Laser rangefinder and monocular camera data fusion for human-following algorithm by pmb-2 mobile robot in simulated gazebo environment

Abstract

The paper presents a human-following algorithm for an autonomous mobile robot, which is equipped with a 2D laser rangefinder (LRF) and a monocular camera. As a rule, quality of a human tracking by a LRF is reduced in cluttered environments. We used a monocular camera to increase a human-tracking reliability. In contradiction with popular human-tracking algorithms that apply only a 2D LRF, our algorithm does not impose any restrictions on a type of human’s clothes, and our approach does not require a human head and an upper body to be located within a monocular camera field of view. Several human trackers and variations of our algorithm were compared in the Gazebo virtual experiments within a free corridor and an office room environment. The virtual experiments demonstrated that our method successfully improved a human-tracking quality being employed with the human-following virtual PMB-2 robot.