• The proposed system enables real-time localization calibration between robots without relying on fixed infrastructure. • A master–slave architecture with mobile UWB transmitters and EKF allows accurate localization correction through inter-robot collaboration. • Compared to LiDAR- and camera-based methods, the system achieves competitive accuracy with significantly lower computational cost and configuration cost. • Experimental results in both indoor and outdoor settings show a localization error reduction of over 91% per calibration on average. Large-scale multi-robot deployment in smart factories and logistics facilities improves coverage and throughput, but it exposes a critical cost–accuracy trade-off. Equipping every robot with high-end sensors (e.g., LiDAR/vision) can yield accurate SLAM, yet the fleet-wide cost becomes prohibitive. In contrast, low-cost robots are economically scalable, but odometry-driven localization suffers from drift due to floor friction variations, wheel slippage, and sensor noise, degrading efficiency and safety. This study proposes an infrastructure-independent UWB-based collaborative localization error calibration system that suppresses drift while preserving the cost advantage of low-cost fleets. The system adopts a master–slave architecture: a master robot with two UWB transmitters (mobile anchors) shares its pose, and a slave robot measure ranges to the transmitters, estimates its position using a baseline-constrained geometric solution, and fuses it with odometry via an Extended Kalman Filter. Hardware experiments in indoor and outdoor environments demonstrate effectiveness: the method achieved 6.8 cm RMSE indoors (LiDAR: 4.2 cm; camera: 5.6 cm) and reduced the slave error to 2.3 cm RMSE outdoors when RTK-GNSS provided the master reference. The average processing time per calibration was 12.6 ms, confirming real-time feasibility.
Yoon et al. (Wed,) studied this question.