Optimization of an airport security selectee lane system using simulation
MetadataShow full item record
In this thesis, we present an optimization approach for an airport security selectee lane system using simulation. A typical U.S. airport has two kinds of passenger checkpoint screening lanes: a normal lane and a selectee lane which has enhanced scrutiny. The selectee lane is not efficiently utilized in some airports due to the small number of passengers selected to go through it. Previous research (Nie, 2008) has suggested a selectee lane queueing design model that leads to more efficient resource utilization than the current practice, assuming that the model’s assumption of steady-state arrivals is valid. Nie’s model is based on an objective function that maximizes the passenger checkpoint screening system’s probability of true alarm. In reality, passengers at an airport do not arrive via a time-homogeneous Poisson process and the queue rarely reaches steady-state conditions, so the assumptions of Nie’s model are not valid in practice. To conclusively evaluate the model and to further improve its results, we developed a simulation of an airport security system with a selectee lane under realistic assumptions. The first goal of our simulation model is to verify the effectiveness of Nie’s results. To do this the simulation model is first run assuming that only "selectee" passengers are sent to the "selectee" lane; it is then run with the assignment matrix generated by the model. Our findings suggest that the model improves the probability of true alarm as reported in Nie (2008). Our second goal is to improve upon the assignment matrix suggested by the model in Nie (2008). To do this, a simulation based optimization approach is proposed. The output of selectee lane queueing design model is assumed to be a good starting solution and is evaluated using the simulation model. A neighborhood search technique is used to conduct a local search for an improved solution for the assignment matrix; each assignment matrix solution is evaluated through multiple simulation runs. Our findings suggest that the simulation based optimization approach leads to a new assignment matrix which yields an overall increase in performance with respect to the probability of true alarm. Our third and final goal is to examine the tradeoff between the objectives of maximizing the probability of true alarm and minimizing the average time in system. To do this, we evaluate and plot each of the solutions in our local search scheme with respect to these two objectives using simulation. Our findings yield a small set of non-dominating solutions.