Notice and Invitation

Department of Systems Engineering and Operations Research
Volgenau School of Engineering, George Mason University 

Keith W. DeGregory


Bachelor of Science, Systems Engineering, United States Military Academy, 1997 
Master of Arts, Management, Webster University, 2002

Masters of Science, Operations Research, Massachusetts Institute of Technology, 2007 



An Approximate Dynamic Program for Allocating

Federal Air Marshals in Real-Time Under Uncertainty

Wednesday, April 30, 2014, 2:00 pm
Engineering Bldg., Room 2091


Rajesh Ganesan, Chair 
Alexander Brodsky

Andrew Loerch

Lance Sherry



The Federal Air Marshal Service provides front line security in homeland defense by protecting civil aviation from potential terrorist attacks.  Unique challenges arise in maximizing effective deployment of a limited number of air marshals to cover the risk posed by potential terrorists on nearly 30,000 daily domestic and international flights.  Some risk presents in a stochastic nature (i.e., a last minute ticket sale to suspicious individual).  Response to near-real time risk is not possible for pre-scheduled air marshal deployments.  This dissertation proposes the formation of a quick reaction force to specifically address the stochastic risk and a means for near-real time force allocation to optimize risk coverage.


The dynamic allocation of reactionary air marshals would involve sequential decision making under uncertainty with limited lead time.  This dissertation proposes implementing an approximate dynamic program (ADP) to assist schedulers dynamically allocating air marshals in near-real time.  Approximate dynamic programming is a form of reinforced learning that seeks optimal decisions by incorporating future impacts rather than optimizing short-term rewards.  The system is modeled as a Markov decision process.  Due to the many variables and environment complexity, explicit storage of all states and their values is not possible.  A value function approximation scheme mitigates the scalability issue by alleviating the need for storage of state values.  The study demonstrates that marshal allocation in near-real time is possible using an ADP methodology and results in improved coverage of stochastic risk over the myopic approach or pre-scheduling.