Skip to content

Latest commit

 

History

History
93 lines (57 loc) · 3.92 KB

README.md

File metadata and controls

93 lines (57 loc) · 3.92 KB
Raw

Deep Reinforcement Learning Nanodegree: Project 2 - Continuous Control

This project includes the code for an implementation of the deep deterministic policy gradient(DDPG) algorithm which I wrote to solve the Project 2 - Continuous Control of the Deep Reinforcement Learning Nanodegree @ Udacity. My version of the DDPG algorithm is inspired by chapter 14 of Maxim Lapan’s book called "Deep Reinforcement Learning Hands-On".

For more information on the implemented features refer to "Report.ipynb". The notebook includes a summary of all essential concepts used in the code.

Project 2 - Continuous Control - Details:

The goal of this project was to train an agent, represented by a double-jointed arm, to maintain its position at the target location(great green sphere) for as many time steps as possible.

Random Agent

Random Agent

Trained Agent

Trained Agent

Reward:
  • a reward of +0.1 is provided for each step that the agent's hand is in the goal location
Search Space
  • the state space has 33 dimensions
    • corresponding to the position, rotation, velocity, and angular velocities of the arm
  • the action space has four dimensions
    • every action is a continuous number between -1 and 1
Task
  • the task is episodic
  • the agent has to maintain its position at the target location(great green sphere) for as many time steps as possible
  • to solve the environment, the agent must get an average score of +30 over 100 consecutive episodes

Getting Started

  1. Create (and activate) a new environment with Python 3.6.

conda create --name env_name python=3.6
source activate env_name

  1. Download the environment from one of the links below and place it into \p2_continuous-control\Reacher_One_Linux

\Reacher_One_Linux
     \Reacher_Data
     \Reacher.x86
     \Reacher.x86_64

  1. Install Sourcecode dependencies

conda install -c pytorch pytorch
conda install -c anaconda numpy
pip install tensorboardX

  • unityagents is also required
    • an easy way to get this is to install the Deep Reinforcement Learning Nanodegree with its dependencies

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

How to run the project

You can run the project by running the main.py file through the console.

  • open the console and run: python main.py -c "your_config_file.json"
  • to train the agent from scratch set "run_training" in the config file to true
  • to run the pre-trained agent set "run_training" in the config file to false

optional arguments:

-h, --help

- show help message

-c , --config

- Config file name - file must be available as .json in ./configs

Example: python main.py -c "reacher_one.json"