Sistema de navegação autônoma de cadeira de rodas motorizada em ambiente conhecido

Abstract

There is a large number of people who are facing difficulties in their daily lives, due to physical limitations, this reality motivates several researches focused on assistive technologies, which are resources used to expand functional abilities of people with disability, promoting better quality of life. In this context, this work presentes an autonomous navigation system of a wheelchair between corridors and doors in a known indoor environment. This system uses a Dijkstra algorithm to generate a trajectory from a grid map of the place, and the destination selected by the user through a laptop interface. The relative position of the wheelchair during the trajectory is estimated through Odometry using encoders on the wheels, accelerometer and gyroscope. In addition of Odometry, an absolute location system, by artificial landmarks, is applied, which uses RGB-D images captured by a Kinect v2 camera, to determine the position of the wheelchair by locating rectangles placed in the environment in previously known positions. The location of the rectangles is performed by applying the Canny edge detector and the probabilistic Hough transform to extract lines in RGB images, and through the information of image depth, it is verified which lines belong to the landmark. The depth images captured by Kinect are also used for the process of determining the position of the wheelchair in relation to the center of the door to be passed. The depth images captured by Kinect are also used for the process of determining the position of the wheelchair in relation to the center of the door to be passed. The door detection consists of locating its edges through the variation of the depth between the wall and its gap. The control of the wheelchair's trajectory in the dislocation between corridors, and in the passage through doors it is performed using controllers, based on fuzzy logic that act on the linear and angular velocity of the wheelchair. After carrying out the tests, it was observed that there is a strong potential for research in the application of odometry together with Kinect in the autonomous navigation of mobile robots. However, the absolute location by artificial landmarks during the dislocation still performs unreliable, once that there are a lack of synchronism between the capturing the RGB images of depth. [resumo fornecido pelo autor]