Hi everyone, I’d like to present my project GaussianRPG, the first open-source autonomous driving closed-loop simulator demo using 3D Gaussian Splatting tech.
The whole project is base on the codes from this paper:
Street Gaussians: Modeling Dynamic Urban Scenes with Gaussian Splatting
Yunzhi Yan, Haotong Lin, Chenxu Zhou, Weijie Wang, Haiyang Sun, Kun Zhan, Xianpeng Lang, Xiaowei Zhou, Sida Peng
ECCV 2024
In this project, a simulator has been developed. Street_Gaussians is responsible for rendering images based on the camera position. A dummy AEB controller applies YOLOv5 to recognize cars in the images and determine the distance, issuing braking commands when appropriate. The car dynamics component within the simulator adjusts the camera position according to the braking commands and provides this information to Street_Gaussians as the basis for rendering the next frame. For convenience, ROS2 serves as the middleware providing the framework for the simulator.
Examples:
The simulator runs in two modes:
Separate perception mode:
Not separate perception mode:
