CN-121982934-A - Automatic driving-oriented highway signalless intersection driving adaptability evaluation method

CN121982934ACN 121982934 ACN121982934 ACN 121982934ACN-121982934-A

Abstract

The invention provides an automatic driving-oriented highway signalless intersection suitability assessment method which comprises the steps of matching perception, braking and takeover characteristic parameters of corresponding grades based on driving automation grades of target vehicles, calculating required parking vision distances of the corresponding grades in a differentiated mode, obtaining automatic driving effective detection distances corresponding to target scenes, obtaining available vision distances through combination of intersection geometric parameters, obtaining static suitability indexes through vision distance function functions, calculating intrusion time PET after time dimension conflict indexes of conflict scenes and space dimension redundancy index parking distance ratio PSD aiming at automatic driving vehicle-dynamic interactive vehicle collision scenes, and obtaining dynamic suitability indexes by taking a comparison result of PET and PET safety thresholds as a first judgment level and taking a parking distance ratio PSD as a second rechecking judgment level. And according to the static driving adaptability index and the dynamic driving adaptability index, the driving adaptability evaluation is realized.

Inventors

WANG SHUYI
WEI YUXIN
LAI Yuanwen
XU LING
YAN GUIFENG
LUO QINGLIN
Dong Zhefu
LIU WEI

Assignees

福州大学

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. An automatic driving-oriented highway signalless intersection suitability evaluation method is characterized by comprising the following steps of: obtaining geometric parameters and weather environmental conditions of a highway signalless intersection to be evaluated, and functional characteristics and driving automation level of a target automatic driving vehicle; based on the driving automation level of the target vehicle, matching the perception, braking and taking over characteristic parameters of the corresponding level, and differentially calculating the required parking sight distance of the corresponding level; acquiring an effective automatic driving detection distance corresponding to a target scene, and acquiring an available viewing distance by combining geometric parameters of an intersection; calculating the vision distance failure probability from the available vision distance and the required parking vision distance through a vision distance function constructed based on the structural reliability theory to obtain a static driving adaptability index; Aiming at an automatic driving vehicle-dynamic interactive vehicle collision scene, calculating intrusion time PET after a time dimension conflict index of the conflict scene and a space dimension redundancy index parking distance ratio PSD, wherein the parking distance ratio PSD is the ratio of the effective remaining distance of a vehicle when the collision risk is detected to the required parking sight distance of a corresponding driving automation level; and according to the static driving adaptability index and the dynamic driving adaptability index, the driving adaptability evaluation is realized.
2. The automated driving-oriented highway signalless intersection suitability assessment method of claim 1, wherein: the required parking sight distance of the corresponding grade adopts differentiated calculation logic according to different driving automation grades, and specifically comprises the following steps: Aiming at the primary and secondary driving automation level, the calculation of the required parking sight distance simultaneously comprises the sensing-braking reaction time of a driver and the sensing-braking reaction time of an automatic driving system; aiming at the three-level driving automation level, the calculation of the required parking sight distance is incorporated into the sensing reaction time, the driver sensing-braking reaction time, the driver taking over reaction time and the preset deceleration of the taking over transition period of an automatic driving system; aiming at a four-level driving automation level, the calculation of the required parking sight distance only includes the sensing-braking reaction time of an automatic driving system, and does not include the related characteristic parameters of a driver; the calculation formula of the required parking sight distance RSD j corresponding to different driving automation levels is as follows: wherein j is the driving automation level of the vehicle, and j values 1, 2,3 and 4 respectively correspond to one-level to four-level driving automation; The design speed of the intersection to be evaluated is set; sensing-braking reaction time for the driver; sensing-braking reaction time for an autopilot system; sensing a reaction time for the autopilot system; I is the gradient of the longitudinal slope; Take over the reaction time for the driver; The preset deceleration of the transition period is taken over for the three-level driving automation system.
3. The method for evaluating the suitability of a highway signalless intersection for automatic driving according to claim 1, wherein the vision distance function is constructed by taking available vision distance as vision distance supply and the vision distance required by stopping at a corresponding level as vision distance required, and the vision distance failure probability is the probability that the available vision distance is smaller than the required stopping vision distance and is calculated by adopting a Monte Carlo method.
4. The method for evaluating the suitability of the road signalless intersection for automatic driving according to claim 1, wherein the post-intrusion time PET is a time interval when the own vehicle and the target vehicle pass through the collision area in sequence, and the parking distance ratio PSD is an effective remaining distance when the own vehicle detects collision risk and takes a value of a track distance from a detection point of the own vehicle to a collision point of tracks of the two vehicles.
5. The automated driving-oriented highway signalless intersection suitability assessment method of claim 1, wherein: The PET safety threshold value pre-determining method specifically comprises the following steps: generating a plurality of groups of conflict scene samples by simulation aiming at various typical working conditions of automatic driving vehicle-dynamic interaction vehicle collision, and recording a post-intrusion time PET calculated value of each group of samples and a corresponding collision occurrence label; clustering samples by adopting a self-adaptive clustering method, counting collision probability in each clustering cluster, and determining a post-intrusion time PET value corresponding to a cluster boundary with the collision probability exceeding a preset risk tolerance threshold for the first time as a PET safety threshold of a corresponding working condition; and traversing all typical working conditions, constructing a mapping relation library of the working condition types and the PET safety threshold values, and directly calling the PET safety threshold values of the corresponding working conditions during evaluation.
6. The automated driving-oriented highway signalless intersection suitability assessment method of claim 1, wherein: the specific execution rules of the first judging level and the second rechecking judging level are as follows: If the post intrusion time PET is greater than or equal to the corresponding PET safety threshold, directly judging that the current dynamic scene runs safely, and if the corresponding dynamic driving adaptability index is safe; If the post-intrusion time PET is smaller than the corresponding PET safety threshold, further checking and judging through the parking distance ratio PSD, if the PSD is larger than or equal to 1, judging that the current dynamic scene runs safely, and if the PSD is smaller than 1, judging that the current dynamic scene does not meet the safety running requirement, and the corresponding dynamic driving adaptability index is a risk.
7. The method for evaluating the suitability of the automatic-driving-oriented highway signalless intersection according to claim 1 is characterized in that the automatic driving effective detection distance corresponding to the target scene is obtained through matching of a preset automatic driving effective detection distance data set, the data set is constructed through a machine learning method, a mapping relation of automatic driving functional characteristics, intersection geometric parameters, weather environment conditions and effective detection distances is established, and the mapping relation comprises paired data of a plurality of groups of scene elements and the corresponding effective detection distances.
8. The method for evaluating the suitability of an automatic-driving-oriented highway signalless intersection according to claim 1 is characterized in that the automatic driving vehicle-dynamic interaction vehicle collision scene comprises three typical working conditions of straight-going-straight-going collision, left-turning-straight-going collision and right-turning-straight-going collision, and the spatial positions of a vehicle detection point, two vehicle track collision points and a target vehicle effective detection point are determined according to vehicle driving tracks, intersection geometric parameters and effective detection distances aiming at different typical working conditions.
9. An automated driving-oriented highway signalless intersection suitability assessment system for implementing the method of claim 1, comprising: The parameter acquisition module is used for acquiring geometric parameters of the road no-signal intersection to be evaluated, weather environmental conditions, functional characteristics of the target automatic driving vehicle and driving automation level; The reference calculation module is used for matching the perception, braking and takeover characteristic parameters of the corresponding grade based on the driving automation grade of the target vehicle, and differentially calculating the required parking sight distance of the corresponding grade; The static evaluation module is used for acquiring an effective automatic driving detection distance corresponding to a target scene, combining with geometrical parameters of an intersection to obtain an available sight distance, and calculating the sight distance failure probability from the available sight distance and a required parking sight distance through a sight distance function constructed based on a structural reliability theory to obtain a static driving adaptability index; The dynamic evaluation module is used for calculating intrusion time PET after a time dimension conflict index and a space dimension redundancy index parking distance ratio PSD of the conflict scene aiming at the collision scene of the automatic driving vehicle-dynamic interaction vehicle, acquiring a PET safety threshold corresponding to the current conflict working condition, taking a comparison result of the intrusion time PET and the PET safety threshold as a first judgment level, and taking the parking distance ratio PSD as a second rechecking judgment level to obtain a dynamic driving adaptability index.
10. A computer readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the method according to any one of claims 1 to 8.

Description

Automatic driving-oriented highway signalless intersection driving adaptability evaluation method Technical Field The invention belongs to the technical field of road traffic control, and particularly relates to an automatic driving-oriented road signalless intersection driving adaptability evaluation method. Background With rapid iteration and large-scale landing application of the autopilot technology, the running scene of the autopilot vehicle is gradually extended from a closed test park, a high-level structured highway to a suburb trunk highway, a county rural highway and other common highway scenes. The highway signalless intersection is used as an intersection node of different traffic flows, is one of the scenes with the most concentrated traffic conflict and the highest accident risk in the road traffic network, and is also a core scene which the automatic driving vehicle has to overcome from the demonstration operation to the large-scale commercial operation. Currently, the traffic safety, traffic efficiency and environmental adaptability of an automatic driving vehicle at a highway signalless intersection become core research hotspots in the field of intelligent traffic and automatic driving engineering. The existing highway engineering design and traffic safety evaluation system is built on the physiological perception characteristic, psychological response characteristic and operation behavior mode of the traditional manual driving, the core design parameters and the safety evaluation indexes of the system are formulated around the visual perception range, the perception-braking response time delay, the operation control precision and other core characteristics of the human driver, and the essential difference of the perception mechanism, the decision logic and the control characteristic of the automatic driving system and the human driver is not considered from the bottom design logic. The inherent mismatch of the infrastructure design system and the automatic driving operation characteristic directly leads to the frequent problems of mismatch of sight distance supply and demand, insufficient dynamic interaction conflict pre-judgment, mismatch of safety redundancy design and system capacity and the like when the automatic driving vehicle operates at a highway signalless intersection, and causes a large number of traffic safety hazards and efficiency bottlenecks. At present, aiming at the adaptability and safety research of automatic driving in a scene without a signalized intersection, the method is still in a starting and exploring stage. The existing related research focuses on the optimization of an automatic driving traffic control algorithm and a collision avoidance strategy in the intelligent dimension of a bicycle, and is mostly based on the ideal environment and simulation scenes of fixed parameters for verification, and systematic analysis and quantitative evaluation under the coupling action of the characteristics of an automatic driving system, geometrical and physical characteristics of an intersection and a dynamic traffic running environment are lacking. Meanwhile, most of the existing automatic driving road adaptability assessment methods still use a traditional manual driving assessment framework, focus on parking sight distance compliance verification under a static obstacle scene, and the existing researches incorporate dynamic vehicle interaction behavior and multi-vehicle conflict evolution process in a signalless intersection into a adaptability assessment system, so that comprehensive adaptability of an automatic driving vehicle in a real running environment cannot be comprehensively reflected. In the related research of automatic driving safety assessment related to dynamic interaction scenes, the prior art scheme still has the technical defect of multiple dimensions. In the existing evaluation, single safety replacement indexes based on time dimension are adopted, although the indexes can reflect conflict urgency among vehicles to a certain extent, braking space redundancy requirements of different automatic driving systems due to perception-decision-control link differences cannot be quantified, even if the indexes are identical in time conflict, the automatic driving systems with different automation levels and different sensor configurations have obvious differences in actual collision risk and safety coping capacity, and the single indexes cannot effectively distinguish the differences. Meanwhile, the existing safety evaluation method based on the parking sight distance is characterized in that the core applicable scene is an emergency braking working condition under a static obstacle, a dynamic vehicle interaction scene under a signalless intersection yielding rule cannot be adapted, and the automatic driving vehicle is safe to pass when serving as a yielding party, so that the method depends on whether parking can be completed before a conflict point, and further