IILABS 3D: iilab Indoor LiDAR-based SLAM Dataset

The IILABS 3D dataset is a rigorously designed benchmark intended to advance research in 3D LiDAR-based Simultaneous Localization and Mapping (SLAM) algorithms within indoor environments. It provides a robust and diverse foundation for evaluating and enhancing SLAM techniques in complex indoor settings. The dataset was retrived in the Industry and Innovation Laboratory (iiLab) and comprises synchronized data from a suite of sensors—including four distinct 3D LiDAR sensors, a 2D LiDAR, an Inertial Measurement Unit (IMU), and wheel odometry—complemented by high-precision ground truth obtained via a Motion Capture (MoCap) system.

Project Webpage https://jorgedfr.github.io/3d_lidar_slam_benchmark_at_iilab/

Dataset Toolkit https://github.com/JorgeDFR/iilabs3d-toolkit

Data Collection Method

Sensor data was captured using the Robot Operating System (ROS) framework’s rosbag record tool on a LattePanda 3 Delta embedded computer. Post-processing involved timestamp correction for the Xsens MTi-630 IMU via custom Python scripts.

Ground-truth data was captured using an OptiTrack MoCap system featuring 24 high-resolution PrimeX 22 cameras. These cameras were connected via Ethernet to a primary Windows computer running the Motive software (https://optitrack.com/software/motive), which processed the camera data. This Windows computer was then connected via Ethernet to a secondary Ubuntu machine running the NatNet 4 ROS driver (https://github.com/L2S-lab/natnet_ros_cpp). The driver published the data as ROS topics, which were recorded into rosbag files. Additionally, temporal synchronization between the robot platform and the ground-truth system was achieved using the Network Time Protocol (NTP). Finally, the bag files were processed using the EVO open-source Python library (https://github.com/MichaelGrupp/evo) to convert the data into TUM format and adjust the initial position offsets for accurate SLAM odometry benchmarking.

Type of Instrument
Mobile Robot Platform: INESC TEC MRDT Modified Hangfa Discovery Q2 Platform. R.B. Sousa, H.M. Sobreira, J.G. Martins, P.G. Costa, M.F. Silva and A.P. Moreira, "Integrating Multimodal Perception into Ground Mobile Robots," 2025 IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC2025), Madeira, Portugal, 2025, pp. TBD, doi: TBD [Manuscript accepted for publication].https://sousarbarb.github.io/inesctec_mrdt_hangfa_discovery_q2/

Sensor data: Livox Mid-360, Ouster OS1-64 RevC, RoboSense RS-HELIOS-5515, and Velodyne VLP-16 (3D LiDARs); Hokuyo UST-10LX-H01 (2D LiDAR); Xsens MTi-630 (IMU); and Faulhaber 2342 wheel encoders (64:1 gear ratio, 12 Counts Per Revolution (CPR)).

Ground Truth data: OptiTrack Motion Capture System with 24 PrimeX 22 cameras installed in Room A, Floor 0 at iilab