sgillen / policy_refinement

Direct Random Search for Fine Tuning of Deep Reinforcement Learning Policies

This Repo contains code to accompany the paper "Direct Random Search for Fine Tuning of Deep Reinforcement Learning Policies" submitted to RA-L/ICRA 2022

It contains one script to run the experiments used in then paper, and several notebooks to load and process data from those experiments. Some of the code (most notably the random search implementation we use) is contained in a library of mine hosted here.

Installation:

First, clone and install seagul:

git clone https://github.com/sgillen/seagul
pip install ./seagul

You will also need to install the rl baselines 3 zoo, to access the baseline agents

git clone --recursive https://github.com/DLR-RM/rl-baselines3-zoo

If you want to use the panda environments you will also need to clone the latest from here:

git clone https://github.com/qgallouedec/panda-gym
pip install ./panda_gym

Finally clone this repo and install the remaining requirements

git clone https://github.com/sgillen/policy_refinement/
cd policy_refinment/
pip install -r requirements.txt

You should now be able to train new agents, load the data used in the paper, and run baseline agents. To access our own agents without re running the training, see the keep_agents/ folder, which has a link to download the trained agents at.