Technical advances are making it possible to deploy increasingly large numbers of sensors to monitor the physical world. These deployments give rise to exciting new applications, but they also raise great data management challenges.
In this seminar, we will read papers presenting the main types of sensor deployments and application domains and the key data management challenges that need to be addressed to enable the "World-Wide Sensor Web".
This seminar is inspired by the UW MSR Institute on the World-Wide Sensor Web that took place in August 2006 in Blaine, WA. http://www.cs.washington.edu/mssi/2006/index.html
Day | Topic(s) | Presenter(s) |
---|---|---|
02 October | Introduction and paper assignment | Magda Balazinska |
09 October | Sensor networks data management In order for everyone to be on the same page, we will first review one of the main systems for query processing in sensor networks.
TinyDB: An Acquisitional Query Processing System for Sensor Networks.
Additional references:
|
YongChul Kwon |
16 October | Application domain 1: Scientific data management Increasingly, scientists need to deploy and share common sensor infrastructures to monitor and study the physical world. We will discuss two short papers that present some of the challenges of scientific sensor data management.
Life Under your Feet: A Wireless Soil Ecology Sensor Network.
Batch is Back: CasJobs, Serving Multi-TB Data on the Web
. Additional references:
|
Nicholas Murphy |
23 October | Application domain 2: Monitoring cities and homes Different from scientific deployments are sensor networks designed to monitor humans in their daily activities. These networks can be deployed at the scale of a city or at the scale of individual homes.
CarTel: A Distributed Mobile Sensor Computing System.
An Assisted Living Oriented Information System Based on a Residential Wireless Sensor Network.
Additional references:
|
Brian DeRenzi |
27 October |
Spatio-temporal data management [Note: Rescheduled to Friday, October 27th at 2pm in CSE 605] Sensor data requires powerful spatio-temporal data management capabilities. The spatio-temporal research area is quite wide. We will only discuss a specific example based on RFID data.
Temporal Management of RFID Data. |
Evan Welbourne |
06 November | Sensor data cleaning (RFID data) An important challenge when dealing with sensor data is handling erroneous inputs due to noise and sensor failures. We start the discussion by examining how to clean data produced by an RFID infrastructure.
Additional references:
|
Nodira Khoussainova |
13 November | Model-based sensor data management Continuing on the idea of noisy sensor data, we will discuss the use of correlations and models to handle uncertainty, missed reading, outliers, and more.
Using Probabilistic Models for Data Management in Acquisitional Environments. Additional references:
|
Chris Re Abhay Kumar Jha |
20 November | Internet-scale infrastructures We do not spend too much time on infrastructures because most of us studied them last Winter. Here are two good overview papers that we can use to discuss push-based vs pull-based infrastructures and their challenges.
The Design of the Borealis Stream Processing Engine. |
Stefan Ekerfelt Hanna Filipsson |
27 November | Data integration and lineage The world-wide sensor web requires data integration across autonomous organizations. Data integration is one of the main database challenges today. There isn't any work done specifically on integrating sensor data. Instead, we will discuss the more general problem of scientific data integration.
Additional references:
|
Mike Cafarella Roxana Geambasu |
04 December | Security and privacy Finally, monitoring the physical world raises serious privacy and security issues.
RFID Security and Privacy: A Research Survey.
Security and Privacy in Sensor Networks |
Nilesh Dalvi Jeffrey Bigham |
Please sign up for the course mailing list here. Send mail to that list at cse590q at cs