Efficient, Robust RFID Stream Processing for Tracking and Monitoring
University of Massachusetts, Amherst
Recent advances in Radio Frequency Identification (RFID) technology and ubiquitous networking are facilitating the emergence of an information infrastructure that collects real-time data associated with physical objects and delivers high-value content to a variety of user communities. Emerging user communities include supply chain management, healthcare, postal services, to name just a new.
Data stream management is central to such an RFID-based information infrastructure---it allows the relevant information to be sifted out of the flood of RFID data immediately after it emerges. Despite recent advances in related areas such as relational stream processing and sensor data management, RFID data---inherently noisy data for identification of individual objects---raises many new questions. The significant mismatch between raw RFID data and meaningful, actionable information required by RFID applications requires complex processing beyond the capabilities of existing stream systems. The incomplete and noisy nature of RFID data further complicates such data-information translation. The volumes of data generated from large RFID deployments can also stress or overwhelm existing stream systems.
The goal of this research project is to design and develop an efficient, robust stream processing system that addresses the challenges posed by the data-information mismatch, incomplete and noisy data, and high data volumes, and enables real-time tracking and monitoring. This project has several main components.
- SPIRE: inference and compression over RFID streams . This low-level substrate offers accurate interpretation of incomplete and noisy raw data. In particular, it infers locations of unobserved objects and inter-object relationships using probabilistic algorithms. To handle high data volume, it performs online interpretation, enabling online compression by identifying and discarding redundant data. (Project completed.)
- CLARO: probabilistic relational query processing over streams. After inference over raw RFID data streams, our system then evaluates relational queries on tuple streams that carry probability distributions to describe input data uncertainty. We focus on data that is naturally modeled as continuous random variables, such as object locations inferred from RFID raw readings. Our system employs a unique data model to capture a wide variety of data uncertainties and efficient techniques to evaluate relational operators by exploring advanced statistical theory and approximation.
- SASE: complex event processing over streams. This higher-level event processing system addresses the data-information mismatch by encoding application information needs as temporal patterns and evaluating these patterns continuously over event streams. The SASE system offers a compact, expressive event language, automata-based mechanisms for efficient event pattern evaluation, and advanced techniques for robust event processing.
- SCALLA: scalable one-pass analytics on high volumes of data. Recently, MapReduce has emerged as a popular programming model for processing large datasets using a cluster of machines. However, the MapReduce model is geared towards batch processing and requires the data set to be fully loaded into the cluster before running analytical queries. In this project, we examine, from a systems standpoint, what architectural design changes are necessary to bring the benefits of the MapReduce model to fast one-pass analytics. Our work includes theoretical and empirical analyses of existing MapReduce systems and the proposal of a new data analysis platform that employs advanced hashing and frequency analysis to enable scalable fast one-pass analytics.
Project Members
Team at UMass Amherst
- Yanlei Diao (faculty)
- Neil Immerman (faculty)
- Andrew McGregor (faculty)
- Prashant Shenoy (faculty)
- Liping Peng (grad)
- Haopeng Zhang (grad)
- Boduo Li (grad)
- Abhishek Roy (grad)
- Yunmeng Ban (grad)
- Wenzhao Liu (grad)
Collaborator(s)
- Charles Sutton (University of Edinburgh)
Alumni
- Thanh Tran (Twitter)
- Ed Mazur (Google)
- Zhao Cao (HP Labs)
- Yanming Nie (Chinese Northwest A&F University)
- Richard Cocci (Harvard Law School)
- Jagrati Agrawal
- Daniel Gyllstrom
- Hee-Jin Chae
- Eugene Wu (MIT)
Sponsor
National Science Foundation | |
CAREER: Efficient, Robust RFID Stream Processing for Tracking and Monitoring.
Yanlei Diao (PI).
National Science Foundation IIS-0746939.
Award abstract.
This grant supports our research on both the low-level inference and compression over RFID streams and the high-level complex event processing. Any opinions, findings, and conclusions or recommendations expressed at this web site are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. |
Last Updated: July 29, 2014