Driving risk identification and avoidance area generation method and system under multiple constraints

Abstract

The invention discloses a driving risk identification and avoidance area generation method and system under multiple constraints. The method comprises the following steps: step 1, acquiring traffic environment state information of a current road where a vehicle is located in real time; 2, acquiring traffic rules and all obstacles of the current traffic environment in real time, and judging whether the traffic rules and the obstacles acquired in the current environment are validity constraints or not; step 3, identifying the malignant driving risk of the other vehicle to the own vehicle under the hard constraint by using an automatic driving multi-constraint risk field model considering the risk source and the risk gradient change; 4, the driving danger of the vehicle is evaluated, if the evaluation result reaches or exceeds a preset value, the soft constraint needs to be broken through, and the step 5 is executed; and step 5, obtaining a risk area, a low-risk area and a safe and feasible area of the vehicle through the driving safety field, performing path planning in the safe and feasible area, and selecting an optimal path. According to the invention, the driving safety can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, in particular to a method and system for identifying and generating an evading area of a driving risk under multiple constraints. BACKGROUND

[0002] When a vehicle encounters a collision with another vehicle under strong constraints of traffic rules, it often causes a vicious accident, such as Figure 3 As shown in the figure, the front vehicles are respectively constrained by red lights and solid lines in the longitudinal and lateral directions and cannot change lanes at will to avoid. Since the driver's attention is mostly focused on the front of the vehicle, and the traffic regulations determine that the rear-end collision accident of the rear vehicle is the full responsibility of the rear vehicle, the driver usually does not deliberately pay attention to the rear vehicle in this scenario, and it is difficult to identify the rear danger in time. The automatic driving vehicle traffic safety white paper published by China Automotive Technology Research Center Co., Ltd., Tongji University and Baidu Apollo shows that the proportion of accidents caused by other road traffic participants is close to 70%, and the main reason is that the automatic driving vehicle is kept at the intersection and is hit by other road traffic participants. It can be seen that although the automatic driving vehicle can detect the rear vehicle in time through the sensor, it is difficult to take measures to accelerate or change lanes to avoid rear-end collision under strong constraints of traffic rules (red light and solid line).

[0003] From the existing technology, the current driving risk modeling theory and method cannot predict the driving risk loss and accurately quantify the driving risk danger degree. When dealing with the vicious risk under the constraints of traffic rules and roadblocks, the existing technology cannot determine whether it is necessary to temporarily break through the constraints of traffic regulations in some cases, nor can it effectively determine when to break through to avoid potential fatal accidents.

[0004] Therefore, it is necessary to study the integrated technology of identification, quantification and evaluation of vicious driving risk, solve the problem of judgment of soft constraint breaking of traffic rules in high-risk environment, and reduce the possibility of fatal accidents. It has become a problem that the technical staff in the current technical field urgently needs to solve. SUMMARY

[0005] The purpose of the present application is to provide a method and system for identifying and generating an evading area of a driving risk under multiple constraints of traffic rules and roadblocks, and improving driving safety.

[0006] To achieve the above purpose, the present application provides a method for identifying and generating an evading area of a driving risk under multiple constraints, which comprises:

[0007] Step 1, real-time acquisition of the current road traffic environment state information of the ego vehicle;

[0008] Step two, according to the current road traffic environment state information, real-time acquisition of the traffic rules and all obstacles involved in the current traffic environment, judge whether the traffic rules and obstacles obtained by the current environment are effective constraints;

[0009] Step three, according to the current road traffic environment state information obtained in step one and the judgment result of the traffic rules and obstacle effectiveness constraints obtained by the current environment in step two, obtain the soft constraint related to the traffic rules and the hard constraint related to the roadblock, use the automatic driving multi-constraint risk field model considering the risk source and the risk gradient change to identify the malignant driving risk of the other car to the car under the hard constraint, wherein the automatic driving multi-constraint risk field model is described by formula (1) or (2) or (3):

[0010]

[0011]

[0012]

[0013] In the formula, v i and v j are the speed vectors of the car i and the car j, F ji is the risk field force of the car j to the car i, by converting the earth coordinate system XOY to the coordinate system x ji jy ji , the direction of F ji is set to the direction of the vector relative speed v ji , the coordinate system x ji jy ji is the coordinate system with the car j as the origin, the positive direction of the horizontal axis x ji j is the direction of the vector relative speed v ji , the positive direction of the vertical axis jy ji is the direction of the acute angle with the speed vector v j of the car j and perpendicular to the horizontal axis x ji j, m i and m j are the masses of the car i and the car j, r e is the unit length, r0 is the following distance, k x and k y are the longitudinal and lateral gradient coefficients, x′ ji and y′ ji are the longitudinal and lateral relative distances of the car i and the car j in the coordinate system x ji jy ji , r ji is the distance between the car i and the car j in the longitudinal direction, r max and r minrespectively, are the maximum and minimum distances of the vehicle in free flow;

[0014] Step four, F ji is obtained according to F ji obtained in step three, the driving risk of the ego vehicle is evaluated, and if the evaluation result reaches or exceeds a preset value, the soft constraint needs to be broken through and step five is entered;

[0015] Step five, F ji is obtained according to F ji obtained in step three, the risk region, the low-risk region and the safe feasible region of the ego vehicle are obtained through the driving safety field, and the optimal path is selected in the safe feasible region.

[0016] Further, the method of "evaluating the driving risk of the ego vehicle" in step four specifically includes: ji

[0017] F i is calculated, a i is obtained according to F ji / m i , a i is compared with a i，max preset value a i , if a i，max ≥ a si , it is determined that the evaluation result reaches or exceeds the preset value.

[0018] Another method specifically includes:

[0019] F si is calculated, F si is obtained as a preset value, F ji is compared with a preset value F si , if F ji > wF si , it is determined that the evaluation result exceeds the preset value, wherein w is an adjustment coefficient obtained by calibrating real road traffic accident data in the same scene.

[0020] Further, a i，max represents a virtual acceleration corresponding to a serious traffic accident.

[0021] Further, F si is calculated by formula (5):

[0022]

[0023] In the formula, R si represents the constraint force of the red light of the traffic signal on the vehicle i, and is described as R si =m i aib，max , a ib，max represents the maximum deceleration of the ego vehicle i in the braking process, x si represents the longitudinal distance between the ego vehicle i and the traffic signal light, k sx represents the longitudinal gradient adjustment coefficient, and the negative sign represents that the direction of F si is opposite to the speed direction of the other vehicle j.

[0024] Further, step five specifically comprises:

[0025] First, according to the low-risk value F ji , the coordinates (x, y) of a series of corresponding points in the geodetic coordinate system XOY are obtained, a polynomial model described by formula (6) is used for fitting, the powers p and q of x and y are repeatedly adjusted, and then a set of smooth fitting curves are obtained, and one of the fitting curves is selected as the lowest risk boundary:

[0026]

[0027] In the formula, F(x, y) is the corresponding F ji at (x, y); n and m are the orders of the polynomial, a pq are polynomial coefficients;

[0028] Subsequently, the data points (x, y) and the corresponding low-risk values F ji are substituted into the polynomial model to obtain an equation group AX=F ji , and the coefficient matrix A is solved; wherein A is a coefficient matrix with a size of mn*mn, and the elements thereof are composed of polynomial term coefficients a pq ; X is an unknown polynomial coefficient vector with a size of mn*1; and F ji is a risk value vector with a size of mn*1, which is composed of risk contour data.

[0029] The application also provides a method and system for identifying and avoiding a downlink driving risk area under multiple constraints, which comprises:

[0030] An environmental information acquisition module is configured to acquire current road traffic environment state information of a self-vehicle in real time.

[0031] A traffic rule and barrier multi-constraint judgment module is configured to acquire traffic rules and all obstacles involved in the current traffic environment according to the current road traffic environment state information, and judge whether the acquired traffic rules and obstacles in the current environment are effective constraints.

[0032] The driving risk identification module is used to obtain soft constraints related to traffic rules and hard constraints related to road obstacles based on the current road traffic environment status information obtained by the environmental information acquisition module and the judgment results of the traffic rules and obstacle effectiveness constraints obtained by the traffic rules and obstacle multi-constraint judgment module. Using the autonomous driving multi-constraint risk field model that considers the risk source and risk gradient change, it identifies the malicious driving risk of other vehicles to its own vehicle under hard constraints. The autonomous driving multi-constraint risk field model is described by equation (1) or (2):

[0033]

[0034]

[0035]

[0036] In the formula, v i and v j F represents the velocity vectors of vehicle i and vehicle j, respectively. ji For the risk field force of vehicle j on vehicle i, by translating the geodetic coordinate system XOY to x ji jy ji The coordinate system is transformed to convert F ji The direction is set as the vector relative velocity v ji The direction, x ji jy ji The coordinate system is based on car j as the origin, with the x-axis as the horizontal axis. ji The positive direction of j is the vector relative velocity v. ji The direction, the vertical axis jy ji The positive direction is the velocity vector v of the other vehicle j. j Forming an acute angle and perpendicular to the horizontal axis x ji The direction of j, m i and m j These are the masses of vehicle i and other vehicle j, respectively, r e r0 is the following distance per unit length, and k is the following distance per vehicle. x and k y These are the longitudinal and transverse gradient coefficients, x′, respectively. ji and y′ ji Represent the coordinate system x of vehicle i and vehicle j respectively. ji jy ji The relative distance between the longitudinal and transverse directions in the middle, r ji r is the longitudinal distance between vehicle i and other vehicle j. max and r min These represent the maximum and minimum distances between vehicles in free flow;

[0037] The driving hazard assessment module is used to assess the F risk obtained from the driving risk identification module. jievaluate the driving risk of the ego vehicle;

[0038] The path planning generation module is configured to, when the evaluation result of the driving risk evaluation module reaches or exceeds the preset value, generate a path planning according to the F ji obtained by the driving risk identification module, and obtain the risk region, the low-risk region and the safe feasible region of the ego vehicle through the driving safety field, and plan a path in the safe feasible region and select an optimal path.

[0039] Further, the driving risk evaluation module specifically includes:

[0040] a virtual acceleration calculation unit configured to calculate the virtual acceleration a ji caused by the F i , a i = F ji / m i ;

[0041] a judgment unit configured to compare the size of a i with a preset value a i，max , and if a i ≥ a i，max , it is determined that the evaluation result reaches or exceeds the preset value a i，max .

[0042] Further, the driving risk evaluation module specifically includes:

[0043] An equivalent constraint resistance calculation unit configured to calculate the equivalent constraint resistance caused by the red light of the traffic signal to the ego vehicle as the preset value;

[0044] a judgment unit configured to compare the size of F ji with a preset value F si , and if F ji > wF si , it is determined that the evaluation result exceeds the preset value, wherein w is an adjustment coefficient obtained by calibrating real road traffic accident data in the same scene.

[0045] Further, F si is calculated by formula (5);

[0046]

[0047] In the formula, R si represents the constraint force of the traffic signal on vehicle i, which is described as R si = m i a ib，max , a ib，max represents the maximum deceleration of the ego vehicle i during braking, and x si represents the longitudinal distance between the ego vehicle i and the traffic signal, ksx denotes the longitudinal gradient adjustment coefficient, and the negative sign represents F si The direction of F is opposite to the speed direction of the jth vehicle.

[0048] Further, the path planning generation module specifically comprises:

[0049] a lowest risk boundary acquisition unit configured to acquire a low risk value F ji obtain the coordinates (x, y) of a series of corresponding points in the geodetic coordinate system XOY, fit the polynomial model described by formula (6), and obtain a smooth fitting curve set by repeatedly adjusting the powers p and q of x and y, and select one of the fitting curves as the lowest risk boundary:

[0050]

[0051] In the formula, F(x, y) is the corresponding F ji at (x, y), and n and m are the orders of the polynomial, a pq is the polynomial coefficient;

[0052] a coefficient matrix construction unit configured to substitute the data points (x, y) and the corresponding low risk values F ji into the polynomial model to obtain an equation group AX=F ji , and solve the coefficient matrix A; wherein A is a coefficient matrix of mn x mn, the elements of which are composed of the polynomial term coefficients a pq ; X is an unknown polynomial coefficient vector of mn x 1; F ji is a risk value vector of mn x 1, which is composed of the risk contour data.

[0053] Since the present application considers the relative risk vector expression of the driving interaction, the traffic rule related soft constraint and the roadblock related hard constraint, and sets the direction of the risk as the vector relative speed direction through the coordinate conversion from the geodetic coordinate system to the Cartesian coordinate system, the kinematics relationship between the other vehicle and the ego vehicle is more consistent, and the risk value obtained by the risk modeling is also more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 is a flowchart of the malignant driving risk identification and avoidance area generation method under the traffic rule-roadblock multi-constraint of the present application;

[0055] Figure 2 is a vehicle-vehicle interaction relationship schematic diagram in a crossroad scene;

[0056] Figure 3 and Figure 4 are risk schematic diagrams obtained by using the automatic driving multi-constraint sharing field model of the embodiment of the present application, respectively;

[0057] Figure 5 safety feasible region schematic diagram. DETAILED DESCRIPTION

[0059] In the drawings, the same or similar notations used in the drawings represent the same or similar elements or elements having the same or similar functions. The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0060] In the description of the present application, the terms "center", "longitudinal", "transverse", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the scope of protection of the present application.

[0061] As Figure 1 shown, the multi-constraint downlink vehicle risk identification and avoidance region generation method provided by the embodiments of the present application comprises:

[0062] Step one, real-time acquisition of current road traffic environment state information of the ego vehicle.

[0063] According to the automatic driving vehicle configured with the front, surround, rear and side cameras, the current road traffic environment posture information is acquired in real time. Specifically, the front camera acquires the front traffic environment state information, such as traffic sign recognition and front vehicle running condition, such as traffic lights, pedestrian crosswalk lines, road traffic signs, etc.; the surround camera acquires panoramic traffic environment state information, which provides auxiliary function for other cameras; the rear camera acquires rear traffic environment state information, such as rear vehicle running condition, reversing image, etc.; the side camera acquires blind spot area.

[0064] Step two, according to the current road traffic environment state information, real-time acquisition of traffic rules and all obstacles involved in the current traffic environment, and judgment on whether the traffic rules and obstacles acquired in the current environment are effective constraints. Among them, the traffic rules mainly include traffic signal lights and traffic signs; the obstacles mainly include other road users, static obstacles and traffic facilities in the traffic environment.

[0065] Specifically, according to step one, the posture information of the ego vehicle is expressed as (x, y), wherein (x, y) represents the coordinates of the ego vehicle in the driving environment.

[0066] The expression of the spatial pose information of the obstacle is as follows:

[0067] (P i , Card(A))

[0068] P i =(x i ,y i )

[0069] Card(A)={D 1i ,D 2i ,D 3i ,D 4i ,D 5i ,D 6i ,D 7i ,D 8i}

[0070] wherein (x i , y i ) represents the coordinates of the ith obstacle in the driving environment; D 1i , D 8i respectively represent the ith obstacle in the east, south, west, north, southeast, northeast, southwest and northwest directions of the ego vehicle.

[0071] Based on the attitude information of the ego vehicle and the spatial pose information of the obstacles, distance judgments are made for the obstacles in the eight directions.

[0072] The definition of the distance of the obstacles to the ego vehicle is as follows:

[0073]

[0074] R i =min{,......,r i ,......},i=1,2......

[0075] wherein r i is the distance of the obstacle in this direction to the ego vehicle, and R i is the distance of the effective constraint obstacle in this direction to the ego vehicle, i.e., the effective constraint obstacle with the closest distance to the ego vehicle in this direction.

[0076] Through the judgments in the eight directions, the effective constraint obstacles in the current traffic environment can be determined.

[0077] The beneficial effects of the above embodiment are that the current road traffic environment state information is obtained in real time, the traffic rules and all obstacles involved in the current traffic environment are judged based on the current road traffic environment state information, the complexity of the environment is obtained, and preparation is made for the next step of accurately identifying the malignant driving risk under the roadblock hard constraint.

[0078] Step 3: Based on the current road traffic environment status information obtained in Step 1 and the judgment results of the effectiveness constraints of traffic rules and obstacles obtained in Step 2, soft constraint features related to traffic rules and hard constraint features related to road obstacles are extracted to obtain soft constraints related to traffic rules and hard constraints related to road obstacles. The autonomous driving multi-constraint risk field model that considers the risk source and risk gradient changes is used to identify the malicious driving risk of other vehicles to the self-vehicle under the hard constraint of road obstacles.

[0079] Specifically, to quantify the risk between any two moving objects in space, this embodiment analyzes the vehicle-to-vehicle interaction relationships in an intersection scenario, as follows: Figure 2 As shown, where: vehicle i is the vehicle itself, referred to as vehicle i below; vehicle j is the other vehicle, referred to as other vehicle j below; vehicle i and other vehicle j are respectively indicated by v. i and v j The vector velocity of the two vehicles is v as they travel towards the intersection, therefore the relative velocity between the two vehicles is v. ji Represented as: v ji =v j -v i .

[0080] If we take the coordinate of the other car (j) as the origin, along the relative velocity v of the two cars... ji Direction establishes the x-axis ji perpendicular to x ji Establish coordinate axes y in the axial direction ji As you can see, the line d connecting car i and car j... ji With x ji The axis has an included angle θ ji Obviously θ ji The smaller θ is, the greater the risk that vehicle i will be at the expense of vehicle j. ji It depends on the position and speed of the two vehicles.

[0081] In this embodiment, the autonomous driving multi-constraint sharing field model can use the risk field force F of other vehicle j on vehicle i provided by equation (1). ji Describe, F ji The direction is Direction:

[0082]

[0083] In the formula, v i and v j F represents the velocity vectors of vehicle i and vehicle j, respectively. ji For the risk field force of vehicle j on vehicle i, by transforming the geodetic coordinate system XOY to the coordinate system x ji jy ji Perform a conversion to convert F ji The direction is set as the vector relative velocity v jidirection, coordinate system x ji jy ji is the coordinate system with the origin at the ego vehicle i, the horizontal axis x ji is the positive direction of the vector relative speed v ji direction, the vertical axis jy ji is the positive direction of the vector relative speed v j of the ego vehicle j at an acute angle and perpendicular to the horizontal axis x ji direction, m i and m j are the masses of the ego vehicle i and the ego vehicle j, respectively, r e is the unit length, i.e., r e = 1 m, r0 is the following distance, k x and k y are the longitudinal and lateral gradient coefficients, respectively, the values of which can be determined according to the longitudinal and lateral constraints of the vehicle, for example, the vehicle is constrained in the lateral direction by lane lines or lane boundaries, and there is no such constraint in front of the vehicle, so k x is set to 1, and k y is a suitable number greater than 1, so that the lateral and longitudinal risks are matched, x′ ji and y′ ji are the longitudinal and lateral relative distances of the ego vehicle i and the ego vehicle j in the x ji jy ji coordinate system, r ji is the distance between the ego vehicle i and the ego vehicle j in the longitudinal direction, r max and r min are the ranges of influence of the risk in the longitudinal direction, and are the maximum and minimum distances between vehicles in free flow, respectively, the values of which can be obtained by consulting the Road Traffic Manual. Among them, the “longitudinal direction” is the direction of the vehicle, and the “lateral direction” is the horizontal plane of the vehicle body perpendicular to the longitudinal direction.

[0084] defined in formula (1) can ensure that F ji = F ij .

[0085] In another embodiment, the value of F ji is described as formula (2) below:

[0086]

[0087] In yet another embodiment, the value of F ji is described as formula (3) below:

[0088]

[0089] If the actual geodetic coordinates of the ego vehicle j and the ego vehicle i are (x j , yj ) and (x i y i If, in the coordinate system of his vehicle j, the coordinates of his vehicle j and his own vehicle i are (0, 0) and (x, y), respectively. i -x j y i -y j ), in such Figure 2 The coordinate system shown x ji jy ji Below, the coordinates of his car j are still (0, 0), while the coordinates of his car i are (x′). ji y′ ji The description is as follows: (4)

[0090]

[0091] Traditional driving safety field modeling methods typically express the driving risks of each vehicle in the environment in the form of a field, or express the driving risks of other vehicles besides the vehicle itself in the form of a field, without considering the interaction between other vehicles and the vehicle itself. Therefore, the traditional driving risk models are not accurate in modeling real driving risks.

[0092] The autonomous driving multi-constraint shared field model constructed in this embodiment of the invention takes the perspective of the self-vehicle and considers the risk impact of other vehicles on the self-vehicle. As can be seen from the risk modeling method proposed in this invention, this invention considers the mass, vector velocity, and vector relative velocity of the self-vehicle and other vehicles. At the same time, through coordinate transformation, the direction of risk is set as the direction of vector relative velocity. These are more in line with the kinematic relationship between the self-vehicle and other vehicles, so the risk modeling is more accurate.

[0093] The following is a specific example illustrating how, in constructing a multi-constraint shared field model for autonomous driving, this embodiment of the invention transforms the original geodetic coordinate system XXOY to the coordinate system x. ji jy ji The conversion process can make the modeling of the risk impact of other vehicles on one's own vehicle more accurate.

[0094] like Figure 3 and Figure 4 As shown in the figure, the risk obtained by the autonomous driving multi-constraint shared field model constructed according to an embodiment of the present invention is illustrated. In the coordinate system x... ji jy ji In the game, the coordinates of car j are (0, 0), and the coordinates of car i are... Figure 3 and Figure 4 The middle part is represented by the red dot in the figure, with coordinates (12, -12). When the speed v of the other vehicle is set... j =20m / s, vehicle speed i v i When the speed is 10 m / s, the driving risk field posed by other vehicle j to vehicle i is as follows:Figure 3 As shown. And when the speed v of the other vehicle is set... j =20m / s, vehicle speed i v i At a speed of 20 m / s, the driving risk field posed by other vehicle j to vehicle i is as follows: Figure 4 As shown. Obviously, Figure 4 In this situation, vehicle i is at greater risk from other vehicles j because at this time... ji Connecting car i to other car j via line d ji Overlapping areas have the most direct impact on risk.

[0095] On the other hand, an analysis of vehicle motion also reveals that if the vehicles in both scenarios maintain their current state, Figure 3 The car in the middle will not collide with other cars, Figure 4 In this scenario, a collision is inevitable between the vehicle i and another vehicle j. Therefore, this invention enables vehicle i to accurately identify the risk from another vehicle j in advance, making subsequent collision avoidance decisions and control more convenient.

[0096] Step four, based on F obtained in step three ji The driving hazard of the vehicle is assessed. If the assessment result reaches or exceeds the preset value, it is necessary to break through the soft constraints related to traffic rules and proceed to step five.

[0097] In the diagram, the vehicles ahead face soft constraints related to traffic rules, such as red lights and solid lines in both longitudinal and lateral directions. Even if an autonomous vehicle can detect the situation behind it through sensors, it can only comply with these soft constraints and cannot change lanes to avoid them, ultimately leading to a rear-end collision. This invention, however, decides to break through these soft constraints when the evaluation results reach or exceed preset values, thereby preventing traffic accidents.

[0098] Step 5, based on F obtained in Step 3 ji The system obtains the vehicle's risk zone, low-risk zone, and safe and feasible zone through the driving safety field. Path planning is performed in the safe and feasible zone to select the optimal path.

[0099] In one embodiment, step four, "based on F obtained in step three" ji One method for "assessing the driving hazards of a vehicle" specifically includes:

[0100] Calculate F ji The resulting virtual acceleration a i a i =F ji / m i , will a i Compared with the preset value a i，max Compare the sizes of a, if ai a i，max if the evaluation result exceeds the preset value a i，max that is, it can be determined that there is a potential danger in the traffic environment, and in this case, the soft constraint related to the traffic rules can be broken. Preferably, the preset value a i，max can represent the virtual acceleration corresponding to a serious traffic accident, and can be obtained by the virtual acceleration corresponding to a serious traffic accident. The specific value can be calibrated by real road traffic accident data in the same scene. The preset value a i，max can also be obtained by other existing methods.

[0101] In one embodiment, the "determining the driving risk of the ego vehicle according to the F ji of step three" in step four specifically includes:

[0102] calculating the equivalent constraint resistance F si of the red light of the traffic signal on the ego vehicle i si as a preset value, comparing the F ji with the preset value F si , if F ji > wF si , it is determined that the evaluation result exceeds the preset value, that is, it can be determined that there is a potential danger in the traffic environment, and in this case, the soft constraint related to the traffic rules can be broken.

[0103] The preset value F si can be realized by the following formula (5):

[0104]

[0105] In the formula, R si represents the constraint force of the traffic signal on the vehicle i, and is described as R si = m i a ib，max , a ib，max represents the maximum deceleration of the ego vehicle i in the braking process, x si represents the longitudinal distance between the ego vehicle i and the traffic signal, k sx represents a longitudinal gradient adjustment coefficient, and the negative sign represents that the direction of F si is opposite to the speed direction of the vehicle j, and w is an adjustment coefficient obtained by real road traffic accident data in the same scene.

[0106] The preset value F si can also be obtained by existing methods, such as patent documents with publication numbers CN109882587A or CN115792474.

[0107] ​​​​​​​​​​​​​​​​​​​​​​The beneficial effects of the above embodiment are: firstly, the current traffic environment state information is acquired, the obstacles and traffic rules are judged based on the current state environment information, the soft constraint features such as traffic signs and traffic signal lights and the hard constraint features such as road obstacles and traffic facilities are extracted, and the multi-constraint risk field model is established; the virtual acceleration caused by the risk field force of the ego vehicle suffered by other vehicles is calculated based on the multi-constraint risk field model, and the danger degree signal is output according to the calculation result of the virtual acceleration. The roadblock and traffic rules multi-constraint are fused by the risk field model to obtain the multi-constraint risk field model, and the size of the virtual acceleration is calculated according to the model, so that the malignant driving risk can be more comprehensively identified, and the driving safety can be improved.

[0108] In one embodiment, step five specifically comprises:

[0109] As shown in Figure 5 , the F ji obtained according to step three is obtained, the risk area (red area) and the low risk area (yellow area) of the ego vehicle are obtained through the driving safety field, then the safe feasible area (green area) of the vehicle is judged according to the vehicle dynamics constraint, and finally the path planning is carried out in the safe feasible area, and the optimal path is selected.

[0110] Firstly, the low risk value F ji is determined: according to the low risk value F ji , the coordinates (x, y) of a series of corresponding points in the ground coordinate system XOY as shown in Figure 2 are obtained, a polynomial model described by formula (6) is used for fitting, the powers p and q of x and y are repeatedly adjusted, and then a set of smooth fitting curves are obtained, and one of the fitting curves is selected as the lowest risk boundary, for example, the fitting curve corresponding to the minimum sum of square errors of the data points in the fitting curve:

[0111]

[0112] In the formula, F(x, y) is the corresponding F ji at (x, y), n and m are the orders of the polynomial, and a pq is the polynomial coefficient;

[0113] Subsequently, the coefficient matrix A is constructed: the data points (x, y) and the corresponding low risk value F ji are substituted into the polynomial model to obtain an equation group AX=F ji , and the coefficient matrix A is solved; wherein A is a coefficient matrix with a size of mn*mn, and the elements are composed of polynomial term coefficients a pq ; X is an unknown polynomial coefficient vector with a size of mn*1, and the coefficients can be obtained by least square method or other fitting methods; F jiIt is a risk value vector of size mn×1, composed of risk contour data.

[0114] In one embodiment, step five specifically includes:

[0115] Based on the vehicle dynamics constraints, a safe and feasible region is determined. Within this region, multiple polynomial fittings are performed to fit the low-risk boundary, as shown in equation (7).

[0116]

[0117] In the formula, y represents the distance of the low-risk boundary (obtained directly from the distribution of the risk field, while the safe and feasible region is obtained by combining vehicle dynamics constraints under this low-risk boundary) on the horizontal coordinate axis of the plane where the vehicle sensor is located, x represents the distance of the low-risk boundary on the vertical coordinate axis of the plane where the vehicle sensor is located, and a pq These are the coefficients of the curve expression fitted to the low-risk boundary.

[0118] In one embodiment, step five specifically includes:

[0119]

[0120] In the formula, y and x represent the distances of the low-risk boundary within the safe and feasible area on the horizontal and vertical coordinate axes of the plane where the vehicle sensors are located (vehicle coordinate system), respectively, and a pq The coefficients are the coefficients of the fitted curve expression for the low-risk boundary within the safe and feasible region.

[0121] Furthermore, such as Figure 5 As shown, within a safe and feasible area, the optimal path is selected based on the results of path planning, ultimately achieving the avoidance of severe driving risks.

[0122] The beneficial effects of using the above embodiments are: when facing situations such as Figure 3 In the extreme scenario of a rear-end collision caused by a vehicle losing control at a traffic light intersection, as shown, the system can identify and avoid severe driving risks under multiple constraints of traffic rules and road obstacles, which helps improve driving safety.

[0123] The multi-constraint driving risk identification and avoidance area generation method system provided in this embodiment of the invention includes an environmental information acquisition module, a traffic rule and road obstacle multi-constraint judgment module, a driving risk identification module, a driving hazard assessment module, and a route planning generation module, wherein:

[0124] The environmental information acquisition module is used to obtain real-time information on the current road traffic environment where the vehicle is located.

[0125] The traffic rule and barrier multi-constraint judgment module is configured to acquire traffic rules and all barriers involved in the current traffic environment according to the current road traffic environment state information, and judge whether the traffic rules and barriers acquired by the current environment are effective constraints.

[0126] The driving risk identification module is configured to obtain soft constraints related to traffic rules and hard constraints related to barriers by extracting soft constraint features and hard constraint features, the soft constraint features including traffic signs and traffic signal lights, and the hard constraint features including road barriers and traffic facilities, identify malignant driving risk of other vehicles to the ego vehicle under the barrier hard constraints by using an automatic driving multi-constraint risk field model considering risk sources and risk gradient changes, and the automatic driving multi-constraint risk field model is described by formula (1) or (2) or formula (3).

[0127] The driving risk assessment module is configured to assess the driving risk of the ego vehicle according to the F ji obtained by the driving risk identification module.

[0128] The path planning generation module is configured to, in a case where the assessment result of the driving risk assessment module reaches or exceeds a preset value, obtain the F ji obtained by the driving risk identification module, obtain a risk region, a low-risk region and a safe feasible region of the ego vehicle by using a driving safety field, and plan a path in the safe feasible region and select an optimal path.

[0129] In one embodiment, the driving risk assessment module specifically includes a virtual acceleration calculation unit and a first judgment unit, wherein:

[0130] The virtual acceleration calculation unit is configured to calculate a virtual acceleration a ji of the ego vehicle i caused by the risk field force F i of the other vehicle j. a i =F ji / m i .

[0131] The first judgment unit is configured to compare the virtual acceleration a i，max corresponding to a serious traffic accident with a i , and if a i ≥ a i，max , it is determined that the assessment result reaches or exceeds the preset value.

[0132] In one embodiment, the driving risk assessment module specifically includes an equivalent constraint resistance calculation unit and a second judgment unit, wherein:

[0133] The equivalent constraint resistance calculation unit is configured to calculate the equivalent constraint resistance F caused by the red light of the traffic signal to the ego vehicle i by formula (5) si .

[0134] The second judging unit is configured to compare F si and F ji , and determine that the evaluation result reaches or exceeds the preset value if F ji > wF si .

[0135] In one embodiment, the path planning generation module specifically comprises a minimum risk boundary acquisition unit and a coefficient matrix construction unit, wherein:

[0136] The minimum risk boundary acquisition unit is configured to obtain a series of corresponding point coordinates (x, y) in the terrestrial coordinate system XOY according to the low risk value F ji , fit a polynomial model described by formula (6), and obtain a smooth fitting curve set by repeatedly adjusting the powers p and q of x and y, and select one of the fitting curves as the minimum risk boundary.

[0137] The coefficient matrix construction unit is configured to substitute the data points (x, y) and the corresponding low risk value F ji into the polynomial model to obtain an equation group AX=F ji , and solve the coefficient matrix A; wherein A is a coefficient matrix with a size of mn x mn, and the elements of A are composed of polynomial term coefficients a pq ; X is an unknown polynomial coefficient vector with a size of mn x 1; and F ji is a risk value vector with a size of mn x 1, which is composed of the risk contour data.

[0138] Finally, it should be pointed out that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit it. Those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for multi-constraint downlink vehicle risk identification and avoidance zone generation, characterized in that, Comprise: Step one, real-time acquisition of the current road traffic environment state information of the ego vehicle; Step two, according to the current road traffic environment state information, real-time acquisition of the traffic rules and all obstacles involved in the current traffic environment, and judgment of whether the traffic rules and obstacles obtained in the current environment are effective constraints; Step three, according to the current road traffic environment state information obtained in step one and the judgment result of the traffic rules and obstacle effectiveness constraints obtained in the current environment in step two, obtain the soft constraint related to the traffic rules and the hard constraint related to the roadblock, and use the automatic driving multi-constraint risk field model considering the risk source and the risk gradient change to identify the malignant driving risk of the other vehicle to the ego vehicle under the hard constraint, wherein the automatic driving multi-constraint risk field model is described by formula (1) or (2) or (3): where v i and v j are the velocity vectors of ego vehicle i and other vehicle j, respectively, F ji is the risk field force of other vehicle j on ego vehicle i, which is transformed from the earth coordinate system XOY to the coordinate system x ji jy ji to set the direction of F ji to the direction of the relative velocity vector v ji , the coordinate system x ji jy ji is a coordinate system with other vehicle j as the origin, the positive direction of the horizontal axis x ji j is the direction of the relative velocity vector v ji , and the positive direction of the vertical axis jy ji is the direction of the acute angle with the velocity vector v j of other vehicle j and perpendicular to the horizontal axis x ji j, m i and m j are the masses of ego vehicle i and other vehicle j, respectively, r e is the unit length, r0 is the following distance, k x and k y are the longitudinal and lateral gradient coefficients, respectively, x' ji and y' ji are the longitudinal and lateral relative distances of ego vehicle i and other vehicle j in the coordinate system x ji jy ji , r ji is the distance between ego vehicle i and other vehicle j in the longitudinal direction, r max and r min are the maximum and minimum distances between vehicles in free flow, respectively. Step four, F is obtained according to the F obtained in step three ji If the evaluation result reaches or exceeds the preset value, the soft constraint needs to be broken through, and step five is entered. Step five, F is obtained according to the F obtained in step three ji The risk area, the low-risk area and the safe feasible area of the ego vehicle are obtained through the driving safety field, path planning is performed in the safe feasible area, and the optimal path is selected.

2. The method of claim 1, wherein the method further comprises: "the F obtained in step three" in step four ji The method of "evaluating the driving risk of the ego vehicle" specifically includes: F ji virtual acceleration a i , a i = F ji / m i , and a i is compared with a preset value a i,max , if a i ≥ a i,max , it is determined that the evaluation result reaches or exceeds the preset value; Another method specifically comprises: The equivalent constraint resistance F caused by the red light of the traffic signal to the ego vehicle i is calculated si si As a preset value, F ji is compared with the size of the preset value F si , if F ji > wF si , it is determined that the evaluation result exceeds the preset value, wherein w is an adjustment coefficient obtained by calibrating the real road traffic accident data in the same scene.​ 3. The method of claim 2, wherein the method further comprises: a i,max represents a virtual acceleration corresponding to a serious traffic accident.

4. The method of claim 2, wherein the method further comprises: determining a plurality of risk avoidance zones for the vehicle based on the plurality of constraints and the plurality of risk zones. F si By calculation from equation (5): where R si represents the constraint force of the red light of the traffic signal on vehicle i, described as R si = m i a ib,max , a ib,max represents the maximum deceleration of vehicle i during braking, x si represents the longitudinal distance between vehicle i and the traffic signal, k sx represents the longitudinal gradient adjustment coefficient, and the negative sign represents that the direction of F si is opposite to the direction of the speed of vehicle j.

5. The method of claim 1 or 2 or 3 or 4, wherein, Step five specifically comprises: First, according to the low risk value F ji First, the coordinates (x, y) of a series of corresponding points in the geodetic coordinate system XOY are obtained, a polynomial model described by equation (6) is used for fitting, the powers p and q of x and y are repeatedly adjusted, and then a set of smooth fitting curves are obtained, and one of the fitting curves is selected as the lowest risk boundary: where F(x, y) is the corresponding F at (x, y) ji , n and m are the orders of the polynomials, and a pq are the polynomial coefficients; Subsequently, the data points (x, y) and the corresponding low-risk values F ji are substituted into the polynomial model to obtain an equation set Ax = F ji , and the coefficient matrix A is solved; wherein A is a coefficient matrix of mn x mn, and the elements thereof are composed of polynomial term coefficients a pq ; X is an unknown polynomial coefficient vector of mn x 1; and F ji is a risk value vector of mn x 1, which is composed of the risk contour data.

6. A method system for multi-constraint downlink vehicle risk identification and evasion zone generation, characterized in that, Comprise: The environment information acquisition module is used for real-time acquisition of the current road traffic environment state information of the ego vehicle; The traffic rule and roadblock multi-constraint judgment module is used for real-time acquisition of the traffic rules and all obstacles involved in the current traffic environment according to the current road traffic environment state information, and judgment of whether the traffic rules and obstacles obtained in the current environment are effective constraints; The driving risk identification module is used for obtaining the soft constraint related to the traffic rules and the hard constraint related to the roadblock according to the current road traffic environment state information obtained by the environment information acquisition module and the judgment result of the traffic rules and obstacle effectiveness constraints obtained in the current environment by the traffic rule and roadblock multi-constraint judgment module, and identifying the malignant driving risk of the other vehicle to the ego vehicle under the hard constraint by using the automatic driving multi-constraint risk field model considering the risk source and the risk gradient change, wherein the automatic driving multi-constraint risk field model is described by formula (1) or (2): where v i and v j are the velocity vectors of ego vehicle i and other vehicle j, respectively, F ji is the risk field force of other vehicle j on ego vehicle i, which is transformed from the earth coordinate system XOY to the x ji jy ji coordinate system to set the direction of F ji to the direction of the relative velocity v ji , the x ji jy ji coordinate system is with other vehicle j as the origin, the positive direction of the horizontal axis x ji j is the direction of the relative velocity v ji , the positive direction of the vertical axis jy ji is the direction of the acute angle with the velocity vector v j of other vehicle j and perpendicular to the horizontal axis x ji j, m i and m j are the masses of ego vehicle i and other vehicle j, respectively, r e is the unit length, r0 is the following distance, k x and k y are the longitudinal and lateral gradient coefficients, respectively, x' ji and y' ji are the longitudinal and lateral relative distances of ego vehicle i and other vehicle j in the x ji jy ji coordinate system, r ji is the distance between ego vehicle i and other vehicle j in the longitudinal direction, r max and r min are the maximum and minimum distances between vehicles in free flow, respectively. a driving risk assessment module for assessing driving risk of the ego vehicle according to the F ji , the driving risk of the ego vehicle The path planning generation module is configured to, in the case that the evaluation result of the driving risk assessment module reaches or exceeds a preset value, generate a path according to the F ji , the risk area, the low-risk area and the safe feasible area of the ego vehicle are obtained from the driving safety field, and the path planning is performed in the safe feasible area, and the optimal path is selected.

7. The method system of claim 6, wherein, The driving risk assessment module specifically comprises: a virtual acceleration calculation unit for calculating F ji a virtual acceleration a i , a i = F ji / m i ; The judgment unit is used to determine a i Compared with the preset value a i,max Compare the sizes of a, if a i ≥a i,max If the evaluation result reaches or exceeds the preset value a, then it is determined that the evaluation result has reached or exceeded the preset value a. i,max .

8. The method system of claim 6, wherein, The driving risk assessment module specifically comprises: The equivalent constraint resistance calculation unit is used for calculating the equivalent constraint resistance caused by the red light of the traffic signal lamp to the ego vehicle as a preset value; a judging unit for comparing F ji with a preset value F si , and determining that the evaluation result exceeds the preset value if F ji > wF si , where w is an adjustment coefficient obtained by using real road traffic accident data in the same scene for calibration.

9. The method system of claim 8, wherein, F si calculated by equation (5); where R si represents the constraint force of traffic signal light on vehicle i, described as R si = m i a ib,max , a ib,max represents the maximum deceleration of ego vehicle i during braking, x si represents the longitudinal distance between ego vehicle i and traffic signal light, k sx represents the longitudinal gradient adjustment coefficient, and the negative sign represents that the direction of F si is opposite to the direction of the speed of other vehicle j.

10. The multi-constrained downhill vehicle risk identification and evasion zone generation system of claim 6 or 7 or 8 or 9, wherein, The path planning generation module specifically comprises: a lowest risk boundary acquisition unit for acquiring the lowest risk boundary according to the low risk value F ji The coordinates (x, y) of a series of corresponding points in the geodetic coordinate system XOY are fitted by using the polynomial model described by formula (6), the powers p and q of x and y are repeatedly adjusted, and then a set of smooth fitting curves are obtained, and one of the fitting curves is selected as the lowest risk boundary. where F(x, y) is the corresponding F at (x, y) ji , n and m are the orders of the polynomials, and a pq are the polynomial coefficients; a coefficient matrix constructing unit, configured to construct a coefficient matrix A from the data points (x, y) and the corresponding low-risk values F ji , and obtain an equation group AX=F ji , solve the coefficient matrix A; wherein, A is a coefficient matrix with a size of mn×mn, and elements of A are composed of polynomial term coefficients a pq ; X is an unknown polynomial coefficient vector with a size of mn×1; F ji is a risk value vector with a size of mn×1, and F is composed of the risk contour data.

Images