A spatiotemporal coupled driving risk description method

Abstract

The present invention discloses a spatiotemporally coupled driving risk description method, comprising: S1, obtaining the position information and dynamic information of all sensed objects within the perception area of ​​the autonomous driving vehicle and initializing and placing all sensed information in a map; S2, simultaneously performing risk probability calculation and risk intensity calculation; S3, performing a quantitative coupled risk calculation using the received risk probability grid map and risk intensity grid map, and outputting a risk value as the spatial risk distribution within a future observation time domain under the current motion state of the vehicle. According to the present invention, the actual risk situation in a specific autonomous driving scenario can be displayed from the two dimensions of risk probability and intensity, and is applicable to complex traffic scenarios with horizontal and vertical coupling, achieving a more realistic and universal characterization of risk.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile safety and test verification, and in particular to a time-space coupled driving risk description method. Background Art

[0002] These requirements are increasingly high. Autonomous vehicles, based on computers and modern automotive industry technologies, are enabled by digitalization and intelligence. They represent a new era in the automotive industry and are expected to become a vital component of future transportation. Therefore, research into safety assessment techniques within the development of autonomous driving technology can significantly promote technological iteration and product upgrades, while also significantly increasing consumer trust in high-level autonomous vehicles. Ensuring the safety of autonomous driving functions has always been a fundamental requirement, particularly quantifying risks under various conditions.

[0003] Currently, various approximate surrogate indicators are used for safety assessment, which can be divided into time-based indicators, behavior-based indicators and distance indicators. Different indicators have different optimal application ranges, depending on the type of indicator. For example, time-based indicators such as time to collision (TTC) and travel time (TH) are the most commonly used indicators because they are easy to use and have universal applicability. Indicators such as MTTC, TIT and TET are derived from TTC by considering factors such as acceleration and hazard duration. Behavior-based indicators such as the deceleration rate to avoid collision (DRAC) are ideal for describing braking behavior, while the potential collision severity index (PCSI) takes into account the impact of the collision angle and is an excellent indicator for measuring collision severity. Distance indicators such as the Responsibility Sensitive Safety (RSS) model proposed by Intel are effective in evaluating longitudinal and lateral safety distances.

[0004] While the aforementioned proximal surrogate metrics have been widely used, certain limitations remain unaddressed. These metrics are all related to evaluating potential collision objects, which can be challenging in complex multi-object scenarios. Specifically, many methods lack evaluation of lateral scenarios and show poor consistency across multiple metrics. As test evaluation requirements become increasingly complex, a universal evaluation method is necessary. A key solution is to shift from evaluating the relationship between the ego vehicle and objects as a proxy metric to evaluating the risk distribution within the space available to the ego vehicle. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a spatiotemporal coupled driving risk description method that can display the actual risk situation in a specific autonomous driving scenario from the two dimensions of risk probability and intensity. It is applicable to complex traffic scenarios with horizontal and vertical coupling, and can achieve a more realistic and universal characterization of risks. In order to achieve the above-mentioned purpose and other advantages of the present invention, a spatiotemporal coupled driving risk description method is provided, comprising:

[0006] S1. Obtain the position information and dynamic information of all sensed objects in the sensing area of ​​the autonomous driving vehicle and initialize and place all sensed information in the map;

[0007] S2. Calculate risk probability and risk intensity simultaneously;

[0008] S3. Quantitatively couple the risk calculations using the accepted risk probability grid map and the risk intensity grid map, and output the risk value as the spatial risk distribution within the future observation time domain under the current motion state of the vehicle.

[0009] It also includes a map rasterization module, a risk intensity calculation module and a risk probability calculation module connected to the map rasterization module signal, and a spatiotemporally coupled risk quantification description module connected to the risk intensity calculation module and the risk probability calculation module signal.

[0010] Preferably, the map rasterization module is used to map the position information and dynamic information of all sensed objects within the sensing range of the autonomous driving vehicle into a grid by occupying a grid grid method, initialize and place all sensed information in the map, and after processing the rasterization method, pass the grid map information to the risk probability calculation module and the risk intensity calculation module;

[0011] The risk probability calculation module is used to estimate the probability of a moving object appearing in the future space and time based on the information obtained from the rasterization of the map. It also obtains information about all other traffic participants in the raster map and calculates the updated distribution of their spatial positions in a short time window in the future.

[0012] The risk intensity calculation module is used to calculate the intensity of the collision between an obstacle at a specified spatial position and the current motion state of the vehicle based on the information of all perceived objects received from the grid map.

[0013] Preferably, the risk intensity calculation module calculates the collision risk assessment of the traffic participant's motion state information and the vehicle's motion state at each spatial position with occupancy probability based on the current traffic participant's state information and dynamic information in the perception area; wherein the intensity distribution in the entire space is the coupled superposition of risk intensities under different targets.

[0014] Preferably, the spatiotemporal coupled risk quantification description module will construct a coupled risk quantification model based on the risk quantification algorithm provided by ISO 26262, and perform a comprehensive evaluation and calculation of the risk probability and risk intensity calculated in step S2.

[0015] Preferably, the risk probability in step S2 is calculated as follows:

[0016]

[0017] Among them, P (x,y) is the sum of the probabilities of all traffic participants appearing at the specified spatial location (x, y), T is the observation time window, and t is each time step in the time window.

[0018] Preferably, the risk intensity in step S2 is calculated as follows:

[0019]

[0020] Among them, I (x,y) is the risk intensity at the current spatial position (x, y), K is the set of traffic participants, k is the currently selected traffic participant, v1 is the speed of the vehicle, v k is the selected traffic participant, d is the spatial distance between the two vehicles, Δv is the relative collision speed of the two vehicles, ε and b are adjustment factors.

[0021] Preferably, the quantitative coupling calculation of risk in step S3 is as follows:

[0022] R (x,y) =P (x,y) ×I (x,y)

[0023] Among them, R (x,y) is the coupling risk at the current spatial position (x, y), P (x,y) is the sum of the probabilities of all traffic participants appearing at the specified spatial location (x, y), I (x,y) is the risk intensity at the current spatial position (x, y).

[0024] Compared to existing technologies, the present invention offers the following advantages: a comprehensive risk description using a coupled risk probability and risk intensity model accurately reflects the actual risk situation. It obtains information about the status of traffic participants within the current perception range and establishes a risk assessment metric that is easy to calculate and solve. This method, modeled from two dimensions, namely, the probability of an accident and its intensity, provides a comprehensive risk description method. This method can replace traditional risk assessment metrics, such as TTC, that describe conventional risk scenarios. Another key advantage is its ability to assess longitudinal and transverse coupled risks, making it particularly intuitive in cut-in and cross-cutting scenarios. The temporal-spatial risk distribution method constructed in this invention can identify potential risks more quickly than traditional metrics, demonstrating its superiority in the field of hazard assessment. By implementing a map rasterization module, a risk probability calculation module, a risk intensity calculation module, and a spatiotemporal coupled risk module, it effectively and efficiently describes the risk distribution of the driving process from multiple dimensions and perspectives. This method can be applied to the field of autonomous driving, helping to address motion planning and control in complex spatiotemporal coupled scenarios and risk assessment in highly dynamic, multi-objective scenarios. This method is well-suited to the current development and testing needs of high-level autonomous driving, accelerating the implementation of autonomous driving features. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a flow chart of the spatiotemporal coupled driving risk description method according to the present invention;

[0026] Figure 2 A schematic diagram of map rasterization of the spatiotemporal coupled driving risk description method according to the present invention;

[0027] Figure 3 A schematic diagram of risk probability accumulation calculation of the spatiotemporal coupled driving risk description method according to the present invention;

[0028] Figure 4 Schematic diagram of the spatiotemporal coupling risk distribution of an intersection scenario according to the spatiotemporal coupling driving risk description method of the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0030] Reference Figure 1-4, a spatiotemporal coupled driving risk description method, comprising: S1, a map rasterization module obtains the position information and dynamic information of all sensed objects in the vehicle's perception area, divides the current map into 1m*1m grid map information according to the position of the autonomous driving vehicle, and transmits the grid map information to the risk probability calculation module S2 and the risk intensity calculation module S3;

[0031] S2. The risk probability calculation module receives all the sensed object information of the grid map, deduces the motion state evolution process in the future prediction time domain based on the sensed information, calculates the probability of occupancy at each spatial location based on the motion state evolution, and outputs the output probability grid map to the spatiotemporal coupled risk quantification description module;

[0032] S3. The risk intensity calculation module calculates the intensity of an obstacle that would collide with the current motion state of the autonomous vehicle if an obstacle appears at a specific spatial location based on the information of all sensed objects in the received grid map, and outputs the intensity grid map to the risk coupling calculation module.

[0033] S4. Based on the accepted risk probability grid map and risk intensity grid map, perform a quantitative coupled calculation of the risk, and output the risk value as the spatial risk distribution in the future observation time domain under the current motion state of the vehicle.

[0034] It also includes a map rasterization module, a risk probability calculation module, a risk intensity calculation module and a time-space coupled risk description module. The map rasterization module obtains information within the perception control area, divides the spatial attributes into spatial grids, and inputs the divided spatial information into the risk probability calculation and risk intensity calculation modules; the risk probability calculation will perform time domain deduction on the target objects in the grid map, and output a risk probability quantization grid based on the occupancy situation; the risk intensity calculation module receives the vehicle status and other target object status in the perception control area, performs multi-dimensional risk coupling modeling based on the motion situation, and outputs a risk intensity quantization grid; the time-space coupled risk description module will integrate the risk probability grid and risk intensity grid information, make a comprehensive judgment, and output the quantitative result of the risk.

[0035] Furthermore, the step S1 specifically includes the following steps:

[0036] S11, the map rasterization module obtains the position information and dynamic information of all sensed objects in the sensed area;

[0037] S12. Divide the map into grids around the autonomous vehicle, with each grid defined as 1m*1m in size. Place the traffic participant objects identified by the perception module into the grid map, where grids occupied by traffic participants are represented as 1 and unoccupied grids are represented as 0. Figure 2As shown in the left figure, in a given intersection scene, the opposite lane of the autonomous vehicle is blocked by a median strip, there is a truck on the west side, and a passenger car on the east side. Figure 2 The grid map status on the right shows the light-colored parts representing autonomous vehicles and the black parts representing the current occupation of traffic participants on the map.

[0038] Furthermore, the step S2 specifically includes the following steps:

[0039] S21. The risk probability calculation module performs kinematic calculations based on the current state and dynamic information of traffic participants in the sensing area to evaluate the probability distribution over a future time scale. The probability distribution over the entire space is a coupled superposition of different time granularities. The risk probability calculation method is:

[0040]

[0041] Among them, P (x,y) is the sum of the probabilities of all traffic participants appearing at the specified spatial location (x, y). T is the time window of observation, and t is the length of each time step within the time window. Figure 3 As shown in the figure, if T is set to 1s and the time step t is 0.2s, the position states at the time of 0-0.2s, 0.2-0.4s, and 0.4-0.6s are deduced respectively according to the kinematic formula, where the kinematic formula is expressed as:

[0042] x t+1 =x+v k,x *t+0.5*a x *t^2

[0043] y t+1 =y+v k,y *t+0.5*a y *t^2

[0044] Among them, x t+1 and y t+1 Respectively mark the farthest position that can be reached in the vertical and horizontal directions in the next time step, v k,x x, v k,y are the velocity components of the target object k in the horizontal and vertical directions, a x and a y Represents the horizontal and vertical accelerations respectively. t is the time step length in the time window, which is 0.2s here. Figure 3 In the above equation, assuming that the probability of each part of the movement is equal, that is, the probability of the vehicle in each space-time segment is 0.2, the above results can be accumulated and calculated based on the calculation results to obtain the comprehensive probability distribution situation.

[0045] Furthermore, the step S3 specifically includes the following steps:

[0046] S31. The risk intensity calculation module calculates the collision risk assessment based on the current state and dynamic information of traffic participants within the perception area, combining the motion state information of traffic participants with the motion state of the ego vehicle at each spatial location with occupancy probability. The intensity distribution across the entire space is a coupled superposition of risk intensities under different targets. The risk intensity calculation method is as follows:

[0047]

[0048] Among them, I (x,y) is the risk intensity at the current spatial position (x, y). K is the set of traffic participants, k is the currently selected traffic participant, v1 is the speed of the vehicle, v k is the selected traffic participant, d is the spatial distance between the two vehicles, Δv is the relative collision speed of the two vehicles, ε and b are adjustment factors.

[0049] Furthermore, the step S4 specifically includes the following steps:

[0050] S41. The spatiotemporal coupled risk description module will construct a coupled risk quantification model based on the risk quantification algorithm provided by ISO 26262, and comprehensively evaluate and calculate the risk probability and risk intensity calculated by S2 and S3. The spatiotemporal coupled risk probability calculation method is:

[0051] R (x,y) =P (x,y) ×I (x,y)

[0052] Among them, R (x,y) is the coupling risk at the current spatial position (x, y), P (x,y) is the sum of the probabilities of all traffic participants appearing at a specified spatial location (x, y). (x,y) is the risk intensity at the current spatial position (x, y). Figure 4 This diagram shows the spatiotemporal coupled driving risk distribution for an intersection scenario, calculated using the inventive method. The white vehicles in the diagram represent the driver's vehicle, and the blue vehicles represent other traffic participants. The heat map color blocks indicate the spatial distribution of risk. The diagram calculates the risk distribution within the reach of other traffic participants 1 second into the current motion. Redder locations in the diagram indicate higher risk, while greener locations indicate lower risk. Figure 4 It intuitively shows that in an intersection scenario, the coupling risk is higher on the side close to the vehicle, and lower on the side far from the vehicle, verifying that this method can display the distribution of risks in horizontal and vertical coupling scenarios.

[0053] In summary, this invention uses risk intensity and risk probability as indicators to calculate the combined risk of spatial coupling, supported by a grid map. This approach overcomes the object-oriented nature of traditional risk quantification assessment indicators and, through a spatially oriented risk distribution approach, faithfully restores the risk distribution state within a spatial state. This indicator calculation method is simple, fast, and applicable to multi-target, highly dynamic scenarios, making it particularly well-suited for risk quantification assessment in autonomous driving environments.

[0054] By implementing a map rasterization module, a risk probability calculation module, a risk intensity calculation module, and a spatiotemporal coupled risk module, this method effectively and efficiently describes the risk distribution of the driving process from multiple dimensions and perspectives. This method can be applied to the field of autonomous driving, helping to solve motion planning and control in complex spatiotemporal coupled scenarios and risk assessment in highly dynamic multi-target scenarios. This method is well-suited to the current development and testing needs of high-level autonomous driving, accelerating the implementation of autonomous driving capabilities.

[0055] The number of devices and processing scales described herein are intended to simplify the description of the present invention, and applications, modifications, and variations of the present invention will be apparent to those skilled in the art.

[0056] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A spatiotemporal coupled driving risk description method, characterized in that: The following steps are involved: S1. Obtain the position information and dynamic information of all sensed objects in the sensing area of ​​the autonomous driving vehicle and initialize and place all sensed information in the map; S2. Calculate the risk probability and risk intensity simultaneously. The risk probability calculation in step S2 is as follows: ； in, To specify a spatial location The sum of the probabilities of all traffic participants appearing on the time window, T is the time window of observation, and t is the length of each time step in the time window; The risk intensity is calculated as follows: ； in, At the current spatial position The risk intensity under the condition, K is the set of traffic participants, For the currently selected traffic participant, is the vehicle speed, is the selected traffic participant, d is the spatial distance between the two vehicles, is the relative collision speed of the two vehicles, and b are adjustment factors; S3. Quantitative coupling calculation of risk is performed using the accepted risk probability grid map and risk intensity grid map. The output risk value is used as the spatial risk distribution in the future observation time domain under the current motion state of the vehicle. The quantitative coupling calculation of risk is as follows: ； in At the current spatial location The coupling risk under To specify a spatial location The sum of the probabilities of all traffic participants appearing on At the current spatial position the intensity of risk under It also includes a map rasterization module, a risk intensity calculation module and a risk probability calculation module that are signal-connected to the map rasterization module, and a spatiotemporally coupled risk quantification description module that is signal-connected to the risk intensity calculation module and the risk probability calculation module.

2. The spatiotemporal coupled driving risk description method according to claim 1, characterized in that: The map rasterization module is used to map the position information and dynamic information of all sensed objects within the sensing range of the autonomous vehicle to a grid by occupying the grid grid method, and initialize all sensed information in the map. After processing the rasterization method, the grid map information is passed to the risk probability calculation module and the risk intensity calculation module; The risk probability calculation module is used to estimate the probability of a moving object appearing in the future space and time based on the information obtained from the rasterization of the map. It also obtains information about all other traffic participants in the raster map and calculates the updated distribution of their spatial positions in a short time window in the future. The risk intensity calculation module is used to calculate the intensity of the collision between an obstacle at a specified spatial position and the current motion state of the vehicle based on the information of all perceived objects received from the grid map.

3. The spatiotemporal coupled driving risk description method according to claim 2, characterized in that: The risk intensity calculation module calculates the collision risk assessment of the traffic participant motion state information and the vehicle motion state at each spatial location with occupancy probability based on the current traffic participant status information and dynamic information in the perception area. Among them, the intensity distribution in the entire space is the coupled superposition of risk intensities under different targets.

4. The spatiotemporal coupled driving risk description method according to claim 3, characterized in that: The spatiotemporal coupled risk quantification description module will construct a coupled risk quantification model based on the risk quantification algorithm provided by ISO 26262, and conduct a comprehensive evaluation and calculation of the risk probability and risk intensity calculated in step S2.

Images