Neuronav SLAM SDK

The SLAM Abstraction Layer for Robotics Platforms

Write once, run on any sensor. Simple abstraction layer over ROS2 SLAM pipelines.

Quick Start

RealSense Camera:

from neuronav import RealSenseSensor, RTABMapSLAM, run_slam

sensor = RealSenseSensor()
run_slam(sensor, RTABMapSLAM())

OAK-D Camera:

from neuronav import OAKDSensor, RTABMapSLAM, run_slam

sensor = OAKDSensor()
run_slam(sensor, RTABMapSLAM())

With Visualization:

run_slam(sensor, RTABMapSLAM(), visualize=True)
# Opens Foxglove at ws://localhost:8765

Features

• 2-Line API - Start SLAM instantly
• Sensor Agnostic - RealSense, OAK-D, or add your own
• RTAB-Map Integration - Production-ready visual SLAM
• Foxglove Visualization - Web-based 3D viewing
• Extensible - Add sensors in 20 min, SLAM algorithms in 1 hour

Installation

Docker (Recommended):

./docker_build.sh
./docker_run.sh

Local:

pip install -e .

Requires: ROS2 Humble, Python 3.8+, camera drivers (setup guide)

Supported Hardware

Sensors:

• Intel RealSense D435i, D455, D415
• Luxonis OAK-D Pro/W

SLAM:

• RTAB-Map RGB-D SLAM

License

Licensed under Apache 2.0.

Credits

Built on RTAB-Map (Copyright (c) 2010-2025, Mathieu Labbé - IntRoLab - Université de Sherbrooke, BSD 3-Clause License).

See ACKNOWLEDGMENTS.md for complete credits and licenses.

---

Issues? → GitHub Issues