LiV-GS: LiDAR-Vision Integration for 3D Gaussian Splatting SLAM in Outdoor Environments

TL;DR

To reliably and stably update Gaussians outside the LiDAR field of view, a novel conditional Gaussian constraint is introduced that aligns these Gaussians closely with the nearest reliable ones.

Abstract

We present LiV-GS, a LiDAR-visual SLAM system in outdoor environments that leverages 3D Gaussian as a differentiable spatial representation. Notably, LiV-GS is the first method that directly aligns discrete and sparse LiDAR data with continuous differentiable Gaussian maps in large-scale outdoor scenes, overcoming the limitation of fixed resolution in traditional LiDAR mapping. The system aligns point clouds with Gaussian maps using shared covariance attributes for front-end tracking and integrates the normal orientation into the loss function to refines the Gaussian map. To reliably and stably update Gaussians outside the LiDAR field of view, we introduce a novel conditional Gaussian constraint that aligns these Gaussians closely with the nearest reliable ones. The targeted adjustment enables LiV-GS to achieve fast and accurate mapping with novel view synthesis at a rate of 7.98 FPS. Extensive comparative experiments demonstrate LiV-GS's superior performance in SLAM, image rendering and mapping. The successful cross-modal radar-LiDAR localization highlights the potential of LiV-GS for applications in cross-modal semantic positioning and object segmentation with Gaussian maps.

Figures (8)

• Figure 2: Overview of the system: The SLAM system comprises a tracking and optimization process that together support the visual representation of the Gaussian map. The map update process uses LiDAR depth and color supervision to adjust the new Gaussians.
• Figure 3: Relationship between Density and weight: Gaussians based on only color supervision result in isotropic and sparse Gaussians (top left). Regions with dense depth input from LiDAR typically show Gaussians in higher density (bottom left).
• Figure 4: Effect of Normal Restriction:Top: Ellipsoid visualization. Middle: Render images. Bottom: Magnified details of the render. The left comparison (in red) illustrates uncontrolled Gaussian growth leading to significant artifacts. The right comparison (in green) shows gaps in the rendered images caused by isotropic Gaussians. Our method effectively prevents these issues.
• Figure 5: Effect of Splitting via conditional Gaussian constraints (CGC). Our approach enhances the representation of Gaussians for objects in the images that lack LiDAR depth input via the introduced CGC.
• Figure 6: Comparison of trajectories using different SLAM algorithms on four sequences of NTU4DRadLM dataset.