MOOS-IvP Agent
Overview
Model agnostic machine learning tooling for MOOS-IvP. Inspired by moos-ivp-pLearn this project was designed to allow for training of python ML agents for Project Aquaticus while also providing generalized software to use with any MOOS-IvP project.
Note
This project is still very young. As such feel free to any ask questions not answered by the documentation via a github issues. There, you can also see the current backlog.
Below you can see an example MOOS-IvP mission AgentAquaticus based on Project Aquaticus. The agents being trained are rewarded by successfully grabbing and returning a flag protected by robots which circle it. The agent's actions are being determined by q-table written in NumPy.
Installation
This project currently supports both OSX and Linux operating systems. See the main README.
Pretty Please
PLEASE Raise a GitHub issue if there is ~any~ issue with these instructions.
Getting Started
See the getting started page to start you on your ML path.