bobiblazeski / tennis

Two agents control rackets to bounce a ball. Project 3 in Udacity Deep Reinforcement Learning Nanodegree.

Tennis

Collaboration and Competition project from Udacity Deep Reinforcement Learning Nanodegree.

Description

In this environment, two agents control rackets to bounce a ball over a net. If an agent hits the ball over the net, it receives a reward of +0.1. If an agent lets a ball hit the ground or hits the ball out of bounds, it receives a reward of -0.01. Thus, the goal of each agent is to keep the ball in play.

The observation space consists of 8 variables corresponding to the position and velocity of the ball and racket. Each agent receives its own, local observation. Two continuous actions are available, corresponding to movement toward (or away from) the net, and jumping.

The task is episodic, and in order to solve the environment, your agents must get an average score of +0.5 (over 100 consecutive episodes, after taking the maximum over both agents). Specifically,

After each episode, we add up the rewards that each agent received (without discounting), to get a score for each agent. This yields 2 (potentially different) scores. We then take the maximum of these 2 scores. This yields a single score for each episode. The environment is considered solved, when the average (over 100 episodes) of those scores is at least +0.5.

Installation

Install deep reinforcement learning repository

1. Clone deep reinforcement learning repository
2. Fallow the instructions to install necessary dependencies

Download the Unity Environment

1. Download environment for your system into this repository root

• Linux: click here

• Mac OSX: click here

• Windows (32-bit): click here

• Windows (64-bit): click here

• Headless: click here

2. Unzip (or decompress) the archive

Run the project

1. Start the jupyter server
2. Open the Tennis.ipynb notebook
3. Change the kernel to drlnd
4. You should be able to run all the cells

Weights

The directory saves contains saved weights for 2 different agents:

• 1039_96_96_actor.pth & 1039_96_96_critic.pth - Agent that learned in 1039 episodes
• 1411_256_256_actor.pth & 1411_256_256_critic.pth - Agent that learned in 1411 episodes

Naming convention EpisodesFully connected layer 1Fully connected layer 2_[actor|critic].pth

Credits

Most of the code is based on the Udacity code for DDPG. I've adapted some of the code by akhiadber, which adds batch normalization & training function.