A self driving car with reinforcement learning

Machine Learning Project

A.A. 2018/2019

Antonio Lategano

Salvatore Visaggi

Usage

Donkey simulator

If you have a Mac the DonkeyCar simulator starts automatically when you run the python file ddqn.py. Instead, if you have a Windows PC you can find the simulator in the donkey_windows folder. You have to start the executable DonkeySim.exe before running the Python file.

Setup

You need Python 3.6 or higher and you have to install all the requirements.

pip install -r requirements.txt

Second, you have to install donkey_gym python package, which extends the OpenAI gym class to allow RL developers to interact with Donkey environment using the familiar OpenAI gym like interface.

To install the package, navigate to ml_project/donkey_gym folder and type the following command

$ cd ml_project/donkey_gym
$ pip install -e .

Run

In the config.py file you can find lots of variables you can modify. The most important are:

• EPISODES number of epochs
• TRAIN if set to True you can train your model, otherwise you can use the pretrained model
• LANE_DETECTION_TYPE 1 = bw raw images | 2 = lane detection | 3 = points detection
• MODEL_TYPE 1 = atari | 2 = custom_nn

To start the execution

$ python ddqn.py