A method for generating and evaluating the effectiveness of expected functional safety trigger conditions
By building a sensor knowledge base and establishing a mapping matrix, systematically identifying and evaluating the expected functional safety trigger conditions of autonomous vehicles, the problems of low analysis efficiency and weak interpretability in the existing technology are solved, and efficient and highly interpretable trigger conditions are achieved.
Patent Information
- Application Number
- CN202111267470.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-29
- Publication Date
- 2025-02-28
- Estimated Expiration
- 2041-10-29
AI Technical Summary
During driving, due to the limitations of perception systems and complex algorithms, autonomous vehicles may not be able to achieve the expected functions, resulting in expected functional safety problems. The prior art analytical methods are inefficient, weak interpretability, and it is difficult to target the evaluation of the safety of a single autonomous driving function.
By building a hierarchical model-based sensor trigger source knowledge base, sensor trigger mechanism knowledge base and sensor trigger effect knowledge base, a trigger condition mapping matrix and effectiveness judgment matrix are established, and the effectiveness level of trigger conditions are systematically identified and quantified.
It realizes efficient and highly interpretable generation and evaluation of expected functional safety trigger conditions, reduces the amount of analysis and human subjective factors, has a wide range of application, and is suitable for people who lack professional knowledge.
Smart Images
Figure CN114021329B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of autonomous driving vehicle testing and evaluation, and in particular to a method for generating and evaluating the effectiveness of an expected functional safety trigger condition. Background Art
[0002] Due to the introduction of perception systems and complex algorithms, autonomous driving vehicles may fail to achieve their expected functions without hardware or software failures during driving. Such problems belong to expected functional safety issues. In expected functional safety issues, the complex and changeable driving environment exacerbates the uncertainty of environmental perception, which is specifically manifested in the perception uncertainty caused by hardware limitations and cognitive uncertainty caused by algorithm limitations under the influence of external factors. There are knowledge-based or data-based analysis methods for trigger condition analysis of perception system performance limitations. The knowledge-based analysis method refers to analyzing the factors that may affect the hardware and software based on the physical structure and working principle of the hardware and software. This method has the advantages of being intuitive and highly interpretable. However, when the working principles of the hardware and software are complex or unknown, this method has certain limitations. The data-based analysis method refers to extracting factors that may cause accidents based on autonomous driving data or simulation experiment data. This method does not need to know the physical structure and working principle of the hardware and software in advance, and can be regarded as a "black box" treatment, but requires a large amount of data input, low analysis efficiency, and weak interpretability. In addition, although the data-based method has a high coverage, it is less targeted for a single autonomous driving function. Summary of the invention
[0003] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method for generating and evaluating the effectiveness of expected functional safety trigger conditions, which has high efficiency, strong interpretability, high reliability and wide application range.
[0004] The purpose of the present invention can be achieved by the following technical solutions:
[0005] A method for generating and evaluating the effectiveness of an expected functional safety trigger condition, comprising:
[0006] Construct sensor trigger source knowledge base, sensor trigger mechanism knowledge base and sensor trigger effect knowledge base based on hierarchical model;
[0007] The trigger condition mapping matrix is constructed through the mapping relationship between the sensor trigger source knowledge base, the sensor trigger mechanism knowledge base and the sensor trigger effect knowledge base;
[0008] Acquire the trigger condition according to the trigger condition mapping matrix;
[0009] Evaluate the effectiveness level of the triggering conditions;
[0010] By constructing a trigger condition mapping matrix, the trigger source of the perception system is mapped to a trigger condition marked with a validity level, so that people who lack professional knowledge can use the trigger condition mapping matrix and the trigger condition validity judgment matrix to directly obtain various trigger conditions for the expected functional safety and the trigger condition validity level, thereby achieving systematic identification and direct quantitative evaluation of trigger conditions, reducing the analysis workload and human subjective factors, with high efficiency and strong intuitiveness. At the same time, the trigger conditions are analyzed from the trigger source, trigger mechanism and trigger effect of the sensor, with strong interpretability, high reliability and a wide range of applications.
[0011] Furthermore, the process of obtaining the trigger condition according to the trigger condition mapping matrix includes:
[0012] According to the mapping relationship, the lowest level elements of each knowledge base are combined to generate trigger conditions.
[0013] Furthermore, the evaluation process of the effectiveness level of the trigger condition includes the following steps:
[0014] 31) Obtain effectiveness evaluation indicators;
[0015] 32) Evaluate the effectiveness level of the trigger condition based on the effectiveness evaluation index.
[0016] Furthermore, the effectiveness evaluation index includes an exposure rate level of the sensor trigger source and a severity level of the sensor trigger effect.
[0017] Furthermore, the step 32) includes:
[0018] Construct an effectiveness judgment matrix with the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect as judgment conditions and the effectiveness level as the judgment result;
[0019] According to the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect, the effectiveness level of the trigger condition is obtained through the effectiveness judgment matrix.
[0020] Furthermore, the evaluation process of the exposure rate level of the sensor trigger source includes:
[0021] Obtain the occurrence probability of the sensor trigger source based on the historical driving data of the car;
[0022] Construct an exposure rate level judgment matrix with the occurrence probability as the judgment condition and the exposure rate level as the judgment result;
[0023] According to the occurrence probability of the sensor trigger source, the exposure rate level of the sensor trigger source is obtained through the exposure rate level judgment matrix.
[0024] Furthermore, the evaluation process of the severity level of the sensor triggering effect includes:
[0025] Construct a severity level judgment matrix with the severity of the sensor triggering effect as the judgment condition and the severity level as the judgment result;
[0026] According to the severity of the sensor triggering effect, the severity level of the sensor triggering effect is obtained through the severity level judgment matrix.
[0027] Furthermore, the sensor trigger source knowledge base includes three levels, the first level nodes include target entities, non-target entities and environmental conditions, the second level nodes are element types, and the third level nodes are specific element individuals;
[0028] The target entities include road conditions, traffic facilities, target objects and non-target objects; the non-target entities include tiny objects and floating objects; the environmental conditions include precipitation, atmospheric particles, light and energy sources.
[0029] Furthermore, the sensor trigger mechanism knowledge base includes two levels, the first level nodes include attribute guide words, factor guide words and feature guide words, the attribute guide words are used to describe that the trigger source is triggered by a specific attribute, the factor guide words are used to describe that a specific attribute affects perception, and the feature guide words are used to describe changes in object features;
[0030] The second-level nodes include specific guide words;
[0031] The specific guide words of the attribute guide words include material, color, shape, surface structure, speed, composition, concentration, particle size, lighting angle, lighting intensity, vibration frequency and temperature; the specific guide words of the factor guide words include reflectivity, reflection area, refresh frequency, transmittance, brightness and inherent characteristics; the specific guide words of the feature guide words include variability, observability, separability and difference.
[0032] Furthermore, the sensor trigger effect knowledge base includes three levels. The first-level nodes include active sensors and passive sensors, the second-level nodes include perception processes and cognitive processes, and the third-level nodes include perception effects and cognitive effects, which record the effects of the trigger conditions on the sensors during the perception and cognitive processes of different sensors.
[0033] Compared with the prior art, the present invention has the following beneficial effects:
[0034] The present invention maps the trigger source of the perception system to the trigger condition marked with the effectiveness level by constructing the trigger condition mapping matrix and the trigger condition effectiveness judgment matrix, so that personnel lacking professional knowledge can use the trigger condition mapping matrix and the trigger condition effectiveness judgment matrix to directly obtain various trigger conditions for the expected functional safety and the effectiveness level of the trigger conditions, realize the systematic identification and rapid and quantitative evaluation of the trigger conditions, and use them as the design basis of the sensor performance indicators, reduce the analysis workload and human subjective factors, have high efficiency and strong intuitiveness, and analyze the trigger conditions from the trigger source, trigger mechanism and trigger effect of the sensor respectively, have strong interpretability, high reliability, and a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 It is a schematic diagram of the method flow of the present invention;
[0036] Figure 2 It is a structural diagram of the sensor trigger source knowledge base;
[0037] Figure 3 It is a schematic diagram of the structure of the sensor trigger mechanism knowledge base;
[0038] Figure 4 It is a structural diagram of the sensor trigger effect knowledge base;
[0039] Figure 5 It is the trigger condition mapping matrix. DETAILED DESCRIPTION
[0040] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.
[0041] A method for generating and evaluating the effectiveness of expected functional safety trigger conditions, such as Figure 1 , including the following steps:
[0042] 1) Construct a sensor trigger source knowledge base, a sensor trigger mechanism knowledge base, and a sensor trigger effect knowledge base based on a hierarchical model;
[0043] 2) Constructing a trigger condition mapping matrix through the mapping relationship between the sensor trigger source knowledge base, the sensor trigger mechanism knowledge base, and the sensor trigger effect knowledge base;
[0044] 3) Obtaining trigger conditions according to the trigger condition mapping matrix;
[0045] 4) Evaluate the effectiveness level of the triggering conditions.
[0046] By constructing a trigger condition mapping matrix, the trigger source of the perception system is mapped to a trigger condition marked with a validity level, so that people who lack professional knowledge can use the trigger condition mapping matrix and the trigger condition validity judgment matrix to directly obtain various trigger conditions for the expected functional safety and the trigger condition validity level, thereby achieving systematic identification and direct quantitative evaluation of trigger conditions, reducing the analysis workload and human subjective factors, with high efficiency and strong intuitiveness. At the same time, the trigger conditions are analyzed from the trigger source, trigger mechanism and trigger effect of the sensor, with strong interpretability, high reliability and a wide range of applications.
[0047] like Figure 2 ,The sensor trigger source knowledge base includes three levels, the first level nodes include target entities, non-target entities and environmental conditions, the second level nodes are element types, and the third level nodes are specific element individuals;
[0048] A target entity refers to an element with a certain volume, specific shape, and certain physical mass that imposes certain constraints on vehicle travel. The types of elements included in the target entity include road conditions, traffic facilities, target objects, and non-target objects. Non-target entities exist in space, have negligible volume and mass, and will not pose obstacles to vehicle travel. The types of elements included in non-target entities include tiny objects and floating objects. Environmental conditions include precipitation, atmospheric particles, light, and energy sources.
[0049] The sensor trigger mechanism knowledge base includes two levels. The first level nodes include attribute guide words, factor guide words and feature guide words. Attribute guide words are used to describe the trigger source to be triggered by specific attributes, factor guide words are used to describe the specific attributes that affect perception, and feature guide words are used to describe the changes in object features.
[0050] The second-level nodes include specific guide words;
[0051] like Figure 3 The specific guide words of attribute guide words are divided into entity attribute guide words and environmental condition attribute guide words. Entity attribute guide words include material, color, shape, surface structure and speed. Environmental condition attribute guide words are divided into precipitation and atmospheric particle attribute guide words, illumination attribute guide words and capability source attribute guide words. Precipitation and atmospheric particle attribute guide words include composition, concentration and particle size. Illumination attribute guide words include illumination angle and illumination intensity. Capability source attribute guide words include vibration frequency and temperature.
[0052] The specific guide words of factor guide words include reflectivity, reflective area, refresh rate, transmittance, brightness and intrinsic characteristics, and the specific guide words of feature guide words include variability, observability, separability and difference;
[0053] Take LiDAR as an example. LiDAR is an active sensor that emits detection signals to the target and then analyzes the signal information reflected from the target to obtain relevant information about the target. Analyzing the performance limitations of LiDAR, during the perception and cognition process of LiDAR, it can interfere with LiDAR by affecting the laser propagation process, the target reflection process, and the sensor itself. The attribute guide words in the laser propagation process include composition, concentration, and particle size, the factor guide words include reflectivity and transmittance, and the feature guide words include variability, observability, separability, and difference.
[0054] like Figure 4 ,The sensor trigger effect knowledge base includes three levels.,The first level nodes include active sensors and passive sensors, the second level nodes include ,perception process and cognitive process, and the third level nodes include perception ,effects and cognitive effects, which record the effects of trigger conditions on sensors ,during the perception and cognitive processes of different sensors;
[0055] Taking lidar as an example, in the perception process of lidar, due to the triggering of trigger conditions, there are triggering effects such as point cloud missing, point cloud distortion and point cloud noise. In the cognitive process, due to the triggering of trigger conditions, there are triggering effects such as semantic segmentation error, recognition error, non-recognition, reduced recognition accuracy, tracking loss and tracking error.
[0056] The trigger condition mapping matrix generated in this embodiment is as follows: Figure 5 As shown, the trigger condition mapping matrix has the trigger source and the trigger mechanism guide word as columns and the trigger effect as rows, so as to realize orthogonal combination mapping.
[0057] Step 3) The process includes:
[0058] According to the mapping relationship, the lowest level elements of each knowledge base are combined to generate trigger conditions.
[0059] The attribute guide words in the sensor trigger mechanism knowledge base map the impact of the trigger source on the perception and cognitive process of the vehicle perception system through factor guide words and feature guide words respectively.
[0060] The process of evaluating the effectiveness level of the triggering conditions includes the following steps:
[0061] 31) Obtain effectiveness evaluation indicators;
[0062] 32) Evaluate the effectiveness level of the trigger condition based on the effectiveness evaluation index.
[0063] The effectiveness evaluation indicators include the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect.
[0064] Step 32) includes:
[0065] Construct an effectiveness judgment matrix with the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect as judgment conditions and the effectiveness level as the judgment result;
[0066] According to the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect, the effectiveness level of the trigger condition is obtained through the effectiveness judgment matrix.
[0067] The process of evaluating the exposure rate level of the sensor trigger source includes:
[0068] Obtain the occurrence probability of the sensor trigger source based on the historical driving data of the car;
[0069] Construct an exposure rate level judgment matrix with the occurrence probability as the judgment condition and the exposure rate level as the judgment result;
[0070] According to the occurrence probability of the sensor trigger source, the exposure rate level of the sensor trigger source is obtained through the exposure rate level judgment matrix.
[0071] The severity level of the sensor triggering effect is assessed by:
[0072] Construct a severity level judgment matrix with the severity of the sensor triggering effect as the judgment condition and the severity level as the judgment result;
[0073] According to the severity of the sensor triggering effect, the severity level of the sensor triggering effect is obtained through the severity level judgment matrix.
[0074] The exposure rate level judgment matrix obtained in this embodiment is shown in Table 1:
[0075] Table 1 Exposure rate level judgment matrix
[0076] E1 p≤0.05％ E2 0.05＜p≤1％ E3 (1％<p<10％) E4 (p>10％)
[0077] Among them, p is the probability of occurrence of the trigger condition, and the exposure rate levels are divided from low to high into E1 (extremely low exposure rate), E2 (low exposure rate), E3 (moderate exposure rate)) and E4 (high exposure rate).
[0078] The severity level judgment matrix obtained in this embodiment is shown in Table 2:
[0079] Table 2 Severity level judgment matrix
[0080] S0 No harm S1 Minor or moderate injury S2 Serious but non-life-threatening injuries S3 Life-threatening injuries
[0081] The severity levels are divided into S0, S1, S2 and S3 from low to high.
[0082] The validity judgment matrix is shown in Table 3:
[0083] Table 3 Validity judgment matrix
[0084]
[0085] The effectiveness levels are divided from low to high into R1, R2, R3, R4 and R5.
[0086] By using the effectiveness evaluation method proposed in this embodiment, it is concluded that the effectiveness level of the trigger condition that there is an obstruction on the laser radar surface is R5. Relevant simulation test cases are designed to verify the effectiveness level of the trigger condition. This embodiment designs a test case for the trigger condition that there is an obstruction on the laser radar surface based on a SAE L3 autonomous driving car. Three tests were conducted on the trigger condition that there is an obstruction on the laser radar surface. In all three tests, the vehicle deviated from the expected driving behavior, which proves the reliability of the effectiveness evaluation method proposed in this embodiment.
[0087] This embodiment proposes a method for generating and evaluating the effectiveness of expected functional safety trigger conditions. By constructing a trigger condition mapping matrix, any form of actual trigger conditions of the vehicle are mapped into trigger conditions marked with effectiveness levels, so that people who lack professional knowledge can use the trigger condition mapping matrix to directly obtain the effectiveness levels of various trigger conditions for expected functional safety, thereby achieving rapid and quantitative evaluation of trigger conditions, reducing analysis workload and human subjective factors, and having high efficiency and strong intuitiveness. At the same time, the trigger conditions are analyzed from the trigger source, trigger mechanism and trigger effect of the sensor respectively, and have strong interpretability, high reliability and a wide range of applications.
[0088] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.
Claims
1. A method for generating and evaluating the effectiveness of an expected functional safety trigger condition, characterized in that: include: Construct sensor trigger source knowledge base, sensor trigger mechanism knowledge base and sensor trigger effect knowledge base based on hierarchical model; The trigger condition mapping matrix is constructed through the mapping relationship between the sensor trigger source knowledge base, the sensor trigger mechanism knowledge base and the sensor trigger effect knowledge base; Acquire the trigger condition according to the trigger condition mapping matrix; Evaluate the effectiveness level of the triggering conditions, The process of obtaining the trigger condition according to the trigger condition mapping matrix includes: According to the mapping relationship, the lowest level elements of each knowledge base are combined to generate trigger conditions. The evaluation process of the effectiveness level of the trigger condition includes the following steps: 31) Obtain effectiveness evaluation indicators; 32) Evaluate the effectiveness level of the triggering conditions based on the effectiveness evaluation indicators. The effectiveness evaluation index includes the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect. The step 32) comprises: Construct an effectiveness judgment matrix with the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect as judgment conditions and the effectiveness level as the judgment result; According to the exposure rate level of the sensor trigger source and the severity level of the sensor trigger effect, the effectiveness level of the trigger condition is obtained through the effectiveness judgment matrix.
2. The method for generating and evaluating the effectiveness of an expected functional safety trigger condition according to claim 1, characterized in that: The evaluation process of the exposure rate level of the sensor trigger source includes: Obtain the occurrence probability of the sensor trigger source based on the historical driving data of the car; Construct an exposure rate level judgment matrix with the occurrence probability as the judgment condition and the exposure rate level as the judgment result; According to the occurrence probability of the sensor trigger source, the exposure rate level of the sensor trigger source is obtained through the exposure rate level judgment matrix.
3. The method for generating and evaluating the effectiveness of an expected functional safety trigger condition according to claim 1, characterized in that: The evaluation process of the severity level of the sensor triggering effect includes: Construct a severity level judgment matrix with the severity of the sensor triggering effect as the judgment condition and the severity level as the judgment result; According to the severity of the sensor triggering effect, the severity level of the sensor triggering effect is obtained through the severity level judgment matrix.
4. The method for generating and evaluating the effectiveness of an expected functional safety trigger condition according to claim 1, characterized in that: The sensor trigger source knowledge base includes three levels. The first level nodes include target entities, non-target entities and environmental conditions, the second level nodes are element types, and the third level nodes are specific element individuals.
5. The method for generating and evaluating the effectiveness of an expected functional safety trigger condition according to claim 1, characterized in that: The sensor trigger mechanism knowledge base includes two levels, the first level nodes include attribute guide words, factor guide words and feature guide words, and the second level nodes include specific guide words.
6. The method for generating and evaluating the effectiveness of an expected functional safety trigger condition according to claim 1, characterized in that: The sensor trigger effect knowledge base includes three levels. The first level nodes include active sensors and passive sensors, the second level nodes include perception processes and cognitive processes, and the third level nodes include perception effects and cognitive effects.