Fully Autonomous Wheelchair for Indoor Mobility: A Hybrid Approach to Mapping and Navigation with ROS

Abstract

People with severe physical disabilities often find existing mobility solutions inadequate, necessitating the need for reliable autonomous systems. This paper introduces a ROS-based autonomous wheelchair that utilizes the Kinect Xbox sensor for visual perception, RTAB-Map for mapping, and AMCL for localization. Additionally, it incorporates additional sensors that supply odometry and obstacle detection data to the ROS navigation stack, enabling precise path planning and real-time obstacle avoidance. Initial tests using the Gmapping SLAM algorithm revealed synchronization and accuracy issues. However, a hybrid RTAB-Map and AMCL approach resolved these issues, delivering superior detail in the maps and reliable navigation indoors. The system consistently reached its target destinations, proving robustness and accuracy. The Qt framework-built user-friendly interface enables destination selection, and AI-powered voice recognition integration is currently in progress to facilitate hands-free control.