MOVING: A MOdular and Flexible Platform for Embodied VIsual NaviGation

Marco Rosano, Francesco Ragusa, Antonino Furnari, Giovanni Maria Farinella

Abstract: We present MOVING, a flexible and modular hardware and software platform for visual mapping and navigation in the real world. The platform comprises a flexible sensor configuration consisting of an RGB-D camera, a tracking camera for odometry, and a 2D Lidar, along with a compact processing unit that is equipped with a GPU for running deep learning models. The software is based on ROS, utilizing the RGB-D RTAB-Map SLAM system for mapping and localization and the move base package for path planning and robot movement control. The platform is easily detachable and can be installed on any robot with minimal adaptation required, enabling the reuse of the same robotic software regardless of the robot employed. The effectiveness of the proposed platform was verified through mapping sessions of a large indoor environment, leveraging a Loomo robot. The proposed platform can represent a reasonable solution to speed up the design and testing of new software for autonomous navigation systems, minimizing deployment time in the real world.