Automatic driving blind area compensation method based on vehicle-road cooperation

Abstract

The invention relates to the technical field of vehicle-road cooperation communication, in particular to an automatic driving blind area compensation method based on vehicle-road cooperation, which comprises the following steps: acquiring coordinate speeds and blind area boundary points uploaded by a vehicle and an RSU, screening boundaries to construct a closed area, calculating an overlapping ratio, extracting track segments, clustering and screening overlapped segments to generate a track set, and compensating the track set. And calculating the overlap ratio and the blind area exposure value, combining with speed normalization sorting to generate a risk avoiding priority, calculating a trajectory intersection difference and a density judgment direction, and correcting a blind area structure to generate a path pre-scheduling instruction set. According to the method, a closed region is constructed through Euclidean distance screening, an overlapping proportion is calculated in combination with a blind area, a trajectory overlapping section is extracted to analyze vector and distance change, a blind area exposure value and a function level are fused to evaluate threats, a trajectory density sequence is constructed to discriminate a risk direction, and a blind area structure is corrected to generate a pre-scheduling instruction. And the path avoidance adaptability and the cooperation efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-road cooperative communication technology, and in particular to a method for compensating blind spots in autonomous driving based on vehicle-road cooperation. Background Art

[0002] The field of vehicle-road cooperative communication technology encompasses technologies related to information exchange and coordinated control between vehicles and road infrastructure. Its core objective is to achieve efficient information sharing and collaborative operation between road traffic participants and the environment through wireless communications, edge computing, and sensor fusion, thereby improving traffic safety and operational efficiency. This field encompasses subsystems such as vehicle-to-vehicle communication, vehicle-to-road communication, and vehicle-to-cloud communication, with key research areas focusing on multi-source sensor data fusion, network communication protocols, collaborative path planning, and dynamic scheduling and control. It has widespread application in intelligent transportation systems and autonomous driving systems.

[0003] Among them, the blind spot compensation method for autonomous driving based on vehicle-road collaboration refers to a technical solution that uses vehicle-road collaborative communication technology to supplement the information of the vehicle's perceived blind spots during autonomous driving. The key involves obtaining dynamic target information in the blind spot in real time through roadside perception equipment, and transmitting it to the autonomous driving vehicle through low-latency communication. After receiving the supplementary information, the vehicle combines it with its own sensor data for fusion processing to improve the accuracy of cognition of traffic participants in the blind spot area. This method usually relies on a vehicle-road integrated deployment architecture, uses fixed-position road test radars and cameras to obtain perception blind spot data, and combines it with the on-board V2X communication unit to receive remote perception results, thereby compensating for the lack of blind spot information caused by the autonomous driving vehicle due to reasons such as sensor installation position and field of view obstruction.

[0004] Existing technologies generally rely on roadside sensing devices to acquire blind spot information and transmit it directly to vehicle terminals via communication. These sensing data are often based on static area judgments and single-moment perception results, ignoring the dynamic characteristics of vehicle behavior patterns as their trajectories evolve. This results in a lag in identifying risky targets. Blind spot coverage areas are often described using fixed geometric models, which fail to fully capture the complex spatial overlap between multiple vehicles in real-world road conditions, thus impacting the adaptability and matching accuracy of the sensing data. Vehicle priority is often based on functional level or positional relationships, lacking a dynamic ranking mechanism based on multi-factor fusion. This can lead to low-risk targets being prematurely avoided, increasing the system's computational burden. Path scheduling relies heavily on rule-based matching and simple avoidance strategies, making it difficult to adapt to the continuous trajectory intersections and complex interactions in high-density traffic scenarios. This can lead to untimely avoidance strategies, inaccurate path reconstruction, and even scheduling failures in certain emergency scenarios due to insufficient predictions. If multiple vehicles enter the blind spot intersection area at the same time, the existing solutions are unable to complete priority sorting and path correction in a short time, which may easily cause traffic congestion or collision risks, exposing the obvious lack of dynamic perception and path planning linkage processing capabilities. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a blind spot compensation method for autonomous driving based on vehicle-road collaboration.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for compensating blind spots in an autonomous driving system based on vehicle-road collaboration, comprising the following steps:

[0007] S1: Obtain vehicle coordinates, speed, and blind spot boundary points periodically uploaded by the vehicle-side and roadside RSUs, compare the Euclidean distances of multiple boundary points, select boundary point pairs to construct a contour closed area, calculate the area ratio of the overlapping area based on the blind spot area of ​​the vehicle itself, and generate the area overlap parameter;

[0008] S2: Based on the area overlap parameter, a vehicle trajectory point set is obtained and continuous trajectory segments are extracted, the vector angle and distance shortening amplitude between the trajectory points are calculated, the vehicle trajectory point set is temporally clustered, and overlapping trajectory segments are screened to generate a trajectory overlapping segment set;

[0009] S3: Extracting the duration of each trajectory from the set of trajectory overlap segments to calculate the trajectory overlap degree, calculating the blind spot exposure value and functional level of the corresponding vehicle based on the area overlap parameter, constructing a four-factor set based on the vehicle speed and normalizing it, sorting it using the fuzzy hierarchical analysis method, and outputting a risk avoidance priority list;

[0010] S4: Based on the vehicle trajectory point set in the risk avoidance priority list, calculating the difference between the extension direction of the multi-trajectory segment vector and the coordinate of the corresponding intersection point, counting the number of trajectory intersections within a certain period of time to form a trajectory density sequence and determining whether it is monotonically increasing, and generating a trust deflection direction;

[0011] S5: Based on the trust deflection direction, the blind spot direction is corrected, the blind spot structure is reconstructed using the Delaunay triangulation algorithm, and the warning compensation area is marked. The reconstructed blind spot structure and the current vehicle coordinates are combined to construct a path adjustment instruction, and a path pre-scheduling instruction set is generated in combination with a preset vehicle spacing extension value.

[0012] As a further solution of the present invention, the area overlap parameters are specifically the overlapping area value, the boundary closure index, and the boundary point pair set; the trajectory overlap segment set includes the overlapping trajectory number, the overlapping start and end time period, and the trajectory vector difference; the risk avoidance priority list is specifically the vehicle number sequence, the normalization factor value, and the priority level label; the trusted deflection direction includes the dominant extension angle, the trajectory intersection growth trend, and the offset direction deviation value; the path pre-scheduling instruction set specifically refers to the path correction vector, the vehicle spacing compensation value, and the blind spot number matching index.

[0013] As a further solution of the present invention, the specific steps of S1 include:

[0014] S101: Obtain vehicle coordinate information, speed, and blind spot boundary point sets periodically uploaded by the vehicle-side and roadside RSUs, pair the vehicle position at the same moment with the corresponding blind spot boundary points, calculate the Euclidean distance of each pair of boundary points, and select boundary point combinations within the boundary convergence interval to generate boundary point distance value screening results;

[0015] S102: Based on the boundary point distance value screening result, boundary point combinations are sequentially connected to construct contour edge curves and perform contour closing operations, coordinates of closed area boundaries are regularized, and area overlap calculations are performed using the vehicle body blind spot area parameters to generate the area intersection area;

[0016] S103: Obtaining the blind spot area of ​​the vehicle body based on the area of ​​the region boundary, calculating the degree of overlap between the closed contour region and the vehicle blind spot, and generating a region overlap parameter.

[0017] As a further solution of the present invention, the boundary convergence interval is adaptively adjusted and set with the vehicle's current position as the center as the vehicle speed changes.

[0018] As a further solution of the present invention, the specific steps of S2 include:

[0019] S201: Based on the area overlap parameter, all trajectory points recorded by the current vehicle in the overlap area are obtained and a trajectory point sequence is constructed. The coordinate values ​​between adjacent trajectory points are differentially calculated, and then the vector angle change value and distance shortening rate between all adjacent points are calculated to generate a trajectory change amount set.

[0020] S202: Calling the trajectory change amount set, based on the synchronous fluctuation characteristics between the vector angle change value and the distance shortening rate, selecting trajectory point segments with consistent vector directions and continuously shortening distances, and aggregating and grouping them to generate a trajectory continuous segment set;

[0021] S203: Based on the set of continuous trajectory segments, overlapping segment detection is performed on multiple trajectory segments. Double constraint judgment is performed using the overlapping range of trajectory segments in spatial position and the overlapping interval of timestamps. Coincident trajectory segments are screened and integrated to obtain a set of overlapping trajectory segments.

[0022] As a further solution of the present invention, the specific steps of S3 include:

[0023] S301: Based on the set of trajectory overlap segments, extract the start timestamps and end timestamps of multiple trajectory segments, calculate the corresponding duration values, and calculate the ratio of the total duration in the current path, measure the degree of trajectory overlap in the form of time proportion, and obtain the trajectory overlap degree;

[0024] S302: The trajectory overlap and the area overlap parameter are called, and combined with the speed of the corresponding vehicle, the blind spot position coverage rate, and the functional level scoring standard, a four-factor set including overlap, blind spot exposure rate, functional level value, and speed is constructed, and interval normalization is performed to obtain a normalized indicator set;

[0025] S303: Based on the normalized indicator set, the membership degree is calculated for each of the four factors: trajectory overlap, blind spot exposure rate, functional level value, and speed. The evaluation values ​​under the four factor dimensions are calculated using the fuzzy analytic hierarchy process, and the vehicles are sorted to generate a sorted list.

[0026] As a further solution of the present invention, the specific steps of S4 include:

[0027] S401: Based on the vehicle trajectory point set in the risk avoidance priority list, extract the end vector direction of multiple trajectory segments and calculate the difference between the end vector direction and the coordinates of the trajectory intersection point, and measure the offset distance and angular error between the vector extension direction and the intersection point in a two-dimensional coordinate system to obtain a trajectory offset difference sequence;

[0028] S402: Calling the trajectory offset difference sequence, extracting the cumulative number of intersections within a fixed time period, counting the number of trajectory intersections in time series and constructing a time series of intersections, using a continuous difference method to determine whether the trend of number changes in adjacent intervals is non-negative, and obtaining an increasing trend in trajectory density;

[0029] S403: According to the increasing trend of trajectory density, a dominant direction with a good trajectory extension direction deviation difference in a time segment with a continuous increasing feature is identified as a key direction of the group movement trend, and a trusted deflection direction is obtained.

[0030] As a further solution of the present invention, the specific steps of S5 include:

[0031] S501: adjusting the current blind spot boundary direction data based on the trusted deflection direction, rotating all boundary point coordinates according to the deflection direction, calculating the blind spot region composition relationship using a Delaunay triangulation algorithm, marking the corresponding warning compensation region, and obtaining reconstructed region structure parameters;

[0032] S502: Calling the reconstructed region structure parameters and the current vehicle coordinate set, determining whether the vehicle's current position falls within the marked compensation region, calculating the path offset between the vehicle and the boundary point in the reconstructed structure, and obtaining path adjustment instruction data;

[0033] S503: Based on the path adjustment instruction and the preset vehicle spacing extension value, the adjustment trajectory boundary within the vehicle step length is calculated, the path segment is offset outward by adding the vehicle spacing extension length, a scheduling mapping table indexed by the vehicle number is established, and a path pre-scheduling instruction set is obtained.

[0034] As a further solution of the present invention, the preset vehicle spacing extension value is set by the minimum safe braking distance, average following vehicle interval time and front and rear vehicle speed difference of vehicles under multiple traffic density levels in road operation data.

[0035] Compared with the prior art, the advantages and positive effects of the present invention are:

[0036] In this paper, closed regions are constructed by filtering vehicle contour boundary points using Euclidean distances. The overlap ratio is then calculated based on the vehicle blind spot area, improving the accuracy of spatial overlap measurements and forming a more detailed regional overlap determination mechanism. Based on this overlap result, a trajectory point set is further extracted. Time series clustering is achieved through analysis of vector angle and distance changes. The selected trajectory overlap segments are then combined with the duration of the overlap segments and the regional overlap parameter. The calculated blind spot exposure value and functional level form a more hierarchical target threat assessment system. A multi-factor normalization mechanism, combined with speed factors, is introduced and sorted using fuzzy hierarchical analysis, improving the accuracy of risk response and the rationality of priority scheduling. Track intersection statistics are combined with vector direction differences to construct a trajectory density sequence, which automatically identifies potential risk concentration directions through trend analysis. The blind spot structure is modified by trusting the deflection direction, and a more accurate blind spot topology is constructed using Delaunay triangulation. The structured results are then mapped into path adjustment instructions, and a pre-schedule set is generated based on vehicle spacing parameters. This significantly improves the foresight and adaptability of path avoidance strategies in autonomous driving scenarios. In the overall solution, multi-stage trajectory screening and spatial reconstruction jointly construct a perception compensation mechanism that integrates multi-factor fusion judgment and dynamic blind spot correction, forming a more efficient collaborative closed loop between perception cognition and path planning, thereby significantly improving traffic safety warning capabilities and the overall operating efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the main steps of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0039] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0040] See also Figure 1 The present invention provides a technical solution: a method for compensating blind spots in an autonomous driving system based on vehicle-road collaboration, comprising the following steps:

[0041] S1: Obtain vehicle coordinates, speed, and blind spot boundary points periodically uploaded by the vehicle-side and roadside RSUs, compare the Euclidean distances of multiple boundary points, select boundary point pairs to construct a contour closed area, calculate the area ratio of the overlapping area based on the blind spot area of ​​the vehicle itself, and generate the area overlap parameter;

[0042] S2: Based on the regional overlap parameters, the vehicle trajectory point set is obtained and continuous trajectory segments are extracted. The vector angle and distance shortening amplitude between the trajectory points are calculated. The vehicle trajectory point set is temporally clustered and overlapping trajectory segments are screened to generate a set of trajectory overlapping segments.

[0043] S3: Extract the duration of each trajectory from the set of overlapping segments to calculate the trajectory overlap. The blind spot exposure value and functional level of the corresponding vehicle are calculated based on the area overlap parameter. A four-factor set is constructed and normalized based on the vehicle speed. The four-factor set is sorted using the fuzzy analytic hierarchy process to output a risk avoidance priority list.

[0044] S4: Based on the vehicle trajectory point set in the risk avoidance priority list, calculate the difference between the extension direction of the multi-trajectory vector and the coordinate of the corresponding intersection point, count the number of trajectory intersections within a certain period of time to form a trajectory density sequence, and determine whether it is monotonically increasing, and generate the trust deflection direction;

[0045] S5: Correct the blind spot direction based on the trust deflection direction, reconstruct the blind spot structure through the Delaunay triangulation algorithm, and mark the warning compensation area. Build the path adjustment instructions based on the reconstructed blind spot structure and the current vehicle coordinates, and generate the path pre-scheduling instruction set by combining the preset vehicle spacing extension value.

[0046] The regional overlap parameters specifically include the overlapping area value, boundary closure index, and boundary point pair set. The trajectory overlap segment set includes the overlapping trajectory number, overlapping start and end time period, and trajectory vector difference. The risk avoidance priority list specifically includes the vehicle number sequence, normalization factor value, and priority level label. The trust deflection direction includes the dominant extension angle, trajectory intersection growth trend, and offset direction deviation value. The path pre-scheduling instruction set specifically refers to the path correction vector, vehicle spacing compensation value, and blind spot number matching index.

[0047] See also Figure 1 , the specific steps for obtaining S1 are:

[0048] S101: Obtain vehicle coordinate information, speed, and blind spot boundary point sets periodically uploaded by the vehicle-side and roadside RSUs, pair the vehicle position at the same moment with the corresponding blind spot boundary points, calculate the Euclidean distance of each pair of boundary points, and select boundary point combinations within the boundary convergence interval to generate boundary point distance value screening results;

[0049] Obtain vehicle coordinate information (longitude 118.798°E, latitude 32.062°N), speed (72 km / h), and blind spot boundary point set (including left front A (x1, y1) = (5.2, 1.8), right front B (6.3, 1.7), left rear C (4.9, -2.1), right rear D (6.1, -2.0) meter coordinate system data) from the vehicle-side and roadside RSU. Pair the vehicle position at the timestamp 07-07, 14:30:00 with the blind spot points and calculate the multi-point Euclidean distance: d B =6.48m,d C =5.38m,d D =6.40m,d A d B d C and d D Represent the Euclidean distances of point A, point B, point C, and point D respectively. Set the convergence interval threshold θ = 5.0 ± 0.5m, and select point C (5.38m) and point A (5.51m) to enter the screening results. The boundary point distance value screening result table is shown in Table 1, which displays the coordinates and distance data of the valid boundary point combination. The screening process ensures that only boundary points that meet the spatial constraints are used in the subsequent contour construction.

[0050] Table 1 Boundary point distance value screening results

[0051] Boundary Point X coordinate (m) Y coordinate (m) Distance (m) Filter results A 5.2 1.8 5.51 reserve C 4.9 -2.1 5.38 reserve

[0052] S102: Filtering boundary point combinations in the boundary point distance value result, sequentially connecting them to construct a contour edge curve and performing a contour closing operation, regularizing the coordinates of the closed area boundary, and using the vehicle body blind spot area parameter to perform area overlap calculation to generate the area intersection area;

[0053] Take points A (5.2, 1.8) and C (4.9, -2.1) from the screening results, calculate the intermediate interpolation point E (5.05, -0.15), connect AEC to form a broken line trajectory, add the symmetrical point F (5.35, 1.75) to complete the contour closure, and regularize the vertex coordinates of the closed polygon (retain one decimal place) to obtain the vehicle body blind spot area parameter S0 = 22.6m 2 , when calculating the area of ​​the overlapping region, substitute the coordinates of the contour vertices into the formula: Substituting A(5.2, 1.8), E(5.05, -0), C(4.9, -2), F(5.35, 1.75) into the equation, we get S=8.7m 2 , S0 represents the base area of ​​the vehicle's blind spot, and S represents the area of ​​the overlapping area, which is measured by the vehicle's onboard sensors. i ,yi Represents the coordinates of the contour vertices (e.g. point A x1 = 5.2, y1 = 1.8), and n = 4 represents the number of polygon vertices. The overlap calculation process accurately measures the area of ​​the regional intersection through the sequential connection of coordinate points and area integration.

[0054] S103: Obtaining the vehicle's blind spot area based on the area of ​​the region boundary, calculating the degree of overlap between the closed contour area and the vehicle's blind spot, and generating a region overlap parameter;

[0055] Based on overlapping area S = 8.7m 2 , calculate the blind area coverage S0 represents the base area of ​​the vehicle's blind spot, and sets the safety threshold λ crit = 60%, the current coverage exceeds the critical value by 1.5 percentage points, generating the overlap parameter α = λ - λ crit =1.5%. The parameter is calculated by the difference between the actual measurement data and the preset safety standard, reflecting the safety margin of the blind spot monitoring system. When α>0, the early warning mechanism is triggered, including when a single calculation obtains λ=58%, it generates α=-2.0%. At this time, the system automatically initiates the emergency response protocol.

[0056] See also Figure 1 , the specific steps for obtaining S2 are:

[0057] S201: Based on the area overlap parameter, all trajectory points recorded by the current vehicle in the overlap area are obtained and a trajectory point sequence is constructed. The coordinate values ​​between adjacent trajectory points are differentially calculated, and then the vector angle change value and distance shortening rate between all adjacent points are calculated to generate a trajectory change value set.

[0058] Based on the area overlap parameter α = 1.5%, obtain the five trajectory points P1 (10.2, 5.3), P2 (10.5, 5.1), P3 (10.8, 4.9), P4 (11.0, 4.7), and P5 (11.3, 4.5) within the time stamp of 14:30:00-14:30:10, and calculate the coordinate difference of adjacent points: Δx = 0.3m, Δy = -0.2m of P1-P2, and the vector angle change value Distance reduction rate (The benchmark distance of 1.0m is the mean value of the previous cycle). The data of all adjacent points are calculated to generate a set of trajectory changes, as shown in Table 2. The trajectory change parameter table fully records the coordinate change, angle change and shortening rate data of multiple points.

[0059] Table 2 Trajectory change parameters

[0060] Point pair X difference Y difference Angle changes Shortening rate P1-P2 0.3 -0.2 -33.7 36.1 P2-P3 0.3 -0.2 -33.7 36.1

[0061] S202: Calling the trajectory change set, based on the synchronous fluctuation characteristics between the vector angle change value and the distance shortening rate, selecting trajectory point segments with consistent vector directions and continuously shortening distances, and aggregating and grouping them to generate a trajectory continuous segment set;

[0062] Set the synchronous fluctuation threshold to angle change deviation ±5° and shortening rate deviation ±5%, select the P1-P2-P3 segment (angle change -33.7°±1°, shortening rate 36.1%±0.5%), and calculate the average vector direction angle of the three points The distance shortening rate sequence is 36.1%, 36.1%, and 36.0%, which is determined to be in the same direction and continuously shortened. It is aggregated into trajectory segment G1 (time stamp 14:30:00-14:30:06). Similarly, segment P4-P5 is excluded due to the sudden change of angle +15°, generating a continuous segment set including G1.

[0063] S203: Based on the continuous trajectory segment set, overlapping segment detection is performed on multiple trajectory segments. The overlapping range of the trajectory segments in spatial position and the overlapping interval of the timestamp are used to perform dual constraint judgment, and the overlapping trajectory segments are screened and integrated to obtain the trajectory overlapping segment set;

[0064] Overlap detection was performed on G1 (spatial range X: 10.2-10.8m, time 14:30:00-14:30:06) and the newly added segment G2 (X: 10.5-11.0m, time 14:30:03-14:30:07), and the spatial overlap area X: 10.5-10.8m (accounting for 50%) and the time overlap interval 14:30:03-14:30:06 (3 seconds) were calculated. The dual constraint thresholds were set as spatial overlap ≥ 30% and time overlap ≥ 2 seconds. After the conditions were met, it was integrated into the overlapping segment H1 (X: 10.2-11.0m, time 14:30:00-14:30:07), and the intersection calculation was performed by comparing the coordinate extreme values ​​10.5 and 10.8 with the timestamp.

[0065] See also Figure 1 , the specific steps for obtaining S3 are:

[0066] S301: Based on the set of trajectory overlap segments, extract the start and end timestamps of multiple trajectory segments, calculate the corresponding duration values, and calculate the ratio of the total duration in the current path. Measure the degree of trajectory overlap in the form of time percentage to obtain the trajectory overlap degree;

[0067] Based on the trajectory overlap segments H1 (time stamp 14:30:00-14:30:07) and H2 (14:30:05-14:30:10), the duration T1 = 7s, T2 = 5s are extracted, and the total duration T total =10s, calculate the overlap The overlap time T overlap =2s is obtained by calculating the intersection of timestamps. When there is a third segment H3 (contributing 3 seconds without overlap in the total duration of 15 seconds), The calculation process quantifies the degree of trajectory overlap by comparing the numerical values ​​of the timestamps (including the 2-second overlap between the end time of H1 14:30:07 and the start time of H2 14:30:05).

[0068] Table 3 Trajectory coincidence calculation table

[0069] trajectory segment Start time End Time Duration Overlap time H1 14：30：00 14：30：07 7 2 H2 14：30：05 14：30：10 5 2

[0070] S302: Calling trajectory overlap and area overlap parameters, combined with the corresponding vehicle speed, blind spot position coverage, and function level scoring criteria, constructing a four-factor set including overlap, blind spot exposure rate, function level value, and speed, and performing interval normalization processing to obtain a normalized indicator set;

[0071] Take vehicle V1 data: speed v = 60km / h, blind spot exposure rate 98.5%, functional level value L = 4 (graded according to ISO26262 standard), construct a four-factor set {γ = 100%, β = 98.5%, L = 4, v = 60}, and use the extreme value method for normalization: speed term (Assuming the maximum speed is 120km / h), blind spot exposure rate (Benchmark interval 90%-100%), functional level (Level 1-5), trajectory overlap γ′ = 1.0, generating a normalized set {0.5, 0.85, 0.75, 1.0}.

[0072] S303: Based on the normalized indicator set, the membership degree is calculated for each of the four factors: trajectory overlap, blind spot exposure rate, functional level value, and speed. The evaluation values ​​under the four factor dimensions are calculated using the fuzzy analytic hierarchy process, and the vehicles are ranked to generate a ranked list.

[0073] Set target value c k ={1.0, 1.0, 1.0, 0.8} (k = 1-4 corresponds to four factors), allowing fluctuation b k ={0.2, 0.15, 0.25, 0.3}, calculate the V1 membership and substitute it into the formula Calculate trajectory overlap Blind spot exposure rate Functional level speed a ik Represents the normalized value of the i-th vehicle under the k-th factor, calculated by the extreme value method Trajectory overlap a i1 ∈[0, 1], obtained by calculation, blind area exposure rate a i2 ∈[0, 1], transformation-function level, r ik represents the membership degree of the i-th vehicle under the k-th factor, reflecting the degree to which the indicator is close to the target value, a i3 ∈[0, 1], according to IS26262 hierarchical quantization-speed, a i4 ∈

[0074] [0, 1], c k The target value setting is based on c1=1.0, which means the trajectory is completely overlapped; c2=1.0, which means the blind spot is zero exposed; c3=1.0, which means the highest functional level 5 corresponds to 1.0; c4=0.8, which means the economic speed is 96km / h; b k b1 = 0.2 represents the allowed fluctuation range, b2 = 0.15 represents the allowed fluctuation of ±15% in blind spot exposure rate, b3 = 0.25 represents the allowed fluctuation of 1.25 levels in function level, b4 = 0.3 represents the allowed fluctuation of ±36 km / h in speed, and the evaluation value The weights w = 0.4, 0.3, 0.2, 0.1, so R = 0.4×1.0+0.3×0.5+0.2×0.5+0.1×0.5=0.4+0.15+0.1+0.05=0.7. After calculation for 10 vehicles, a sorted list is generated.

[0075] See also Figure 1 , the specific steps for obtaining S4 are:

[0076] S401: Based on the vehicle trajectory point set in the risk avoidance priority list, extract the end vector direction of multiple trajectory segments and calculate the difference between the end vector direction and the coordinates of the trajectory intersection point. Measure the offset distance and angular error between the vector extension direction and the intersection point in the two-dimensional coordinate system to obtain a trajectory offset difference sequence.

[0077] Based on the trajectory data of the first three vehicles V1-V3 in the priority list, extract the coordinates of the last two points of the multi-trajectory segment V1: P4 (11.0, 4.7) and P5 (11.3, 4.5), and calculate the end vector direction At the same time, the offset distance of the intersection point J (12.1, 3.9) is calculated as Angle error Δθ=|-33.7°-(-26.6°)|=7.1°(J point direction ), generate an offset sequence, as shown in Table 4. The trajectory offset parameter table, Table 4 fully records the spatial deviation data of multiple vehicle terminal vectors and intersection points.

[0078] Table 4 Track offset parameters

[0079]

[0080] S402: Calling the trajectory offset difference sequence, extracting the cumulative number of intersections within a fixed time period, counting the number of trajectory intersections in time series and constructing a time series of intersections, using the continuous difference method to determine whether the trend of number changes in adjacent intervals is non-negative, and obtaining the increasing trend of trajectory density;

[0081] Count the number of intersections between 14:30:00 and 14:30:30: [8, 12, 15, 18, 22, 25] and calculate the continuous difference value Δ n We get [4, 3, 3, 4, 3], and set the non-negative trend threshold to be a difference value ≥ 0 for more than 3 cycles. In the example, intervals 1-5 all meet Δ n ≥0, it is judged as a continuous increasing trend. When Δ=-2 appears in a certain interval, it needs to be re-accumulated, which is achieved by counting the timestamp groups (including 12 intersections detected in the interval 14:30:05-14:30:10) and comparing them in sequence.

[0082] S403: Based on the increasing trend of trajectory density, the dominant direction with the best trajectory extension direction deviation difference in the time segment with the continuous increasing feature is identified as the key direction of the group movement trend, and the trusted deflection direction is obtained;

[0083] Calculate the V1-V3 dominant direction: V1 average deviation in the continuous increasing interval between 14:30:10 and 14:30:25 Angular error V2 corresponds to 1.5m, 18.4°, V3 corresponds to 0.8m, 5.2°, and the optimal threshold d is set. max =1.2m, Δθ max =10°, select V3 as the optimal direction (0.8m<1.2m and 5.2°<10°), and obtain the trusted deflection direction θ trust =-28.5°±2.0°, where the angle is calculated by weighted averaging the preferred vehicle trajectory direction (V3 weight 0.6, V1 weight 0.3).

[0084] See also Figure 1 , the specific steps for obtaining S5 are:

[0085] S501: Adjust the current blind spot boundary direction data based on the trusted deflection direction, rotate the coordinates of all boundary points according to the deflection direction, calculate the blind spot area composition relationship using the Delaunay triangulation algorithm, mark the corresponding warning compensation area, and obtain the reconstructed area structure parameters;

[0086] Based on the trust deflection direction θ trust=-28.5°, rotate the blind spot boundary point B1 (10.2, 5.3) to obtain the new coordinates B′1 (10.2cosθ-5.3sinθ, 10.2sinθ+5.3cosθ) = (9.8, 4.1), perform Delaunay decomposition on the 8 rotated boundary points, generate 12 triangles, and calculate the area of ​​the triangle. The vertices of triangle t1 are (9.8, 4.1), (10.5, 3.9), and (11.0, 3.5), and the area is A n Represents the total area of ​​the nth blind spot after reconstruction, which is obtained by accumulating the areas of all triangles. n represents the number of triangles generated by the subdivision, which is determined by the Delaunay empty circle criterion (the circumscribed circle of three points does not include other points), (x kt ,y kt ): Coordinates of the k-th vertex of the t-th triangle, rotation calculation: Among them, θ = -28.5° is the trust deflection direction. The parameter assignment example is calculated by rotating the boundary point B1 (10.2, 5.3), x′1 = 10.2×cos(-28.5°)-5.3×sin(-28.5°) = 9.8m, y′1 = 10.2×sin(-28.5°)+5.3×cos(-28.5°) = 4.1m, select the vertices of triangle t2: -(11.0, 3.5)-(11.5, 3.2)-(10.5, 3.9), area calculation, Corrected to the measured reasonable value, the threshold is set to the safe distance d safe =0.5m Setting basis: According to ISO15622 standard, the 0.5-second braking distance at a vehicle speed of 30km / h (8.33m / s) is 8.33×0.5=4.17m. The compensation area margin is 12% of the main braking distance (4.17×0.12=0.5m), and the total subdivision area A1=8.3m 2 Compared with the original blind area of ​​22.6m 2 The reduction was 63.3%, indicating that the compensation area effectively reduced the monitoring scope.

[0087] S502: Calling the reconstructed region structure parameters and the current vehicle coordinate set to determine whether the vehicle's current position falls within the marked compensation region, calculating the path offset between the vehicle and the boundary points in the reconstructed structure, and obtaining path adjustment instruction data;

[0088] The current position of vehicle V4 (10.8, 4.0) is calculated to offset the boundary point of the compensation area (10.5, 3.9). Set safety threshold d safe=0.5m, because 0.32<0.5, a 15° turn left instruction is generated. When the offset between the vehicle V5 position (11.2, 3.3) and the boundary point (11.5, 3.2) is 0.36m<0.5m, a 20% deceleration instruction is generated.

[0089] S503: Based on the path adjustment instruction and the preset vehicle spacing extension value, the adjustment trajectory boundary within the vehicle step length is calculated, the path segment is offset outward by adding the vehicle spacing extension length, a scheduling mapping table indexed by the vehicle number is established, and a path pre-scheduling instruction set is obtained;

[0090] When setting traffic density level 2 (20-40 vehicles / km), the extension length ΔL = 1.2m × (1 + 0.1 × 2) = 1.44m. The V4 trajectory point (10.8, 4.0) is normal offset to generate the new coordinates (10.8 + 1.44cos(90° - 28.5°), 4.0 + 1.44sin(90° - 28.5°)) = (11.1, 4.3). The dispatch mapping table is generated:

[0091] Table 5 Vehicle dispatch mapping table

[0092] Vehicle number Original coordinates Expanded coordinates Adjust direction Extended length V4 (10.8，4.0) (11.1，4.3) 61.5° 1.44m

[0093] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. The blind spot compensation method for autonomous driving based on vehicle-road collaboration is characterized by: The following steps are involved: S1: Obtain vehicle coordinates, speed, and blind spot boundary points periodically uploaded by the vehicle-side and roadside RSUs, compare the Euclidean distances of multiple boundary points, select boundary point pairs to construct a contour closed area, calculate the area ratio of the overlapping area based on the blind spot area of ​​the vehicle itself, and generate the area overlap parameter; S2: Based on the area overlap parameter, a vehicle trajectory point set is obtained and continuous trajectory segments are extracted, the vector angle and distance shortening amplitude between the trajectory points are calculated, the vehicle trajectory point set is temporally clustered, and overlapping trajectory segments are screened to generate a trajectory overlapping segment set; S3: Extracting the duration of each trajectory from the set of trajectory overlap segments to calculate the trajectory overlap degree, calculating the blind spot exposure value and functional level of the corresponding vehicle based on the area overlap parameter, constructing a four-factor set based on the vehicle speed and normalizing it, sorting it using the fuzzy hierarchical analysis method, and outputting a risk avoidance priority list; S4: Based on the vehicle trajectory point set in the risk avoidance priority list, calculate the difference between the extension direction of the multi-trajectory segment vector and the coordinate of the corresponding intersection point, count the number of trajectory intersections within a certain period of time to form a trajectory density sequence, and determine whether it is monotonically increasing, and generate a trust deflection direction.

2. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 1, characterized in that: The area overlap parameters specifically include the overlapping area value, boundary closure index, and boundary point pair set. The trajectory overlap segment set includes the overlapping trajectory number, overlapping start and end time period, and trajectory vector difference. The risk avoidance priority list specifically includes the vehicle number sequence, normalization factor value, and priority level label. The trust deflection direction includes the dominant extension angle, trajectory intersection growth trend, and offset direction deviation value.

3. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 1, characterized in that: The specific steps of S1 include: S101: Obtain vehicle coordinate information, speed, and blind spot boundary point sets periodically uploaded by the vehicle-side and roadside RSUs, pair the vehicle position at the same moment with the corresponding blind spot boundary points, calculate the Euclidean distance of each pair of boundary points, and select boundary point combinations within the boundary convergence interval to generate boundary point distance value screening results; S102: Based on the boundary point distance value screening result, boundary point combinations are sequentially connected to construct contour edge curves and perform contour closing operations, coordinates of closed area boundaries are regularized, and area overlap calculations are performed using the vehicle body blind spot area parameters to generate the area intersection area; S103: Obtaining the blind spot area of ​​the vehicle body based on the area of ​​the region boundary, calculating the degree of overlap between the closed contour region and the vehicle blind spot, and generating a region overlap parameter.

4. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 3, characterized in that: The boundary convergence interval is adaptively adjusted and set with the current position of the vehicle as the center as the vehicle speed changes.

5. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 1, characterized in that: The specific steps of S2 include: S201: Based on the area overlap parameter, all trajectory points recorded by the current vehicle in the overlap area are obtained and a trajectory point sequence is constructed. The coordinate values ​​between adjacent trajectory points are differentially calculated, and then the vector angle change value and distance shortening rate between all adjacent points are calculated to generate a trajectory change amount set. S202: Calling the trajectory change amount set, based on the synchronous fluctuation characteristics between the vector angle change value and the distance shortening rate, selecting trajectory point segments with consistent vector directions and continuously shortening distances, and aggregating and grouping them to generate a trajectory continuous segment set; S203: Based on the set of continuous trajectory segments, overlapping segment detection is performed on multiple trajectory segments. Double constraint judgment is performed using the overlapping range of trajectory segments in spatial position and the overlapping interval of timestamps. Coincident trajectory segments are screened and integrated to obtain a set of overlapping trajectory segments.

6. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 1, characterized in that: The specific steps of S3 include: S301: Based on the set of trajectory overlap segments, extract the start timestamps and end timestamps of multiple trajectory segments, calculate the corresponding duration values, and calculate the ratio of the total duration in the current path, measure the degree of trajectory overlap in the form of time proportion, and obtain the trajectory overlap degree; S302: The trajectory overlap and the area overlap parameter are called, and combined with the speed of the corresponding vehicle, the blind spot position coverage rate, and the functional level scoring standard, a four-factor set including overlap, blind spot exposure rate, functional level value, and speed is constructed, and interval normalization is performed to obtain a normalized indicator set; S303: Based on the normalized indicator set, the membership degree is calculated for each of the four factors: trajectory overlap, blind spot exposure rate, functional level value, and speed. The evaluation values ​​under the four factor dimensions are calculated using the fuzzy analytic hierarchy process, and the vehicles are sorted to generate a sorted list.

7. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 1, characterized in that: The specific steps of S4 include: S401: Based on the vehicle trajectory point set in the risk avoidance priority list, extract the end vector direction of multiple trajectory segments and calculate the difference between the end vector direction and the coordinates of the trajectory intersection point, and measure the offset distance and angular error between the vector extension direction and the intersection point in a two-dimensional coordinate system to obtain a trajectory offset difference sequence; S402: Calling the trajectory offset difference sequence, extracting the cumulative number of intersections within a fixed time period, counting the number of trajectory intersections in time series and constructing a time series of intersections, using a continuous difference method to determine whether the trend of number changes in adjacent intervals is non-negative, and obtaining an increasing trend in trajectory density; S403: According to the increasing trend of trajectory density, a dominant direction with a good trajectory extension direction deviation difference in a time segment with a continuous increasing feature is identified as a key direction of the group movement trend, and a trusted deflection direction is obtained.

8. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 1, characterized in that: The method further comprises: S5: Correcting the blind spot direction based on the trust deflection direction, reconstructing the blind spot structure using a Delaunay triangulation algorithm, and marking the warning compensation area. Path adjustment instructions are constructed using the reconstructed blind spot structure and the current vehicle coordinates. A path pre-scheduling instruction set is generated in combination with a preset vehicle spacing extension value. The path pre-scheduling instruction set specifically refers to a path correction vector, a vehicle spacing compensation value, and a blind spot number matching index.

9. The blind spot compensation method for autonomous driving based on vehicle-road collaboration according to claim 1, characterized in that: The specific steps of S5 include: S501: adjusting the current blind spot boundary direction data based on the trusted deflection direction, rotating all boundary point coordinates according to the deflection direction, calculating the blind spot region composition relationship using a Delaunay triangulation algorithm, marking the corresponding warning compensation region, and obtaining reconstructed region structure parameters; S502: Calling the reconstructed region structure parameters and the current vehicle coordinate set, determining whether the vehicle's current position falls within the marked compensation region, calculating the path offset between the vehicle and the boundary point in the reconstructed structure, and obtaining path adjustment instruction data; S503: Based on the path adjustment instruction and the preset vehicle spacing extension value, the adjustment trajectory boundary within the vehicle step length is calculated, the path segment is offset outward by adding the vehicle spacing extension length, a scheduling mapping table indexed by the vehicle number is established, and a path pre-scheduling instruction set is obtained.

10. The method for compensating blind spots in autonomous driving based on vehicle-road collaboration according to claim 9, characterized in that: The preset vehicle spacing extension value is set by the minimum safe braking distance, average following vehicle interval time and front and rear vehicle speed difference of vehicles under multiple traffic density levels in road operation data.

Images