Smart Car Project

OpenGL based self-driving car (with DQN algorithm LibTorch neural network)

Installation

Make sure you have (Graphics part)

1. Git
2. Powershell
3. OpenGL

Make sure you have (Neural network part)

1. CUDA (~3.5 GB) [Tested version 12.2.1_536.67 on Windows 11]
2. cuDNN (~700 MB) [Tested version 8.9.3.28 for CUDA 12 on Windows 11]
3. NVIDIA Nsight Graphics (~1 GB) [Tested version 2023.2.1.23178 on Windows 11]
4. LibTorch for C++ (~3 GB) [Tested DEBUG version 2.0.1 for CUDA 11.8 on Windows 11]

and copied all files from

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\<version>\extras\visual_studio_integration\MSBuildExtensions

to

C:\Program Files (x86)\Microsoft Visual Studio\<version>\<product>\MSBuild\Microsoft\VC\<version>\BuildCustomizations

where <product> is the MSVC product you use: { BuildTools, Enterprise, Community, etc. }

Execution

git clone https://github.com/BiscuitsLayer/SmartCarProject.git
cd ./SmartCarProject
git submodule update --init --recursive
./run.ps1

Sport car model: link

Road model: link

Checked flag model: link

Useful sources:

LibTorch RL implementation

DQN Theory

DQN Basics

DQN for Continuous Control Tasks