Simulation Dataset for Risk-Aware Autonomous Driving with Quantile-Calibrated Safety Envelopes

This dataset contains CARLA-based simulation data used to evaluate a risk-aware supervisory control framework with quantile-calibrated safety envelopes. Contents: 1. Raw Simulation Logs: - stageA_normal.zip: Baseline (normal driving) runs used for risk calibration - stageB_eval.zip: Evaluation runs under different control modes (Normal, TTC-based, Proposed) 2. Processed Dataset: - eval_master.csv: Aggregated per-episode metrics including: • collision indicators • minimum time-to-collision (TTC) • risk values • violation rates 3. Statistical Outputs: - stats_effect_sizes.csv: Effect size comparisons between control strategies 4. Configuration: - calibration.json: Quantile-based calibration parameters used to define the safety boundary Reproducibility: All results reported in the paper can be reproduced using the dataset and the analysis scripts available at:https://github.com/autism-researcher/Risk-Envelope-Simulation (Release v1.0) Environment:- CARLA 0.9.13- Python 3.7- numpy==1.21.6- pandas==1.1.5