Naslov
Human Localization in Warehouse Environments based on a Wearable Camera Sensor Suite and Dynamic Ultra-Wide Band Nodes

Autori
Popović, Goran ; Cvišić, Igor ; Marković, Ivan ; Petrović, Ivan

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
International Conference on Advanced Robotics (ICAR) / - , 2021, 1-6

Skup
International Conference on Advanced Robotics (ICAR)

Mjesto i datum
Ljubljana, Slovenija, 07.12.2021. - 10.12.2021

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
indoor localization ; pose-graph optimization ; visual odometry

Sažetak
Modern warehouses, equipped with an autonomous robots fleet, suffer from an efficiency deficit during human interventions in the shop floor area, since, during such interventions, all the robots must stop for safety reasons. This particularly affects large warehouses which could extremely benefit from a solution that would allow human-robot collaboration during such interventions. This solution should only stop the robots that are in the vicinity of the human and reroute the others, and it requires the following prerequisites: (i) reliable relative ranging of robots with respect to human workers that would ensure safety, and (ii) reliable localization of human workers in the warehouse. In this paper we propose a system for reliable human localization that is based on fusing cues from measured relative ranges with respect to moving robots and a camera system that is worn by the human worker. Since global locations of the robots in the warehouse are known, we use the relative ranges to robots in an indoor GPS-like fashion to provide global human location updates, while the wearable cameras provide visual odometry, i.e., the relative motion of the human, and another global cue by detecting ground fiducial markers used by robots. The fusion is performed within a graph-based optimization framework that also includes a visual odometry error model. We carried out an experimental analysis on three different datasets to validate the performance of the proposed approach and compared it to a state-of-the-art visual SLAM solution, namely ORB-SLAM2. The results demonstrate that our approach yields reliable human location within warehouse like environments.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Računarstvo, Temeljne tehničke znanosti

POVEZANOST RADA