Welcome to the third Udacity reinforcement Project.
In this Project we train two agents to play tennis.
👇🏼These are the resulting agents
This project contains a solution to the third project of the Udacity Deep Reinforcement Learning Course. This Project uses a multi agent DDPG Algorithm to train the agent.
If an agent hits the ball over the net, it receives a reward of 0.1, if however the ball is dropped or is thrown out of bounds it receives an reward of -0.01.
The task is episodic, and in order to solve the environment, the agent must get an average score of +0.5 over 100 consecutive episodes.
The state space consists of 8 dimensions and is continuous. The observations contain velocity and position of the ball and the racket.
The action space consists of 2 continuous actions, which control the racket. These correspond to a vertical and horizontal movement.
The Agents are trained using a DDPG algorithm with a shared replay buffer.
For further information on training please read the Report.md.
###Prerequisites
Python 3.6
Unity
Conda
##Installation:
- Clone the repository
https://github.com/schmiJo/p3_collabl-compet
- Install Jupyter Notebook
pip install jupyter
- Create and activate a new environment for Python 3.6
- Linux or Mac
conda create --name drlnd python=3.6
source activate drlnd
- Windows
conda create --name drlnd python=3.6
activate drlnd
- Install several dependencies
pip install -r requirements.txt
- Before running the Tennis.ipynb change the kernel to match the drlnd environment by using the drop down Kernel menu.
Download the unity environment using the following link for macOs:
https://s3-us-west-1.amazonaws.com/udacity-drlnd/P3/Tennis/Tennis.app.zip
More instructions for the installation can be found under:
https://github.com/udacity/deep-reinforcement-learning#dependencies