Enabling Autonomous Indoor Navigation with an ROS2/Nav2 Framework

Abstract

Autonomous indoor service robots require reliable navigation to perform room-to-room delivery, person guidance, and human-robot interaction data collection during deployment. Factory software for the AgileX Scout Mini does not provide an integrated workflow for user-directed autonomous navigation through a touchscreen graphical interface. An opensource Robot Operating System 2 (ROS2) framework was extended using the Nav2 navigation stack to support waypointbased autonomous navigation. Floor maps were generated using Simultaneous Localization and Mapping (SLAM), refined through keep-out zones and waypoint annotation, and integrated into Nav2 for localization and path planning. A PyQt touchscreen interface enabled users to select destinations by room name, initiate autonomous navigation, and submit post-navigation feedback. Adaptive Monte Carlo Localization (AMCL) and Nav2 planning parameters were tuned to improve localization accuracy and navigation reliability. System performance was evaluated through repeated autonomous trials between a fixed starting location and multiple destinations. The system successfully completed 46 of 50 autonomous trials (92%). All routes up to approximately 65 ft were completed successfully, whereas failures on longer routes were primarily associated with localization drift rather than path-planning limitations. Across representative routes, the mean waypoint error was 0.204 ± 0.038 m (95% CI: 0.188–0.219 m). These results demonstrate that an opensource ROS2/Nav2 framework can be extended into a reliable indoor navigation system and provide a foundation for future autonomous service robot deployment and human-robot interaction research.