TY - JOUR
T1 - Experimental Evaluation and Modeling of the Accuracy of Real-Time Locating Systems for Industrial Use
AU - Bisheh, Sepideh Valiollahi
AU - Rodriguez, Ignacio
AU - Zhang, Weifan
AU - E. Mogensen, Preben
PY - 2024
Y1 - 2024
N2 - Real-time locating systems (RTLSs) have proven to be a practical and effective solution for monitoring positions/status of humans and other entities in industrial environments, ensuring safe and efficient automated operations, by responding in real-time to unexpected events, such as the proximity of human workers to autonomous mobile robots (AMRs). This work focuses on evaluating and modeling the performance of two complementary RTLSs targeting human localization in both static and mobile conditions within a realistic industrial environment in operational conditions; with the aim of integrating live positioning information into digital twins (DT) for industrial use. Both the primary RTLS examined, an ultra-wide band (UWB) radio-based system; and the secondary RTLS, an optimized camera vision (CV)-based system with three surveillance cameras introduced as a backup RTLS, exhibited a similar median accuracy of 12-13 cm in static conditions, being the one of the UWB system slightly degraded down to 19 cm in presence of human shadowing. In human mobility conditions, the median accuracy values were further debased by 4 and 13 cm for the UWB and CV systems, respectively, indicating a limited real-time fluctuation, sufficiently bounded to guarantee the safety of human workers based on the readings of either of the primary or secondary RTLS systems. Based on the observed performance, a safety protocol for human detection in operational production scenarios was established, considering operational safety margins around humans of 1-2 m, which could be further leveraged by centralized monitoring and control entities such as industrial digital twins. The localization accuracy of the systems is characterized by means of error functions quantifying the distance to ground truth (GT) points through Gamma distribution functions using maximum likelihood estimates (MLEs). The proposed models are practical for implementation in system level simulators or industrial digital twin tools considering akin industrial environments. The different observations presented along the paper are useful for advanced industrial operation planning and optimization considerations.
AB - Real-time locating systems (RTLSs) have proven to be a practical and effective solution for monitoring positions/status of humans and other entities in industrial environments, ensuring safe and efficient automated operations, by responding in real-time to unexpected events, such as the proximity of human workers to autonomous mobile robots (AMRs). This work focuses on evaluating and modeling the performance of two complementary RTLSs targeting human localization in both static and mobile conditions within a realistic industrial environment in operational conditions; with the aim of integrating live positioning information into digital twins (DT) for industrial use. Both the primary RTLS examined, an ultra-wide band (UWB) radio-based system; and the secondary RTLS, an optimized camera vision (CV)-based system with three surveillance cameras introduced as a backup RTLS, exhibited a similar median accuracy of 12-13 cm in static conditions, being the one of the UWB system slightly degraded down to 19 cm in presence of human shadowing. In human mobility conditions, the median accuracy values were further debased by 4 and 13 cm for the UWB and CV systems, respectively, indicating a limited real-time fluctuation, sufficiently bounded to guarantee the safety of human workers based on the readings of either of the primary or secondary RTLS systems. Based on the observed performance, a safety protocol for human detection in operational production scenarios was established, considering operational safety margins around humans of 1-2 m, which could be further leveraged by centralized monitoring and control entities such as industrial digital twins. The localization accuracy of the systems is characterized by means of error functions quantifying the distance to ground truth (GT) points through Gamma distribution functions using maximum likelihood estimates (MLEs). The proposed models are practical for implementation in system level simulators or industrial digital twin tools considering akin industrial environments. The different observations presented along the paper are useful for advanced industrial operation planning and optimization considerations.
KW - Smart factories
KW - autonomous mobile robots
KW - camera-vision
KW - digital twins
KW - human workers
KW - localization error models
KW - real-time locating systems
KW - ultra-wideband
U2 - 10.1109/ACCESS.2024.3405393
DO - 10.1109/ACCESS.2024.3405393
M3 - Journal article
SN - 2169-3536
VL - 12
SP - 75366
EP - 75383
JO - IEEE Access
JF - IEEE Access
ER -