Autonomous Vehicle Risk Exposure Estimation for Virtual ADS Safety Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated driving systems face challenges in accurately estimating risk exposure, particularly in scenarios beyond conventional traffic rules and uncertainties, such as corner cases involving vulnerable road users, aggressive maneuvers, and adverse weather conditions, making it difficult to ensure safety without extensive real-world data collection.
Innovation Solution
A computer-implemented method and system that estimates risk exposure in a virtual test environment by analyzing candidate states, predicted scenes, and events using probabilistic reliability methods, incorporating risk thresholds and accounting for perceived and non-perceived objects, as well as the ADS's capabilities, to determine high-risk or acceptable-risk states, thereby enhancing safety and operational reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If extensive real-world data collection is used to ensure safety, then reliability is improved, but loss of time and productivity deteriorate due to the intractable amounts of data required
Solution Approach 1:
The patent applies preliminary action by performing risk estimation through virtual test environment simulations before deploying the ADS to real-world operations. The system pre-identifies high-risk states and scenarios through probabilistic reliability methods, allowing safety validation to occur in advance rather than requiring extensive post-deployment data collection. This enables safety assessment without waiting for rare adverse events to occur in real traffic.
Solution Approach 2:
The patent uses copying by creating virtual replicas of real-world driving scenarios in a simulated test environment. Instead of collecting and analyzing actual real-world data directly, the system generates synthetic test cases that replicate dangerous situations, allowing risk assessment to be performed on copies of real scenarios rather than requiring extensive real data collection.
2Measurement precision
If conventional testing methods are used, then device complexity is reduced, but measurement precision deteriorates due to inability to accurately estimate risk in corner cases
Solution Approach 1:
The patent introduces an intermediary layer between simple testing and complex real-world data collection by implementing a virtual test environment with probabilistic reliability methods. This intermediary system uses limit state functions and simulation models to bridge the gap between conventional testing and accurate risk measurement, enabling precise risk estimation for corner cases without requiring the full complexity of extensive real-world deployment.
Solution Approach 2:
The patent applies parameter changes by transforming the risk estimation problem from direct real-world observation to virtual simulation with controlled parameters. The system changes the measurement parameters from actual accident frequencies to probabilistic risk metrics derived from limit state functions, allowing accurate risk assessment for rare events by changing how risk is quantified rather than requiring more complex data collection infrastructure.
3Reliability
If safety margins are increased to ensure safety, then reliability is improved, but productivity deteriorates due to unnecessary safety margins reducing system performance
Solution Approach 1:
The patent applies local quality by implementing risk-specific safety measures rather than uniform safety margins across all operating conditions. The system identifies specific high-risk states and scenarios through virtual testing and applies targeted safety constraints only where needed, rather than imposing blanket safety margins that would degrade overall system performance. This allows safety to be enhanced locally in critical situations without globally reducing productivity.
Data Source
Figure 1
Figure 2
Figure 3a~3c
AI summary
The present invention relates to methods and systems for estimating a risk exposure of a vehicle equipped with an automated driving system (ADS) in a virtual test environment. In particular the present invention relates to estimations of the risk exposure of an ADS-equipped vehicle that includes for the second order risk, i.e. the risk which an adverse event imposes on the ADS-equipped vehicle while accounting for the ADS's capability of avoiding an incident (e.g. nearcollision or collision) should the adverse event take place.