potential field index to assess safety risk;At the same time, according to the dynamic change of vehicle relative position, the interaction process is subdivided into at least eight categories of scenarios, and combined analysis is carried out according to vehicle type, so as to realize the fine and comprehensive characterization of the micro-behavior characteristics of different vehicles in different strong interaction scenarios.The application improves the accuracy and scene pertinence of mixed traffic flow micro-behavior modeling.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation technology, and in particular relates to a method for representing hybrid traffic interaction behavior based on safety field theory. Background Technology

[0002] With the evolution of vehicle-to-everything (V2X) and autonomous driving technologies, the analysis of mixed traffic flow characteristics, including driverless vehicles (HDVs), infrastructure support vehicles (ISVs), and connected autonomous vehicles (CAVs), has become a key cornerstone for the development of intelligent transportation systems. In highly dynamic interaction scenarios such as merging and diverging zones on highways, frequent lane-changing, merging, and merging behaviors between vehicles constitute a complex micro-interaction system. However, traditional theoretical methods such as car-following models and lane-changing models are insufficient to accurately characterize the interaction mechanisms among multiple types of traffic participants in such highly interactive scenarios.

[0003] Existing technologies have limitations at the data foundation level. Currently, mainstream trajectory datasets used for traffic behavior research generally suffer from problems such as limited scene diversity, insufficient data density, or lack of contextual information, hindering the refinement of models. As shown in Table 1, the NGSIM dataset, based on fixed-point camera data, has a relatively short recording duration (approximately 1.5 hours) and primarily depicts basic road sections, failing to adequately characterize merging / diversion zones. While the highD dataset utilizes drone photography to increase data scale, it lacks precise geographic reference information, and the trajectories are mainly straight-line travel, making it difficult to reflect complex interactions. The Automatum dataset has a longer recording duration, but its traffic flow density is low (approximately 33 vehicles / minute), failing to fully reflect high-density interaction conflicts.

[0004] Table 1 Comparison of existing road user trajectory datasets in the highway sector

[0005]

[0006] In summary, existing data acquisition and representation methods have shortcomings in terms of scene complexity, data density, and characterization of vehicle interaction relationships, making it difficult to support accurate and reliable modeling and analysis of the micro-level interactive behavior of mixed traffic flows. Therefore, there is an urgent need to develop a new method for micro-level behavioral representation and modeling that can integrate high-density trajectory data, distinguish vehicle type differences, and is applicable to highly interactive scenarios. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for characterizing mixed traffic interaction behavior based on safety field theory, so as to solve the problem of quantitative characterization of the microscopic behavior of mixed traffic flow of multiple types of vehicles in merging zone by introducing equivalent mass correction and differentiated potential field function.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is a method for representing hybrid traffic interaction behavior based on safety field theory, comprising the following steps:

[0009] Acquire trajectory data of each vehicle in the mixed traffic flow within the target area;

[0010] A comprehensive safety field model is constructed based on the trajectory data of each vehicle, wherein the equivalent mass of a vehicle in the comprehensive safety field model increases with the increase of the vehicle's speed.

[0011] Based on the comprehensive safety field model, a potential field index characterizing the degree of safety risk at the location of the target vehicle is calculated.

[0012] Furthermore, the construction of the comprehensive safety field model based on the trajectory data of each vehicle includes:

[0013]

[0014] in, for surrounding vehicles Equivalent quality, for surrounding vehicles The actual quality for surrounding vehicles The current speed, and These are parameters determined by fitting accident data.

[0015] Furthermore, based on the aforementioned comprehensive safety field model, potential field indicators characterizing the degree of safety risk at the location of the target vehicle are calculated, including:

[0016] Determine the potential field indicators for target vehicle j:

[0017]

[0018] For the target vehicle Potential market indicators For the potential energy at the road boundary, Let i be the sum of the potential energy of the surrounding vehicles. and These are the weighting coefficients for the road boundary and vehicle potential energy, respectively; where:

[0019]

[0020] In the formula, i represents surrounding vehicles, and j represents the target vehicle. for surrounding vehicles At the location of target vehicle j The potential field strength generated at that location.

[0021] Furthermore, the specific potential field strength of the vehicle is as follows:

[0022]

[0023] In the formula: i represents surrounding vehicles, j represents the target vehicle. For vehicle i, the location of target vehicle j is... The potential field strength generated at that location; λ is the driving characteristic parameter. For equivalent quality; Let Euclidean distance be the distance from the point to the vehicle's center of mass. The angle between the vehicle's direction of motion and the line connecting the vehicle and the spatial point; This is the acceleration influence coefficient; It is acceleration.

[0024] Furthermore, after acquiring the trajectory data, the method further includes: determining the interaction scenario of the target vehicle into one of the following eight categories based on the relative position changes of the target vehicle with respect to the vehicles in front and behind it:

[0025] Scenario A: No car behind and no car in front;

[0026] Scenario B: No car behind but car in front;

[0027] Scenario C: There is no following vehicle, and the original following vehicle overtakes the target vehicle and becomes the leading vehicle;

[0028] Scenario D: There is a car behind but no car in front;

[0029] Scenario E: There is a car behind and a car in front;

[0030] Scenario F: There is a following vehicle, and the original following vehicle overtakes the target vehicle and becomes the leading vehicle;

[0031] Scenario G: There is a following vehicle, and the target vehicle overtakes the original preceding vehicle, making it the following vehicle. At this time, there is no preceding vehicle.

[0032] Scenario H: There is a following vehicle, and the target vehicle overtakes the original leading vehicle, making it the following vehicle. At this time, there is a leading vehicle.

[0033] Furthermore, acquiring the trajectory data of each vehicle in the mixed traffic flow within the target area includes:

[0034] Obtain a high-precision map of the target area and map the trajectory data to the geographic coordinate system corresponding to the high-precision map;

[0035] The interaction scenario of the target vehicle is determined into one of eight categories based on the relative position changes of the target vehicle with the vehicles in front and behind it, including:

[0036] When the longitudinal distance between the target vehicle and the following vehicle exceeds a set threshold, it is determined that there is no following vehicle.

[0037] When the longitudinal distance between the target vehicle and the vehicle in front exceeds a set threshold, it is determined that there is no vehicle in front.

[0038] Furthermore, after determining the interaction scenario, it also includes:

[0039] The vehicle types are combined with the eight types of interaction scenarios for analysis, wherein the vehicle types include cars, trucks and buses;

[0040] For each type of vehicle and its interaction behavior in each interaction scenario, the merging speed, merging distance, lateral speed, lateral acceleration, and the potential field index are extracted and analyzed.

[0041] Furthermore, the comprehensive safety field model also includes a road boundary field; the potential field intensity generated by the road boundary field at the target vehicle j. for:

[0042]

[0043] For road boundary indexing, Corresponding to the left boundary, Corresponding to the right boundary; For the first The constraint strength coefficient of the road boundary; For the target vehicle To the The distance vector of the road boundary; This corresponds to the Euclidean distance modulus.

[0044] Furthermore, acquiring the trajectory data of each vehicle in the mixed traffic flow within the target area includes:

[0045] High-precision vehicle trajectory data, including merging and / or diverging zones on highways, is acquired by drone photography. The mixed traffic flow includes manually driven vehicles (HDVs) and connected autonomous vehicles (CAVs).

[0046] Furthermore, the method also includes:

[0047] Based on the potential field indicators, target vehicles and corresponding interaction scenarios with safety risks exceeding a preset threshold are identified;

[0048] Based on the identification results, differentiated safety warnings or driving decision support information are generated.

[0049] Compared with the prior art, the beneficial effects of the present invention include the following points:

[0050] (1) This invention subdivides the interactive scenarios of the merging / diverting zones of highways into eight categories (Scenario A to Scenario H), and determines the scenarios based on the existence status and dynamic position changes of the vehicles in front and behind (such as the rear vehicle becoming the front vehicle, and the front vehicle becoming the rear vehicle). Compared with the rough scenario division method in the prior art, this invention achieves a more refined classification and representation of the micro-behavior of vehicles in strong interactive scenarios.

[0051] (2) Based on eight types of interactive scenarios, this invention conducts a quantitative comparative analysis of the potential field index (PFI) under different scenarios. For the first time, it reveals that the PFI values ​​of scenario C (no following vehicle, the original following vehicle becomes the preceding vehicle) and scenario F (with following vehicle, the original following vehicle becomes the preceding vehicle) are significantly lower than those of other scenarios, indicating higher safety risks. It clarifies that the moment when the vehicle relationship changes dynamically is the critical window period for risk management in the merging zone, providing scenario-specific data support for differentiated safety warnings and driving decisions.

[0052] (3) This invention combines vehicle types (cars, trucks, buses) with eight types of interactive scenarios to form 24 sub-situations. It extracts multi-dimensional indicators such as merging speed, merging distance, lateral speed, lateral acceleration and PFI respectively, realizing a comprehensive quantitative characterization of the micro-behavioral characteristics of different vehicle types in different interactive scenarios, making up for the lack of fine-grained scenario analysis in the existing technology.

[0053] (4) This invention constructs a comprehensive safety field model that includes the road boundary field and the vehicle potential field, and introduces an equivalent mass correction formula based on the velocity-mass coupling effect. This solves the problem that existing safety field models treat vehicle mass as a constant value and cannot reflect the amplification effect of high-speed driving risks. It makes the risk assessment of the safety field model in the high-interaction scenario of highways more in line with the statistical laws of accidents, and improves the physical rationality and accuracy of the model output. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of the method of this embodiment;

[0056] Figure 2 This is a schematic diagram of the vehicle safety field distribution in this embodiment;

[0057] Figure 3 This is a schematic diagram of the electric field strength around the target vehicle in this embodiment;

[0058] Figure 4 This is a schematic diagram of the interaction behavior analysis framework for this implementation method;

[0059] Figure 5 This is a schematic diagram of trajectory data acquisition scenarios in this embodiment, where (a) is trajectory acquisition in scenario 2, (b) is trajectory acquisition in scenario 3, (c) is trajectory acquisition in scenario 5, and (d) is trajectory acquisition in scenario 6.

[0060] Figure 6 This is a diagram showing the mapping effect of trajectory data on a high-precision map in this embodiment. (a) is the total trajectory mapping, (b) is the car trajectory mapping, (c) is the truck trajectory mapping, and (d) is the bus trajectory mapping.

[0061] Figure 7 The following are histograms showing the distribution of PFI in different scenarios according to this embodiment: (a) is the PFI in scenario 2, (b) is the PFI in scenario 3, (c) is the PFI in scenario 5, and (d) is the PFI in scenario 6.

[0062] Figure 8 This is a comparison chart of different vehicle interaction characteristics in this embodiment, where (a) is the merging speed, (b) is the merging distance, (c) is the merging distance ratio, (d) is the merging duration, (e) is the maximum lateral velocity during merging, and (f) is the maximum lateral acceleration during merging. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] like Figure 1 As shown, this embodiment provides a method for representing hybrid traffic interaction behavior based on safety field theory; specifically, it is a method for representing and modeling hybrid traffic micro-interaction behavior based on safety field theory, and the specific implementation of the method is as follows:

[0065] Example 1

[0066] S1. Select the merging area of ​​the highway and the adjacent diverging area as the highly interactive scenario. For example... Figure 5As shown in (a) to (d), this embodiment selects different merging / diverging zones on a highway in a certain country as the research scenario, with locations 2, 3, 5, and 6 being the focus of analysis, covering typical highly interactive environments such as mainline-ramp weaving areas and continuous merging areas. The exiD dataset is used as the trajectory data source. Compared to existing datasets such as NGSIM and highD, exiD is based on drone photography, recording for 16.1 hours, extracting high-precision trajectories for 69,172 vehicles, with an average traffic flow density of approximately 71 vehicles / minute. Figure 6 As shown in (a) to (d), the trajectory data is mapped to the real geographic coordinate system using the Lanelet2 or OpenDRIVE format high-precision map provided by exiD, establishing a basis for visualization analysis, and the study area is subdivided into the main line area and the ramp area.

[0067] S2. Construct a comprehensive safety field model based on the collected trajectory data, such as... Figure 2 As shown, in this embodiment, the safety potential field around a single vehicle is regarded as a spatial distribution field centered on the vehicle. Its field strength decreases with increasing distance and exhibits an asymmetrical distribution in different directions (the potential field in front is stronger than that behind).

[0068] In some specific implementations, the integrated safety field model includes a road boundary field and a vehicle potential field.

[0069] S201. The formula for calculating the potential field strength of the road boundary field is as follows:

[0070]

[0071] In the formula: This represents the potential field generated by the road boundary for the target vehicle. The intensity of the impact is a dimensionless risk value. For road boundary indexing, Corresponding to the left boundary, Corresponding to the right boundary; For the first The constraint strength coefficient (equivalent quality parameter) of the road boundary, in units of ; For the target vehicle To the The distance vector of the road boundary; To correspond to the Euclidean distance modulus, the unit is... .

[0072] This formula is analogous to the classical conservative field potential function (such as that of a gravitational field). The inverse square decay method is used to describe the change of the safety potential field with distance. This invention applies this method to road boundaries, taking the scalar sum of the potential fields of the left and right boundaries as the total boundary potential field, since safety risks do not cancel each other out. Dimensionally, Quantity , and the denominator After cancellation, it becomes dimensionless, consistent with the left side.

[0073] S202. Specifically, the vehicle potential field reflects the range of safety influence of the moving vehicle on the surrounding space. Its construction first considers the velocity-mass coupling effect of the vehicle. The derivation process of the vehicle's equivalent mass formula is as follows:

[0074] S2021, Kinetic Energy Equivalence Principle. The kinetic energy of a vehicle. The greater the kinetic energy, the greater the destructive force generated when a vehicle collides, and the higher the safety risk.

[0075] In the formula: for surrounding vehicles The actual physical mass, in units of ; for surrounding vehicles Instantaneous velocity, in units of ; and These are empirical parameters fitted using accident data. Dimensions are , Dimensionless. This formula is derived from the kinetic energy formula. The power-law relationship between accident severity and vehicle speed was determined through regression analysis to identify the parameters. Dimensionally, Dimensionless, both sides of the equation are dimensionless. Therefore, the equivalent mass should reflect the amplifying effect of speed on safety risks, thus satisfying consistency requirements.

[0076] It should be noted that in this embodiment, the indexes i and j are the indices of the surrounding vehicles and the target vehicle, respectively.

[0077] S2022, Accident Data Statistical Analysis. Based on the statistical analysis of highway vehicle speed and accident data, the severity of accidents exhibits a non-linear relationship with vehicle speed. Through regression analysis of a large amount of accident data, a significant power function relationship between accident severity and vehicle speed was found. Therefore, [the following is introduced].

[0078] S2023, Parameter Calibration. Parameters are determined by fitting data from measured highway speeds and corresponding accident statistics, using the least squares method. and The value of . Specifically, using accident incidence or accident severity as the dependent variable and vehicle speed as the independent variable, a power function regression model is established. , Indicates vehicle speed of The power of, where The parameter is a power-law parameter (dimensionless) obtained by fitting accident data. The parameter is obtained through data fitting. and to determine , .

[0079] S2024. Verification and Correction. The calibrated equivalent mass formula is applied to the safety field model, and the accuracy of the model is verified by comparing it with actual traffic flow data. If there is a discrepancy between the safety risks predicted by the model and the actual accident data, the parameters are fine-tuned until the model output matches the actual observation data.

[0080] S2025, The final formula for calculating equivalent mass is:

[0081]

[0082] in, For vehicles Equivalent quality, For vehicles The actual quality For vehicles The current speed, and These are parameters determined by fitting accident data.

[0083] S203, such as Figure 3 As shown, in a multi-vehicle scenario within a merging zone, the target vehicle is simultaneously affected by both the road boundary field and the potential field of surrounding vehicles. To differentiate the driving characteristics of connected autonomous vehicles (CAVs) and manually driven vehicles (HDVs), a differentiated potential field strength is constructed. Driving characteristic parameters are introduced. For CAV vehicles with better information accessibility and more sensitive interactive responses, set up , Driving characteristic parameters are used to distinguish between CAVs and HDVs. (Vehicle) The potential field strength at any point in the surrounding space is calculated as follows:

[0084]

[0085] In the formula: For vehicles At point The potential field strength generated at that location is dimensionless. For driving characteristic parameters, distinguish between CAV driving characteristic parameters and HDV driving characteristic parameters ( ), dimensionless; For equivalent quality, the unit is... ; The Euclidean distance from the point to the vehicle's center of mass is given by units of 1. ; The angle between the vehicle's direction of motion and the line connecting the vehicle and the spatial point; This is the acceleration influence coefficient, in units of... ; For acceleration, the unit is . .

[0086] The formula uses an exponential decay model. Instead of inverse square law, it avoids singular values ​​at close range. This is the potential field attenuation coefficient, in units of Multiply Achieve an asymmetric distribution of forward potential enhancement and backward suppression; multiply by Reflects the dynamic correction of acceleration to the potential field range; introduces Differentiate the interaction sensitivity of vehicles with different levels of intelligence. In practical applications, this can be achieved... Divided by reference mass To maintain the dimensionless nature.

[0087] S3. Based on the above safety field model, the Potential Field Index (PFI) is defined as the standard for assessing vehicle interaction risks. Target vehicle The safety field index is calculated as follows:

[0088] ,

[0089] In the formula: For the target vehicle Potential field indicators, dimensionless risk values; The potential energy at the road boundary is calculated using formula S201. The sum of the potential energy of the surrounding vehicles. , calculated and superimposed by formula S203; and These are the weighting coefficients for the road boundary and the vehicle potential energy, respectively; they are dimensionless and satisfy... It can be dynamically adjusted according to driving scenarios (such as lowering the speed when changing lanes). ,improve ).

[0090] PFI directly performs a weighted summation based on the potential field function outputs of S201 and S203, demonstrating the specific application of the safety field model at the risk assessment level. Without the potential field definition in S2, PFI cannot obtain computable numerical inputs.

[0091] S4, such as Figure 4 As shown, this embodiment establishes a complete interactive behavior analysis framework, including the following steps:

[0092] S401, High-precision map application and trajectory data preprocessing

[0093] S4011. The exiD dataset was used as the basis for analysis. This dataset contains over 16 hours of video data from seven different measurement locations at the entrances and exits of German highways, from which the trajectories of 69,172 vehicles were extracted, of which over 99% were visible in the video. Vehicle types included cars (52,621) and trucks (12,622).

[0094] S4012, Coordinate System Establishment

[0095] Using the OpenDRIVE and Lanelet2 high-definition maps provided by the exiD dataset, a precise geographic coordinate system is established. The trajectory data is then registered with the high-definition map, mapping different types of vehicle trajectories onto the map to achieve visualization and spatial positioning of the trajectory data.

[0096] S4013, Regional Division

[0097] Based on the lane location information of the study area, the weaving zone is divided into a mainline area (main lanes and auxiliary lanes) and a ramp area (acceleration lanes and deceleration lanes). The weaving zone is further subdivided into multiple smaller areas for more refined analysis.

[0098] S4014, Data Cleaning and Filtering

[0099] Outlier detection and removal are performed on the raw trajectory data. Kalman filtering or moving average filtering methods are used to smooth the trajectory data, eliminating sensor noise and tracking errors. Missing data is interpolated to ensure the continuity and integrity of the trajectory data.

[0100] S4015, Vehicle Matching

[0101] Based on deep learning-based target detection and tracking algorithms, the system matches vehicles around the target vehicle and identifies the surrounding vehicles (front vehicle, rear vehicle, and side vehicle) of each vehicle, providing a foundation for subsequent interaction behavior analysis.

[0102] S402, Extraction of Interactive Behavior Indicators

[0103] S4021. Remittance / Exit Point Identification

[0104] Based on the Lanelet2 high-precision map, the precise locations of vehicles entering and leaving the weaving zone are determined. The merging point is defined as the position where a vehicle first enters the main lane from the acceleration lane, and the merging point is defined as the position where a vehicle first enters the deceleration lane from the main lane.

[0105] S4022, Calculation of inbound and outbound distances

[0106] Calculate the distance from the starting point of the vehicle entering the acceleration lane to the merging point (merging distance), and the distance from the merging point to the end point of the vehicle leaving the deceleration lane (merging distance). Through statistical analysis, determine the distribution characteristics of the merging and merging distances.

[0107] S4023, Calculation of Inbound and Outbound Transfer Times

[0108] Calculate the time required for merging and retrieving vehicles based on the merging and retrieving distances and vehicle speeds. Analyze the differences in merging and retrieving times for different vehicle types (cars, trucks).

[0109] S4024, Lateral Motion Feature Extraction

[0110] Extract the lateral velocity and lateral acceleration of the vehicle to analyze its lateral motion characteristics during merging and merging processes. The lateral velocity and acceleration of passenger cars are typically significantly higher than those of trucks and vans.

[0111] S4025, Calculation of Safety Field Indicators

[0112] Based on the safety field model, the PFI (Potential Field Index) value of each vehicle at each time step is calculated. The PFI value comprehensively considers the superposition effect of the road boundary field and the potential field of surrounding vehicles, and quantitatively assesses the safety risk level of the vehicle.

[0113] S403, fine-grained parallel vehicle recognition and scene classification

[0114] S4031, Parallel Vehicle Recognition

[0115] Based on the relative position of the vehicle with surrounding vehicles, identify the vehicle in front, behind, and alongside the target vehicle. The vehicle in front is defined as a vehicle located in front of the target vehicle with a longitudinal distance of less than a threshold (100m / 150m / 200m), and the vehicle behind is defined as a vehicle located behind the target vehicle with a longitudinal distance of less than a threshold.

[0116] S4032, Interaction Scenario Definition

[0117] Based on the presence and positional changes of the vehicles in front and behind, the interaction scenarios are divided into eight categories:

[0118] Scenario A: No following vehicle, no preceding vehicle (None, None)

[0119] Scenario B: No following vehicle, but one in front (None, Exist)

[0120] Scenario C: No following vehicle, the original following vehicle becomes the leading vehicle (None, Exist(reartolead))

[0121] Scenario D: There is a car behind, but no car in front (Exist, None)

[0122] Scenario E: There is a car behind and a car in front (Exist, Exist)

[0123] Scenario F: There is a car behind, and the original car behind becomes the car in front (Exist, Exist(reartolead)).

[0124] Scenario G: The car in front becomes the car behind, and there is no car in front (Exist(leaderear)).

[0125] Scenario H: The car in front becomes the car behind, and there is a car in front (Exist(leaderear),Exist).

[0126] S4033. Determine the scene determination logic. This implementation method dynamically divides the scene based on the relative positional changes of the vehicle and its surrounding environment. The specific determination rules are as follows:

[0127] When the longitudinal distance between the target vehicle and the following vehicle exceeds a set threshold, it is determined that there is no following vehicle.

[0128] When the longitudinal distance between the target vehicle and the vehicle in front exceeds a set threshold, it is determined that there is no vehicle in front.

[0129] When the vehicle that was originally behind overtakes the target vehicle and becomes the vehicle in front, it is determined to be scenario C or F.

[0130] When the target vehicle overtakes the vehicle in front, making it the vehicle behind, it is determined to be scenario G or H.

[0131] In the vehicle type and scenario combination analysis, based on vehicle type (car, truck, bus) and eight interaction scenarios, there are a total of 24 sub-cases (3 vehicle types × 8 interaction scenarios). For each sub-case, it is necessary to analyze its interaction behavior characteristics and safety risks separately.

[0132] S404, Security Analysis

[0133] S4041, Scenario-based PFI analysis

[0134] The PFI (Pressure Factor Index) values ​​of vehicles were calculated under eight interaction scenarios to analyze the differences in safety risks across different scenarios. The analysis results show that scenarios C and F have significantly lower PFI values ​​than other scenarios at each distance threshold, for different vehicle types, and at each location, indicating that these two scenarios have higher safety risks. Figure 7 As shown in (a) to (d), Figure 7 The average PFI distribution for eight interaction scenarios is presented in histogram form. The graph clearly shows that the PFI values ​​for scenarios C (no following vehicle, the original following vehicle becomes the preceding vehicle) and F (a following vehicle exists, the original following vehicle becomes the preceding vehicle) are significantly lower than those for scenarios B, D, and E. According to safety field theory, a lower PFI value indicates a higher superimposed potential field strength at the vehicle's location, and a greater risk of collision. In locations 2, 3, and 6, the above conclusions are consistent with... Figure 7 The low PFI distributions shown corroborate each other, jointly indicating that the moment when the vehicle relationship undergoes a dynamic transformation (the following vehicle overtakes and becomes the preceding vehicle) is the critical window period for risk management in the merging zone.

[0135] S4042, Vehicle Type Difference Analysis

[0136] A comparative analysis was conducted on the safety performance differences between cars and trucks under different scenarios. The results show that:

[0137] The average and median merging speeds of cars were significantly higher than those of trucks and vans;

[0138] The merging distance for trucks is significantly longer than that for cars and vans;

[0139] Cars have a shorter preparation time and a significantly higher lateral speed than trucks;

[0140] The extreme values ​​for cars are concentrated in lateral velocity and acceleration, indicating that car driving behavior is more aggressive; for example... Figure 8 As shown in (a) to (f), Figure 8 By combining box plots and scatter plots, the differences in key kinematic indicators such as merging speed, merging distance, lateral velocity, and lateral acceleration among the different vehicle types are visually displayed. The figures show that cars have higher median and outlier merging speeds than trucks, and their lateral velocity distribution is more dispersed, further confirming that cars exhibit more aggressive driving behavior and more agile maneuvering in the merging zone. Trucks and vans, on the other hand, have longer merging distances, reflecting the greater space required for safe merging. The analysis results provide data support for subsequent differentiated safety field modeling based on vehicle type.

[0141] S4043, Macro-traffic flow analysis

[0142] A macroscopic analysis of traffic flow, density, and speed is conducted to assess the overall operational status of the weaving zone. By combining microscopic interactive behaviors and macroscopic traffic flow characteristics, the traffic operation characteristics of the weaving zone are comprehensively characterized.

[0143] Based on the above scenario classification, a differentiated quantitative analysis of the micro-decision-making behaviors of three types of vehicles—HDV, ISV, and CAV—was conducted. The analysis dimensions included merging and merging characteristics (merging point location, merging distance, merging time), lateral movement characteristics (lateral velocity, lateral acceleration distribution), and risk exposure levels (PFI values ​​under different scenarios). The results show that cars can complete lane changes across the entire acceleration lane, while trucks / vans tend to merge in the middle or later sections. Cars have significantly higher merging speeds and lateral speeds than trucks, but trucks have longer merging distances. In scenarios C and F, the PFI values ​​are significantly lower than in other scenarios, indicating higher risk during vehicle relationship transitions, while the risk of traffic conflict is relatively lower in scenarios B, D, and E. Through comparative analysis, the differences in the micro-behavioral characteristics of different vehicle types on different road segments were clarified.

[0144] The comprehensive safety field model constructed in this embodiment possesses asymmetric spatial distribution characteristics. Existing safety field models typically assume a symmetrical distribution of the safety potential field around a vehicle. This invention treats the safety potential field around a single vehicle as a spatially distributed field centered on the vehicle, where the field strength decreases with increasing distance and exhibits an asymmetric distribution in different directions, specifically, the potential field in front is stronger than that behind. This aligns with the characteristic that drivers pay greater attention to the space in front during actual driving. Furthermore, existing models treat vehicle mass as a constant value. This invention introduces the concept of equivalent mass, considering the vehicle's velocity-mass coupling effect, and uses a formula... The equivalent mass is calculated, where it increases with speed. At high speeds, vehicles with the same mass face significantly greater safety risks than at low speeds, consistent with statistical patterns of highway accidents. It's worth noting that existing technologies lack differentiated modeling for CAVs and HDVs. This invention introduces driving characteristic parameters to distinguish between CAVs and HDVs. Since CAV information is more readily available and safer than HDVs, CAVs are more sensitive during interactions. Therefore, different sensitivity coefficients are used in the CAV potential field strength calculation to achieve refined modeling of mixed traffic flows. This implementation defines the PFI (Potential Field Index) as a safety assessment indicator, comprehensively considering the superposition effect of the road boundary field and the vehicle potential field. Through formula calculation, it can quantitatively assess the safety risk level of vehicles at specific locations and states. This implementation uses the exiD dataset (containing over 69,000 high-precision vehicle trajectories, representing a high-density traffic flow of approximately 71 vehicles per minute on average) for model validation. Compared to datasets such as NGSIM and highD, the selection of specific locations in the exiD dataset generates a higher density of interactive and challenging scenarios, which fully validates the applicability of the model in highly interactive scenarios (such as merging and diverging zones).

[0145] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0146] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for representing hybrid traffic interaction behavior based on safety field theory, characterized in that, Includes the following steps: Acquire trajectory data of each vehicle in the mixed traffic flow within the target area; A comprehensive safety field model is constructed based on the trajectory data of each vehicle, wherein the equivalent mass of a vehicle in the comprehensive safety field model increases with the increase of the vehicle's speed. Based on the comprehensive safety field model, a potential field index characterizing the degree of safety risk at the location of the target vehicle is calculated.

2. The method for representing hybrid traffic interaction behavior based on safety field theory according to claim 1, characterized in that, The construction of the comprehensive safety field model based on the trajectory data of each vehicle includes: in, for surrounding vehicles Equivalent quality, for surrounding vehicles The actual quality for surrounding vehicles The current speed, and These are parameters determined by fitting accident data.

3. The method for representing hybrid traffic interaction behavior based on safety field theory according to claim 2, characterized in that, Based on the comprehensive safety field model, potential field indices characterizing the degree of safety risk at the location of the target vehicle are calculated, including: Determine the potential field indicators for target vehicle j: For the target vehicle Potential market indicators For the potential energy at the road boundary, Let i be the sum of the potential energy of the surrounding vehicles. and These are the weighting coefficients for the road boundary and vehicle potential energy, respectively; where: In the formula, i represents surrounding vehicles, and j represents the target vehicle. for surrounding vehicles At the location of target vehicle j The potential field strength generated at that location.

4. The hybrid traffic interaction behavior representation method based on safety field theory according to claim 3, characterized in that, The specific potential field strength of the vehicle is as follows: In the formula: i represents surrounding vehicles, j represents the target vehicle. For vehicle i, the location of target vehicle j is... The potential field strength generated at that location; λ is the driving characteristic parameter. For equivalent quality; Let Euclidean distance be the distance from the point to the vehicle's center of mass. The angle between the vehicle's direction of motion and the line connecting the vehicle and the spatial point; This is the acceleration influence coefficient; It is acceleration.

5. The method for representing hybrid traffic interaction behavior based on safety field theory according to claim 1, characterized in that, After acquiring the trajectory data, the method further includes: determining the interaction scenario of the target vehicle into one of the following eight categories based on the relative position changes of the target vehicle with respect to the vehicles in front and behind it: Scenario A: No car behind and no car in front; Scenario B: No car behind but car in front; Scenario C: There is no following vehicle, and the original following vehicle overtakes the target vehicle and becomes the leading vehicle; Scenario D: There is a car behind but no car in front; Scenario E: There is a car behind and a car in front; Scenario F: There is a following vehicle, and the original following vehicle overtakes the target vehicle and becomes the leading vehicle; Scenario G: There is a following vehicle, and the target vehicle overtakes the original leading vehicle, making it the following vehicle. At this time, there is no leading vehicle. Scenario H: There is a following vehicle, and the target vehicle overtakes the original leading vehicle, making it the following vehicle. At this time, there is a leading vehicle.

6. The hybrid traffic interaction behavior representation method based on safety field theory according to claim 5, characterized in that, The acquisition of trajectory data for each vehicle in the mixed traffic flow within the target area includes: Obtain a high-precision map of the target area and map the trajectory data to the geographic coordinate system corresponding to the high-precision map; The interaction scenario of the target vehicle is determined into one of eight categories based on the relative position changes of the target vehicle with the vehicles in front and behind it, including: When the longitudinal distance between the target vehicle and the following vehicle exceeds a set threshold, it is determined that there is no following vehicle. When the longitudinal distance between the target vehicle and the vehicle in front exceeds a set threshold, it is determined that there is no vehicle in front.

7. The hybrid traffic interaction behavior representation method based on safety field theory according to claim 5, characterized in that, After determining the interaction scenario, the following is also included: The vehicle types are combined with the eight types of interaction scenarios for analysis, wherein the vehicle types include cars, trucks and buses; For each type of vehicle and its interaction behavior in each interaction scenario, the merging speed, merging distance, lateral speed, lateral acceleration, and the potential field index are extracted and analyzed.

8. The method for representing hybrid traffic interaction behavior based on safety field theory according to claim 1, characterized in that, The integrated safety field model also includes a road boundary field; the potential field intensity generated by the road boundary field at the target vehicle j. for: For road boundary indexing, Corresponding to the left boundary, Corresponding to the right boundary; For the first The constraint strength coefficient of the road boundary; For the target vehicle To the The distance vector of the road boundary; This corresponds to the Euclidean distance modulus.

9. The method for representing hybrid traffic interaction behavior based on safety field theory according to claim 1, characterized in that, The acquisition of trajectory data for each vehicle in the mixed traffic flow within the target area includes: High-precision vehicle trajectory data, including merging and / or diverging zones on highways, is acquired by drone photography. The mixed traffic flow includes manually driven vehicles (HDVs) and connected autonomous vehicles (CAVs).

10. The method for representing hybrid traffic interaction behavior based on safety field theory according to claim 1, characterized in that, The method further includes: Based on the potential field indicators, target vehicles and corresponding interaction scenarios with safety risks exceeding a preset threshold are identified; Based on the identification results, differentiated safety warnings or driving decision support information are generated.

Images