Multi-dimensional sensing with computational RFID tags

Abstract

Radio Frequency Identification (RFID) technology is one of the key technologies to support automatic identification in many ubiquitous applications. An RFID tag compromises of an integrated circuit and antenna, which enable to wirelessly communicate with the RFID reader in a non-line-of-sight way. The RFID tag, especially the passive tag, has constrained capabilities in computation, communication, and storage, due to the extremely low production cost. For instance, passive tags can only report their unique IDs to the reader for identification. On the other hand, modern applications usually require more rich information for better management. For example, the system requires a timely detection on the abnormal event, including the improper leaning, rolling, or accidently falling off the conveyer, etc. Recently, researchers develop a new technology equipping the tag with sensors and more powerful computational capacity, termed as Computational RFID (CRFID). CRFID tag is able to sense the temperature, humidity, and acceleration of objects, etc. Hence, they can provide for fine-grained monitoring and control over targeted objects. In this thesis, I have focused on detecting abnormal events of targets based on the multi-dimensional sensing of CRFID tags. Furthermore, I have developed practical methods to support the fine-grained trajectory management. A detail description of this technology will be presented in my thesis. Compared to tags, the RFID reader is very expensive. It would be cost-inefficient, sometimes impractical, if fully-covering the entire logistic processing flow using readers. To detect the abnormal events along the whole processing flow, we propose to attach CRFID tags to the objects and deploy RFID readers only in critical areas. We design a tree-indexed Markov Chain scheme, which leverages statistical methods to achieve fine-grained abnormal event detection and dynamic trajectory management. We have implemented a prototype system onto a passenger luggage handling system and conduct extensive experiments. The result shows that our system can effectively detect the anomalous event with low cost and high accuracy. On the other hand, collaboratively sensing multi-dimensional information is important when processing or transporting the items with server requirements on the environment and condition. For example, chemical shipment has a rigorous constrain on both the temperature and vibration. Thus, multi-dimensional information, such as the temperature and vibration, should be collaboratively sensed for comprehensive analysis and detection on the abnormal event. We propose a multi-dimensional information based surveillance scheme to timely detect the abnormal event that occurs in cold chain logistics. Cold chain is a supply-chain which maintains the important parameters, e.g., the temperature or vibration, in a given range. We adopt the minimum entropy of accelerometer readings and classification algorithms to distinguish various statuses of items. Through multi-dimensional information processing, the scheme can gain a guaranteed accuracy with low false alarm rate and false detection rate. We also demonstrate the effectiveness of the framework by performing a preliminary implementation and trace-driven simulations.

Date of Award	2013
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology