Seoul National University

Humanoid and physical AI, built and tested by students

SHAPE builds service robots that work in ordinary human spaces, and takes them to public evaluation so the results are checked by someone other than us.

2026
Founded
70
Members
3
Tracks
2
Competitions in 2026–27

What we work on

SHAPE (SNU Humanoid And Physical Embodied AI) is a student research team at Seoul National University. We study how learning systems can perceive, reason about and act in the physical world, and we build the robots that put those systems to the test.

The team runs three tracks in parallel. An education track takes new members from robot kinematics and URDF through imitation and reinforcement learning to recent vision-language-action models. A research track, SHAPE-UP, groups members into small teams that pick their own problem and publish a project page for it each semester. A competition track takes the resulting system to public evaluation, so that our claims are checked by someone other than us.

We are affiliated with the Department of Mechanical Engineering and work out of Room 109, Building 301, on the Gwanak campus.

Research areas

Six areas the team works across. Each SHAPE-UP project sits in one or more of them.

Perception and state estimation

Computer vision · 3D perception · SLAM · tactile and force sensing

Robot learning and generalist models

Vision-language-action · imitation learning · reinforcement learning · sim-to-real · world models

Reasoning and action planning

Task planning · motion planning · failure detection and recovery

Manipulation

Dexterous manipulation · deformable objects · contact-rich tasks · compliant control

Humanoids and control

Whole-body control · loco-manipulation · locomotion and balance · motion retargeting

People, data and hardware

Human-robot interaction · teleoperation · data pipelines · hardware fabrication

Where to look next

The material a reviewer usually needs, in the order they usually need it.