SLABIM: A SLAM-BIM Coupled Dataset in HKUST Main Building

TL;DR

This work introduces SLABIM, the first open-source dataset that couples SLAM data with BIM for the HKUST main building, addressing the gap between robotic mapping and architectural reasoning. It provides a decomposed as-designed BIM into meshes and floor plans alongside a multi-session, multi-sensor SLAM dataset with timestamped data and ground-truth poses, enabling cross-modal benchmarking. The authors validate SLABIM on three tasks—global LiDAR-to-BIM registration, robot pose tracking on BIM, and semantic mapping evaluation—demonstrating practical viability and highlighting strengths and limitations of current cross-modal approaches. By making the dataset publicly available, SLABIM aims to spur interdisciplinary research in robotics, BIM, and digital twin applications for indoor environments.

Abstract

Existing indoor SLAM datasets primarily focus on robot sensing, often lacking building architectures. To address this gap, we design and construct the first dataset to couple the SLAM and BIM, named SLABIM. This dataset provides BIM and SLAM-oriented sensor data, both modeling a university building at HKUST. The as-designed BIM is decomposed and converted for ease of use. We employ a multi-sensor suite for multi-session data collection and mapping to obtain the as-built model. All the related data are timestamped and organized, enabling users to deploy and test effectively. Furthermore, we deploy advanced methods and report the experimental results on three tasks: registration, localization and semantic mapping, demonstrating the effectiveness and practicality of SLABIM. We make our dataset open-source at https://github.com/HKUST-Aerial-Robotics/SLABIM.

Figures (9)

• Figure 1: Screenshot of SLABIM in the HKUST main building. The top image shows a SLAM-generated colored map aligned with the BIM. The bottom images in boxes show the same views of real-world colored map and rendered images generated by SLAM and BIM, respectively.
• Figure 2: (a) HKUST BIM and the four typical types of elements extracted from the decomposition and conversion: walls, floors, doors, and columns. (b) 2D floor plan map exported from HKUST BIM, which could be used for map-based navigation. The column on the right shows the different departments (regions) of the university, e.g. ECE (Electronic and Computer Engineering).
• Figure 3: Sensor suite for data collection. A real-time SLAM system is integrated to produce colored 3D point cloud maps, with offline map refinement available to enhance mapping quality.
• Figure 4: Scenarios in the HKUST main (academic) building. Challenging conditions, such as long corridors and glass windows, pose difficulties for SLAM and applications with BIM.
• Figure 5: Trajectories and SLAM-generated maps of our self-collected session in the 1F of the HKUST BIM, shown with bird's eye view. The other sessions are distributed across different regions of the 3F, 4F and 5F. All trajectories can not be displayed due to the page limit, which will be open sourced for reference.