Radar-Camera Fusion via Representation Learning in Autonomous Driving
X. Dong, B. Zhuang, Y. Mao, L. Liu
CVPR-W · 2021 · Autonomous driving
Xu "Joe" Dong · 董旭
Real-world AI is won by systems, not models.
Founder & CTO of Chestnut Robotics (previously TetherIA), building dexterous robots for the real world. Tech lead for the planning model behind XPeng's production end-to-end driving; planning foundation models at Waymo. Physicist by training, computer vision by trade. Author of Physical AI Deep Dives.
Building — Chestnut Robotics
The data-driven dexterous hand, and the pipeline that trains it.
We build the Aero Hand and the data engine behind it — egocentric, high-fidelity capture that turns human demonstrations into robot skill. Aero Hand (18-DOF, sub-mm) · Aero UMI (capture exoskeleton) · GAIN3D (3D reconstruction) · a robotic foundation model on top.
Selected publications
Physics → computer vision → autonomous driving → physical AI
Selected work, from medical-image reconstruction to radar-camera perception; the full record is on Google Scholar.
Probabilistic Oriented Object Detection in Automotive Radar
X. Dong, P. Wang, P. Zhang, L. Liu
CVPR-W · 2020 · Autonomous driving
A Sparse-View CT Reconstruction Method Based on Combination of DenseNet and Deconvolution
Z. Zhang, X. Liang, X. Dong, Y. Xie, G. Cao
IEEE TMI · 2018 · Medical imaging
Writing — Physical AI Deep Dives
I think in public about hard problems in physical AI.
- System Thinking and Model Thinking in Robot Learning2026 · essay
- The right hand is the one you can capture2026 · essay
- Robot Learning Should Be Goal-Driven and Data-First2026 · essay
The most useful split in robot learning isn't VLA vs. world model — it's system thinking vs. model thinking.Subscribe →
About
A researcher who builds in production.
I trained as a physicist (USTC, 2010–14), then moved into computer vision for my PhD, working on medical image analysis (Virginia Tech, 2014–19). From 2019 to 2025 I carried vision into autonomous driving through three of the field's paradigm shifts — monocular 3D detection, learning-based planning, then end-to-end foundation models, the last shipped to production at XPeng. In 2025 I cofounded Chestnut Robotics to do the same for robots. Every move has followed one read — a field AI had just cracked open.
Focus — dexterous manipulation, data engines, foundation models for the physical world.
Belief — real-world AI is a systems problem before it's a model problem.
Now — the Aero Hand, and teaching robots to use it.