Welcome to the Continuous Control Project.
In this Project we train an agent to follow a region with a robotic arm
👇🏼This is the resulting agent
This project contains a solution to the second project of Udacity Deep Reinforcement Learning. This Project uses a DDPG Algorithm to train the agent.
A reward of ~0.04 is provided for being within the ball, and a reward of 0 is provided for being outside the goal at any given timestep. Therefore the goal is to follow the goal area with the robotic arm.
The state space consists of 33 dimensions and is continuous.
The action space consists of 4 continuous actions, which control the arm.
The Agent is trained using a DDPG algorithm. For further information on training please read the Report.md.
The task is episodic, and in order to solve the environment, the agent must get an average score of +30 over 100 consecutive episodes.
Check the following link for more details:
https://github.com/udacity/deep-reinforcement-learning/tree/master/p1_navigation
###Prerequisites Python 3.6 Unity Conda
##Installation:
- Clone the repository
https://github.com/schmiJo/p2_Continuos-control
- Install Jupyter Notebook
pip install jupyter
- Create and activate a new environment for Python 3.6
- Linux or Mac
conda create --name drlnd python=3.6
source activate drlnd
- Windows
conda create --name drlnd python=3.6
activate drlnd
- Install several dependencies
pip install -r requirements.txt
- Before running the Continuous_Control.ipynb change the kernel to match the drlnd environment by using the drop down Kernel menu.
Download the unity environment using the following link for macOs:
https://s3-us-west-1.amazonaws.com/udacity-drlnd/P2/Reacher/one_agent/Reacher.app.zip
Mre instructions for the installation can be found under:
https://github.com/udacity/deep-reinforcement-learning#dependencies