Addressing passenger risk uncertainty for aviation security screening

Within the aviation security system, the preeminent objective of passenger and baggage screening is to prevent prohibited items from entering the airport terminal and getting put onboard a commercial aircraft. Because of limited budget and personnel resources, as well as constraints on the screening device capacities, only a fraction of the passengers may be screened at the highest security levels. Sequential passenger assignment policies have been formulated through dynamic programming and nonlinear control. However, both of these approaches rely on a known distribution of passenger risk. This paper presents estimation algorithms that address various levels of uncertainty in the passenger risk distribution, which can be applied to existing passenger assignment policies. Simulation results are reported to illustrate the sensitivity to variations in the unknown distribution parameter and to demonstrate that the prudent practice of overestimating the overall population risk level produces a larger number of improperly screened passengers and a lower level of security in comparison to underestimating passenger risk. The key contribution of this work is the finding that integrating online estimation of passenger risk into security screening assignment decisions increases the overall expected security and decreases the sensitivity to variations in the overall population risk level.