Safety performance quantitative evaluation method for autonomous driving system and perception layer, electronic device, and storage medium

Abstract

A safety performance quantitative evaluation method for an autonomous driving system, comprising: establishing a vehicle accident rate decomposition model to decompose hazardous events at a vehicle level into hazardous events caused by functional deficiencies at a component level, the component level comprising a perception layer, a positioning layer, a planning layer, and a control layer; acquiring statistical probabilities of hazardous events caused by functional deficiencies in the planning layer, the control layer, and the positioning layer; performing scenario modeling on the probability of the hazardous events caused by the functional deficiency in the perception layer; obtaining confirmed objectives of functions at least on the basis of a natural driving data set, the vehicle accident rate decomposition model, the statistical probabilities of the hazardous events, and the scenario modeling of the perception layer; and using the confirmed objectives of the functions as standards for safety performance quantitative evaluation.

Description

Safety performance quantification evaluation method for automatic driving system and perception layer, electronic device and storage medium TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of automatic driving, in particular to a safety performance quantification evaluation method for automatic driving system and perception layer, an electronic device and a storage medium. BACKGROUND

[0002] The core of the unmanned system can be summarized as three parts: perception, positioning, planning and control. Among them, perception refers to the ability of the unmanned system to collect information from the environment and extract relevant knowledge from it. Perception can include environmental perception and positioning. The national standard divides the driving automation system into six levels: L0 (emergency assistance), L1 (partial driving assistance), L2 (combined driving assistance), L3 (conditional automatic driving), L4 (high automatic driving) and L5 (complete automatic driving).

[0003] Currently, the safety design of the automatic driving system is either to design a total safety quantification target at the vehicle level, or to design some experience-based safety quantification targets at the system level, or some safety requirements based on experience or theoretical analysis, without specific quantitative indicators and poor theoretical basis. The safety design of the perception module of the automatic driving system is mostly based on qualitative design, without quantitative design reference. A small part of the quantitative automatic driving system safety design is based on the estimated value of engineering experience or expert experience, without theoretical basis. The current performance development of the automatic driving system cannot guarantee the usability of the performance without a safety target, and it is unclear to what extent the performance needs to be developed and improved.

[0004] SUMMARY

[0005] Embodiments of the present application provide a safety performance quantification evaluation method for automatic driving system and perception layer, an electronic device and a storage medium, which at least solve one of the above technical problems.

[0006] In a first aspect, the embodiments of the present application provide a safety performance quantitative evaluation method for an automatic driving system, comprising: establishing a whole vehicle accident rate decomposition model, decomposing hazard events at a whole vehicle layer into hazard events caused by functional deficiencies at component layers, wherein the component layers include a perception layer, a positioning layer, a planning layer, and a control layer; obtaining statistical probabilities of hazard events caused by functional deficiencies at the planning layer, the control layer, and the positioning layer; performing scenario modeling on probabilities of hazard events caused by functional deficiencies at the perception layer; obtaining confirmation targets of each function based at least on a natural driving dataset, the whole vehicle accident rate decomposition model, the statistical probabilities of hazard events, and the scenario modeling of the perception layer; and taking the confirmation targets of each function as a standard for safety performance quantitative evaluation.

[0007] In a second aspect, the embodiments of the present application provide a safety performance quantitative evaluation method for a perception layer of an automatic driving system, comprising: establishing a perception layer accident rate decomposition model, decomposing hazard events at the perception layer into hazard events caused by functional deficiencies at the perception layer; performing scenario modeling on probabilities of hazard events caused by functional deficiencies at the perception layer; and obtaining confirmation targets of each function of the perception layer based at least on a natural driving dataset, the perception layer accident rate decomposition model, and the scenario modeling of the perception layer.

[0008] In a third aspect, the embodiments of the present application provide an electronic device, comprising: at least one processor, and a memory connected to the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any one of the safety performance quantitative evaluation methods for an automatic driving system and a perception layer described above.

[0009] In a fourth aspect, the embodiments of the present application provide a storage medium, wherein the storage medium stores one or more programs including execution instructions, and the execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to perform any one of the safety performance quantitative evaluation methods for an automatic driving system and a perception layer described above.

[0010] In a fifth aspect, the embodiments of the present application further provide a computer program product, comprising a computer program stored on a storage medium, and the computer program comprises program instructions, when the program instructions are executed by a computer, the computer executes any one of the safety performance quantitative evaluation methods for an automatic driving system and a perception layer described above.

[0011] This application provides a theoretically sound and complete method for quantitatively assessing the safety of autonomous driving systems. Using the method described in this application, quantitative safety design indicators for autonomous driving systems can be obtained, providing clear quantitative references for the safety design, development, and testing phases of autonomous driving systems. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 is a flowchart of a safety energy assessment method for an autonomous driving system provided in an embodiment of this application;

[0014] Figure 2 is a flowchart of a safety energy assessment method for the perception layer of an autonomous driving system provided in an embodiment of this application;

[0015] Figure 3 shows a vehicle failure decomposition scheme provided in an embodiment of this application;

[0016] Figure 4 shows a p provided in an embodiment of this application. i The probability distribution diagram;

[0017] Figure 5 is a probability distribution diagram of ps1 provided in an embodiment of this application;

[0018] Figure 6 shows a p provided in an embodiment of this application. d，persist ×pdTTC, probability distribution plot of persist;

[0019] Figure 7 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] The application is a safety performance quantitative evaluation method for each algorithm module of an automatic driving system perception module. Mathematical modeling, theoretical analysis, and natural driving data analysis methods are used to design safety performance index quantitative schemes for each algorithm module such as perception, planning, control, and positioning, and to obtain safety quantitative targets for each algorithm module, providing a reference for the safety design and evaluation of the automatic driving system perception module. The mathematical modeling mainly includes kinematics model analysis and dangerous scene modeling. The theoretical analysis mainly includes theoretical analysis of the harm caused by each algorithm module and the rationality of mathematical modeling. The natural driving data analysis mainly obtains driver behavior by statistical natural driving data.

[0022] The existing safety design of the automatic driving system perception module cannot give reasonable theoretical derivation, resulting in no quantitative safety targets for each algorithm module. The embodiments of the application obtain safety performance target quantitative schemes for each algorithm module based on reasonable mathematical modeling and natural driving data, which are used for the safety design and evaluation of the automatic driving system perception module.

[0023] Please refer to FIG. 1, which shows a flowchart of a safety performance quantitative evaluation method for an automatic driving system according to an embodiment of the application. The above method can be applied to the field of safety design and evaluation of an automatic driving system or the field of automatic driving system testing. As part of automatic driving safety, it can be applied to the safety design and evaluation of automatic driving system perception modules of various levels or automatic driving system testing.

[0024] As shown in FIG. 1, in step 101, an entire vehicle accident rate decomposition model is established to decompose the hazard events at the entire vehicle layer into those caused by functional deficiencies at the component layer;

[0025] In step 102, the statistical probability of hazard events caused by functional deficiencies at the planning layer, the control layer, and the positioning layer is obtained;

[0026] In step 103, the probability of hazard events caused by functional deficiencies at the perception layer is modeled by scene;

[0027] In step 104, the confirmation targets of each function are obtained based on at least a natural driving data set, the entire vehicle accident rate decomposition model, the statistical probability of hazard events, and the scene modeling of the perception layer;

[0028] In step 105, the confirmation targets of each function are used as the standard for safety performance quantitative evaluation.

[0029] In the embodiments of the present application, for step 101, by establishing a vehicle accident rate decomposition model, the hazard events of the vehicle layer are further refined into the functional deficiencies of the component layer, and the functional deficiencies of the component layer ultimately lead to the accident of the vehicle layer, which can be quantified as the probability of different functions failing ultimately leading to the probability of accidents. And the functional failure leading to the accident can be specifically to the different event probabilities, and then the probability of functional failure can be obtained by back calculation from the known time probability, so as to obtain the quantitative index. The component layer includes a perception layer, a positioning layer, a planning layer and a control layer. In other embodiments, the positioning layer can be included in the perception layer.

[0030] For step 102, the planning layer, the control layer and the positioning layer, the statistical probability of the hazard events of the functional deficiencies of these component layers can be obtained by other ways. The specific ways include but are not limited to simulation test, real vehicle test, data set extraction and / or experience value setting, etc. Among them, the simulation test can use simulation software or HIL (Hardware-in-the-Loop) simulator to establish related scenes, and in this scene, through the form of fault injection or scene injection, the performance of system functional deficiency is simulated, and the probability of occurrence of hazard events is counted. The real vehicle test is similar to the simulation test, only in the real vehicle test environment, the fault scene is simulated for test statistics. The data set extraction can use the data set to establish the driver model and the scene model, and extract the probability of occurrence of hazard events. The experience value setting is to set the probability of occurrence of hazard events of the related dangerous scene according to experience.

[0031] For step 103, the probability of hazard events caused by functional deficiencies of the perception layer can be obtained by scene modeling. In some specific examples, dangerous scenes in different scenes can be modeled, for example, different scenes can include cut in scene, following scene, cut out scene, etc., which is not limited herein. Among them, cut in means that the host vehicle is following a vehicle, and suddenly a vehicle inserts, the system needs to reselect the following object. Cut out means that the object followed by the host vehicle suddenly deviates from the lane, and the system needs to reselect the following object. Since it needs to be substituted into the above vehicle accident rate decomposition model, therefore, the modeling is mainly to model the dangerous scenes in these scenes, such as collision with the front obstacle or being rear-ended by the rear obstacle. By modeling the probability of hazard events caused by functional deficiencies, the vehicle accident rate decomposition model can be further decomposed into specific hazard events.

[0032] Afterwards, for step 104, the confirmation target of each function is obtained based on at least the natural driving data set, the vehicle accident rate decomposition model, the hazard event statistical probability and the scene modeling of the perception layer. By obtaining some specific parameters required by the vehicle accident rate decomposition model and the natural driving data set, and then substituting the related specific parameters and specific scene modeling into the vehicle accident rate decomposition model, the vehicle accident rate decomposition model can give the probability of specific functional deficiency. The specific parameters can vary according to actual conditions, for example, each company sets according to specific needs, which is not limited herein.

[0033] Finally, for step 105, when the confirmation target of each function is obtained, the confirmation target of each function can be used as a safety performance quantitative evaluation standard, which is used as an evaluation index in each scene. For example, it can be used as an evaluation index in the design and development stage, or as an evaluation index in the test stage, which is not limited herein.

[0034] The embodiment of the present application provides a complete automatic driving system safety performance quantitative evaluation method with theoretical basis. By using the method of the embodiment of the present application, the quantitative safety design index of the automatic driving system can be obtained, which provides a clear quantitative reference for the safety design and development stage and the test stage of the automatic driving system.

[0035] In some optional embodiments, the hazard event of the whole vehicle layer is decomposed into a functional deficiency of the component layer, including: in the case where the functional deficiency of the component layer occurs, a hazard behavior exposure due to the functional deficiency, and the hazard behavior is uncontrollable, and an accident with a preset severity is caused, then the functional deficiency of the component layer is considered to cause the hazard event of the whole vehicle layer. Further, the functional deficiency can include a deficiency of the expected function specification definition of the whole vehicle layer, or a deficiency of the specification definition or performance limitation of the electrical / electronic elements in the system. The hazard behavior exposure refers to the occurrence of a hazard behavior. The hazard behavior is uncontrollable, which means that the hazard behavior cannot be controlled (for example, the specific operation of the driver in response to the hazard behavior cannot be predicted). The preset severity refers to the accident that can cause property damage, personal injury, death, etc., and can also be other severities, which are not described here. Among them, the functional deficiency of the component layer, the hazard behavior exposure due to the functional deficiency, the uncontrollability of the hazard behavior after the hazard behavior exposure, the accident caused by the uncontrollable hazard behavior, and the accident with the preset severity can be characterized by corresponding quantifiable parameters, so that the hazard event of the whole vehicle layer can be decomposed to some specific scenarios with quantifiable characterization. Then, the probability of the occurrence of the corresponding scenario can be obtained through relevant statistical data, so that the final probability of the functional deficiency of each function under different actual requirements can be obtained by parameter assignment and setting of some parameters according to actual requirements. Through the above decomposition, the whole vehicle accident rate decomposition model can be visualized, and finally the quantifiable index can be obtained.

[0036] In some optional embodiments, obtaining the statistical probability of the hazard event caused by the functional deficiency of the planning layer, the control layer and the positioning layer includes: obtaining the statistical probability of the hazard event caused by the functional deficiency of the planning layer, the control layer and the positioning layer by at least one of simulation testing, real vehicle testing, data set extraction and / or experience value setting. Thus, the statistical probability of the hazard event caused by the functional deficiency of the planning layer, the control layer and the positioning layer can be obtained by the above-mentioned manner. For example, a simulation software or a HIL simulator can be used to establish a related scenario, and in this scenario, the performance of the functional deficiency of the system can be simulated by means of fault injection or scene injection, for example, the performance of the functional deficiency of the planning layer is simulated, and the statistical probability of the hazard event caused by the functional deficiency of the planning layer is counted. The control layer functional testing can be performed by establishing a real vehicle testing dangerous scene, and the statistical probability of the hazard event caused by the functional deficiency of the control layer is obtained. The driver model can be established by using the data set, and the scene model of the functional deficiency of the positioning layer is established, and the statistical probability of the hazard event caused by the functional deficiency of the positioning layer is extracted. Of course, in other examples, the statistical probability of the hazard event caused by the functional deficiency of the planning layer, the control layer and the positioning layer can also be obtained by the above-mentioned manner.

[0037] In some optional embodiments, the perception layer function deficiency includes missed detection and false detection, and the scenario modeling of the probability of the hazard event caused by the perception layer function deficiency includes: scenario modeling of the probability of collision with the front obstacle caused by the missed detection; and scenario modeling of the probability of rear-end collision after false braking caused by the false detection. The scenario modeling is mainly mathematical modeling, and the mathematical modeling is implemented through a program. For example, if the distance between the two vehicles is too close, it can be mathematically expressed as TTC < 3s of the two vehicles, and such scenarios in the data set can be extracted by writing a program. Wherein, TTC is the time to collision between the ego vehicle and the front vehicle, defined as the distance between the ego vehicle and the obstacle divided by the relative speed. By using the method of the embodiments of the present application, the performance indicators of each algorithm of the perception module of the autonomous driving system can be obtained, which can be used to evaluate whether the current safety performance of the perception module of the autonomous driving system meets the requirements.

[0038] In further optional embodiments, the relationship between the missed detection rate and the false detection rate is set, and the confirmation target of each function obtained based on at least the natural driving data set, the whole vehicle accident rate decomposition model, the statistical probability of the hazard event, and the scenario modeling of the perception layer includes: extracting parameters in the scenario modeling of the perception layer using the natural driving data set; obtaining the missed detection rate and the false detection rate of the perception layer based on the extracted parameters, the statistical probability of the hazard event, the set relationship between the missed detection rate and the false detection rate, the whole vehicle accident rate decomposition model, and the assignment of related parameters in the whole vehicle accident rate decomposition model; and setting that the missed detection rate and the false detection rate both meet a preset distribution, and obtaining the confirmation target of the missed detection rate or the false detection rate based on the missed detection rate or the false detection rate, respectively, in combination with the preset distribution. By setting the relationship between the missed detection rate and the false detection rate, for example, as a multiple relationship, the specific missed detection rate and false detection rate can be obtained by substituting the multiple relationship into the above whole vehicle accident rate decomposition model. Further, with the specific values of the missed detection rate and the false detection rate, the corresponding confirmation target can be obtained in combination with the preset distribution. The preset distribution is, for example, a Poisson distribution, or other distributions that can represent vehicle collision events, which are not limited in the present application.

[0039] Further optionally, the relationship between the missed detection rate and the false detection rate includes setting the missed detection rate to be lower than the false detection rate. Since the hazard probability caused by false detection is lower than that caused by missed detection, the system missed detection rate can be set to be lower than the false detection rate. In other embodiments, the relationship between the missed detection rate and the false detection rate includes setting the missed detection rate to be higher than the false detection rate, or setting the missed detection rate to be equal to the false detection rate, and the specific relationship can be set according to specific requirements.

[0040] In some optional embodiments, the function deficiency of the perception layer includes a detection accuracy error, and the method includes: obtaining the detection accuracy error by simulation statistics through establishing a driver model; and taking the detection accuracy error as a detection accuracy error constraint of the autonomous driving system. Optionally, the detection accuracy error can be taken as a constraint of the autonomous driving system, for example, to constrain the accuracy of each sensor used by the autonomous driving system, and the present application does not limit this. The detection accuracy error can include distance, speed, acceleration, yaw angle, missed detection, false detection, etc. The detection accuracy error is only taken as a constraint because the errors of the planning, control and other component layers have relatively small impact on the perception detection error, and the errors of the planning, control and other component layers can be obtained by other means and are not considered in the present embodiment. Since the impact of the detection accuracy error is much smaller than that of the missed detection and false detection, the impact of the detection accuracy error can be ignored or the system is assumed to meet the detection accuracy error requirement in the vehicle accident rate decomposition model.

[0041] In some optional embodiments, the above method further includes: taking the confirmation target of each function as a design and development target in the initial design and development of the autonomous driving system; and / or taking the confirmation target of each function as a threshold for judging whether a test result is qualified when the vehicle is tested. The confirmation target is a defined test target of the system, and the system design needs to meet acceptance criteria. The confirmation target provides evidence for verifying whether the system meets the acceptance criteria. The specific logic is: first, obtain the acceptance criteria as the design and development target of the system, and then obtain the confirmation target of each function of the system for testing and verifying that the design of the system meets the acceptance criteria. Thus, the confirmation target of each function obtained by the method of the present embodiment can be used as a design and development target or a test target.

[0042] In some optional embodiments, the formula of the vehicle accident rate decomposition model λ is as follows:

[0043] wherein, p pl，j，k represents the probability of function deficiency of each component layer, p j，k represents the probability of exposure of the harmful behavior in the scene due to the function deficiency, p c|ps represents the probability of uncontrollability of the harmful behavior assuming that the harmful behavior is exposed in the scene, p s|c represents the probability of accident severity assuming that the harmful behavior is uncontrollable, and p irepresents the probability under each speed interval, j represents each component layer, k represents the functional deficiency of each component layer, l represents the vehicle accident rate, sence represents the perception system (perception layer), plan represents the planning system (planning layer), position represents the positioning system (positioning layer), and act represents the control execution system (control layer). Thus, the vehicle accident rate decomposition model can be used to quantitatively represent the failure scenarios of the vehicle.

[0044] Referring to FIG. 2, a flowchart of a safety performance quantitative evaluation method for a perception layer of an automatic driving system is shown. The method can be applied to the safety design and evaluation of the perception module of the automatic driving system or the testing of the perception module of the automatic driving system. As part of the safety of the automatic driving, the method can be applied to the safety design and evaluation of the perception module of the automatic driving system or the testing of the perception module of the automatic driving system.

[0045] As shown in FIG. 2, in step 201, a perception layer accident rate decomposition model is established, and the hazard events of the perception layer are decomposed into events caused by the functional deficiency of the perception layer.

[0046] In step 202, the probability of the hazard events caused by the functional deficiency of the perception layer is modeled.

[0047] In step 203, the confirmation target of each function of the perception layer is obtained based on at least the natural driving data set, the perception layer accident rate decomposition model, and the modeling of the hazard events of the perception layer.

[0048] In this embodiment, a perception layer accident rate decomposition model is established. The perception layer accident rate decomposition model is obtained by removing the influence of the planning layer and the control layer from the vehicle layer accident rate decomposition model of the foregoing embodiment. The specific scheme is similar to that of the foregoing embodiment, and thus is not described herein.

[0049] The embodiments of the present application provide a complete safety performance quantitative evaluation method for the perception module of the automatic driving system with a theoretical basis. The method can be used to obtain the quantitative safety design index of the perception module of the automatic driving system, and provide a clear quantitative reference for the safety design and development of the perception module of the automatic driving system.

[0050] In some optional embodiments, decomposing the hazard event of the perception layer into a function deficiency of the perception layer includes: in the case that the function deficiency of the perception layer occurs, a hazard behavior exposure due to the function deficiency, the hazard behavior being uncontrollable, and an accident with a preset severity, the function deficiency of the perception layer is considered to cause the hazard event of the perception layer. Further, the function deficiency can include a deficiency of an expected function specification defined at a vehicle level or a deficiency or performance limitation of an electrical / electronic element in the system. The hazard behavior exposure means that the hazard behavior occurs. The hazard behavior being uncontrollable means that the hazard behavior cannot be controlled. The preset severity means that the accident can cause property damage, personal injury, death, and the like, and can also be other severities, which are not described here. The hazard event of the perception layer can include a hazard event due to detection accuracy error, a hazard event due to missed detection, a hazard event due to false detection, and a hazard event due to positioning error.

[0051] In some optional embodiments, the function deficiency of the perception layer includes missed detection and false detection, and the scenario modeling of the probability of the hazard event caused by the function deficiency of the perception layer includes: scenario modeling of the probability of a collision with a front obstacle caused by the missed detection; and scenario modeling of the probability of a rear-end collision after false braking caused by the false detection. The scenario modeling is mainly mathematical modeling, and then the mathematical modeling can be implemented through a program. For example, the distance between two vehicles is too close, which can be mathematically expressed as TTC<5s of the two vehicles, and the data set can be extracted through a program. The TTC is the time to collision between the ego vehicle and the front vehicle, defined as the distance between the ego vehicle and the obstacle divided by the relative speed. Since the consequences of detection accuracy error and positioning error are much smaller than those of missed detection and false detection, the detection accuracy error and the positioning error can be ignored, and only the consequences of missed detection and false detection are considered. By scenario modeling of the missed detection and the false detection, the perception layer accident rate model is further refined, so that quantitative indicators of the missed detection rate and the false detection rate can be obtained.

[0052] In a further optional embodiment, the relationship between the missed detection rate and the false detection rate is set, and the confirmation target of each function of the perception layer based on at least the natural driving data set, the perception layer accident rate decomposition model and the scene modeling of the perception layer comprises: extracting the parameters in the scene modeling of the perception layer based on the natural driving data set; obtaining the missed detection rate and the false detection rate of the perception layer based on the extracted parameters, the set relationship between the missed detection rate and the false detection rate, the whole vehicle accident rate decomposition model and the assignment of the related parameters in the perception layer accident rate decomposition model; setting that the missed detection rate and the false detection rate meet a preset distribution, and obtaining the confirmation target of the missed detection rate or the false detection rate based on the missed detection rate or the false detection rate, respectively, in combination with the preset distribution. The confirmation target is to define the test target of the system, to provide evidence that the system design meets the acceptance criteria. The specific logic is: first, obtain the acceptance criteria as the design and development target of the system, and then obtain the confirmation target for testing and verifying that the system design meets the acceptance criteria. By substituting the related parameters set according to the specific situation into the above model, extracting the parameters in the scene modeling of the perception layer based on the natural driving data set, and setting the relationship between the missed detection rate and the false detection rate, for example, setting a preset multiple relationship, the missed detection rate and the false detection rate can be obtained by solving the whole vehicle accident rate decomposition model, and then the specific confirmation target of the missed detection rate and the false detection rate can be obtained in combination with the corresponding distribution. Finally, the confirmation target can be used as the target of the design and development stage or the target of the test stage to guide the specific quantitative parameters of design and development and testing.

[0053] Further optionally, the relationship between the missed detection rate and the false detection rate comprises setting the missed detection rate to be lower than the false detection rate. Since the harm probability caused by false detection is lower than that caused by missed detection, the system missed detection rate can be required to be lower than the false detection rate. Thus, by setting the specific relationship between the missed detection rate and the false detection rate, the relationship between the missed detection rate and the false detection rate is quantified, which is substituted into the perception layer accident rate decomposition model to obtain the specific missed detection rate and false detection rate, thereby providing specific data for subsequent calculation of the confirmation target of the missed detection rate and the false detection rate.

[0054] In some optional embodiments, the function deficiency of the perception layer comprises detection accuracy error, and the method further comprises: obtaining the detection accuracy error by establishing a driver model for simulation and statistics; and using the detection accuracy error as a constraint on the detection accuracy error of the automatic driving system. Thus, the detection accuracy error of the perception layer can be obtained by simulation, statistics and the like, and the detection accuracy error of the automatic driving system can be constrained by using the detection accuracy error.

[0055] In some optional embodiments, the whole vehicle accident rate decomposition model λ sence is expressed as follows:

[0056] wherein, pFN，j，1 p represents the perception layer miss rate, p FP，j，2 ps represents the perception layer false alarm rate, ps j，1 p represents the probability that a hazardous behavior is exposed in a scenario due to the miss rate, p FP，j，2 p represents the probability that a hazardous behavior is exposed in a scenario due to the false alarm rate, p c|ps p represents the probability that a hazardous behavior is uncontrollable given that the hazardous behavior is exposed in a scenario, p s|c p represents the probability of accident severity given that a hazardous behavior is uncontrollable, p i p represents the probability in each speed interval, j represents the perception layer, k represents the functional deficiency of each component layer, λ sence sence represents the perception system (perception layer). Thus, the vehicle failure caused by the functional deficiency (false alarm and miss) of the perception layer can be quantitatively characterized.

[0057] Please refer to FIG. 3, which shows a vehicle failure decomposition scheme provided by an embodiment of the present application.

[0058] As shown in FIG. 3, the embodiment of the present application adopts a layer-by-layer decomposition method to decompose the vehicle failure problem to each algorithm module. The vehicle failure is caused by a hazardous behavior and a related scenario. The hazardous behavior can include unexpected braking, unexpected steering, braking too late, etc. The core idea is that the safety of the perception module of the automatic driving system cannot be lower than that of a human. First, the cause of the vehicle failure is analyzed. The vehicle failure is caused by a hazardous behavior and a related scenario. For the related scenario, scenario modeling can be performed to obtain the region of interest of different scenarios. The scenarios that need to be modeled include cut in scenarios, following scenarios, cut out scenarios, etc. The corresponding region of interest is a dangerous scenario. Then, the hazardous behavior is analyzed, and specific hazardous behaviors are listed. After that, the cause of the hazardous behavior is analyzed, and the performance limitations that cause the hazardous behavior are identified, including the performance limitations of perception, planning, control, etc. The existence of uncertainty in perception is the main influencing factor, which mainly includes the miss performance limitation and the false alarm performance limitation of perception. The existence of uncertainty in perception is mainly due to the uncertainty of various detections of the perception layer, which can include dynamic obstacle detection, static obstacle detection, VRU (Vulnerable Road User) detection, etc.

[0059] All performance limitations are caused by different algorithm modules. The performance limitations are analyzed and modeled, and the natural driving dataset is calculated to obtain the quantitative indicators of the performance limitations of different algorithm modules as the safety performance quantitative targets of the corresponding algorithm modules.

[0060] In view of the existence of uncertainty in the performance limitation related to perception, it is necessary to decompose the accident rate of the whole vehicle system layer to each key subsystem. The conditional automatic driving system can be split into component layers such as perception, planning and control. An accident rate decomposition model of the whole vehicle layer is established as follows:

[0061] In the formula, p pl，j，k represents the probability of functional deficiency of each component layer; p j，k s represents the probability of exposure of hazardous behavior in the scene due to functional deficiency; p c|ps represents the probability that the hazardous behavior is uncontrollable assuming that the hazardous behavior is exposed in the scene; p s|c represents the probability of accident severity caused by the uncontrollable hazardous behavior; p i represents the probability in each speed interval; j represents each component layer, including the perception layer, the positioning layer, the planning layer and the control layer; and k represents the functional deficiency of each component layer.

[0062] The meaning of the above-mentioned accident rate decomposition model of the whole vehicle layer is that the hazardous event of the whole vehicle layer is caused by the hazardous behavior of each component layer, and the hazardous behavior of the component layer is caused by the functional deficiency of the component layer corresponding to the component layer. In the case of functional deficiency of the component layer, the hazardous behavior is exposed in the scene caused by the functional deficiency, and the hazardous behavior is uncontrollable. The accident caused by the hazardous behavior has a certain severity, and it is considered that the functional deficiency causes the hazardous event of the whole vehicle layer.

[0063] The model is expanded as follows:

[0064] In the formula, p poserr,j,k is the probability that the position error does not meet the requirements.

[0065] λ includes the whole vehicle hazard probability caused by the positioning component layer, the planning component layer and the control component layer, and the hazardous event probability caused by the functional deficiency of the positioning layer, the planning layer and the control layer. The hazardous event probability caused by the functional deficiency of the positioning layer, the planning layer and the control layer can be obtained by simulation test or real vehicle test, which is not limited in the present application. It should be noted that although various specific parameters are introduced in the subsequent examples, the specific parameters can also be changed according to the actual situation, which is not limited in the present application.

[0066] The statistical probability of the hazardous event caused by the functional deficiency of the planning layer, the control layer and the positioning layer can be obtained by a large number of simulation tests, which is λ HPnC , assuming that the statistical result is 4e-5 / h, which can also be set to other values, which is not limited in the present application.

[0067] The detection accuracy requirements of the position detection error, speed detection error and the like of the perception layer can be obtained by establishing a driver model, such as an RSS (Responsibility Sensitive Safety) model, an FSM (Finite State Machine) model and the like. The detection accuracy error of the perception layer obtained by the driver model can be used as a system detection accuracy constraint. Compared with the detection accuracy error of distance, speed and the like, the consequences caused by the missed detection and false detection of perception are more serious, that is, p FN,j，1 and p FP,j,2 under p c|ps and p s|c are much greater than the detection accuracy p c|ps and p s|c of the perception, so in the embodiments of the present application, the probability of the hazard event caused by the dangerous behavior due to the inaccuracy of the system detection accuracy can be ignored. Specifically, the accuracy detection error can include distance, speed, acceleration, yaw angle, missed detection, false detection and the like. Among them, the missed detection and false detection can be calculated by the subsequent related model of the present solution, and the other detection errors can be obtained by the above driver model. On the other hand, since the detection accuracy error other than the missed detection and false detection has a small impact, the impact of these errors can be ignored in the embodiments. Further, the accuracy detection error can be used as a system constraint, and these accuracy detection errors can be determined in advance by the above model. The reason why only the detection accuracy error is used as a constraint is that the error of the planning, control and the like layer has a relatively small impact on the perception detection error, or it is not within the scope of the embodiments.

[0068] Therefore, the model can be simplified as:

[0069] At the same time, the region of interest when the perception occurs with uncertainty needs to be considered, and the dangerous scene is modeled. The hazard event caused by the missed detection of the perception is the collision with the obstacle in front of the vehicle, and the probability ps j，1 of the occurrence of the related traffic scene can be further expressed as: ps j，1 ={p|0<T<5s}.

[0070] The above value 5s can also be set to other values, for example, 4s, which is not limited in the present application.

[0071] The hazard event caused by the false detection rate of the perception is that the vehicle brakes incorrectly, and the rear vehicle rear-ends the front vehicle, and the probability ps ，j，2 of the occurrence of the related traffic scene can be further expressed as: ps j，2 =pdj,persist×pdTTC,j,persist.

[0072] By using the method of the embodiments of the present application, some specific dangerous scenarios and driver models in dangerous scenarios can be obtained, which can provide input and reference for the development of the perception module of the automatic driving system. In the formula, pdj,persist represents the probability of the host vehicle continuously decelerating in the speed range i, assuming that the duration is greater than 1 s. The duration can also be set to other numerical values, which are not limited in the present application. pdTTC,j,persist represents that in the case of continuous deceleration of the host vehicle, the rear vehicle is close enough to the host vehicle, i.e., the rear vehicle can cause danger to the host vehicle, taking TTC<5s as an example to represent the scenario in which the rear vehicle is close enough to the host vehicle. Wherein, TTC is the time to collision between the ego vehicle and the front vehicle, defined as the distance between the ego vehicle and the obstacle divided by the relative speed. It should be noted that TTC can also be set to other numerical values, which are not limited in the present application.

[0073] Therefore, the model is further optimized as follows:

[0074] Considering L3 and above automatic driving, people do not participate in monitoring, then p c|ps = 100%. Considering that the severity of the accident is the severity of personnel injury or death, which is unacceptable, assuming that the probability of accident severity is a constant value under all dangerous behaviors, based on the data of Chinese automobile collision accidents, assuming that the proportion of fatal accidents and injury accidents is 45%, p s|c = 0.45. Each manufacturer can define it according to different data sets, which are not limited in the present application.

[0075] Therefore, the model can be further expanded as follows:

[0076] At the same time, by using the natural driving data set, the parameters in the scenario modeling are extracted, and the probability distribution of p i is obtained, as shown in FIG. 4, the probability distribution of ps1 is shown in FIG. 5, and the probability distribution of p d,persist × pdTTC,persist is shown in FIG. 6. The natural driving data set can use various data sets in the industry. For example, the aerial survey data set collected by the enterprise itself, the CitySim data set, the HighD data set, the VisDrone data set, etc., which are not limited in the present application.

[0077] According to the calculation of the data in FIG. 4, FIG. 5 and FIG. 6, it can be obtained that (∑ i p i × ps1) = 0.00253 and (∑ i p i × pdj,persist × pdTTC,j,persist) = 0.00014.

[0078] Further, in the relevant dangerous scene, the harm probability caused by the false detection is lower than the harm probability caused by the missed detection, and the system missed detection rate should be ensured to be lower than the false detection rate. In the embodiment, 2p FN，1 =p FP，2 is obtained, p FN，1 = 0.047 / h, and p FP,2 = 0.094 / h. The relationship between the missed detection rate and the false detection rate can also be set to other values, for example, 3p FN，1 =p FP，2 , which is not limited in the application. Assuming that the event obeys the Poisson distribution, the confidence is set to 80%, and a = 0.8. It should be noted that, in addition to the Poisson distribution, other statistical probability methods can also be used to obtain, for example, other distributions that can represent vehicle collisions, which are not limited in the application. For the confidence, other values can also be taken, for example, 85%, which is not described here. According to the above setting, the confirmation target τ FN，1 of the missed detection rate is -ln(1-a) / p FN，1 = 1.6 / 0.047 = 34.1h, which means that if there is no missed detection in the continuous 34.1h of highway driving time, there is 80% confidence that the probability of missed detection of the autonomous driving system or the perception module is less than 0.047 / h. The confirmation target τ Ep，2 of the false detection rate is -ln(1-a) / p FP，2 = 1.6 / 0.094 = 17h. This means that if there is no false detection in the continuous 17h of highway driving time, there is 80% confidence that the probability of false detection of the autonomous driving system or the perception module is less than 0.094 / h.

[0079] In some other embodiments, the application also provides a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions are used to execute the safety performance quantitative evaluation method for the autonomous driving system in any of the above method embodiments.

[0080] As an implementation manner, the non-volatile computer storage medium of the application stores computer executable instructions, and the computer executable instructions are set to:

[0081] establishing a vehicle accident rate decomposition model to decompose the hazard events of the vehicle layer into the hazard events caused by the functional deficiency of the component layer, wherein the component layers include a perception layer, a positioning layer, a planning layer, and a control layer;

[0082] obtaining the statistical probability of the hazard events caused by the functional deficiency of the planning layer, the control layer, and the positioning layer;

[0083] modeling a probability of the hazard event caused by the function deficiency of the perception layer;

[0084] obtaining a confirmation target of each function based on at least the natural driving data set, the vehicle accident rate decomposition model, the hazard event statistical probability, and the scenario modeling of the perception layer;

[0085] using the confirmation target of each function as a standard of the safety performance quantitative evaluation.

[0086] As another implementation, the non-volatile computer storage medium of the present application stores computer executable instructions, which are configured to:

[0087] establishing a perception layer accident rate decomposition model to decompose hazard events of the perception layer into function deficiencies of the perception layer;

[0088] modeling a probability of the hazard event caused by the function deficiency of the perception layer;

[0089] obtaining a confirmation target of each function of the perception layer based on at least the natural driving data set, the perception layer accident rate decomposition model, and the scenario modeling of the perception layer.

[0090] The non-volatile computer readable storage medium can include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by the function; the data storage area can store data created according to the use of the safety performance quantitative evaluation device for an automatic driving system, etc. In addition, the non-volatile computer readable storage medium can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state memory device. In some embodiments, the non-volatile computer readable storage medium can optionally include a memory remotely arranged relative to the processor, and these remote memories can be connected to the safety performance quantitative evaluation device for an automatic driving system through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0091] The embodiments of the present application also provide a computer program product, which includes a computer program stored on a non-volatile computer readable storage medium, and the computer program includes program instructions, which, when executed by a computer, cause the computer to execute any one of the above safety performance quantitative evaluation methods for an automatic driving system.

[0092] FIG. 7 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. As shown in FIG. 7, the device includes one or more processors 710 and a memory 720, and FIG. 7 takes one processor 710 as an example. The device for the safety performance quantification evaluation method of an automatic driving system can also include an input device 730 and an output device 740. The processor 710, the memory 720, the input device 730, and the output device 740 can be connected through a bus or other means, and FIG. 7 takes the connection through a bus as an example. The memory 720 is the non-volatile computer readable storage medium described above. The processor 710 performs various functional applications and data processing of the server by running the non-volatile software programs, instructions, and modules stored in the memory 720, that is, implements the safety performance quantification evaluation method for an automatic driving system of the method embodiments described above. The input device 730 can receive input digital or character information, and generate key signal inputs related to user settings and functional control of the safety performance quantification evaluation device for an automatic driving system. The output device 740 can include a display device such as a display screen.

[0093] The above product can perform the method provided by the embodiments of the present application, and has the corresponding functional modules and beneficial effects of performing the method. Technical details not described in detail in the embodiments can be referred to the method provided by the embodiments of the present application.

[0094] As an implementation form, the above electronic device is applied to a safety performance quantification evaluation device for an automatic driving system, and includes at least one processor, and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0095] establish a whole vehicle accident rate decomposition model to decompose hazard events at a whole vehicle layer into those caused by functional deficiencies at a component layer, wherein the component layers include a perception layer, a positioning layer, a planning layer, and a control layer;

[0096] obtain statistical probabilities of hazard events caused by functional deficiencies at the planning layer, the control layer, and the positioning layer;

[0097] perform scenario modeling on probabilities of hazard events caused by functional deficiencies at the perception layer;

[0098] obtain confirmation targets of each function based at least on a natural driving data set, the whole vehicle accident rate decomposition model, the statistical probabilities of hazard events, and the scenario modeling of the perception layer;

[0099] use the confirmation targets of each function as a standard for safety performance quantification evaluation.

[0100] As another implementation, the electronic device is applied to a safety performance quantification evaluation device for an automatic driving system, and includes at least one processor, and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0101] establish a perception layer accident rate decomposition model, decompose hazard events of the perception layer into hazards caused by functional deficiencies of the perception layer;

[0102] scene model the probability of hazard events caused by functional deficiencies of the perception layer;

[0103] obtain the confirmation targets of each function of the perception layer based on at least a natural driving data set, the perception layer accident rate decomposition model, and the scene modeling of the perception layer.

[0104] The electronic device of the embodiments of the present application exists in various forms, including but not limited to:

[0105] (1) Mobile communication device: The feature of this type of device is to have mobile communication function and to provide voice and data communication as the main target. This type of terminal includes smart phones (such as iPhone), multimedia phones, functional phones, and low-end phones, etc.

[0106] (2) Ultra-mobile personal computer device: This type of device belongs to the category of personal computers and has computing and processing functions, and generally also has the feature of mobile Internet. This type of terminal includes PDA, MID and UMPC devices, such as iPad.

[0107] (3) Portable entertainment device: This type of device can display and play multimedia content. This type of device includes audio and video players (such as iPod), handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0108] (4) Server: A device that provides computing services. The components of a server include a processor, a hard disk, a memory, a system bus, etc. The server is similar to a general-purpose computer in architecture, but requires higher processing capability, stability, reliability, security, scalability, and manageability due to the need to provide high-reliability services.

[0109] (5) Other electronic devices with data interaction function.

[0110] The device embodiments described above are merely illustrative, wherein the units illustrated as separate components can or can not be physically separate, and the components illustrated as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0111] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, and the computer software products can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods of the embodiments or some parts of the embodiments.

[0112] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for safety performance quantification evaluation of an autonomous driving system, comprising: establishing a vehicle accident rate decomposition model to decompose hazard events at a vehicle level into those caused by functional deficiencies at component levels, wherein the component levels include a perception level, a localization level, a planning level, and a control level; obtaining statistical probabilities of hazard events caused by functional deficiencies at the planning level, the control level, and the localization level; modeling scenarios of hazard events caused by functional deficiencies at the perception level; obtaining confirmation targets of each function based on at least a natural driving dataset, the vehicle accident rate decomposition model, the statistical probabilities of hazard events, and the scenario modeling of the perception level; and using the confirmation targets of each function as criteria for safety performance quantification evaluation.

2. The method of claim 1, wherein, The decomposing hazard events at the vehicle level into those caused by functional deficiencies at the component levels comprises: in the case of functional deficiencies at the component levels, hazard behaviors are exposed due to the functional deficiencies, the hazard behaviors are uncontrollable, and accidents caused thereby have a preset severity, then it is considered that the functional deficiencies at the component levels cause hazard events at the vehicle level.

3. The method of claim 1, wherein, The obtaining statistical probabilities of hazard events caused by functional deficiencies at the planning level, the control level, and the localization level comprises: obtaining the statistical probabilities of hazard events caused by functional deficiencies at the planning level, the control level, and the localization level by at least one of simulation testing, real vehicle testing, dataset extraction, and / or experience value setting.

4. The method of claim 1, wherein, The functional deficiencies at the perception level include missed detection and false detection, and the modeling scenarios of hazard events caused by functional deficiencies at the perception level comprises: modeling scenarios of a probability of collision with a front obstacle caused by missed detection; and modeling scenarios of a probability of rear-end collision after false braking caused by false detection.

5. The method of claim 4, wherein, Setting a relationship between a missed detection rate and a false detection rate, the obtaining confirmation targets of each function based on at least a natural driving dataset, the vehicle accident rate decomposition model, the statistical probabilities of hazard events, and the scenario modeling of the perception level comprises: extracting parameters in the scenario modeling of the perception level from the natural driving dataset; obtaining the missed detection rate and the false detection rate of the perception level based on the extracted parameters, the statistical probabilities of hazard events, the set relationship between the missed detection rate and the false detection rate, the vehicle accident rate decomposition model, and assignment of related parameters in the vehicle accident rate decomposition model; and setting the missed detection rate and the false detection rate to conform to a preset distribution, and obtaining a confirmation target of the missed detection rate or the false detection rate based on the missed detection rate or the false detection rate, respectively, in combination with the preset distribution. The setting a relationship between a missed detection rate and a false detection rate comprises setting the missed detection rate to be lower than the false detection rate.

6. The method of claim 5, wherein, The functional deficiencies at the perception level include detection accuracy errors, and the method comprises:

7. The method of claim 1, wherein, obtaining the detection accuracy errors by establishing a driver model for simulation statistics; and using the detection accuracy errors as constraints on detection accuracy errors of the autonomous driving system. The method further comprises:

8. The method according to any one of claims 1-7, characterized in that, The formula of the vehicle accident rate decomposition model λ is expressed as follows: where p pl，j，k represents the probability of a functional deficiency in each component layer, ps j，k represents the probability of exposure of a harmful behavior in a scenario due to a functional deficiency, p c|ps represents the probability of uncontrolled harmful behavior given exposure of a harmful behavior in a scenario, p s|c represents the probability of severity of an accident given uncontrolled harmful behavior, p i represents the probability in each speed interval, j represents each component layer, and k represents a functional deficiency in each component layer.

9. The method according to any one of claims 1-7, characterized in that, using the confirmation targets of each function as design and development targets in an initial stage of design and development of the autonomous driving system; and / or ​ The confirmation target of each function is used as a threshold for judging whether a test result is qualified in vehicle testing.

10. A safety performance quantification evaluation method for a perception layer of an autonomous driving system, comprising: establishing a perception layer accident rate decomposition model to decompose a hazard event of the perception layer into a function deficiency of the perception layer; modeling a scenario of a probability of the hazard event caused by the function deficiency of the perception layer; obtaining a confirmation target of each function of the perception layer based on at least a natural driving data set, the perception layer accident rate decomposition model, and the scenario modeling of the perception layer.

11. The method of claim 10, wherein, The decomposing of the hazard event of the perception layer into the function deficiency of the perception layer comprises: in the case of the function deficiency of the perception layer, a hazard behavior is exposed due to the function deficiency, the hazard behavior is uncontrollable, and an accident with a preset severity is caused, and it is considered that the function deficiency of the perception layer causes a perception layer hazard event.

12. The method of claim 10, wherein, The function deficiency of the perception layer comprises a missed detection and a false detection, and the scenario modeling of the probability of the hazard event caused by the function deficiency of the perception layer comprises: modeling a scenario of a probability of a collision with a front obstacle caused by the missed detection; modeling a scenario of a probability of a rear-end collision after a false brake caused by the false detection.

13. The method of claim 12, wherein, Setting a relationship between a missed detection rate and a false detection rate, the obtaining of the confirmation target of each function of the perception layer based on at least the natural driving data set, the perception layer accident rate decomposition model, and the scenario modeling of the perception layer comprises: extracting a parameter in the scenario modeling of the perception layer by using the natural driving data set; obtaining a missed detection rate and a false detection rate of the perception layer based on the extracted parameter, the set relationship between the missed detection rate and the false detection rate, the whole vehicle accident rate decomposition model, and an assignment of a related parameter in the perception layer accident rate decomposition model; setting that the missed detection rate and the false detection rate both conform to a preset distribution, and obtaining a confirmation target of the missed detection rate or the false detection rate based on the missed detection rate or the false detection rate, respectively, in combination with the preset distribution.

14. The method of claim 13, wherein, The setting of the relationship between the missed detection rate and the false detection rate comprises setting that the missed detection rate is lower than the false detection rate.

15. The method of claim 10, wherein, The function deficiency of the perception layer comprises a detection accuracy error, and the method comprises: obtaining the detection accuracy error by establishing a driver model for simulation statistics; using the detection accuracy error as a detection accuracy error constraint of the autonomous driving system.

16. The method according to any one of claims 10-15, characterized in that, A vehicle accident rate decomposition model λ sence is expressed by the following formula: where p FN，j，1 represents the perception layer false negative rate, p FP，j，2 represents the perception layer false positive rate, p j，1 represents the probability of exposure of harmful behavior in the scene due to the false negative rate, p FP，i，2 represents the probability of exposure of harmful behavior in the scene due to the false positive rate, p c|ps represents the probability that the harmful behavior is uncontrollable assuming that the harmful behavior is exposed in the scene, p s|c represents the probability of accident severity caused assuming that the harmful behavior is uncontrollable, p i represents the probability under each speed interval, j represents the perception layer, and k represents the functional deficiency of each component layer.

17. An electronic device comprising: At least one processor and a memory connected to the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the method of any one of claims 1-16.

18. A storage medium having stored thereon a computer program, characterized in that The program is executed by the processor to implement the steps of the method of any one of claims 1-16.

19. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instructions are executed by the processor to implement the steps of the method of any one of claims 1-16.

Images