We would like to thank members of the Stanford PAIR Group for their support and feedback on this project. These people in particular have made the following contributions at different stages of this project:
Rohun Kulkarni (assistance with collecting real robot datasets and running real robot experiments)
Jim Fan (providing support for running experiments on lab clusters)
Our Config class (see
config/config.py) was adapted from addict.
models/distributions.pywas adapted from rlkit.
Support for training distributional critics (see
algos/bcq.py) was adapted from Acme. It also served as a useful reference for implementing Gaussian Mixture Model (GMM) policies.
We wholeheartedly welcome the community to contribute to our project through issues and pull requests. New contributors will be added to the list above.