Back to the map
Robotics2026
FastSLAM & Autonomous Navigation
A TurtleBot that maps a room, localizes in it, and drives itself to a target.
Overview
A full ROS 2 Humble autonomy stack for a TurtleBot, taken from simulation to the real robot. It builds a map with grid-based FastSLAM (a Rao-Blackwellized particle filter carrying one occupancy grid per particle), localizes with Monte Carlo Localization, and navigates with A* planning and pure-pursuit control, closing with a red-cone perception mission on the physical TurtleBot4. ROS-free math (geometry, likelihood field, odometry) is isolated into a shared library so the core can be unit-tested without booting ROS.
Highlights
- Grid-based FastSLAM (RBPF) with one occupancy grid per particle, from reactive obstacle avoidance up to full mapping.
- Autonomous navigation: MCL localization plus A* planning plus pure-pursuit tracking.
- Sim-to-real: red-cone detection and mission executed on the real TurtleBot4, with a shared testable math library.