This repository contains code to accompany the paper "Switched Soft Actor Critic For the Acrobot Swingup and Balance Task"
This is a python package, to use it please install it into your python environment (highly recommend some sort of virtual environment) with
pip install -e ssac/
You can then install the requirements with
$ pip install -r requirements.txt
- rl-baselines-zoo/ contains a fork of https://github.com/araffin/rl-baselines-zoo, we modified the code to use our custom acrobot environment, and added a script to launch the algorithms we wanted to compare with their specific hyper parameters.
- switched_rl/gate_generation.ipynb/py notebook/script walking through how we train the gating function
- switched_rl/ssac.py This contains the implementation of SSAC as described in the paper
- switched_rl/make_figures.py This is the script that actually generates the images found in the paper (stored in figs/)
- switched_rl/run_ssac.py This will launch SAC with the parameters specified in the paper