Expected function safety risk assessment method

Abstract

The invention discloses an expected function safety risk assessment method, and relates to the technical field of automatic driving. The method is based on an SOTIF scene architecture, collision severity S is calculated through collision object types and relative speed, corresponding information is acquired from external environment by using a current mainstream BEV + Occupancy Network perception algorithm, collision risk and detection capability of the BEV + Occupancy Network perception algorithm are calculated, and the collision severity S is calculated according to the detection capability of the BEV + Occupancy Network perception algorithm and the detection capability of the BEV + Occupancy Network perception algorithm and the detection capability of the BEV + Occupancy Network perception algorithm and the detection capability of the BEV + Occupancy Network perception algorithm. Finally, the risk priority number of the vehicle in the current environment is output in real time according to the three indexes, whether a takeover signal is sent out or not is judged according to the total risk priority number obtained after normalization, and improvement measures are made. According to the method, the accuracy of a sensing system is improved, and the unsafe risk that expected function safety is known and unknown is reduced.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method for assessing expected functional safety risks. Background Art

[0002] With the rapid development of autonomous driving technology, ensuring the safe and reliable driving of vehicles in complex environments has become the current research focus.

[0003] There is no public definition standard for the driving risks of existing autonomous vehicles. First of all, there is no consistent definition of the concept of risk. There are many different ways to understand the concept of risk. Some definitions are based on its origin, some are based on adverse events or dangers, and others are based on uncertainty. In terms of the definition and specific process of risk assessment, the International Organization for Standardization defines risk assessment as "the whole process of risk identification, risk analysis and risk assessment". In the risk identification stage, all possible risks are identified. Risk identification is a process of understanding the nature of risk and determining risks. At the same time, the main concern of expected functional safety is the hazards caused by the limitations of sensors and algorithms in open and complex environments. The ultimate goal of the expected function is to reduce the probability of known and unknown unsafe scenarios to a reasonable and acceptable level. In order to achieve this goal, it is necessary to analyze and identify risks in combination with expected functional safety, further enhance the perception ability, decision logic and execution mechanism of the autonomous driving system, and avoid the occurrence of false detection, missed detection and other phenomena. However, there is currently a lack of methods that can conduct real-time risk assessment and quantification in combination with expected functional safety requirements.

[0004] In terms of risk assessment, functional safety and safety of intended functions (SOTIF) have become important research areas for autonomous driving perception systems. Among them, intended functional safety enhances the overall safety of autonomous driving by predicting possible system failures and risk scenarios. In particular, when encountering emergencies or unexpected environmental changes, the system must have the ability to accurately identify risks and respond quickly. Therefore, when an autonomous driving vehicle is running, the system needs to conduct effective risk assessment. When the detected risk is higher than expected, it must be able to prompt the driver to take over in time, which is crucial to the safety of the autonomous driving system. At the same time, the perception algorithm, as the core of the environmental perception of the autonomous driving system, relies on sensors such as cameras and LiDAR, and analyzes environmental information through deep learning algorithms. In the current autonomous driving risk assessment technology, the calculation of external autonomous driving risks is obtained by the vehicle's perception system to obtain corresponding information, so as to evaluate the current risk of the vehicle from different angles.

[0005] SOTIF emphasizes whether the system can effectively cope with complex or unknown environments when the system hardware or software does not fail. It focuses on the completeness of system functions, especially whether it can provide expected functions in unexpected situations. For example, in rainy and snowy weather, when pedestrians or road obstacles suddenly appear, the perception system must detect these risks in a timely and accurate manner and make safe driving decisions. The ASIL risk level assessment system proposed by ISO 26262 classifies functional safety through three parameters: severity (S), exposure rate (E) and controllability (C), ensuring that the autonomous driving system can conduct effective risk assessment when encountering emergencies. Most of the existing expected functional safety risk assessment methods rely on exhausting all possible requirements and calculating their risks, but the factors of the real environment cannot be fully listed. Therefore, it is necessary to find solutions that can be implemented in practice and ensure the safety and reliability of perception algorithms in various complex scenarios, especially how to effectively avoid the error risks of perception algorithms in practical applications to ensure the comprehensive safety of autonomous driving systems. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a method for evaluating the expected functional safety risk under a perception algorithm in view of the shortcomings of the above-mentioned prior art.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0008] An expected functional safety risk assessment method comprises the following steps:

[0009] Step 1: Based on the SOTIF scenario architecture, the collision severity S is calculated by the collision object type and relative speed, and the collision severity S is normalized and given a corresponding score;

[0010] Step 1.1: Determine the type of collision object and set different types of fixed scores for different collision objects;

[0011] Step 1.2: Determine the collision speed and collision target category and their corresponding weights, adjust the impact of different factors on severity according to the weights, and calculate the collision severity S;

[0012] The collision severity S can be expressed by the following formula:

[0013] S=ω1*ObjectType+ω2*RelativeSpeed

[0014] Among them, ω1 and ω2 represent the weights of the collision target category and the collision speed. The weight of the collision speed is set in segments according to the size of the collision speed. ObjectYype represents the collision target category. The collision target category and the corresponding weight are set with reference to the expected functional safety standard and expert guidance. RelativeSpeed ​​represents the collision speed, which refers to the relative speed between the vehicle and the collision object.

[0015] Step 1.3: normalize the calculated collision severity S, and assign corresponding scores to the collision severity S with values ​​in different ranges after the normalization;

[0016] Step 2: Calculate the collision risk O based on a single collision object of a simplified physical model, normalize the collision risk O and give a corresponding score;

[0017] Step 2.1: Obtain relevant video frame data Data of the vehicle from multiple sensors of the vehicle perception system;

[0018] Step 2.2: The video frames captured by different cameras at the same time are fused to generate a BEV image. At the same time, the depth map of the video frame at the same time is generated and projected into the BEV image to obtain the depth value of the corresponding pixel, which is sent to the BEV+OccupancyNetwork perception algorithm to obtain the relevant output information at the i-th second, including the distance between the collision target and the vehicle, the moving distance of the collision target's horizontal coordinate, and the moving distance of the collision target's vertical coordinate;

[0019] Based on the relevant output information at the i-th second obtained by the BEV+OccupancyNetwork perception algorithm, the time when the vehicle collides with the vehicle in front is first calculated. The calculation formula is as follows;

[0020]

[0021] Among them, TTC i is the predicted time of collision between the vehicle and the front vehicle at the i-th second, d i is the distance between the collision target and the vehicle at the i-th second, RelativeSpeed i is the collision velocity at the ith second,

[0022] Step 2.3: Calculate the position change and movement direction of the collision target in the occupied network. The calculation formula is as follows;

[0023] θ i =arctan2(Δx i ,Δy i )

[0024] Among them, θ iis the heading angle of the moving direction of the collision target at the i-th second, Δx i , Δy i are the moving distance of the horizontal coordinate and the moving distance of the vertical coordinate of the collision target at the i-th second respectively;

[0025] Step 2.4: Calculate the collision avoidance deceleration rate, which represents the deceleration speed required to avoid a collision at the current distance between the vehicle and the collision target. The calculation formula is as follows:

[0026]

[0027] Among them, DRAC i represents the collision avoidance deceleration rate at the i-th second, v ei represents the speed of the vehicle at the i-th second, and e represents the vehicle;

[0028] Step 2.5: Calculate the steering threat coefficient, which represents the steering threat coefficient based on the heading angle difference at the current distance between the vehicle and the collision target. It is used in scenarios such as lane change and turning. The calculation formula is as follows:

[0029]

[0030] Among them, STN i represents the turning threat coefficient at the i-th second, θ ei ,θ oi Respectively represent the heading angles of the ego vehicle and the collision target at the i-th second;

[0031] Step 2.6: Calculate the minimum safe distance, which means the minimum safe distance under the current speed and acceleration. The calculation formula is as follows:

[0032]

[0033] Among them, MSD i represents the minimum safe distance in the ith second, a ei represents the acceleration of the vehicle at the i-th second;

[0034] Step 2.7: Calculate the collision distance, which is the difference between the current distance and the minimum safe distance. The calculation formula is as follows:

[0035] DTC i =d i -MSD i

[0036] Among them, DTC i represents the collision distance at the i-th second;

[0037] Step 2.8: By considering the three indicators of collision time, collision avoidance deceleration rate and collision distance, combined with the occupation probability of each collision target in the spatial volume unit output by the occupation network, the collision risk of the corresponding unit is calculated, and the total collision risk O is calculated based on the collision risk of the corresponding unit;

[0038] The total collision risk O is calculated as follows:

[0039]

[0040]

[0041] Among them, O collision,b represents the collision risk of the bth spatial volume unit, b represents the spatial volume unit, P occ,b represents the probability that the bth spatial volume unit is occupied, and N is the total number of spatial volume units;

[0042] Step 2.9: normalize the calculated total collision risk O, and assign corresponding scores to the total collision risk O with values ​​in different ranges after the normalization;

[0043] Step 3: Calculate and evaluate the detection capability D of the BEV+OccupancyNetwork perception algorithm and give a corresponding score;

[0044] Step 4: Calculate the total risk priority number based on the collision severity S, collision risk O, and the detection capability D of the BEV+OccupancyNetwork perception algorithm;

[0045] Step 5: Normalize the total risk priority number and give corresponding weights according to the corresponding levels. Finally, determine whether to issue a takeover signal based on the size of the normalized total risk priority number.

[0046] Furthermore, in step 3, the detection capability D of the BEV+OccupancyNetwork perception algorithm is evaluated using the F1 score that can represent the emphasis on missed detection and false detection.

[0047] Furthermore, in step 4, the formula for calculating the total risk priority number is as follows:

[0048] R 总 =S*O*D

[0049] Where R 总 Indicates the total risk priority number.

[0050] The beneficial effect of adopting the above technical solution is that: the invention provides an expected functional safety risk assessment method, which utilizes the current mainstream BEV+OccupancyNetwork perception algorithm of autonomous driving vehicles to obtain corresponding information about the external environment, and can output the risk value of the vehicle in the current environment in real time, thereby improving the accuracy of the perception system and reducing the known and unknown unsafe risks of expected functional safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flowchart of a method for evaluating expected functional safety risks provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0053] like Figure 1 As shown, the method of this embodiment is described as follows.

[0054] In the risk assessment method for expected functional safety proposed in the present invention, risk assessment is divided into the following five processes:

[0055] (1) Collision severity assessment: The impact of failure modes (such as failure to correctly identify obstacles) on the safety of the automated driving system is assessed by calculating the collision severity (Severity, S). The calculation of the collision severity S is based on the type of collision object and the relative speed.

[0056] (2) Collision risk assessment: Determine the probability of occurrence (Occurrence, O) of potential failure modes, i.e., collision risk, through the output of the BEV+Occupancy Network perception algorithm;

[0057] (3) Algorithm performance evaluation: The false detection rate and missed detection rate of the current BEV+Occupancy Network perception algorithm are evaluated by calculating the F1 score as the detection capability (Detection, D) of the BEV+Occupancy Network perception algorithm. The F1 score reflects the precision and recall rate of the detection algorithm.

[0058] (4) Calculate the Risk Priority Number (RPN): Consider the severity and probability of collision risk and the detection capability of the BEV+Occupancy Network perception algorithm to calculate the risk priority number RPN.

[0059] (5) Develop improvement measures: When the risk priority number (RPN) value exceeds the preset threshold, the system triggers a takeover prompt to remind the driver to take over vehicle control, thereby reducing risks.

[0060] In this embodiment, the expected functional safety risk assessment method includes the following steps:

[0061] Step 1: Calculate the collision severity S, normalize the collision severity S and assign a corresponding score; the collision severity is used to reflect the potential impact of the failure mode on the vehicle and occupants. This embodiment is based on the SOTIF scenario architecture and selects the following factors for setting:

[0062] Step 1.1, determine the type and score of the collision object; collision objects include pedestrians, vehicles, fixed obstacles, etc. Different collision objects are set with different types of fixed scores, and are listed with reference to the severity in the expected functional safety. For example, the score of pedestrians is 10, the score of vehicles is 8, and the score of fixed obstacles is 6. Unrecognizable objects are inferred through the occupancy network, and the score is set to a basic value of 5;

[0063] Step 1.2, determine the collision speed, collision target category and corresponding weights, adjust the impact of different factors on severity according to the weights, and calculate the collision severity S;

[0064] The collision severity S can be expressed by the following formula:

[0065] S=ω1*ObjectType+ω2*RelativeSpeed ​​(1)

[0066] Among them, ω1 and ω2 represent the weights of the collision target category and the collision speed. The weight of the collision speed is set in stages according to the size of the collision speed (for example, the weight of 0-10km / h is 1, and the weight of 10-20km / h is 2, and they increase in sequence); ObjectType represents the collision target category. The collision target category and the corresponding weight are set with reference to the intended functional safety standard and expert guidance. RelativeSpeed ​​represents the collision speed, which refers to the relative speed between the vehicle and the collision object;

[0067] Step 1.3, normalize the calculated collision severity S, and assign corresponding scores to the collision severity S with values ​​in different ranges after the normalization, as shown in Table 1;

[0068] Table 1 Relationship between collision severity and corresponding scores

[0069]

[0070]

[0071] Step 2: Calculate the collision risk O, normalize the collision risk O and give a corresponding score; calculate the collision risk based on a single collision object of the simplified physical model, and then superimpose the collision risks of all collision objects in the current scene to obtain the current collision risk;

[0072] Step 2.1, obtain the relevant video frame data Data of the vehicle from multiple sensors such as the camera and radar of the vehicle perception system, and the calculation method is as follows:

[0073] Data = {I 1t ,···,I nt} (2)

[0074] Among them, I nt represents the tth video frame captured by the nth camera;

[0075] Step 2.2, the video frames captured by different cameras at the same time are fused to generate a BEV (Bird's Eye View) image, and the depth map of the video frame at the same time is generated and projected into the BEV image to obtain the depth value of the corresponding pixel; then it is sent to the BEV+Occupancy Network perception algorithm to obtain the relevant output information at the i-th second, including the distance between the collision target and the vehicle, the moving distance of the collision target's horizontal coordinate, and the moving distance of the collision target's vertical coordinate;

[0076] Based on the relevant output information at the i-th second obtained by the BEV+OccupancyNetwork perception algorithm, the time to collision (TTC) between the vehicle and the vehicle in front is calculated first. The smaller the TTC value, the higher the collision risk O. Therefore, TTC can be used as a warning indicator to remind the driver or the autonomous driving system to take measures (such as slowing down or changing lanes). The calculation method is as follows:

[0077]

[0078] Among them, TTC i is the predicted time of collision between the vehicle and the front vehicle at the i-th second, d i is the distance between the collision target and the vehicle at the i-th second, RelativeSpeed i is the collision velocity at the i-th second, and the collision velocity is calculated as follows:

[0079]

[0080] Where Δa is the time interval, Δx i , Δy iare the moving distance of the horizontal coordinate and the moving distance of the vertical coordinate of the collision target at the i-th second respectively;

[0081] Step 2.3, calculate the position change and movement direction of the collision target in the occupied network;

[0082] The moving direction of the collision target in the occupied network is determined by the following formula:

[0083] θ i =arctan2(Δx i ,Δy i ) (5)

[0084] Among them, θ i is the heading angle of the moving direction of the collision target at the i-th second;

[0085] Step 2.4: Calculate the collision avoidance deceleration rate, which represents the deceleration speed required to avoid a collision at the current distance between the vehicle and the collision target. The calculation formula is as follows:

[0086]

[0087] Among them, DRAC i represents the collision avoidance deceleration rate at the i-th second, v ei represents the speed of the vehicle at the i-th second, and e represents the vehicle;

[0088] Step 2.5: Calculate the steering threat coefficient, which represents the steering threat coefficient based on the heading angle difference at the current distance between the vehicle and the collision target. It is used in scenarios such as lane change and turning. The calculation formula is as follows:

[0089]

[0090] Among them, STN i represents the turning threat coefficient at the i-th second, θ ei ,θ oi Respectively represent the heading angles of the ego vehicle and the collision target at the i-th second;

[0091] Step 2.6, calculate the minimum safe distance, which represents the minimum safe distance of the vehicle at the current speed and acceleration. The calculation formula is as follows:

[0092]

[0093] Among them, MSD i represents the minimum safe distance in the ith second, a ei represents the acceleration of the vehicle at the i-th second;

[0094] Step 2.7, calculate the collision distance, which represents the difference between the current distance and the minimum safe distance. The calculation formula is as follows:

[0095] DTC i =d i -MSD i (9)

[0096] Among them, DTC i represents the collision distance at the i-th second;

[0097] Step 2.8: By considering the three indicators of collision time, collision avoidance deceleration rate and collision distance, combined with the occupation probability of each collision target in the spatial volume unit output by the occupation network, the collision risk of the spatial volume unit corresponding to the ego vehicle and the collision target is calculated. The calculation formula is as follows:

[0098]

[0099] Among them, O collision,b represents the collision risk of the b-th spatial volume unit, P occ,b represents the probability that the bth spatial volume unit is occupied, Represents the inverse of the collision time in the ith second. The larger it is, the higher the collision risk is. The larger it is, the greater the deceleration required to avoid a collision, and the higher the risk of collision. The larger it is, the more difficult it is to avoid a collision by steering, and the higher the risk of collision;

[0100] The three indicators of collision time, collision avoidance deceleration rate and collision distance are used as collision risk coefficient weights to calculate the collision risk of the corresponding unit. Then, the risk of all traffic participants and the vehicle in the scene is calculated, and then the total collision risk O of the current scene is obtained by summing them up. The calculation formula is as follows:

[0101]

[0102] Where N is the total number of spatial volume units;

[0103] Step 2.9, normalize the calculated total collision risk O, and assign corresponding scores to the total collision risk O with values ​​in different ranges after normalization, as shown in Table 2;

[0104] Table 2 Relationship between total collision risk and corresponding scores

[0105]

[0106] Step 3: Calculate and evaluate the detection capability D of the BEV+Occupancy Network perception algorithm and give a corresponding score. According to the characteristics of the BEV+Occupancy Network perception algorithm, it is crucial to reduce false detections and missed detections. Therefore, compared with the confidence level, the F1 score, which can represent the emphasis on missed detections and false detections, should be used to evaluate the detection capability D of the BEV+Occupancy Network perception algorithm.

[0107]

[0108]

[0109]

[0110] Among them, Precisio represents the accuracy of the algorithm, Recall represents the recall rate of the algorithm, TP true positive represents the number of positive samples correctly detected by the algorithm; FP false positive represents the number of positive samples incorrectly detected by the algorithm; FN false negative represents the number of positive samples that the algorithm failed to detect;

[0111] Afterwards, the corresponding scores are given to the detection capabilities D with values ​​in different ranges, as shown in Table 3;

[0112] Table 3 Relationship between detection ability and score

[0113]

[0114] Step 4: Based on the collision severity S, collision risk O, and the detection capability D of the BEV+OccupancyNetwork perception algorithm, calculate the total risk priority number, as follows:

[0115] R 总 =S*O*D (15)

[0116] Among them, R 总 Indicates the total risk priority number;

[0117] Step 5: Prioritize the total risk R 总 Normalize and give corresponding weights according to the corresponding levels. Finally, according to the normalized R 总 The size of is used to determine whether to send a takeover signal, as shown in Table 4;

[0118] Table 4 Visually perceived risks and takeover reminders

[0119]

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A method for assessing the risk of expected functional safety. Features: The following steps are involved: Step 1: Based on the SOTIF scenario architecture, the collision severity S is calculated by the collision object type and relative speed, and the collision severity S is normalized and given a corresponding score; Step 2: Calculate the collision risk O based on a single collision object of a simplified physical model and the BEV+Occupancy Network perception algorithm, normalize the collision risk O and give a corresponding score; Step 3: Calculate and evaluate the detection capability D of the BEV+Occupancy Network perception algorithm and give a corresponding score; Step 4: Calculate the total risk priority number based on the collision severity S, collision risk O, and the detection capability D of the BEV+Occupancy Network perception algorithm; Step 5: Normalize the total risk priority number and give corresponding weights according to the corresponding levels. Finally, determine whether to issue a takeover signal based on the size of the normalized total risk priority number.

2. According to claim 1, a method for evaluating the expected functional safety risk, It is characterized in that The step 1 further comprises the following steps: Step 1.1: Determine the type of collision object and set different types of fixed scores for different collision objects; Step 1.2: Determine the collision speed and collision target category and their corresponding weights, adjust the impact of different factors on severity according to the weights, and calculate the collision severity S; Step 1.3: The calculated collision severity S is normalized, and corresponding scores are given to the collision severity S having values ​​within different ranges after the normalization.

3. According to claim 2, a method for evaluating the expected functional safety risk, Features: The collision severity S can be expressed by the following formula: S=ω 1 *ObjectType+ω 2 *RelativeSpeed Among them, ω 1 ,ω 2 Represents the collision target category and the weight of the collision speed. The weight of the collision speed is set in segments according to the size of the collision speed. ObjectType represents the collision target category. The collision target category and the corresponding weight are set with reference to the expected functional safety standard and expert guidance. RelativeSpeed ​​represents the collision speed. The collision speed refers to the relative speed between the vehicle and the collision object.

4. According to claim 1, a method for evaluating the expected functional safety risk, Features: The step 2 further comprises the following steps: Step 2.1: Obtain relevant video frame data Data of the vehicle from multiple sensors of the vehicle perception system; Step 2.2: The video frames captured by different cameras at the same time are fused to generate a BEV image. At the same time, the depth map of the video frame at the same time is generated and projected into the BEV image to obtain the depth value of the corresponding pixel, which is sent to the BEV+Occupancy Network perception algorithm to obtain the relevant output information at the i-th second, including the distance between the collision target and the vehicle, the moving distance of the collision target's horizontal coordinate, and the moving distance of the collision target's vertical coordinate; Based on the relevant output information at the i-th second obtained by the BEV+OccupancyNetwork perception algorithm, the time when the vehicle collides with the vehicle in front is first calculated. The calculation formula is as follows; Among them, TTC i is the predicted time of collision between the vehicle and the front vehicle at the i-th second, d i is the distance between the collision target and the vehicle at the i-th second, RelativeSpeed i is the collision velocity at the i-th second; Step 2.3: Calculate the position change and movement direction of the collision target in the occupied network. The calculation formula is as follows; θ i =arctan2(Δx i ,Δy i ) Among them, θ i is the heading angle of the moving direction of the collision target at the i-th second, Δx i , Δy i are the moving distance of the horizontal coordinate and the moving distance of the vertical coordinate of the collision target at the i-th second respectively; Step 2.4: Calculate the collision avoidance deceleration rate, which represents the deceleration speed required to avoid a collision at the current distance between the vehicle and the collision target. The calculation formula is as follows: Among them, DRAC i represents the collision avoidance deceleration rate at the i-th second, v ei represents the speed of the vehicle at the i-th second, and e represents the vehicle; Step 2.5: Calculate the steering threat coefficient, which represents the steering threat coefficient based on the heading angle difference at the current distance between the vehicle and the collision target. It is used in scenarios such as lane change and turning. The calculation formula is as follows: Among them, STN i represents the turning threat coefficient at the i-th second, θ ei ,θ oi Respectively represent the heading angles of the ego vehicle and the collision target at the i-th second; Step 2.6: Calculate the minimum safe distance, which means the minimum safe distance under the current speed and acceleration. The calculation formula is as follows: Among them, MSD i represents the minimum safe distance in the i-th second, a ei represents the acceleration of the vehicle at the i-th second; Step 2.7: Calculate the collision distance, which is the difference between the current distance and the minimum safe distance. The calculation formula is as follows: DTC i =d i -MSD i Among them, DTC i represents the collision distance at the i-th second; Step 2.8: By considering the three indicators of collision time, collision avoidance deceleration rate and collision distance, combined with the occupation probability of each collision target in the spatial volume unit output by the occupation network, the collision risk of the corresponding unit is calculated, and the total collision risk O is calculated based on the collision risk of the corresponding unit; The total collision risk O is calculated as follows: Among them, O cllision，b represents the collision risk of the bth spatial volume unit, b represents the spatial volume unit, P occ，b represents the probability that the bth spatial volume unit is occupied, and N is the total number of spatial volume units; Step 2.9: The calculated total collision risk O is normalized, and corresponding scores are given to the total collision risk O having values ​​within different ranges after the normalization.

5. The method for assessing expected functional safety risk according to claim 1, Features: The step 3 uses the F1 score that can represent the emphasis on missed detection and false detection to evaluate the detection capability D of the BEV+Occupancy NetWork perception algorithm.

6. The method for evaluating the expected functional safety risk according to claim 1, It is characterized in that In step 4, the formula for calculating the total risk priority number is as follows: R 总 =S*O*D Among them, R 总 Indicates the total risk priority number.

Images