PCL-Based Autonomous Wheelchair Navigation in Unmapped Indoor Environments

Abstract

Autonomous navigation in an unmapped indoor environment involves several challenges. Notable among these is the detection of various obstacles encountered in the navigation environment. Recent developments in image processing and computer vision have allowed for the detection of objects using different image processing algorithms. Point Cloud Localization (PCL) is one such method. Using PCL, an image is divided into a grid with a distance value is assigned to each cell. This method enables a robotic system to estimate distances to various objects it "sees" in its environment. This paper presents an autonomously navigating wheelchair which operates in an unmapped indoor environment using Point Cloud Localization data. The wheelchair uses a LIDAR, a Stereo Camera, and an Ultrasonic Proximity sensor to achieve collision-less navigation.