|
NSF and NASA support CS research |
|
|
|
D r. Adrian Sandu, associate professor at the CS department at Virginia Tech in collaboration with Dr. I.M. Navon of Florida University, have received funding of $350,000 from NSF for research entitled "Solution of Inverse Problems with Adaptive Models." $186,000 of this award is assigned to Virginia Tech. Another $211,000 is allotted to the CS department by NASA-JPL for the development of a Sensor-web Operations Explorer, in collaboration with PI M. Lee and Co-Investigators K. Bowman and C.E. Miller of Jet Propulsion Laboratory.
Inverse problems like parameter estimation, data assimilation, and optimal control for large scale systems are of considerable importance in many fields, ranging from the optimal control of engineering flows to atmospheric and oceanic sciences, and from climate change to homeland security.
State-of-the-art solvers for large scale systems adaptively refine the time step and the mesh to control the numerical errors. In contrast, most inverse problems to date have been solved using non-adaptive methods.
The research entitled "Solution of Inverse Problems with Adaptive Models" aims to develop new mathematical concepts and new computational tools needed to fill the gap between the state-of-the-art adaptive methods used in forward simulation and the methods currently available for the solution of inverse problems. The new tools will be demonstrated on the comprehensive three-dimensional adaptive oceanic model ICOM.
SOX (the Sensor-web Operations Xplorer) seeks to enable the design of next negation sensor networks by providing the capability to rapidly simulate their performance.
The SOX system simulates observations of the Earth's atmosphere from instruments deployed aboard NASA satellites and sub-orbital platforms that sample reference models of the Earth system. These virtual observations are used to retrieve geophysical parameter estimates, which are then assimilated. The estimated parameters are compared with the "true"
parameters from the reference model to assess the quality of the virtual sensor network.
For further information on Dr. Sandu's research, please see http://csl.cs.vt.edu/ |
|
