DRL Continuous Control

Introduction

For this project, I trained a DDPG agent to solve 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.

The setting contains 20 identical agents, each with its own copy of the environment. It is useful for algorithms like PPO, A3C, and D4PG that use multiple (non-interacting, parallel) copies of the same agent to distribute the task of gathering experience.

Solving the Environment

To take into account the presence of many agents, 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. Download the environment from one of the links below. You need only select the environment that matches your operating system:

   • 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.)

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

Setup of repository

Apart from the README.md file this repository consists of the following files:

1. model.py: Actor and Critc Network classes;
2. ddpg_agent.py: Agent, ReplayBuffer and OUNoise classes; The Agent class makes use of the Actor and Critic classes from model.py, the ReplayBuffer class and the OUNoise class;
3. run.py: Script which will train the agent. Can be run directly from the terminal;
4. checkpoint_actor.pth: Contains the weights of successful Actor Networks;
5. checkpoint_critic.pth: Contains the weights of successful Critic Networks.

Todos:

1. Add config.py: a configuration files for training the model;
2. report.ipynb: As an alternative to the run.py script this Jupyter Notebook has a step-by-step structure. Here the learning algorithm is described in detail;
3. Improve plots;
4. Add more screen outputs during the training to better monitor the learning;
5. Add specific dirs for saving plots and model weights;
6. Add functions of loading weights and resume training;
7. Add gif visualizing the performance of the trained model;

Instruction

To train the model, simply run the following command

python run.py