🏎 DDPG design for lane keeping in TORCS Environment

My absolutely first repository on github!

This repository contains my bachelor's degree thesis project: Designing a DDPG algorithm for approaching the lane keeping problem in the autonomous ground vehicle driving.

Pre-requisites

• Python 3 with Keras and Tensorflow
• TORCS game with SCR patch

Structure

The work is structured as follows.

├── src                         #The main folder of all the code
│   ├── DDPG                    #Folder containing all the classes and functions for implementing the algo.
│   |   ├── noise               #Foler containing the noise classes used fro the DDPG
|   |   |   ├── OU.py           #Implementation of the Ornstein–Uhlenbeck noise with the stochastic brake mod
|   |   |   ├── GN.py           #Implementation of the Gaussian Noise
|   |   |   └── TVNoise.py      #Implementation of the Time Variant Noise
│   │   ├── conditioning.py     #Implementation of a utility function for interacting with TORCS env.
│   │   ├── networks.py         #Implementation of Actor and Critic networks
│   │   ├── replayBuffer.py     #Implementation of the experience replay required by DDPG
│   │   ├── reward.py           #Implementation of the three env. reward functions
│   │   ├── config.py           #Config file you need to look before running any experiment with DDPG
│   │   └── agent.py            #Implementation of the agent class and all of its functions
│   ├── launch.py               #Main python script to launch the experiment
│   └── snakeoil.py             #Implementation of the client interface for the interaction with TORCS game
├── experiment                  #A copy of the src folder but with the final driver model in it for testing purpose
│   ├── DDPG                    #Folder containing all the classes and functions for implementing the algo.
│   |   ├── noise               #Foler containing the noise classes used fro the DDPG
|   |   |   ├── OU.py           #Implementation of the Ornstein–Uhlenbeck noise with the stochastic brake mod
|   |   |   ├── GN.py           #Implementation of the Gaussian Noise
|   |   |   └── TVNoise.py      #Implementation of the Time Variant Noise
|   ├── models                  #Folder containing different version of the driver. 
│   │   ├── conditioning.py     #Implementation of a utility function for interacting with TORCS env.
│   │   ├── networks.py         #Implementation of Actor and Critic networks
│   │   ├── replayBuffer.py     #Implementation of the experience replay required by DDPG
│   │   ├── reward.py           #Implementation of the three env. reward functions
│   │   ├── config.py           #Config file you need to look before running any experiment with DDPG
│   │   └── agent.py            #Implementation of the agent class and all of its functions
│   ├── launch.py               #Main python script to launch the experiment
│   └── snakeoil.py             #Implementation of the client interface for the interaction with TORCS game
└── docs                        #Folder containing documents
    ├── Tesi.pdf                #Thesis on DDPG design for lane keeping in TORCS Environment. Italian language
    └── latex_template          #Latex template folder of my thesis project