A risk evaluation method for cross-sight multi-targets

Abstract

The application discloses a kind of risk evaluation methods of cross sight distance multi-target, comprising: risk perception prediction module and global risk assessment module;And according to current perception information, blind area is calculated, and according to blind area space, blind area hypothetical target obstacle is set, and the motion range of hypothetical target and visible target is predicted, the intersection of the motion range of self-vehicle and the motion range of all target objects is calculated according to the motion range, finally, the comprehensive risk evaluation is carried out in combination with the normalized evaluation index.According to the application, various risks can be effectively evaluated, the correct perception of surrounding environment and risk of automatic driving function is realized, and the safety capability of automatic driving system is improved.

Description

Technical Field

[0001] This invention relates to the technical field of automotive safety and testing verification, and in particular to a risk assessment method for multiple targets across line of sight. Background Technology

[0002] The application of autonomous driving is becoming increasingly widespread; however, a long-standing challenge is the current lack of safety, with frequent accidents caused by autonomous driving. In this context, traditional risk assessment methods have proven inadequate in addressing the numerous challenges of current autonomous driving operations. Therefore, a forward-looking risk assessment method specifically designed for autonomous driving is urgently needed.

[0003] Traditional risk assessment methods suffer from two major drawbacks. First, traditional indicators assess risk based on objects within the sensor's visible range, while many accidents are caused by vehicles or pedestrians suddenly appearing in blind spots. Therefore, there is a pressing need for an assessment method capable of predicting and evaluating invisible or occluded objects. Second, traditional indicators are poor at multi-object assessment. Theoretically, indicators such as Time-of-Collision (TTC) are based on the vehicle's own state and the state of other objects. Multi-object assessments can only be derived by comparing multiple values, making them difficult to apply directly in scenarios with an increased number of objects and complex horizontal and vertical dynamics. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a risk assessment method for multiple targets across line of sight, ensuring effective assessment of various risks, enabling autonomous driving functions to accurately perceive the surrounding environment and risks, and improving the safety capabilities of autonomous driving systems. To achieve the above-mentioned objectives and other advantages of the present invention, a risk assessment method for multiple targets across line of sight is provided, comprising:

[0005] S1. To predict the behavior and dynamics of perceived traffic participants and potential traffic participants in blind spots.

[0006] S2. Calculate the distribution area of ​​the perception blind spot in the current environmental space based on the perception situation;

[0007] S3. Based on some traffic constraints and potential risk assumptions, hypothetical target objects are set. The target object dynamic prediction module performs kinematic and dynamic modeling based on the perceived target object and the state information of the hypothetical target object, and predicts the set of reachable locations of the target object.

[0008] S4. Determine the intersection of the prediction space of all targets in the environment and the feasible space of the vehicle through a unified set expression method;

[0009] S5. By retrieving the set of hypothetical targets in the environment, assess the potential risks arising from the presence of the hypothetical targets;

[0010] S6. By retrieving the set of visible objects in the environment, assess the potential risks arising from the presence of visible objects.

[0011] Preferably, the method for calculating the perception blind spot uses the principle of visual field and employs a recursive shadow projection algorithm to calculate all invisible areas within the grid space.

[0012] Preferably, the space reachable by the vehicle within a short time t can be represented by a closed interval P consisting of the following points, and the specific calculation results are as follows:

[0013]

[0014] c x (t k ) for t k The position of the center of the circle in the x-direction at time c x (t k+1 ) for t k+1 The position of the center of the circle in the x-direction at time c; y (t k ) for t k The position of the center of the circle in the y-direction at time c y (t k+1 ) for t k+1 The position of the center of the circle in the y-direction at time t. k ) for t k The radius of the circle that can be reached at any time, r(t) k+1 ) for t k+1 The radius of the circle that can be reached at any time, b x b y t respectively k With t k+1 The points where the circle intersects in the x and y directions at any given moment. And, without considering steering, the reachable range of the vehicle is the envelope interval set P.

[0015] Preferably, the reachable range needs to be extended in a two-dimensional planar space, and the initial turning range is defined as ψ(0)∈[-ψ max ,ψ max ], by rotating ψ counterclockwise max Rotate q1, q2, and q3 counterclockwise to the desired angle. Similarly, rotate ψ max Rotate q4, q5, and q6 by the angle to The position of the envelope curve at ψ = 0 is the farthest point of the approximate envelope interval.

[0016] Preferably, the range of values ​​on the envelope can be approximated by the following envelope. Set representation:

[0017]

[0018] Because rotation results in a significant loss of sampling points on the outer arc, an approximate envelope curve is constructed by rotating q0. The rotation angle can be defined as a function of q0 and ψ. The entire envelope interval considering the turning direction can be defined as...

[0019]

[0020]

[0021] The preferred method for calculating the risk intersection is as follows:

[0022]

[0023] in For the intersection region of interest, Indicates that the vehicle is at t k Feasible space within time, Indicates other target objects at t k Feasible space within a given time frame.

[0024] Preferably, using a generalized evaluation system and any evaluation index that conforms to the additive principle, the risk assessment method for a specific location space is as follows:

[0025]

[0026] R T Where R represents the total risk at a spatial location. o Risks arising from blind spot participants; No represents the collection of blind spot participants; R v The risk generated by visible participants, where Nv is the set of visible participants; w p ,w v This is for adjusting the coefficient.

[0027] Compared with existing technologies, the beneficial effects of this invention are as follows: by designing a risk expression framework, it can integrate blind spot risk assessment capabilities on the one hand, and support multi-objective risk coupling assessment on the other. The method of this invention is a universal risk assessment method that can be well matched with real-world functional applications. It achieves both blind spot perception and prediction, and expands the comprehensive description capability of multi-objective risks. This method overcomes the limitations of traditional risk assessment in terms of limited scope and single objective. Based on the expression of spatial risk, it supplements the ability to predict blind spot risks and multi-objective coupled risks, making it particularly suitable for risk assessment in complex traffic scenarios. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the risk assessment method for multiple targets across line of sight according to the present invention;

[0029] Figure 2 A schematic diagram illustrating the blind zone calculation method according to an embodiment of the cross-line-of-sight multi-target risk assessment method of the present invention;

[0030] Figure 3 A schematic diagram illustrating the placement of dummy targets according to an embodiment of the cross-line-of-sight multi-target risk assessment method of the present invention;

[0031] Figure 4 This is a schematic diagram of blind zone risk distribution according to an embodiment of the cross-line-of-sight multi-target risk assessment method of the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Reference Figure 1-4 A multi-target risk assessment method across line-of-sight includes: a risk perception prediction module, which identifies all risk types within the perceptible range by predicting the behavior and dynamics of perceptible traffic participants and potential traffic participants in blind spots. Simultaneously, the spatial location expressed by the envelope algorithm is input into a global risk assessment module to evaluate the safety of each spatial location; the global risk assessment module evaluates the risk status of the vehicle's feasible location space based on the risk perception prediction input from the risk perception prediction module. This module realizes the calculation of the intersection of multi-target risks, achieving a unified assessment of blind spot risks and visible risks.

[0034] The specific steps include: S1, predicting the behavior and dynamics of perceived traffic participants and potential traffic participants in blind spots;

[0035] S2. Calculate the distribution area of ​​the perception blind zone in the current environmental space based on the perception situation; the perception blind zone calculation method will use a recursive shadow projection algorithm to calculate all invisible areas in the grid space based on the field of vision principle. Figure 2 The recursive algorithm for shadow casting is demonstrated, consisting of three main steps. First, the shadow casting algorithm divides the FOV calculation into eight octagonal circles. Second, the algorithm visits map cells row-by-row or column-by-column, starting from the nearest row or column and working outwards. For east- and west-facing sectors, a column-by-column approach is used. Figure 2 As shown in the left figure, the shadow projection will start scanning from (0,0), then from (1,1) to (1,0), and so on, until the sixth column ends. When the scan encounters an obstructing cell, it calculates which other cells in the row / column are invisible due to obstruction. In the third step, when an obstructing cell is found, a new scan will recursively begin, covering the area in the next row / column further away until the first cell in the shadow of the obstructing cell. The rest of the initial row / column is scanned, skipping subsequent obstructing cells directly adjacent to the initial obstruction. In this figure, one line defines the top of the sector, and another line defines the bottom; the displayed value is the slope of the line, and circled squares are considered visible. For each column, it scans down from the top tile within the sector to the bottom tile within the sector. In this figure, the first three columns have been scanned, and an obstruction was found in the fourth column, so the slope value was adjusted downwards. Figure 2 In the right figure, the transition from the feasible area to the obstacle has been found, and the bottom slope has been adjusted. The algorithm stops when the maximum line of sight is reached or all sectors become empty, i.e., the bottom slope is greater than the top slope.

[0036] S3. Based on some traffic constraints and potential risk assumptions, hypothetical target objects are set. The target object dynamic prediction module performs kinematic and dynamic modeling based on the perceived target object and the state information of the hypothetical target object, predicting the set of reachable locations of the target object. The placement of the hypothetical target follows three assumptions: First, a hypothetical target object exists at the edge of the occluded area, mimicking the same motion posture as the vehicle. Second, the hypothetical target object will follow traffic rules, i.e., it will not travel against traffic, will not change lanes arbitrarily, and will typically travel along the center line of the lane. Third, the hypothetical target object can be a vehicle, a non-motorized vehicle, or a pedestrian, but must comply with traffic constraints; for example, a hypothetical target object appearing under regular lane occlusion is a vehicle, and a hypothetical target object appearing on a crosswalk is a pedestrian. Figure 3 For example, vehicles may appear in the blind spots at shadows ① and ② based on the traffic network conditions. Blind spot ③ is a lane traveling from east to west, and the possibility of vehicles appearing from this point in the opposite direction is not considered; that is, potential risks are not assumed at ③ when calculating blind spot risk. It is assumed that the location of the appearing vehicle is at the boundary between the blind spot and the visible area, and the speed of the vehicle appearing in the blind spot will follow the vehicle's own speed. Based on the above information, with the blind spot boundary and the vehicle's speed in the blind spot clearly defined, this information is input into the global risk assessment module for dynamic reachability area prediction, calculating the risk domain generated by the blind spot target and the perceived target.

[0037] The space reachable by a vehicle within a short time t can be represented by a closed interval P consisting of the following points, and the specific calculation results are as follows:

[0038]

[0039] c x (t k ) for t k The position of the center of the circle in the x-direction at time c x (t k+1 ) for t k+1 The position of the center of the circle in the x-direction at time c; y (t k ) for t k The position of the center of the circle in the y-direction at time c y (t k+1 ) for t k+1 The position of the center of the circle in the y-direction at time t. k ) for t k The radius of the circle that can be reached at any time, r(t) k+1 ) for t k+1 The radius of the circle that can be reached at any time, b x b y t respectively k With t k+1 The point where the circle intersects in the x and y directions at any given moment.

[0040] Without considering steering, the reachable range of a vehicle is the aforementioned envelope interval set P. However, vehicle predictions often involve steering, thus requiring an extension of the reachable range in a two-dimensional planar space. The initial steering range is defined as ψ(0)∈[-ψ max ,ψ max The steering operation involves rotating the entire space P around the origin, depending on the vehicle's initial orientation. This is achieved by rotating ψ counterclockwise. max Rotate q1, q2, and q3 counterclockwise to the desired angle. Similarly, rotate ψ max Rotate q4, q5, and q6 by the angle to The position of the envelope curve at ψ = 0 is the farthest point of the approximate envelope interval. The range of values ​​on the envelope can be approximated by the following envelope. Set representation:

[0041]

[0042] Because rotation results in a significant loss of sampling points on the outer arc, an approximate envelope curve is constructed by rotating q0. The rotation angle can be defined as a function of q0 and ψ. The entire envelope interval considering the turning direction can be defined as...

[0043]

[0044]

[0045] S4. Determine the intersection of the prediction space of all targets in the environment and the feasible space of the vehicle using a unified set representation method; the risk intersection calculation method is as follows:

[0046]

[0047] in For the intersection region of interest, Indicates that the vehicle is at t k Feasible space within time, Indicates other target objects at t k Feasible space within a given time frame.

[0048] S5. By retrieving the set of hypothetical targets in the environment, assess the potential risks arising from the presence of the hypothetical targets;

[0049] S6. By retrieving the set of visible targets within the environment, assess the potential risks arising from the presence of these visible targets. The assessments in steps S5 and S6 are a generalized assessment system that can be calculated using any assessment index that conforms to the additive principle. The risk assessment method for a specific location space is as follows:

[0050]

[0051] R T Where R represents the total risk at a spatial location. o Risks arising from blind spot participants; No represents the collection of blind spot participants; R v The risk generated by visible participants, where Nv is the set of visible participants; w o ,w v This is for adjusting the coefficient. Figure 4 This diagram illustrates the risk distribution under blind spot occlusion, where all risks originate from vehicles in the blind spot. The black border represents the area the vehicle can reach within the next 2 seconds; the area within the border is considered safe. The selected risk calculation method is a spatiotemporal coupled driving risk description method based on occupancy grid network, where the darker the color, the greater the risk, which can prompt the autonomous driving control system to decelerate and avoid collisions.

[0052] In summary, this invention proposes a framework that enables both the perception and prediction of blind spots and expands the comprehensive description capabilities of multi-objective risks. This method overcomes the limitations of traditional risk assessment, such as narrow field of vision and singular objectives. Based on the expression of spatial risks, it supplements the ability to predict blind spot risks and multi-objective coupled risks, making it particularly suitable for risk assessment in complex traffic scenarios.

[0053] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention, and applications, modifications and variations thereof will be apparent to those skilled in the art.

[0054] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A risk assessment method for multiple targets across line of sight, characterized in that, Includes the following steps: S1. Behavior and dynamic prediction of perceived traffic participants and potential traffic participants in blind spots; the blind spot calculation method uses the field of view principle and a recursive shadow projection algorithm to calculate all invisible areas in the grid space; the space reachable by a vehicle in a short time t can be defined by the following closed intervals. The specific calculation results are as follows: ； in, for The position of the center of the circle in the x-direction at time [time]. for The position of the center of the circle in the x-direction at time 1; for The position of the center of the circle in the y-direction at time [time]. for The position of the center of the circle in the y-direction at that moment; for The radius of the circle can be reached at any time. for The radius of the circle can be reached at any time. respectively and The intersection points of the circle in the x and y directions at any given moment; and the reachable range of the vehicle without considering steering is the envelope interval set P; S2. Calculate the distribution area of ​​the perception blind spot in the current environmental space based on the perception situation; S3. Based on some traffic constraints and potential risk assumptions, hypothetical target objects are set. The target object dynamic prediction module performs kinematic and dynamic modeling based on the perceived target object and the state information of the hypothetical target object, and predicts the set of reachable locations of the target object. S4. Determine the intersection of the prediction space of all targets in the environment and the feasible space of the vehicle through a unified set expression method; S5. By retrieving the set of hypothetical targets within the environment, assess the potential risks arising from the presence of these hypothetical targets; the risk intersection calculation method is as follows: ； in For the intersection region of interest, Indicates that the car is Feasible space within time, Indicates other target objects in Feasible space within a given timeframe; S6. By retrieving the set of visible objects within the environment, assess the potential risks arising from the presence of these visible objects; using a generalized assessment system and any assessment index conforming to the additive principle, the risk assessment method for a specific location space is as follows: ； This represents the total risk of a single spatial location. Risks arising for participants in blind spots A collection of participants from the blind spot; Risks arising for visible participants A collection of visible participants; This is for adjusting the coefficient.

2. The risk assessment method for multiple targets across line of sight as described in claim 1, characterized in that, The reachable range needs to be extended in a two-dimensional planar space, and the initial turning range is defined as follows. Then the turning operation will rotate the entire space P around the origin; The method is to rotate counterclockwise. Angle will , , Rotate counterclockwise to Similarly, rotation Angle will Rotate to ; envelope curve in The position at that time is the farthest point of the approximate envelope interval. .

3. The risk assessment method for multiple targets across line of sight as described in claim 2, characterized in that, The range of values ​​on the envelope can be approximated by the following envelope. Set representation: ； Due to the rotation, there is a significant lack of sampling points on the outer arc. The rotational construction approximates the envelope curve, where the rotation angle can be defined as... and The function can be defined as the set of the entire envelope intervals considering the turning direction. : ； 。

Images