|
2004–2005 Sensor Networks Projects
Sensor Networks for Human Activity Inferencing
Gaetano Borriello
University of Washington
We seek to develop blended sensor networks, that combine sensors carried by a person with those they may encounter in the environment, to determine a user’s context and thereby cause applications to adjust their behavior appropriately. The technical aspects of our work include the development of: wrist-watch and cell-phone platforms for the devices the user carries to be used both for sensing as well as user interface; middleware to fluidly adjust what computation is done and where so as to minimize power consumption and conserve bandwidth while operating within performance constraints; and discovery and binding algorithms for connecting newly encountered sensors to a user application.
Tiered: Development of Tiered Applications Across Motes and Micro-Servers
Deborah Estrin
University of California at Los Angeles
As our experiences with deployed sensor network instruments increase, we have found that the basic unit of operation is a microserver, or master, and its associated flock of motes. The flock consists of a self-configuring shallow multihop tree of motes with a microserver as root. The tree is shallow to avoid long wireless paths that infringe on duty cycling and lead to congestion. Larger area coverage is achieved by deploying multiple such flocks, with the resulting need for coordination among the microservers in order to run applications across the larger mote herd. Currently, the software is readily available to support simple tree building and data extraction to a microserver, but there is no reusable software for construction of applications across a mote herd. This is critical because the collection of motes within a flock is determined by wireless connectivity, and not logical affiliation. Consequently, application level functions are likely to cross flock boundaries.
This research effort would advance current practice by creating public domain, publicly accessible and modifiable, software that implements middleware services for these tiered sensor networks. The middleware services would facilitate both the construction and evaluation of sensor network applications, and the exploration of next generation sensor network algorithms and mechanisms.
Sensor Networks & Web Services
Joe Hellerstein
University of California at Berkeley
Query processing is proving to be an attractive high-level interface for tasking clouds of wireless sensors. In recent years this field has blossomed in the research community, with proposals for a variety of query processing paradigms. However, the only widely available software is our own TinyDB system, which provides an implementation of only one such paradigm, and is difficult to adapt in any significant way.
We propose to design and implement a unified sensornet query processing architecture that can be used to achieve a variety of query processing paradigms. The architecture will be flexible with regard to route construction, communication patterns for in-network processing, scheduling of sensing, communication and sleep. The results of this work should include insight into the architectural fundamentals of sensornet query processing; better understanding of the practical tradeoffs between proposed techniques for sensornet query processing; novel techniques that may arise from the enhanced flexibility of the architecture; and an open and more flexible platform for the research community to explore sensornet query processing, data acquisition, and distributed intelligence.
Web Service
Protocols for Self-Monitoring and Self-Healing Networked Embedded Sensor
Systems
Xenofon Koutsoukos
Vanderbilt University
Reconfiguration and self-adaptation are vital capabilities of networked embedded sensor systems that are required to operate in dynamic environments. Runtime technologies that allow software to evolve as system requirements change are critical to the development and deployment of such systems. The proposed research activities aim at developing adaptive software by establishing a novel direction in software composition for networked embedded sensor systems based on Model-Integrating Computing and Web services. We will investigate scalable solutions that enable reconfiguration in large sensor networks based on distributed algorithms for constraint satisfaction problems. We will develop efficient reconfiguration architectures based on Web services and we will demonstrate the research advances using an experimental test-bed that we have built in the Hybrid and Networked Embedded Systems lab at Vanderbilt University. Furthermore, we will disseminate the results we develop in open-source format available for non-commercial use.
Wireless Sensor Networks for Soil Ecosystem Studies
Katalin Szlavecz
Johns Hopkins University
The proposed research will customize, test, and deploy a network of low-cost wireless sensors to monitor the soil and aboveground conditions along an urban-rural gradient. The data will be collected automatically and uploaded into an online, publicly available database. The project will augment ongoing research in the Baltimore Ecosystem Study (BES), which is part of the NSF funded LTER (Long-Term Ecological Research) network. Elements of the system (wireless modules, multi sensor boards, environmental monitoring boards, Ethernet interface board) are manufactured by Xbow. We also add soil moisture sensors, and CO2 sensors and a few light sensors to the system. The first steps of the research are verify the stability and the accuracy of the sensors, develop a prototype weather-proof enclosure for the motes, calibrate the sensors in a climate chamber, begin development of the data management system, and test the wireless network on 1-20 m scale by deploying it at JHU campus.
We have obtained funds from the Gordon and Betty Moore Foundation to support a graduate student, but those funds cannot support hardware. About $20K would enable us to purchase the first 30 sensor units and build the waterproof enclosures.
Integrating Sensor Networks into Medical Care Using Web Services
Matt Welsh
Harvard University
We propose to develop a Web services-based infrastructure for integrating wireless sensor networks into medical care settings. Sensor networks have the potential to greatly benefit many aspects of medical care, allowing many patients to be continuously monitored using wearable, wireless vital sign sensors. An important challenge that arises in this domain is the integration of real-time sensor data into other information systems, such as hospital patient records and 911/emergency dispatch services. By leveraging Web Services standards, we intend to develop a scalable infrastructure for collecting, filtering, and routing medical sensor data to a broad range of Internet-based applications. This system will be based on the Harvard CodeBlue and Hourglass platforms. All source code will be released under an open license and course materials will be contributed to the MSDN Academic Alliance Curriculum Repository.
back to top
|
