Human detection and tracking with knee-high mobile 2D LIDAR

Abstract

Human-robot interaction and safe operation requires human detection capability from a mobile robot. This paper presents a novel algorithm for detecting and tracking walking humans with a 2D LIDAR (Light Detection and Ranging) on a mobile robot. The horizontal LIDAR receives reflections of human legs taken from approximately the knee level. A list of predefined features are used for classifying the reflections as clusters, which are then tracked separately. If two clusters fulfill predefined conditions to be classified as legs, and if they are close enough, a human target is generated and tracked using the reflections of the legs. The probability of a human target to actually be a human is constantly evaluated based on numerous features. Experimental results performed with a mobile robot verify that the algorithm is able to detect and track people in typical indoor environment.