Project 3. Collaboration and Competition

Project Details

This is my solution to Collaboration and Competition Project of Udacity Deep Reinforcement Learning course. Original project template is available at https://github.com/udacity/deep-reinforcement-learning/tree/master/p3_collab-compet

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, the agents must get an average score of +0.5 (over 100 consecutive episodes, after taking the maximum over both agents). Specifically,

• After each episode, add up the rewards that each agent received (without discounting), to get a score for each agent. This yields 2 (potentially different) scores. 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.

Besides README.md, this repository holds of the following files:

• Report.md provides a description of the implementation
• test.py is the main file for testing
• train.py is the main file for training
• actor.pth is the Actor neural network trained parameters
• agent.py implements an agent for training and testing
• env_agent_factory.py creates an environment and its agent
• neural_nets.py creates neural networks for an Actor and a Critic.
• replay_buffer.py implements a Replay Buffer
• *_test.py unit tests of corresponding modules

All the Python code is pylint-compliant.

Getting Started

Follow the steps, described in https://github.com/udacity/deep-reinforcement-learning/tree/dc65050c8f47b365560a30a112fb84f762005c6b README.md, Dependencies section, to deploy your development environment for this project.

Basically, you will need:

• Python 3.6
• PyTorch 0.4.0
• Numpy and Matplotlib, compatible with PyTorch
• Unity ML Agents. Udacity Navigation Project requires its own version of this environment, available https://github.com/udacity/deep-reinforcement-learning/tree/dc65050c8f47b365560a30a112fb84f762005c6b/python with references to other libraries

The project has been developed and tested on Mac OS Catalina with a CPU version of PyTorch 0.4.0.

Instructions

1. Download the project to your PC
2. Open environment.py in your text editor and set a correct path to Tennis simulator in ENV_PATH variable
3. Open your terminal, cd to the project folder
4. Run test.py to test previously trained agent over 100 episodes
5. Run train.py to retrain the agent
6. Look through Report.md of this repository to learn further details about my solution