This project involved building a real-time indoor mapping system using RTAB-Map SLAM to generate accurate spatial representations of tunnel-like environments. The goal was to reliably capture both the robot’s trajectory and the surrounding map in real time, while ensuring consistency over long traversals.
Using a ZED Mini stereo camera and ROS2, I deployed RTAB-Map to perform visual SLAM and generate 3D occupancy maps. A key challenge was mitigating visual odometry drift caused by uneven motion and dynamic lighting conditions. To address this, I engineered a stabilized rolling platform that ensured smoother motion and more reliable feature tracking.
The system produced high-fidelity occupancy maps that closely matched the physical tunnel layout. Robot trajectories and loop closures were visualized live in RViz, allowing real-time validation and debugging of SLAM behavior. This project demonstrated the practical challenges of deploying SLAM systems outside controlled laboratory conditions.