Hi there! I'm Shuo Cheng (成硕), I'm a Ph.D. student at Georgia Tech advised by Prof. Danfei Xu. I'm generally interested in robotics, computer vision, and machine learning, where my current focus is to enable robots with the ability to reason and perform in complex and highly variable environments for achieving long-horizon tasks. I'm open to discussions and collaborations, please feel free to drop me an email if you are interested.
Before studying at Georgia Tech, I'm also very fortunate to work with Prof. Hao Su, Prof. Ravi Ramamoorthi, Prof. Lin Shao, and Prof. Bingbing Ni.
NOD-TAMP: Generalizable Long-Horizon Planning with Neural Object Descriptors
Shuo Cheng, Caelan Garrett*, Ajay Mandlekar*, Danfei Xu
CoRL 2024
CoRL 2023, Workshop on Learning Effective Abstractions for Planning (Oral presentation)
A TAMP-based framework featuring neural object descriptors, capable of learning from only a handful of brief demonstrations yet exhibiting robust performance in long-horizon tasks involving diverse object shapes, poses, and goal configurations.
A Survey of Optimization-based Task and Motion Planning: From Classical To Learning Approaches
Zhigen Zhao, Shuo Cheng, Yan Ding, Ziyi Zhou, Shiqi Zhang, Danfei Xu, Ye Zhao
T-MECH 2024
[pdf]
A comprehensive review on popular TAMP solutions as well as their integrations with learning components. We also discuss challenges and potential future directions.
Neural Field Dynamics Model for Granular Object Piles Manipulation
Shangjie Xue, Shuo Cheng, Pujith Kachana, Danfei Xu
CoRL 2023
ICRA 2023, Workshop on Representing and Manipulating Deformable Objects (Oral presentation, Best Paper Finalist)
A new field-based representation (occupancy density) to model and optimize granular object manipulation.
LEAGUE: Guided Skill Learning and Abstraction for Long-Horizon Manipulation
Shuo Cheng, Danfei Xu
RA-L 2023 (Best Paper Honorable Mention, 1 of 5 over 1200+ accepted papers)
CoRL 2022, Long Horizon Planning Workshop (Oral presentation, Best Paper Finalist)
We use Task and Motion Planning (TAMP) as guidance for learning generalizable and composable sensorimotor skills.
Learning to Regrasp by Learning to Place
Shuo Cheng, Kaichun Mo, Lin Shao
CoRL 2021
We propose a point-cloud-based system for robots to transform the initial grasp pose to the desired grasp pose, where stable object placements serve as intermediate waypoints. A challenging synthetic dataset is introduced for learning and evaluating the proposed approach.
Deep Stereo using Adaptive Thin Volume Representation with Uncertainty Awareness
Shuo Cheng*, Zexiang Xu*, Shilin Zhu, Zhuwen Li, Li Erran Li, Ravi Ramamoorthi, Hao Su (*equal contribution)
CVPR 2020 (Oral presentation, top 5.7% accepted papers)
We design a novel multi-stage framework that progressively sub-divides the vast scene space with increasing depth resolution and precision, which enables scene reconstruction with high completeness and accuracy in a coarse-to-fine fashion.
Fine-grained Video Captioning for Sports Narrative
Huanyu Yu*, Shuo Cheng*, Bingbing Ni*, Minsi Wang, Jian Zhang, Xiaokang Yang (*equal contribution)
CVPR 2018 (Spotlight presentation, top 6.7% accepted papers)
[pdf]
We develop a well-modularized pipeline for automatic sports game narration. We incorporate human pose and optical flow to depict the movements of each player, which facilitates action recognition and human interaction learning.
Service
Reviewer: RSS, CVPR, ICML, NeurIPS, RA-L, CoRL, ICLR, T-PAMI, ICCV, ECCV, T-ASE, ICRA, IROS, T-MM, AAAI, Neurocomputing, CIKM, WACV, ACCV, T-CSVT
Co‐organizer: Learning for Task and Motion Planning Workshop, RSS 2023