Michel Aractingi

Michel Aractingi
Research Scientist at UMA | Ex-HuggingFace, Naverlabs Europe

Email / CV / Google Scholar / Youtube / Github

Latest Releases

Feb 2026: Released Unfolding Robotics large-scale project for shirt folding.
Jan 2026: The LeRobot paper was accepted at ICLR 2026.
Nov 2025: Gr00T N1.5 in LeRobot, joint Collab with NVIDIA robotics.
Nov 2025: Released Robot Learning: A Tutorial.
Oct 2025: Released LeRobotDataset v3.0 + Streaming for large-scale robotics datasets.
Jun 2025: Released SmolVLA a lightweight, community-drive foundation model for robotics.
Jun 2025: Open-sourced online RL for real-world robotics.

About Me

Feb 2026: Research Scientist at UMA.
2024-2026: Robotics Researcher at Hugging Face 🤗 to work on LeRobot.
2024: Robotics ML engineer at Enchanted Tools.
2020-2023: Ph.D. at Naver Labs Europe and the Gepetto team at LAAS/CNRS, focused on learning control policies for quadruped locomotion.
2018-2020: ML research engineer at Naver Labs Europe working on robotics.
Education: Ph.D. in robotics and computer science, M.Sc. in data science, and B.Sc. in electrical engineering.

Selected Publications

	LeRobot: An Open-Source Library for End-to-End Robot Learning Remi Cadene, Simon Alibert, Francesco Capuano, Michel Aractingi, Adil Zouitine, Pepijn Kooijmans, et al. ICLR, 2026 paper bibtex An open-source library for end-to-end robot learning, spanning robot control, datasets, streaming, and learning algorithms for real-world robotics.
	Robot Learning: A Tutorial Francesco Capuano, Caroline Pascal, Adil Zouitine, Thomas Wolf, Michel Aractingi arXiv, 2025 paper bibtex A tutorial covering modern robot learning, from reinforcement learning and behavioral cloning to generalist language-conditioned robot policies.
	SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics Mustafa Shukor, Dana Aubakirova, Francesco Capuano, Pepijn Kooijmans, Steven Palma, Adil Zouitine, Michel Aractingi, et al. arXiv, 2025 paper bibtex A small, efficient, community-driven vision-language-action model for affordable robotic platforms.
	A Hierarchical Scheme for Adapting Learned Quadruped Locomotion Michel Aractingi, Pierre-Alexandre Leziart, Thomas Flayols, Julien Perez, Tomi Silander, Philippe Soueres IEEE Humanoids, 2023 paper bibtex video A hierarchical approach for learning locomotion policies where several aspects of the low-level behaviour can be comanded.
	Controlling the Solo12 Quadruped with Deep Reinforcement Learning Michel Aractingi, Pierre-Alexandre Leziart, Thomas Flayols, Julien Perez, Tomi Silander, Philippe Soueres Scientific Reports, 2023 paper bibtex video A deep reinforcement learning approach to learn joint-angle control for the solo12 quadruped with a state estimation network.
	DiPCAN: Distilling Privileged Information for Crowd-Aware Navigation Gianluca Monaci Michel Aractingi, Tomi Silander, RSS 2022, Nominated for Best Paper Award paper bibtex video Learning navigation policies in densely crowded environments.
	Learning to Adapt the Trotting Gait of the Solo Quadruped Michel Aractingi, Pierre-Alexandre Leziart, Thomas Flayols, Julien Perez, Tomi Silander, Philippe Soueres Preprint, 2021 paper bibtex We propose to augment the model-based controller of the solo12 quadruped with a policy that modifies the gait sequence learned with deep reinforcement learning.
	Improving the Generalization of Visual Navigation Policies using Invariance Regularization Michel Aractingi, Christopher Dance, Julien Perez, Tomi Silander, ICML 2019 Workshop RL4RealLife paper bibtex We study the generalization ability of visual navigation agents trained with deep RL. We propose a regularization term to improve their generalization ability.

website's template credit goes to Jon Barron