SLH: An Introduction to Multi-agent Driving Simulation with GPUDrive

In this seminar, I will introduce GPUDrive, a high-performance, data-driven driving simulator that operates at 1 million FPS. GPUDrive is built on the Madrona Game Engine and uses GPU acceleration to enable scalable multi-agent simulation. I will discuss how this simulator can be used to train reinforcement learning agents efficiently, drawing from a recent paper where we demonstrate its application to the Waymo Open Motion Dataset. Additionally, I will walk through tutorials on how to set up and use the simulator.

The repository can be found at https://github.com/Emerge-Lab/gpudrive/tree/main