Autonomous Driving Accident Risk Estimation via Manual Data Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in accurately assessing accident risk due to the complexity of driving scenarios, leading to insufficient testing and potential safety issues, as existing methods require extensive autonomous driving cycles to ensure software sophistication and safety.
Innovation Solution
A method that estimates accident risk by comparing autonomous driving patterns with manual driving data, using driving parameters like speed and acceleration to determine an autonomous driving accident rate, allowing for software enhancement without real accidents, and providing a statistical safety rating for insurance and liability purposes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If extensive autonomous driving cycles are used for testing, then software safety and sophistication are improved, but testing time and cost increase significantly
Solution Approach 1:
The system performs preliminary actions by collecting and analyzing manual driving data beforehand to establish baseline accident rates. This preliminary data preparation enables subsequent autonomous driving evaluations to be conducted more efficiently with fewer test cycles, as the framework for risk assessment is already in place before actual autonomous testing begins.
Solution Approach 2:
The patent creates a virtual copy of manual driving behavior patterns by collecting data from multiple manual drivers and synthesizing it into representative accident rate models. These copied manual driving patterns serve as proxies for actual autonomous driving performance, allowing safety assessment without requiring extensive physical autonomous test cycles.
2Measurement precision
If more autonomous driving test cycles are conducted, then accident risk estimation accuracy is improved, but productivity and development speed decrease
Solution Approach 1:
The system introduces an intermediary layer consisting of manual driving data and accident rate models that mediate between actual autonomous driving performance and safety assessment. This intermediary framework allows accurate risk estimation by comparing autonomous driving parameters against established manual driving baselines, eliminating the need for numerous autonomous test cycles while maintaining measurement precision.
Solution Approach 2:
The patent changes the assessment parameters from requiring extensive autonomous driving mileage to using autonomous driving parameter comparisons against manual driving baselines. By transforming the measurement approach from quantity-based (kilometers driven) to quality-based (parameter comparison accuracy), the system achieves high measurement precision with significantly reduced testing requirements.
3Reliability
If autonomous driving software is enhanced based on real accidents, then software capability is improved, but time loss due to retesting increases
Solution Approach 1:
The system implements a feedback mechanism where autonomous driving parameters are continuously compared against manual driving accident rate models. This feedback loop identifies specific parameter deviations that indicate safety risks, allowing targeted software enhancements without requiring actual accidents to occur. The feedback from parameter comparisons guides precise software adjustments, eliminating the need for time-consuming retesting cycles.
4Reliability
If comprehensive testing of all driving situations is performed, then safety coverage is improved, but testing complexity and resource requirements increase
Solution Approach 1:
The patent creates a universal assessment framework that uses manual driving data collected from multiple drivers and various conditions to build comprehensive accident rate models. This universal baseline framework can evaluate any autonomous driving scenario by comparing against the comprehensive manual data, providing broad safety coverage without requiring separate comprehensive testing for each autonomous driving situation.
Data Source
Figure 1~2
Figure 3~4
AI summary
The present invention relates to a method for estimating an accident risk of an autonomous driving unit. To provide a method which produces helpful results with less autonomous driving cycles, the method comprises the steps: - from driving values, which have been determined from monitoring at least one driving parameter of the driving unit during autonomous driving, an autonomous-driving quantity is determined quantifying an autonomous-driving quality of the driving of the driving unit, - the autonomous-driving quantity is associated to a plurality of manual-driving quantities, which have been determined from the same driving parameter during manual driving periods of different driving units, and - from accident rate values associated to those manual-driving quantities an autonomous-driving accident rate value is determined.