COACH-gym

A python implementation of the COACH algorithm for the Cartpole problem in OpenAI gym.

This code is based on the following publications:

1. Interactive learning of continuous actions from corrective advice communicated by humans
2. An Interactive Framework for Learning Continuous Actions Policies Based on Corrective Feedback

Link to paper video:

Installation

To use the code, it is necessary to first install the gym toolkit: https://github.com/openai/gym

Then, the files in the gym folder of this repository should be replaced/added in the installed gym folder on your PC.

Requirements

• NumPy
• PyGame

Usage

To run the code just type in the terminal inside the folder COACH-gym:

python main.py

Along with the rendered environment, a small black window should appear when running the code. To be able to give feedback to the agent, this window must be selected/clicked with the computer mouse.

Comments

The COACH algorithm is designed to work with problems of continuous action spaces. Given that the Cartpole environment of gym was designed to work with discret action spaces, a modified continuous version of this environment is used.

This code has been tested in Ubuntu 16.04 and python >= 3.5.