nvidia2025cc-by-4.0

Arena-GR1-Manipulation-Task

A multimodal dataset of 50 trajectories for humanoid (GR1) manipulation tasks in Isaac Lab, including 10 human teleoperated demonstrations and 40 synthetically generated demonstrations using MimicGen. The dataset provides state, vision, language, and action data for training generalist robot policies for microwave opening tasks.

Downloads124
Episodes50
Likes2

Why This Matters for Physical AI

This dataset enables training of generalist robot manipulation policies through behavior cloning and supports sim-to-real transfer research for humanoid robot control in complex manipulation tasks.

Technical Profile

Modalities
rgbproprioceptionlanguage
Robot Embodiments
humanoidGR1
Action Space
joint_positions
Environment
simulation
Task Types
manipulationopening microwave
Episodes
50
Data Format
HDF5, LeRobot
Annotation Types
language_instructionsaction_labels
License
cc-by-4.0
Part of the Arena-GR1-Manipulation-Task family

Community Signals

Access

Need custom rgb data?

Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets