UDACITY Project 2: Continuous Control

Introduction

For this project, you will work with the Reacher environment.

In this environment, a double-jointed arm can move to target locations. A reward of +0.1 is provided for each step that the agent's hand is in the goal location. Thus, the goal of your agent is to maintain its position at the target location for as many time steps as possible.

The observation space consists of 33 variables corresponding to position, rotation, velocity, and angular velocities of the arm. Each action is a vector with four numbers, corresponding to torque applicable to two joints. Every entry in the action vector should be a number between -1 and 1.

Distributed Training

For this project, we will provide you with two separate versions of the Unity environment:

• The first version contains a single agent.
• The second version contains 20 identical agents, each with its own copy of the environment.

In this repository we solve both versions.

Solving the Environment

Solve the First Version

The task is episodic, and in order to solve the environment, your agent must get an average score of +30 over 100 consecutive episodes.

Solve the Second Version

The barrier for solving the second version of the environment is slightly different, to take into account the presence of many agents. In particular, your agents must get an average score of +30 (over 100 consecutive episodes, and over all 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 20 (potentially different) scores. We then take the average of these 20 scores.
• This yields an average score for each episode (where the average is over all 20 agents).

The environment is considered solved, when the average (over 100 episodes) of those average scores is at least +30.

Getting Started

1. Requirements: To replicate the project on your computer you need to create a conda environment using python 3.6 and activate it. This can be done as follows for Linux or Mac. (For windows, follow Udacity's instructions)

$ conda create --name drlnd python=3.6
$ source activate drlnd
(drlnd) $

Then in your environment, install OpenAI's gym as follows:

(drlnd) $ git clone https://github.com/openai/gym.git
(drlnd) $ cd gym
(drlnd) $ pip install -e .

After installing gym, you must install Udacity's required python packages as follows:

(drlnd) $ git clone https://github.com/udacity/deep-reinforcement-learning.git
(drlnd) $ cd deep-reinforcement-learning/python
(drlnd) $ pip install .

Then you create a Jupiter notebook kernel that can run the Unity environment provided by Udacity as follows:

(drlnd) $ python -m ipykernel install --user --name droned --display-name "drlnd"

2. Download the environment from one of the links below. You need only select the environment that matches your operating system:

   • Version 1: One (1) Agent

     • Linux: click here
     • Mac OSX: click here
     • Windows (32-bit): click here
     • Windows (64-bit): click here

   • Version 2: Twenty (20) Agents

     • Linux: click here
     • Mac OSX: click here
     • Windows (32-bit): click here
     • Windows (64-bit): click here

   (For Windows users) Check out this link if you need help with determining if your computer is running a 32-bit version or 64-bit version of the Windows operating system.

   (For AWS) If you'd like to train the agent on AWS (and have not enabled a virtual screen), then please use this link (version 1) or this link (version 2) to obtain the "headless" version of the environment. You will not be able to watch the agent without enabling a virtual screen, but you will be able to train the agent. (To watch the agent, you should follow the instructions to enable a virtual screen, and then download the environment for the Linux operating system above.)

3. Place the file in the DRLND GitHub repository, in the p2_continuous-control/ folder, and unzip (or decompress) the file.

Instructions

Run the notebook Report.ipynb to train your own agent !

Approach and solution

The notebook Report.ipynb contains the instruction to set up the environment and the coden to solve the reinforcement problem. My solution uses a Deep Deterministic Policy Gradient approach with experience replay, see this paper for more details.

The code for agent, deep Q-Network and memory buffer are implemented in the file agent.py for the mono agent (option 1) and in the file multi_agent.py for the multi agent (option2). The deep neural network architectures for both actor and critic are defined in model.py.