Automatic driving-oriented highway unsignalized intersection suitability evaluation method

Abstract

The application provides a highway unsignalized intersection suitability evaluation method for automatic driving, comprising the following steps: based on the driving automation level of a target vehicle, matching the perception, braking and takeover characteristic parameters of the corresponding level, and differentiating to calculate the required stopping sight distance of the corresponding level; obtaining the effective detection distance of the target scene corresponding to the automatic driving, combining the intersection geometric parameters to obtain the available sight distance; obtaining the static suitability index through the sight distance function function; for the automatic driving vehicle-dynamic interactive vehicle collision scene, calculating the time dimension conflict index post-intrusion time PET and the space dimension redundancy index stopping distance ratio PSD of the conflict scene; taking the comparison result of PET and the PET safety threshold as the first judgment level, and taking the stopping distance ratio PSD as the second review judgment level to obtain the dynamic suitability index. According to the static suitability index and the dynamic suitability index, the suitability evaluation is realized.

Description

Technical Field

[0001] This invention belongs to the field of road traffic control technology, specifically relating to a method for assessing the drivability of unsignalized intersections on highways for autonomous driving. Background Technology

[0002] With the rapid iteration and large-scale application of autonomous driving technology, the operating scenarios for autonomous vehicles have gradually expanded from closed testing parks and high-grade structured highways to ordinary road scenarios such as suburban trunk roads and county and township roads. Unsignalized intersections, as the meeting points of traffic flows from different directions, are among the scenarios with the highest concentration of traffic conflicts and the highest accident risks in road traffic networks. They are also a core scenario that autonomous vehicles must overcome to move from demonstration operation to large-scale commercialization. Currently, the traffic safety, efficiency, and environmental adaptability of autonomous vehicles at unsignalized intersections have become core research hotspots in the fields of intelligent transportation and autonomous driving engineering.

[0003] Current highway engineering design and traffic safety assessment systems are based on the physiological perception characteristics, psychological reaction characteristics, and operational behavior patterns of traditional manual driving. Their core design parameters and safety assessment indicators are formulated around the core characteristics of human drivers, such as visual perception range, perception-braking reaction time delay, and operational control precision. They fail to consider the fundamental differences between autonomous driving systems and human drivers in terms of perception mechanisms, decision-making logic, and control characteristics from the underlying design logic. This inherent mismatch between the infrastructure design system and the operational characteristics of autonomous driving directly leads to frequent problems for autonomous vehicles operating at unsignalized intersections, including a mismatch between sight distance supply and demand, insufficient prediction of dynamic interaction conflicts, and a mismatch between safety redundancy design and system capabilities. This results in numerous traffic safety hazards and efficiency bottlenecks.

[0004] Currently, research on the adaptability and safety of autonomous driving at unsignalized intersections is still in its early stages and exploratory phase. Existing research largely focuses on optimizing autonomous driving traffic control algorithms and conflict avoidance strategies at the single-vehicle intelligence level. Most of these studies are based on simulations with ideal environments and fixed parameters, lacking a systematic analysis and quantitative evaluation of the coupled effects of autonomous driving system characteristics, intersection geometry and physics, and the dynamic traffic environment. Meanwhile, existing methods for assessing the drivability of autonomous vehicles on roads mostly still use the traditional framework for manual driving assessments, focusing primarily on compliance checks for stopping sight distances in static obstacle scenarios. Few studies incorporate dynamic vehicle interaction behavior and multi-vehicle conflict evolution processes at unsignalized intersections into the drivability assessment system, failing to comprehensively reflect the overall adaptability of autonomous vehicles in real-world operating environments.

[0005] In research on autonomous driving safety assessment involving dynamic interaction scenarios, existing technical solutions still suffer from multi-dimensional technical shortcomings. Current assessments often employ single, time-based safety surrogate indicators. While these indicators can reflect the urgency of conflicts between vehicles to some extent, they cannot quantify the braking space redundancy requirements arising from differences in the perception-decision-control link between different autonomous driving systems. Even with identical time-based conflict indicators, autonomous driving systems with different automation levels and sensor configurations exhibit significant differences in actual collision risk and safety response capabilities, which a single indicator cannot effectively differentiate. Furthermore, current safety assessment methods based on stopping sight distance are primarily applicable to emergency braking scenarios with stationary obstacles, and cannot adapt to dynamic vehicle interaction scenarios under yield rules at unsignalized intersections. The safety of autonomous vehicles acting as yielding vehicles depends not only on their ability to stop before the conflict point but also on continuous perception of conflicting vehicles and safe separation of their spatiotemporal trajectories during dynamic driving. Traditional static assessment methods cannot cover this core requirement. More importantly, existing assessment methods generally lack adaptation mechanisms for vehicles of different levels of automation. Different levels of autonomous driving systems have fundamental differences in the division of human and machine responsibilities, perception and response latency, takeover logic design, and control execution accuracy. In particular, the driver takeover latency and safety control logic during the takeover transition period involved in conditional autonomous driving systems are not embedded in the assessment index system in existing methods. As a result, the assessment results cannot truly reflect the actual risk level of different levels of autonomous driving vehicles at unsignalized intersections, nor can they provide scientific support for the road adaptability design of different levels of autonomous driving vehicles. Summary of the Invention

[0006] To address the shortcomings and deficiencies of existing technologies, this invention provides a method for assessing the drivability of unsignalized intersections on highways for autonomous driving. This method first acquires the functional characteristics of autonomous driving, the geometric parameters of the unsignalized intersection to be assessed, weather conditions, and the effective detection distance of autonomous driving in the corresponding scenario. It then constructs a dataset mapping the relationships between multiple scenario elements and effective detection distances using machine learning methods, providing fundamental support for the entire assessment process that adapts to the perception characteristics of autonomous driving. Based on the driving automation level of the target vehicle, it matches the perception, braking, and takeover characteristic parameters of the corresponding level, and calculates the required stopping sight distance for the corresponding level in a differentiated manner, embedding the core characteristic differences of autonomous driving systems at different automation levels into the construction of the assessment benchmark. For autonomous vehicle-stationary obstacle collision scenarios, it combines the effective detection distance and intersection geometric parameters to obtain the available sight distance. Based on structural reliability theory, it constructs a sight distance function function, and calculates the sight distance failure probability using the available sight distance and the required stopping sight distance for the corresponding level, thus completing the assessment. The system performs a tiered drivability assessment for static scenarios. For collision scenarios involving autonomous vehicles and dynamically interacting vehicles, it calculates the Post-Intrusion Time (PET), representing the urgency of the conflict, and the Stop Distance Ratio (PSD), representing the redundancy of braking space. The PSD is calculated based on the required stopping sight distance for the corresponding level of driving automation. Simultaneously, an adaptive clustering data-driven method pre-determines the PET safety thresholds for different conflict conditions. The comparison between PET and the corresponding safety threshold serves as the first judgment level, and the PSD serves as the second verification judgment level, completing a spatiotemporal dual-dimensional drivability assessment for dynamic scenarios. Finally, by integrating the assessment results of static and dynamic scenarios, the system outputs the drivability assessment results for autonomous driving at unsignalized intersections on highways. This achieves a systematic quantitative assessment of drivability across all scenarios (static obstacles and dynamic interactions) and all levels of autonomous driving at unsignalized intersections, providing scientific and technical support for the safety verification of autonomous vehicles and the intelligent adaptation and transformation of highway infrastructure.

[0007] The specific technical solution adopted by this invention to solve its technical problem is as follows:

[0008] A method for evaluating the drivability of unsignalized intersections on highways for autonomous driving includes:

[0009] Obtain the geometric parameters, weather conditions, functional characteristics, and driving automation level of the target autonomous vehicle at the unsignalized intersection of the highway to be evaluated;

[0010] Based on the target vehicle's driving automation level, the corresponding perception, braking, and takeover characteristic parameters are matched, and the required parking sight distance for the corresponding level is calculated differentially.

[0011] The effective detection distance for autonomous driving corresponding to the target scenario is obtained, and the available line of sight is obtained by combining the geometric parameters of the intersection. The line of sight failure probability is calculated by using the line of sight function constructed based on the structural reliability theory, and the static drivability index is obtained by using the available line of sight and the required parking line of sight.

[0012] For autonomous vehicle-dynamic interactive vehicle collision scenarios, the following metrics are calculated: time of intrusion (PET) and spatial redundancy (PSD). The PSD is the ratio of the effective remaining distance when the vehicle detects a collision risk to the required stopping sight distance for the corresponding level of driving automation. A PET safety threshold corresponding to the current collision condition is obtained, pre-determined using an adaptive clustering data-driven method. The comparison between PET and the PET safety threshold serves as the first judgment level, and the PSD serves as the second verification judgment level, resulting in dynamic drivability indicators.

[0013] Based on the aforementioned static and dynamic drivability indicators, a drivability assessment is achieved.

[0014] Furthermore, the required parking sight distance for the corresponding level is calculated using differentiated logic based on the different levels of driving automation, specifically as follows:

[0015] For Level 1 and Level 2 driving automation, the calculation of required parking sight distance includes both the driver's perception-braking reaction time and the autonomous driving system's perception-braking reaction time.

[0016] For Level 3 driving automation, the calculation of required parking sight distance incorporates the autonomous driving system's perception reaction time, driver's perception-braking reaction time, driver takeover reaction time, and preset deceleration during the takeover transition period.

[0017] For Level 4 driving automation, the calculation of required parking sight distance only includes the perception-braking reaction time of the autonomous driving system, and does not include driver-related characteristic parameters.

[0018] The required parking sight distance (RSD) for different levels of driving automation j The calculation formula is:

[0019]

[0020] In the formula, j represents the level of driving automation of the vehicle, and the values ​​of j 1, 2, 3, and 4 correspond to levels one to four of driving automation, respectively. The design speed of the intersection to be evaluated; For driver perception-braking reaction time; For the perception-braking response time of autonomous driving systems; For the perception and reaction time of autonomous driving systems; Where i is the longitudinal friction coefficient; i is the longitudinal slope. This is the driver's reaction time to take over. Preset deceleration for the transition period when the Level 3 driving automation system takes over.

[0021] Furthermore, the sight distance function is constructed with the available sight distance as the sight distance supply and the corresponding level of required parking sight distance as the sight distance demand. The sight distance failure probability is the probability that the available sight distance is less than the required parking sight distance, which is calculated using the Monte Carlo method.

[0022] Furthermore, the post-intrusion time PET is the time interval between the voluntary vehicle and the target vehicle passing through the conflict area one after the other; in the parking distance ratio PSD, the effective remaining distance when the voluntary vehicle detects the conflict risk is taken as the trajectory distance from the voluntary vehicle's detection point to the point of conflict between the two vehicles' trajectories.

[0023] Furthermore, the method for pre-determining the PET safety threshold is as follows:

[0024] For various typical working conditions of collisions between autonomous vehicles and dynamic interactive vehicles, multiple sets of conflict scenario samples are generated through simulation, and the PET calculation value and corresponding collision occurrence label of each set of samples are recorded.

[0025] An adaptive clustering method is used to cluster the samples, and the collision probability within each cluster is calculated. The PET value corresponding to the boundary of the cluster where the collision probability first exceeds the preset risk tolerance threshold is determined as the PET safety threshold for the corresponding working condition.

[0026] By iterating through all typical working conditions, a mapping library between working condition types and PET safety thresholds is constructed, and the PET safety threshold for the corresponding working condition is directly retrieved during the assessment.

[0027] Furthermore, the specific execution rules for the first determination level and the second review determination level are as follows:

[0028] If PET is greater than or equal to the corresponding PET safety threshold, it is directly determined that the current dynamic scenario is safe to drive, and the corresponding dynamic drivability index is safe.

[0029] If PET is less than the corresponding PET safety threshold, the parking distance ratio PSD is used for further verification. If PSD is greater than or equal to 1, the current dynamic scenario is considered safe to drive, and the corresponding dynamic driving suitability index is safe. If PSD is less than 1, the current dynamic scenario does not meet the requirements for safe driving, and the corresponding dynamic driving suitability index is risky.

[0030] Furthermore, the effective detection distance for autonomous driving corresponding to the target scenario is obtained by matching a preset effective detection distance dataset for autonomous driving. The dataset is constructed using machine learning methods to establish a mapping relationship between autonomous driving functional features, intersection geometric parameters, weather environmental conditions, and effective detection distance, and includes pairing data of multiple scene elements and corresponding effective detection distances.

[0031] Furthermore, the autonomous vehicle-dynamic interactive vehicle collision scenario includes three typical working conditions: straight-on-straight-on conflict, left-turn-straight-on conflict, and right-turn-straight-on conflict. For different typical working conditions, based on the vehicle's driving trajectory, intersection geometric parameters, and effective detection distance, the spatial positions of the autonomous vehicle's detection point, the point of trajectory conflict between the two vehicles, and the effective detection point of the target vehicle are determined.

[0032] Furthermore, a drivability assessment system for unsignalized intersections on highways for autonomous driving, adapted to the methods provided above, includes:

[0033] The parameter acquisition module is used to acquire the geometric parameters, weather conditions, functional characteristics and driving automation level of the target autonomous vehicle at the unsignalized intersection of the highway to be evaluated.

[0034] The baseline calculation module is used to match the perception, braking and takeover characteristic parameters of the corresponding level based on the driving automation level of the target vehicle, and to calculate the required parking sight distance for the corresponding level in a differentiated manner.

[0035] The static evaluation module is used to obtain the effective detection distance for autonomous driving corresponding to the target scenario, and to obtain the available line of sight by combining the geometric parameters of the intersection; through the line of sight function constructed based on the structural reliability theory, the line of sight failure probability is calculated from the available line of sight and the required parking line of sight, and the static drivability index is obtained.

[0036] The dynamic evaluation module is used to calculate the time-dimensional collision index Post-Intrusion Time (PET) and the spatial-dimensional redundancy index Parking Distance Ratio (PSD) for autonomous vehicle-dynamic interactive vehicle collision scenarios. It obtains the PET safety threshold corresponding to the current collision condition, uses the comparison result of PET and the PET safety threshold as the first judgment level, and uses the Parking Distance Ratio (PSD) as the second review judgment level to obtain the dynamic drivability index.

[0037] And a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0038] Compared to existing technologies, this invention and its preferred solution overcome the limitation of the separation between static obstacle scenario and dynamic interaction scenario assessment in existing technologies. It constructs a systematic drivability assessment framework covering all conflict scenarios at unsignalized highway intersections, filling the technical gap in existing assessment methods that focus primarily on static sight distance verification and lack consideration of dynamic vehicle interaction behavior. Addressing the dimensionality deficiency of existing single-time-dimensional safety indicators, this invention constructs a time-space dual-dimensional coupled assessment framework. It quantifies the urgency of conflict between vehicles through time-dimensional indicators and quantifies the braking safety redundancy of autonomous driving systems through spatial-dimensional indicators. Simultaneously, it embeds the characteristic differences of autonomous driving systems at different automation levels into the assessment indicator system, solving the core problem that traditional indicators cannot distinguish the differences in safety capabilities of different autonomous driving systems, making the assessment results more closely reflect the actual operational risks of autonomous vehicles. This invention uses a data-driven approach to determine safety assessment thresholds, avoiding the subjective bias caused by traditional human experience-based threshold setting. It can adapt to different types of conflict conditions, effectively improving the objectivity and accuracy of the assessment results. This invention establishes an evaluation mechanism adaptable to all levels of autonomous driving. It addresses the fundamental differences in the perception-decision-control chain among vehicles of different automation levels, particularly the operational characteristics related to the division of human and machine responsibilities and takeover logic. By matching differentiated evaluation benchmarks, it can accurately reflect the actual drivability of autonomous vehicles at unsignalized intersections. Furthermore, this solution possesses good scalability and scenario compatibility, providing a scientific theoretical basis and technical support for the safety verification and engineering application of autonomous vehicles at unsignalized intersections, as well as the intelligent adaptation and transformation of highway infrastructure. Attached Figure Description

[0039] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0040] Figure 1 This is a flowchart illustrating the implementation of the drivability assessment method for unsignalized intersections on highways for autonomous driving provided in this embodiment of the invention.

[0041] Figure 2 This is a schematic diagram of the process for constructing an effective detection distance dataset for autonomous driving in an embodiment of the present invention;

[0042] Figure 3 This is a flowchart illustrating the calculation of the line-of-sight failure probability in an autonomous vehicle-stationary obstacle collision scenario according to an embodiment of the present invention.

[0043] Figure 4 This is a flowchart of the two-dimensional alternative indicator of the ratio of intrusion time to parking distance after an autonomous vehicle-dynamic interactive vehicle collision scenario in an embodiment of the present invention;

[0044] Figure 5This is a flowchart of the PET safety threshold determination based on adaptive clustering in an embodiment of the present invention;

[0045] Figure 6 This is a flowchart of the two-dimensional drivability determination process for an autonomous vehicle-dynamic interactive vehicle collision scenario in an embodiment of the present invention. Detailed Implementation

[0046] To make the features and advantages of the present invention more apparent and understandable, specific embodiments are described below in detail:

[0047] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0048] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0049] The purpose of this invention is to provide a drivability assessment method for autonomous driving at unsignalized intersections. This method comprehensively considers the interaction risks of autonomous vehicles with static obstacles and dynamic interactive vehicles at unsignalized intersections, and can quantitatively assess the drivability of autonomous vehicles at unsignalized intersections. By acquiring autonomous driving functional characteristics, intersection geometric parameters, weather conditions, and effective detection distance of autonomous vehicles, a dataset of effective detection distances for autonomous vehicles is constructed using machine learning methods. A line-of-sight function is constructed to calculate the line-of-sight failure probability in a collision scenario between an autonomous vehicle and a stationary obstacle. For a collision scenario between an autonomous vehicle and a dynamic interactive vehicle, a safe threshold for post-intrusion time is determined based on an adaptive clustering method, and a spatial-temporal dual-dimensional drivability substitution index is established in conjunction with the stopping distance ratio to quantify the collision risk of autonomous vehicles in this scenario. This invention comprehensively realizes the drivability assessment of autonomous vehicles at unsignalized intersections under both static obstacle and dynamic interaction scenarios, providing an effective technical means for autonomous driving safety risk assessment.

[0050] The specific implementation steps of the above solution are as follows:

[0051] Step S1: By acquiring autonomous driving function features, intersection geometric parameters, weather conditions, and effective detection distance of autonomous driving, construct an effective detection distance dataset for autonomous driving using machine learning methods;

[0052] Step S2: Construct the line-of-sight function and calculate the line-of-sight failure probability in the autonomous vehicle-stationary obstacle collision scenario;

[0053] Step S3: For the autonomous vehicle-dynamic interactive vehicle collision scenario, determine the safety threshold of the intrusion time based on the adaptive clustering method, and establish a space-time dual-dimensional drivability substitution index in conjunction with the parking distance ratio to quantify the collision risk of autonomous vehicles in this scenario.

[0054] The specific implementation process of step S1 is as follows:

[0055] Step S11: Obtain the characteristics of the autonomous driving function, intersection geometric parameters, weather conditions, and effective detection distance of the autonomous driving system.

[0056] The features of autonomous driving functions include at least: different levels of automation, the number of LiDAR lines, sensor installation height and other configurations, as well as the perception-braking reaction time of the autonomous driving system, the perception-braking reaction time of the driver, the perception reaction time of the autonomous driving system, the driver takeover reaction time, and the preset deceleration of the Level 3 autonomous driving system during the time interval from issuing a takeover request to the driver performing the takeover action.

[0057] The geometric parameters of an intersection should include at least the following: intersection angle, design speed, lane width, number of lanes, turning radius, longitudinal friction coefficient, and longitudinal slope.

[0058] Weather conditions include at least: weather type (sunny, rainy, foggy, etc.); light intensity, visibility, rainfall intensity, road surface friction coefficient, etc.

[0059] The effective detection distance of autonomous driving is obtained after multi-scenario simulation testing through an autonomous driving simulation platform based on the aforementioned autonomous driving functional characteristics, intersection geometric parameters, and weather conditions. The simulation test needs to reproduce the detection characteristics of autonomous driving sensors under the coupling of different scenario elements and output the measured value of the effective detection distance for the corresponding scenario.

[0060] Step S12: Construct an effective detection distance dataset for autonomous driving using machine learning methods:

[0061] A mapping relationship between autonomous driving functional characteristics, intersection geometric parameters, weather conditions, and effective detection distance was established. Ten-fold cross-validation and grid search were used to optimize hyperparameters. Multiple models were trained, and the model with the best performance was selected to construct an effective detection distance dataset for unsignalized intersections that includes paired data of multiple scene elements and effective detection distances.

[0062] The specific implementation process of step S2 is as follows:

[0063] Step S21: For autonomous vehicle-stationary obstacle collision scenarios, the available line of sight is calculated by combining the effective detection distance with the intersection geometric parameters, and this is used as the supply. The required parking line of sight is calculated based on the autonomous driving level, design speed, longitudinal slope, and road surface friction coefficient, and this is used as the demand. The required parking line of sight function RSD for autonomous vehicles of different automation levels is as follows:

[0064]

[0065] In the formula, RSD j V represents the parking sight distance for Level j autonomous driving, where j = 1, 2, 3, 4; d For design speed; t pb_h For driver perception-braking reaction time; t pb_s For the perception-braking reaction time of an autonomous driving system; t p_s For the perception and reaction time of the autonomous driving system; f L t is the longitudinal friction coefficient; i is the longitudinal slope; t T / This is the driver's takeover reaction time; a dp The preset deceleration for a Level 3 autonomous driving system during the time interval between issuing a takeover request and the driver performing the takeover action;

[0066] Step S22: Construct the line-of-sight function using structural reliability theory:

[0067]

[0068] In the formula, S (supply) represents the supply function, R (require) represents the demand function; ASD represents the available line of sight for autonomous vehicles; RSD j This indicates the required stopping sight distance for autonomous vehicles;

[0069] Step S23: Based on the obtained autonomous driving function features and intersection geometric parameter information, determine t. pb_h t pb_s t p_s t T a dp , i, f L The calculated value, t pb_h t pb_s t p_s t T The mean, standard deviation, and probability distribution form;

[0070] Step S24: Calculate the line-of-sight failure probability P using the Monte Carlo method. f :

[0071] .

[0072] The specific implementation process of step S3 is as follows:

[0073] Step S31: For the autonomous vehicle-dynamic interactive vehicle collision scenario, distinguish different typical working conditions, and determine the spatial location of the autonomous vehicle's detection point, the collision point, and the target vehicle's effective detection point based on the vehicle's driving trajectory, intersection geometric parameters, and the obtained effective detection distance.

[0074] Step S32: Calculate the distance D1 from the vehicle to the conflict point and the distance D2 from the target vehicle to the conflict point. Combining the speeds of both vehicles, define the subsequent intrusion time PET as the time interval between the two vehicles passing through the conflict area. The calculation formula is:

[0075]

[0076] In the formula, t1 is the time when the vehicle leaves the conflict point, t2 is the time when the target vehicle enters the conflict point, and L v Let V be the vehicle length, and V1 and V2 be the speeds of the vehicle and the target vehicle, respectively.

[0077] Step S33: Define the parking distance ratio PSD as the effective remaining distance RD of the vehicle when a conflict risk is detected, and its corresponding required parking sight distance RSD for the autonomous driving level j. j The ratio is calculated using the following formula:

[0078]

[0079] In the formula, RD is the trajectory distance D1 from the vehicle's detection point to the collision point, and RSD is... j This represents the required stopping sight distance for autonomous vehicles, where j = 1, 2, 3, 4;

[0080] Step S34: Determine the safety threshold PET for post-intrusion time based on an adaptive clustering method. threshold ,like Figure 5 As shown, it specifically includes:

[0081] Step S341: For the type of conflict condition to be evaluated, generate multiple conflict scenario samples through simulation, and record the PET calculation value and corresponding collision occurrence label for each sample.

[0082] Step S342: Concatenate the PET values ​​with the corresponding scene features to form a multi-dimensional feature vector, which is used as the clustering input;

[0083] Step S343: Determine the optimal number of clusters K by combining the elbow rule with the silhouette coefficient, and perform K-means clustering on the samples;

[0084] Step S344: Calculate the collision probability within each cluster, i.e., the number of collision samples within the cluster divided by the total number of samples in the cluster. Determine the PET value corresponding to the cluster boundary where the collision probability first exceeds the preset risk tolerance threshold as the safety threshold PET under this operating condition. threshold The preset risk tolerance threshold is set with reference to the safety level standards for highway engineering structures.

[0085] Step S345: Traverse all typical conflicting working conditions and construct a threshold library containing the mapping relationship between working condition type and PET safety threshold. During the evaluation, the corresponding safety threshold is automatically retrieved based on the currently identified conflicting working condition.

[0086] Step S35, as Figure 6 As shown, a two-dimensional criterion for determining drivability is constructed: if PET ≥ PET threshold If PET is deemed to be driving safely; <PET threshold Then, the PSD value is further calculated. If PSD≥1, it is determined to be safe driving; if PSD<1, it is determined to be unsafe driving.

[0087] Compared with the prior art, the differences and advantages of the above solutions of the present invention include:

[0088] (1) A joint evaluation framework of PET and PSD is proposed. It is not a simple superposition of indicators, but a coupled design based on in-depth analysis of the dynamic interaction behavior of autonomous driving: PET quantifies the urgency of conflict, and PSD embeds the autonomous driving level characteristics into the spatial redundancy measurement. The two complement each other and realize the quantitative adaptation of the differences in autonomous driving levels in dynamic interaction scenarios, taking into account both the urgency of conflict and the system response capability.

[0089] (2) An adaptive clustering method is introduced to determine the PET safety threshold. The data-driven generation avoids the subjectivity of human experience setting. Different working conditions correspond to different thresholds, realizing the scenario adaptation of the threshold. The evaluation results are more objective and accurate.

[0090] (3) It comprehensively considers the interaction risks between autonomous vehicles and static obstacles and dynamic interactive vehicles in the scenario of unsignalized intersections, and can comprehensively and quantitatively evaluate the driving suitability level of autonomous vehicles at unsignalized intersections, filling the gap in the field of lack of systematic evaluation methods.

[0091] (4) It has good scalability and compatibility, and can be adapted to different levels of automation, sensor configurations and intersection geometry types, providing theoretical basis and technical support for the engineering application, safety verification and intelligent transformation of road infrastructure of autonomous driving system in unsignalized intersections.

[0092] To make the features and advantages of this patent application more apparent and understandable, specific embodiments are described below in conjunction with the accompanying drawings:

[0093] like Figure 1 The figure shows the driving suitability assessment method for unsignalized intersections on highways for autonomous driving proposed in this invention, which includes the following steps:

[0094] (1) Obtain the features of autonomous driving functions, intersection geometric parameters, weather conditions and effective detection distance of autonomous driving, and use machine learning methods to construct an effective detection distance dataset for autonomous driving;

[0095] 1) Obtain the characteristics of autonomous driving functions, intersection geometric parameters, weather conditions, and effective detection distance of autonomous driving;

[0096] The features of autonomous driving functions include at least: different levels of automation, the number of LiDAR lines, sensor installation height and other configurations, as well as the perception-braking reaction time of the autonomous driving system, the perception-braking reaction time of the driver, the perception reaction time of the autonomous driving system, the driver takeover reaction time, and the preset deceleration of the Level 3 autonomous driving system during the time interval from issuing a takeover request to the driver performing the takeover action.

[0097] The geometric parameters of an intersection should include at least the following: intersection angle, design speed, lane width, number of lanes, turning radius, longitudinal friction coefficient, and longitudinal slope.

[0098] Weather conditions include at least the weather type and corresponding quantitative indicators, such as light intensity, visibility, rainfall intensity, and road surface friction coefficient.

[0099] The effective detection distance for autonomous driving is obtained based on the aforementioned autonomous driving functional characteristics, intersection geometric parameters, and weather conditions. It is obtained through multi-scenario simulation tests conducted on an autonomous driving simulation platform. The simulation tests need to reproduce the detection characteristics of autonomous driving sensors under the coupling of different scenario elements and output the measured effective detection distance values ​​for the corresponding scenarios.

[0100] Among them, the characteristics of autonomous driving functions can be obtained through on-site testing, review of existing research results, or retrieval of product technical data; the geometric parameters of intersections can be obtained through on-site collection, extraction of high-precision maps, or retrieval of road design drawings; weather and environmental information can be obtained through on-site equipment monitoring and collection, or by statistical analysis of publicly available weather information data; and the effective detection distance of autonomous driving can be obtained through multi-scenario simulation testing using an integrated simulation platform of Prescan and MATLAB.

[0101] 2) Construct an effective detection distance dataset for autonomous driving using machine learning methods: Establish the mapping relationship between autonomous driving functional features, intersection geometric parameters, weather conditions, and effective detection distance. Use 10-fold cross-validation + grid search to optimize hyperparameters, train multiple models, select the best-performing model, and construct an effective detection distance dataset for unsignalized intersections containing paired data of multiple scene elements and effective detection distances. The flowchart for this step is shown below. Figure 2 As shown.

[0102] The model used includes multiple models such as random forest, XGBoost, and support vector regression. The model with the best performance is selected to construct an effective detection distance dataset, which serves as the basis for calculating the available line of sight in subsequent steps.

[0103] The above dataset of effective detection distances for autonomous driving serves as the basic input for subsequent static and dynamic scenario evaluations. When conducting drivability assessments, the effective detection distances for the corresponding scenarios are retrieved from the dataset based on the autonomous driving function characteristics, intersection geometric parameters, and weather conditions of the scenario to be evaluated, or the effective detection distances for the corresponding scenarios are predicted and output through machine learning models trained in the dataset.

[0104] (2) Construct a line-of-sight function to calculate the line-of-sight failure probability in a collision scenario between an autonomous vehicle and a stationary obstacle; the flowchart for this step is as follows. Figure 3 As shown;

[0105] 1) For autonomous vehicle-stationary obstacle collision scenarios, the available line of sight is calculated by combining the effective detection distance with the intersection's geometric parameters, and this is used as the supply. The required stopping line of sight is calculated based on the autonomous driving level, design speed, longitudinal slope, and road surface friction coefficient, and this is used as the demand. The required stopping line of sight function RSD for autonomous vehicles of different automation levels is as follows:

[0106]

[0107] In the formula, RSD j (m) represents the stopping sight distance for Level j automated driving, j = 1, 2, 3, 4; V d (km·h) -1 ) represents the design speed; t pb_h (s) represents the driver's perception-braking reaction time; t pb_s (s) represents the perception-braking response time of the autonomous driving system; t p_s (s) represents the perception and reaction time of the autonomous driving system; f L The longitudinal friction coefficient is denoted by i; the longitudinal slope (%) is denoted by t. T (s) represents the driver's takeover reaction time; a dp / (m·s -2() is the preset deceleration of the Level 3 automated driving system during the time interval between issuing a takeover request and the driver performing the takeover action;

[0108] As a further preferred implementation, the above method for calculating the available line-of-sight (ASD) is as follows: based on the effective detection distance of the autonomous vehicle at the detection point, combined with the geometric parameters of the intersection angle, lane width, and turning radius, the path distance from the vehicle's detection point to the obstacle vehicle in a conflict scenario is calculated, which is the available line-of-sight (ASD).

[0109] 2) Construct the line-of-sight function using structural reliability theory:

[0110]

[0111] In the formula, S (supply) represents the supply function, R (require) represents the demand function; ASD represents the available line of sight for autonomous vehicles; RSD represents the available line of sight for autonomous vehicles. j This indicates the required stopping sight distance for autonomous vehicles;

[0112] 3) Based on the obtained autonomous driving function characteristics and intersection geometric parameter information, determine t. pb_h t pb_s t p_s t T a dp , i, f L The calculated value, t pb_h t pb_s t p_s t T The mean, standard deviation, and probability distribution form;

[0113] The typical value range and probability distribution of the above parameters are determined based on publicly available academic research results and industry standards, where i=2%, and the other parameters are shown in Table 1 and Table 2 below;

[0114] Table 1. Characterization parameters of autonomous driving response capabilities for levels 1-4

[0115]

[0116] Table 2. Longitudinal friction coefficient (f) L Design value

[0117]

[0118] 4) The Monte Carlo method was used to calculate the line-of-sight failure probability:

[0119]

[0120] (3) For the autonomous vehicle-dynamic interactive vehicle collision scenario, a safety threshold for the post-intrusion time is determined based on an adaptive clustering method, and a spatial-temporal dual-dimensional drivability substitution index is established in conjunction with the parking distance ratio to quantify the collision risk of autonomous vehicles in this scenario; the flowchart of this step is as follows. Figure 4 As shown;

[0121] 1) For autonomous vehicle-dynamic interactive vehicle collision scenarios, different typical working conditions are distinguished. Based on the vehicle driving trajectory, intersection geometric parameters and the obtained effective detection distance, the spatial positions of the autonomous vehicle detection point, the collision point and the target vehicle's effective detection point are determined.

[0122] Among them, different typical working conditions include straight-on-straight-on conflict, left-turn-straight-on conflict, and right-turn-straight-on conflict. The spatial positions of the three points are the self-vehicle detection point where the self-vehicle can first detect the target vehicle, the conflict point where the trajectories of the two vehicles intersect, and the effective detection point of the target vehicle when the target vehicle is located at the self-vehicle detection point.

[0123] 2) Calculate the distance D1 from the vehicle to the conflict point and the distance D2 from the target vehicle to the conflict point. Combining the speeds of both vehicles, define the subsequent intrusion time PET as the time interval between the two vehicles passing through the conflict area. The calculation formula is:

[0124]

[0125] In the formula, t1 is the time (s) when the vehicle leaves the conflict point, t2 is the time (s) when the target vehicle enters the conflict point, and L v Let V be the vehicle length (m), and V1 and V2 be the speeds of the vehicle and the target vehicle (m / s), respectively.

[0126] 3) Define the parking distance ratio PSD as the effective remaining distance RD of the vehicle when a conflict risk is detected, and its corresponding required parking sight distance RSD for the autonomous driving level j. j The ratio is calculated using the following formula:

[0127]

[0128] In the formula, RD is the trajectory distance D1 from the vehicle's detection point to the collision point, and RSD is... j This represents the required stopping sight distance for autonomous vehicles, where j = 1, 2, 3, 4;

[0129] 4) Determine the safety threshold PET for post-intrusion time based on adaptive clustering method. threshold ;

[0130] Traditional PET thresholds largely rely on empirical values ​​and lack adaptability to scene features. This invention proposes an adaptive clustering method to determine the threshold, with the following specific steps:

[0131] ① Sample generation: For the type of conflict condition to be evaluated, M conflict scenarios are generated in batches through simulation, and the PET value and corresponding collision occurrence label Y of each scenario are recorded;

[0132] Among them, the conflict conditions include straight-on-straight-on conflict, left-turn-straight-on conflict, and right-turn-straight-on conflict. M≥1000, Y=1 indicates that a collision has occurred, and Y=0 indicates that a collision has not occurred.

[0133] ② Feature space construction: The PET value is concatenated with the corresponding scene features to form a multi-dimensional feature vector X, which is used as the clustering input;

[0134] The scene features include intersection angle θ, vehicle speed V1, target vehicle speed V2, relative heading angle α, and multidimensional feature vector X=[PET, θ, V1, V2, α];

[0135] ③ Adaptive K value determination: The elbow rule combined with the silhouette coefficient is used to determine the optimal number of clusters K. The sum of squared clustering errors (SSE) and the average silhouette coefficient are calculated as K changes from 2 to 10. The K value with the slowest decrease in SSE (elbow point) and the largest silhouette coefficient is selected as the number of clusters.

[0136] ④ Threshold extraction: For each cluster, calculate the collision probability P of samples within the cluster. collision P collision The PET value corresponding to the cluster boundary that first exceeds the preset risk tolerance threshold is determined as the safety threshold PET for that operating condition. threshold ;

[0137] Among them, P collision =Number of collision samples within the cluster / Total number of samples in the cluster. The preset risk tolerance threshold is the failure probability, for example, the failure probability of a Class III highway is 20%.

[0138] ⑤ Threshold Library Construction: Repeat the above steps, iterating through all typical conflict conditions, to construct an adaptive threshold library containing the mapping relationship between condition types and PET safety thresholds. During evaluation, the system automatically retrieves the corresponding PET based on the currently identified conflict condition. threshold .

[0139] 5) such as Figure 6 As shown, a two-dimensional criterion for determining drivability is constructed: if PET ≥ PET threshold If PET is deemed to be driving safely; <PET threshold Then, the PSD value is further calculated. If PSD≥1, it is determined to be safe driving; if PSD<1, it is determined to be unsafe driving.

[0140] Among them, PET is used as the primary screening indicator to efficiently eliminate scenarios without urgent conflicts; for potential conflict scenarios, PSD is introduced for spatial redundancy verification, which fully considers the impact of differences in autonomous driving levels on safety.

[0141] Based on the above design, as a preferred embodiment of the present invention, the specific steps for achieving a drivability assessment based on the static drivability index calculated according to the corresponding sight distance failure probability and the dynamic drivability index according to the corresponding dual-dimensional drivability judgment criteria are as follows:

[0142] S1, Preset Driving Suitability Assessment Criteria

[0143] Based on the engineering safety level of the highway to be evaluated, and referring to standards such as the "Unified Standard for Reliability Design of Highway Engineering Structures" (JTG2120-2020), and in conjunction with the design provided in the embodiments of this invention above, a corresponding level of sight failure probability tolerance threshold and dynamic scenario safety coverage requirements are preset to form a driving suitability assessment benchmark. Specifically, highway engineering safety levels are divided into Level 1, Level 2, Level 3, and Level 4, corresponding to different risk tolerance thresholds. For example, the sight failure probability tolerance threshold for Level 3 highways is 20%, and for Level 1 highways it is 5%. The dynamic scenario safety coverage requirement is that all typical conflict conditions must meet safe driving requirements.

[0144] S2. Static Drivability Assessment

[0145] Based on the sight failure probability tolerance threshold preset in step S1, the static driving suitability index is judged as qualified: if the sight failure probability of the scene to be evaluated is less than or equal to the tolerance threshold of the corresponding level, the static scene driving suitability is judged to be qualified; if the sight failure probability is greater than the tolerance threshold of the corresponding level, the static scene driving suitability is judged to be unqualified, and the final conclusion of the driving suitability of the intersection to be evaluated is directly output as unqualified.

[0146] S3, Dynamic Driving Suitability Compliance Assessment

[0147] The typical conflict scenarios correspond one-to-one with the scenarios covered in the aforementioned PET safety threshold determination process. For all typical conflict scenarios at the intersection to be evaluated (straight-straight conflict, left-turn-straight conflict, right-turn-straight conflict), the dynamic drivability indicators corresponding to each scenario are obtained one by one to complete the dynamic drivability qualification judgment for all scenarios: if the dynamic drivability indicators for all typical conflict scenarios are safe, the dynamic scenario drivability is judged to be qualified; if any typical conflict scenario has a dynamic drivability indicator that does not meet the safe driving requirements, the dynamic scenario drivability is judged to be unqualified.

[0148] S4, Overall Driving Suitability Level Output

[0149] Combining the static and dynamic drivability assessment results, the final drivability evaluation is completed, and the corresponding evaluation conclusion is output:

[0150] If the static scenario driving suitability is qualified and the dynamic scenario driving suitability is qualified, it is determined that the driving suitability of the unsignalized intersection on the highway to be evaluated is good and fully meets the safe passage requirements of the corresponding level of autonomous driving vehicles.

[0151] If the driving suitability in static scenarios is qualified, but the driving suitability in dynamic scenarios is not qualified, it is determined that the driving suitability of the unsignalized intersection on the highway to be evaluated needs to be optimized. The driving suitability can be improved through conventional methods in this field, such as intersection traffic organization optimization and intelligent infrastructure transformation.

[0152] If the static scene drivability is not up to standard, the drivability of the unsignalized intersection on the highway to be evaluated will be directly determined to be unsatisfactory, and the traffic safety of the corresponding level of autonomous vehicles cannot be guaranteed.

[0153] As a further refinement of this preferred embodiment, when there are multiple approach lanes with different design speeds and different longitudinal slopes at the intersection to be evaluated, the static drivability index and dynamic drivability index are calculated and judged for each approach lane. The lowest drivability level of all approach lanes is taken as the final drivability evaluation conclusion of the intersection, ensuring that the evaluation results cover all traffic scenarios of the intersection.

[0154] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can implement all the steps of the above-described method for evaluating the drivability of unsignalized intersections for autonomous driving. This computer device can be implemented as a general-purpose computing device, including but not limited to industrial control computers, personal computers, servers, cloud computing nodes, etc. Internally, it can be connected to basic structures such as processors, memory, communication interfaces, and input / output units via a system bus to meet the computing and deployment requirements of this solution.

[0155] This embodiment also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements all the steps of the method described above in this embodiment. The computer-readable storage medium may include volatile and non-volatile memory, including but not limited to random access memory (RAM), read-only memory (ROM), flash memory, hard disk drive (HDD), solid-state drive (SSD), etc., and can be used to store the computer program instructions and related data of this solution, providing a stable storage medium for the implementation of the method.

[0156] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0157] This invention is not limited to the above-described preferred embodiments. Anyone inspired by this invention can derive other various forms of drivability assessment methods for unsignalized intersections on highways for autonomous driving. All equivalent variations and modifications made within the scope of the claims of this invention shall fall within the scope of this invention.

Claims

1. A method for evaluating the drivability of unsignalized intersections on highways for autonomous driving, characterized in that, include: Obtain the geometric parameters, weather conditions, functional characteristics, and driving automation level of the target autonomous vehicle at the unsignalized intersection of the highway to be evaluated; Based on the target vehicle's driving automation level, the corresponding perception, braking, and takeover characteristic parameters are matched, and the required parking sight distance for the corresponding level is calculated differentially. The effective detection distance for autonomous driving corresponding to the target scenario is obtained, and the available line of sight is obtained by combining the geometric parameters of the intersection. The line of sight failure probability is calculated by using the line of sight function constructed based on the structural reliability theory, and the static drivability index is obtained by using the available line of sight and the required parking line of sight. For autonomous vehicle-dynamic interactive vehicle collision scenarios, the intrusion time (PET) and parking distance ratio (PSD) are calculated as the temporal dimension of the conflict scenario and the spatial dimension of the redundancy index. The parking distance ratio (PSD) is the ratio of the effective remaining distance when the autonomous vehicle detects the conflict risk to the required parking sight distance for the corresponding level of driving automation. The PET safety threshold corresponding to the current conflict condition is obtained. The PET safety threshold is predetermined by an adaptive clustering data-driven method. The comparison result between the intrusion time (PET) and the PET safety threshold is used as the first judgment level, and the parking distance ratio (PSD) is used as the second review judgment level to obtain the dynamic drivability index. Based on the aforementioned static and dynamic drivability indicators, a drivability assessment is achieved. The specific execution rules for the first determination level and the second review determination level are as follows: If the post-intrusion time PET is greater than or equal to the corresponding PET safety threshold, it is directly determined that the current dynamic scenario is safe to drive, and the corresponding dynamic drivability index is safe. If the intrusion time PET is less than the corresponding PET safety threshold, the parking distance ratio PSD is used for further verification. If PSD is greater than or equal to 1, the current dynamic scenario is considered safe to drive, and the corresponding dynamic drivability index is safe. If PSD is less than 1, it is determined that the current dynamic scenario does not meet the requirements for safe driving, and the corresponding dynamic drivability index is risky.

2. The method for evaluating the drivability of unsignalized intersections on highways for autonomous driving according to claim 1, characterized in that: The required parking sight distance for the corresponding level is calculated using differentiated logic based on the different levels of driving automation, specifically as follows: For Level 1 and Level 2 driving automation, the calculation of required parking sight distance includes both the driver's perception-braking reaction time and the autonomous driving system's perception-braking reaction time. For Level 3 driving automation, the calculation of required parking sight distance incorporates the autonomous driving system's perception reaction time, driver's perception-braking reaction time, driver takeover reaction time, and preset deceleration during the takeover transition period. For Level 4 driving automation, the calculation of required parking sight distance only includes the perception-braking reaction time of the autonomous driving system, and does not include driver-related characteristic parameters. The required parking sight distance (RSD) for different levels of driving automation j The calculation formula is: In the formula, j represents the level of driving automation of the vehicle, and the values ​​of j 1, 2, 3, and 4 correspond to levels one to four of driving automation, respectively. The design speed of the intersection to be evaluated; For driver perception-braking reaction time; For the perception-braking response time of autonomous driving systems; For the perception and reaction time of autonomous driving systems; Where i is the longitudinal friction coefficient; i is the longitudinal slope. This is the driver's reaction time to take over. Preset deceleration for the transition period when the Level 3 driving automation system takes over.

3. The method for evaluating the drivability of unsignalized intersections on highways for autonomous driving according to claim 1, characterized in that: The sight distance function is constructed with the available sight distance as the sight distance supply and the corresponding level of required parking sight distance as the sight distance demand. The sight distance failure probability is the probability that the available sight distance is less than the required parking sight distance, and is calculated using the Monte Carlo method.

4. The method for evaluating the drivability of unsignalized intersections on highways for autonomous driving according to claim 1, characterized in that: The post-intrusion time PET is the time interval between the voluntary vehicle and the target vehicle passing through the conflict area; the parking distance ratio PSD is the effective remaining distance when the voluntary vehicle detects the risk of conflict, which is the trajectory distance from the voluntary vehicle's detection point to the point of conflict between the two vehicles.

5. The method for evaluating the drivability of unsignalized intersections on highways for autonomous driving according to claim 1, characterized in that: The method for pre-determining the PET safety threshold is as follows: For various typical working conditions of collisions between autonomous vehicles and dynamic interactive vehicles, multiple sets of conflict scenario samples are generated through simulation, and the post-intrusion time PET value and the corresponding collision occurrence label are recorded for each set of samples. An adaptive clustering method is used to cluster the samples, and the collision probability within each cluster is calculated. The post-intrusion time PET value corresponding to the cluster boundary where the collision probability first exceeds the preset risk tolerance threshold is determined as the PET safety threshold for the corresponding working condition. By iterating through all typical working conditions, a mapping library between working condition types and PET safety thresholds is constructed, and the PET safety threshold for the corresponding working condition is directly retrieved during the assessment.

6. The method for evaluating the drivability of unsignalized intersections on highways for autonomous driving according to claim 1, characterized in that: The effective detection distance for autonomous driving corresponding to the target scenario is obtained by matching a preset effective detection distance dataset for autonomous driving. The dataset is constructed using machine learning methods to establish a mapping relationship between autonomous driving functional features, intersection geometric parameters, weather conditions and effective detection distance, and includes multiple sets of paired data of scene elements and corresponding effective detection distances.

7. The method for evaluating the drivability of unsignalized intersections on highways for autonomous driving according to claim 1, characterized in that: The autonomous vehicle-dynamic interactive vehicle collision scenario includes three typical working conditions: straight-on-straight-on conflict, left-turn-straight-on conflict, and right-turn-straight-on conflict. For different typical working conditions, the spatial positions of the autonomous vehicle's detection point, the point of conflict between the two vehicle trajectories, and the target vehicle's effective detection point are determined based on the vehicle's driving trajectory, the intersection's geometric parameters, and the effective detection distance.

8. A drivability assessment system for unsignalized intersections on highways for autonomous driving, used to implement the method as described in claim 1, characterized in that, include: The parameter acquisition module is used to acquire the geometric parameters, weather conditions, functional characteristics and driving automation level of the target autonomous vehicle at the unsignalized intersection of the highway to be evaluated. The baseline calculation module is used to match the perception, braking and takeover characteristic parameters of the corresponding level based on the driving automation level of the target vehicle, and to calculate the required parking sight distance for the corresponding level in a differentiated manner. The static evaluation module is used to obtain the effective detection distance for autonomous driving corresponding to the target scenario, and to obtain the available line of sight by combining the geometric parameters of the intersection; through the line of sight function constructed based on the structural reliability theory, the line of sight failure probability is calculated from the available line of sight and the required parking line of sight, and the static drivability index is obtained. The dynamic evaluation module is used to calculate the time-dimensional conflict index Post-Intrusion Time (PET) and the spatial-dimensional redundancy index Parking Distance Ratio (PSD) for autonomous vehicle-dynamic interactive vehicle collision scenarios. It obtains the PET safety threshold corresponding to the current conflict condition, and uses the comparison result between Post-Intrusion Time (PET) and PET safety threshold as the first judgment level, and Parking Distance Ratio (PSD) as the second review judgment level to obtain the dynamic drivability index.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images