OpenAI Gyms

This repository contains experiments on Reinforcement Learning using OpenAI Gyms.

Environment Setup

1. Setup a Python 3 virtual environment

   python3 -m venv openai
   source openai/bin/activate

2. Install dependencies, including transitive dependencies, using requirements.txt

   pip install -r requirements.txt

Cartpole Agent

The Cartpole problem is a classic control problem, also known as Inverted Pendulum, where the centre of mass lies above the pivot point and the pole (or the pendulum) has to be balanced by moving the pivot under the centre of mass around. Imagine humans trying to balance a stick on their palm.

Cartpole-v0 OpenAI Gym provides us a simulation of this problem. This environment is used in Cartpole Agent to train a Q-learning agent that learns to balance the cartpole after few runs. Read more about the agent here.