保存先:
| 主要な著者: | , , , , , |
|---|---|
| フォーマット: | Recurso digital |
| 言語: | |
| 出版事項: |
Zenodo
2026
|
| 主題: | |
| オンライン･アクセス: | https://doi.org/10.5281/zenodo.20073406 |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
目次:
- <p>This archive contains the full simulation data, calibration artifacts, and analysis outputs accompanying the IEEE T-ITS submission:</p> <p>Hossain, M.B., Kamal, M.A.S., Rahman, S.S., Tayan, O., Mehedi, I.M., Showail, A. (2026). "Real-Time Risk-Aware Supervisory Control for Autonomous Driving with Quantile-Calibrated Safety Envelopes." IEEE Transactions on Intelligent Transportation Systems (under review).</p> <p>The hypotheses, analysis plan, exclusion criteria, and disconfirmation conditions were pre-registered on the Open Science Framework on 2026-05-01, before any Stage B evaluation data was analysed. The pre-registration is at https://osf.io/sfdj2.</p> <p>CONTENTS</p> <p>Main study: 1,800 unique CARLA episodes across three independent random seeds (1000, 2000, 3000) and two urban towns (Town03, Town05). 720 Stage A NORMAL-only calibration episodes plus 1,080 held-out Stage B paired evaluation episodes re-run under eight controllers (NORMAL, four TTC tuning grid points at theta in {2.0, 2.5, 3.0, 3.5} s, and three calibrated controllers at tau in {0.10, 0.15, 0.20}), for a total of 8,640 controller-runs. Three ZIPs: seed_1000.zip, seed_2000.zip, seed_3000.zip.</p> <p>Calibration artifacts: per-seed boundary B_{N,tau} and softness delta values used by the calibrated controllers, in JSON format inside each seed_*.zip under seed_*/calibrations/.</p> <p>Pilot data (April 2026): 999 NORMAL-only Stage A calibration episodes (which informed the Town04 exclusion criterion) and 94 paired Stage B evaluation episodes (referenced in pre-registration section 4.4 power analysis). One ZIP: pilot.zip.</p> <p>Analysis outputs: pre-computed per-hypothesis result CSVs corresponding to paper Tables IX-XIV, plus the master 8,640-row aggregation combined_master.csv. One ZIP: analysis_outputs.zip.</p> <p>Data dictionary: zenodo_README.md describes the per-episode CSV schema, identifier conventions, calibration artifact format, and reproducibility procedures.</p> <p>REPRODUCIBILITY</p> <p>The implementation and analysis code is at https://github.com/autism-researcher/quantile-fca-supervisor. Step-by-step instructions to reproduce every numerical result, table, and figure in the paper from this dataset are in zenodo_README.md.</p> <p>PRE-REGISTRATION COMPLIANCE</p> <p>This dataset is released in accordance with the open-science commitments in the pre-registration section 11.</p> <p>PATENT NOTICE</p> <p>The methods that generated this data relate to U.S. Patent and Trademark Office Application No. 19/533,330 (2026). The CC BY 4.0 license on this dataset does not grant any patent license. See PATENTS.md in the code repository for additional patent notice.</p> <p> </p> <p>Version 1.1 adds a supplementary rare-event power analysis (<code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">rare_event_power_analysis.py</code>) that reproduces all main-text rare-event statistics from the Stage B dataset and computes sample-size requirements for narrowing the rare-event collision-rate confidence intervals.</p>