Human identification and tracking using ultra-wideband-vision data fusion in unstructured environments

Abstract

Nowadays, the importance of working in changing and unstructured environments such as logistics warehouses through the cooperation between Automated Guided Vehicles (AGV) and the operator is increasingly demanded. The challenge addressed in this article aims to solve two crucial functions of autonomy: operator identification, and tracking. These tasks are necessary to enable an AGV to follow the selected operator along his path. This paper presents an innovative, accurate, robust, autonomous, and low-cost operator real-time tracking system, leveraging the inherent complementarity of the uncertainty regions (2D ellipses) between ultra-wideband (UWB) transceivers and cameras. The test campaign shows how the UWB system has higher uncertainty in the angular direction. In contrast, in the case of the vision system, the uncertainty is predominant along the radial coordinate. Due to the nature of the data, a sensor fusion demonstrates improvement in the accuracy and goodness of the final tracking.