Home > News & Events > News

George Biros and collaborators awarded NSF grant:

A Computational Framework for Real-time Identification of Hazardous Events: Application to Dispersion of Airborne Contaminants

George Biros, University of Pennsylvania (PI)
Omar Ghattas, University of Texas at Austin (lead PI)
Karen Willcox, MIT (PI)
Bart van Bloemen Waanders, Sandia National Labs (PI)
Alfio Borzi, University of Graz (co-PI)
Volkan Akcelik, Carnegie Mellon University (co-PI)
Judy Hill, Sandia National Labs (co-PI)
Andrei Draganescu, Sandia National Labs (co-PI)

The National Science Foundation has awarded a team from the University of Pennsylvania, UT-Austin, MIT, Sandia National Labs, and the University of Graz an $825,000 grant to create a data-driven, high performance computational framework for real-time identification of hazardous events from sensor measurements, and consequent prediction of the evolution of the hazard.

The framework will be applied to the identification and prediction of the dispersion of intentionally- or accidentally-released atmospheric contaminants in urban regions. The system consists of four phases that execute continuously: sensors measure contaminant concentrations at points within the atmosphere; the sensor data is inverted to determine initial conditions with built-in uncertainty estimates for contaminant transport models; uncertainty in the reconstructed initial conditions is propagated to provide probabilistic predictions of the contaminant dispersion; and mobile sensors are steered into new locations to reduce uncertainty in the predictions, leading to a repetition of the cycle. The team will consider two time scales of decision-making at which the framework must execute. The seconds-to-minutes decision-making scale is required by first responders to begin immediate response efforts. For such time scales, high-fidelity models are too formidable. Instead, the team will construct reduced-order models to facilitate real-time execution. On the other hand, the minutes-to hours decision-making scale permits more careful and measured response by emergency officials using high-fidelity, high-resolution models. To enable rapid execution of the framework for such models, fast, scalable, parallel algorithms for inversion and prediction will be developed. The framework will be implemented in a software toolkit that permits application to a broader class of simulation-based decision-making problems involving natural disasters, industrial accidents, and terrorist attacks.

(October 2005)