Implementing an Autonomous Navigation Stack in ROS2: Simulation on the Bumperbot Platform
Fernando Ruvalcaba Cruz, Claire L. Walton, Michael T. Frye · IFAC-PapersOnLine · 2025
Autonomous vehicle navigation is a basic functionality for ground robots, with applications within research and real-world scenarios. This paper discusses the configuration and utilization of ROS 2’s native tools and packages to deploy a complete indoor path planning system on a differential-drive robot in a simulation setup. The BumperBot platform, obtained from an open-source GitHub repository, receives simulated odometry and lidar sensor data, and employs Graph SLAM for mapping and Adaptive Monte Carlo Localization (AMCL) for pose estimation through the slam_toolbox package and configuration files. Path planning and control are implemented using the Nav2 stack, with the addition of a custom Python node for executing autonomous waypoint navigation. Experimental results on the Gazebo simulation platform confirm the usefulness of ROS 2’s modular navigation stack for enabling autonomous point-to-point navigation with minimal custom code. One of the main characteristics of this method is its architectural modularity, allowing for clear separation of mapping, localization, and navigation modes and portability, as demonstrated by adapting the same launch and navigation framework to the QCar2 simulated platform with slight changes. The system that is produced offers a reproducible and extensible basis for future research on simulation-based autonomous navigation, educational robotics, and quick prototyping of novel platforms or behaviors.