Auxiliary driving automobile accident key process identification method based on potential field theory

Abstract

According to the auxiliary driving automobile accident key process identification method based on the potential field theory provided by the invention, the relevance between the auxiliary driving automobile operation key process and the accident risk is deeply analyzed, and the ideas of qualitative analysis and quantitative analysis are fused; the accident key process is defined as the space-time partition corresponding to the space-time node from sudden change of the operation risk to the end of the accident in the driving process, and the function of identifying the accident key process is achieved. The accident risk assessment model is constructed by using full-time-domain multivariate refined data, and risk abrupt change points are calibrated on the basis of space-time continuity to identify and extract the key process of an auxiliary driving automobile accident, so that the accuracy of auxiliary driving automobile risk assessment and accident key process identification is improved. The potential traffic risk of the auxiliary driving system is considered, so that the robustness and applicability of the model are improved; and a precise time-space node identification result of the accident occurrence key process is provided for assisting automobile driving accident process reconstruction, fact restoration, hidden danger tracing and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent transportation and safety management and control, and specifically relates to a method for identifying key processes of assisted driving automobile accidents based on potential field theory. Background Art

[0002] The identification and extraction of key processes in assisted driving car traffic accidents are of great significance to the in-depth digital investigation of assisted driving car accidents. When a vehicle is driving on the road, it is composed of a series of longitudinal and lateral movements. Under the influence of each other, new safety risks will continue to arise between vehicles and between vehicles and roads, forming a safety field. Risk management adopts a logical and systematic method to clarify the risks existing in the operation of assisted driving cars, combines the potential field theory in physics, integrates the ideas of qualitative analysis and quantitative analysis, and uses risk assessment methods to analyze risks from the two aspects of consequences and possibilities, determine the time and space points of risk mutation, and realize the identification of key processes, so as to provide accurate time and space node identification results of key processes of accidents for accident process reconstruction, fact restoration, hidden danger tracing, etc.

[0003] Existing accident investigation studies have not clearly defined the critical process of an accident. Since the event data recorder (EDR) of a traditional car can only record driving data about 5 seconds before the accident is triggered, the collision process is usually the focus in the accident scene investigation, trajectory identification, and accident process reconstruction, while ignoring the accumulation process of a certain factor that caused the accident. Assisted driving cars can automatically record and store operating status information from at least 90 seconds before the accident to at least 30 seconds after the accident, and can achieve full-time risk monitoring. It is necessary to define the critical process of an accident from the perspective of risk management and control, and propose a calibration process for the time and space interval of the critical process.

[0004] Due to the complexity of assisted driving car in terms of perception, decision-making and control technology and the particularity of human-machine co-driving environment, its data is multi-source and fine but complex and difficult to control, which makes it difficult to identify key accident processes and analyze factors. Existing research uses field theory to describe the driving risks caused by various traffic factors, which can be used to assess the potential driving risks in real traffic scenarios and applied in the research of assisted driving warning algorithms. It is necessary to couple multi-factors of multi-source fine perception data based on potential field theory, and build a quantitative assessment model of operation risk covering the entire time domain based on spatiotemporal continuity and changes in risk indicator characteristics, so as to realize the identification and extraction of key accident processes of assisted driving cars and improve the accuracy of identification. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for identifying critical processes of an assisted driving automobile accident based on potential field theory, which is used for identifying critical processes of an accident.

[0006] The technical solution adopted by the present invention to solve the above technical problems is: a method for identifying key processes of assisted driving automobile accidents based on potential field theory, comprising the following steps:

[0007] S1: Collect and pre-process the accident data of assisted driving vehicles;

[0008] S2: Based on the preprocessed data, the safety field model of assisted driving vehicle operation is established based on the potential field theory, the driving risk is quantified with the potential energy value, the dynamic risk value covering the entire time domain is calculated and a time series is formed;

[0009] S3: According to the time series, the MK mutation test method is used to determine different risk thresholds at different significance levels, and the corresponding risk mutation points are marked;

[0010] S4: Observe the spatiotemporal segmentation results of the accident pointed to by different risk mutation points, determine the optimal threshold and spatiotemporal interval, and identify the key process of the accident.

[0011] According to the above scheme, in step S1,

[0012] The time range of assisted driving car accident data is from 90 seconds before the accident to 30 seconds after the accident;

[0013] Assisted driving vehicle accident data includes assisted driving vehicle operation data, road condition data and surrounding environment monitoring data; the sources of assisted driving vehicle accident data include assisted driving vehicle event data recording system EDR, CAN, Mobileye, radar, camera, road test equipment and on-site survey;

[0014] The preprocessing methods include using interpolation to fill in missing values ​​of assisted driving vehicle operation data, deleting dirty data, and smoothing data to reduce noise.

[0015] According to the above scheme, in step S2, the specific steps are:

[0016] S21: Assume that the field strength vector of the safety potential field of the assisted driving vehicle is E S , the field strength vector of the road marking field is E R , the field strength vector of the vehicle kinetic energy field is E V ; Assume that the coefficient of different lane markings is T a , lane width is D, r ai =(x i -x a ,y i -y a ) is the coordinate point (x a ,y a ) and the center of mass of vehicle i (x i ,yi ) between them; let the velocity vector of vehicle i be v i ,(x j ,y j ) The velocity vector of vehicle j is v j , r ji =(x i -x j ,y i -y j ) is the distance between vehicle j and vehicle mass center (x i ,y i ) between them; based on the potential field theory, the assisted driving vehicle operation safety potential field model including the road marking field and the vehicle kinetic energy field is established as follows:

[0017] E s =E R +E V (1);

[0018] E R =E R (T a ,D,r ai ) (2);

[0019] E V =E V (r ji , v i , v j , x i , x j ,y i , x j ) (3);

[0020] S22: Let E k is the k-type potential field, m is the actual mass of the target, v is the velocity value of the target, α is the unknown coefficient, and θ is the angle between the distance vector from the field source to the target and the target velocity vector; the field force is constructed according to the electric field force formula F = Eq, and the k-type field force is recorded as F k , the expression is:

[0021] F k =|E k |·m·e α·v·cosθ (4);

[0022] S23: Let F be the field force, r be the distance from the field source to the target center of mass; use the potential energy value SPE to quantify the driving risk of the assisted driving car and calculate the dynamic risk value in the full time domain, and the calculation results form a time series TS; the potential energy calculation formula is:

[0023]

[0024] Furthermore, in step S21, k1 is set to be a constant greater than 0; the road marking field is described as the vehicle i at (x i ,y i ) when the lane marking line a forms the field intensity vector E R，ai :

[0025]

[0026] Furthermore, in step S21, k2 is set to be a constant greater than 0, k3 is a speed weight coefficient, σ x , σ y is the convergence coefficient of the potential field influence range, θ is the angle between the distance vector from the field source to the target and the target velocity vector; the vehicle kinetic energy field is described as the vehicle i at (x i ,y i ) at a certain point on the road (x j ,y j ) formed by vehicle j on it V，ji :

[0027]

[0028] Furthermore, in step S22, it is assumed that m i is the actual mass of vehicle i, α and β are unknown coefficients; the field force of type k is denoted as F k , specifically expressed as follows:

[0029]

[0030]

[0031] Furthermore, in step S23, the potential energy generated by the k-type potential field is recorded as SPE k :

[0032]

[0033] According to the accident scenario and data, multiple fixed values ​​of the model are set, and the dynamic risk value SPE covering the entire time domain is calculated based on the preprocessed data:

[0034] SPE = max{SPE k} (13);

[0035] According to the calculation results, a time series TS is formed, which contains n SPE values ​​(SPE1, SPE2, ... SPE n ).

[0036] Furthermore, in step S3, the specific steps are:

[0037] S31: Let time series TS be a sequential time series, and its reverse time series be TS1; construct the order sequence TS of the sequential time series TS m ; Define the statistic UF m and UB m , when m=1, UF1=0; E(TS m ), Var(TS m ) are TS m The mean and variance of :

[0038]

[0039] E(TS m )=n(n+1) / 4(16),

[0040] Var(TS m )=n(n-1)(2n+5) / 72(17), repeat the above steps for TS1 based on the reverse time series, and get:

[0041] UB m =-UF m (m=n, n-1, ... 1);

[0042] S32: Given different significance levels, determine different confidence intervals, and compare the confidence interval critical value straight line with the two statistical sequences UF obtained in step S31 m and UB m The curves are drawn in a rectangular coordinate system; if the two curves have an intersection and the intersection is between the critical lines, then the intersection is the risk mutation point, and the corresponding risk value is the threshold.

[0043] Furthermore, in step S4, the specific steps are:

[0044] According to step S32, the optimal threshold is selected to calibrate the risk mutation point, the risk mutation point is matched with the key process time-space node, the key process time-space area pointed to by the risk mutation point is identified, and the key accident process of the assisted driving car is identified and extracted.

[0045] A computer storage medium stores a computer program executable by a computer processor, wherein the computer program executes a method for identifying critical processes of an assisted driving automobile accident based on potential field theory.

[0046] The beneficial effects of the present invention are:

[0047] 1. The present invention provides a method for identifying critical accident processes of assisted driving vehicles based on potential field theory, which defines the critical accident process as the space-time partition corresponding to the space-time node from the sudden change of operating risk during driving to the end of the accident, thereby realizing the function of identifying the critical accident process.

[0048] 2. The present invention clearly defines the time and space points of the "critical accident process". By conducting an in-depth analysis of the correlation between the critical operation process of assisted driving vehicles and accident risks, the definition of the "critical accident process" and its relationship with risks are clarified, and the ideas of qualitative analysis and quantitative analysis are integrated to point out the principle of time and space division of the critical accident process.

[0049] 3. The present invention improves the accuracy of risk assessment and identification of key accident processes of assisted driving vehicles: an accident risk assessment model is constructed using multivariate refined data in the entire time domain to achieve quantitative assessment of accident risk values, thereby calibrating risk mutation points based on spatiotemporal continuity to achieve identification and extraction of key accident processes of assisted driving vehicles.

[0050] 4. This invention takes into account the potential traffic risks of the assisted driving system itself, improves the field theory-based automobile accident risk assessment and process identification method, and enhances the robustness and applicability of the model. It provides accurate spatial and temporal node identification results of key processes of accidents for assisted driving automobile accident process reconstruction, fact restoration, and hidden danger tracing. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of an embodiment of the present invention.

[0052] Figure 2 4 is a road vehicle layout diagram according to an embodiment of the present invention.

[0053] Figure 3 It is a schematic diagram of constructing a road marking field according to an embodiment of the present invention.

[0054] Figure 4 It is a schematic diagram of constructing a vehicle kinetic energy field according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0056] See also Figure 1 , an embodiment of the present invention comprises the following steps:

[0057] Figure 2 Shown is a schematic diagram of a two-way six-lane road, vehicles and coordinate layout.

[0058] 1) Data collection of assisted driving vehicle accidents: assisted driving vehicle operation data, road condition data, and surrounding environment monitoring data from 90 seconds before to 30 seconds after the accident are collected through assisted driving vehicle event data recording system (EDR), CAN, Mobileye, radar, camera, road test equipment, on-site survey, etc. The specific data types are shown in Table 1. After collection, the data is preprocessed, including: filling missing values ​​of assisted driving vehicle operation data with interpolation, deleting dirty data, and smoothing data to reduce noise values.

[0059] Table 1 Data types of assisted driving car accidents

[0060]

[0061] 2) Driving risk assessment and quantitative analysis: Based on potential field theory, a potential field model for assisted driving vehicle operation safety is established, which includes road marking field and vehicle kinetic energy field. The field force is constructed according to the electric field force formula F=Eq in physics. The potential energy value SPE is used to quantify the driving risk of assisted driving vehicles and calculate the dynamic risk value in the entire time domain. The calculation results form a time series TS.

[0062] (1) The potential field model of assisted driving vehicle operation safety is expressed as follows:

[0063] E S =E R +EV (6)

[0064] In the formula, E S 、E R 、E V They represent the field strength vectors of the safety potential field of the assisted driving vehicle, the road marking field, and the vehicle kinetic energy field respectively;

[0065] (2) Figure 3 The figure shows a schematic diagram of the road marking field construction. The road marking field can be described as:

[0066]

[0067] Among them, E R，ai is the vehicle i in (x i ,y i ) at the time, the field strength vector formed by the lane marking type α, T a is the marking coefficient of different lanes, D is the lane width, r ai =(x i -x a ,y i -y a ) is the coordinate point (x a ,y a ) and the center of mass of vehicle i (x i,y i ), k1 is a constant greater than 0.

[0068] (3) Figure 4 The figure shows a schematic diagram of the vehicle kinetic energy field construction. The vehicle kinetic energy field can be described as:

[0069]

[0070] Among them, E V，ji is the vehicle i in (x i ,y i ) at a certain point on the road (x j ,y j ) formed at the field strength vector, r ji =(x i -x j ,y i -y j ) is the distance between vehicle j and vehicle mass center (x i ,y i ), the distance vector between i is the velocity vector of vehicle i, v j Yes(x j ,y j ), k2 is a constant greater than 0, k3 is the speed weight coefficient, θ is the angle between the distance vector from the source to the target and the target speed vector, σ x , σ y is the convergence coefficient representing the influence range of the potential field.

[0071] (4) The field force is constructed according to the electric field force formula F = Eq in physics, and the k-type field force is recorded as F k , specifically expressed as follows:

[0072]

[0073] In the formula, m i is the actual mass of vehicle i, and α and β are unknown coefficients.

[0074] (5) The potential energy generated by the k-type potential field is denoted as SPE k , specifically expressed as follows:

[0075]

[0076] SPE = max{SPE k} (13)

[0077] According to the accident scenario and data, multiple fixed values ​​of the model are set, and the dynamic risk value (SPE) covering the entire time domain is calculated based on the preprocessed data. According to the calculation results, a time series TS is formed, which contains n SPE values ​​(SPE1, SPE2, ... SPE n ).

[0078] 3) Calibration of risk mutation points: The Mann-Kendall (MK) mutation test method was used in Matlab to determine different risk thresholds at different significance levels and calibrate the corresponding risk mutation points.

[0079] (1) Based on the sequential time series TS and the reverse time series TS1, construct the order sequence TS of the sequential time series TS m ; Define the statistic UF m and UB m , when m=1, UF1=o; E(TS m ), Var(TS m ) are TS m The mean and variance of , the calculation process is specifically expressed as follows:

[0080]

[0081] E(TS m )=n(n+1) / 4 (16)

[0082] Var(TS m )=n(n-1)(2n+5) / 72 (17)

[0083] Repeating the above steps for TS1 based on the reverse time series yields:

[0084] UB m =-UF m (m=n, n-1, ...1).

[0085] (2) Given different significance levels, different confidence intervals are determined. The confidence interval critical value line and the two statistical sequence curves are plotted in a plane rectangular coordinate system. When plotting, since the two curves should have the same time axis when looking for the mutation point, the result statistic UB is reversed according to the time series. m , and get the UB in positive time sequence m , and then draw the graph. If the two curves have an intersection and the intersection is between the critical lines, then the intersection is the risk mutation point, and the corresponding risk value is the threshold.

[0086] 4) Identification of critical accident processes, observing the time-space segmentation results of accidents pointed to by different risk mutation points, and determining the optimal threshold and time-space interval, i.e., the critical accident process. The specific details are as follows:

[0087] The risk mutation points are calibrated according to the selection of the optimal threshold, matched with the key process time-space nodes, and the key process time-space areas pointed by the risk mutation points are identified to realize the identification and extraction of the key processes of assisted driving car accidents.

[0088] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0089] The above embodiments are only used to illustrate the design ideas and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, any equivalent changes or modifications made based on the principles and design ideas disclosed by the present invention are within the protection scope of the present invention.

Claims

1. A method for identifying critical process of assisted driving vehicle accidents based on potential field theory, characterized by: The following steps are involved: S1: Collect and pre-process the accident data of assisted driving vehicles; S2: Based on the preprocessed data, the safety field model of assisted driving vehicle operation is established based on the potential field theory, the driving risk is quantified with the potential energy value, the dynamic risk value covering the entire time domain is calculated and a time series is formed; S3: According to the time series, the MK mutation test method is used to determine different risk thresholds at different significance levels, and the corresponding risk mutation points are marked; S4: Observe the spatiotemporal segmentation results of the accident pointed to by different risk mutation points, determine the optimal threshold and spatiotemporal interval, and identify the key process of the accident.

2. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 1 is characterized in that: In the step S1, The time range of assisted driving car accident data is from 90 seconds before the accident to 30 seconds after the accident; Assisted driving vehicle accident data includes assisted driving vehicle operation data, road condition data, and surrounding environment monitoring data; The sources of assisted driving vehicle accident data include assisted driving vehicle event data recording system EDR, CAN, Mobileye, radar, camera, road test equipment and on-site survey; The preprocessing methods include using interpolation to fill in missing values ​​of assisted driving vehicle operation data, deleting dirty data, and smoothing data to reduce noise.

3. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 1 is characterized in that: In the step S2, the specific steps are: S21: Assume that the field strength vector of the safety potential field of the assisted driving vehicle is E S , the field strength vector of the road marking field is E R , the field strength vector of the vehicle kinetic energy field is E V ; Assume that the coefficient of different lane markings is T a , lane width is D, r ai =(x i -x a ,y i -y a ) is the coordinate point (x a ,y a ) and the center of mass of vehicle i (x i ,y i ) between them; let the velocity vector of vehicle i be v i ,(x j ,y j ) The velocity vector of vehicle j is v j , r ji =(x i -x j ,y i -y j ) is the distance between vehicle j and vehicle mass center (x i ,y i ) between them; based on the potential field theory, the assisted driving vehicle operation safety potential field model including the road marking field and the vehicle kinetic energy field is established as follows: AND S =And R +E V (1); E R =E R (T a ,D,r ai ) (2); E V =E V (r ji ,v i ,v j ,x i ,x j ,y i ,x j ) (3); S22: Let E k is a potential field of type k, m is the actual mass of the target, v is the velocity value of the target, α is an undetermined coefficient, and θ is the angle between the distance vector from the field source to the target and the target velocity vector; Construct the field force according to the electric field force formula F = Eq, and record the k-type field force as F k , the expression is: F k =|E k |·m·e α·v·cosθ (4); S23: Let F be the field force, r be the distance from the field source to the target center of mass; use the potential energy value SPE to quantify the driving risk of the assisted driving car and calculate the dynamic risk value in the full time domain, and the calculation results form a time series TS; the potential energy calculation formula is:

4. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 3 is characterized in that: In the step S21, k1 is set to be a constant greater than 0; the road marking field is described as the vehicle i at (x i ,y i ) when the lane marking line α forms the field intensity vector E R，ai :

5. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 3 is characterized in that: In step S21, k2 is set to be a constant greater than 0, k3 is the speed weight coefficient, σ x , σ y is the convergence coefficient of the potential field influence range, θ is the angle between the distance vector from the field source to the target and the target velocity vector; the vehicle kinetic energy field is described as the vehicle i at (x i ,y i ) at a certain point on the road (x j ,y j ) formed by vehicle j on it V，ji :

6. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 5 is characterized in that: In the step S22, it is assumed that m i is the actual mass of vehicle i, α and β are unknown coefficients; the field force of type k is denoted as F k , specifically expressed as follows:

7. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 6 is characterized in that: In step S23, the potential energy generated by the k-type potential field is recorded as SPE k : According to the accident scenario and data, multiple fixed values ​​of the model are set, and the dynamic risk value SPE covering the entire time domain is calculated based on the preprocessed data: SPE=max{SPE k } (13); According to the calculation results, a time series TS is formed, which contains n SPE values ​​(SPE1, SPE2, ... SPE n ).

8. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 7 is characterized in that: In the step S3, the specific steps are: S31: Let time series TS be a sequential time series, and its reverse time series be TS1; construct the order sequence TS of the sequential time series TS m ; Define the statistic UF m and UB m , when m=1, UF1=0; E(TS m ), Var(TS m ) are TS m The mean and variance of : E(TS m )=n(n+1) / 4 (16), <h2 style=";text-align:left;direction:ltr">Var(TS<h2 style=";text-align:left;direction:ltr"> m <h2 style=";text-align:left;direction:ltr"> )=n(n-1)(2n+5) / 72 (17), Repeat the above steps for TS1 based on the reverse time series and obtain: UB m =-UP m (m=n,n-1,…1); S32: Given different significance levels, determine different confidence intervals, and compare the confidence interval critical value straight line with the two statistical sequences UF obtained in step S31 m and UB m The curves are drawn in a rectangular coordinate system; if the two curves have an intersection and the intersection is between the critical lines, then the intersection is the risk mutation point, and the corresponding risk value is the threshold.

9. The method for identifying critical process of an assisted driving automobile accident based on potential field theory according to claim 8 is characterized in that: In the step S4, the specific steps are: According to step S32, the optimal threshold is selected to calibrate the risk mutation point, the risk mutation point is matched with the key process time-space node, the key process time-space area pointed to by the risk mutation point is identified, and the key accident process of the assisted driving car is identified and extracted.

10. A computer storage medium, characterized in that: A computer program executable by a computer processor is stored therein, and the computer program executes a method for identifying critical processes of an assisted driving automobile accident based on potential field theory as described in any one of claims 1 to 6.

Images