Autonomous Human Tracking using UWB sensors for mobile robots: An Observer-Based approach to follow the human path

Assistance robots are rising as a promising solution to help workers in everyday life, for many applications. In particular, many outdoor tasks, such as manual harvesting or carrying heavy loads, may benefit from the use of a mobile robot following a human worker. The literature offers various frameworks and systems for target tracking, using several kinds of sensors and control algorithms. Most of them are based on vision, which is sensitive to lighting conditions, or LIDAR. Both of these sensors require to keep the human in line of sight to be tracked, and have limitations in some challenging environmental conditions encountered outdoors, such as in rough weather or in the presence of airborne dust or fog. Moreover, existing approaches are often focused on direct tracking and may lead the robot to cross the human trajectory. This is penalising when considering off-road assistance, especially in agriculture, where the environment is often narrow (because of crop rows), and require many obstacle avoidance situations. In this paper an algorithm based on Ultra-Wide-Band technology is proposed in order to ensure a relative localisation without the need for a direct view of the target. The method approximates locally the trajectory of the human leader as a circle in order to follow the path achieved by the human, without direct communication nor absolute localization system. This is made possible thanks to the on-line reconstruction of the leader's variables (namely, its velocity and angular course), using a state observer. The effectiveness of this adaptive approach is demonstrated through full-scale experiments.