Festivus

policy

Robomimic BC-RNN

Stanford / UT Austin (ARISE) · PyTorch

or hover any field below to flag it

Overview

Name
Robomimic BC-RNN
Author
Stanford / UT Austin (ARISE)
Framework
PyTorch
License
mit
Skill type
manipulation
Evidence level
verified
Task description
Behavioral cloning with recurrent neural network (LSTM) backbone for sequential manipulation. Processes observation histories to model temporal correlations. Optional GMM action head for multi-modal distributions. Key baseline in the Robomimic benchmark suite, consistently outperforms non-recurrent BC on human demos.

Spaces

Action space
end-effector-pose · 7-dim · 20Hz
Observation space
  • type: multimodal
  • · agentview_rgb (84x84)
  • · wrist_rgb (84x84)
  • · joint_pos

Links

HuggingFace repo
null
Paper (arXiv)
https://arxiv.org/abs/2108.03298

Compatible robots

0+1 mentioned but not in catalog yet

No robots list Robomimic BC-RNN as compatible yet. Know of one? Flag it above.

Compatible environments

2
tabletop-cleannot in seedrobomimic-cannot in seed

Datasets that reference this policy

0

No datasets reference Robomimic BC-RNN yet.