A fusion approach for the localization of humans in factory environments

Abstract

Industrial automation increasingly relies on the automation and interconnection of formerly human operated transport machinery. Products are already moved by automated guided vehicles and machines like cranes may soon position themselves just in time based on context information. Human presence is however still required in some of these environments. This raises the question how to allow machines to sense a human's presence to enable context awareness functions and safety around autonomous machines. We propose a combination of Ultra-Wide Band (UWB) localization and Pedestrian Dead Reckoning (PDR) to achieve robust and reliable long- and short-term tracking of humans in an industrial environment. We highlight advantages and drawbacks of both localization methods, as well as current approaches to increase tracking accuracy of the individual technologies and their combination. Additionally, we present measurements using a Zero Velocity Update assisted PDR system and show that, under certain circumstances, the error of PDR can be reduced from 5% to 2% of the travelled distance. We also analyze the applicability of PDR correction methods in industrial environments. Our results show the potential of PDR for accurate short-term localization, supplanted by UWB to ensure long-term precision.