Mixed-reality simulation of RFID systems using iterated learning

摘要

Radio Frequency Identification (RFID) is a key enabling technology for the Internet of Things. Like other information systems, RFID systems evolve to meet new requirements. The behavior (e.g. tag-reading performance) of RFID systems is closely coupled with software, hardware and its deployment environment. An RFID system needs to be rigorously tested before its deployment and re-tested each time when there are changes in software, hardware or environment. The cost of testing an RFID system can be significantly lowered by substituting its hardware and deployment environment with a simulator. However, radio waves can be affected by many environmental factors such as multi-path effects, backing materials and electro-magnetic noises. The effect of these factors on radio waves cannot be precisely modeled algorithmically. In this paper, we propose an effective technique to construct a mixed-reality simulator for RFID system testing using iterated learning based on a Support vector Machine (SVM). We applied our simulator to a real-life RFID object tracking system, which keeps track of the stock items entering or leaving a warehouse. Evaluation results show that our mixed-reality simulator can achieve 90% prediction accuracy with 90% reduction in both the time and storage space as compared with performing the tests using real hardware and environment.