Towards the Development of Autonomous Wheelchair

Abstract

Precise and fast indoor navigation represent well-known challenges that, if solved, can improve many aspects of our lives. In this paper we propose a complex system architecture dedicated to precise indoor navigation in unknown environments with the final purpose of developing an autonomous electric wheelchair. The desired outcome is to use computer vision to detect floors and doors with high accuracy and speed and compute a safe path to be followed. Our solution is based on fully convolutional neural networks (FCN) for image segmentation of corridor images to identify the floor area further used for navigation.