ADS Collision Severity Estimation Using Limit State Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for verifying compliance of Autonomous Driving Systems (ADS) with safety norms for different severity levels are inefficient and resource-intensive, making it challenging to estimate failure probabilities accurately, especially for rare events.
Innovation Solution
A computer-implemented method using a structural reliability approach, such as subset simulation, with an extended Limit State Function (LSF) that distinguishes between failure and severity levels, allowing for the estimation of probability of failure for different severity classes in a virtual test environment, thereby reducing verification and development time and resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If brute force methods are used to estimate failure probability for ADS compliance verification, then comprehensive safety assessment can be achieved, but the method is proven unfeasible due to excessive computational resources and time requirements
Solution Approach 1:
The patent segments the continuous severity spectrum into discrete severity levels (S0-S4) based on damage criteria. This segmentation transforms the intractable continuous safety verification problem into manageable discrete categories, allowing efficient estimation of failure probabilities for each severity level while maintaining comprehensive safety assessment capability
Solution Approach 2:
The patent introduces a Limit State Function that transforms the safety verification problem by changing the parameter representation from direct failure probability estimation to structured severity-based probability estimation. This parameter transformation enables efficient computation by organizing the verification around defined severity thresholds rather than exhaustive scenario analysis
2Measurement precision
If comprehensive safety verification is performed to ensure compliance with safety norms, then accurate failure probability estimation can be achieved, but it requires intractable amounts of data and time (e.g., 100 vehicles for 5 centuries)
Solution Approach 1:
The patent establishes predefined severity levels and damage criteria before conducting the safety verification. This preliminary structuring of the verification framework allows for efficient data collection and analysis by categorizing scenarios according to predetermined severity thresholds, eliminating the need for exhaustive long-term data collection
Solution Approach 2:
By transforming the verification approach to estimate failure probabilities for discrete severity levels rather than requiring precise continuous measurement, the patent achieves accurate safety assessment with significantly reduced data requirements and verification time
3Reliability
If traditional verification methods are used to assess ADS compliance, then general safety assessment can be performed, but detailed failure rate estimates for different severity levels are not provided
Solution Approach 1:
The patent segments the failure probability assessment into distinct severity levels (S0-S4) with specific damage criteria for each level. This segmentation preserves comprehensive safety assessment while simultaneously providing detailed failure rate estimates for each severity category, eliminating the information loss present in traditional aggregate assessment methods
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to a computer-implemented method and processing system for estimating a probability of failure for different severity levels for an Automated Driving System (ADS) feature in a virtual test environment. In more detail, the embodiments of the present disclosure enables estimation of a probability of crash of different severities, by utilizing a limit state function (LSF) that attains increasingly negative or positive values after crash (e.g. when TTC = 0 or PET = 0). This may for example be achieved by defining a function for severity that is more negative for more severe crashes. The LSF may for example comprises a function of the delta speed at collision (i.e. minus delta speed at collision). Being able to generate a probability of failure for different severity classes for a given ADS feature may be advantageous for focusing development and verification activities to the most needed areas/aspects of the system under test (ADS feature under test).