DAVIAN-Robotics2026apache-2.0

RoboLab-BananaInBowl-hard-base-stoch-260802

RL rollouts from a RoboLab (Isaac Lab) BananaInBowl policy under the hard initial-pose distribution, collected via stochastic sampling from a trained RFCL agent for pi0.5 delta-action supervised fine-tuning.

Downloads22
Episodes5941

Why This Matters for Physical AI

Provides large-scale RL-generated manipulation trajectory data with standardized recording conventions and quantile statistics to support reproducible policy learning and imitation learning research.

Technical Profile

Modalities
rgbproprioception
Robot Embodiments
DROID
Action Space
joint_positions
Environment
simulation
Task Types
manipulationpick_and_place
Episodes
5941
Data Format
LeRobot
Annotation Types
reward_labels
License
apache-2.0
Part of the RoboLab 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