Academic Homepage

Alan D. Liu

PhD Candidate  ·  Autonomous Systems & Perception Lab
ETH Zürich

About

Investigating the intersection of 3D scene understanding and dexterous robotic manipulation. Focused on building systems that generalise to novel objects and cluttered workspaces without dense human supervision.

My work bridges geometry-aware perception and contact-rich control — asking how a robot can form an accurate model of the world quickly enough to act on it gracefully.

Research Interests

My current focus spans four interconnected areas:

3D Scene Reconstruction Dexterous Manipulation Object Pose Estimation Sim-to-Real Transfer

I am particularly interested in how sparse, geometry-aware representations can enable robust grasping and placement in scenes the robot has never encountered — closing the gap between controlled lab settings and real deployment without exhaustive re-training.

Current Project

2024 — Present  ·  Autonomous Systems & Perception Lab, ETH Zürich

Category-Level Pose Estimation for Deformable Object Manipulation

An investigation into neural implicit shape representations for estimating the 6-DoF pose of previously unseen object instances at category level. The goal is a manipulation pipeline that requires no CAD model and fewer than ten reference views to achieve reliable pick-and-place in clutter.

ROS 2 NeRF / 3DGS PyTorch C++ RGBD Sensing

Academic Resources

A Guiding Thought

“The scientist does not study nature because it is useful to do so. He studies it because he takes pleasure in it, and he takes pleasure in it because it is beautiful. If nature were not beautiful it would not be worth knowing, and life would not be worth living.”

— Henri Poincaré, Science and Method, 1908

Tools & Philosophy

Good tools lower the cost of good thinking. My stack is deliberately small.

Obsidian Networked knowledge & evergreen notes
Python & C++ Prototyping & performance-critical systems
ROS 2 Robot middleware & sensor integration
The 12-Week Year Execution rhythm over annual planning

Contact

I am open to research collaborations, conversations about robotic manipulation, or a well-placed recommendation for a good monograph.

Reach me at hello@oneoneliu.lol.