Traffic flow operation safety risk assessment method

Abstract

The present invention discloses a method for evaluating the safety risk of traffic flow operation, comprising: obtaining the traffic density and traffic speed of the traffic flow; obtaining and obtaining a traffic safety operation index based on the vehicle distance and relative change information of different vehicles in the traffic flow; judging the traffic safety operation index, and obtaining a safety risk value based on the judgment result; fitting the relationship between the traffic density and traffic speed and the safety risk value to obtain a traffic flow risk prediction model, and identifying the traffic density and traffic speed obtained in real time through the traffic flow risk prediction model to obtain a traffic flow risk value monitored in real time.

Description

Technical Field

[0001] The present invention belongs to the technical field of traffic safety risk assessment, and in particular relates to a traffic flow operation safety risk assessment method. Background Art

[0002] Analyzing accident or collision data and identifying and addressing road infrastructure safety deficiencies are key measures for traffic accident prevention. As accident prevention shifts toward pre-emptive measures, real-time assessment and mitigation of collision risks in traffic flow are crucial. Traditional models that rely on accident data analysis struggle to meet proactive prevention needs due to data lags and missing details. Existing research often relies on event-based approaches, ignoring significant information about safety interactions and hindering a comprehensive assessment of traffic flow safety. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a traffic flow operation safety risk assessment method to solve the problems existing in the above prior art.

[0004] To achieve the above objectives, the present invention provides a traffic flow operation safety risk assessment method, comprising:

[0005] Obtain traffic density and traffic speed of traffic flow;

[0006] Obtain and obtain traffic safety operation indicators based on the vehicle distance and relative change information of different vehicles in the traffic flow;

[0007] Judging the traffic safety operation index, and calculating based on the judgment result to obtain a safety risk value;

[0008] The relationship between traffic density, traffic speed and safety risk value is fitted to obtain a traffic flow risk prediction model. The traffic density and traffic speed obtained in real time are identified through the traffic flow risk prediction model to obtain the final real-time monitored traffic flow risk value.

[0009] Optionally, the process of obtaining the traffic safety operation index includes:

[0010] Distance interaction information is obtained according to the vehicle distance; relative interaction information is obtained according to the relative change information; and a traffic safety operation index is calculated according to the distance interaction information and the relative interaction information.

[0011] Optionally, the process of acquiring the distance mutual information CRF includes:

[0012]

[0013] Where n represents the number of vehicles in the impact area, and The target vehicle and The distance between the vehicles in the lateral and longitudinal directions, and is the lateral and longitudinal stopping distance, , represents speed, g represents acceleration due to gravity, and f represents ground friction.

[0014] Optionally, the process of acquiring the relative interaction information DBF(t) includes:

[0015]

[0016] in, The target vehicle and Car at time The relative distance change, The target vehicle and Car at time The relative speed change, The target vehicle and The current distance between vehicles; It is The speed of the car.

[0017] Optionally, the traffic safety operation index is obtained by calculating the product of the distance interaction information and the relative interaction information.

[0018] Optionally, the process of judging the traffic safety operation index includes:

[0019] According to the upper threshold of the maximum safe operation domain, the traffic safety operation index is judged.

[0020] The upper limit threshold of the maximum safe operating range for:

[0021]

[0022] in, is the maximum safe operating area distance, is the tolerance probability.

[0023] Optionally, the process of obtaining the security risk value TSC(x,t) includes:

[0024]

[0025] in, is the time step, which indicates the resolution of trajectory data. i is the total number of time steps that the i-th vehicle travels in the space-time window, Conf ij is the conflict discrimination index, which is used to determine whether the i-th vehicle is in a conflict state at time j;

[0026] in, .

[0027] Optionally, the traffic flow risk prediction model adopts a deep learning model, wherein: the traffic flow risk prediction model is used to characterize the fitting relationship between traffic density and traffic speed and the safety risk value:

[0028]

[0029] Among them, TSC represents the security risk value, represents the traffic flow risk prediction model, represents traffic density, and V represents traffic speed.

[0030] Optionally, the traffic flow risk prediction model adopts a support vector machine, a random forest regression model or a gradient boosting model.

[0031] Optionally, the process of obtaining the traffic density ρ and traffic speed V of the traffic flow includes:

[0032]

[0033]

[0034] in, and represents the spatial and temporal size of the window, n is the number of vehicles in the window, and tt i and Dt i are the driving time and distance of vehicle i respectively.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] This paper constructs an evaluation framework based on information theory and control theory, proposing a hypothesis for the safety limit of traffic flow. Furthermore, a distributed stochastic predictive control model for single-lane non-free flow conditions is constructed, and the mapping relationship between macroscopic traffic flow characteristics and safety risks is derived. To quantify microscopic interactive risks, the "Safe Operational Domain" (SOD) metric is proposed, which constructs the safety risk value as a function of density and speed. A machine learning-based macroscopic traffic flow risk prediction model is established, enabling effective pre-emptive risk analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0038] Figure 1A schematic diagram of the impact area of ​​a target vehicle and vehicles within the impact area according to an embodiment of the present invention;

[0039] Figure 2 A schematic diagram of a target vehicle and vehicles within its area according to an embodiment of the present invention;

[0040] Figure 3 Schematic diagram of the verification result of SOD according to an embodiment of the present invention;

[0041] Figure 4 Schematic diagram of the trend line of the number of accidents changing with speed when the traffic flow density is 30 veh / h according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0043] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0044] In response to the problem in the existing technology that it is difficult to comprehensively evaluate the safety status of traffic flow based on "safety interaction" information, the present invention proposes a traffic safety operation domain (SOD) indicator to quantify micro-interaction risks, and uses this indicator to identify whether traffic conflicts occur. When traffic conflicts occur, the conflict time is used as an indicator to measure the micro-traffic safety risk, and the safety risk value is constructed as a function of density and speed. A macro-traffic flow risk prediction model based on machine learning is established to effectively respond to traffic flow risk prediction scenarios. The effectiveness of the SOD indicator and the accuracy of the prediction model are learned through relevant cases.

[0045] Regarding the relationship between traffic safety risk and interaction. From a physical perspective, traffic safety risk can be defined at the micro level as the product of the probability of spatial and temporal overlap in an interaction and the consequences. Without interaction, there would be no traffic accidents. Therefore, interactions involving different environments, different vehicles, and different people will result in significant differences in traffic safety risks, that is, different types of interactions. Ideally, every interaction occurring in the traffic flow can be captured, and by integrating the risks of all interactions, the number of traffic accidents and their consequences on a second-by-second scale can be predicted. However, current technical means make it difficult to capture the massive heterogeneous interactions occurring in a vast space and time, making it impossible to use ideal methods to analyze, evaluate, and predict safety risks.

[0046] Considering that interactions are a source of traffic accidents, if traffic accident data is used as the basis for traffic safety assessment, the number of traffic accidents per annual mileage can be used for road segment evaluation, known as the annual traffic accident rate; for intersection evaluation, the number of traffic accidents per annual traffic volume in the weaving direction can be used, known as the annual weaving accident rate. From the perspective of the physical process of accident occurrence, micro-risk arises from the uncertainty of human perception of surrounding information and driving decisions, which is positively correlated with driving speed and negatively correlated with the ability to gather environmental information and make decisions. The risk of a specific traffic flow segment arises from the instability of longitudinal and lateral interactions, which is negatively correlated with vehicle spacing and the level of environmental information supply, and positively correlated with traffic flow speed.

[0047] The present invention provides a traffic flow safety structured analysis framework for analysis, providing a theoretical analysis basis for subsequent safety risk prediction.

[0048] "Interaction" is the core of traffic safety analysis. From a morphological perspective, interactions can be divided into car-following interaction, lane-changing interaction, and crossing interaction. The physical process of car-following interaction is the simplest and easily quantified and modeled. This invention primarily studies car-following, but similar logic exists for other types of interactions. Considering the randomness of the driving controller model parameters, and targeting the dynamic interactions within the platoon formed under single-lane non-free flow conditions, each vehicle optimizes its local objective function based solely on its own situation and the status of the preceding vehicle, establishing a Distributed Stochastic Model Predictive Control (DSMPC) model for traffic flow. Further analysis from the perspectives of stability, randomness, and safety demonstrates that under a specific traffic density, there is a safe upper limit to the average speed of traffic flow, and below this upper limit, the accident rate is low.

[0049] This embodiment discloses a traffic flow operation safety risk assessment method, including:

[0050] Obtain traffic density and traffic speed of traffic flow;

[0051] Obtain and obtain traffic safety operation indicators based on the vehicle distance and relative change information of different vehicles in the traffic flow;

[0052] Judge the traffic safety operation indicators and calculate the safety risk value based on the judgment results;

[0053] The relationship between traffic density, traffic speed and safety risk value is fitted to obtain a traffic flow risk prediction model. The traffic density and traffic speed obtained in real time are identified through the traffic flow risk prediction model to obtain the real-time monitored traffic flow risk value.

[0054] As an implementation of this embodiment, the process of obtaining the traffic safety operation index includes:

[0055] Distance interaction information is obtained according to the vehicle distance; relative interaction information is obtained according to the relative change information; and a traffic safety operation index is calculated according to the distance interaction information and the relative interaction information.

[0056] The theoretical analysis of this embodiment specifically includes:

[0057] First, the distributed stochastic predictive control model is modeled as follows:

[0058] A distributed stochastic predictive control model is established that can simulate human driving characteristics and their randomness, describing the vehicle kinematics, driver perception and reaction, predictive control strategy and its randomness and uncertainty in the convoy. The details are as follows:

[0059] (1) Vehicle kinematic modeling:

[0060] (1)

[0061] in, represents the position of vehicle i at time k, represents the speed of vehicle i at time k, represents the acceleration of vehicle i at time k, Indicates the sampling time interval.

[0062] (2) Driver perception and reaction modeling:

[0063] Considering the reaction time of human drivers, a time delay is introduced , the acceleration is expressed as:

[0064] (2)

[0065] in, Indicates the relationship between vehicle i and the preceding vehicle The distance between the two, f( ) represents the driver’s decision function, represents the reaction time of driver i, assuming it follows a normal distribution .

[0066] (3) Predictive control strategy:

[0067] Assuming that each driver is constantly predicting the behavior of the vehicle ahead and adjusting the target vehicle acceleration based on these predictions, the predictive control strategy can be expressed as:

[0068] (3)

[0069] Among them, N represents the prediction time domain, Indicates the expected vehicle distance, represents the expected speed, Indicates the target weight coefficient of vehicle distance, speed, and acceleration, , represents the acceleration constraint, represents the speed constraint, Represents the vehicle distance constraint.

[0070] (4) Randomness and uncertainty:

[0071] In order to simulate the randomness of human driving behavior, multiple random variables are introduced, including reaction time, state estimation error , prediction error and control error In addition, a random parameter is introduced To represent the driver's personality characteristics (such as aggressive or conservative), this parameter can affect the weight coefficient of choice, where It can be assumed to follow a standard normal distribution N(0, 1).

[0072] Safety analysis based on distributed stochastic predictive control model:

[0073] Based on the established distributed stochastic predictive control model, stability, randomness and safety analysis are carried out to further demonstrate the existence of a safe upper limit for the average speed of traffic flow.

[0074] (1) Stability analysis:

[0075] Analyze the stability of the system. In an ideal situation (no randomness and uncertainty), the stability of the system mainly depends on the parameter selection of the predictive control strategy. The stability condition is determined by analyzing the characteristic equation of the system. Consider a simplified linear model:

[0076] (4)

[0077] in and are the proportional gain and the speed difference gain, then the characteristic equation of the system is:

[0078] (5)

[0079] According to the Routh-Hurwitz stability judgment conditions: , then the sampling time interval When the speed is less than a given value, the system converges to a stable state, and this value is related to the speed difference gain. This indicates that when the human (driver) sampling time is long (such as distracted driving), system stability deteriorates, traffic flow tends to become unstable, or collisions occur. Higher speeds require a higher sampling frequency, and the tolerable driving error decreases.

[0080] (2) Random impact analysis

[0081] Considering randomness and uncertainty, analyze the impact of these factors on system performance. Use random stability theory to analyze the mean square stability of the system:

[0082] Define the state vector , the system can be expressed as:

[0083] (6)

[0084] Among them, A(k) and B(k) are random matrices, is the control input, is a random disturbance. The mean square stability condition of the system can be expressed as:

[0085] (7)

[0086] Where E[] represents the expectation and M is a bounded constant. Through Lyapunov equation analysis, the mean square stability expression of the system is:

[0087] (8)

[0088] Where P is a positive definite matrix and Q is a semi-positive definite matrix. This yields the mean square stability condition for the system. This indicates that in traffic flow, if speed disturbances or position control errors decrease over time, the system tends to a stable state. This means that errors caused by randomness are one of the main factors determining the upper limit of system stability.

[0089] (3) Safety analysis

[0090] Define a prediction-based safety indicator, expressed as:

[0091] (9)

[0092] Among them, N is the prediction time domain, is the minimum safe distance. This safety indicator takes into account the difference between the minimum predicted vehicle spacing and the safe distance in the next N time steps. When , it means that unsafe conditions may occur within the prediction range. Therefore, it is necessary to define the safety probability, which is expressed as:

[0093] (10)

[0094] In the DSMPC framework, this probability is affected by many factors, such as and Distribution characteristics of control strategy parameters choice, the driver's reaction time and personality parameters And the perceived control error. Therefore, the optimization objective of DSMPC is expressed as:

[0095] (11)

[0096] Where E[] represents expectation, is a weight coefficient. By directly considering safety in the control strategy in this way, the controller will balance the pursuit of the desired state and the guarantee of safety when making decisions. is a large value. If the above objective function has a solution, it means that there is a driver prediction control strategy that makes Approaching 0, whether there is a solution depends on the distribution characteristics of the aforementioned random parameters, as well as the expected inter-vehicle distance and expected speed. Therefore, the driver predictive control strategy can be used as a generalized traffic control strategy. Traffic management departments can implement control strategies such as speed limits and inter-vehicle distance warnings to optimize the predictive control strategy. The traffic accident rate can then approach 0, and the expected inter-vehicle distance and speed have a steady-state relationship.

[0097] (4) Analysis of the relationship between average speed and safety based on DSMPC

[0098] Under the DSMPC framework, the relationship between average speed and safety becomes more complicated because they are both functions of control strategies, prediction models, and random factors. Define an average speed based on DSMPC as:

[0099] (12)

[0100] Where M is the number of vehicles in the convoy. At this time, the upper limit of safe speed is Should meet the following requirements:

[0101] (13)

[0102] in, is a small tolerance probability that can approach 0.

[0103] All randomness and uncertainty of the DSMPC system need to be considered, such as the prediction error distribution and , state estimation error distribution and , control error distribution , driver characteristics distribution and , control parameter distribution 、 and Further analysis The relationship between various DSMPC parameters. First, the effect of prediction time domain N. When N is larger, it usually improves security, but also increases computational complexity (attention consumption). Second, the control parameters The impact of different parameter combinations will lead to different Furthermore, the driver's characteristic parameters and Finally, the influence of error distribution parameters (such as variance) also affects the upper limit of safe speed when there is uncertainty in the system. Therefore, a more comprehensive expression:

[0104] (14)

[0105] The above analysis demonstrates that, under a given traffic density, there is indeed a safe upper limit for the average vehicle speed. Below this upper limit, the accident rate is low. This is because the SMPC (driver) can adjust the vehicle's acceleration to maintain stability in local platoon segments and cope with random disturbances. However, when random disturbances exceed a certain threshold, collisions are unavoidable.

[0106] In view of the above theoretical analysis, the traffic flow operation safety evaluation method based on the traffic safety operation domain provided by the present invention is subsequently elaborated in detail:

[0107] The conceptual model of traffic safety analysis provides a goal and path for the quantitative analysis of traffic safety. In response to the above analysis results, the present invention provides a traffic flow operation safety evaluation method based on the traffic safety operation domain. On the basis of the aforementioned analysis of traffic flow safety mechanisms, this paper attempts to use quantitative analysis methods to construct a relationship model between traffic flow speed, density and safety. First, an alternative measurement index for the micro-interaction risk of traffic flow is proposed to achieve the quantitative calculation of risk value; then, based on key traffic flow parameters, a machine learning prediction model for traffic flow operation safety risk is established. Furthermore, traffic management departments can use high-risk warning values ​​of traffic flow to actively adjust the traffic flow status through means such as variable speed limits, lane control, vehicle type control, and induction prompts to improve high-risk status and achieve dynamic traffic safety risk management and control.

[0108] Quantification of traffic flow operation safety risks based on micro-interaction:

[0109] This paper addresses the effectiveness and limitations of different conflict indicators in assessing traffic safety. Based on the macroscopic state of traffic flow, this paper proposes the Safety Operational Domain (SOD), a safety assessment metric. This metric quantifies the safe interaction area between vehicles based on macroscopic characteristics of traffic flow, such as traffic density. It also incorporates features such as lane switching and traffic flow instability to achieve a microscopic interactive safety risk measurement for traffic flow. Specifically, based on macroscopic traffic flow trajectory data, the macroscopic state variables traffic density ρ, traffic flow Q, and speed V are calculated as shown in the following formula:

[0110] (15)

[0111] in, and represents the spatial and temporal size of the window, n is the number of vehicles in the window, and tt i and Dt i are the travel time and distance of vehicle i respectively. The definition of Edie’s macroscopic state variables is consistent and automatically satisfies the fluid dynamics relationship:

[0112] (16)

[0113] According to formula (16), the macro-state variable essentially contains the traffic flow efficiency characteristics reflected by all the underlying vehicle-following interactions. These interactions between vehicles can cause traffic conflicts, which are usually micro-characteristics and essentially depend on the behavior patterns of drivers and pedestrians. However, in traffic safety evaluation, it is not feasible to obtain the micro-behaviors of vehicles or pedestrians (such as steering wheel operation, temporary change of walking direction, etc.) in real time. Therefore, safety assessment can be performed based on the macro-traffic flow state (such as traffic density, traffic flow and speed, etc.) reflected by these micro-behaviors. In order to quantify traffic flow safety risks from a macro-aggregate perspective, "time to conflict" (TSC) can be used as an indicator to measure micro-traffic safety risks. TSC refers to the total time experienced by all vehicles when a conflict occurs, which is calculated based on a specific conflict metric. For a specific traffic flow condition, a higher TSC value means that the vehicle stays in the conflict situation for a longer time under this condition and the risk of collision is higher. Considering the vehicle interactions within the spatiotemporal window, TSC can be expressed by the following formula:

[0114] (17)

[0115] in, is the time step (seconds), which indicates the resolution of the trajectory data. i is the total number of time steps that the i-th vehicle travels in the space-time window, Conf ijIt is a conflict discrimination index used to determine whether the i-th vehicle is in a conflict state at time j. x represents the interacting vehicle number, and t represents the corresponding time step.

[0116] The present invention's calculation of TSC is based on two key considerations: First, the TSC calculated based on conflict metrics should maintain a significant correlation with macro-traffic metrics; second, TSC calculated using different conflict metrics should possess clear physical meaning and interpretability. The Time to Collision (TTC) and Deceleration Rate to Avoid Crash (DRAC) metrics are based on the assumption that traffic conflicts are limited to interactions between preceding and following vehicles within the same lane. This assumption limits their applicability in macro-level traffic safety assessments, particularly in effectively reflecting potential traffic safety hazards arising from interactions between vehicles in adjacent lanes. Specifically, when using TTC or DRAC as conflict metrics, they only consider situations where the following vehicle's speed exceeds that of the leading vehicle (i.e., rear-end collision scenarios). Consequently, they only reflect interactions along the collision path and fail to fully capture other potentially dangerous interactions between vehicles at high speeds. In contrast, the TSC indicator based on Potential Stopping Distance (PSD) can more effectively establish the correlation between macro traffic flow variables and traffic safety levels by measuring the time when the stopping distance is less than the relative distance between vehicles.

[0117] However, moderate congestion is often associated with a series of complex traffic phenomena, such as increased speed variability, traffic instability, and hysteresis, which are of great significance in traffic safety assessment. Therefore, to further understand the intrinsic relationship between traffic safety and traffic flow, TSC, calculated based on arbitrary conflict metrics, should be able to effectively account for this phenomenon. For example, if time to collision (TTC) is used as a traffic conflict metric, congestion is considered the most critical influencing factor. Under congested conditions, due to the significant reduction in vehicle spacing, time-based conflict metrics (such as TTC, TIT, and TET) typically identify congestion as the most significant safety risk factor. However, the potential stopping distance (PSD), a distance-based conflict metric, considers congestion to be less important because vehicles have longer stopping distances at low speeds. Moderate congestion, or the transition from non-congested to congested, is the most critical stage for safety. Under moderate congestion conditions, the potential risk of conflict will indeed increase significantly due to the smaller vehicle spacing and more frequent dynamic behaviors of traffic participants. In order to improve the operational stability and safety of dynamic traffic flow, a new operational safety risk assessment method is urgently needed. By comprehensively considering characteristics such as traffic density, speed, lane switching and traffic flow instability, a more comprehensive portrayal of the relationship between macro traffic flow and safety can be achieved.

[0118] (1) Safety analysis of vehicle-to-vehicle interaction

[0119] In homogeneous traffic conditions with high lane-observing standards, driver interactions primarily occur in the longitudinal dimension, such as speed adjustments to follow the vehicle ahead, while interactions in the lateral dimension are relatively rare. Lateral interactions between vehicles typically occur during lane changes or overtaking. This means that the target vehicle's response is primarily influenced by the vehicle immediately preceding it. However, in heterogeneous traffic conditions with no lane discipline, drivers interact not only in the longitudinal dimension but also in the lateral dimension. In this scenario, the target vehicle's response is influenced not only by the lead vehicle but also by other surrounding vehicles, such as multiple lead vehicles, following vehicles, and lateral neighbors, and even by other vehicles within the target vehicle's zone of influence. The target vehicle's "zone of influence" is a hypothetical area within which the surrounding traffic environment influences its response. To further investigate this phenomenon, zones of influence of varying sizes and shapes have been explored and field investigated to obtain reasonable behavioral predictions. This invention employs a sector-shaped zone of influence because the driver's field of vision during driving encompasses the sector and the rearview mirror, and the road boundary can be easily approximated as the dimensions of the two sides of the sector-shaped zone of influence.

[0120] Figure 1The hypothetical influence zone and the vehicles within it at a specific time step are shown. With the target vehicle at the center, the target vehicle's influence zone is divided into five regions: the left adjacent zone, the right adjacent zone, the front zone, the left rear leading zone, and the right rear leading zone. The front zone includes the leading zone, the left leading zone, and the right leading zone. It should be noted that if the target vehicle moves to the left or right, the boundaries of the influence zone will shorten accordingly. Furthermore, the influence zone and the vehicles within it are evaluated at each time step and are therefore subject to the traffic environment at that time step. As the target vehicle moves at each time step, the vehicle composition within the influence zone changes accordingly.

[0121] (2) Construction of traffic safety operation domain indicators based on conflict time

[0122] Considering that the target vehicle interacts with other vehicles in the impact zone simultaneously in both the lateral and longitudinal dimensions, the target vehicle may collide with surrounding vehicles in the lateral or longitudinal direction, or in both directions simultaneously. This chapter models and calculates the conflict time using the newly constructed SOD indicator.

[0123] like Figure 2 As shown in Figure 1, assuming that the target vehicle interacts with its left leading vehicle, the SOD at a certain time step (𝜏) will present different values ​​in the lateral and longitudinal directions based on the longitudinal and lateral velocities of the target vehicle and the relative longitudinal and lateral spacing between the target vehicle and different vehicles in the impact area. Its mathematical expression is:

[0124] (18)

[0125] in = , and The target vehicle and The distance between the vehicles in the lateral and longitudinal directions, = , and Indicates the lateral position and longitudinal position of the vehicle, the subscript i indicates that it corresponds to vehicle i, and the subscript target indicates that it corresponds to the target vehicle. and It is the lateral and longitudinal stopping distance, usually referring to the distance required for the tested vehicle to come to a complete stop to avoid a collision, i.e. , v represents velocity, g represents acceleration due to gravity, and f represents ground friction; In the longitudinal dimension, it is 3.92 m / s², and in the lateral dimension, it is 1.47 m / s². These values ​​represent the deceleration required for the tested vehicle to avoid a collision. n is the number of vehicles in the impact zone. In congested conditions, the potential risk of conflict increases significantly due to the close spacing between vehicles. The volatility and instability of local traffic flow are crucial to the risk of traffic conflict. Therefore, the SOD is optimized based on traffic flow density characteristics.

[0126] When traffic density suddenly increases, the volatility and instability of local traffic flow, such as sudden acceleration of the target vehicle, can increase the risk of traffic conflicts. The risk of conflicts caused by the sudden and unexpected behavior of the target vehicle is even higher when traffic density is high. Considering that it is impossible to obtain the target vehicle's speed changes in real time, but it can be represented by relative speed changes and relative distance changes, the present invention measures this instability by calculating the relative speed changes and relative distance changes between the target vehicle and surrounding vehicles, as shown below:

[0127] (19)

[0128] in, The target vehicle and Car at time The relative distance change, The target vehicle and Car at time relative speed changes. The target vehicle and The current distance between vehicles. It is The speed of the car.

[0129] In summary, in order to more comprehensively reflect the impact of the dynamic behavior of target vehicles on the collision risk in a high-density environment, the instability characteristics caused by the sudden microscopic behavior of the target vehicle are introduced. The SOD is expressed as:

[0130] (20)

[0131] A high SOD value indicates that the target vehicle's driving safety is low in the current traffic environment, with a high risk of conflict. This may be due to increased instability caused by high traffic density or increased uncertainty caused by failure to maintain a safe braking distance. A low SOD value indicates that the target vehicle's driving environment is relatively stable, with the surrounding traffic environment having a minimal impact on safety, and a low risk of conflict. This paper introduces the Safety Operational Domain (SOD) evaluation metric to more comprehensively assess traffic safety, especially in complex traffic environments, and improve the accuracy and effectiveness of traffic safety management.

[0132] According to the theoretical analysis of the previous traffic flow safety structured analysis framework, there is a safety threshold upper limit. When the SOD value exceeds the safety upper limit, it means that the driving environment of the target vehicle is very close to a conflict or has already occurred. The SOD upper limit is closely related to the stability and safety of the system. Taking into account random disturbances and uncertainties, it can be concluded that the upper limit condition of SOD is and , respectively represent the expected security and expected stability of the system. Combined with the relationship between speed and security in the DSMPC framework, the safe operation domain is defined The upper limit is:

[0133] (twenty one)

[0134] in, is the maximum safe operating area distance, It is a small tolerance probability that can approach 0, representing an extremely low probability of conflict.

[0135] use To identify whether a vehicle conflict occurs. If the SOD is greater than or equal to the target vehicle in the interaction with its surrounding vehicles , then the interacting vehicle and the target vehicle are marked as in conflict. The conflict calculation is:

[0136] (twenty two)

[0137] Traffic conflicts are identified at each time step for all vehicles in the traffic flow. Once traffic conflicts are identified in the traffic flow, the TSC value is calculated using Equation (17).

[0138] Operational safety risk prediction model based on macroscopic traffic flow status:

[0139] According to the established traffic flow operation safety risk quantitative calculation model, the safety risk value of a certain section of traffic flow can be calculated in real time. However, it is difficult to obtain a full amount of vehicle motion trajectory data in traffic management practice, and it cannot be used for actual traffic flow safety management. According to the theoretical analysis of the traffic flow safety structured analysis framework, traffic flow density and speed are key factors affecting the safety status of traffic flow. They are also common traffic flow parameters that are relatively easy to obtain in traffic management scenarios. Therefore, the present invention attempts to establish a prediction model for traffic flow safety risk value based on macro traffic flow parameters. The machine learning model is used to directly predict TSC from the macro traffic flow state parameters. Its expression is as follows:

[0140] (twenty three)

[0141] The machine learning model should not only have high prediction accuracy but also good generalization performance. Therefore, the selected machine learning model is evaluated from the following aspects:

[0142] a. Goodness of Fit (GoF): How well the model's predicted values ​​map to the actual values. For this metric, use R-squared and a scatter plot between TSC and density.

[0143] b. Measures of Performance (MoP): The accuracy of model predictions is mainly evaluated using different error metrics such as root mean square error (RMSE), mean absolute error (MAE), and mean square error (MSE).

[0144] c. Model reliability: The reliability of a model is mainly evaluated by the error distribution of the model.

[0145] d. Meaningful predictability: The significance of the model prediction results mainly comes from the non-negative nature of TSC. If the predicted SOD value is negative, the model is considered to have poor suitability.

[0146] Based on a preliminary analysis of the problem, this paper selected three machine learning algorithms that have been widely validated in traffic safety research: support vector machines (SVM), random forest regression (RF), and gradient boosting (GB). By optimizing the hyperparameters of different machine learning models, K-fold cross-validation and grid search techniques were used to determine the optimal hyperparameter values ​​for each model. Five-fold cross-validation and grid search were also used for hyperparameter tuning.

[0147] (1) Support Vector Machine (SVM) algorithm

[0148] The basic idea of ​​SVM is to convert the regression problem into a classification problem that can be handled by the support vector machine method. The core goal is to find a hyperplane to maximize the geometric interval between two types of data points, so the optimization objective is to minimize the norm of the weight vector:

[0149] (twenty four)

[0150] The constraints are: .in is the category label, and Define the hyperplane. For nonlinearly separable or noisy data, SVM introduces slack variables and regularization parameter C, the data expression of the optimization objective is:

[0151] (25)

[0152] The constraints are: . Therefore, the corresponding loss function is:

[0153] (26)

[0154] The solution to this optimization problem can be expressed by the Lagrange multiplier method as:

[0155] (27)

[0156] in, The present invention adopts the radial basis kernel function (RBF) as the Lagrange multiplier, which is expressed as follows:

[0157] (28)

[0158] in, is a hyperparameter that determines the influence of a single sample on the training process.

[0159] (2) Random Forest (RF) algorithm

[0160] Random forest (RF) is a tree-based ensemble learning method that improves the final prediction accuracy by combining multiple weak learners. In the random forest (RF) method, each tree is independently constructed through bootstrap sampling, which is called the bagging method. In this process, each data point has an equal probability of being selected from different random subsets. Compared with a single decision tree, random forest has stronger predictive ability, mainly due to its improved node splitting strategy. In a standard decision tree, all features are considered when splitting a node, while in a random forest, each node is split only based on a randomly selected subset of features. The RF algorithm has many hyperparameters, and the key hyperparameters that usually affect model performance include the number of trees ( ), the minimum number of samples required for each leaf node and the maximum depth of the tree ( ).

[0161] (3) Gradient Boosting (GB) Algorithm

[0162] The gradient boosting (GB) algorithm is also a tree-based ensemble learning method. It uses a step-by-step iterative approach to build decision trees, with the results of each tree used to improve the prediction performance of subsequent trees. The optimization goal of the GB algorithm is to minimize the prediction error. Its objective function is expressed as follows:

[0163] (29)

[0164] in, Representative A decision tree, is the splitting variable, The weight of the tree determines the degree of influence of the tree on the final prediction result. The core of GB is to optimize To minimize the loss function between the predicted value and the true value. The training process can be summarized into three steps: First, random initialization ; Then, build each decision tree in turn, calculate the residual of each sample, calculate the gradient based on the loss function, and use the residual to fit the new decision tree estimate , update the model prediction value Finally, the optimal model performance is achieved. In the GB algorithm, the key hyperparameters that affect model performance include the number of trees and the learning rate.

[0165] In view of the above technical solution, the present invention conducts traffic flow operation safety simulation experiments and results analysis based on relevant data examples:

[0166] Dataset:

[0167] (1) Traffic conditions: The road scenario is set as a two-way four-lane urban expressway, with each lane 3.5 meters wide and a road length of 5 km. At the same time, the expected speed is set to conform to the Gaussian distribution. Due to the limitations of the vehicle dynamics engine version, the models deployed are mainly cars.

[0168] (2) Experimental data collection: Under different working conditions, as shown in Table 1, Monte Carlo simulation experiments were conducted after setting different working condition parameters and scenarios based on the microscopic traffic safety simulation platform to generate traffic flow and vehicle motion trajectory data under each working condition.

[0169] Table 1

[0170]

[0171] The technical solution provided by this invention uses simulated vehicle trajectory data to simultaneously assess traffic safety risks from both microscopic and macroscopic perspectives. Macroscopic traffic flow characteristic parameters, namely traffic volume, speed, and density, are extracted at 5-minute aggregation intervals. Vehicle trajectories are extracted at a resolution of 0.5 seconds, covering a 200-meter road section for a total of 60 minutes.

[0172] Trajectory data preprocessing:

[0173] In the processing of vehicle motion trajectory data, local errors may occur during the calculation process, resulting in drastic fluctuations in instantaneous acceleration, deceleration or speed values, which in turn affects the accuracy of the data. Therefore, the present invention needs to smooth the trajectory data to reduce noise. In order to reduce the noise in the data set, the seven-point moving average method is used to smooth the vehicle motion trajectory extracted by simulation to reduce fluctuations caused by errors in the calculation process. This method smoothes the data by taking the average value of each point in the trajectory data and its six adjacent points, thereby effectively reducing instantaneous fluctuations and improving the stability and reliability of the trajectory data. Specifically, for a given trajectory point , its smoothed value Calculated by the following formula:

[0174] (30)

[0175] in, This smoothing process can effectively remove high-frequency noise from the data, maintain the overall trend of the vehicle trajectory, and improve the accuracy and reliability of subsequent analysis.

[0176] Traffic flow characteristics analysis based on traffic safety operation domain:

[0177] Based on the established traffic flow operation safety risk quantification method, a brief analysis of traffic flow safety characteristics is conducted through simulation experimental data. First, the limit of the traffic safety operation domain is determined. The traffic conflict time TSC is calculated based on the traffic safety operation domain. The results are as follows: Figure 3 As shown in Figure 2, when CRF and DBF are less than 1.09, TSC is relatively flat. When CRF and DBF continue to increase, TSC rises sharply. The upper limit is set at 1.188. When the SOD is higher than 1.188, it is considered that the risk of an accident increases sharply.

[0178] Calculations and statistical analysis of the simulated trajectory data indicate that safety risks are primarily concentrated in medium-density areas. As density increases, speed gradually decreases, while low speeds in high-density areas are associated with reduced risk. Notably, high-risk areas are primarily located at the intersection of medium density and medium-to-high speeds, potentially reflecting the increased risk of accidents due to frequent interactions between vehicles, such as overtaking and lane changes.

[0179] In addition, taking the traffic flow density of 30veh / h as an example, the data related to the number of collision accidents and speed are intercepted. The trend line of the change of the number of collision accidents with speed is as follows: Figure 4When the average traffic speed was between 18 km / h and 30 km / h, the number of accidents increased slightly. Analysis of vehicle trajectory data revealed that a small number of vehicles chose to overtake or change lanes when the vehicle ahead was slower, resulting in collisions with slower vehicles and causing the increase in accidents.

[0180] It can be inferred that under certain traffic flow density conditions, there is a safe range of traffic flow speed with a low accident rate, within which efficiency and safety are balanced.

[0181] Comparative analysis of the effectiveness of traffic safety operation domains:

[0182] This section explores the relationship between TSC results based on different conflict metrics such as TET, TIT, and SOD and macro traffic flow state parameters, and compares the relationship between different TSC values ​​and macro state parameters in the dataset. Different TSCs show different relationship patterns with macro state parameters.

[0183] The TSC values ​​derived from TTC-based metrics show a significant correlation with density, with varying ranges of TSC values, primarily significant in high-density situations. This experimental conclusion can be explained by two mechanisms: First, in high-density situations, vehicle spacing decreases, making TTC-based conflicts more likely with smaller speed differences. Second, as density increases, the total travel time of all vehicles within each spatiotemporal window also increases, leading to the accumulation of more instances of such conflicts and, ultimately, an increase in the TTC-based TSC value. Furthermore, in highly congested conditions, TSC values ​​derived from TTC-based metrics may be unstable because vehicle speeds are low and their SOD values ​​are above a critical threshold, meaning that vehicles are more likely to stop safely before a collision. Even in highly congested conditions, the SOD-based TSC values ​​are lower, and even in these conditions, the total travel time of all vehicles within each spatiotemporal window increases due to increased density.

[0184] As an event-based conflict indicator, TTC-based metrics have certain limitations in their scope of application. This leads to significant deficiencies in TSC based on TTC in the following two aspects: First, the incidence, frequency, and severity of conflict events are largely influenced by drivers' car-following behavior, and driving behavior characteristics often have significant regional differences. Second, TSC-based TTC assessments only focus on a single car-following process within a specific car-following scenario and cannot fully reflect the safety status of other driving periods within each spatiotemporal window. Therefore, although TTC-based metrics perform well in assessing single-vehicle collision risk at the micro level, they have limitations in evaluating traffic flow safety risks.

[0185] When analyzing the relationship between the SOD-based TSC and macro-state variables, the present inventors found that the SOD-TSC correlation was more significant and substantively meaningful than the correlations between other event-based TSC and macro-state variables. Experimental results indicate that moderate congestion, due to its highest TSC value, is identified as a key high-risk state. This phenomenon is closely related to a variety of complex traffic phenomena, such as increased speed fluctuations, traffic flow instability, and stop-and-go traffic wave propagation, which are widely considered to be the main factors leading to rear-end collisions. Furthermore, multiple studies have further confirmed that moderate congestion is one of the traffic conditions with the highest risk of rear-end collisions. Therefore, the SOD-based TSC, when reflecting real moderate congestion traffic conditions, has higher validity and authenticity than traditional TTC-based alternative metrics. SOD is essentially a non-event-driven metric, and its input parameters (including vehicle speed and distance) are significantly correlated with macro-traffic flow parameters (such as average speed and density). This characteristic enables the relationship model between the TSC value and macro traffic flow parameters based on SOD to maintain good generalization performance in other different traffic scenarios with similar flow-density curve characteristics.

[0186] In summary, from the perspective of practical application of quantitative assessment, compared with TTC-type alternative evaluation indicators, the traffic safety status identification and classification method based on SOD has outstanding advantages: first, its quantitative characterization results can accurately reflect the essential characteristics of traffic flow; second, this indicator has better applicability and effectiveness in actual traffic control scenarios.

[0187] Performance analysis of traffic flow operation safety risk prediction model:

[0188] This paper establishes a data-driven machine learning model and evaluates its predictive capabilities. As previously mentioned, three machine learning algorithms are used to model TSC as a function of macroscopic traffic flow parameters (density and speed): random forest (RF), support vector machine (SVM), and gradient boosting (GB). These machine learning algorithms are implemented in Python. The hyperparameters of each machine learning model are adjusted or optimized using k-fold cross-validation and grid search techniques. The optimal hyperparameter values ​​are shown in Table 2.

[0189] Table 2

[0190]

[0191] An analysis of the error distribution of the prediction results shows that the error distribution of the RF model approximately follows a normal distribution. This is mainly due to its unique feature randomness mechanism. During the model construction process, each decision tree is trained using only part of the features, thereby enhancing the diversity of the model and improving its robustness to noisy data and redundant features.

[0192] Table 3 shows the results of the machine learning models on the training and test datasets. The experimental results in Table 3 demonstrate that all machine learning models exhibit consistent performance on both the training and test sets. Comparative analysis reveals that among the three machine learning models—random forest (RF), support vector machine (SVM), and gradient boosting (GB)—the RF model achieves the lowest values ​​in evaluation metrics such as root mean square error, mean absolute error, and mean square error, indicating slightly superior predictive performance to the other models.

[0193] Table 3

[0194]

[0195] Among them, RMSE stands for root mean square error; MAE stands for mean absolute error; MSE stands for mean square error.

[0196] In order to deeply evaluate the performance differences of different machine learning models, the present invention verifies the correlation between the TSC values ​​predicted by the comparative analysis model and the actual observed values. Specifically, based on the distribution characteristics of the TSC-density relationship and the TSC-speed relationship, the relationship between the observed values ​​and the predicted values ​​is visualized and analyzed. The results show that under given traffic flow density and speed conditions, the TSC values ​​predicted by the model are highly consistent with the actual observed values. It is worth noting that although all the machine learning models constructed by the present invention show good consistency, the SVM and GB models predict negative TSC values ​​in some cases, which is neither in line with physical meaning nor has practical application value. In contrast, the RF model did not have negative values ​​in all predictions. This result further confirms the effectiveness and reliability of the RF model in predictive performance.

[0197] Traffic flow operation safety risk assessment and intervention empirical analysis:

[0198] Empirical analysis of traffic safety operation domain:

[0199] The macroscopic traffic flow evaluation indicators SOD and TSC are calculated based on the MFTD dataset. When dividing the time-space window, the appropriate time size can capture the dynamic instability in the traffic flow, and the appropriate space size can ensure that the window is large enough when calculating traffic variables to avoid noise. Therefore, the time-space plane is discretized into multiple sizes of and In the traffic scenario of Huaxiang Bridge on the South Fourth Ring Road of Beijing, the macro traffic flow safety risk evaluation index TSC of the spatiotemporal window is calculated according to the method proposed in this chapter.

[0200] TSC varies with traffic density and speed, with higher TSC values ​​indicating higher risk. For traffic densities less than 300 veh / km, traffic is in a free-flow or metastable state, with high vehicle speeds and high dispersion. TSC increases with increasing density. This is due to drivers having insufficient time to react to unexpected events (such as sudden braking of the preceding vehicle). Despite a large absolute distance, TSC increases with density, reflecting speed-dominated risk. In moderately congested traffic (the transition from non-congested to congested), speed differences increase significantly, triggering frequent acceleration and deceleration. Because moderately congested traffic is associated with more chaotic traffic conditions, collision risk increases. When traffic density exceeds 500 veh / km, traffic enters a fully congested state, with vehicles operating at very low speeds, TSC values ​​decrease, reflecting a risk mitigation phenomenon under the density saturation effect. This is consistent with the consensus of relevant research and indirectly validates the validity of the SOD indicator.

[0201] Feasibility analysis of risk intervention measures:

[0202] Traffic safety control is a key responsibility of traffic management departments. Taking intervention measures based on dynamic risk assessment and early warning is a viable means of preventing traffic accidents. For example, in medium-density traffic conditions, a real-time traffic risk classification can be released to the public. Using red, yellow, and green to indicate high, medium, and low risk, drivers can be prompted to pay attention and adjust their driving behavior. When traffic flow becomes unstable, dynamic gradient speed limit strategies can be implemented to improve traffic flow. For example, the speed limit could be gradually reduced from 80 km / h to 60 km / h and then to 40 km / h.

[0203] To evaluate the feasibility and effectiveness of risk intervention measures, a baseline scenario replicated the aforementioned real-world traffic conditions, calculating baseline SOD and TSC values. High-risk areas accounted for 12.3% of the total area, with an average TSC value of 22.52. An intervention scenario simulated a gradient speed limit policy, comparing the proportion of high-risk areas and average TSC values, and calculating a pre- and post-intervention comparison. The experimental results, shown in Table 4, show that after the intervention, the proportion of high-risk areas decreased to 8.5%, and the average TSC value in high-risk areas decreased to 20.41.

[0204] Table 4

[0205]

[0206] The above results show that risk intervention measures can effectively reduce the proportion of high-risk areas, improve the stability of traffic flow, and reduce the risk of traffic accidents. The experimental results show that traffic safety management based on traffic flow operation safety risk assessment results has broad application prospects.

[0207] Based on this, this paper proposes a traffic safety analysis method that integrates macroscopic characteristics. First, a safety evaluation theoretical framework that integrates information theory and cybernetics is constructed based on the mechanism of traffic flow instability. A traffic safety limit hypothesis is proposed, and a single-lane non-free flow distributed stochastic predictive control model is established to reveal the intrinsic relationship between macroscopic traffic flow parameters and safety risks. An innovative quantitative indicator, "Traffic Safety Operational Domain" (SOD), is proposed to quantitatively characterize microscopic interactive risks. A macroscopic risk prediction model is constructed by combining machine learning. The optimal prediction performance of the random forest model is verified by comparing the SVM, RF, and GB algorithms. Experiments with multiple density-speed combinations are conducted on a microscopic traffic safety simulation platform to verify the rationality and effectiveness of the SOD indicator and prediction model. Empirical studies demonstrate that the proposed evaluation method can effectively assess traffic flow safety trends. Its application value in dynamic traffic flow safety management and control is verified through a comparison of intervention measures. This provides theoretical methods and technical tools to support traffic management departments in the "pre-emptive" prevention of traffic accidents.

[0208] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. Traffic flow operation safety risk assessment method, characterized by: include: Obtain traffic density and traffic speed of traffic flow; Obtain and obtain traffic safety operation indicators based on the vehicle distance and relative change information of different vehicles in the traffic flow; Judging the traffic safety operation index, and calculating based on the judgment result to obtain a safety risk value; The relationship between traffic density, traffic speed and safety risk value is fitted to obtain a traffic flow risk prediction model. The traffic density and traffic speed obtained in real time are identified by the traffic flow risk prediction model to obtain the real-time monitored traffic flow risk value. The process of obtaining the traffic safety operation index includes: Obtaining distance interaction information based on the vehicle distance; obtaining relative interaction information based on the relative change information; and calculating a traffic safety operation index based on the distance interaction information and the relative interaction information; The process of obtaining the distance mutual information CRF includes: Where n represents the number of vehicles in the impact area, and The target vehicle and The distance between the vehicles in the lateral and longitudinal directions, and is the lateral and longitudinal stopping distance, the stopping distance , v represents velocity, g represents acceleration due to gravity, and f represents ground friction; The process of obtaining the relative interaction information DBF(t) includes: in, The target vehicle and Car at time The relative distance change, The target vehicle and Car at time The relative speed change, The target vehicle and The current distance between vehicles; It is The speed of the vehicle; By calculating the product of distance interaction information and relative interaction information, the traffic safety operation index is obtained; The process of judging the traffic safety operation index includes: According to the upper threshold of the maximum safe operation domain, the traffic safety operation index is judged. The upper limit threshold of the maximum safe operating range for: in, is the maximum safe operating area distance, is the tolerance probability; The process of obtaining the security risk value TSC(x,t) includes: in, is the time step, which indicates the resolution of trajectory data; nT i is the total number of time steps that the i-th vehicle travels in the space-time window, Conf ij is the conflict discrimination index, which is used to determine whether the i-th vehicle is in a conflict state at time j; in, .

2. The method according to claim 1, characterized in that The traffic flow risk prediction model adopts a deep learning model, wherein: the traffic flow risk prediction model is used to characterize the fitting relationship between traffic density and traffic speed and safety risk value: Among them, TSC represents the security risk value, represents the traffic flow risk prediction model, represents traffic density, and V represents traffic speed.

3. The method according to claim 2, characterized in that The traffic flow risk prediction model adopts a support vector machine, a random forest regression model or a gradient boosting model.

4. The method according to claim 1, wherein The process of obtaining the traffic density ρ and traffic speed V of the traffic flow includes: in, and represents the spatial and temporal size of the window, n is the number of vehicles in the window, and tt i and Dt i are the driving time and distance of vehicle i respectively.

Images