Reference cage architecture for autonomous docking of mobile robots in automotive production systems

This paper addresses the need for high-accuracy docking in material handling processes using autonomous mobile robots in factories. To satisfy this need for the tasks, such as loading, unloading, and reaching the charging station, traditional navigation methods often rely on physical restraints at the docking stations, which limits flexibility in production lines. To achieve high-level accuracy without such restrictions, this study proposes the reference cage architecture, which utilizes multi-reference points to maintain scan-matching-based localization performance during docking. The contributions of this research include achieving sub-centimeter accuracy in pose estimation near the target pose and the development of a real-time reference selection decision mechanism. To verify the effectiveness of the proposed approach, extensive testing and validation have been conducted on the automotive production lines of the Ford Otosan Golcuk Plant. These tests consider real-world operational conditions, such as noises, disturbances, and outliers, setting this study apart from similar publications in the literature. The results demonstrate the potential of the reference cage architecture in enabling high-accuracy docking in autonomous mobile robot applications within factory environments.