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.
The agent is represented by a friendly rat with the objective of catching the cheese. Which will have eight move action options.
The cheese won't move and will always be visible to the Agent.
The Hunter is represented by a cat and will move differently depending on the environment.
pandas >= 1.4.1
numpy >= 1.20.3
gym >= 0.23.0
pygame >= 2.1.2
opencv-python
To install the package into your machine you can use the code bellow
pip install gym-rat-runner
The open environment has 10 by 10 spaces and the cat is sleeping and won't move independently what the rat does.
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:
- The Agent is outside the field vision of the hunter
- The cat will move randomly by the maze.
- 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.