Splat-LOAM: Gaussian Splatting LiDAR Odometry and Mapping

2025-03-21

AI Summary

To address the longstanding trade-off between high-fidelity geometric representation and low computational overhead in LiDAR SLAM, this paper introduces the first pure LiDAR odometry and mapping framework based on 3D Gaussian primitives. Methodologically, it departs from conventional representations—such as voxels, meshes, and NeRF—and pioneers the integration of differentiable Gaussian splatting into LiDAR SLAM. By leveraging spherical projection, it enables end-to-end optimization of Gaussian parameters directly from raw point clouds, supporting native, neural-network-free, lightweight LiDAR rendering and incremental map updates. Experiments demonstrate that the method achieves localization accuracy comparable to state-of-the-art approaches and attains SOTA mapping quality. Crucially, it operates in real time on low-end GPUs, substantially lowering the barrier for embedded deployment. The core contribution lies in a synergistic breakthrough: simultaneously achieving high geometric fidelity and exceptional computational efficiency.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Intelligent Robots: Localization, Mapping, and Navigation
• Computer Vision: Low Level & Physics-based Vision

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

LiDARs provide accurate geometric measurements, making them valuable for ego-motion estimation and reconstruction tasks. Although its success, managing an accurate and lightweight representation of the environment still poses challenges. Both classic and NeRF-based solutions have to trade off accuracy over memory and processing times. In this work, we build on recent advancements in Gaussian Splatting methods to develop a novel LiDAR odometry and mapping pipeline that exclusively relies on Gaussian primitives for its scene representation. Leveraging spherical projection, we drive the refinement of the primitives uniquely from LiDAR measurements. Experiments show that our approach matches the current registration performance, while achieving SOTA results for mapping tasks with minimal GPU requirements. This efficiency makes it a strong candidate for further exploration and potential adoption in real-time robotics estimation tasks.

Problem

• Achieve accurate LiDAR odometry with lightweight Gaussian Splatting
• Balance mapping accuracy and computational efficiency in LiDAR systems
• Enable real-time robotics tasks with minimal GPU requirements

Innovation

• Uses Gaussian Splatting for LiDAR mapping
• Refines primitives via spherical projection
• Achieves SOTA mapping with minimal GPU