Autonomous Driving Evaluation via Time-Inverted Scene Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving evaluation methods fail to effectively assess the performance of autonomous driving algorithms by not considering past traffic scenarios, leading to inadequate evaluation of the algorithm's ability to prevent critical traffic situations.
Innovation Solution
An autonomous driving evaluation apparatus and method that simulate initial traffic scenes, calculate past traffic scenes by retracing time, and evaluate algorithm performance based on past and comparison scenes to recognize convergence or divergence tendencies, ensuring the algorithm does not lead to critical situations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the autonomous driving algorithm is expected to control the vehicle object such that the situation does not fall into the initial traffic scene from the beginning, then the evaluation of the autonomous driving algorithm can be improved, but the related art methods only evaluate from the initial traffic scene forward which is insufficient
Solution Approach 1:
The patent applies time inversion by calculating past traffic scenes from the initial traffic scene backward in time. Instead of only simulating forward from the initial scene, the system traces back to generate past scenes and evaluates whether the autonomous driving algorithm would have prevented the critical situation, thereby improving evaluation accuracy by examining the algorithm's preventive capability.
Solution Approach 2:
The patent performs preliminary calculation of past traffic scenes before the actual evaluation of autonomous driving performance. By pre-generating the past traffic scene data and comparison scenes, the system prepares the necessary evaluation framework in advance, allowing for a more comprehensive assessment of the algorithm's ability to prevent critical situations.
2Measurement precision
If past traffic scene calculation is added to the evaluation process, then the performance evaluation of the autonomous driving algorithm becomes more comprehensive, but the calculation time and processing complexity increase
Solution Approach 1:
The patent creates comparison traffic scenes by copying and modifying the past traffic scene. Instead of performing multiple independent complex simulations, the system generates variant scenes by adjusting specific parameters of the base past scene, thereby reducing calculation time while maintaining comprehensive evaluation precision.
Solution Approach 2:
The patent evaluates autonomous driving performance by changing specific parameters in the past traffic scene to generate comparison scenes. By systematically varying key parameters while keeping others constant, the system achieves comprehensive evaluation with reduced computational burden compared to full re-simulation.
Data Source
AI summary
An autonomous driving evaluation apparatus includes: an initial traffic scene setting unit configured to set an initial traffic scene, an initial state of the moving object model, and a road environment in which the autonomous driving vehicle model and the moving object model are disposed; a past traffic scene calculation unit configured to calculate a past traffic scene in which the autonomous driving vehicle model and the moving object model are involved at a past time point back traced from a time point of the initial traffic scene; and a performance evaluation unit configured to evaluate a performance of the autonomous driving algorithm.