Radio Frequency Identification (RFID) technology is gaining acceptance in an increasing number of applications for tracking and monitoring purpose. Despite its promise to provide unprecedented visibility in various domains, RFID technology presents numerous challenges:

  • Incomplete and noisy data:
    • Read rates of RFID readers in actual deployments are significantly below 100%, due to the intrinsic sensitivity of radio frequencies (RFs) to environmental factors such as interference from nearby metal objects and contenttion among tags.
    • Furthermore, mobile readers have lower read rates than fixed readers becauses they tend to read objects from arbitrary orientatuons, and certain orientations can result in poor read rates.
  • Insufficient information: Raw RFID readings only containtag identifications and do not contain additional high-level information such as object locations, containment and co-location relationships. Such information, however, is important to object tracking and monitoring, e.g., to ensure that perishable food is contained in a cooling box, flammable objects are secured in a fire-proof container, and foods with and without pernuts are not packaged in the same container.
  • High volume with redundancy: Large deployments could create high volumes of data, e.g., over terabytes of data in a single day. Such data, however, may encode significant amounts of redundant information such as an unchanged object location. Hence, it is crucial that data be filtered and compressed close to the hardware while preserving all useful information.


In this project, we design and develop an efficient inference, compression and query processing system over RFID data streams. Our system provides the following functionalities:

  • Inference: It provieds accurate interpretation of incomplete and insufficient raw data; in particular, it infer locations of unobserved objects and inter-object relationships such as collocation and containment.
  • Compression: To handle high data volumes, SPIRE performs online interpretation, enabling online compression by identifying and discarding redundant data close to the hardware..
  • Query processing: Our system further provides a query processor that runs over the infered event stream and returns application-specific high level information such as complex events or anomalies.

We gratefully acknowledge the funding provided by National Science Foundation.

ww National Science Foundation

CAREER: Efficient, Robust RFID Stream Processing for Tracking and Monitoring. Yanlei Diao (PI). National Science Foundation IIS-0746939.

III-COR-small: Capturing Data Uncertainty in High-Volume Stream Processing. Yanlei Diao (PI) and Anna Liu. National Science Foundation IIS-0812347.