A safety assessment method based on a dangerous scene

Abstract

The application discloses a kind of safety evaluation methods based on dangerous scene, belong to the technical field of collision avoidance algorithm, comprising: S1 extracts use case from database, adds scene label;S2 constructs emergency braking model and emergency steering model according to use case state parameter;S3 by substituting multiple scenes into dangerous degree index formula, obtain critical condition threshold;S4 according to critical condition threshold, the safety of scene is evaluated, select appropriate collision avoidance method.The present application relies on CIDAS accident library, contains various road types, collision form, the scene working condition of collision speed, extracts key function scene and key logic parameter from it, re-labels, forms multi-classification accident library for dangerous evaluation.Establish driver emergency braking model and emergency steering model, convert into optimization problem, solve the parameters of motion trajectory.Through whether it can be solved, judge what kind of strategy should be taken at the current time.When real-time system parameters are transmitted, the safety of current vehicle can be quickly evaluated.

Description

Technical Field

[0001] This invention belongs to the field of collision avoidance algorithm technology, specifically a safety assessment method based on dangerous scenarios. Background Technology

[0002] Safety assessments determine the level of danger in a given scenario and provide corresponding collision avoidance strategies based on the severity level to prevent accidents. Existing algorithms mainly fall into two categories: ① fitting a collision probability distribution to the entire dataset; ② directly calculating the characteristic parameters. Deterministic methods are computationally efficient and easy to develop. However, existing deterministic methods fail to fully consider the behavioral characteristics of Chinese drivers, rarely consider the function of braking combined with steering, and lack sufficient support from a large amount of natural driving data and accident scenario data. Furthermore, the design process does not fully reflect the current accident distribution, resulting in results that lack realism and persuasiveness. Summary of the Invention

[0003] To address the above problems, this invention provides a safety assessment method based on hazardous scenarios. Utilizing the CIDAS accident database, which contains scenario conditions for various road types, collision types, and collision speeds, key functional scenarios and key logical parameters are extracted and re-labeled to form a multi-classified accident database for hazard assessment. Emergency braking and emergency steering models for the driver are established separately, transforming them into optimization problems to solve for the parameters of the trajectory. The solvability of the problem determines the appropriate strategy for the current moment. Furthermore, when real-time system parameters are input, this strategy can quickly assess the safety of the vehicle at that moment.

[0004] The technical solution of this invention is as follows: a safety assessment method based on hazardous scenarios, comprising the following steps:

[0005] S1 extracts use cases from the database and adds scenario tags;

[0006] S2 constructs an emergency braking model and an emergency steering model based on the use case state parameters;

[0007] S3 derives the critical condition threshold by substituting multiple scenarios into the danger level index formula.

[0008] S4 performs a safety assessment of the scene based on critical condition thresholds and selects an appropriate collision avoidance method.

[0009] Furthermore, in step S1, the scenario is a rear-end collision scenario, and the extraction conditions are: the following vehicle is either following or rear-ending the vehicle in front, and the initial speed of the following vehicle is greater than or equal to the speed of the vehicle in front. The initial speed, initial acceleration, relative position, and speed at the moment of collision of the following and the vehicle in front are extracted.

[0010] Furthermore, in step S1, the scenario is a steering collision avoidance scenario, where the absolute value of the steering wheel angular velocity δ is greater than the set value δ.min or the lateral acceleration a y the absolute value of which is greater than a set value a ymin is taken as a screening condition.

[0011] Further, when the vehicle speed satisfies 0<V≤10km / h, δ min is 35° / s; when the vehicle speed satisfies 10<V≤35km / h, δ min is (46.39-1.1V)° / s; when the vehicle speed satisfies V>35km / h, δ min is (8.82-0.044V)° / s; when the vehicle speed satisfies 0<V≤10km / h, a ymin is 4m / s 2 ; when the vehicle speed satisfies 10<V≤40km / h, a ymin is (3.17-0.034V)m / s 2 ; when the vehicle speed satisfies 40<V≤90km / h, a ymin is (1.99-0.005V)m / s 2 ; when the vehicle speed satisfies V≥90km / h, a ymin is (0.03V-1.17)m / s 2 .

[0012] Further, the driving time is equally divided into ten parts, and the initial speed, initial acceleration, relative position and collision moment speed of the vehicle at the ten marked sampling points are respectively recorded.

[0013] Further, in step S2, the emergency braking model is established in the following manner:

[0014] A longitudinal displacement equation based on the quintic polynomial rule is established:

[0015] s(t)=a5t 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0

[0016] wherein a0 to a5 are polynomial coefficients, t is time, and s(t) is the longitudinal relative displacement from the initial position; set the initial state of the vehicle and the end state as follows,

[0017]

[0018]

[0019] wherein, are respectively the position / speed / acceleration of the vehicle;

[0020] An optimization problem is solved iteratively, with the acceleration gradient as the optimization objective. The objective function is set as follows:

[0021]

[0022] The constraints are:

[0023]

[0024] Where J is the integral of the square of the derivative of acceleration, j is the derivative of acceleration, D is the braking distance, and υ max For the maximum speed, a max To determine the maximum acceleration, data from braking collision scenarios in the CIDAS accident database was analyzed. max and a max The speeds are 120 km / h and 9.8 m / s respectively. 2 By obtaining all parameters of the equation of motion, the driver's emergency braking model can be obtained.

[0025] Furthermore, the emergency steering model satisfies the following conditions to avoid collisions and minimize passenger impact:

[0026] ω max＝ k * δ max

[0027] Where, ω max δ represents the maximum steering wheel angular velocity, k is the steering coefficient, and based on statistical data of steering collision scenarios in the CIDAS accident database, the value of k ranges from [1-3], typically 1.5. max This is the maximum steering wheel angle.

[0028] Furthermore, the position of the rear vehicle on its right rear is (x0 y0), the position of the front vehicle on its left rear is (x1 y1), and the turning angle of the vehicles is... The longitudinal direction of the car is the x-axis, and the lateral direction is the y-axis, where y1 > y0. Draw a perpendicular line from the left rear point of the front car to the rear car, and let the distance be D. Assume the safe distance is D. safe D was identified safe =0.5m, D must satisfy D>D safe Right now:

[0029]

[0030] Set the initial longitudinal position of the rear axle of the rear vehicle to x. i If the distance between the following car and the preceding car is Δx, then:

[0031]

[0032]

[0033] Among them, υ M υ A These are the real-time speeds of the following vehicle and the vehicle in front, respectively.

[0034] Furthermore, the lateral acceleration a of the vehicle throughout the collision avoidance process y It should be less than a specific threshold a ymax ; Vehicle lateral acceleration gradient j1 Less than a specific threshold j max ,Right now:

[0035] a y ≤a ymax

[0036] j1≤j max

[0037] In summary, this optimization problem can be denoted as:

[0038] Min[ω max ] = min[k * δ max ]

[0039] The constraints are:

[0040]

[0041] Furthermore, by analyzing steering collision scenario data from the CIDAS accident database, the vehicle's lateral acceleration a y Specific threshold a ymax The value is 9.8 m / s 2 A specific threshold j of the vehicle's lateral acceleration gradient j1 max The value is 0.01 m / s 3 .

[0042] Furthermore, in step S3, the formula for the risk level index is:

[0043]

[0044] Among them, TTC -1 As an indicator of risk level, υ M It is the speed of the following vehicle, υ A It is the speed of the car in front, x M It's the rear vehicle position, x A That's the position of the car in front.

[0045] Furthermore, the speed of the following vehicle υ M The horizontal axis represents the degree of danger (TTC). -1 Using the vertical axis as the ordinate, multiple scenes extracted from the database are used to draw a collision avoidance boundary map through the model in step S2.

[0046] The beneficial effects of this invention are as follows:

[0047] This invention relies on the CIDAS accident database, which contains scenario conditions for various road types, collision types, and collision speeds. Key functional scenarios and key logical parameters are extracted and re-labeled to form a multi-classified accident database for hazard assessment. Emergency braking and emergency steering models for the driver are established separately, transforming them into optimization problems to solve for the parameters of the trajectory. The solveability of these problems determines the appropriate strategy for the current moment. Furthermore, when real-time system parameters are input, this strategy can quickly assess the safety of the vehicle. Attached Figure Description

[0048] Figure 1 This is a flowchart of the present invention.

[0049] Figure 2 This is a diagram illustrating an emergency collision avoidance process.

[0050] Figure 3 This is a schematic diagram of the collision avoidance boundary of the present invention.

[0051] In the picture:

[0052] M - Rear vehicle; A - Front vehicle; A1 - Braking to avoid collision; B - Steering to avoid collision; C - Both braking and steering can avoid collision; D - Neither braking nor steering can avoid collision. Detailed Implementation

[0053] It should be noted that in the description of this invention, the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", "clockwise", "counterclockwise", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation.

[0054] In this invention, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; a connection can be a mechanical connection or an electrical connection; a link can be a direct connection or an indirect connection through an intermediate medium, and can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0055] The method of this invention is as follows:

[0056] First, specify the scenario extraction rules, extract use cases applicable to this invention patent from the CIDAS incident database, and add scenario tags;

[0057] Then, the construction methods of the emergency braking model and the emergency steering model are respectively described according to the state parameters;

[0058] Finally, a large number of models are trained offline to obtain the critical condition threshold. Based on this critical condition, the safety assessment of the scene can be quickly realized.

[0059] The present invention mainly analyzes the driving condition where the host vehicle follows and rear-ends the preceding vehicle, and judges whether an automatic driving strategy can successfully avoid collision starting from the same state to prevent traffic accidents. Therefore, in the CIDAS database, scenes where the target vehicle is located ahead and the speed of the host vehicle at the initial moment is greater than or equal to the speed of the target vehicle are extracted. During the process, the target vehicle may decelerate or maintain a constant speed.

[0060] The initial speed, initial acceleration, relative position, and speed at collision moment of the host vehicle and the target vehicle are extracted. If the number of samples is insufficient, the whole process can be divided into 10 parts according to the driving time, and the speed, acceleration and relative position of the vehicle at 10 sampling points are respectively marked. In this way, the number of use cases will be expanded by 10 times.

[0061] Extract collision avoidance by steering scenes: according to the steering wheel angular velocity δ and the lateral acceleration a y as the screening conditions. Both of them shall be greater than the given threshold δ min and a ymin , which is recorded as:

[0062] δ≥δ min

[0063] a y ≥a ymin

[0064] Thus, a set of scene characteristic parameters is formed.

[0065] The screening condition is that when the vehicle speed 0<V≤10km / h, δ min is 35° / s; when the vehicle speed 10<V≤35km / h, δ min is (46.39-1.1V)° / s; when the vehicle speed V>35km / h, δ min is (8.82-0.044V)° / s; when the vehicle speed 0<V≤10km / h, a ymin is 4m / s 2 ; when the vehicle speed 10<V≤40km / h, a ymin is (3.17-0.034V)m / s 2 ; when the vehicle speed 40<V≤90km / h, a ymin is (1.99-0.005V)m / s 2 ; when the vehicle speed V≥90km / h, aymin =(0.03V-1.17)m / s 2 .

[0066] Two collision avoidance strategies are proposed: braking collision avoidance and steering collision avoidance. Emergency braking and emergency steering models are established for each strategy. Vehicle kinematic expressions are constructed, and the ability to solve for the parameters determines whether collision avoidance can be successfully achieved theoretically.

[0067] The emergency braking model is established based on the longitudinal displacement equation according to the law of fifth-order polynomials:

[0068] s(t)=a5t 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0

[0069] Where a0 to a5 are polynomial coefficients, t is time (in seconds, s), and s(t) is the longitudinal relative displacement from the initial position (in meters, m).

[0070] Assume the initial state of the vehicle and the end state As shown in the following formula. Where, These represent the vehicle's position, speed, and acceleration, respectively.

[0071]

[0072]

[0073] The problem should be transformed into an optimization problem and solved iteratively.

[0074] The objective function is set with acceleration gradient as the optimization objective:

[0075]

[0076] The constraints are:

[0077]

[0078] Where J is the integral of the square of the derivative of acceleration, j is the derivative of acceleration, D is the braking distance, and υ max For the maximum speed, a max To determine the maximum acceleration, data from braking collision scenarios in the CIDAS accident database was analyzed. max and a max The speeds are 120 km / h and 9.8 m / s respectively. 2 Based on this, all parameters of the motion equation can be obtained, thus obtaining the driver's emergency braking model.

[0079] Another strategy is for the vehicle to make an emergency turn, i.e., to veer into another lane and avoid a collision by moving laterally. Figure 2 This is a diagram illustrating an emergency maneuver. M represents the following vehicle, and A represents the vehicle directly in front that needs to be avoided.

[0080] In this operating condition, three conditions must be met: "collision avoidance," "minimizing passenger impact," and "vehicle performance meeting dynamic constraints." Among these, dynamic constraints are objective conditions, collision avoidance is the most basic requirement, and minimizing passenger impact is an additional requirement. Therefore, using a two-degree-of-freedom vehicle model, the relationship between the maximum angular velocity of the steering wheel and motion parameters is established. With the relative minimum safe distance and its own limits as constraints, an optimization problem is solved. Based on this, a driver emergency steering model is established.

[0081] Steering angular velocity cannot be directly obtained in real time, so it needs to be calculated from other parameters. Based on accident scene slice data, the correlation coefficients between the maximum steering wheel angular velocity and other motion parameters (such as vehicle longitudinal velocity, vehicle lateral velocity, vehicle mass, etc.) were analyzed in each slice. It was ultimately found that it has a linear relationship with the maximum steering wheel angle during collision avoidance, and this ratio has a steady-state value at a specific level. This is denoted as:

[0082] ω max =k*δ max

[0083] Where, ω max δ represents the maximum steering wheel angular velocity, k is the steering coefficient, and based on statistical data of steering collision scenarios in the CIDAS accident database, the value of k ranges from [1-3], typically 1.5. max This represents the maximum steering wheel angle. Constraint 1:

[0084] Throughout the lateral collision avoidance maneuver, the steering angle gradually increases. For example... Figure 2 The area within the middle circle represents the moment when the following vehicle's turning angle is at its maximum, which is also the most dangerous moment. Simply ensure that the minimum distance between the two vehicles meets the safe distance requirement.

[0085] Let the position of car M to its right rear be (x0, y0), and the position of car A to its left rear be (x1, y1). Let the turning angle of the vehicles be... The longitudinal direction of the cars is the x-axis, and the lateral direction is the y-axis. At this point, y1 > y0 must be satisfied; otherwise, a collision will not occur. Draw the perpendicular line from the left rear point of car A to car M, and let the distance be D. Let the safe distance be D. safe Preliminary identification of D safe =0.5m. D > D must be satisfied. safe . Recorded as:

[0086]

[0087] Set the initial longitudinal position of the rear axle of car M to x. iLet the distance between car M and car A be Δx. Then we have:

[0088]

[0089]

[0090] Among them, υ M υ A These are the real-time speeds of car M and car A, respectively.

[0091] Constraint 2:

[0092] To ensure passenger comfort, the vehicle's lateral acceleration 'a' throughout the collision avoidance process is... y It should be less than a specific threshold a ymax ; Vehicle lateral acceleration gradient j1 Less than a specific threshold j max . Recorded as:

[0093] a y ≤a ymax

[0094] j1≤j max

[0095] In summary, this optimization problem can be denoted as:

[0096] Min[ω max ]=min[k*δ max ]

[0097] The constraints are:

[0098]

[0099] The above strategies can be used to calculate vehicle motion parameters.

[0100] By analyzing steering collision scenario data from the CIDAS accident database, the vehicle's lateral acceleration a y Specific threshold a ymax The value is 9.8 m / s 2 A specific threshold j of the vehicle's lateral acceleration gradient j1 max The value is 0.01 m / s 3 .

[0101] At the initial moment of a dangerous scenario, based on the vehicle's initial states, both an emergency braking model and an emergency steering model are constructed simultaneously. The feasibility of a collision avoidance strategy is determined by whether a solution can be found. However, the model construction speed does not support real-time online processing and can only be performed offline. Therefore, we propose to construct vehicle emergency models offline based on numerous dangerous scenarios; label the initial level of danger to form a two-dimensional "danger level - collision avoidance strategy" graph. Then, as the vehicle moves in real time, the danger level is calculated in real time, and the optimal collision avoidance strategy for the current moment is selected by looking up a table.

[0102] The parameters required for hazard level indicators must be readily available and easily calculated (i.e., the calculation process must be indirect); otherwise, real-time requirements will not be met. TTC is proposed. -1 And the speed of this car υ M Both represent the degree of danger. TTC is defined as the time it will take to collide at the current speed. For ease of study, its reciprocal is used, i.e., TTC. -1 The larger the value, the higher the level of danger. This can be expressed as:

[0103]

[0104] Among them, υ M It is the speed of the following vehicle, υ A It is the speed of the vehicle in front, in m / s; x M It's the rear vehicle position, x A It indicates the position of the vehicle ahead, in meters (m). Despite the TTC... -1 The expression includes the speed υ of the rear vehicle. M , but υ M It still has its independent significance and should not be confused with it. Applying the controlled variable method, fixing υ... M (Case studies were conducted at speeds of 10 m / s, 20 m / s, and 30 m / s), with different TTC values ​​selected. -1 The results show that with TTC -1 As speed increases and vehicle spacing decreases, the required maximum steering wheel angular velocity and maximum lateral acceleration continuously increase, leading to a continuous increase in the level of danger. At low speeds, the maximum braking deceleration gradient changes significantly, requiring a smaller longitudinal distance for braking; at high speeds, the maximum braking deceleration has a greater impact, requiring a smaller longitudinal distance for steering.

[0105] The horizontal axis is to be represented by the vehicle speed υ. M The vertical axis represents the degree of danger (TTC). -1 Draw the boundary lines for the two collision avoidance strategies on the diagram. Based on numerous use cases from the CIDAS incident database, draw the collision avoidance boundary diagram as follows: Figure 3 .

[0106] As the values ​​on the x and y axes increase, the risk level rises sharply. Two boundary lines divide the coordinate graph into four regions:

[0107] Zone A1 – Braking and Collision Avoidance Zone

[0108] Zone B – Steering and Collision Avoidance Zone

[0109] Zone C – The collision avoidance zone for both braking and steering.

[0110] Zone D – Area where neither braking nor steering can avoid a collision.

[0111] It can be seen that braking is more effective at lower speeds (A1 area), so a better braking strategy should be chosen; while steering is more effective at higher speeds (B area), so a better steering strategy should be chosen; if TTC... -1 If the value is relatively small (in region C), either strategy is acceptable.

[0112] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be included within the scope of protection of the present invention. Furthermore, all content not described in detail in this specification is prior art known to those skilled in the art.

[0113] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

Claims

1. A safety assessment method based on hazardous scenarios, characterized in that, Includes the following steps: S1 extracts use cases from the database and adds scenario tags; S2 constructs an emergency braking model and an emergency steering model based on the use case state parameters; S3 derives the critical condition threshold by substituting multiple scenarios into the danger level index formula. S4 performs a safety assessment of the scene based on critical condition thresholds and selects an appropriate collision avoidance method. In step S1, the scenario is a rear-end collision scenario. The extraction conditions are: the following vehicle is following or rear-ending the vehicle in front, the initial speed of the following vehicle is greater than or equal to the speed of the vehicle in front, and the initial speed, initial acceleration, relative position and speed at the moment of collision of the following vehicle and the vehicle in front are extracted. In step S1, the scenario is a steering collision avoidance scenario, where the absolute value of the steering wheel angular velocity δ is greater than the set value δ min or the lateral acceleration a y the absolute value is greater than the set value a ymin as the screening condition, the screening condition is that when the vehicle speed 0<V≤10km / h, δ min is 35 o / s; when the vehicle speed 10<V≤35km / h, δ min is (46.39-1.1V) o / s; when the vehicle speed V>35km / h, δ min is (8.82-0.044V) o / s; when the vehicle speed 0<V≤10km / h, a ymin is 4m / s 2 ; when the vehicle speed 10<V≤40km / h, a ymin is (3.17-0.034V)m / s 2 ; when the vehicle speed 40<V≤90km / h, a ymin is (1.99-0.005V)m / s 2 ; when the vehicle speed V≥90km / h, a ymin is (0.03V-1.17)m / s 2 ; In step S2, the emergency steering model satisfies the following condition: minimizing collision and passenger impact; therefore, the emergency steering model is: oh max= k * d max Where, ω max The maximum angular velocity of the steering wheel is given by k, where k is the steering coefficient and δ is the maximum value. max Let φ be the maximum steering wheel angle. To avoid a collision, let the position of the rear right rear point of the following vehicle be (x0, y0), the position of the front left rear point be (x1, y1), the vehicle turning angle be φ, the longitudinal direction of the car be the x-axis, the lateral direction be the y-axis, and y1 > y0. Draw a perpendicular line from the front left rear point to the rear vehicle, and let the distance be D. Let the safe distance be D. safe D was identified safe =0.5m, D must satisfy D>D safe Right now: Set the initial longitudinal position of the rear axle of the rear vehicle to x. i If the distance between the following car and the preceding car is Δx, then: Among them, υ M υ A These are the real-time speeds of the following vehicle and the vehicle in front, respectively. Based on the condition of minimizing passenger impact, the vehicle's lateral acceleration 'a' during the entire collision avoidance process... y It should be less than a specific threshold a ymax ; Vehicle lateral acceleration gradient Less than a specific threshold j max ,Right now: a y ≤a ymax j1≤j max The optimization problem is: Min[ω max ]=min[k * d max ] The constraints are: An emergency steering model was obtained.

2. The safety assessment method based on hazardous scenarios as described in claim 1, characterized in that, In step S2, the emergency braking model is: Establish the longitudinal displacement equation based on the law of fifth-order polynomials: Where a0~a5 are polynomial coefficients, t is time, and s(t) is the longitudinal relative displacement from the initial position; let the initial state of the vehicle be... and the end state as follows, in, These refer to the vehicle's position, velocity, and acceleration, respectively. An optimization problem is solved iteratively, with the acceleration gradient as the optimization objective. The objective function is set as follows: The constraints are: Where J is the integral of the square of the derivative of acceleration, j is the derivative of acceleration, D is the braking distance, and υ max For the maximum speed, a max To obtain the maximum acceleration, all parameters of the motion equation are obtained, thus yielding the driver's emergency braking model.

3. The safety assessment method based on hazardous scenarios as described in claim 1, characterized in that, In step S3, the formula for the risk level index is: Among them, TTC -1 As an indicator of risk level, υ M It is the speed of the following vehicle, υ A It is the speed of the car in front, x M It's the rear vehicle position, x A That's the position of the car in front.

4. The safety assessment method based on hazardous scenarios as described in claim 1, characterized in that, The critical condition threshold is obtained through a collision avoidance boundary map, and the following vehicle speed υ M The horizontal axis represents the degree of danger (TTC). -1 Using the vertical axis as the ordinate, multiple scenes are extracted from the database, and a collision avoidance boundary map is drawn using the model from step S2.

Images