LIVE-GS: Online LiDAR-Inertial-Visual State Estimation and Globally Consistent Mapping with 3D Gaussian Splatting

TL;DR

LIVE-GS, an online LiDAR-Inertial Visual SLAM framework that tightly couples 3D Gaussian Splatting with LiDAR-based surfels to ensure high-precision map consistency through global geometric optimization, is proposed.

Abstract

While 3D Gaussian Splatting (3DGS) enabled photorealistic mapping, its integration into SLAM has largely followed traditional camera-centric pipelines. As a result, they inherit well-known weaknesses such as high computational load, failure in texture-poor or illumination-varying environments, and limited operational range, particularly for RGB-D setups. On the other hand, LiDAR emerges as a robust alternative, but its integration with 3DGS introduces new challenges, such as the need for tighter global alignment for photorealistic quality and prolonged optimization times caused by sparse data. To address these challenges, we propose LIVE-GS, an online LiDAR-Inertial Visual SLAM framework that tightly couples 3D Gaussian Splatting with LiDAR-based surfels to ensure high-precision map consistency through global geometric optimization. Particularly, to handle sparse data, our system employs a depth-invariant Gaussian initialization strategy for efficient representation and a bounded sigmoid constraint to prevent uncontrolled Gaussian growth. Experiments on public and our datasets demonstrate competitive performance in rendering quality and map-building efficiency compared with representative 3DGS SLAM baselines.

Figures (8)

• Figure 1: LIVE-GS. Bird's-eye view of a reconstructed 3D Gaussian map (left) and rendered images from selected viewpoints (right). Our globally optimized LiDAR-inertial-visual mapping produces photorealistic renderings in large-scale outdoor environments.
• Figure 2: LIVE-GS System Overview. Dual-map architecture fuses LiDAR, camera, and IMU streams, tracking poses on a 2D surfel map while optimizing a depth-invariant 3D Gaussian map for photorealistic rendering. Loop closure then performs global refinement that jointly optimizes poses and Gaussians, yielding a geometrically consistent high-resolution map.
• Figure 3: Depth-Invariant Gaussian Initialization. (a) illustrates the overall process of depth-invariant Gaussian initialization, while (b) shows its relationship with conics
• Figure 4: Sliding-window 3DGS Optimization. At index $i$, we jointly optimize the most recent $K$ camera keyframes $\{\mathbf{T}_{i-K+1},\dots,\mathbf{T}_{i}\}$ together with all Gaussians visible from them.
• Figure 5: Experimental handheld platforms. The Livox AVIA (left) and MID-360 (right) are integrated with IMU and camera sensors.