SLAM2REF: Advancing Long-Term Mapping with 3D LiDAR and Reference Map Integration for Precise 6-DoF Trajectory Estimation and Map Extension

TL;DR

SLAM2REF addresses the problem of long-term indoor mapping by integrating mobile 3D LiDAR-IMU data with a reference map (BIM or point cloud) to achieve precise 6-DoF trajectory estimation and automatic map extension. It introduces a three-stage pipeline: (i) automatic generation of synthetic session data from large-scale reference maps, (ii) reference-map-based multi-session anchoring to align drifted SLAM data with the reference using ISC loop detection and YawGICP, and (iii) change detection with OctoMap to produce a revised, visually interpretable map. Key contributions include a robust ISCD for indoor place recognition, the YawGICP registration module, and a multi-session anchoring formulation with anchor nodes enabling global alignment without starting inside the map, plus end-to-end evaluation on ConSLAM showing centimeter-level pose retrieval and reliable map updates. The framework supports various reference representations, handles Scan-Map deviations, and provides practical tools for digital twins, construction monitoring, and emergency response. Overall, SLAM2REF advances SLAM research by enabling automatic, centimeter-level ground-truth Pose retrieval and coherent map extension in challenging GPS-denied indoor environments.

Abstract

This paper presents a pioneering solution to the task of integrating mobile 3D LiDAR and inertial measurement unit (IMU) data with existing building information models or point clouds, which is crucial for achieving precise long-term localization and mapping in indoor, GPS-denied environments. Our proposed framework, SLAM2REF, introduces a novel approach for automatic alignment and map extension utilizing reference 3D maps. The methodology is supported by a sophisticated multi-session anchoring technique, which integrates novel descriptors and registration methodologies. Real-world experiments reveal the framework's remarkable robustness and accuracy, surpassing current state-of-the-art methods. Our open-source framework's significance lies in its contribution to resilient map data management, enhancing processes across diverse sectors such as construction site monitoring, emergency response, disaster management, and others, where fast-updated digital 3D maps contribute to better decision-making and productivity. Moreover, it offers advancements in localization and mapping research. Link to the repository: https://github.com/MigVega/SLAM2REF, Data: https://doi.org/10.14459/2024mp1743877.

Figures (16)

• Figure 1: Overview of SLAM2REF. The pipeline consists of three steps: map-based session data generation, Reference map-based multi-session anchoring, and Change detection and map update.
• Figure 2: Generated OGM from the BIM model. On the left, the different layers, and on the right, the merged final OGM.
• Figure 3: Calculated locations for scan simulation. On the left are the main steps, and on the right are all the calculated positions in the entire OGM.
• Figure 4: Synthetic session data from the reference map. On the left, from top to bottom: Top view of one LiDAR scan, its corresponding polar scan context (SC) descriptor, and the descriptor in the matrix form. In the middle, a set of simulated scans and the STL mesh from the BIM model are used. Right, corresponding SC descriptors for the simulated scans.
• Figure 5: Comprehensive flowchart illustrating the multi-session anchoring process within SLAM2REF. This process includes the generation of session data from the reference map $\mathcal{S_R}$, creation of the real-world query session $\mathcal{S_Q}$, inter-session loop detection using Indoor Scan Context and YawGICP, and pose refinement with KNN loops and final ICP. The outcome includes the anchor node $\Delta_Q^*$, optimized 6-DoF poses $\textbf{x}_Q^*$, and a confidence level list $\mathbf{\nu}_Q$ for each pose in the query session.