Project 2: Continuous Control

Introduction

This project is to meet the requirements of project 2 - Continuous Control. It uses the Reacher environment.

In this environment, a double-jointed arm can move to target locations. A reward of +0.1 (+0.1 is advertised, but I found it varies between 0 and .0399_) is provided for each step that the agent's hand is in the goal location. Thus, the goal of the 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

The assignment is to pick one of two versions of the assignment. V1 is a single arm. V2 has 20 identical arms. I used V2, but failed to get any distributed algorithms to work. Vanilla DDPG was used to train this agent.

Criteria

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.

Instructions for using

1. Clone this repo.

2. If you are not running windows 10, you will need to Download the environment from one of the links below. (I'm amused that Win 32 is supported. Are there any win 32 machines left in the world.)

   • 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.)

2a. Place the file in the p2_continuous-control/ folder, and unzip (or decompress) the file.

2b. You will need to change the path in DDPG_main.py according to your plaform. Toward the top, you will find the following code. You should be able to change "Reacher_Windows_x86_64" and "Reacher.exe" to your flavor to make this work.

dir = os.getcwd()
dir = dir + os.sep + "Reacher_Windows_x86_64"
env = UnityEnvironment(file_name=dir + os.sep + "Reacher.exe")

Instructions

From a command line: Activate your environment. In my case activate drlnd40

type python DDPG_main.py or run from within PyCharm as I did.

requirements:

1. PyTorch 0.4.0 or 0.4.1 (Tested in both although I switched to 0.4.0 for reasons that have since escaped me)
2. Numpy (I used latest as far as I know.)
3. https://github.com/Unity-Technologies/ml-agents