This repository includes several reinforcement learning algorithms for parameterised action space MDPs:
- PA-DDPG [Hausknecht & Stone 2016]
All implementations of algorithms are based on
Stable-baselines3 (https://github.com/DLR-RM/stable-baselines3). The plot.py
borrows some code from Spinning Up (https://github.com/openai/spinningup)
Experiment scripts are provided to run each algorithm on the following domains with parameterised actions:
- Platform (https://github.com/cycraig/gym-platform)
- Robot Soccer Goal (https://github.com/cycraig/gym-goal)
- Half Field Offense (https://github.com/cycraig/gym-soccer)