This map of New Orleans shows the sectors created by the optimized evacuation plan. Blue lines indicate secondary roads used in evacuation routes. Green lines indicate roads used as one-way contraflow evacuation routes. The orange shading indicates population density, with darker shading indicating greater density. Image courtesy: Pascal Van Hentenryck and K’Kio HardinANN ARBOR—As the 10th U.S. hurricane season since Katrina rolls on, a University of Michigan professor is using advanced data analytics and optimization techniques to find better ways to evacuate regions before disaster strikes.

The approach revealed a plan that researchers say could have evacuated New Orleans 37 percent faster in the days before Katrina.

The plan used to evacuate New Orleans was state-of-the-art for 2005; it was one of the first major evacuations to use what’s called “contra-flow traffic routing” that turns major roads into one-way routes out of the city. And today, some coastal cities use even more advanced techniques to plan evacuations, including zone-based approaches that divide cities up and signal those closest to the storm to leave first.

“Evacuation planning has come a long way since Katrina, which didn’t use zone-based evacuation,” said Pascal Van Hentenryck, U-M professor of industrial and operations engineering. “But even today, zone-based evacuations are fairly coarse. The zones are large, and detailed routing and timing aren’t provided. Plans are developed manually, with computer simulations run late in the process to predict how they’ll play out on the ground.”

Van Hentenryck’s approach differs in that it uses computer modeling from the get-go. It crunches a staggering number of variables, from population density to infrastructure and even the projected behavior of evacuees.

“The number of possible outcomes in these scenarios is greater than the number of atoms in the universe,” Van Hentenryck said. “It’s important to eliminate as many possibilities as possible early in the process, so we apply some very sophisticated analysis techniques to do that.”

The approach then generates a complete evacuation plan, dividing the city into intricate zones and planning a detailed route out for each zone. It also calculates the ideal time for each zone to evacuate, maximizing the number of cars on the road but avoiding congestion. The method is the first to incorporate such a wide variety of variables into a single model. It’s the also the first model of its kind that can be scaled to a city the size of New Orleans.

“When you see these models, it’s clear that no human could have come up with them,” Van Hentenryck said. “They can be counterintuitive, but they’re very effective at using infrastructure effectively and minimizing congestion.”

As a test case, Van Hentenryck applied the approach to a hypothetical repeat of Hurricane Katrina, essentially turning it into a city-sized math problem. He acknowledges that such an approach would require more detailed on-the-ground planning and communication than conventional evacuation tactics. But he believes that simple, inexpensive tools could be used to make an optimized New Orleans evacuation go smoothly.

“Keeping people informed and mobilizing them efficiently is a really important part of the process, but it doesn’t have to be complicated or expensive,” Van Hentenryck said. “Cities like Sydney, Australia, for example, use simple tools like text messages and door-to-door volunteers to keep people informed. They’re very effective and I think they could work here, too.”

Once evacuees are on the road, they’d follow optimized paths that are mapped out by a computer using what’s called “convergent routing.” This minimizes forks and intersections and funnels traffic into a single smooth flow. Van Hentenryck says this is one area where human behavior comes into play.

“We know that every time a driver faces a choice, like an intersection or a fork in the road, he or she slows down,” he said. “Even if each driver only slows for a few seconds, the combined effect can have a big impact on traffic. So the algorithm plans a route that eliminates these choices whenever possible, enabling a given stretch of road to handle many more cars.

Van Hentenryck says his team’s methods could be applied in any city to virtually any large-scale incidents, including wildfires, floods, earthquakes and industrial hazards. The next step is to move the system beyond one-off optimizations and ultimately develop a comprehensive and integrated software tool that cities could use to optimize their own evacuation plans and make more informed decisions about disaster planning.

“The world is urbanizing at a faster pace than we’ve ever seen before, and that’s going to make evacuation planning increasingly important in the years ahead,” Van Hentenryck said. “There’s an implicit promise between city planners and residents that people will be kept safe in the event of a disaster, and I’m looking forward to helping cities do a better job of keeping that promise.”

Recent findings from the team are detailed in a paper that’s been accepted in Transportation Research: Part B. The paper is titled “A Conflict-Based Path-Generation Heuristic for Evacuation Planning.”

More information:

• Related news story: Hurricane Joaquin: 14 million could lose power
• Pascal Van Hentenryck