(Skip Navigation Menu)


The Virtual Research Vessel
Model Coupling

National Science Foundation
VRV Home · Contact Us
 

Virtual Research Vessel

About
People
Database Support
Tools
Composition
Data Access
Model Coupling
Docs & Pubs
Links

The East Pacific Rise in 3-D, from active source tomographic imaging of the subsurface magmatic system. Image by R. Dunn and D. Toomey.

Click image to enlarge

 
Faculty: Jan Cuny and Doug Toomey (U of O)
Grad Students: Tom Bulatewicz, Maureen Ford, Matt Sottile (U of O)

See the main U of O Model Coupling site at www.csi.uoregon.edu/projects/coupling/ .

Increasingly, scientists rely on computational models to understand complex phenomena having multiple, interacting components whose aggregate behavior can be understood only by the simultaneous analysis of their individual behaviors. In approaching such complexity, scientists have naturally adopted a "divide-and-conquer" approach, first creating models of isolated subprocesses. Many of these models are now well understood and captured in robust programs; they constitute the important first steps towards the ultimate goal of understanding complex interactions among physical or biological processes. The challenge now is for the ridge community as well as many other communities is to couple their isolated models into self-consistent representations of more complex processes. Coupling is more complex than the mere composition of computational elements. Scientists are faced with the poorly understood task of establishing sophisticated time-varying relationships between models and large, multi-dimensional data that are heterogeneous in quantity, quality, scale, type, and ultimately importance. To accomplish this, they will need more than standard coupling mechanisms; they will need support for dynamically exploring model correlations and relationships at a very high, domain- specific level.

We are building infrastructure to support the fast prototyping of model couplings. Robust, abstract descriptions of existing computational models (existing code) will be made available in a common database, a graphical interface will allow scientist to easily specify couplings between models, and the required interfaces and runtime monitor will be automatically generated. We have a rudimentary design of the entire system and are currently focusing on the first step, building an automated assistant to support the development of appropriate abstract, program descriptions with minimal effort on the part of the user.


Computational Model (source code is annotated with coupling potential)

Automatically-Generated Intermediate Form (the program's control flow graph is automatically compressed with respect to the annotations)

Database Representation


BACK TO TOP





Last update: September 9, 2005
© 2001, D. Wright and OSU Webworks
http://oregonstate.edu/dept/vrv/models.html