Autonomous robotic aerial tracking, avoidance, and seeking of a mobile human subject

Abstract

This paper presents a methodology to achieve Robotic Aerial Tracking of a mobile – human – subject within a previously-unmapped environment, potentially cluttered with unknown structures. The proposed system initially employs a high-end Unmanned Aerial Vehicle, capable of fully-autonomous estimation and flight control. This platform also carries a high-level Perception and Navigation Unit, which performs the tasks of 3D-visual perception, subject detection, segmentation, and tracking, which allows the aerial system to follow the human subject as they perform free unscripted motion, in the perceptual – and equally importantly – in the mobile sense. To this purpose, a navigation synthesis which relies on an attractive/repulsive forces-based approach and collision-free path planning algorithms is integrated into the scheme. Employing an incrementally-built map model which accounts for the ground subject’s and the aerial vehicle’s motion constraints, the Robotic Aerial Tracker system is capable of achieving continuous tracking and reacquisition of the mobile target.