Engineering better airline passenger security

Long lines of passengers waiting to go through Transportation Security Administration screenings at airports can be a source of dread. The time it takes to complete this security process varies, and this can add extra stress for both passengers wanting to get to their flights and TSA officials wanting to provide a smooth and secure operation.

The Center for Accelerating Operational Efficiency, a Department of Homeland Security Center of Excellence led by Arizona State University, is tackling the challenge of ensuring that TSA screenings are maximizing the efficiency of their resources. 

Researchers at the CAOE have been developing decision-support tools to simulate, optimize and visualize TSA security screening checkpoint operations with ever-changing passenger demands.

The goal of this project at CAOE is to improve the performance of the airport screening process without compromising safety and security. The team is engineering operations research techniques to reduce the wait times for passengers as they go through the screening process to get onto an airplane.

CAOE’s work is meant to help TSA strategize when and where to have TSA agents assigned to best serve passengers and maintain a secure airport.

“Our first concentration was to come up with a better estimation of when passengers will actually be arriving at the screening checkpoints,” says Ron Askin, professor of industrial engineering in the School of Computing, Informatics, and Decision Systems Engineering, one of the six Ira A. Fulton Schools of Engineering. “We based that on flight schedules, how many people are originating at the airport and the allocation of flights to various concourses in Phoenix.”

Building prediction models

The research team put together a mechanistic model, which is a method of examining the workings of individual parts in a system. It looked at flight schedules, the capacity of the planes that are scheduled for each gate and a statistical curve of when people tend to arrive for flights.

The team then used machine learning techniques to compensate for errors based on data that were not specific to the day of week and time of day. They also integrated a time series analysis model, which predicts future values based on previously observed data.

“We did an ensemble of the different approaches to get a better sense overall, and we were able to enhance the methodology that TSA is currently using,” says Askin.

The statistical curve for the length of time passengers arrive before a flight was drastically altered this year as the COVID-19 pandemic greatly diminished the number of people flying. So, researchers at CAOE had to update the passenger arrival curve to keep the models current.

Part of that update was delivered courtesy of computer science capstone projects using the video captured by airport TSA checkpoints to estimate when passengers arrive. Using that data is potentially more accurate than the counts used by TSA now as it captures when passengers actually arrive instead of when they complete the screening process after waiting in queue.

When COVID-19 started spreading this spring, TSA wanted to know about changes in passenger behavior. Were people arriving closer to their flight times? Or were they arriving even earlier expecting extra waits due to the need for social distance?

“We found that passengers were arriving closer to their flight time,” says Jorge A. Sefair, assistant professor of industrial engineering in the Fulton Schools and principal investigator of the project. “The challenge was that we couldn’t interview passengers to ask them the times of their flights. So, we wanted to create an indirect method of calculating those arrival profiles, without interacting with people because of COVID-19.”

The team was able to provide multiple tools to TSA to help direct resources to maximize efficiency. One of those tools was a spreadsheet built from the work the team had done previously.

Using the number of screening lanes that will be open at each checkpoint for PreCheck, standard lines and how many X-ray machines are available. Given the variables, equations in the spreadsheet estimate wait times and predict the lines passengers will experience throughout the day.

The TSA can use this tool to decide whether and when to open an additional lane. It also can evaluate the impact of closing certain screening lanes and opening others elsewhere. Taken as a whole, the agency can determine how to improve passenger experience.

Optimizing resources

With the prediction models created, CAOE is working with TSA analysts in Phoenix to develop best practices for optimizing and deploying staffing resources.

The new tools will help TSA decide how many security officer shifts to plan for the various days of the week, and schedule the correct number of agents to provide the best experience for the arriving passengers.

The researchers at ASU have also been working with colleagues at the University of Texas at El Paso on simulations that allow them to validate the models they have created.

“We have deterministic optimization models and they have simulation models to insert more stochasticity (randomness) into it,” says Askin. “That’s also another way of validating the model.”

Those simulation models display and can track individual passengers moving through the system. It’s detailed enough to see them walking through the turnstiles.

The models are now being tested with data from Phoenix and Las Vegas, but the goal is to create systems that can help TSA nationwide. Also, some of the tools might be commercialized and used on a widespread basis across the country in different capacities.

“We want to improve the efficiency of the security checkpoint,” says Sefair. “But we’re doing that in many layers. It’s not only focused on the passenger experience in terms of wait times, but it’s also from the perspective of efficiently allocating TSA’s resources.”

Those resources are the transportation security officers, but also their technology. CAOE is able to see what happens if there is a change in technology, such as a new more efficient scanner.

“It’s a scan of efficiency, but in a broad sense,” says Sefair. “It’s beyond just the public’s perspective. It’s also about internal operational excellence.”

TSA staffing is determined by the United States Congress, but it’s up to the agency to decide the best way to use them. The agency also has a desired level of service requirement in terms of the maximum amount of time passengers should be waiting to pass through a checkpoint.

Educating the future

One benefit of having CAOE housed at ASU is it creates opportunities for students to learn and understand how engineering and computer science impacts the world around us — even the seemingly mundane experience of waiting in airports.

“Part of research is developing tools for them [TSA] to improve their operation, but part is training students in the analytic tools for decision-making in general,” says Sefair. “We use the airports as a case study for them to go collect information and learn about the operation of TSA. Our goal is to train the next generation of decision-makers for the homeland security enterprise.”

More than 20 undergraduates and several master’s students and doctoral students have been involved in the project over the past three years at ASU, UTEP, the University of Nevada, Las Vegas, Texas State University and San Diego State University.

Deploying tools

Now that the tools have been developed, the next step is transitioning them to TSA for everyday use.

In an era of new social distancing standards, new questions will likely arise, but the tools will help make the decisions easier.

Questions include the future possibility of incorporating machines capable of doing document checks so one security officer can float between multiple stations. Another question relates to the advantages of having one officer working at the front of the X-ray machine belt to assist people in getting their luggage ready.

Across all possible scenarios, the tools created by CAOE can help TSA distribute their staff effectively and efficiently.

“They may have more questions coming,” says Sefair. “But we’re here to help them with their decisions.”

CAOE wants to provide TSA with tools that improve decisions, but they are not trying to replace decision-makers. The idea is to reduce the burden of calculations that TSA officials make in real time.

“They need the autonomy to make crucial decisions,” says Askin. “And they can incorporate factors we may not know about for our model, so we just want to support their work. Our models are meant to be advisory, and ideally easy to use and accurate. We just need to convince people that we have factors built into these models that are relevant and are helpful.”