A method for calculating risk indicators based on reachable envelope partitioning

By predicting the reachable location set of autonomous vehicles through the reachable envelope partitioning method, and combining envelope partitioning estimation and risk assessment modules, the problems of insufficient lateral assessment and risk prediction in existing technologies are solved, and the risk description and quantification of complex motion situation scenarios are realized.

CN117196912BActive Publication Date: 2026-11-17TONGJI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311056914.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-08-22
Publication Date
2026-11-17
Estimated Expiration
2043-08-22

AI Technical Summary

Technical Problem

Existing risk quantification assessment indicators are mainly designed for rear-end collision scenarios, lacking the ability to assess risks across different directions and lacking the ability to predict risks in future motion states, thus failing to provide a comprehensive risk description in complex motion scenarios.

Method used

A risk index calculation method based on reachability envelope partitioning is adopted. Kinematic and dynamic modeling is performed through the reachability space calculation module. Combined with the envelope partitioning estimation module and the risk assessment module, the set of vehicle reachable locations is predicted. The envelope pattern is expressed through an approximation method, and a physically meaningful risk threshold is set to construct a generalized risk assessment framework.

Benefits of technology

It enables risk description of complex motion scenarios, predicts future motion states, provides risk assessment with horizontal and vertical coupling, has a risk benchmark with real physical significance, and supports risk quantification in autonomous driving scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117196912B_ABST
    Figure CN117196912B_ABST
Patent Text Reader

Abstract

This invention discloses a risk index calculation method based on reachable envelope partitioning, applied to the testing, verification, and evaluation of autonomous vehicles. The core objective is to achieve accurate prediction of the reachable location and motion state of a target object within a future time window through rapid computation. The method includes: calculating the reachable space of the target object and constraining the reachable space using approximation methods; providing a simpler representation of the reachable range space through envelope patterns; calculating the state representation and state estimation of each partition through state accumulation; integrating the envelope space and state; and performing a risk assessment for each independent location. According to this invention, reachable area prediction and representation can be achieved rapidly, spatial risk index calculation is supported, and it is suitable for describing risk scenarios with complex motion states, achieving a more comprehensive characterization of risk.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the technical field of automotive safety and testing verification, and in particular to a method for calculating risk indicators based on reachable envelope partitioning. Background Technology

[0002] The application of autonomous driving is showing an unstoppable trend. As more and more vehicles with these functions enter the market, the need for safety assessments is also gradually increasing. Ensuring the safety of autonomous driving remains a fundamental requirement, especially in terms of risk quantification under various conditions.

[0003] Existing risk quantification assessment metrics, such as Time to Collision (TTC) and Headway (TH), all focus on the interaction between the vehicle and other road users, aiming to describe the direct relationship between the vehicle's current motion state and that of other road users. While these metrics are widely used, some limitations remain unresolved. First, traditional metrics are designed for rear-end collision scenarios, thus lacking the ability to assess lateral collisions. Second, these metrics assess the current motion state and rarely express risk prediction information for that current state. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a risk index calculation method based on reachable envelope partitioning. This method can quickly predict and represent reachable regions, supports spatial risk index calculation, and is suitable for describing risk scenarios with complex motion states, thus achieving a more comprehensive characterization of risk. To achieve the above-mentioned objectives and other advantages of the present invention, a risk index calculation method based on reachable envelope partitioning is provided, comprising:

[0005] S1. The reachability space calculation module will perform kinematic and dynamic modeling based on the perceived target object state information to predict the set of reachable locations of the vehicle.

[0006] S2. The precise reachable location information is processed through an approximation method to give a spatial set expression of the envelope pattern. Thus, all state information can be approximated by the coordinates of a few points. The reachable space obtained from step S1 is a non-convex curve space, which has a large computational load when used to describe the envelope. Therefore, it is necessary to process the curve with an outer envelope through an approximation method.

[0007] S3. By characterizing the minimum distance belonging state of each spatial location point, the velocity and probability of each location interval within the envelope are described; where the risk probability can be expressed as the length of time that a target object appears at a spatial location (x, y) within time t.

[0008] S4. By combining spatial locations, perform expected calculations on all states at a specified location to obtain the state expression of the specified location;

[0009] S5. Construct a generalized risk indicator description mechanism to benchmark all risks. To provide an objective assessment of the safety of a spatial location, it is necessary to establish its real-world physical reference significance. Simply relying on motion state to calculate risk intensity cannot intuitively evaluate location safety. Therefore, a reasonably recognized risk threshold must be set. One of the main drawbacks of using traditional indicators is the lack of a threshold with clear physical meaning. Taking TTC as an example, different research standards suggest a threshold of 1–5 seconds. However, there is no consensus on choosing among these different scenarios, and applying the same threshold at different speeds may be ambiguous. Therefore, the threshold adopted is the potential danger that can be mitigated by the maximum effort of the autonomous vehicle when the space ahead is occupied.

[0010] S6. Assess whether a designated location is dangerous using a generalized risk assessment framework.

[0011] Preferably, it includes an reachable envelope calculation module, an envelope partition estimation module that is signal-connected to the reachable envelope calculation module, and a risk assessment module that is signal-connected to the envelope partition estimation module;

[0012] The reachable envelope calculation module is used to model the set of all locations that a target can reach within a future time period using general vehicle dynamics and kinematics methods. At the same time, it matches the target with the occupancy grid network method commonly used in vehicle motion prediction through envelope and approximation methods to quickly predict the reachable area of ​​the target. The occupancy space location expressed by the envelope algorithm is transmitted to the envelope partition estimation module, so that the envelope partition estimation module can calculate the set state represented by each partition.

[0013] The envelope partition estimation module is used to estimate the state of the target object at each location based on the vehicle position iteration and motion state prediction. It uses the rasterized information input from the reachable envelope calculation module to perform coupled calculation of the velocity and probability of the specified spatial location by predicting the expectation of the whole process. The state coupled with the partition estimation information is then transmitted to the risk assessment module.

[0014] The risk assessment module is used to calculate the corresponding risk value based on the state within the acceptance envelope partition and the minimum acceptable risk assessment model, and to assess whether a specified location is dangerous.

[0015] Preferably, the reachable envelope calculation method will predict the spatial position in the future within a short period of time under the ideal tire friction state. The calculation method for the core variables, the center position c(t) and the radius r(t), is as follows:

[0016]

[0017]

[0018] Among them, s x s y The positions in the x and y directions are respectively, and v x v y The velocities in the x and y directions are respectively, a max This is the vehicle's maximum braking speed.

[0019] Preferably, without considering the vehicle's size, the reachable set's envelope space can be represented by six vertices q1-q6, and the specific calculation results are as follows:

[0020] q1 = [c x (t k )-r(t k ),c y (t k )+r(t k )] T

[0021] q2=[b x (t k ),c y (t k+1 )+r(t k+1 )] T

[0022] q3=[c x (t k+1 )+r(t k+1 ),c y (t k+1 )+r(t k+1 )] T

[0023] q4=[c x (t k+1 )+r(t k+1 ),c y (t k+1 )-r(t k+1 )] T

[0024] q5=[b x (t k ),c y (t k+1 )-r(t k+1 )] T

[0025] q6=[c x (t k )-r(t k ),c y (t k )-r(t k)] T

[0026] Where q1-q6 are the coordinates of the vertices in the envelope space, and 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 These are the intersection points in the x and y directions, respectively.

[0027] The preferred method for calculating the boundary points of reachable spatial lines is as follows:

[0028]

[0029]

[0030] Where v0 is the vehicle speed, a max b is the vehicle's maximum braking speed. x b y These are the results of solving for the coordinate positions of the tangent at two time points.

[0031] Preferably, the method for calculating the hazard in step S5 is as follows:

[0032]

[0033] Among them, I b This is the risk baseline expression at the current spatial position (x, y), where v1 is the vehicle's speed, Δv is the speed relative to the vehicle, and d... b The reference distance is denoted as a, and c are coefficients. The reference distance used is:

[0034]

[0035] Preferably, the generalized risk assessment framework in step S6 is as follows:

[0036]

[0037] Where T is the risk value, P b I b P is the baseline expression at the current spatial position (x,y). b =1, The actual risk probability and risk intensity are represented by I. b The base is multiplied by the velocity of the target object.

[0038] Compared with existing technologies, the beneficial effects of this invention are as follows: This method predicts the state of each target object in a scene through a reachable envelope partitioning method, and then uses a risk benchmark comparison method with real physical meaning as a risk indicator. This risk indicator method has considerable predictive ability, can assess the risk of horizontal and vertical coupling, and balances ease of use and interpretability, thus describing the risk situation well. This invention obtains the motion state of the vehicle and other participants at the current moment, predicts the set of reachable locations within the next t time period through a reachable envelope algorithm, and uses the calculated risk state at the reachable location as an indicator to assess the risk. This method can solve the shortcomings of current traditional risk assessment indicators, such as the inability of TTC to perform short-term risk prediction and the lack of reference for the abstract threshold definition, providing a general risk description indicator. By setting up a reachable envelope calculation module, an envelope partitioning estimation module, and a risk assessment module, a simple and effective indicator covering predictive ability is described, uniformly describing the risk state of the driving process from a spatiotemporal dimension. This method can be applied to the field of autonomous driving, helping to solve the risk description of high-dynamic scenarios. This method can well meet the development and testing needs of current high-level autonomous driving, accelerating the application and implementation of autonomous driving functions. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the risk index calculation method based on reachable envelope partitioning according to the present invention;

[0040] Figure 2 This is a schematic diagram of reachability envelope estimation based on the risk index calculation method of reachability envelope partitioning according to the present invention;

[0041] Figure 3 This is a schematic diagram of an envelope grid considering vehicle size in the risk index calculation method based on reachability envelope partitioning according to the present invention.

[0042] Figure 4 This is a schematic diagram of the risk distribution at different times according to the risk index calculation method based on reachable envelope partitioning of the present invention. Detailed Implementation

[0043] 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.

[0044] Reference Figure 1-4 A risk indicator calculation method based on reachable envelope partitioning includes: a reachable envelope calculation module, an envelope partitioning estimation module signal-connected to the reachable envelope calculation module, and a risk assessment module signal-connected to the envelope partitioning estimation module;

[0045] The reachable envelope calculation module is used to model the set of all locations that a target can reach within a future time period using general vehicle dynamics and kinematics methods. At the same time, it matches the target with the occupancy grid network method commonly used in vehicle motion prediction through envelope and approximation methods to quickly predict the reachable area of ​​the target. The occupancy space location expressed by the envelope algorithm is transmitted to the envelope partition estimation module, so that the envelope partition estimation module can calculate the set state represented by each partition.

[0046] The envelope partition estimation module is used to estimate the state of the target object at each location based on the vehicle position iteration and motion state prediction. It uses the rasterized information input from the reachable envelope calculation module to perform coupled calculation of the velocity and probability of the specified spatial location by predicting the expectation of the whole process. The state coupled with the partition estimation information is then transmitted to the risk assessment module.

[0047] The risk assessment module is used to calculate the corresponding risk value based on the state within the acceptance envelope partition and the minimum acceptable risk assessment model, and to assess whether a specified location is dangerous.

[0048] The risk indicator calculation method includes the following steps: S1. The reachable space calculation module performs kinematic and dynamic modeling based on the perceived target object state information to predict the set of reachable positions of the vehicle; the specific calculation method is as follows: based on the Kammm circle method of ideal tire friction state, the spatial position in the future short time under the specified state is predicted, where the core variable, the circle center position c(t), and the radius r(t) are calculated as follows:

[0049]

[0050]

[0051] Where c(t) is the position of the center of the circle, r(t) is the radius, and s x s y The positions in the x and y directions are respectively, and vx v y The velocities in the x and y directions are respectively, a max This represents the vehicle's maximum braking speed. The method for calculating the reachable spatial boundary points is as follows:

[0052]

[0053]

[0054] Where v0 is the vehicle speed, a max b is the vehicle's maximum braking speed. x b y These are the results of solving for the coordinate positions of the tangent at two time points. Figure 2 The common envelope region of the middle circular part is the initially reachable envelope of the Kammm circle.

[0055] S2. The precise reachable location information is processed using an approximation method to give a spatial set representation of the envelope pattern. Thus, the entire reachable space information can be approximated using the coordinates of a few points. The reachable space obtained from S1 is a non-convex curve space, which involves a large computational load when used to describe the envelope. Therefore, an approximation method is needed to process the curve into an outer envelope. Without considering the vehicle's dimensions, the reachable set can be represented by six vertices q1-q6, and the specific calculation results are as follows:

[0056] q1 = [c x (t k )-r(t k ),c y (t k )+r(t k )] T

[0057] q2=[b x (t k ),c y (t k+1 )+r(t k+1 )] T

[0058] q3=[c x (t k+1 )+r(t k+1 ),c y (t k+1 )+r(t k+1 )] T

[0059] q4=[c x (t k+1 )+r(t k+1 ),c y (tk+1 )-r(t k+1 )] T

[0060] q5=[b x (t k ),c y (t k+1 )-r(t k+1 )] T

[0061] q6=[c x (t k )-r(t k ),c y (t k )-r(t k )] T

[0062] Where c is the center of the circle, r is the radius of the circle, and b x b y These are the intersection points in the x and y directions, respectively. Figure 2 The Chinese bounding box is the approximated spatial envelope, and q1-q6 are the vertices of the entire bounding box. This method can completely cover all spatial locations described by the Kamm circle in S1, while also being easier to compute and more easily represented in an occupied grid network. Figure 3 To consider the application of the vehicle size envelope space in the occupied grid network, the points outside the vehicle size envelope are considered to be Q1-Q6.

[0063] S3. By characterizing the minimum distance belonging state of each spatial location point, the velocity and probability of each location interval within the envelope are described. The risk probability can be expressed as the length of time a spatial location (x, y) appears in time t. For example, if the prediction duration of a scene is t = 1s and the prediction granularity is τ = 0.1s, then if a location is occupied within 0.6s, the risk probability is 0.6.

[0064] S4. By combining spatial locations, the expected value of all states at a specified location is calculated to obtain the state expression of that location. For example, if a scene prediction duration is t = 1s and the prediction granularity is τ = 0.1s, and a location is occupied within 0.6s, the velocity of the occupied location is 12m / s at 0.2s and 10m / s at 0.4s. That is, the velocity state of that location is...

[0065]

[0066] S5. Construct a generalized risk indicator description mechanism to benchmark all risks. To provide an objective assessment of the safety of a spatial location, it is necessary to establish its real-world physical reference significance. Simply relying on motion state to calculate risk intensity cannot intuitively evaluate location safety. Therefore, a reasonably recognized risk threshold must be set. One of the main drawbacks of using traditional indicators is the lack of a threshold with clear physical meaning. Taking TTC as an example, different research standards suggest a threshold of 1–5 seconds. However, there is no consensus on choosing among these different scenarios, and applying the same threshold at different speeds may be ambiguous. Therefore, the threshold adopted is the potential danger that can be mitigated by the maximum effort of the autonomous vehicle when the space ahead is occupied. The calculation method for this danger is as follows:

[0067]

[0068] Among them, I b This is the risk baseline expression at the current spatial position (x, y), where v1 is the vehicle's speed, Δv is the speed relative to the vehicle, and d... b Let a and c be the reference distances, and a and c be the coefficients, respectively. The reference distance used is:

[0069]

[0070] Where, d b v1 is the base distance, v1 is the vehicle speed, and a is the speed of the vehicle. max This is the vehicle's maximum braking speed.

[0071] S6. Using a generalized risk assessment framework, assess whether a specified location is hazardous, where the risk baseline value is calculated as follows:

[0072]

[0073] Where T is the risk value, P b I b P is the baseline expression at the current spatial position (x,y). b =1, The actual risk probability and risk intensity are represented by I. b The base is multiplied by the velocity of the target object.

[0074] Figure 4 This section demonstrates the risk distribution during a vehicle approach scenario from 0 to 0.4 seconds using this metric calculation method. The risk is presented as a heatmap, effectively illustrating the real-time assessment of the autonomous vehicle's risk distribution during the approach process. Darker colors indicate greater risk. Locations with a risk value greater than 1 indicate a collision risk that is difficult to avoid even with maximum braking capacity.

[0075] 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.

[0076] 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 method for calculating risk indicators based on reachable envelope partitioning, characterized in that, Includes the following steps: S1. The reachability space calculation module will perform kinematic and dynamic modeling based on the perceived target object state information to predict the set of reachable locations of the vehicle. S2. The precise reachable location information is processed using an approximation method to give a spatial set representation of the envelope pattern; The method for calculating the boundary points of reachable space lines is as follows: ; ; wherein, is the vehicle speed, is the maximum braking speed of the vehicle, is the solution of the tangent coordinate position at two times; S3. By characterizing the minimum distance belonging state of each spatial location point, the velocity and probability of each location interval within the envelope are described. S4. By combining spatial locations, perform expected calculations on all states at a specified location to obtain the state expression of the specified location; S5. Construct a generalized risk indicator description mechanism to express all risks in a benchmark form; the calculation method for the benchmark risk in step S5 is as follows: ; in At the current spatial location Risk benchmark expression below, For the vehicle's speed, The speed is relative to the vehicle's speed. The reference distance is denoted as a, and c are coefficients. The reference distance used is: ; in, As the reference distance, For the vehicle's speed, This refers to the vehicle's maximum braking speed. S6. Using a generalized risk assessment framework, assess whether a specified location is hazardous. The calculation method for the generalized risk assessment framework in step S6 is as follows: ; in This is the risk value. At the current spatial location The baseline expression below, =1, , The actual risk probability and risk intensity are given by: The base is multiplied by the velocity of the target object.

2. The risk index calculation method based on reachable envelope partitioning as described in claim 1, characterized in that, It includes an reachable envelope calculation module, an envelope partition estimation module that is signal-connected to the reachable envelope calculation module, and a risk assessment module that is signal-connected to the envelope partition estimation module; The reachable envelope calculation module is used to model the set of all locations that a target can reach within a future time period using general vehicle dynamics and kinematics methods. At the same time, it matches the target with the occupancy grid network method commonly used in vehicle motion prediction through envelope and approximation methods to quickly predict the reachable area of ​​the target. The occupancy space location expressed by the envelope algorithm is transmitted to the envelope partition estimation module, so that the envelope partition estimation module can calculate the set state represented by each partition. The envelope partition estimation module is used to estimate the state of the target object at each location based on the vehicle position iteration and motion state prediction. It uses the rasterized information input from the reachable envelope calculation module to perform coupled calculation of the velocity and probability of the specified spatial location by predicting the expectation of the whole process. The state coupled with the partition estimation information is then transmitted to the risk assessment module. The risk assessment module is used to calculate the corresponding risk value based on the state within the acceptance envelope partition and the minimum acceptable risk assessment model, and to assess whether a specified location is dangerous.

3. The risk index calculation method based on reachable envelope partitioning as described in claim 1, characterized in that, The reachability envelope calculation method in step S1 will predict the spatial position in the future short time under the ideal tire friction state, where the core variable is the center position. With radius size The calculation method is as follows: ; ; in The center position of the variable. For the radius size, These represent the positions in the x and y directions, respectively. The velocities are in the x and y directions, respectively. This is the vehicle's maximum braking speed.

4. The risk index calculation method based on reachable envelope partitioning as described in claim 1, characterized in that, Without considering the vehicle's dimensions, the reachable space set in step S2 can be represented by six vertices q1-q6, and the specific calculation results are as follows: ; Where q1-q6 are the coordinates of the vertices in the envelope space. 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. These are the intersection points in the x and y directions, respectively.