A crowd stampede that caused at least 769 deaths at the annual Muslim pilgrimage to Mecca provided a stark reminder of the safety challenges at the most congested public space in the world. The millions of pilgrims who visit Mecca and other holy sites in Saudi Arabia each year often crowd into spaces as densely packed as six people per square meter. Such crowd densities present a huge challenge for both public safety authorities and computer simulation software designed to help model and predict crowd disasters.
In 2010, Saudi Arabia hosted an international design competition aimed at safely accommodating more pilgrims at Mecca’s Grand Mosque, which houses the cube-shaped Kaaba building representing Islam’s holiest site. One of the main proposals came from Gensler, a U.S. architectural design firm. Gensler brought along some special outside expertise in the form of Cenk Tunasar, a principal at the strategic innovation group of Booz Allen Hamilton. Tunasar sees a combination of infrastructure design, technology and operational planning as the best way to manage huge crowds and avoid potential disaster.
“I don’t believe you can directly control crowds efficiently,” Tunasar says. “You can control capacity, but crowds can figure out [efficient crowd movement] themselves with the right information.”
Tunasar and the Gensler team figured out how to double the Grand Mosque’s peak visitor rate from 40,000 to 102,000 people per hour. They also calculated how to reduce crowd densities from six to four people per square meter. Typically, swarms of pilgrims must perform a series of “Tawaf” rituals requiring them to circle around the Kaaba building within the Grand Mosque seven times in a counter-clockwise direction.
The new proposal suggested adding a separate floor ringing the Kaaba to segregate the usual ground floor crowd into smaller, more manageable crowds. That design philosophy helped retain some measure of safety rather than simply expanding the Grand Mosque’s ground floor and having more people in the same space. “If something did happen, a separate space means you can at least stop feeding people into the channels leading to that space,” Tunasar said.
But simulating the huge Hajj crowds was no cakewalk. Tunasar soon realized that commercial off-the-self software made for simulating crowds was unable to handle the crowd densities at the Hajj. (To put six people per square meter into perspective, imagine five other people squeezed next to you in a space bounded by dimensions shorter than your arm span.) He eventually ended up coding people’s movements in multiple simulation software languages and using several different types of software, as detailed in a paper presented at the 2013 Simulation Conference.
The Hajj pilgrimage also represents a very different scenario from the usual crowds being simulated by commercial software. Such software typically assumes the scenario involves a high throughput or fast rate of people passing through a given space; one example would be airline passengers trying to get through airport security as fast as possible. But Hajj pilgrims tend to take their time in performing rituals and savoring the religious experience; many can spend decades on the waiting lists to visit Mecca.
Eventually, the crowd simulations ran up against their limits: Simulating the huge Hajj crowds with collision modeling—accounting for each individual person’s physical space—proved impractical for assessing all the design possibilities for the project. Just one day’s worth of simulations on an Intel 16 Core server with 64 gigabytes main memory took over two weeks to run; simulation run times were about 10 to 20 times slower than real time crowd movement.
“We needed to constantly try alterations but to also have a very logical way of testing,” Tunasar explained. “If you want to do detailed simulations, they’re too expensive to try for all the designs.”
Tunasar’s solution was to avoid using the simulation software during the project’s design phase, which required quick calculations to assess the viability of each possible design for the Hajj. Instead, he turned to a much simpler crowd modeling based on queuing theory, which considers the arrival of certain entities in a system and their behavior related to time spent in the system. In this case, the “entities” are the pilgrims and the “behavior” involves their pilgrimage path and rituals surrounding the holy sites.
The key formula from queuing theory that proved most helpful is called Little’s Law. The law simply calculates the number of pilgrims present as the rate of arrival to the Grand Mosque multiplied by the time spent in the Grand Mosque. That allowed Tunasar and the Gensler team to quickly test different designs and then use the crowd simulation software to validate the best choices.
The recent deadly disaster at Mecca did not take place at the Grand Mosque. Instead, it took place at a different point along the pilgrimage route that extends between the Grand Mosque and multiple holy sites. The stampede occurred in an area with two huge lines of pilgrims converging at right angles on an intersection near the Jamarat Bridge, about 5 kilometers from Mecca, BBC News reports.
It’s tough to say exactly what could have been done better to prevent this particular stampede. But technology could make it easier to safely manage the huge crowds at the Hajj in the future, Tunasar says. He suggested that surveillance cameras and computer software could keep an eye on crowds to detect early signs of crowd movement trouble. An early warning might give authorities time to deal with congestion by closing certain entrances or opening alternative exits.
The Saudi Ministry of Hajj has already begun considering plans to give all pilgrims an electronic bracelet with individual identification, GPS, a prayer times alert, a compass and a guide to the Hajj, according to Al Arabiya News. Tunasar pointed out that such bracelets could theoretically also include a vibrating or alert signal to tell pilgrims when it’s time to slow down and possibly prevent a crowd situation from getting worse. A simple signal could be much easier than shouting at a multilingual crowd from all over the world through loudspeakers.
Tunasar’s research into crowds has given him trust in how they can efficiently “self-regulate” their movements and find the most efficient entrances and exits. But in order to self-regulate, crowds must have vital sources of information about possible travel routes, such as an alternative airport security checkpoint or pilgrimage path.
Of course, crowds also can’t do much if the existing infrastructure—building passages, doorways, and roads—don’t give them any movement options. Danger can often occur at big crowd events because of such inflexible, fixed infrastructure, Tunasar says. Changing the local infrastructure can be as simple as opening a closed doorway or removing road barriers. That’s a lesson that could apply equally to either the Hajj or a music festival.
“My thinking is that 21st century architecture should be able to flexibly expand and contract,” Tunasar says.
Jeremy Hsu has been working as a science and technology journalist in New York City since 2008. He has written on subjects as diverse as supercomputing and wearable electronics for IEEE Spectrum. When he’s not trying to wrap his head around the latest quantum computing news for Spectrum, he also contributes to a variety of publications such as Scientific American, Discover, Popular Science, and others. He is a graduate of New York University’s Science, Health & Environmental Reporting Program.