Aryaman Gupta

Hi, I’m Aryaman, a Ph.D. candidate at Stanford University, where I feel grateful to be advised by Prof. Somil Bansal in the SIA Lab. Prior to Stanford, I spent four wonderful years at IIT (BHU) Varanasi earning my B.Tech. in Electronics Engineering.

My research focuses on uncovering failure patterns in robotic systems for their continuous improvement. I aim to: (a) bake safety mechanisms into their design, and (b) ensure reliable and robust operation at runtime through lightweight guardrails.

Get an overview of my research, grab my resume, or find my work on Google Scholar. Feel free to reach out — I’m always happy to chat about robotics, AI, and tech in general.

news

May 2026	My team won the Qualcomm Innovation Fellowship (North America) 2026! 🎉
Jun 2025	Presented my paper on MLLM-based failure interpretability at the OOD Generalization and Robot Evaluation Workshops at RSS 2025.
Jun 2025	New paper on MLLM-based semantic interpretation of robotic failure data is out on arXiv.
Jan 2025	Started my Ph.D. in Aeronautics and Astronautics at Stanford University.
Oct 2024	My paper on efficient data routing in small-world IoT networks got accepted into IEEE ANTS 2024 and won the Best Paper Award.

selected publications

Preprint

Unsupervised Discovery of Failure Taxonomies from Deployment Logs

Aryaman Gupta^*, Yusuf Umut Ciftci^*, and Somil Bansal

arXiv preprint arXiv:2506.06570, 2025

arXiv Bib PDF Video Code Website

@article{gupta2025failure,
  title = {Unsupervised Discovery of Failure Taxonomies from Deployment Logs},
  author = {Gupta, Aryaman and Ciftci, Yusuf Umut and Bansal, Somil},
  journal = {arXiv preprint arXiv:2506.06570},
  year = {2025},
}

ICRA

Detecting and Mitigating System-Level Anomalies of Vision-Based Controllers

Aryaman Gupta^*, Kaustav Chakraborty^*, and Somil Bansal

In IEEE International Conference on Robotics and Automation (ICRA), 2024

arXiv Bib PDF Video Code

@inproceedings{gupta2024detecting,
  title = {Detecting and Mitigating System-Level Anomalies of Vision-Based Controllers},
  author = {Gupta, Aryaman and Chakraborty, Kaustav and Bansal, Somil},
  booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
  year = {2024},
}

ASOC
Deep Reinforcement Learning for Sim-to-Real Policy Transfer of VTOL-UAVs Offshore Docking Operations

Ali Mohamed Ali, Aryaman Gupta, and Hashim A. Hashim

Applied Soft Computing, 2024

Abs DOI Bib HTML Code

This paper proposes a novel Reinforcement Learning (RL) approach for sim-to-real policy transfer of Vertical Take-Off and Landing Unmanned Aerial Vehicles (VTOL-UAV). The proposed approach is designed for VTOL-UAV landing on offshore docking stations in maritime operations. The landing procedure is decomposed into an approach phase, handled by a model-based control scheme, and a landing phase, where DRL agents are trained offline to learn the optimal docking policy. The JONSWAP spectrum model generates a wave model for each episode to enhance policy generalization for sim2real transfer. Numerical experiments show that the PPO agent learns efficient policies to land in uncertain environments, enhancing the likelihood of successful sim-to-real transfer.
@article{ali2024deep, title = {Deep Reinforcement Learning for Sim-to-Real Policy Transfer of VTOL-UAVs Offshore Docking Operations}, author = {Ali, Ali Mohamed and Gupta, Aryaman and Hashim, Hashim A.}, journal = {Applied Soft Computing}, volume = {162}, pages = {111843}, year = {2024}, issn = {1568-4946}, doi = {10.1016/j.asoc.2024.111843}, }