Driving test scenarios from real-world driving data is considered as an effective solution for ICV on-road test. The main problem is how to select valid test scenarios from a large number of scenario snapshots. In this paper, a novel test scenario selecting method is proposed. Both the collision risk factor and the traffic factor are considered, and criticality evaluation factors for three typical applications - FCR, LCR and ICR, are defined. A LSTM-AE-Attention model is designed to identify critical scenarios. Experimental results show that the LSTM-AE-Attention based method has rapid convergence and acceptable accuracy while providing critical scenarios is reasonable.
The Critical Scenario Extraction and Identification Method for ICV Testing
2023-09-24
2593383 byte
Conference paper
Electronic Resource
English
LIGHTWEIGHT IN-VEHICLE CRITICAL SCENARIO EXTRACTION SYSTEM
European Patent Office | 2022
|CRITICAL SCENARIO IDENTIFICATION FOR VERIFICATION AND VALIDATION OF VEHICLES
European Patent Office | 2023
|CRITICAL SCENARIO IDENTIFICATION FOR VERIFICATION AND VALIDATION OF VEHICLES
European Patent Office | 2022
|CRITICAL SCENARIO IDENTIFICATION FOR VERIFICATION AND VALIDATION OF VEHICLES
European Patent Office | 2024
|