HR: 17:15h
AN: IN24A-06 [Abstracts]
TI: Goal-Directed Planning for Sensor Webs
AU: * Morris, R
EM: morris@ptolemy.arc.nasa.gov
AF: NASA, MS 269-1
Ames Research Center, Moffett Field, CA 94035-1000, United States
AU: Dungan, J L
EM: Jennifer.L.Dungan@nasa.gov
AF: NASA, MS 242-4
Ames Research Center, Moffett Field, CA 94035-1000, United States
AU: Khatib, L
EM: lina@email.arc.nasa.gov
AF: Perot Systems Government Services, MS 269-1
Ames Research Center, Moffett Field, CA 94035-1000, United States
AU: Votava, P
EM: pvotava@mail.arc.nasa.gov
AF: California State University, Monterey Bay, MS 242-4
Ames Research Center, Moffett Field, CA 94035-1000, United States
AB:
An Earth-observing sensor web is an organization of space, airborne, or in situ sensing devices for collecting
measurements of the Earth's processes. Sensor web coordination involves formulating Earth science goals and
transforming them into sensor web workflows, i.e., sequences of data acquisition and processing tasks that
satisfy the specified goals. Automating parts of this process using recent advances in intelligent control software
technology will offer improved sensor web effectiveness.
Our approach to the coordination problem applies architectural concepts of workflow management systems by
identifying two phases in workflow generation. In the first phase, users formulate high-level campaign goals that
are automatically transformed into abstract workflow plans. An abstract workflow plan represents the organization
of data acquisition and processing actions that fulfills the goals specified by the user, but leaves out details such
as how requests for access to a data resource are formatted. Abstracting away these details improves the
usability of sensor web resources by scientists. To implement the first phase, we utilize the Labeled Transition
System Analyzer (LTSA), a model-checking software tool. LTSA contains a concise process-based language,
FSP (Finite State Processes) for designing and modeling software programs. We will use LTSA and FSP to
automate the process of building executable plans for accessing resources on a sensor web. FSP has the
constructs for representing conditional dependencies, iterations, and parallel actions, all of which are common
features in Earth science campaigns.
The second phase of the process consists of the automatic transformation of an abstract plan into a concrete
plan, i.e., a sequence of actions that can be autonomously executed on a sensor web. The transformation in
phase two might require further decomposition of actions in the abstract plan into a sequence of lower-level data
acquisition requests. It may also involve the selection of resources to accomplish a given action and the
representation of data acquisition tasks in a format that is recognized by the targeted resource (e.g. a sensor
control command or a data archive query). The second phase relies on a service-layer information infrastructure
for accessing sensor web resources. Standardizing requirements for such a service layer through the Open
Geospatial Consortium Sensor Web Enablement (OGC/SWE) effort should allow access to numerous and
diverse sensor web resources. For the purpose of demonstrating a prototype of our workflow management
concepts, our system currently utilizes a simpler information infrastructure layer for servicing requests. This layer
controls access to TOPS (Terrestrial Observation and Prediction System), a modeling software system that
brings together technologies in information technology, weather/climate forecasting, ecosystem modeling, and
satellite remote sensing to enhance management decisions related to floods, droughts, forest fires, human
health, and crop, range, and forest production. We provide examples of concrete plans for accessing TOPS data
and modeling resources and how they are generated from abstract plans.
DE: 1622 Earth system modeling (1225)
DE: 1640 Remote sensing (1855)
DE: 1694 Instruments and techniques
SC: Earth and Space Science Informatics [IN]
MN: 2007 Fall Meeting