Identification of Safety-critical Scenarios for Autonomous Cars Using Feature Extraction

This repository contains data for reproducibility of experimental results listed in the paper titled as "Identification of Safety-critical Scenarios for Autonomous Cars Using Feature Extraction".

1. Scenarios.csv contains a set of 60K test scenarios generated by search-based testing of Apollo using SVL simulator

2. feature-extraction-code.zip contains java code for extraction of features from scenarios

3. complete-feature-set.csv contains a complete set of features extracted from the test scenarios

4. metadata.csv contains a subset of features selected and used in our experiments

5. classification.py contains the python implementation of classification algorithms to learn scenario outcomes using the selected features.