Code implementation for the RA-L and ICRA 21 paper 'Learning from Imperfect Demonstrations from Agents with Varying Dynamics' Arxiv. (Under developping)

Installation

pip install -r requirement.txt

cd imperfect_envs

pip ininstall -e .

Demonstrations

Download demonstrations here

Extract the demonstrations here. tar -zxf codes.tar.gz

Training

Traning commands are in train.sh. We annotate the experiment name as a comment before the command.

Testing

Acknowledgement

The TRPO part is hugely based on: https://github.com/ikostrikov/pytorch-trpo