An Thai Le

I defended my Ph.D. at TU Darmstadt in July 2025 and am now an Assistant Professor at VinUniversity and Director of Foundation AI at VinRobotics. I continue to collaborate with IAS as a Visiting Professor at TU Darmstadt. Please visit my personal website https://anindex.github.io/ for current research, publications, and opportunities to work together!

An Thai Le

An Thai Le joined the Intelligent Autonomous Systems lab on November 1st, 2021, and completed his Ph.D. in 2025 under Prof. Jan Peters, defending his thesis, Tensor Search Methods for Vectorizing Motion Planning, in August. He is grateful to Prof. Peters and the IAS community for their guidance and support. He now works with students and collaborators at VinUniversity and VinRobotics, while continuing to co-advise MSc and PhD students at IAS.

His research interests lie in robot planning, learning, and control, particularly when data and computational resources are limited. A recurring question in his work is how geometry, problem representation, and parallel computation can help make these methods more practical on robot hardware.

Planning and optimization. He works on tensor search methods for parallel motion planning and trajectory optimization, with an interest in how problem structure can make computation more efficient. This work brings together optimal transport, geometric methods, and learned trajectory priors such as diffusion and flow matching.

Robot foundation models. Together with students and collaborators, he studies how robot policies can learn from fewer demonstrations and run efficiently on the hardware available on a robot. Topics include supervision from future observations, model pruning, low-bit quantization, and native inference kernels. Joint projects include FOCA, FoldQuantVLA, vla.simd, and vla.cpp.

Humanoid control. He also works with colleagues on coordinating locomotion, manipulation, and compliance under physical contact and changing payloads. This includes studying how robots can estimate external forces from their own sensors and combine parallel foothold planning with reinforcement learning. Related projects include CompliantWBC and TACT-ful.

His interests also include applications of optimal transport to molecular conformer aggregation, vision-language alignment, and training non-differentiable networks. In his spare time, he explores computational general relativity through numerical tools such as warpax.

He welcomes conversations with students and colleagues interested in these topics. More information about the group and opportunities to work together is available on the group and join us pages.

Before his Ph.D., An Thai Le worked on forceful imitation learning during an internship at Bosch AI. He also worked on task and motion planning for human-robot collaboration. During his master's research, he was advised by Dr. Meng Guo and Dr. Jim Mainprice.

Selected Recent Publications

• FOCA: Future-Oriented Conditioning for Data-Efficient Vision-Language-Action Adaptation. ICML 2026.
• Start Right, Arrive Right: Asynchronous Execution via Initial Noise Selection (PAINT). CoRL 2026.
• Training Non-Differentiable Networks via Optimal Transport (PolyStep). TMLR 2026.
• DoublyAware: Dual Planning and Policy Awareness for Temporal Difference Learning in Humanoid Locomotion. IEEE RA-L 2026.

For the full publication list, including current preprints, see my personal website. Research software is available on the open-source page.

IAS Publications

Optimal Transport in Robotics

1. •       Bib
     Le, A. T.; Chalvatzaki, G.; Biess, A.; Peters, J. (2023). Accelerating Motion Planning via Optimal Transport, Advances in Neural Information Processing Systems (NIPS / NeurIPS).
   •     Bib
     Le, A. T.; Chalvatzaki, G.; Biess, A.; Peters, J. (2023). Accelerating Motion Planning via Optimal Transport, IROS 2023 Workshop on Differentiable Probabilistic Robotics: Emerging Perspectives on Robot Learning, [Oral].
   •     Bib
     Le, A. T.; Chalvatzaki, G.; Biess, A.; Peters, J. (2023). Accelerating Motion Planning via Optimal Transport, NeurIPS 2023 Workshop Optimal Transport and Machine Learning, [Oral].
2. •       Bib
     Le, A. T.; Hansel, K.; Peters, J.; Chalvatzaki, G. (2023). Hierarchical Policy Blending As Optimal Transport, 5th Annual Learning for Dynamics & Control Conference (L4DC), PMLR.

Optimal Transport in Machine Learning

1. •     Bib
     Nguyen, D.M.H.; Lukashina, N.; Nguyen, N.; Le, A.T.; Nguyen, T.T.; Ho, N.; Peters, J.; Sonntag, D.; Zaverkin, V.; Niepert, M. (2024). Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks, Proceedings of the International Conference on Machine Learning (ICML).
2. •     Bib
     Nguyen, D.H.M.*; Le, A.T.*; Nguyen, T.Q.; Nghiem, T.D.; Duong-Tran, D. ; Peters, J.; Li, S.; Niepert, M.; Sonntag, D. (2024). Dude: Dual Distribution-Aware Context Prompt Learning For Large Vision-Language Model, Asian Conference on Machine Learning (ACML).

Generative Models for Imitation Learning and Motion Planning

1. •       Bib
     Carvalho, J.; Le, A.T.; Kicki, P. ; Koert, D.; Peters, J. (2025). Motion Planning Diffusion: Learning and Adapting Robot Motion Planning with Diffusion Models, IEEE Transactions on Robotics (T-Ro), 41, pp.4881-4901.
2. •     Bib
     Nguyen, K.; Le, A. T.; Pham, T.; Manfred, H.; Peters, J.; Vu, M.N. (2025). FlowMP: Learning Motion Fields for Robot Planning with Conditional Flow Matching, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).
3. •       Bib
     Urain, J.*; Le, A.T.*; Lambert, A.*; Chalvatzaki, G.; Boots, B.; Peters, J. (2022). Learning Implicit Priors for Motion Optimization, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).
   •   Bib
     Le, A. T.; Urain, J.; Lambert, A.; Chalvatzaki, G.; Boots, B.; Peters, J. (2022). Learning Implicit Priors for Motion Optimization, RSS 2022 Workshop on Implicit Representations for Robotic Manipulation.
4. •       Bib
     Le, A. T.; Guo M.; Duijkeren, N.; Rozo, L.; Krug, R.; Kupcsik, A.G.; Buerger, M. (2021). Learning forceful manipulation skills from multi-modal human demonstrations, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

Tensor Methods and Geometric Motion Planning

1. •       Bib
     Le, A. T.; Nguyen, K.; Vu, M.N.; Carvalho, J.; Peters, J. (2025). Model Tensor Planning, Transactions on Machine Learning Research (TMLR).
   •   Bib
     Le, A. T.; Nguyen, K.; Vu, M.N.; Carvalho, J.; Peters, J. (2025). Model Tensor Planning, ICRA @ RoboARCH: Robotics Acceleration with Computing Hardware and Systems.
2. •     Bib
     Le, A. T.; Hansel, K.; Carvalho, J.; Watson, J.; Urain, J.; Biess, A.; Chalvatzaki, G.; Peters, J. (2025). Global Tensor Motion Planning, IEEE Robotics and Automation Letters (RA-L), and ICRA 2026 (RA-L Track), 10, 7, pp.7302-7309.
   •   Bib
     Le, A. T.; Hansel, K.; Carvalho, J.; Urain, J.; Biess, A.; Chalvatzaki, G.; Peters, J. (2024). Global Tensor Motion Planning, CoRL 2024 Workshop on Differentiable Optimization Everywhere.
3. •     Bib
     Pompetzki, K.; Le, A. T.; Gruner, T.; Watson, J.; Chalvatzaki, G.; Peters, J. (2025). Geometrically-Aware Goal Inference: Leveraging Motion Planning as Inference, German Robotics Conference (GRC).
   •     Bib
     Pompetzki, K.; Le, A. T.; Gruner, T.; Watson, J.; Chalvatzaki, G.; Peters, J. (2025). Geometrically-Aware Goal Inference: Leveraging Motion Planning as Inference, Multi-disciplinary Conference on Reinforcement Learning and Decision Making (RLDM).

Kinematic Correspondence

1. •     Bib
     Le, A.T.; Pompetzki, K.; Peters, J.; Biess, A. (2025). Kinematics Correspondence As Inexact Graph Matching, German Robotics Conference (GRC).
   •     Bib
     Le, A.T.; Pompetzki, K.; Peters, J.; Biess, A. (2025). Kinematics Correspondence As Inexact Graph Matching, Multi-disciplinary Conference on Reinforcement Learning and Decision Making (RLDM).