Using Dynamic Time Warping to Identify RFID Tag Movement in a Logistics Scenario with and without Additional Process Knowledge

In recent years, Radio Frequency Identification (RFID) has been widely adopted as a solution for the fully automatic identification of physical objects. However, technological and physical constraints may sometimes hamper a productive use. One of the key issues is the problem of false-positive RFID tag reads, that is, RFID transponders that are detected unintentionally by the reader hardware. The present paper proposes and evaluates a comprehensive time-series analysis technique to identify and filter false-positives from the RFID data stream. Furthermore, we investigate the value of additional knowledge about the business process to be monitored. We empirically test our approach using a large set of RFID data collected at the distribution center of a large European retailer.