This is an implementation on Windows 10 of the Navigation Project (Project 1) of Udacity's Deep Reinforcement Learning Nanodegree (DRLND) program. Original information about this project and the starter code can be found in Udacity's GitHub repository for this project.

For this project, the agent is to navigate and collect bananas in a large square world.

A reward of +1 is provided for collecting a yellow banana, and a reward of -1 for a blue banana. Thus, the goal of the agent is to collect as many yellow bananas as possible while avoiding blue bananas.

The state space has 37 dimensions and contains the agent's velocity, along with ray-based perception of objects around agent's forward direction. Given this information, the agent has to learn how to best select actions. Four discrete actions are available, corresponding to:

• 0 - move forward.
• 1 - move backward.
• 2 - turn left.
• 3 - turn right.

The task is episodic, and in order to solve the environment, the agent must get an average score of +13 over 100 consecutive episodes.

1. Download the environment from Udacity using this link (for 64-bit Windows only): https://s3-us-west-1.amazonaws.com/udacity-drlnd/P1/Banana/Banana_Windows_x86_64.zip.

2. Place the file in the repository and unzip it there creating a Banana_Windows_x86_64 folder.

The repository contains three modules:

• model.py: Neutral network for implementing the model used by the agent to select its actions.
• ddqn_agent.py: Code for implementing agent's behavior using the neutral network model.
• train_test.py: Code for training and testing the agent.

Step by step instructions for using the modules to test the environment and experiment with training and testing the agent are provided in Navigation.ipynb. Detail discussions of the implementation are in Report.md.

To set up the Python environment to run the code in the repository, follow the instructions below.

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

   conda create --name ENVNAME python=3.6
   conda activate ENVNAME

2. Install PyTorch and TorchVision into the newly created environment.

   conda install -c pytorch pytorch torchvision

3. Navigate to the python/ folder in the repository and install dependencies specified there.

   cd Udacity_DRLND_Navigation/python
   pip install .