Navegação assistiva de cadeira de rodas em passagem por portas com o uso do sensor Kinect

Abstract

There is a large number of people who face difficulties in their daily lives due to the limitations imposed by their physical condition. This estimate motivates research focused on assistive technologies, all for autonomy and functional independence in their social and family interactions. Bearing this scenario in mind the work presents an assistive navigation system for a motorized wheelchair, in the task of safely crossing either doors or narrow passages. The system detects and locates doors through the acquisition of depth images obtained by a RGB-D Kinect v2 camera, and with this information it controls the trajectory until the crossing is completed. The techniques present in this work are based on research of works already implemented, drawing on concepts about robot navigation and depth sensors The technique implemented in this work consists of detecting a door through the difference in depth between the wall and its span, identifying its left and right ends. In this way, the location of the wheelchair is determined in x, y coordinates and an angle of orientation in relation to a point of passage in the center of the door and from there the trajectory is controlled to this point. The trajectory control process uses a fuzzy-logic based algorithm acting on the angular speed of the wheelchair, controlling its position along the trajectory, and thus performing the movement in a smooth manner until it passes through the detected door. Functional tests of the door detection algorithm were performed in order to validate the operation and establish its limitations. Subsequently, simulations were performed from different initial positions of the wheelchair, checking whether the system is able to identify the door and perform the collision-free movement. Finally, testing of the system implemented in a controlled environment was performed, with the chair arranged in various positions in front of a door, in order to evaluate the performance of the detection system and location of the door as well as the path taken by the fuzzy controller. The system was able to detect doors and control the movement of the wheelchair to the point of passage, even if the trajectory performed by the chair has a deviation from the simulated trajectory, the controller was able to correct the movement without significantly impacting the final result (sic).