Cover Feature Management Newswire

Sharing Elevators During COVID-19

With COVID-19 vaccines now slowly being distributed, city officials, business administrators and high-rise building managers are planning how to safely and efficiently open offices as people come back to work. But with social distancing and masking likely still needed, the capacity of both elevators and building lobbies will remain limited — it could take workers so long time to get to their desks that it makes no sense to come in at all.

Elevators are the key component of the “vertical transportation industry” that so many people use to get to their offices each day, but social distancing has reduced elevator capacity by at least half and by as much as two-thirds of the amount of pre-pandemic times. When buildings fully reopen with the usual amount of foot traffic, the length of queues will likely explode during rush hour, thus posing a significant public health risk.

Researchers at Columbia Engineering, the engineering school at Columbia University, worked on solutions, using real-world data and context provided by the Office of the Mayor of New York City. The resulting report, “Queuing Safely for Elevator Systems Amidst a Pandemic,” is posted on the Social Science Research Network.

“There are obvious advantages for using the status quo, as it ensures fairness and requires no management of the queue. However, even pre-COVID-19, especially during rush hours, such as morning and lunchtime, the lobby may be crowded with passengers waiting to get to their floor,” the report said. “Meanwhile, elevators are fully-loaded and can make many stops during the trip. With a physical distancing rule during a pandemic, such as COVID-19, the dramatically reduced elevator capacity could cause congestion in the lobby and thus increase the risk of disease spread.”

To find a safer method, the team used mathematical modeling and epidemiological principles to design interventions for queuing safely in elevators during a pandemic, without having to program elevators.

“Our interventions basically reduce congestion by explicitly or implicitly trying to get people going to the same floor to travel together,” said Adam N. Elmachtoub, associate professor of industrial engineering and operations research (IEOR) and a member of Columbia’s Data Science Institute (DSI). The team worked with Charles C. Branas, Gelman Professor and chair of epidemiology at the Mailman School of Public Health, to translate their algorithms into safe and practical interventions.

The goal of the team, which included IEOR and computer science Professor Clifford Stein and IEOR Ph.D. students Sai Mali Ananthanarayanan and Yeqing Zhou, was to manage elevator lines by drastically reducing the waiting time and length of lobby queues. In simulations based on an actual 25-story building, the researchers compared the traditional “first-come, first-served” method of loading elevators with two different interventions they call “cohorting” and “queue splitting.”

The report noted that it only considers the issue of moving people upwards in a building from the lobby.

“We focus on analyzing solutions that work for high-volume periods, e.g. morning rush hour, lunchtime, etc. where social distancing is a challenge,” the report said. “Without elevator AI, it is near-impossible to do any interventions for downward and inter-floor movement.”

Under the traditional loading approach during the pandemic, the queue length in the lobby keeps growing. Using the cohorting intervention, the team grouped passengers together who were going to the same floor as the first person in the queue.

“In this intervention, passengers line up in a queue in order of arrival. When an elevator arrives, the first passenger boards. Then, the QM asks if anyone in the queue going to the same floor as the first passenger, and they board as well (according to their arrival order), in order to create a cohort going to the same floor (such passengers are allowed to “cut in line”). If there is still capacity in the elevator, then the passenger at the front of the queue enters, and the QM again allows passengers going to the same floor to board the elevator. This process is repeated until the elevator is full or the queue is empty.”

Cohorting is the best-performing intervention to improve efficiency, but it requires a queue manager (QM), the report said, to interact effectively with the queue to learn where passengers are going. Other considerations include the ease of understanding for the passengers and perceived inequity when passengers cut the line.

Under the queue splitting approach, the researchers created different lines for people going to different groups of floors. Floors are assigned to different groups, where each group consists of consecutive floors, such as one through eight or nine through 16.

“Arriving passengers join a queue corresponding to their floor group, and elevators are boarded from the queues in a round-robin fashion (possibly with the help of a QM),” according to the report. “By creating queues for every floor group, the travel time of elevators is naturally reduced since passengers are likely to be going to the same or nearby floors, which achieves an effect similar to cohorting.”

Unlike cohorting, however, queue splitting requires reorganizing lobby space to allow room for the queues.

In Allocation, a third simulation discussed in the report, each elevator is assigned to only go to predetermined floors (which might be a range or splitting into odd and even floors). This intervention can be accomplished by changing the elevator control system or simply by adding signs on each elevator door. However, the allocation of floors into different ranges needs careful design. On one hand, the traffic to different floors may vary a lot. On the other hand, it is clear that higher floors will need more elevators allocated, as the travel time to higher floors are naturally longer than other floors. The second issue is that it may not be possible to reprogram the elevators.

“An allocation decision needs to be based on solving an optimization problem that may be difficult to solve,” the report said.

Based on their simulations, the researchers found that both the cohorting and queue splitting interventions not only significantly reduced queue length and wait time but also enabled workers to keep safe distances from viral transmissions in what would otherwise be overcrowded elevators, building lobbies and entrances. These proposed interventions are simple and easy to implement, relying on signage and/or a queue manager to guide passengers, the group said. They could work especially well in cities like New York, where many elevator systems are older, and changing their algorithms and technology would require long-term planning and expensive modifications.

“This project shows the critical role of interdisciplinary collaboration, intellectually between data science and other areas like public health, [and] between city government and academic institutions in working on applied problems grounded in real-world constraints,” said Neal Parikh, director of artificial intelligence for the City of New York. “This is a key way to facilitate the responsible and beneficial use of algorithms and other mathematical and computational tools in society.”

Additional efficiencies could occur if passengers are willing and able to get off the elevator one floor above or below the desired destination and walk the rest of the way. Passengers could then join the shortest queue going one floor from their destination.

Each building may choose a different intervention, depending on its particular simulation results, physical layout, personnel and epidemiological principles. The researchers are refining their simulation model and testing out the policies on many different building designs.

“The interventions we study apply to generic buildings, and we have provided open-source code that can be customized and studied for other building types,” said Ananthanarayanan, another co-lead author of the report.

“Our simulation model is not only general, in that it can be adapted to any number of elevators, floors and general traffic patterns, but it also takes into account potential issues in implementation, such as the impact of limited lobby space,” added co-lead author Zhou. “It’s very flexible.”

While these initial approaches offer simple, low-tech solutions, the team is also looking at more advanced elevator systems that would enable them to embed AI technologies to more efficiently and safely manage elevators. They are developing algorithms for moving elevators that can depend on the global state of the system, which can be measured via sensors in the elevators and waiting areas.

“We’ve simulated for different building types and calibrated data from a large New York City building,” said Stein, who is also the associate director of research at the Data Science Institute. “Our proposed interventions will be useful to any high-rise building managers who are formulating reopening plans. We’re excited to be part of engineering a speedy recovery for New York City and locations around the world with a vertical transportation solution.”

Sign Up for Newswire

    [ctct ctct-156 type:hidden 'Mann Report Management Newswire::#159']