Robotics research has introduced Splat-Nav, a real-time navigation pipeline based on Gaussian Splatting maps, an emerging 3D scene representation. This novel approach consists of two parts: a safe planning module and a robust state estimation module. The key features include:
All these operate on CPU to prioritize GPU resources for other tasks, boasting impressive re-planning at 5 Hz and pose estimation at 20 Hz, which are significantly faster than NeRF-based methods. This approach could revolutionize how robots interact with 3D environments, especially in real-time applications where both speed and accuracy are critical.