StrataSLAM releases open-source multi-floor 3D SLAM for ROS 2 Jazzy

The Intelligent Robotics Lab at Universidad Rey Juan Carlos has released StrataSLAM, an Apache-2.0 licensed multi-floor 3D SLAM system for ROS 2 Jazzy on Ubuntu 24.04. The package addresses a persistent failure mode in conventional SLAM: robots losing localization or generating merged maps when traversing stairs, ramps, elevators, or visually identical floors. StrataSLAM maintains discrete submaps per floor and detects transition types automatically, using IMU data to continue odometry through elevator rides even when LiDAR or cameras only see the cabin interior.

The system supports RGB-D cameras, 3D LiDAR, or fused sensor configurations, with optional IMU, barometer, and wheel odometry inputs. A YAML profiling system covers ground robots, hand-held devices, and aerial platforms. A strataslam doctor diagnostic tool validates topic rates, TF trees, and synchronization before mapping begins. Offline fusion aligns floors in a gravity-aligned 4-DoF frame without requiring visual overlap between levels, and maps export to a versioned open format with per-floor and global point clouds. Demonstrations include simulated buildings and real-world data from Summit XL, TIAGo, and KITTI datasets.

Architecture and floor-transition handling

• Floor gate mechanism: Loop closure is explicitly prohibited between different floors, preventing identical floor geometries from being merged into a single incorrect map; the project video demonstrates behavior with the gate enabled and disabled.
• Elevator odometry fallback: When camera or LiDAR sensors observe only the moving cabin interior, the system continues pose estimation using IMU integration to measure vertical displacement.
• Multi-sensor input flexibility: Core SLAM operates on RGB-D, 3D LiDAR, or combined streams; IMU, barometer, and wheel odometry are optional augmentations rather than hard dependencies.
• Pre-mapping diagnostics: The strataslam doctor command checks ROS 2 topic publication rates, TF frame connectivity, and message synchronization to catch configuration errors before data collection.
• Offline gravity-aligned fusion: A post-processing stage places floor submaps relative to each other in a 4-DoF frame aligned to gravity, enabling multi-floor reconstruction even without inter-floor visual features.

Why it matters

StrataSLAM fills a tooling gap for indoor mobile robotics deployments in warehouses, hospitals, and multi-level infrastructure where existing SLAM packages routinely fail during vertical transitions. Its ROS 2 Jazzy targeting and Apache-2.0 licensing align with the trajectory of modern robotics middleware, making it directly integrable with Nav2 and the broader ROS 2 navigation stack. The open map format and platform-agnostic YAML profiles lower barriers for researchers and systems engineers working across heterogeneous robot fleets.