ajaysri

LEGO flat final-goal policy

200 robot demonstrations (100 teleoperated + 100 correction) in simulation with 79,956 frames at 20 Hz for learning assembly tasks from visual and textual goal specifications.

Downloads0
Episodes200

Why This Matters for Physical AI

This dataset enables research into learning assembly tasks from multimodal goal specifications (images and language), demonstrating how robots can be trained to follow structured final-goal representations rather than step-by-step instructions.

Technical Profile

Modalities
rgbproprioceptionlanguage
Robot Embodiments
robot
Action Space
joint_positions
Environment
simulation
Task Types
assemblymanipulation
Episodes
200
Data Format
LeRobot
Annotation Types
language_instructionsgoal_imagesaction_labels
Part of the LeRobot family

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