Gaussian Semantic Field for One-shot LiDAR Global Localization

2025-10-13

AI Summary

To address semantic landmark ambiguity arising from semantic repetition in LiDAR-based global localization, this paper proposes a robust localization method leveraging Gaussian semantic fields and a three-tier scene graph. The core contributions are: (1) a lightweight, hierarchical 3D scene graph that explicitly encodes semantic–geometric–metric relationships; (2) a Gaussian process–driven Gaussian semantic field, which models discrete semantic labels as continuous spatial distributions to jointly represent fine-grained geographic semantics and metric information; and (3) a semantic–metric joint optimization algorithm that enhances both accuracy and robustness in correspondence establishment. Extensive experiments on public urban-scale datasets demonstrate that the proposed method achieves significantly higher single-shot localization success rates compared to state-of-the-art approaches—particularly excelling in landmark-dense and structurally repetitive regions.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Search and Optimization: Local Search
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

We present a one-shot LiDAR global localization algorithm featuring semantic disambiguation ability based on a lightweight tri-layered scene graph. While landmark semantic registration-based methods have shown promising performance improvements in global localization compared with geometric-only methods, landmarks can be repetitive and misleading for correspondence establishment. We propose to mitigate this problem by modeling semantic distributions with continuous functions learned from a population of Gaussian processes. Compared with discrete semantic labels, the continuous functions capture finer-grained geo-semantic information and also provide more detailed metric information for correspondence establishment. We insert this continuous function as the middle layer between the object layer and the metric-semantic layer, forming a tri-layered 3D scene graph, serving as a light-weight yet performant backend for one-shot localization. We term our global localization pipeline Outram-GSF (Gaussian semantic field) and conduct a wide range of experiments on publicly available data sets, validating the superior performance against the current state-of-the-art.

Problem

• Solving landmark ambiguity in LiDAR global localization
• Modeling semantic distributions with continuous Gaussian processes
• Creating lightweight tri-layered scene graph for localization

Innovation

• Uses Gaussian processes for semantic distribution modeling
• Builds tri-layered scene graph with continuous semantic layer
• Creates lightweight backend for one-shot LiDAR localization