Quick Start Guide

Quick-Start Guide

This guide should get you up and running with the base rover domain code.

Setup

A requirements.txt is provided, which contains all the necessary dependencies: pip install -r requirements.txt generate these requirements

Running

You can run any of the main experiment files directly through Python 3.X: python experiment_1.py Example experiment files can be found in the baselines directory.

Writing Experiments

We recommend copying one of the baseline experiments and working off that as a template. The experiments are defined through a series of function hooks, which pass around a reference to a single dictionary containing the current state of the world. By writing the order in which functions are called, you define the behavior of the environment and the learning of the agents.

Gym Interface

If you are more interested in testing learning algorithms, an OpenAI Gym interface is provided. It provides an interface for the "standard" rover domain problem, with rewards of G or D. As this is a multiagent domain, the joint action, and joint observation space and joint reward space are provided back in the Gym interface functions.