Problem
Enable Segway Loomo robots to navigate crowded, dynamic indoor environments; most academic navigation systems never leave simulation.
Built
A complete ROS autonomy stack — SLAM (GMapping/Cartographer) producing occupancy grids, an A* global planner with dynamic replanning, costmap-based reactive obstacle avoidance, and vision-based person-following with multi-object tracking and trajectory prediction.
Why hard
Real-time performance across several computationally intensive perception modules running simultaneously, validated among actual pedestrians, not just in simulation.
What happened
Validated end-to-end on physical robots (Segway Loomo, EPFL VITA Lab research platform); the full stack was open-sourced on GitHub (github.com/thrmnn/ROS_Autonomous_Driving).