zrgongmit

RMBench++ Put Back Block Relocate v1

A pilot dataset containing 100 scripted-expert demonstrations for the put_back_block_relocate task, where a robot must pick up a block, press a button, observe feedback, and place the block on the correct mat based on memory and visual signals.

Downloads0
Episodes100

Why This Matters for Physical AI

This dataset advances physical AI by providing demonstrations for temporally-reasoned manipulation tasks that require memory and observation of environmental feedback, essential for training robots that must adapt behavior based on dynamic scene changes.

Technical Profile

Modalities
rgbproprioception
Robot Embodiments
humanoid
Action Space
joint_positions
Environment
lab
Task Types
manipulationgraspingpick_and_placetemporal_reasoningmemory
Episodes
100
Data Format
LeRobot
Annotation Types
language_instructionsmemory_keyframesaction_labels
License
mit
Part of the RMBench++ family

Access

Need custom rgb data?

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

Request a Sample Pack

Related Datasets