Autonomous Metro Cleaning Robot

This project focused on developing a robust autonomous navigation system for a mobile robot designed to clean metro train cars in Barcelona. The primary challenge was the highly symmetrical and narrow environment, which posed significant difficulties for precise localization and effective full-coverage path planning.

My work involved a rigorous benchmarking process to select the optimal software components. For localization and SLAM, I evaluated five state-of-the-art methods using a real-world dataset. After analyzing the trade-offs of Lidar Localization (NDT-OMP), GLIM, ORB-SLAM3, and RTAB-Map, Fast-LIO was selected for its robust, real-time performance and its ability to generate accurate maps without drift or corruption.

For path planning, I benchmarked several coverage planners. Existing solutions like Fields2Cover and the Nobleo Full Coverage Planner proved unsuitable for the metro’s complex and narrow spaces. Ultimately, a custom ROS2 wrapper for a Coverage Planner was selected and successfully adapted to the metro’s constraints by carefully tuning its parameters.

The navigation stack was built on Nav2, with the controller changed to the Regulated Pure Pursuit algorithm for smoother path following. Final experiments conducted in a metro carriage demonstrated the system’s success, as the robot completed both obstacle avoidance and full-coverage patterns without collisions. The project concluded with a robust and validated system, and identified key areas for future improvement, including enhanced sensor redundancy and filtering to handle challenges like LiDAR occlusion and reflective surfaces.