CN114372708A - Unmanned vehicle-oriented driving safety field model construction method - Google Patents
Unmanned vehicle-oriented driving safety field model construction method Download PDFInfo
- Publication number
- CN114372708A CN114372708A CN202210025985.8A CN202210025985A CN114372708A CN 114372708 A CN114372708 A CN 114372708A CN 202210025985 A CN202210025985 A CN 202210025985A CN 114372708 A CN114372708 A CN 114372708A
- Authority
- CN
- China
- Prior art keywords
- road
- unmanned vehicle
- driving
- risk
- expression
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
Abstract
The invention relates to a construction method of a driving safety field model for an unmanned vehicle. The method analyzes the influence of complex driving risk factors on the driving safety of the unmanned vehicle, and establishes a driving safety field model for the unmanned vehicle. It comprises three steps: (1) analyzing and determining the complex driving risk factors of the road according to the real driving environment of the unmanned vehicle; (2) analyzing the parameters of the complex driving risk factors of the road, and calibrating unknown undetermined coefficients; (3) and constructing a dynamic potential energy field and a safety behavior field of the unmanned vehicle to form a driving safety field of the unmanned vehicle. The method can provide basis for the driving safety evaluation of the unmanned vehicle.
Description
Technical Field
The invention provides a driving safety field model construction method for an unmanned vehicle, and aims to accurately evaluate the driving safety of the unmanned vehicle in a comprehensive system, so that a basis is provided for the design of the reliability and safety of the unmanned vehicle. The invention belongs to the technical field of reliability engineering.
Background
With the progress of the autonomy and the intelligent technology, the development of the unmanned vehicle gradually becomes an important index for measuring the technological progress of the country. The reduction of the driving safety risk of the unmanned vehicle becomes the core for guaranteeing the road traffic safety in spite of the safety accidents of the unmanned vehicle which occur in recent years, and the comprehensive and systematic accurate evaluation of the driving safety is a key technology. However, most of the existing driving safety evaluation methods are established based on motion parameters such as distance and time, such evaluation methods cannot adapt to complex traffic situations, the existing driving safety field model is established for manned vehicles or driving auxiliary systems, and the behavior field construction method taking people as the center in the original driving safety field is difficult to apply to unmanned vehicle driving risk evaluation. In view of the above, the invention aims to accurately evaluate the driving risk of the unmanned vehicle, considers the characteristics of three modules of perception, decision and control in an Artificial Intelligence (AI) system of the unmanned vehicle, and designs a construction method for an unmanned vehicle driving safety field model, thereby providing a basis for the reliability and safety design of the unmanned vehicle.
Disclosure of Invention
The invention aims to analyze the influence of complex driving risk factors on the driving safety of an unmanned vehicle, integrate the methods of influence factor analysis and parameter modeling, establish a driving safety field model facing the unmanned vehicle and provide a basis for the driving safety evaluation of the unmanned vehicle.
The invention provides a construction method of a driving safety field model for an unmanned vehicle. The invention mainly comprises the following steps:
the method comprises the following steps: and analyzing and determining the complex driving risk factors of the road according to the real driving environment of the unmanned vehicle.
According to the driving traffic situation of the unmanned vehicle in the real scene, determining a complex driving risk factor influencing the driving safety of the unmanned vehicle as a key index for evaluating the driving risk.
The complex driving risk factor of the unmanned vehicle refers to an index which influences the safe driving of the unmanned vehicle in a real driving scene, and comprises the following steps: equivalent quality, road influence factor, artificial intelligence driving risk factor. This step comprises 3 sub-steps:
step 1: the equivalent mass is determined. The equivalent mass is used for representing the attribute of an object on a road and is related to the physical mass of the object, the geometric dimension of the vehicle and the included angle between the object and the unmanned vehicle.
Step 2: a road impact factor is determined. Road impact factors are used to characterize potential hazards to driving caused by road conditions, including road surface adhesion coefficients, road curvature, grade, and visibility.
And step 3: and determining an artificial intelligence driving risk factor. The artificial intelligence driving risk factor is used for representing the influence of the three processes of perception, decision and control of the AI system on the driving risk, and comprises a perception risk factor, a decision risk factor and a control risk factor.
Step two: and analyzing the parameters of the road complex driving risk factors, and calibrating unknown undetermined coefficients.
And carrying out parameterization processing on the complex driving risk factors of the unmanned lane, collecting real traffic accident data, and solving unknown undetermined coefficients in the parameters. This step comprises 2 sub-steps:
step 1: and carrying out parameterization processing on the complex driving risk factors of the unmanned lane.
(1) Equivalent mass omega of object i on unmanned vehicle lanei(i ═ 1.., n). Severity of collision between unmanned vehicle and object i and equivalent mass omega thereofiRelated, equivalent mass ΩiThe larger the loss caused by collision of an unmanned vehicle with the vehicle.
The equivalent mass contains the following parameters:
firstly, according to the common characteristics, use purpose and function of the vehicle, the vehicle is divided into C types, which are respectively marked as { T }1,T2,...,TcAnd the wheel base of the vehicle type and the vehicle is { alpha }1,α2,...,αcVehicle length { beta }1,β2,...,βcAnd (c) are related. Wherein alpha isc(c=1,2,…C,αc> 0) represents the wheelbase length, beta, of the c-th vehiclec(c=1,2,…C,βc> 0) represents the vehicle length of the c-th vehicle.
Secondly, according to the principle that the larger the physical mass of the object on the road is, the larger the risk caused by collision is, constructing an actual physical mass parameter set { m) of the object on the road1,m2,...,mn}。Wherein m isiRepresenting the actual mass of the object i on the road.
The vehicle body corner of the unmanned vehicle can influence the driving safety when the unmanned vehicle is driven, so that a parameter set of the included angle between an object and the unmanned vehicle is constructed and recorded as { theta [ [ theta ] ]i1, …, n, where θiAnd represents an angle formed by a direction vector of the unmanned vehicle relative to the object i and the speed direction of the object i. And establishing a rectangular coordinate system by taking the intersection point of the road starting point and the road center line as an origin. Marking the tangent line of the road center line passing through the origin as an x-axis, enabling a y-axis to be vertical to the x-axis and to be intersected at the origin, and taking a straight line l of a coordinate plane0Direction vector representing unmanned vehicle,/iDirection vector representing the speed of the object,/0And liThe included angle is the included angle theta between the direction vector of the unmanned vehicle relative to the object i and the speed direction of the objecti。θiThe value range of theta is more than or equal to minus 90 degrees and less than or equal to 90 degrees.
And fourthly, the speed of the object moving on the road surface influences the driving safety of the unmanned vehicle, and the greater the speed is, the greater the risk and loss caused by collision are. Thus, a set of on-road moving object velocity parameters { v ] is constructed1,v2,...,vn}. Wherein v isiIs the velocity of the object i.
Based on this, the equivalent mass Ω of the object i is definediThe specific functional form of (a) is shown in the following formula (1) for describing the influence on the driving risk:
λkthe undetermined coefficient in the equivalent quality function expression can be determined through the substep 2 of the step; k represents a power exponent and is a natural number from 0 to K, and the greater the value of K is, the more accurate the mathematical expression representing the equivalent mass is.
(2) Road influence factor Qj. The severity of the collision between the unmanned vehicle and the object on the road and the condition of the road are determined, and the worse the road condition is, the greater the possibility of the collision between the unmanned vehicle and the object on the road is.
First pathSurface adhesion index muj(j=1,2,…,m,μj> 0) represents the static friction coefficient between the wheel and the road surface of the jth road section. According to the precision requirement of the experiment, the actual driving road is averagely divided into m road sections, and the road adhesion index of each road section is enabled to be mu1,μ2,...,μj,...μm. In the actual traffic environment, the optimal road surface condition is a dry cement road surface, the adhesion coefficient is large, and vehicles are not easy to slip. Therefore, let the standard road surface adhesion coefficient μ*The adhesion coefficient of the j-th road in the actual driving road is set to be mu for the adhesion coefficient of the dry cement road surfacej. The risk evaluation function for obtaining the road adhesion coefficient is shown in (2) below.
Wherein, the adhesion coefficient mu of the j-th road of the actual driving roadjIt can be determined by measuring the braking force at which the wheel brakes are locked and dividing by the load on the wheel.
② road curvature rhoj(j=1,2,…,m,ρj> 0) represents the degree of curve of the j-th driving road, and is used for measuring the degree of flatness of the road. And establishing a rectangular coordinate system by taking the intersection point of the starting road section of the jth road and the road center line as an origin. Recording a tangent line of the road center line passing through the origin as an x-axis, wherein a y-axis is perpendicular to the x-axis and intersects with the origin, and setting a function expression of the j-th road center line as S based on the established road rectangular coordinate systemjIs provided with SjThe expression of (c) is shown in the following formula (3).
Sj=c0+c1x+c2x2+c3x3 (3)
Wherein, c1,c2,c3Is the undetermined coefficient. Road center line SjCan be expressed by sampling points s on the w centerlines1,s2,...,swS coordinate value ofw(xw,yw) Further fitting and determining undetermined coefficients to obtain the road inAnd (4) functional expression of the center line. Then from the formula of curvaturePoint sw(xw,yw) Curvature ρ ofjAs shown in the following formula (4).
Wherein the undetermined coefficient of the road center line of the straight road is c0=0,c1=0,c30, i.e. the straight road always coincides with the x-axis and the curvature is always 0. Thus, the risk assessment function f is exponential as a road curvature factorρ(ρj) As shown in (5) below, namely
③ road gradient τj(j=1,2,…,m,τjGreater than 0) is used for indicating the degree of steepness of the jth road section, the gradient is generally indicated by the ratio of the vertical height of a slope surface between two points to the horizontal width of the slope surface, the vertical height of the slope surface is H, the horizontal width of the slope surface is B, the actual running road is averagely divided into m road sections, and the road gradient of each road section is tau1,τ2,...,τj,...τmThe calculation formula is shown in (6) below.
Wherein, the slope vertical height of the horizontal road is always 0, namely the road slope is 0. Thus, the risk assessment function f is exponential as the road gradientτ(τj) As shown in (7) below, that is
Wherein, the gradient tau of the jth road sectionjThe vertical height of the slope between the j section and the j +1 section of road can be measured and divided by the horizontal width of the slope to obtain the height.
Road visibility deltaj(j=1,2,…,m,δj> 0) is the expression of the influence of the normal bad weather on the j-th road on the driving risk. Let standard visibility delta*For the farthest distance visibility in dry, haze-free and dust-free weather, the road (x) is actually drivenj,yj) Has a visibility of deltaj. The risk evaluation function for obtaining road visibility is shown in the following equation (8).
Wherein, the actual driving road (x)j,yj) Visibility delta ofjThe measurement can be carried out by an atmosphere transmission instrument, and the principle is that a light beam penetrates through an atmosphere column between two fixed points to measure the transmissivity of the atmosphere column, so that the visibility value is calculated.
To sum up, the road impact factor QjThe functional expression of (2) is shown in (9):
(3) artificial intelligence driving risk factor Di
Firstly, according to the principle that the perception of the unmanned vehicle is to classify the objects according to the characteristics of the road objects and the system module, perception risk factors are described by adopting the precision ratio in the machine learning model evaluation method. And defining the precision ratio as the correct classification proportion of the neural network model of the unmanned vehicle perception system. The precision ratio P is expressed by the following formula (10):
wherein, TP is a true example, namely the neural network of the unmanned vehicle perception system identifies the A object as the sample number of the A object; FP is a false positive example, namely the B object is recognized as the number of samples of other objects by the neural network of the unmanned vehicle perception system. Wherein, TP and FP can be obtained by testing the trained machine learning model through the test set. In order to make the change trend of the perception risk factor consistent with the change trend of the driving risk, the perception risk factor is expressed by the difference between 1 and precision ratio, and the expression of the perception risk factor is shown as (11):
secondly, determining the driving safety principle of the unmanned vehicle facing a complex traffic situation according to the accuracy of the unmanned vehicle decision system, evaluating the influence of the unmanned vehicle decision system on the driving safety by adopting a simulation collision rate, and defining the influence as a decision risk factor. The expression of the decision risk factor is shown in (12).
Wherein the content of the first and second substances,the simulation times of the track of the unmanned vehicle are shown,indicating the number of times a collision would occur in the unmanned vehicle simulation.
Analyzing the actual running route and the planned running route of the unmanned vehicle, dividing the track into eta sections according to (eta-1) fixed time differences delta t on the actual running route and the planned running route, obtaining the relative pose difference between each section of actual track and the planned track, and counting errors by adopting mean square error. Based on the above, the influence of the control system of the unmanned vehicle on the driving safety is evaluated by adopting the response accuracy rate, and the influence is defined as a control risk factor. The expression for controlling the risk factor is shown below (13).
Wherein η represents the granularity of trace sampling; trans (d)q) And representing the relative pose difference of each segment of the actual track and the planned track. The larger the mean square error of the relative pose difference is, the larger the deviation of the actual driving track and the planned driving track is, and the larger the driving risk is.
In conclusion, the artificial intelligence driving risk factor D can be obtainediThe functional expression of (c) is shown in (14):
step 2: and determining an unknown undetermined coefficient in the driving risk factor function.
(1) K undetermined coefficients lambda existing in equivalent quality function expressionkAnd (6) calibrating. Equivalent mass omegaiIs the collision risk loss of the unmanned vehicle caused by the object on the road due to the property of the object. Therefore, the risk severity is expressed in terms of the loss caused in the traffic accident, and the average number of deaths from the accident N is defined as the risk loss index, and the expression is shown in (15) below.
Wherein N is0Number of deaths in an accident, NsFor the number of accidents, more deaths from an accident represent greater loss of the accident.
(2) Acquiring accident data including speed, number of accidents and accident death of actual traffic in recent yearsThe number of people, for the unknown undetermined coefficient contained in the expression of the equivalent quality function in the formula (1)It is a polynomial expression used to describe the effect of the velocity component on the driving risk, and thus to establish the relationship between the average number of deaths from accident N and the velocity, as shown in the following equation (16).
(3) Selecting a value of K according to the requirement of experimental precision, wherein the value of K determines the number of polynomial items, the value is a natural number from 0 to K, and the greater the value of K is, the more accurate the mathematical expression representing the equivalent mass is; calculating the average number of dead people N of the accident according to the formula (15), eliminating noise data and generating available data; finally fitting the polynomial back to the polynomial containing the road speed component, and solving to obtain the coefficient lambda to be determined1,λ2...λKThe value of (c).
Step three: and constructing a dynamic potential energy field and a safety behavior field of the unmanned vehicle to form a driving safety field of the unmanned vehicle.
And establishing an unmanned vehicle driving safety field based on the function expression of the complex driving risk factors obtained in the first step and the second step.
This step comprises 3 sub-steps:
step 1: and constructing a dynamic potential energy field for representing the influence of static and moving objects on the road on traffic safety.
Firstly, parameters of influence of static and moving objects on a road on driving safety are defined.
The magnitude and direction of the dynamic potential field intensity is determined by the vector distance r of the objectijEquivalent mass omegaiVelocity viAcceleration aiRoad condition QiThe safe distance l between the vehicle and the vehicle in front of the vehicle and the included angle theta between the direction vector of the unmanned vehicle relative to the object i and the speed direction of the objectiAnd (6) determining.
Wherein, define (x)i,yi) Is the centroid coordinate of the object i, let rij=(xj-xi,yj-yi) Is the vector distance between the objects; a isiIs the acceleration of vehicle i; v. ofiIs the speed of vehicle i; thetaiThe included angle between the direction vector of the unmanned vehicle relative to the object and the speed direction of the object is formed; qiIs a road condition influence factor; omegaiIs equivalent mass; k is a radical of1,k2The undetermined coefficient plays a role in unifying dimensions; l is the safe distance of the vehicle from the preceding vehicle, and rijL and l always satisfy rijA relational expression that l is less than or equal to l.
Secondly, analogy to Hooke's law and influence parameters to obtain a calculation formula of the dynamic potential energy field intensity.
Based on the above analysis of the risk characteristics of the object formation and the characteristics of the vehicle that the closer the vehicle is to the obstacle, the greater the elastic force, an analogy to hooke's law proposes an object i (x)i,yi) Around it (x)j,yj) The calculation formula of the formed dynamic potential field strength is shown as the following formula (17):
step 2: and constructing a safety behavior field for representing the influence of the unmanned vehicle AI system on driving safety.
And on the basis of the mathematical expression of the artificial intelligent driving risk factor in the step two, considering that the safety action field also has the characteristic of field intensity. Definition of unmanned vehicle i (x)i,yi) Around it (x)j,yj) The field strength calculation formula of the formed safety action field is shown as the following formula (18):
and step 3: and constructing an unmanned vehicle driving safety field.
And (3) constructing an unmanned vehicle driving safety field by the dynamic potential energy field formed by the static object and the moving object on the road and the unmanned vehicle safety behavior field based on the step 1 and the step 2.
ES_ij=EV_ij+ED_ij (19)
Wherein E isS_ijRepresenting an object i (x)i,yi) Around it (x)j,yj) The field intensity of the formed driving safety field EV_ijRepresenting an object i (x)i,yi) Around it (x)j,yj) The dynamic potential field strength formed at ED_ijIndicates unmanned vehicle i (x)i,yi) Around it (x)j,yj) The field strength of the security action field formed.
All n objects on the road are in (x)j,yj) The field strength of the generated driving safety field is shown as the following formula (20).
The formula (20) describes the magnitude of the driving risk degree caused by each traffic factor in the actual unmanned vehicle driving scene, and can be used for evaluating the driving risk.
Drawings
FIG. 1 shows a process of a driving safety field model construction method for an unmanned vehicle
Fig. 2 case scenario diagram
FIG. 3 is a schematic diagram of the field intensity of the driving safety field
Detailed Description
The method comprises the following steps: and analyzing and determining the complex driving risk factors of the road according to the real driving environment of the unmanned vehicle.
According to the driving traffic situation of the unmanned vehicle in the real scene, determining a complex driving risk factor influencing the driving safety of the unmanned vehicle as a key index for evaluating the driving risk.
The complex driving risk factor of the unmanned vehicle refers to an index which influences the safe driving of the unmanned vehicle in a real driving scene, and comprises the following steps: equivalent quality, road influence factor, artificial intelligence driving risk factor. This step comprises 3 sub-steps:
step 1: the equivalent mass is determined. The equivalent mass is used for representing the attribute of an object on a road and is related to the physical mass of the object, the geometric dimension of the vehicle and the included angle between the object and the unmanned vehicle.
Step 2: a road impact factor is determined. Road impact factors are used to characterize potential hazards to driving caused by road conditions, including road surface adhesion coefficients, road curvature, grade, and visibility.
And step 3: and determining an artificial intelligence driving risk factor. The artificial intelligence driving risk factor is used for representing the influence of the three processes of perception, decision and control of the AI system on the driving risk, and comprises a perception risk factor, a decision risk factor and a control risk factor.
According to the real driving scene of the unmanned vehicle, factors which may cause risks to the driving safety of the unmanned vehicle are analyzed, the factors are classified, and driving risk factors are determined as shown in the following table 1.
TABLE 1 road Complex Driving Risk factors
Equivalent mass | Road influence factor | Artificial intelligence driving risk factor |
Physical quality | Coefficient of road surface adhesion | Perceptual risk factors |
Vehicle geometry | Curvature of road | Decision risk factor |
Angle between object and unmanned vehicle | Slope of slope | Controlling risk factors |
/ | Visibility | / |
Step two: and analyzing the parameters of the road complex driving risk factors, and calibrating unknown undetermined coefficients.
And carrying out parameterization processing on the complex driving risk factors of the unmanned lane, collecting real traffic accident data, and solving unknown undetermined coefficients in the parameters. This step comprises 2 sub-steps:
step 1: and carrying out parameterization processing on the complex driving risk factors of the unmanned lane.
(1) Equivalent mass omega of object i on unmanned vehicle lanei(i ═ 1.., n). Severity of collision between unmanned vehicle and object i and equivalent mass omega thereofiRelated, equivalent mass ΩiThe larger the loss caused by collision of an unmanned vehicle with the vehicle.
The equivalent mass contains the following parameters:
firstly, according to the common characteristics, use purpose and function of the vehicle, the vehicle is divided into C types, which are respectively marked as { T }1,T2,...,TcAnd the wheel base of the vehicle type and the vehicle is { alpha }1,α2,...,αcVehicle length { beta }1,β2,...,βcAnd (c) are related. Wherein alpha isc(c=1,2,…C,αc> 0) represents the wheelbase length, beta, of the c-th vehiclec(c=1,2,…C,βc> 0) represents the vehicle length of the c-th vehicle.
Secondly, according to the principle that the larger the physical mass of the object on the road is, the larger the risk caused by collision is, constructing an actual physical mass parameter set { m) of the object on the road1,m2,...,mn}. Wherein m isiRepresenting the actual mass of the object i on the road.
The vehicle body corner of the unmanned vehicle can influence the driving safety when the unmanned vehicle is driven, so that a parameter set of the included angle between an object and the unmanned vehicle is constructed and recorded as { theta [ [ theta ] ]i1, …, n, where θiAnd represents an angle formed by a direction vector of the unmanned vehicle relative to the object i and the speed direction of the object i. And establishing a rectangular coordinate system by taking the intersection point of the road starting point and the road center line as an origin. Marking the tangent line of the road center line passing through the origin as an x-axis, enabling a y-axis to be vertical to the x-axis and to be intersected at the origin, and taking a straight line l of a coordinate plane0Direction vector representing unmanned vehicle,/iDirection vector representing the speed of the object,/0And liThe included angle is the included angle theta between the direction vector of the unmanned vehicle relative to the object i and the speed direction of the objecti。θiThe value range of theta is more than or equal to minus 90 degrees and less than or equal to 90 degrees.
And fourthly, the speed of the object moving on the road surface influences the driving safety of the unmanned vehicle, and the greater the speed is, the greater the risk and loss caused by collision are. Thus, a set of on-road moving object velocity parameters { v ] is constructed1,v2,...,vn}. Wherein v isiIs the velocity of the object i.
Based on this, the equivalent mass Ω of the object i is definediIs shown as the following formula (21) for describing the influence on the driving risk:
λkthe undetermined coefficient in the equivalent quality function expression can be determined through the substep 2 of the step: k represents a power exponent and is a natural number from 0 to K, and the greater the value of K is, the more accurate the mathematical expression representing the equivalent mass is.
(2) Road influence factor Qj. The severity of the collision between the unmanned vehicle and the object on the road and the condition of the road are determined, and the worse the road condition is, the greater the possibility of the collision between the unmanned vehicle and the object on the road is.
Road surface adhesion index muj(j=1,2,…,m,μj> 0) represents the static friction coefficient between the wheel and the road surface of the jth road section. According to the precision requirement of the experiment, the actual driving road is averagely divided into m road sections, and the road adhesion index of each road section is enabled to be mu1,μ2,...,μj,...μm. In the actual traffic environment, the optimal road surface condition is a dry cement road surface, the adhesion coefficient is large, and vehicles are not easy to slip. Therefore, let the standard road surface adhesion coefficient μ*The adhesion coefficient of the j-th road in the actual driving road is set to be mu for the adhesion coefficient of the dry cement road surfacej. The risk evaluation function for obtaining the road surface adhesion coefficient is shown in (22) below.
Wherein, the adhesion coefficient mu of the j-th road of the actual driving roadjIt can be determined by measuring the braking force at which the wheel brakes are locked and dividing by the load on the wheel.
② road curvature rhoj(j=1,2,…,m,ρj> 0) represents the degree of curve of the j-th driving road, and is used for measuring the degree of flatness of the road. And establishing a rectangular coordinate system by taking the intersection point of the starting road section of the jth road and the road center line as an origin. Recording a tangent line of the road center line passing through the origin as an x-axis, wherein a y-axis is perpendicular to the x-axis and intersects with the origin, and setting a function expression of the j-th road center line as S based on the established road rectangular coordinate systemjIs provided with SjThe expression of (c) is shown in the following formula (23).
Sj=c0+c1x+c2x2+c3x3 (23)
Wherein, c1,c2,c3Is the undetermined coefficient. Road center line SjCan be expressed by sampling points s on the w centerlines1,s2,...,swS coordinate value ofw(xw,yw) Further fitting to determine the coefficients to be determined, therebyAnd obtaining a function expression of the central line of the road. Then from the formula of curvaturePoint sw(xw,yw) Curvature ρ ofjAs shown in the following formula (24).
Wherein the undetermined coefficient of the road center line of the straight road is c0=0,c1=0,c30, i.e. the straight road always coincides with the x-axis and the curvature is always 0. Thus, the risk assessment function f is exponential as a road curvature factorρ(ρj) As shown in (25) below, that is
③ road gradient τj(j=1,2,…,m,τjGreater than 0) is used for indicating the degree of steepness of the jth road section, the gradient is generally indicated by the ratio of the vertical height of a slope surface between two points to the horizontal width of the slope surface, the vertical height of the slope surface is H, the horizontal width of the slope surface is B, the actual running road is averagely divided into m road sections, and the road gradient of each road section is tau1,τ2,...,τj,...τmThe calculation formula is shown in (26) below.
Wherein, the slope vertical height of the horizontal road is always 0, namely the road slope is 0. Thus, the risk assessment function f is exponential as the road gradientτ(τj) As shown in (27) below, that is
Wherein, the gradient tau of the jth road sectionjThe vertical height of the slope between the j section and the j +1 section of road can be measured and divided by the horizontal width of the slope to obtain the height.
Road visibility deltaj(j=1,2,…,m,δj> 0) is the expression of the influence of the normal bad weather on the j-th road on the driving risk. Let standard visibility delta*For the farthest distance visibility in dry, haze-free and dust-free weather, the road (x) is actually drivenj,yj) Has a visibility of deltaj. The risk evaluation function for obtaining road visibility is shown as the following equation (28).
Wherein, the actual driving road (x)j,yj) Visibility delta ofjThe measurement can be carried out by an atmosphere transmission instrument, and the principle is that a light beam penetrates through an atmosphere column between two fixed points to measure the transmissivity of the atmosphere column, so that the visibility value is calculated.
To sum up, the road impact factor QjThe functional expression of (2) is shown in (29):
(3) artificial intelligence driving risk factor Di
Firstly, according to the principle that the perception of the unmanned vehicle is to classify the objects according to the characteristics of the road objects and the system module, perception risk factors are described by adopting the precision ratio in the machine learning model evaluation method. And defining the precision ratio as the correct classification proportion of the neural network model of the unmanned vehicle perception system. The precision ratio P is expressed by the following formula (30):
wherein, TP is a true example, namely the neural network of the unmanned vehicle perception system identifies the A object as the sample number of the A object; FP is a false positive example, namely the B object is recognized as the number of samples of other objects by the neural network of the unmanned vehicle perception system. Wherein, TP and FP can be obtained by testing the trained machine learning model through the test set. In order to make the change trend of the perception risk factor consistent with the change trend of the driving risk, the perception risk factor is expressed by the difference between 1 and precision ratio, and the expression of the perception risk factor is shown as (31):
secondly, determining the driving safety principle of the unmanned vehicle facing a complex traffic situation according to the accuracy of the unmanned vehicle decision system, evaluating the influence of the unmanned vehicle decision system on the driving safety by adopting a simulation collision rate, and defining the influence as a decision risk factor. The expression of the decision risk factor is shown in (32).
Wherein the content of the first and second substances,the simulation times of the track of the unmanned vehicle are shown,indicating the number of times a collision would occur in the unmanned vehicle simulation.
Analyzing the actual running route and the planned running route of the unmanned vehicle, dividing the track into eta sections according to (eta-1) fixed time differences delta t on the actual running route and the planned running route, obtaining the relative pose difference between each section of actual track and the planned track, and counting errors by adopting mean square error. Based on the above, the influence of the control system of the unmanned vehicle on the driving safety is evaluated by adopting the response accuracy rate, and the influence is defined as a control risk factor. The expression for controlling the risk factor is shown below (33).
Wherein η represents the granularity of trace sampling; trans (d)q) And representing the relative pose difference of each segment of the actual track and the planned track. The larger the mean square error of the relative pose difference is, the larger the deviation of the actual driving track and the planned driving track is, and the larger the driving risk is.
In conclusion, the artificial intelligence driving risk factor D can be obtainediThe functional expression of (2) is shown in (34):
step 2: and determining an unknown undetermined coefficient in the driving risk factor function.
(1) K undetermined coefficients lambda existing in equivalent quality function expressionkAnd (6) calibrating. Equivalent mass omegaiIs the collision risk loss of the unmanned vehicle caused by the object on the road due to the property of the object. Therefore, the risk severity is expressed in terms of the loss caused in the traffic accident, and the average number of deaths from the accident N is defined as a risk loss index, and the expression is as shown in (35) below.
Wherein N is0Number of deaths in an accident, NsFor the number of accidents, more deaths from an accident represent greater loss of the accident.
(2) Acquiring accident data including speed and accident of actual traffic in recent yearsNumber, number of accident deaths, unknown undetermined coefficient contained in the expression for equivalent quality function in formula (21)It is a polynomial expression used to describe the effect of the velocity component on the driving risk, and thus to establish the relationship between the average number of deaths from accident N and the velocity, as shown in equation (36) below.
(3) Selecting a value of K according to the requirement of experimental precision, wherein the value of K determines the number of polynomial items, the value is a natural number from 0 to K, and the greater the value of K is, the more accurate the mathematical expression representing the equivalent mass is; calculating the average number of dead people N of the accident according to the formula (35), eliminating noise data and generating available data; finally fitting the polynomial back to the polynomial containing the road speed component, and solving to obtain the coefficient lambda to be determined1,λ2...λKThe value of (c).
According to the expression of the equivalent quality, the road influence factor and the artificial intelligence driving risk factor, parameterizing the equivalent quality, the road influence factor and the artificial intelligence driving risk factor to obtain an actual value of the parameter, as shown in table 2 below.
Table 2 parameterized example
The average number of dead people N in recent years, and the vehicle running speed v and the included angle θ between the direction vector of the object and the speed direction of the object at the time of the incident are collected, and the data in table 3 below are obtained.
TABLE 3 average number of deaths in recent years, vehicle speed and included angle
N/man | v/(m/s) | θ/° |
11 | 104.6 | 3.3 |
11 | 100.6 | 33.4 |
7 | 72.2 | 21.4 |
6 | 60.1 | 43.9 |
6 | 61.5 | 26.7 |
6 | 63.5 | 24.8 |
14 | 116.6 | 0.5 |
19 | 120.2 | 28.8 |
15 | 113.3 | 13.4 |
9 | 87.2 | 33.7 |
K is taken as 2, quadratic function fitting is carried out on the data in the table above to obtain lambda2,λ1,λ0The final equivalent masses are expressed by the following equation (37), which is-0.36, 13.44 and 3.49 respectively:
Ωi=Ti·mi·[(-0.36vi 2+13.44vi+3.49)cosθi+1] (37)
step three: and constructing a dynamic potential energy field and a safety behavior field of the unmanned vehicle to form a driving safety field of the unmanned vehicle.
And establishing an unmanned vehicle driving safety field based on the function expression of the complex driving risk factors obtained in the first step and the second step.
This step comprises 3 sub-steps:
step 1: and constructing a dynamic potential energy field for representing the influence of static and moving objects on the road on traffic safety.
Firstly, parameters of influence of static and moving objects on a road on driving safety are defined.
The magnitude and direction of the dynamic potential field intensity is determined by the vector distance r of the objectijEquivalent mass omegaiVelocity viAcceleration aiRoad condition QiThe safe distance l between the vehicle and the vehicle in front of the vehicle and the included angle theta between the direction vector of the unmanned vehicle relative to the object i and the speed direction of the objectiAnd (6) determining.
Wherein, define (x)i,yi) Is the centroid coordinate of the object i, let rij=(xj-xi,yj-yi) Is the vector distance between the objects; a isiIs the acceleration of vehicle i; v. ofiIs the speed of vehicle i; thetaiThe included angle between the direction vector of the unmanned vehicle relative to the object and the speed direction of the object is formed; qiIs a road condition influence factor; omegaiIs equivalent mass; k is a radical of1,k2The undetermined coefficient plays a role in unifying dimensions; l is the safe distance of the vehicle from the preceding vehicle, and rijL and l always satisfy rijA relational expression that l is less than or equal to l.
Secondly, analogy to Hooke's law and influence parameters to obtain a calculation formula of the dynamic potential energy field intensity.
Based on the above analysis of the risk characteristics of the object formation and the characteristics of the vehicle that the closer the vehicle is to the obstacle, the greater the elastic force, an analogy to hooke's law proposes an object i (x)i,yi) Around it (x)j,yj) The calculation formula of the formed dynamic potential field strength is shown as the following formula (38):
step 2: and constructing a safety behavior field for representing the influence of the unmanned vehicle AI system on driving safety.
And on the basis of the mathematical expression of the artificial intelligent driving risk factor in the step two, considering that the safety action field also has the characteristic of field intensity. Definition of unmanned vehicle i (x)i,yi) Around it (x)j,yj) The field strength calculation formula of the formed safety action field is shown as the following formula (39):
and step 3: and constructing an unmanned vehicle driving safety field.
And (3) constructing an unmanned vehicle driving safety field by the dynamic potential energy field formed by the static object and the moving object on the road and the unmanned vehicle safety behavior field based on the step 1 and the step 2.
ES_ij=EV_ij+ED_ij (40)
Wherein E isS_ijRepresenting an object i (x)i,yi) Around it (x)j,yj) The field intensity of the formed driving safety field EV_ijRepresenting an object i (x)i,yi) Around it (x)j,yj) The dynamic potential field strength formed at ED_ijIndicates unmanned vehicle i (x)i,yi) Around it (x)j,yj) The field strength of the security action field formed.
All n objects on the road are in (x)j,yj) The field strength of the traveling safety field generated at this point is shown by the following formula (41).
Formula (41) describes the magnitude of the driving risk degree caused by each traffic factor in the actual unmanned vehicle driving scene, and can be used for evaluating the driving risk.
To establish an unmanned vehicle driving safety field, case scenes are designed and explained on a standard two-lane road surface. The schematic diagram of the case scene is shown in fig. 2, in which the light-colored vehicle is an unmanned vehicle and is located at (x)j,yj). The dark vehicle is another vehicle 1 running on the road, the vehicle 1 and the unmanned vehicle respectively run on adjacent lanes, and a static road barrier 2 exists in front of the vehicle 1, so that the vehicle 1 tries to turn right to avoid collision. Vector distance r between vehicle 1 and roadblock 2 and unmanned vehicleijEquivalent mass omegaiVelocity viAcceleration aiRoad condition QjThe set safe distance l and the included angle theta between the direction vector of the unmanned vehicle relative to the vehicle 1 and the speed direction of the objectiThe values of (b) are shown in tables 4 and 5 below.
Table 4 selected case scenarios the values of the parameters associated with vehicle 1
Parameter(s) | Value of |
|
3m |
Ω1 | 106057.54 |
|
20m/s |
a1 | 2m/s2 |
Qj | 2.91 |
l | 6m |
θ1 | 30° |
Table 5 values of parameters associated with selected case scenario barrier 2
Parameter(s) | Value of |
r2j | 5m |
Ω2 | 60.98 |
v2 | 0m/s |
a2 | 0m/s2 |
Qj | 2.91 |
Let k1,k2To 1, the vehicle 1 is calculated as being in (x)j,yj) The field strength of the dynamic potential energy field is:
and the field intensity of the safety action field is calculated as follows:
ED_1j=EV_1j·Dj=1.575×108 (43)
get vehicle 1 at (x)j,yj) The field intensity of the driving safety field generated at the position is as follows:
ES_1j=ED_1j+EV_1j=8.089×108 (44)
similarly, the calculation results in that the roadblock 2 is in (x)j,yj) The field strength of the dynamic potential energy field is:
and calculates the position (x) of the roadblock 2 according to the calculated positionj,yj) The field strength of the safety action field is as follows:
ED_2j=EV_2j·Dj=46.137468 (46)
get the roadblock 2 at (x)j,yj) The field intensity of the driving safety field generated at the position is as follows:
ES_2j=ED_2j+EV_2j=223.5893 (47)
in summary, the object on the road can be found to be (x)j,yj) The field strength of the driving safety field generated at the position is shown in a formula (48), and the schematic diagram of the driving safety field is shown in fig. 3.
Claims (1)
1. A construction method of a driving safety field model for an unmanned vehicle is characterized by comprising the following steps:
the method comprises the following steps: analyzing and determining the complex driving risk factors of the road according to the real driving environment of the unmanned vehicle;
determining a complex driving risk factor which has influence on the driving safety of the unmanned vehicle according to the driving traffic situation of the unmanned vehicle in a real scene, and taking the complex driving risk factor as a key index for evaluating the driving risk;
the complex driving risk factor of the unmanned vehicle refers to an index which influences the safe driving of the unmanned vehicle in a real driving scene, and comprises the following steps: equivalent quality, road influence factors and artificial intelligence driving risk factors;
step two: analyzing the parameters of the complex driving risk factors of the road, and calibrating unknown undetermined coefficients;
carrying out parameterization processing on the complex driving risk factors of the unmanned lane, collecting real traffic accident data, and solving unknown undetermined coefficients in the parameters;
step 1: carrying out parameterization processing on the complex driving risk factors of the unmanned lane;
(1) equivalent mass omegaiParameterizing;
firstly, according to the common characteristics, use purpose and function of the vehicle, the vehicle is divided into C types, which are respectively marked as { T }1,T2,...,Tc};
Secondly, constructing a parameter set (m) of the actual physical quality of the object on the road1,m2,...,mnIn which m isiRepresenting the actual mass of an object i on the road;
thirdly, a parameter set of the included angle between the object and the unmanned vehicle is constructed and is recorded as thetai1, …, n, where θiRepresenting an included angle formed by a direction vector of the unmanned vehicle relative to the object i and the speed direction of the object i;
based on this, the equivalent mass Ω of the object i is definediThe specific functional form of (a) is shown in the following formula (1) for describing the influence on the driving risk:
λkis undetermined coefficient in equivalent quality function expression, K represents power exponent, and the value is natural number from 0 to K;
(2) influence factor Q on roadjParameterizing;
road surface adhesion index muj(j=1,2,…,m,μj> 0) represents the static friction coefficient between the wheel and the road surface of the jth road section, the actual running road is averagely divided into m road sections, and the road adhesion index of each road section is enabled to be mu1,μ2,...,μj,...μm(ii) a Make the standard road surface adhesion coefficient mu*The adhesion coefficient of the j-th road in the actual driving road is set to be mu for the adhesion coefficient of the dry cement road surfacejThe risk evaluation function of the road adhesion coefficient is as follows (2):
fμ(μj)=μ*/μj (2)
② road curvature rhoj(j=1,2,…,m,ρjGreater than 0) represents the bending degree of the jth section of road, a rectangular coordinate system is established by taking the intersection point of the jth section of road starting section and the road center line as the origin, the tangent line of the road center line passing through the origin is recorded as an x-axis, a y-axis is perpendicular to the x-axis and intersects at the origin, and the function expression of the jth section of road center line is SjIs provided with SjThe expression of (c) is shown in the following formula (3):
Sj=c0+c1x+c2x2+c3x3 (3)
wherein, c1,c2,c3For undetermined coefficients, the coordinate values s of points on w central lines are sampledw(xw,yw) Fitting to determine undetermined coefficient, obtaining a function expression of the road center line, and obtaining a point s by a curvature formulaw(xw,yw) Curvature ρ ofjAnd taking the index form as a risk evaluation function f of the road curvature factorρ(ρj) As shown in the following formula (4):
③ road gradient τj(j=1,2,…,m,τjGreater than 0) is used for indicating the degree of steepness of the jth road section, the vertical height of the slope is H, the horizontal width is B, and the road gradient of each road section is tau1,τ2,...,τj,...τmAnd exponentially as a risk evaluation function f of the road gradientτ(τj) The calculation formula is shown as (5) below:
road visibility deltaj(j=1,2,…,m,δjGreater than 0) shows the influence of normal severe weather on the driving risk on the jth road, and ensures the standard visibility delta*For the farthest distance visibility in dry, haze-free and dust-free weather, the road (x) is actually drivenj,yj) Has a visibility of deltajThe risk evaluation function of road visibility is shown as the following formula (6):
fδ(δj)=δ*/δj (6)
to sum up, the road impact factor QjThe functional expression of (2) is shown in (7):
(3) artificial intelligence driving risk factor Di
Defining precision ratio as the correct classification proportion of a neural network model of the unmanned vehicle perception system, wherein the expression of the precision ratio P is shown as the following formula (8):
wherein, TP is a true case, namely the unmanned vehicle perception system neural network recognizes the A object as the number of samples of the A object, FP is a false case, namely the unmanned vehicle perception system neural network recognizes the B object as the number of samples of the A object; the TP and the FP can be obtained by testing a trained machine learning model through a test set, the perception risk factor is expressed by the difference between 1 and precision ratio, and the expression of the perception risk factor is shown as (9):
secondly, evaluating the influence of a decision system of the unmanned vehicle on driving safety by adopting a simulation collision rate, and defining the influence as a decision risk factor, wherein the expression of the decision risk factor is shown as (10):
wherein the content of the first and second substances,the simulation times of the track of the unmanned vehicle are shown,representing the number of times of collision in the unmanned vehicle simulation;
thirdly, the influence of the control system of the unmanned vehicle on the driving safety is evaluated by adopting the response accuracy rate, and is defined as a control risk factor, and the expression of the control risk factor is shown as the following (11):
where η represents the granularity of the trace sample, trans (d)q) Representing the relative pose difference of each actual track and the planning track;
in conclusion, the artificial intelligence driving risk factor D can be obtainediThe functional expression of (2) is shown in (12):
step 2: determining an unknown undetermined coefficient in the driving risk factor function;
(1) defining the average number of death people N in the accident as a risk loss index, and expressing the following expression (13):
wherein N is0Number of deaths in an accident, NsThe number of accidents;
(2) acquiring accident data of actual traffic, wherein the accident data comprises speed, accident number and accident death number, and establishing the relation between the average accident death number and the speed, as shown in the following formula (14):
(3) selecting the value of K, calculating the average number of dead people N of the accident, eliminating noise data, generating available data, fitting back to a polynomial containing the road speed component, and solving to obtain a waiting coefficient lambda1,λ2...λKA value of (d);
step three: constructing a dynamic potential energy field and a safety behavior field of the unmanned vehicle to form a driving safety field of the unmanned vehicle;
step 1: constructing a dynamic potential energy field for representing the influence of static and moving objects on a road on driving safety;
firstly, defining parameters of influence of static and moving objects on a road on driving safety;
the magnitude and direction of the dynamic potential field intensity is determined by the vector distance r of the objectijEquivalent mass omegaiVelocity viAcceleration aiRoad condition QiThe safe distance l and the included angle theta between the direction vector of the unmanned vehicle relative to the object i and the speed direction of the objectiDetermining;
wherein, define (x)i,yi) Is the centroid coordinate of the object i, let rij=(xj-xi,yj-yi) Is the vector distance between the objects; a isiIs the acceleration of vehicle i; v. ofiIs the speed of vehicle i; k is a radical of1,k2Is the undetermined coefficient; and rijL and l always satisfy rijA relational expression that l is less than or equal to l;
secondly, analogy to Hooke's law and influence parameters to obtain a calculation formula of the dynamic potential energy field intensity;
analogy Hooke's law proposes that an object i (x)i,yi) Around it (x)j,yj) The calculation formula of the potential energy field intensity formed is shown as the following formula (15):
step 2: constructing a safety behavior field for representing the influence of the unmanned vehicle AI system on driving safety;
defining unmanned vehicle i (x) based on artificial intelligence driving risk factor mathematical expressioni,yi) Around it (x)j,yj) The field strength calculation formula of the formed safety action field is shown as the following formula (16):
and step 3: constructing an unmanned vehicle driving safety field;
with ESIndicating the field strength of the driving safety field, EVRepresenting the field strength of the dynamic potential field, EDThe field intensity of the safe behavior field is represented, a mathematical model of the driving safe field can be represented by an expression (17), and the mathematical model (17) describes the driving risk degree caused by each traffic factor in the actual unmanned vehicle driving scene, and can be used for evaluating the driving risk.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210025985.8A CN114372708A (en) | 2022-01-11 | 2022-01-11 | Unmanned vehicle-oriented driving safety field model construction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210025985.8A CN114372708A (en) | 2022-01-11 | 2022-01-11 | Unmanned vehicle-oriented driving safety field model construction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114372708A true CN114372708A (en) | 2022-04-19 |
Family
ID=81143136
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210025985.8A Pending CN114372708A (en) | 2022-01-11 | 2022-01-11 | Unmanned vehicle-oriented driving safety field model construction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114372708A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114842660A (en) * | 2022-06-29 | 2022-08-02 | 石家庄铁道大学 | Unmanned lane track prediction method and device and electronic equipment |
-
2022
- 2022-01-11 CN CN202210025985.8A patent/CN114372708A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114842660A (en) * | 2022-06-29 | 2022-08-02 | 石家庄铁道大学 | Unmanned lane track prediction method and device and electronic equipment |
CN114842660B (en) * | 2022-06-29 | 2022-10-11 | 石家庄铁道大学 | Unmanned lane track prediction method and device and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104164829B (en) | Detection method of road-surface evenness and intelligent information of road surface real-time monitoring system based on mobile terminal | |
Baffet et al. | An observer of tire–road forces and friction for active security vehicle systems | |
CN112389440B (en) | Vehicle driving risk prediction method in off-road environment based on vehicle-road action mechanism | |
Deng et al. | A probabilistic model for driving-style-recognition-enabled driver steering behaviors | |
CN109649396B (en) | Safety detection method for commercial vehicle driver | |
CN115027464A (en) | Automatic driving collision detection method based on direction bounding box | |
CN114372708A (en) | Unmanned vehicle-oriented driving safety field model construction method | |
Hu et al. | Autonomous vehicle’s impact on traffic: Empirical evidence from Waymo Open Dataset and implications from modelling | |
CN108445866B (en) | LDW false negative report test method and test system based on convolutional neural network | |
CN111813083A (en) | Scene task complexity quantification model | |
Tota et al. | An intelligent predictive algorithm for the anti-rollover prevention of heavy vehicles for off-road applications | |
CN108528453A (en) | It is a kind of towards collaborative truck information uncertainty with control method for vehicle of speeding | |
Qi et al. | Driving risk detection model of deceleration zone in expressway based on generalized regression neural network | |
CN114898042A (en) | Time-space urgency-based human-vehicle collision danger situation prediction method | |
Widner et al. | Framework for vehicle dynamics model validation | |
Ramezani-Khansari et al. | Estimating lane change duration for overtaking in nonlane-based driving behavior by local linear model trees (LOLIMOT) | |
CN109406076A (en) | A method of beam bridge structure damage reason location is carried out using the mobile principal component of displacement sensor array output | |
CN113642845B (en) | Quality evaluation method for road traffic perception track data | |
CN114987495A (en) | Man-machine hybrid decision-making method for highly automatic driving | |
CN113743483B (en) | Road point cloud error scene analysis method based on spatial plane offset analysis model | |
Jian et al. | A robust bridge weigh‐in‐motion algorithm based on regularized total least squares with axle constraints | |
Zhou et al. | A hybrid virtual–real traffic simulation approach to reproducing the spatiotemporal distribution of bridge loads | |
CN113515813A (en) | On-site verification method for simulation reliability of automobile dynamics simulation software | |
Mi et al. | Binocular vision vehicle environment collision early warning method based on machine learning | |
CN111310309A (en) | Unmanned aerial vehicle-based thermal infrared image temperature inversion correction method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |