Festivus

policy

HULC

University of Freiburg · PyTorch

or hover any field below to flag it

Overview

Name
HULC
Author
University of Freiburg
Framework
PyTorch
License
mit
Skill type
manipulation
Evidence level
reported
Task description
Hierarchical Universal Language Conditioned policy for long-horizon manipulation. Multimodal transformer encoder organizes behaviors into categorical latent plans, local policies execute low-level actions. Learns a single 7-DoF policy for sequential manipulation tasks from language instructions. CALVIN benchmark baseline.

Spaces

Action space
joint-position · 7-dim · 30Hz
Observation space
  • type: multimodal
  • · static_rgb
  • · gripper_rgb
  • · joint_pos
  • · language_instruction

Links

HuggingFace repo
null
Paper (arXiv)
https://arxiv.org/abs/2204.06252

Compatible robots

0+1 mentioned but not in catalog yet

No robots list HULC as compatible yet. Know of one? Flag it above.

Compatible environments

1
tabletop-cleannot in seed

Datasets that reference this policy

0

No datasets reference HULC yet.