tristanbester / options

Implementation of the options framework in the four-rooms domain.

Options Framework for the Four-Rooms Domain

Ovierview

• This repository contains the code used to train and evaluate the options framework on the four-rooms domain. The options framework is a hierarchical reinforcement learning framework that allows agents to learn temporally extended actions, called options, in addition to primitive actions.
• The agent is trained and evaluated on the four-rooms domain. The four-rooms domain is a simple grid-world environment that consists of four rooms connected by narrow corridors. The agent is tasked with navigating to a goal location in the environment. The agent is provided with a set of primitive actions that allow it to move in the four cardinal directions.
• The agent is also provided with a set of options that allow it to move to the goal location in each of the four rooms. The agent is rewarded for reaching the goal location.

Prerequisites

• Python: 3.11+
• Poetry:

curl -sSL https://install.python-poetry.org | python3 -

Getting Started

Clone the repository:

git clone https://github.com/TristanBester/options.git
cd options

Install dependencies:

poetry install

Run the simulation:

cd src && poetry run main.py