This project is based on a project created during the first semester of my year at the University of Sherbrooke (Canada). Named "PetitMoteur3D", it was a way to learn how game engine works by creating our own engine in C++ using DirectX 11 and PhysX. I kept the "Core" and "Api" part that does the rendering and added IA stuff to it to test and experiment.
- Clone this repository.
- Build PhysX from NVIDIA's repository, and copy the following .dll files in the
dllfolder :
- PhysX_64.dll
- PhysXCommon_64.dll
- PhysXCooking_64.dll
- PhysXFoundation_64.dll
- Build PhysX from NVIDIA's repository in debug mode, and copy the following .lib files in the
libs/debugfolder:
- PhysX_64.lib
- PhysXCommon_64.lib
- PhysXCooking_64.lib
- PhysXExtensions_static_64.lib
- PhysXFoundation_64.lib
- PhysXPvdSDK_static_64.lib
- Build PhysX from NVIDIA's repository in release mode, and copy the same .lib files in the
libs/releasefolder. - Copy PhysX include files from the cloned repo (/physx/include) into the
libs/physx_includefolder. - Build FX11 form Microsoft's repository and copy the
effects11.libfile inlibs/releaseand theeffects11d.libfile inlibs/debug. - Download libtorch from here. Extract the zip file at the project's root.
- If you downloaded libtorch with CUDA support, install CUDA if you don't already have it. (https://developer.nvidia.com/cuda-downloads)
- Run the
rebuild.batfile to generate the solution files. You need cmake and Visual Sutdio 2022 to be installed. If you have an issue like "Failed to find nvToolsExt", have a look at pytorch/pytorch#116242. - Open the
build/AIPlayground.slnfile with your preferred IDE. - Build the solution
// TODO
- I used Eyyu's blog to help me understand how to code the PPO algorithm. You can find the blog post here.