Click for more!

Adam Hung

Email: adamhung@andrew.cmu.edu
Google Scholar
LinkedIn
I am a first year Robotics PhD student at Carnegie Mellon University advised by Jeffrey Ichnowski. My interests include learning from human videos and manipulation with multi-fingered hands. I received a Bachelors in Robotics Engineering from the University of Michigan in 2025. I also worked on reinforcement learning at Symbotic as an intern. My research is supported by the NSF Graduate Research Fellowship Program.
Carnegie Mellon Robotics Institute logo University of Michigan Robotics logo National Science Foundation logo

Publications (*indicates equal contribution)

3PoinTr: 3D Point Tracks for Learning Manipulation from Unconstrained Human Videos
Adam Hung, Bardienus P. Duisterhof, Jeffrey Ichnowski
Under Review
[Website] [Paper] [Video]
3PoinTr learns manipulation from unconstrained human videos: videos where the human demonstrator can act freely rather than mimicking target robot kinematics. 3PoinTr first predicts dense 3D point tracks, and then conditions a closed-loop multitask policy on these tracks.
PointZero: 3D Point Track Completion for Dynamics and Robot Manipulation
Bardienus P. Duisterhof, Kaifeng Zhang, Adam Hung, Bowen Wen, Stan Birchfield, Yunzhu Li, Deva Ramanan, Jeffrey Ichnowski
Under Review
PointZero proposes 3D point track completion as a pretraining objective: given an RGB-D image and a few sparse point trajectories, it predicts future dense 3D motion. Pretrained on our custom synthetic dataset, PointZero learns transferable 3D dynamics priors that improve real-world action-conditioned dynamics prediction and robot manipulation after fine-tuning.
RoPotter: Toward Robotic Pottery and Deformable Object Manipulation with Structural Priors
Uksang Yoo*, Adam Hung*, Jonathan Francis, Jean Oh, Jeffrey Ichnowski
IEEE International Conference on Humanoid Robots (Humanoids), 2024
Best Oral Presentation Finalist Award
[Website] [Paper] [Video]
RoPotter learns to create bowls on a pottery wheel by using structural priors to inform a continuously deforming mesh representation of the clay.
AVO: Amortized Value Optimization for Contact Mode Switching in Multi-Finger Manipulation
Adam Hung, Fan Yang*, Abhinav Kumar*, Sergio Aguilera Marinovic, Soshi Iba, Rana Soltani Zarrin, Dmitry Berenson
arXiv, 2025
[Paper] [Video]
Long-horizon trajectory optimization can be made tractable by decomposing the problem into contact-mode-specific subtasks. AVO bridges these subtasks with a learned value-function ensemble that guides each subtask toward states that are advantageous for future subtasks, and simultaneously reduces online computation.
MapExRL: Learning Efficient Indoor Mapping and Exploration using Predicted Global Map Context
Narek Harutyunyan, Brady Moon, Seungchan Kim, Adam Hung, Cherie Ho, Sebastian Scherer
International Conference on Advanced Robotics (ICAR), 2025
[Website] [Paper] [Video]
MapExRL uses global indoor map predictions and a budget-aware RL policy to exploit structure in indoor environments, outperforming greedy algorithms and other prior methods on long-horizon indoor exploration.
SKOOTR: A SKating, Omni-Oriented, Tripedal Robot
Adam Hung, Challen Enninful Adu, Talia Moore
IEEE International Conference on Robotics and Automation (ICRA), 2025
[Website] [Paper] [Video]
SKOOTR is a brittlestar-inspired tripedal robot that explores the advantages of radially symmetric form-factors. SKOOTR is highly stable and maneuverable, and enables several novel forms of locomotion.