Autonomous navigation with simultaneous localization and mapping in/outdoor

Abstract

The necessity and use of wheelchairs are increasing with the constant rise of the world population and human needs.Admittedly, the advancements in technology over the years have helped to push forward from manual to electrical powered wheelchairs.However, these are still inaccessible to many individuals with specific disabilities.With the technological improvements over the years, electrical powered wheelchairs have helped filled the restrictions that come with manual wheelchairs, but only to an extent.Furthermore, the impact of intelligent wheelchairs can be recognized by allowing access to them to a more significant number of users, as they integrate smart and autonomous systems capable of behaviors such as obstacle detection and avoidance, apprehension of objects and autonomous navigation.Besides, given the complexity intelligent wheelchairs can obtain, it reaches a point where it is not feasible to properly test the system completely without any risks involved.One of the most studied areas in the field of mobile robotics is the SLAM problem, where the robot achieves a certain level of perception of its surroundings with the use of external sensors like cameras and lasers.Therefore, this dissertation explores a platform for a wheelchair capable of simultaneous localization and mapping (SLAM) with autonomous navigation, applied on a simulation environment.The simulation allowed to test the wheelchair in different situations with a distinct set of sensors, as to conclude what achieves better results for indoor and outdoor environments.With the tests done it was possible to conclude that only sensors like the ZED 2 or the RPLidar A3 are capable of obtaining reliable results in all use case scenarios studied.Additionally, the system is part of the IntellWheels 2.0 project, an intelligent wheelchair platform that strives to be capable of being easily adapted to any commercial electric wheelchair, aiding any person with any disabilities with different methods of inputs and a robust multimodal interface.