Sim-to-Real Transfer Through Adaptive Post-Training for Autonomous Driving in Unseen Environments
Keywords:
sim-to-real transfer; adaptive post-training; autonomous driving; domain shift; robustness; governance; socio-technical systems; sustainabilityAbstract
Autonomous driving systems increasingly depend on simulation for scalable training, yet the transfer of learned policies to unseen real-world environments remains a central obstacle. This paper examines sim-to-real transfer through the lens of adaptive post-training, understood not merely as a final fine-tuning step but as a socio-technical capability embedded in the lifecycle of autonomous vehicle deployment. The argument is that unseen environments expose structural limits in simulation-based development because they combine perceptual novelty, behavioral ambiguity, infrastructure irregularity, and governance uncertainty. Adaptive post-training can narrow these gaps by recalibrating representations, control policies, and safety constraints after initial training, using limited field data, targeted simulations, and operational feedback. However, such adaptation also introduces new system-level trade-offs involving architectural complexity, verification burden, data governance, fairness, sustainability, and accountability. The paper develops a conceptual framework for analyzing these trade-offs and connects them to practical concerns in autonomous driving, including operational design domains, edge-case exposure, model auditing, and infrastructure dependence. It also compares autonomous driving with robotics, aviation, and large-scale machine learning systems to clarify how post-training changes reliability engineering and policy expectations. The conclusion is that adaptive post-training is necessary but insufficient by itself; it must be paired with transparent validation, continuous monitoring, and institutional mechanisms that distribute responsibility across manufacturers, operators, regulators, and affected communities. The paper contributes a system-level research agenda for making sim-to-real transfer robust, equitable, and sustainable in unseen environments.
References
1. Tobin, J., Fong, R., Ray, A., Schneider, J., Zaremba, W., & Abbeel, P. (2017). Domain randomization for transferring deep neural networks from simulation to the real world. In 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 23–30). IEEE. https://doi.org/10.1109/IROS.2017.8202133
2. Sadeghi, F., & Levine, S. (2016). CAD2RL: Real single-image flight without a single real image. arXiv preprint arXiv:1611.04201.
3. Peng, X. B., Andrychowicz, M., Zaremba, W., & Abbeel, P. (2018). Sim-to-real transfer of robotic control with dynamics randomization. In 2018 IEEE International Conference on Robotics and Automation (ICRA) (pp. 3803–3810). IEEE. https://doi.org/10.1109/ICRA.2018.8460528
4. Dosovitskiy, A., Ros, G., Codevilla, F., Lopez, A., & Koltun, V. (2017). CARLA: An open urban driving simulator. In Proceedings of the 1st Annual Conference on Robot Learning (CoRL) (pp. 1–16).
5. Codevilla, F., Müller, M., López, A., Koltun, V., & Dosovitskiy, A. (2018). End-to-end driving via conditional imitation learning. In 2018 IEEE International Conference on Robotics and Automation (ICRA) (pp. 4693–4700). IEEE. https://doi.org/10.1109/ICRA.2018.8460487
6. Favarò, F., Nader, N., Eurich, S., Tripp, M., & Varadaraju, N. (2017). Examining accident reports involving autonomous vehicles in California. PLoS ONE, 12(9), e0184952. https://doi.org/10.1371/journal.pone.0184952
7. Bojarski, M., Del Testa, D., Dworakowski, D., Firner, B., Flepp, B., Goyal, P., Jackel, L. D., Monfort, M., Muller, U., Zhang, J., Zhang, X., Zhao, J., & Zieba, K. (2016). End to end learning for self-driving cars. arXiv preprint arXiv:1604.07316.
8. Chen, D., Zhou, B., Koltun, V., & Krähenbühl, P. (2020). Learning by cheating. In Proceedings of the 4th Conference on Robot Learning (CoRL) (pp. 66–75).
9. Amini, A., Paull, L., Balakrishnan, H., & Rus, D. (2020). Learning robust control policies for end-to-end autonomous driving from data-driven simulation. IEEE Robotics and Automation Letters, 5(2), 1143–1150. https://doi.org/10.1109/LRA.2020.2966414
10. Bewley, A., Rigley, J., Liu, Y., Hawke, J., Shen, R., Lam, R., & Kendall, A. (2019). Learning to drive from simulation without real world labels. In 2019 International Conference on Robotics and Automation (ICRA) (pp. 4818–4824). IEEE. https://doi.org/10.1109/ICRA.2019.8793668
11. Kendall, A., Hawke, J., Janz, D., Mazur, P., Reda, M., Allen, J., Lam, V.-D., Qureshi, A., & Shah, A. (2019). Learning to drive in a day. In 2019 International Conference on Robotics and Automation (ICRA) (pp. 8248–8254). IEEE. https://doi.org/10.1109/ICRA.2019.8793742
12. Xu, Q., Zhang, S., & Shen, S. (2022). MetaDrive: Composing diverse driving scenarios for generalizable reinforcement learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(3), 3461–3475. https://doi.org/10.1109/TPAMI.2022.3197916
13. Toromanoff, M., Wirbel, E., & Moutarde, F. (2020). End-to-end model-free reinforcement learning for urban driving using implicit affordances. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 7153–7162).
14. Shalev-Shwartz, S., Shammah, S., & Shashua, A. (2017). On a formal model of safe and scalable self-driving cars. arXiv preprint arXiv:1708.06374.
15. Kalra, N., & Paddock, S. M. (2016). Driving to safety: How many miles of driving would it take to demonstrate autonomous vehicle reliability? Transportation Research Part A: Policy and Practice, 94, 182–193. https://doi.org/10.1016/j.tra.2016.09.010
16. Selbst, A. D., Boyd, D., Friedler, S. A., Venkatasubramanian, S., & Vertesi, J. (2019). Fairness and abstraction in sociotechnical systems. In Proceedings of the Conference on Fairness, Accountability, and Transparency (FAT) (pp. 59–68). ACM. https://doi.org/10.1145/3287560.3287598
17. Raji, I. D., Smart, A., White, R. N., Mitchell, M., Gebru, T., Hutchinson, B., Smith-Loud, J., Theron, D., & Barnes, P. (2020). Closing the AI accountability gap: Defining an end-to-end framework for internal algorithmic auditing. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (FAT) (pp. 33–44). ACM. https://doi.org/10.1145/3351095.3372873
18. Mitchell, M., Wu, S., Zaldivar, A., Barnes, P., Vasserman, L., Hutchinson, B., Spitzer, E., Raji, I. D., & Gebru, T. (2019). Model cards for model reporting. In Proceedings of the Conference on Fairness, Accountability, and Transparency (FAT) (pp. 220–229). ACM. https://doi.org/10.1145/3287560.3287596
19. Sculley, D., Holt, G., Golovin, D., Davydov, E., Phillips, T., Ebner, D., Chaudhary, V., Young, M., Crespo, J.-F., & Dennison, D. (2015). Hidden technical debt in machine learning systems. In Advances in Neural Information Processing Systems (NeurIPS) (Vol. 28, pp. 2503–2511).
20. Amodei, D., Olah, C., Steinhardt, J., Christiano, P., Schulman, J., & Mané, D. (2016). Concrete problems in AI safety. arXiv preprint arXiv:1606.06565.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Advanced Artificial Intelligence Research
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
