September 20, 2021
How AI Enables Rapid, Safe Evacuations

by Isabelle Barany
On Saturday, August 28th, my girlfriend and I woke up at 5 AM, crowded as much food as we could into a cooler, locked and relocked our windows, stowed our handwritten letters from friends and family next to half a week’s worth of clothes in our duffels, and began our drive from our neighborhood Gentilly, in northeastern New Orleans, to Knoxville, Tennessee. We thought about, in no particular order, flash floods, winds over 100 mph, lines of cars curling around multiple street corners for gas, our neighbors who’d chosen to stay behind, when we’d come back, and traffic. 

We were one of thousands of families leaving. Even at this hour, we encountered congestion as we drove North. Throughout the day, these conditions only worsened: there were significant delays on all of the highways leading out of New Orleans, and there were even reports of gas stations running dry

Drivers experienced these delays even though New Orleans only ordered mandatory evacuation orders for those outside of the levee system: those inside the city itself were given voluntary evacuation orders instead. While at first this decision seems puzzling given the magnitude of Ida, mandatory evacuations must be issued 72 hours in advance. Ida intensified too rapidly for officials to guarantee everyone could be evacuated safely before the storm hit. As New Orleans Mayor LaToya Cantrell clarified, “We are not calling for a mandatory evacuation because time is simply not on our side. We do not want to have people on the road, and therefore in greater danger.”

Less than 24 hours later, what Louisiana Governor John Bel Edwards called “one of the strongest hurricanes to hit in modern times” made landfall. All of New Orleans lost power. Over one million Louisiana residents waited weeks to gain back electricity during life threatening heat waves. While New Orleans’ levees held, there were fatalities in communities outside of the levee system, and as the storm moved Northeast it claimed more lives.

Evidence suggests hurricane Ida is not an outlier.

Experts now predict that climate change is causing hurricanes to become stronger, both in their winds speeds and rainfall. Even more troubling, leading scientists also hypothesize that global warming is causing tropical storms and depressions to intensify into major hurricanes much more rapidly. Cities and coastal areas will need to increasingly face storms which arrive too quickly to evacuate people under our current framework of managing traffic flow.

Addressing this terrifying possibility will require a complex, multifaceted approach including investment in levee expansion, fortification of storm shelters, and increasing collaboration between national, state, and local officials. However, better traffic management can unquestionably be part of how we confront these new realities. As I've read the news about my home, I’m particularly proud to serve as the Lead Writer for Flow Labs. Our integrated traffic management platform, powered by artificial intelligence, can be utilized to facilitate faster and safer evacuations.

Evacuation plans require time and reliable data.

In the event of an evacuation, especially a mandatory one, officials rely on two strategies in particular: phasing and contraflow measures. For phasing, officials evacuate a large area in phases, determining who, when, and where citizens evacuate. For contraflow, officials identify major roadways which will serve as escape routes, and reverse traffic flow on lanes leading into the evacuation area. Both phasing and contraflow have proven to be effective: as an example, contraflow was successfully deployed in the days before Hurricane Katrina to quicken the evacuation. However, contraflow is expensive and typically requires 12 to 24 hours to implement. Additionally, contraflows can impede emergency vehicles from accessing evacuation zones.

To make the best possible decisions about when and where to establish phases and contraflows, traffic management professionals must be able to access comprehensive, reliable data on all their roadways. They must be able to see what’s occurring during the days leading up to a hurricane, and they must be able to see what happened during past evacuations, so they can rapidly analyze their previous approaches and make adjustments.

However, today, the majority of agencies don’t have sufficient information to understand what’s happening currently or historically on their networks. Flow Labs recently conducted a large-scale study of over 1,000 detectors, and found that only 28% of them provide reliable data in day to day field conditions, let alone inclement weather.

Our artificial intelligence provides agencies with accurate, comprehensive data for these critical decisions.

At Flow Labs, we leverage artificial intelligence to integrate multiple traffic data sources together to give agencies information they can trust. In contrast to hardware detectors, our proprietary virtual sensors are 94.4% accurate. With our reliable data, agencies can immediately see the evacuation routes their citizens have taken historically and are taking now. They can instantly see where their networks and intersections are most congested, and factor this information into when and where they create phases and contraflows. They can alleviate delays which can become deadly in themselves, as was the case for Hurricane Rita, when over 100 people died during the evacuation itself.

They can also see where their roadways are the most unsafe, which is a critical factor to successful evacuations. On congested roadways where citizens are fatigued and panicked, they become vulnerable to crashes. Flow Labs is leading the industry in safety metrics: we track the key leading indicators of unsafe driving, such as aggressive acceleration or deceleration on roadways. Moving forward, we’re creating solutions to alert drivers to accidents in real time, and to rapidly divert drivers to accident free routes.

This information can also be utilized to help citizens who don’t have access to a vehicle. Currently, cities plan to use public transportation in the event of an evacuation. As an example, New Orleans has created 17 “evacuspots,” where citizens should gather in order to be transported out by buses if a mandatory order is issued. With better evacuation data, agencies can better evaluate whether these spots are in the correct places and how many spots there should be.

Our artificial intelligence optimizes all of the traffic signals in the city, a valuable solution in the event of an evacuation.

Finally, Flow Labs signal optimization software could be a powerful tool for agency officials faced with a multitude of life and death choices in an unfathomably complex, rapidly evolving situation. Our multimodal optimizations leverage our proprietary data to generate plans that account for both mobility and safety, so that agencies can ensure their road users move through their roadways as quickly and as safely as possible. Agencies can rely on our software to create signal timing plans they can immediately use.

What’s particularly exciting to me, as a proud resident of New Orleans, is that our optimization plans have been proven to work. In a partnership with Utah’s Department of Transportation, we reduced travel times by up to 24%. If we’re moving citizens through roadways 24% quicker, that’s 24% more time to evacuate citizens. That’s 24% more time leadership gains to make decisions. And, since a mandatory evacuation primarily requires 72 hours in order to transport citizens without vehicles, that’s 24% more time to evacuate the people who need to leave the most. How many more people could officials save if they had more time?

Our artificial intelligence mitigates emissions.

There always have been, and always will be, hurricanes in the Gulf. However, the catastrophic consequences of Harvey, Laura, and Ida within the last five years alone prove these storms are significantly more dangerous and frequent. Yet we can’t ignore the link between global warming and hurricanes any more than we can ignore the link between global warming and carbon emissions from human activity.

It’s in this way that Flow Labs can perhaps make its greatest contribution to hurricane safety. In the US, transportation emissions are the largest contributor to carbon emissions out of any sector. Roadway transportation alone accounts for 24% of all emissions. Flow Labs is fighting this trend with our technology. By slashing travel times, especially the amount of time cars idle, we’ve already cut emissions by 21%. And we refuse to be satisfied. We’re committed to reducing emissions by 50% by the year 2050, creating the potential to reduce 12% of all US emissions by this year.

Ultimately, by protecting our climate, we’re rendering these rapid evacuation plans less, instead of more, relevant as the years pass.