Akademik Veri Yönetim Sistemi
Expert Systems with Applications, cilt.267, 2025 (SCI-Expanded)
Autonomous mobile robots are one of the critical technologies in the realization of Industry 4.0. They are used in smart factories to perform transport tasks in internal logistics. Positioning system accuracy of robots is important for task efficiency. So, it is important to monitor positioning systems and detect possible anomalies without human intervention. In this study, an online learning-based anomaly detection approach is proposed for positioning systems of autonomous mobile robots. This study uniquely combines the use of both reinforcement learning and control charts in online learning. Within the scope of reinforcement learning, the Upper Confidence Bound (UCB) algorithm was used together with statistical calculations to instantly evaluate the positioning data, and labeling of the unaffected data was carried out. In this approach, the data that are obtained from three different position calculation methods in an autonomous mobile robot, are learned with the UCB algorithm. And then possible abnormal changes in the reward of the algorithm are determined with the proposed Statistical Process Control approach. The development of the formulation for the creation of study-specific breakpoints is also included in the creation of the control chart. The proposed approach has been tested with autonomous mobile robot in laboratory.