Weikang Wan /ˈweɪˌkɑːŋ ; wɑːn/

PhD Student, UC San Diego

I am a Ph.D. student in CSE at UC San Diego, co-advised by Prof. Hao Su and Prof. Henrik I. Christensen. Previously, I received my B.S. in Computer Science (with honors) from Peking University. I have also spent time as a research intern at NVIDIA GEAR and NVIDIA Seattle Robotics Lab, and as a visiting student at CMU and UT Austin.

I'm interested in building embodied agents that can perceive, think, reason, and act like humans in open-world environments. I develop algorithms that facilitate robot learning via visual imitation, world models, simulation, and RL, and build scalable, continually learning systems for physical agents.

📢 Feel free to reach out if you'd like to discuss or collaborate!

Publications ( show all by date / show selected )

*/† denotes equal contribution.

Experience

Jun 2025 – Sep 2025

NVIDIA Seattle Robotics Lab

Research Intern · Seattle, WA

Advised by Dr. Caelan Garrett, Dr. Xuning Yang, and Prof. Fabio Ramos. Developed learning-based bimanual manipulation systems for long-horizon, contact-rich tasks via planning and reinforcement learning.

Sep 2024 – Present

UC San Diego

Research Assistant · La Jolla, CA

Co-advised by Prof. Hao Su and Prof. Henrik I. Christensen. Developing policy- and world-model-learning algorithms for contact-rich real-world manipulation.

Jan 2024 – Aug 2024

NVIDIA GEAR Lab

Research Intern · Santa Clara, CA

Advised by Dr. Jim Fan and Prof. Yuke Zhu. Built a large-scale automated synthetic-data-generation pipeline for bimanual dexterous manipulation.

Jun 2023 – Dec 2023

UT Austin, RPL Lab

Visiting Research Assistant · Austin, TX

Advised by Prof. Yuke Zhu. Developed a continual-learning algorithm for lifelong, long-horizon robot manipulation via unsupervised skill discovery.

Dec 2022 – Jul 2023

Carnegie Mellon University, Robotics Institute

Visiting Research Assistant · Pittsburgh, PA

Co-advised by Prof. Zackory Erickson and Prof. David Held. Designed an end-to-end unified differentiable trajectory-optimization algorithm for deep reinforcement and imitation learning.

May 2021 – May 2024

Peking University

Research Assistant · Beijing, China

Advised by Prof. He Wang. Built a learning-based generalizable dexterous-grasping system using generative models and large-scale reinforcement learning; designed vision-language-action (VLA) models with spatial intelligence.

Weikang Wan /ˈweɪˌkɑːŋ ; wɑːn/

Publications ( show all by date / show selected )

Learning to Plan & Schedule with Reinforcement-Learned Bimanual Robot Skills

LodeStar: Long-horizon Dexterity via Synthetic Data Augmentation from Human Demonstrations

RoboVerse: Towards a Unified Platform, Dataset and Benchmark for Scalable and Generalizable Robot Learning

NaVid-4D: Unleashing Spatial Intelligence in Egocentric RGB-D Videos for Vision-and-Language Navigation

DexMimicGen: Automated Data Generation for Bimanual Dexterous Manipulation via Imitation Learning

DiffTORI: Differentiable Trajectory Optimization for Deep Reinforcement and Imitation Learning

LOTUS: Continual Imitation Learning for Robot Manipulation Through Unsupervised Skill Discovery

UniDexGrasp++: Improving Dexterous Grasping Policy Learning via Geometry-aware Curriculum and Iterative Generalist-Specialist Learning

UniDexGrasp: Universal Robotic Dexterous Grasping via Learning Diverse Proposal Generation and Goal-Conditioned Policy

Learning Category-Level Generalizable Object Manipulation Policy via Generative Adversarial Self-Imitation Learning from Demonstrations

HOI4D: A 4D Egocentric Dataset for Category-Level Human-Object Interaction

Experience