A method and system for early warning of vehicle collision at a curved intersection

By calculating vehicle state disorder factors, collision risk factors and abnormal factors, combined with the extended Kalman filtering algorithm, predicting vehicle collision coefficients and state estimates, the accuracy problem of traditional methods when distinguishing mutation data from noise data is solved, and the curve is improved Accuracy of traffic collision warning at intersections.

CN117711207BActive Publication Date: 2024-11-15BEIJING SINOITS TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311580944.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-24
Publication Date
2024-11-15
Estimated Expiration
2043-11-24

AI Technical Summary

Technical Problem

When traditional extended Kalman filtering processes vehicle operation data, it is difficult to effectively distinguish between mutation data and noise data, resulting in a reduced accuracy of vehicle collision warning at curved intersections.

Method used

By collecting vehicle operation data sequences of target vehicles and historical vehicles, the vehicle state disorder factor, collision risk factor, abnormal factor and relative noise factor are calculated, combined with the extended Kalman filtering algorithm, the vehicle collision coefficient and state estimate value are predicted, and early warning is made.

Benefits of technology

It improves the accuracy of vehicle collision warning, reduces the possibility that normal data is mistaken for noise, and enhances the noise removal effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117711207B_ABST
    Figure CN117711207B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of intelligent transportation systems, and specifically to a method and system for warning of vehicle collision at a curved intersection, comprising: collecting a number of vehicle operation data sequences and a degree of environmental interference at a curved intersection; obtaining a vehicle state disorder factor according to the vehicle operation data sequence; obtaining a collision risk factor according to the vehicle state disorder factor; obtaining an abnormal factor according to the degree of environmental interference at the curved intersection; obtaining a number of control operation data; obtaining a relative noise factor according to the control operation data and the collision risk factor; obtaining a predicted collision coefficient according to the relative noise factor and the abnormal factor; obtaining a vehicle collision state estimation value according to the predicted collision coefficient; and issuing a warning according to the vehicle collision state estimation value. The present invention reduces the possibility of normal vehicle operation data being mistaken for noise, and improves the accuracy of vehicle collision warning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation systems, and in particular to a method and system for early warning of vehicle collision at a curved intersection. Background Art

[0002] Curved intersections are important nodes in the traffic network and are high-incidence locations for traffic accidents. In order to reduce potential safety hazards when vehicles are driving at curved intersections and make vehicle traffic at curved intersections more intelligent, vehicle operation data from on-board sensors in vehicles are usually transmitted to the roadside unit (RSU) at the curved intersection. The roadside unit (RSU) predicts the probability of a vehicle collision and issues corresponding collision warnings.

[0003] Traditional methods usually use extended Kalman filtering for collision warning on vehicle operation data, but vehicle operation data will change rapidly due to actual driving conditions, and there will be a lot of sudden changes in vehicle operation data. The traditional extended Kalman filtering will mistake some normal vehicle operation data for noise data, reducing the accuracy of vehicle collision warning. Summary of the invention

[0004] The present invention provides a method and system for warning of vehicle collision at a curved intersection to solve the existing problem that vehicle operation data may change rapidly due to actual driving conditions, and there may be a lot of sudden changes in vehicle operation data. Traditional extended Kalman filtering cannot effectively distinguish the sudden changes in vehicle operation data from noise, thereby reducing the accuracy of vehicle collision warning.

[0005] A method and system for early warning of vehicle collision at a curved intersection of the present invention adopts the following technical solutions:

[0006] An embodiment of the present invention provides a method for early warning of vehicle collision at a curved intersection, the method comprising the following steps:

[0007] Collecting a plurality of vehicle operation data sequences of a target vehicle, a plurality of vehicle operation data sequences of a plurality of historical vehicles, and a degree of interference of a curved intersection environment, wherein the vehicle operation data sequence includes a plurality of vehicle operation data;

[0008] According to the numerical difference of the vehicle operation data between the vehicle operation data sequences of the target vehicle and the historical vehicle, a vehicle state disorder factor of each vehicle operation data sequence is obtained, wherein the vehicle state disorder factor is used to describe the probability that the vehicle operation data in the vehicle operation data sequence is mistakenly identified as noise; according to the difference of the vehicle state disorder factor between the vehicle operation data sequences, a collision risk factor of each vehicle operation data sequence is obtained;

[0009] According to the interference degree of the curve intersection environment and the difference in the change of the vehicle operation data of the target vehicle between adjacent positions, the abnormal factor of each vehicle operation data of the target vehicle is obtained, and the abnormal factor is used to describe the probability that the vehicle operation data is affected by the driver's operation; for any vehicle operation data in any vehicle operation data sequence, a number of control operation data of each vehicle operation data are screened out from the vehicle operation data sequence; according to the distribution discreteness of the control operation data in the vehicle operation data sequence of the target vehicle and the collision risk factor, the relative noise factor of each vehicle operation data of the target vehicle is obtained, and the relative noise factor is used to describe the initial difference between the vehicle operation data and the normal vehicle operation data;

[0010] A predicted collision coefficient of each vehicle operation data of the target vehicle is obtained according to the relative noise factor and the abnormal factor, and the predicted collision coefficient is used to describe the final difference between the vehicle operation data and the normal vehicle operation data; an estimated vehicle collision state value of the target vehicle is obtained according to the predicted collision coefficient, and the estimated vehicle collision state value is used to describe the error of the collision result of the target vehicle; and an early warning is issued to the target vehicle according to the estimated vehicle collision state value.

[0011] Preferably, the method of obtaining the vehicle state disorder factor of each vehicle operation data sequence according to the numerical difference of the vehicle operation data between the vehicle operation data sequences of the target vehicle and the historical vehicle includes the following specific methods:

[0012] The target vehicle and the historical vehicle are both recorded as a reference vehicle, and for any vehicle operation data sequence of any reference vehicle, the vehicle operation data sequence is fitted using the least square method to obtain a fitting curve;

[0013] For any vehicle operation data in the vehicle operation data sequence, the value of the vehicle operation data on the fitting curve is recorded as the original smoothed operation data of the vehicle operation data, the absolute value of the difference between the original smoothed operation data and the vehicle operation data is recorded as the original operation data difference value of the vehicle operation data, the original operation data difference values ​​of all vehicle operation data in the vehicle operation data sequence are obtained, all the original operation data difference values ​​are linearly normalized, and each normalized original operation data difference value is recorded as the smoothed operation difference value;

[0014]

[0015] Where β represents the initial vehicle state disorder factor of any vehicle operation data sequence of any reference vehicle; N represents the number of all vehicle operation data in the vehicle operation data sequence; a n Indicates the nth vehicle operation data in the vehicle operation data sequence; represents the mean of all vehicle operation data in the vehicle operation data sequence; Δa1n represents the smoothed operation difference value of the nth vehicle operation data in the vehicle operation data sequence; || represents taking the absolute value; obtaining the initial vehicle state disorder factors of all vehicle operation data sequences of the reference vehicle, obtaining the initial vehicle state disorder factors of all vehicle operation data sequences of all reference vehicles, linearly normalizing all the initial vehicle state disorder factors, and recording the normalized initial vehicle state disorder factors as vehicle state disorder factors.

[0016] Preferably, the method of obtaining the collision risk factor of each vehicle operation data sequence according to the difference in vehicle state disorder factors between vehicle operation data sequences includes:

[0017]

[0018] Where, β3 z represents the collision risk factor of the zth vehicle operation data sequence of the target vehicle; β2 z represents the vehicle state disorder factor of the zth vehicle operation data sequence of the target vehicle; Represents the vehicle state disorder factor of the zth vehicle operation data sequence of all historical vehicles.

[0019] Preferably, the abnormal factor of each vehicle operation data of the target vehicle is obtained according to the interference degree of the curve intersection environment and the difference in the change of the vehicle operation data of the target vehicle between adjacent positions, including the specific method of:

[0020] Perform DTW matching on the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle to obtain the DTW distance, and record the DTW distance as the sequence data similarity between the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle;

[0021] Record the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle as the target data, record the vehicle operation data sequence whose vehicle operation data type is position data as the vehicle distance operation data sequence, and record the wth vehicle operation data in the vehicle distance operation data sequence of the target vehicle as the target position data of the target data;

[0022] In the vehicle distance operation data sequence of the i-th historical vehicle, the vehicle operation data with the smallest absolute value of difference with the target position data is recorded as the minimum vehicle distance operation data of the target position data in the vehicle distance operation data sequence of the i-th historical vehicle; the absolute value of the difference between the target position data and the minimum vehicle distance operation data is recorded as the vehicle distance difference value of the target data; the vehicle operation data corresponding to the minimum vehicle distance operation data in the j-th vehicle operation data sequence of the i-th historical vehicle is recorded as the final reference operation data of the target data in the j-th vehicle operation data sequence of the i-th historical vehicle;

[0023]

[0024] In the formula, g j,w represents the initial abnormal factor of the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; I represents the number of all historical vehicles; a2 j,w A represents the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; i,j,w represents the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle and the final reference operation data in the jth vehicle operation data sequence of the ith historical vehicle; B i,j,w represents the vehicle distance difference value of the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; d i,j represents the sequence data similarity between the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the ith historical vehicle; τ represents the environmental interference degree of the curve intersection; || represents taking the absolute value; exp() represents an exponential function with a natural constant as the base; the initial abnormal factors of all vehicle operation data in all vehicle operation data sequences of the target vehicle are obtained, all initial abnormal factors are linearly normalized, and each normalized initial abnormal factor is recorded as an abnormal factor.

[0025] Preferably, the specific method of selecting a plurality of control operation data of each vehicle operation data from the vehicle operation data sequence includes:

[0026] A vehicle operation data quantity T1 is preset, and the j1th vehicle operation data sequence of the target vehicle is recorded as the first target sequence; among all vehicle operation data sequences of the target vehicle, the vehicle operation data sequences except the first target sequence are recorded as the first reference sequence; in the first target sequence, the T1 vehicle operation data before the w1th vehicle operation data and the T1 vehicle operation data after the w1th vehicle operation data are both recorded as reference operation data of the w1th vehicle operation data in the first target sequence.

[0027] Preferably, the relative noise factor of each vehicle operation data of the target vehicle is obtained according to the distribution discreteness of the control operation data in the vehicle operation data sequence of the target vehicle and the collision risk factor, including the specific method of:

[0028]

[0029] Where, L j1,w1 represents the initial relative noise factor of the w1th vehicle operation data in the j1th vehicle operation data sequence of the target vehicle; V j1 represents the number of all first reference sequences of the j1th vehicle operation data sequence of the target vehicle; ε j1,w1 represents the standard deviation of all control operation data of the w1th vehicle operation data in the j1th vehicle operation data sequence of the target vehicle; ε j1,v,w1 β3 represents the standard deviation of all control operation data of the w1th vehicle operation data in the vth first reference sequence of the j1th vehicle operation data sequence of the target vehicle; j1 represents the collision risk factor of the j1th vehicle operation data sequence of the target vehicle; β3 j1,v represents the collision risk factor of the vth first reference sequence of the j1th vehicle operation data sequence of the target vehicle; exp() represents an exponential function with a natural constant as the base; the initial relative noise factors of all vehicle operation data in all vehicle operation data sequences of the target vehicle are obtained, all initial relative noise factors are linearly normalized, and each normalized initial relative noise factor is recorded as a relative noise factor.

[0030] Preferably, the method of obtaining the predicted collision coefficient of each vehicle operation data of the target vehicle according to the relative noise factor and the abnormal factor includes the following specific methods:

[0031] f j2,w2 =1-(g1 j2,w2 ×L j2,w2 )

[0032] In the formula, f j2,w2 g1 represents the predicted collision coefficient of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle; j2,w2 represents the abnormal factor of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle; L j2,w2 Represents the relative noise factor of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle.

[0033] Preferably, the method of obtaining the estimated value of the vehicle collision state of the target vehicle according to the predicted collision coefficient includes the following specific methods:

[0034] The predicted value obtained after the extended Kalman filter of the w3th vehicle operation data in the j3th vehicle operation data sequence of the target vehicle is recorded as the traditional predicted value, and the absolute value of the difference between the traditional predicted value and the w3th vehicle operation data is recorded as the comparative filtered value of the w3th vehicle operation data;

[0035]

[0036] Where P represents the estimated value of the vehicle collision state of the target vehicle; Y represents the number of all vehicle operation data sequences of the target vehicle; R j3 represents the number of all vehicle operation data in the j3rd vehicle operation data sequence of the target vehicle; ω j3,w3 represents the comparison filtering value of the w3th vehicle operation data in the j3th vehicle operation data sequence of the target vehicle; f j3,w3 Represents the predicted collision coefficient of the w3th vehicle operation data in the j3th vehicle operation data sequence of the target vehicle.

[0037] Preferably, the specific method of giving an early warning to the target vehicle according to the estimated value of the vehicle collision state includes:

[0038] The state estimation value of the extended Kalman filter algorithm is replaced with the vehicle collision state estimation value of the target vehicle, and the prediction error covariance matrix is ​​obtained according to all vehicle operation data of the target vehicle and the vehicle collision state estimation value, which is recorded as the vehicle collision prediction matrix; the vehicle collision prediction matrix is ​​used as the prediction error covariance matrix to iteratively optimize the prediction model, and the model parameters corresponding to the minimum error matrix are used as the optimal Kalman filter model to predict the vehicle trajectory and perform vehicle collision warning.

[0039] The present invention also proposes a curve intersection oncoming collision warning system, comprising a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the steps of the curve intersection oncoming collision warning method.

[0040] The beneficial effects of the technical solution of the present invention are: obtaining a vehicle state disorder factor according to the numerical difference of vehicle operation data between vehicle operation data sequences, obtaining a collision risk factor according to the vehicle state disorder factor, obtaining an abnormal factor according to the change difference of vehicle operation data between adjacent positions, obtaining a relative noise factor according to the collision risk factor, obtaining a predicted collision coefficient according to the relative noise factor and the abnormal factor, obtaining a vehicle collision state estimation value according to the predicted collision coefficient, and issuing an early warning according to the vehicle collision state estimation value; the vehicle state disorder factor of the present invention reflects the probability that the vehicle operation data in the vehicle operation data sequence is mistakenly identified as noise, the abnormal factor reflects the probability that the vehicle operation data is affected by the driver's operation, the relative noise factor reflects the initial difference between the vehicle operation data and the normal vehicle operation data, the predicted collision coefficient reflects the final difference between the vehicle operation data and the normal vehicle operation data, and the vehicle collision state estimation value reflects the error of the collision result of the target vehicle; the possibility of normal vehicle operation data being mistaken for noise is reduced, the denoising effect is improved, and the accuracy of vehicle collision early warning is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 The present invention is a flowchart of the steps of a method for early warning of vehicle collision at a curved intersection according to the present invention. DETAILED DESCRIPTION

[0043] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a method and system for warning of vehicle collision at a curved intersection according to the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0044] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0045] The specific scheme of the method and system for early warning of vehicle collision at a curved intersection provided by the present invention is described in detail below with reference to the accompanying drawings.

[0046] See also Figure 1, which shows a flowchart of a method for early warning of vehicle collision at a curved intersection provided by an embodiment of the present invention, the method comprising the following steps:

[0047] Step S001: Collecting a number of vehicle operation data sequences of a target vehicle, a number of vehicle operation data sequences of a number of historical vehicles, and the degree of interference of the curve intersection environment.

[0048] It should be noted that the traditional method usually uses extended Kalman filtering for collision warning on vehicle operation data, but the vehicle operation data will change rapidly due to the actual driving situation, and there will be a lot of sudden changes in vehicle operation data. The traditional extended Kalman filter will mistake some normal sudden changes in vehicle operation data as noise data, reducing the accuracy of vehicle collision warning. To this end, this embodiment proposes a collision warning method for vehicles meeting at a curved intersection.

[0049] Specifically, in order to implement a method for warning of vehicle collision at a curved intersection proposed in this embodiment, it is first necessary to collect several vehicle operation data sequences of the target vehicle, several vehicle operation data sequences of several historical vehicles, and the environmental interference degree of the curved intersection. The specific process is: a vehicle equipped with an on-board sensor is taken as the target vehicle, and the target vehicle is parked at a location 300 meters away from the curved intersection where the roadside unit RSU is installed, the target vehicle is started, and communication is established between the target vehicle and the roadside unit RSU at the same time, and then the target vehicle is driven freely toward the curved intersection; the roadside unit records four types of vehicle operation data, namely, speed data, pressure data, acceleration data, and position data, which are transmitted and communicated with the target vehicle once every 2 seconds until the vehicle leaves the maximum communication distance Q of the roadside unit, and stops collecting the vehicle operation data types; the speed data, pressure data, acceleration data, and position data recorded each time are all recorded as one vehicle operation data; taking any type of vehicle operation data as an example, all vehicle operation data of the vehicle operation data type are arranged from small to large according to the recording time, and the arranged sequence is recorded as a vehicle operation data sequence; and several vehicle operation data sequences of the vehicle are obtained. The maximum communication distance Q of the roadside unit RSU used in this embodiment is 300 meters, and each vehicle operation data sequence corresponds to a vehicle operation data type.

[0050] Furthermore, several vehicle operation data of several vehicles in the past week are obtained from the database of the roadside unit RSU, and these vehicles are recorded as historical vehicles; and several vehicle operation data sequences of several historical vehicles are obtained by referring to the method for obtaining several vehicle operation data sequences of the target vehicle.

[0051] Furthermore, several meteorological data of four meteorological data types, namely, temperature, humidity, wind speed and rainfall, within the past week are obtained from the database of the roadside unit RSU; taking any meteorological data type as an example, all meteorological data of this meteorological data type are linearly normalized, and the normalized meteorological data are recorded as initial meteorological data, and the mean of all meteorological data of this meteorological data type is recorded as initial characterization environmental data of this meteorological data type; the initial characterization environmental data of all meteorological data types are obtained, and the mean of the initial characterization environmental data of all meteorological data types is recorded as the environmental interference degree of the curved intersection.

[0052] So far, several vehicle operation data sequences of the target vehicle, several vehicle operation data sequences of several historical vehicles and the interference degree of the curve intersection environment are obtained through the above method.

[0053] Step S002: According to the numerical difference of the vehicle operation data between the vehicle operation data sequences of the target vehicle and the historical vehicle, the vehicle state disorder factor of each vehicle operation data sequence is obtained; according to the difference of the vehicle state disorder factors between the vehicle operation data sequences, the collision risk factor of each vehicle operation data sequence is obtained.

[0054] It should be noted that when the vehicle is driving towards a curved intersection, the curved intersection will block part of the road condition. Normally, in order to ensure the safety of the vehicle when passing through a curved intersection, the vehicle speed will change in a relatively gentle trend; but in actual driving conditions, there may be interference factors such as some roads being relatively uneven or relatively smooth, and these interference factors will to a certain extent destroy the original gentle trend of the vehicle speed change; this embodiment obtains the vehicle state disorder factor through the change difference between the vehicle operation data in the vehicle operation data sequence, and obtains the collision risk factor according to the changes in the local vehicle state disorder factor and the overall vehicle state disorder factor, so as to facilitate subsequent analysis and processing.

[0055] Specifically, the target vehicle and the historical vehicle are both recorded as a reference vehicle, and any vehicle operation data sequence of any reference vehicle is taken as an example, and the least square method is used to fit the vehicle operation data sequence to obtain a fitting curve; taking any vehicle operation data in the vehicle operation data sequence as an example, the value of the vehicle operation data on the fitting curve is recorded as the original smoothed operation data of the vehicle operation data, and the absolute value of the difference between the original smoothed operation data and the vehicle operation data is recorded as the original operation data difference value of the vehicle operation data, and the original operation data difference values ​​of all vehicle operation data in the vehicle operation data sequence are obtained, and all original operation data difference values ​​are linearly normalized, and each normalized original operation data difference value is recorded as a smoothed operation difference value. The least square method is a well-known technology, and this embodiment will not be repeated.

[0056] Furthermore, according to the smoothed operation difference values ​​of all vehicle operation data in the vehicle operation data sequence, the initial vehicle state disorder factor of the vehicle operation data sequence is obtained. The calculation method of the initial vehicle state disorder factor of the vehicle operation data sequence is:

[0057]

[0058] Where β represents the initial vehicle state disorder factor of the vehicle operation data sequence; N represents the number of all vehicle operation data in the vehicle operation data sequence; a n Indicates the nth vehicle operation data in the vehicle operation data sequence; represents the mean of all vehicle operation data in the vehicle operation data sequence; Δa1 n Indicates the smoothed running difference value of the nth vehicle running data in the vehicle running data sequence; || indicates taking the absolute value. The larger the initial vehicle state disorder factor of the vehicle running data sequence, the more irregular the change of the type of vehicle running data represented by the vehicle running data sequence, reflecting the greater possibility that the vehicle running data in the vehicle running data sequence is mistakenly identified as noise. Obtain the initial vehicle state disorder factors of all vehicle running data sequences of the reference vehicle, obtain the initial vehicle state disorder factors of all vehicle running data sequences of all reference vehicles, linearly normalize all initial vehicle state disorder factors, and record the normalized initial vehicle state disorder factors as vehicle state disorder factors.

[0059] Furthermore, according to the vehicle state disorder factor of the zth vehicle operation data sequence of all historical vehicles, the collision risk factor of the zth vehicle operation data sequence of the target vehicle is obtained. The calculation method of the collision risk factor of the zth vehicle operation data sequence of the target vehicle is:

[0060]

[0061] Where, β3 z represents the collision risk factor of the zth vehicle operation data sequence of the target vehicle; β2 z represents the vehicle state disorder factor of the zth vehicle operation data sequence of the target vehicle; Represents the vehicle state disorder factor of the zth vehicle operation data sequence of all historical vehicles. If the collision risk factor of the zth vehicle operation data sequence of the target vehicle is larger, it means that the change law of the vehicle operation data in the zth vehicle operation data sequence of the target vehicle is less consistent with the normal driving law. From the perspective of the target vehicle in the zth vehicle operation data sequence, the target vehicle is more likely to collide and is more affected by noise. Obtain the collision risk factor of all vehicle operation data sequences of the target vehicle.

[0062] So far, the collision risk factors of all vehicle operation data sequences of the target vehicle are obtained through the above method.

[0063] Step S003: Obtain the abnormal factor of each vehicle operation data of the target vehicle according to the interference degree of the curve intersection environment and the difference in the change of the vehicle operation data of the target vehicle between adjacent positions; select a number of control operation data of each vehicle operation data from the vehicle operation data sequence; obtain the relative noise factor of each vehicle operation data of the target vehicle according to the distribution discreteness of the control operation data in the vehicle operation data sequence of the target vehicle and the collision risk factor.

[0064] It should be noted that the collision risk factor is obtained based on the overall vehicle operation data sequence, and can characterize the risk level of the overall vehicle operation data sequence of the target vehicle; however, noise is usually distributed on the local vehicle operation data in the vehicle operation data sequence, and at this time, the collision risk factor is not sufficient to characterize the risk level of the local vehicle operation data; under normal circumstances, the roughness of the road surface in different areas is not the same, and there will be large or small differences. The corresponding road conditions will have hidden dangers of being bumpy or smooth, and when the vehicle travels on these roads with hidden dangers, in order to ensure the safety of the vehicle, the driving state of the vehicle will be greatly changed; therefore, under normal circumstances, for any position data, if all the vehicle operation data corresponding to the position data change significantly, then these vehicle operation data with significant changes are very likely to be caused by the position data; if some of the vehicle operation data corresponding to the position data change significantly, then it means that the possibility that these vehicle operation data with significant changes are caused by the position data is low, and at the same time, because noise will cause the local vehicle operation data to change significantly, so these vehicle operation data with significant changes are more likely to be caused by noise interference. To this end, this embodiment obtains an abnormal factor by analyzing the difference relationship between the vehicle operation data corresponding to the position data, obtains a relative noise factor based on the collision risk factor, and obtains a predicted collision coefficient based on the abnormal factor and the relative noise factor for subsequent analysis and processing.

[0065] Specifically, DTW matching is performed on the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle to obtain a DTW distance, and the DTW distance is recorded as the sequence data similarity between the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle; wherein the DTW matching algorithm is a well-known technology and will not be repeated in this embodiment.

[0066] Further, the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle is recorded as the target data, the vehicle operation data sequence of which the vehicle operation data type is position data is recorded as the vehicle distance operation data sequence, and the wth vehicle operation data in the vehicle distance operation data sequence of the target vehicle is recorded as the target position data of the target data; in the vehicle distance operation data sequence of the i-th historical vehicle, the vehicle operation data with the smallest absolute value of difference with the target position data is recorded as the minimum vehicle distance operation data of the target position data in the vehicle distance operation data sequence of the i-th historical vehicle; the absolute value of the difference between the target position data and the minimum vehicle distance operation data is recorded as the vehicle distance difference value of the target data; the vehicle operation data corresponding to the minimum vehicle distance operation data in the jth vehicle operation data sequence of the i-th historical vehicle is recorded as the final reference operation data of the target data in the jth vehicle operation data sequence of the i-th historical vehicle. Each vehicle corresponds to a vehicle distance operation data sequence, and for any vehicle operation data in the vehicle distance operation data sequence of any vehicle, the vehicle operation data corresponds to one vehicle operation data in each vehicle operation data sequence of the vehicle. It should also be noted that if there are multiple minimum vehicle distance running data of the target position data in the vehicle distance running data sequence of the i-th historical vehicle, then the serial number of the target position data in the vehicle distance running data sequence of the target vehicle is recorded as the first serial number, the serial number of each minimum vehicle distance running data in the vehicle distance running data sequence of the i-th historical vehicle is recorded as the second serial number, the absolute value of the difference between the second serial number and the first serial number is recorded as the first absolute value, and the minimum vehicle distance running data with the smallest first absolute value is used as the final minimum vehicle distance running data of the target position data in the vehicle distance running data sequence of the i-th historical vehicle.

[0067] Further, according to the sequence data similarity between the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle, the interference degree of the curved road intersection environment, and the minimum vehicle distance operation data of the w-th vehicle operation data in the vehicle distance operation data sequence of the target vehicle in the vehicle distance operation data sequence of the i-th historical vehicle, the initial abnormal factor of the w-th vehicle operation data in the j-th vehicle operation data sequence of the target vehicle is obtained. The calculation method of the initial abnormal factor of the w-th vehicle operation data in the j-th vehicle operation data sequence of the target vehicle is:

[0068]

[0069] In the formula, g j,w represents the initial abnormal factor of the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; I represents the number of all historical vehicles; a2 j,w A represents the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; i,j,wrepresents the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle and the final reference operation data in the jth vehicle operation data sequence of the ith historical vehicle; B i,j,w represents the vehicle distance difference value of the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; d i,j Represents the sequence data similarity between the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle; τ represents the interference degree of the curve intersection environment; || represents taking the absolute value; exp() represents an exponential function with a natural constant as the base; the embodiment uses the exp(-) function to present the inverse proportional relationship and normalization processing, and the implementer can select the inverse proportional function and normalization function according to the actual situation. If the initial abnormal factor of the w-th vehicle operation data in the j-th vehicle operation data sequence of the target vehicle is larger, it means that the w-th vehicle operation data is less affected by the curve intersection environment, and the correlation with the driver's driving operation is lower, reflecting that the w-th vehicle operation data is more likely to be disturbed by noise. Obtain the initial abnormal factors of all vehicle operation data in all vehicle operation data sequences of the target vehicle, linearly normalize all initial abnormal factors, and record each normalized initial abnormal factor as an abnormal factor.

[0070] Further, a vehicle operation data quantity T1 is preset, wherein this embodiment is described by taking T1=5 as an example, and this embodiment is not specifically limited, wherein T1 can be determined according to specific implementation conditions; the j1th vehicle operation data sequence of the target vehicle is recorded as the first target sequence; among all vehicle operation data sequences of the target vehicle, the vehicle operation data sequences other than the first target sequence are recorded as the first reference sequence; in the first target sequence, the vehicle operation data before the w1th vehicle operation data and the vehicle operation data after the w1th vehicle operation data are recorded as the comparison operation data of the w1th vehicle operation data in the first target sequence; and the comparison operation data of the w1th vehicle operation data in each first reference sequence of the first target sequence is obtained. In the process of obtaining the comparison operation data of the w1th vehicle operation data, if the number of vehicle operation data actually existing before and after the w1th vehicle operation data does not meet the preset T1, then the comparison operation data is obtained up to the number of vehicle operation data actually existing before and after the w1th vehicle operation data.

[0071] Furthermore, according to the collision risk factor of the vehicle operation data sequence of the target vehicle and the control operation data of the w1th vehicle operation data in the corresponding vehicle operation data sequence, the initial relative noise factor of the w1th vehicle operation data in the j1th vehicle operation data sequence of the target vehicle is obtained. The calculation method of the initial relative noise factor of the w1th vehicle operation data in the j1th vehicle operation data sequence of the target vehicle is:

[0072]

[0073] Where, L j1,w1 represents the initial relative noise factor of the w1th vehicle operation data in the j1th vehicle operation data sequence of the target vehicle; V j1 represents the number of all first reference sequences of the j1th vehicle operation data sequence of the target vehicle; ε j1,w1 represents the standard deviation of all control operation data of the w1th vehicle operation data in the j1th vehicle operation data sequence of the target vehicle; ε j1,v,w1 β3 represents the standard deviation of all control operation data of the w1th vehicle operation data in the vth first reference sequence of the j1th vehicle operation data sequence of the target vehicle; j1 represents the collision risk factor of the j1th vehicle operation data sequence of the target vehicle; β3 j1,v Represents the collision risk factor of the vth first reference sequence of the j1th vehicle operation data sequence of the target vehicle; exp() represents an exponential function with a natural constant as the base; the embodiment adopts the exp(-) function to present the inverse proportional relationship and normalization processing, and the implementer can select the inverse proportional function and the normalization function according to the actual situation. The larger the initial relative noise factor of the w1th vehicle operation data in the j1th vehicle operation data sequence of the target vehicle, the more obvious the abnormal numerical change characteristics of the w1th vehicle operation data are, reflecting that the degree of noise interference to the w1th vehicle operation data is greater. The initial relative noise factors of all vehicle operation data in all vehicle operation data sequences of the target vehicle are obtained, all initial relative noise factors are linearly normalized, and each normalized initial relative noise factor is recorded as a relative noise factor.

[0074] So far, the abnormal factors and relative noise factors of all vehicle operation data in each vehicle operation data sequence of the target vehicle are obtained through the above method.

[0075] Step S004: obtaining a predicted collision coefficient of each vehicle operation data of the target vehicle according to the relative noise factor and the abnormal factor; obtaining an estimated value of the vehicle collision state of the target vehicle according to the predicted collision coefficient; and issuing an early warning to the target vehicle according to the estimated value of the vehicle collision state.

[0076] Specifically, according to the relative noise factor and abnormal factor of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle, the predicted collision coefficient of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle is obtained. The calculation method of the predicted collision coefficient of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle is:

[0077] f j2,w2=1-(g1 j2,w2 ×L j2,w2 )

[0078] In the formula, f j2,w2 g1 represents the predicted collision coefficient of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle; j2,w2 represents the abnormal factor of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle; L j2,w2 Represents the relative noise factor of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle. The larger the predicted collision coefficient of the w2th vehicle operation data in the j2th vehicle operation data sequence of the target vehicle, the higher the predicted value of the w2th vehicle operation data, reflecting that the degree to which the w2th vehicle operation data is ultimately disturbed by noise is lower. Obtain the predicted collision coefficient of each vehicle operation data in each vehicle operation data sequence of the target vehicle.

[0079] Further, the predicted value obtained after the extended Kalman filter of the w3th vehicle operation data in the j3th vehicle operation data sequence of the target vehicle is recorded as the traditional predicted value, and the absolute value of the difference between the traditional predicted value and the w3th vehicle operation data is recorded as the comparative filtered value of the w3th vehicle operation data. The process of obtaining the predicted value of the data is a well-known content of the extended Kalman filter algorithm, and will not be repeated in this embodiment.

[0080] Furthermore, according to the predicted collision coefficients of all vehicle operation data in each vehicle operation data sequence of the target vehicle and the comparison filter value, the vehicle collision state estimation value of the target vehicle is obtained. The vehicle collision state estimation value of the target vehicle is calculated as follows:

[0081]

[0082] Where P represents the estimated value of the vehicle collision state of the target vehicle; Y represents the number of all vehicle operation data sequences of the target vehicle; R j3 represents the number of all vehicle operation data in the j3rd vehicle operation data sequence of the target vehicle; ω j3,w3 represents the comparison filtering value of the w3th vehicle operation data in the j3th vehicle operation data sequence of the target vehicle; f j3,w3 Represents the predicted collision coefficient of the w3th vehicle operation data in the j3th vehicle operation data sequence of the target vehicle.

[0083] Furthermore, the state estimation value of the traditional extended Kalman filter algorithm is replaced with the vehicle collision state estimation value of the target vehicle, and the prediction error covariance matrix is ​​obtained according to all vehicle operation data of the target vehicle and the vehicle collision state estimation value, which is recorded as the vehicle collision prediction matrix; the vehicle collision prediction matrix is ​​used as the prediction error covariance matrix to iteratively optimize the prediction model, and the model parameters corresponding to the minimum error matrix are used as the optimal Kalman filter model to predict the vehicle trajectory and perform a vehicle collision warning. The process of predicting the vehicle trajectory and performing a collision warning according to the optimal Kalman filter model is the known content of patent CN116526454A, and will not be repeated in this embodiment. In addition, the process of performing a collision warning according to the prediction error covariance matrix is ​​the known content of "Research on Simulation of Vehicle Collaborative Collision Warning Based on Trajectory Prediction", and will not be repeated in this embodiment.

[0084] Through the above steps, a method for early warning of vehicle collision at a curved intersection is completed.

[0085] Another embodiment of the present invention provides a curved intersection collision warning system, the system comprising a memory and a processor, and when the processor executes a computer program stored in the memory, the processor performs the following operations:

[0086] Collecting a plurality of vehicle operation data sequences of a target vehicle, a plurality of vehicle operation data sequences of a plurality of historical vehicles, and a degree of interference of a curved intersection environment, wherein the vehicle operation data sequence includes a plurality of vehicle operation data;

[0087] According to the numerical difference of the vehicle operation data between the vehicle operation data sequences of the target vehicle and the historical vehicle, a vehicle state disorder factor of each vehicle operation data sequence is obtained, wherein the vehicle state disorder factor is used to describe the probability that the vehicle operation data in the vehicle operation data sequence is mistakenly identified as noise; according to the difference of the vehicle state disorder factor between the vehicle operation data sequences, a collision risk factor of each vehicle operation data sequence is obtained;

[0088] According to the interference degree of the curve intersection environment and the difference in the change of the vehicle operation data of the target vehicle between adjacent positions, the abnormal factor of each vehicle operation data of the target vehicle is obtained, and the abnormal factor is used to describe the probability that the vehicle operation data is affected by the driver's operation; for any vehicle operation data in any vehicle operation data sequence, a number of control operation data of each vehicle operation data are screened out from the vehicle operation data sequence; according to the distribution discreteness of the control operation data in the vehicle operation data sequence of the target vehicle and the collision risk factor, the relative noise factor of each vehicle operation data of the target vehicle is obtained, and the relative noise factor is used to describe the initial difference between the vehicle operation data and the normal vehicle operation data;

[0089] A predicted collision coefficient of each vehicle operation data of the target vehicle is obtained according to the relative noise factor and the abnormal factor, and the predicted collision coefficient is used to describe the final difference between the vehicle operation data and the normal vehicle operation data; an estimated vehicle collision state value of the target vehicle is obtained according to the predicted collision coefficient, and the estimated vehicle collision state value is used to describe the error of the collision result of the target vehicle; and an early warning is issued to the target vehicle according to the estimated vehicle collision state value.

[0090] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for early warning of vehicle collision at a curved intersection, characterized in that: The method comprises the following steps: Collecting a plurality of vehicle operation data sequences of a target vehicle, a plurality of vehicle operation data sequences of a plurality of historical vehicles, and a degree of interference of a curved intersection environment, wherein the vehicle operation data sequence includes a plurality of vehicle operation data; According to the numerical difference of the vehicle operation data between the vehicle operation data sequences of the target vehicle and the historical vehicle, a vehicle state disorder factor of each vehicle operation data sequence is obtained, wherein the vehicle state disorder factor is used to describe the probability that the vehicle operation data in the vehicle operation data sequence is mistakenly identified as noise; according to the difference of the vehicle state disorder factor between the vehicle operation data sequences, a collision risk factor of each vehicle operation data sequence is obtained; According to the interference degree of the curve intersection environment and the difference in the change of the vehicle operation data of the target vehicle between adjacent positions, the abnormal factor of each vehicle operation data of the target vehicle is obtained, and the abnormal factor is used to describe the probability that the vehicle operation data is affected by the driver's operation; for any vehicle operation data in any vehicle operation data sequence, a number of control operation data of each vehicle operation data are screened out from the vehicle operation data sequence; according to the distribution discreteness of the control operation data in the vehicle operation data sequence of the target vehicle and the collision risk factor, the relative noise factor of each vehicle operation data of the target vehicle is obtained, and the relative noise factor is used to describe the initial difference between the vehicle operation data and the normal vehicle operation data; A predicted collision coefficient of each vehicle operation data of the target vehicle is obtained according to the relative noise factor and the abnormal factor, and the predicted collision coefficient is used to describe the final difference between the vehicle operation data and the normal vehicle operation data; an estimated vehicle collision state value of the target vehicle is obtained according to the predicted collision coefficient, and the estimated vehicle collision state value is used to describe the error of the collision result of the target vehicle; and an early warning is given to the target vehicle according to the estimated vehicle collision state value; The specific method of selecting a plurality of comparison operation data of each vehicle operation data from the vehicle operation data sequence includes: Preset a vehicle operation data quantity T1, and The vehicle operation data sequence is recorded as the first target sequence; among all the vehicle operation data sequences of the target vehicle, the vehicle operation data sequences other than the first target sequence are recorded as the first reference sequence; in the first target sequence, The vehicle operation data before T1 vehicle operation data and the The vehicle operation data after T1 are recorded as the first target sequence. The comparison operation data of each vehicle operation data.

2. The method for early warning of vehicle collision at a curved intersection according to claim 1, characterized in that: The method of obtaining the vehicle state disorder factor of each vehicle operation data sequence according to the numerical difference of the vehicle operation data between the vehicle operation data sequence of the target vehicle and the historical vehicle includes: The target vehicle and the historical vehicle are both recorded as a reference vehicle, and for any vehicle operation data sequence of any reference vehicle, the vehicle operation data sequence is fitted using the least square method to obtain a fitting curve; For any vehicle operation data in the vehicle operation data sequence, the value of the vehicle operation data on the fitting curve is recorded as the original smoothed operation data of the vehicle operation data, the absolute value of the difference between the original smoothed operation data and the vehicle operation data is recorded as the original operation data difference value of the vehicle operation data, the original operation data difference values ​​of all vehicle operation data in the vehicle operation data sequence are obtained, all the original operation data difference values ​​are linearly normalized, and each normalized original operation data difference value is recorded as the smoothed operation difference value; , In the formula, represents the initial vehicle state disorder factor of any vehicle operation data sequence of any reference vehicle; Indicates the number of all vehicle operation data in the vehicle operation data sequence; Indicates the nth vehicle operation data in the vehicle operation data sequence; Represents the mean of all vehicle operation data in the vehicle operation data sequence; represents the smoothed running difference value of the nth vehicle running data in the vehicle running data sequence; It means taking the absolute value; obtaining the initial vehicle state disorder factor of all vehicle operation data sequences of the reference vehicle, obtaining the initial vehicle state disorder factor of all vehicle operation data sequences of all reference vehicles, linearly normalizing all the initial vehicle state disorder factors, and recording the normalized initial vehicle state disorder factor as the vehicle state disorder factor.

3. The method for early warning of vehicle collision at a curved intersection according to claim 1, characterized in that: The specific method of obtaining the collision risk factor of each vehicle operation data sequence according to the difference in vehicle state disorder factors between vehicle operation data sequences includes: , In the formula, represents the collision risk factor of the zth vehicle operation data sequence of the target vehicle; represents the vehicle state disorder factor of the zth vehicle operation data sequence of the target vehicle; Represents the vehicle state disorder factor of the Zth vehicle operation data sequence of all historical vehicles.

4. The method for early warning of vehicle collision at a curved intersection according to claim 1, characterized in that: The specific method of obtaining the abnormal factor of each vehicle operation data of the target vehicle according to the interference degree of the curve intersection environment and the change difference of the vehicle operation data of the target vehicle between adjacent positions includes: Perform DTW matching on the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle to obtain the DTW distance, and record the DTW distance as the sequence data similarity between the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the i-th historical vehicle; Record the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle as the target data, record the vehicle operation data sequence whose vehicle operation data type is position data as the vehicle distance operation data sequence, and record the wth vehicle operation data in the vehicle distance operation data sequence of the target vehicle as the target position data of the target data; In the vehicle distance operation data sequence of the i-th historical vehicle, the vehicle operation data having the smallest absolute value of difference with the target position data is recorded as the minimum vehicle distance operation data of the target position data in the vehicle distance operation data sequence of the i-th historical vehicle; The absolute value of the difference between the target position data and the minimum vehicle distance operation data is recorded as the vehicle distance difference value of the target data; the vehicle operation data corresponding to the minimum vehicle distance operation data in the jth vehicle operation data sequence of the i-th historical vehicle is recorded as the final reference operation data of the target data in the jth vehicle operation data sequence of the i-th historical vehicle; , In the formula, represents the initial abnormal factor of the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; Represents the number of all historical vehicles; represents the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; represents the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle and the final reference operation data in the jth vehicle operation data sequence of the ith historical vehicle; represents the vehicle distance difference value of the wth vehicle operation data in the jth vehicle operation data sequence of the target vehicle; represents the sequence data similarity between the j-th vehicle operation data sequence of the target vehicle and the j-th vehicle operation data sequence of the ith historical vehicle; Indicates the degree of environmental interference at the curve intersection; Indicates taking the absolute value; represents an exponential function with a natural constant as the base; obtaining initial abnormal factors of all vehicle operation data in all vehicle operation data sequences of the target vehicle, linearly normalizing all initial abnormal factors, and recording each normalized initial abnormal factor as an abnormal factor.

5. The method for early warning of vehicle collision at a curved intersection according to claim 1, characterized in that: The specific method of obtaining the relative noise factor of each vehicle operation data of the target vehicle according to the distribution discreteness of the control operation data in the vehicle operation data sequence of the target vehicle and the collision risk factor includes: , In the formula, The target vehicle The vehicle operation data series The initial relative noise factor of the vehicle operation data; The target vehicle The number of all first reference sequences of vehicle operation data sequences; The target vehicle The vehicle operation data series The standard deviation of all control operation data of each vehicle operation data; The target vehicle The first reference sequence of the vth vehicle operation data sequence The standard deviation of all control operation data of each vehicle operation data; The target vehicle The collision risk factor of a vehicle operation data series; The target vehicle The collision risk factor of the vth first reference sequence of the vehicle operation data sequence; represents an exponential function with a natural constant as base; The initial relative noise factors of all vehicle operation data in all vehicle operation data sequences of the target vehicle are obtained, all the initial relative noise factors are linearly normalized, and each normalized initial relative noise factor is recorded as a relative noise factor.

6. The method for early warning of vehicle collision at a curved intersection according to claim 1, characterized in that: The specific method of obtaining the predicted collision coefficient of each vehicle operation data of the target vehicle according to the relative noise factor and the abnormal factor includes: , In the formula, The target vehicle The vehicle operation data series Predicted collision coefficient of vehicle operation data; The target vehicle The vehicle operation data series Abnormal factors of vehicle operation data; The target vehicle The vehicle operation data series The relative noise factor of vehicle operation data.

7. The method for early warning of vehicle collision at a curved intersection according to claim 1, characterized in that: The specific method of obtaining the estimated value of the vehicle collision state of the target vehicle according to the predicted collision coefficient includes: The target vehicle's The vehicle operation data series The predicted value obtained after the extended Kalman filter of the vehicle operation data is recorded as the traditional predicted value. The absolute value of the difference between the vehicle operation data is recorded as Comparative filtering value of vehicle operation data; , In the formula, P represents an estimated value of a vehicle collision state of the target vehicle; Y The number of all vehicle operation data sequences representing the target vehicle; The target vehicle The number of all vehicle operation data in a vehicle operation data sequence; The target vehicle The vehicle operation data series Comparative filtering value of vehicle operation data; The target vehicle The vehicle operation data series The predicted collision coefficient of each vehicle operation data.

8. The method for warning vehicle collision at a curved intersection according to claim 1, characterized in that: The specific method of giving an early warning to the target vehicle according to the estimated value of the vehicle collision state includes: The state estimation value of the extended Kalman filter algorithm is replaced with the vehicle collision state estimation value of the target vehicle, and the prediction error covariance matrix is ​​obtained according to all vehicle operation data of the target vehicle and the vehicle collision state estimation value, which is recorded as the vehicle collision prediction matrix; the vehicle collision prediction matrix is ​​used as the prediction error covariance matrix to iteratively optimize the prediction model, and the model parameters corresponding to the minimum error matrix are used as the optimal Kalman filter model to predict the vehicle trajectory and perform vehicle collision warning.

9. A vehicle collision warning system at a curved intersection, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the method for warning of vehicle collision at a curved intersection as described in any one of claims 1 to 8 are implemented.