Gym-Rat-Runner

Introduction

Gym Rat Runner is an open-source environment implemented over OpenAI Gym.It is a 2-D discrete environment intended for Reinforcement learning applications with collision treatments. For now, the package has two distinct environments; the open and the maze environments.

Agent

The agent is represented by a friendly rat with the objective of catching the cheese. Which will have eight move action options.

Target

The cheese won't move and will always be visible to the Agent.

Hunter

The Hunter is represented by a cat and will move differently depending on the environment.

Installation

Requirements

pandas >= 1.4.1
numpy >= 1.20.3
gym >= 0.23.0
pygame >= 2.1.2
opencv-python

pip

To install the package into your machine you can use the code bellow

  
pip install gym-rat-runner  

Environments

Open Environment

The open environment has 10 by 10 spaces and the cat is sleeping and won't move independently what the rat does.

Maze Environment

The maze environment has 16 by 32 spaces based on the Bank Heist from Atari. Like the police in Bank Heist, the hunter won't be able to move back but it will have two options of movement:

1. The Agent is outside the field vision of the hunter

• The cat will move randomly by the maze.

2. The agent is inside the field vision of the hunter.

• The cat will start to run after the rat and will move twice the movement of the agent.