DIMACS/DyDAn Research Project on Sensor Management for Nuclear Detection

Project Leaders:

Fred S. Roberts, PI, Rutgers University

Warren Powell, co-PI, Princeton University

Project Faculty:

Tamra Carpenter, Rutgers University

Siddhartha Dalal, Rand Corporation

Savas Dayanik, Princeton University

Nathaniel Dean, Texas State University - San Marcos

Rick Mammone, Rutgers University

Dimitris Metaxas, Rutgers University

William Pottenger, Rutgers University

Warren Powell, Princeton University

Fred S. Roberts, Rutgers University

Minge Xie, Rutgers University

Postdoctoral Researcher:

Alantha Newman, Rutgers University

Graduate Student Researchers:

Jerry Cheng, Rutgers University

Jill Cochran, Texas State University - San Marcos

Peter Frazier, Princeton University

Participating National Laboratories:

Los Alamos National Laboratory

Pacific Northwest National Laboratory

Sandia National Laboratory

Project Support: This project is supported by a grant from the National Science Foundation funded by the US Department of Homeland Security Domestic Nuclear Detection Office.

Project Description

The effective use of sensors in nuclear detection requires choosing the right type of sensor, putting it in the right place and activating it at the right times. It also involves interpreting the results of sensor alarms and making decisions that balance various types of risk and uncertainty based on those results, sometimes in real time. These are the multiple dimensions of sensor management that we seek to address in this project.

We will begin to address sensor management for nuclear threat detection by formulating the related problems using precise mathematical language and then developing tools of the mathematical sciences to solve them. Our project will consider several major problem settings for nuclear detection, both to ensure that we have the greatest possible impact and to demonstrate that our research will handle a suitably diverse set of conditions, spanning signals (the nature of the data being collected), economics (identifying the best technology to use in a particular setting), and responses (detaining a stationary vehicle or container, intercepting a moving vehicle or person, etc.).

In particular, we will:

• apply our previous work on Bayesian Binary and Multinomial regression, including online methods;
• investigate the use of sensor management tools based on our previous algorithm development in clustering, location theory, and equipment placing algorithms;
• apply innovative dynamic programming methods to solve resource allocation problems in the presence of uncertain forecasts, drawing on advances in approximate dynamic programming for on-line learning; and
• investigate new data sampling strategies that we have used successfully in applications ranging from transportation problems, to communication problems, to adverse event detection.

Our project will connect to nuclear detection problems in three specific contexts:

• Risk assessment for containers and trucks at borders and seaports - through collaboration with PNNL's Radiation Portal Monitoring Project, we will analyze data that is archived once a day and is used to evaluate trends, get early warning of faulty detectors, and plan manpower and equipment needs.
• Special events - We will develop methods to determine where to locate sensors so as to optimize detection.
• Moving vehicles or individuals - We will study detection of devices and material in transit, where we lack the ability to choose what to screen. We will study the problem of identifying which vehicle or person was a source and how to integrate information and make further measurements or interventions.

Dissemination and Outreach

We aim to bring the results of this project to a larger research community by hosting workshops and developing educational materials on project themes. Workshops will include talks of a tutorial nature aimed at students and non-specialists. In addition, we hope to integrate with the wide variety of activities already underway at DIMACS and DyDAn, in particular: expanding our REU program by adding a research project on nuclear detection; developing examples from nuclear detection for our "math and homeland security" high school program; and expanding our Homeland Security seminar series to include speakers on problems in nuclear detection.

Document last modified on January 29, 2008.