Vehicle Ego State Estimation Using Extreme Value Theory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for validating and verifying vehicle ego state estimations in autonomous drive and advanced driver assistance systems are cumbersome, costly, and time-consuming, particularly when modeling rare events, and require large amounts of data.
Innovation Solution
A method that reduces the data required for quantifying vehicle ego state estimation performance by using a pre-determined statistical extreme value distribution, such as Generalized Pareto or Generalized Extreme Value distribution, to model and parameterize vehicle ego state estimation performance, allowing for efficient analysis of rare events and improved model verification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional validation and verification methods are used for vehicle ego state estimators, then high integrity and reliability are achieved, but the process becomes technically cumbersome, costly, and time consuming
Solution Approach 1:
The patent transforms the validation approach by changing from traditional statistical methods to extreme value theory (EVT). This parameter change in the mathematical framework allows for more efficient modeling of rare events, reducing the computational burden and time required while maintaining high reliability standards for ego state estimation validation
Solution Approach 2:
The patent replaces the mechanical approach of extensive physical testing and traditional statistical validation with a computational method based on extreme value theory. This substitution allows virtual validation that is both faster and more targeted at critical rare events, reducing both time and cost while maintaining reliability
2Reliability
If traditional validation methods are used to model rare events, then adequate coverage is achieved, but the process becomes costly and time consuming
Solution Approach 1:
The patent applies extreme value theory which specifically models the tails of distributions where rare events occur. By changing the statistical parameters and methods to focus on extreme values rather than overall distribution, the system achieves better rare event coverage with significantly reduced computational resources and faster development cycles
Solution Approach 2:
The patent extracts and focuses specifically on the extreme tail events from the overall data distribution using EVT. By taking out only the critical rare events for specialized modeling rather than attempting to model all events uniformly, the system achieves high reliability for rare events while improving development productivity
3Measurement precision
If large amounts of data are collected for validation, then comprehensive coverage is achieved, but the data processing becomes technically cumbersome and costly
Solution Approach 1:
The patent extracts only the extreme value samples from the overall data set that are relevant for validating rare event handling. By taking out only these critical samples for EVT analysis rather than processing the entire data set, the system maintains high measurement precision for rare events while significantly reducing data processing complexity
Solution Approach 2:
The patent changes the validation parameters to focus on extreme value statistics rather than overall statistical properties. This parameter change allows validation to be performed on a smaller subset of extreme samples while maintaining or improving accuracy for the critical rare events that matter most for safety
Data Source
Figure 1
Figure 2~4
Figure 5~7
AI summary
A method for quantifying vehicle ego state estimation performance, the method comprising; obtaining samples of ego state estimation performance pii=1n−1, selecting a subset of the ego state estimation performance samples such that the selected samples follow a pre-determined statistical extreme value distribution, parameterizing the pre-determined statistical extreme value distribution based on the selected samples of ego state estimation performance, and quantifying vehicle ego state estimation performance based on the parameterized statistical extreme value distribution.