All Tools
R
AgentsFreeOpen Source
ROBOMIMIC
Robot learning from demonstration with shared datasets
MIT
ABOUT
Robot imitation papers used incompatible datasets and algorithm code, so results were hard to reproduce. robomimic packages demonstration datasets with BC, offline RL, and Diffusion Policy so researchers can train visuomotor policies on the same lift, can, and square tasks.
INSTALL
pip install robomimicINTEGRATION GUIDE
1. Train a Diffusion Policy or BC-Transformer on robomimic manipulation datasets
2. Benchmark offline RL algorithms like IQL or CQL on shared robot demos
3. Fine-tune a language-conditioned robot policy across multiple datasets
4. Reproduce published robomimic study results on simulated and real robot tasks
TAGS
pythonroboticsimitation-learningoffline-rlrobosuitepytorch