HR: 08:45h
AN: SF51A-04    [Abstracts]
TI: On-Board Mining in the Sensor Web
AU: * Tanner, S
EM: stanner@itsc.uah.edu
AF: Information Technology and Systems Center, University of Alabama in Huntsville, Huntsville, AL 35899 United States
AU: Conover, H
EM: hconover@itsc.uah.ede
AF: Information Technology and Systems Center, University of Alabama in Huntsville, Huntsville, AL 35899 United States
AU: Graves, S
EM: sgraves@itsc.uah.edu
AF: Information Technology and Systems Center, University of Alabama in Huntsville, Huntsville, AL 35899 United States
AU: Ramachandran, R
EM: ramachan@itsc.uah.edu
AF: Information Technology and Systems Center, University of Alabama in Huntsville, Huntsville, AL 35899 United States
AU: Rushing, J
EM: jrushing@itsc.uah.edu
AF: Information Technology and Systems Center, University of Alabama in Huntsville, Huntsville, AL 35899 United States
AB: On-board data mining can contribute to many research and engineering applications, including natural hazard detection and prediction, intelligent sensor control, and the generation of customized data products for direct distribution to users. The ability to mine sensor data in real time can also be a critical component of autonomous operations, supporting deep space missions, unmanned aerial and ground-based vehicles (UAVs, UGVs), and a wide range of sensor meshes, webs and grids. On-board processing is expected to play a significant role in the next generation of NASA, Homeland Security, Department of Defense and civilian programs, providing for greater flexibility and versatility in measurements of physical systems. In addition, the use of UAV and UGV systems is increasing in military, emergency response and industrial applications. As research into the autonomy of these vehicles progresses, especially in fleet or web configurations, the applicability of on-board data mining is expected to increase significantly. Data mining in real time on board sensor platforms presents unique challenges. Most notably, the data to be mined is a continuous stream, rather than a fixed store such as a database. This means that the data mining algorithms must be modified to make only a single pass through the data. In addition, the on-board environment requires real time processing with limited computing resources, thus the algorithms must use fixed and relatively small amounts of processing time and memory. The University of Alabama in Huntsville is developing an innovative processing framework for the on-board data and information environment. The Environment for On-Board Processing (EVE) and the Adaptive On-board Data Processing (AODP) projects serve as proofs-of-concept of advanced information systems for remote sensing platforms. The EVE real-time processing infrastructure will upload, schedule and control the execution of processing plans on board remote sensors. These plans provide capabilities for autonomous data mining, classification and feature extraction using both streaming and buffered data sources. A ground-based testbed provides a heterogeneous, embedded hardware and software environment representing both space-based and ground-based sensor platforms, including wireless sensor mesh architectures. The AODP project explores the EVE concepts in the world of sensor-networks, including ad-hoc networks of small sensor platforms.
UR: http://eve.itsc.uah.edu
DE: 9820 Techniques applicable in three or more fields
DE: 6979 Space and satellite communication
DE: 3360 Remote sensing
DE: 3394 Instruments and techniques
DE: 1630 Impact phenomena
SC: Special Focus: Advances in Data Acquisition, Management, Analysis and Display [SF]
MN: 2004 AGU Fall Meeting