Towns, cities turn to computers to help with snowplowing
Before snowfall started Tuesday night in Sudbury, Mass., salt trucks had been pretreating roads for more than two hours. When snow began to stick to the ground, it was time for the 11-footwide snowplows to take to the streets, cleaning main roads and residential areas. Operators would be plowing and laying down salt for the next several hours until the storm, a modest one this time, subsided.
“The only way that I can describe plowing to someone who’s never done it is you have to learn to get comfortable being uncomfortable,” said Brian Hawes, a foreman at the Sudbury Department of Public Works who was operating one of the snowplows during this week’s storm. “It’s snowing out or it’s raining out really hard, and you can’t see, but you still have to.”
A snowplow operator’s goal sounds simple: Plow where the snow is. But how do drivers know what route to take? How many snowplows are too many on the road? And why do they always seem to clean my street last?
It turns out that today’s snowplowing practices are more old school than you might think: hand-drawn routes, printed maps. But some cities are showing there’s a more advanced way. Recently, more towns, cities and states are turning to software that can outline the most efficient snowplow routes for an area, providing automated turn-by-turn directions. The result is faster, safer cleanup and fewer snowplows on the streets. It also begs a common question: Who can get the job done better? Human or computer?
Hawes likened the work to a battle with the elements — one that requires more resources than people may realize. “I think [people] would be surprised about the kind of the science that’s going into snow fighting,” he said.
Dan Nason, director of the Sudbury Public Works Department in Massachusetts, said most cities and towns have routes planned and printed out. The routes, which drivers end up memorizing after repeated runs, have typically been used for years despite changes in town growth or resources. “But are you doing it in the most optimal fashion? We’d have to do that analysis,” he added.
To answer that question for Sudbury, Nason decided to analyze the routes himself when he joined the department. Like a 15th-century explorer (except it was circa 2018), he sat down with the map of the community and plotted out a snowplow’s journey around town. He outlined which main roads took priority over secondary roads and broke up each path into equal lengths as best as he could. He knew preferred snowplowing practices, such as avoiding left-hand turns into intersections. In the end, he reduced the number of snowplow routes from 52 to 41, printed and laminated the maps and gave them to each driver.
The problem, Nason said, is handing a physical map to a driver on duty can be inefficient as well. “You’re in the middle of the night. The snow is coming down. It’s hard to see. You can’t really follow a map,” Nason said.
He also still wasn’t sure if his planned pathways cleared a “route in the most optimal fashion.”
Enter the mathematicians. The mathematics behind any route optimization dates back to a famous brain teaser from the 18th century. In the town of Königsberg, Prussia, seven bridges spanned a river and island, but one man wanted to walk along each bridge — and only cross each bridge once. Swiss mathematician Leonard Euler found a solution is impossible, but the investigation gave rise to the field of graph theory, which models the relationship between lines and points.
A similar math problem appeared in mail delivery centuries later. In the early 1960s, Chinese mathematician Kwan Mei-Ko discovered the route inspection problem, also called the Chinese Postman Problem, in which a postal worker wanted to travel along every road in a city to deliver mail while covering the least possible distance.
Mathematicians found the most efficient routes occur when there’s an even number of entries to a street — one way to enter and another way to exit. When intersections have an odd number of streets, you have to backtrack. As a result, one approach is to find the most efficient routes among odd streets.
The postal worker problem is “similar to snow plowing in the sense that you now have a vehicle that needs to traverse every street at least once,” said mathematician and computer scientist Joris Kinable, who published algorithms that improved snowplow routes for the city of Pittsburgh.
“Unfortunately, if you start adding additional, what we would call side constraints, that you would typically see in snow plowing, then the problem becomes much more complex to solve,” he said.
The list of snowplowing constraints are long. Certain critical roads, such as those going to hospitals, need to be cleaned first (and residential areas may be prioritized lower). Other constraints include one-way streets or road restrictions for vehicles weighing thousands of pounds. Recent driver shortages have also stressed many departments, causing some workers to extend their routes to cover more ground.
“It’s a different kind of person who wants to become a highwayman,” said Hawes, who has worked at the Sudbury Department of Public Works for 30 years. “The hours aren’t for everyone. The commitment isn’t for everyone.”
Several commercial companies have entered the field of route optimization. After trying to optimize his routes by hand, Nason hired a company called RouteSmart Technologies, which began operations in the 1980s, to analyze his routes. The algorithm reduced his fleet from 41 to around 32 snowplows, where each route has a twohour maximum. If one driver called out that day, Nason can also redistribute that driver’s route among the other snowplow operators, so no one person has to carry the brunt of the extra work.
The software also increased the number of trucks that treat the road with salt, which operate as a separate fleet from snowplows, from five trucks to nine. Each salt truck and some snowplows are equipped with turn-byturn navigation systems, which Nason said makes it easier to train new employees or one-time contractors on their route.
The city of Centennial, Colo., which used route algorithms from a company called Jacobs, reduced the driving time of its snowplow operators by 40 percent.
Nick Repekta, the highway division manager in Shrewsbury, Mass., hired a company called Quetica, which better balanced the length of the routes from his salt trucks. Previously, the shortest truck route was 18 miles, but the longest was 34 miles. Now, all salt trucks, which also use the turn-by-turn navigation, cover roughly 24 miles and arrive within 10 minutes of one another at the station.
Many places, including winter wonderlands such as Minneapolis, do not use route-optimization software. Others, such as Waukesha County located outside of Milwaukee, used the software to gather new routes years ago but haven’t updated them. One reason is using route-optimization algorithms requires a lot of data input into the software, setup time and testing the routes in real life.
Repekta also cautions the computer isn’t always a better option. His crew in Shrewsbury did not use the new computerized routes for their snowplows. For one, the number of routes didn’t change, still requiring 33 snowplows. The computer-generated pathways also didn’t prioritize main roads as well as the status quo. In the end, he felt his original routes were the most efficient.
Other snowplow operators and managers agree: it’s the human driving the snowplow, not the computer.
“No matter what these technologies bring, it’s great information and fantastic tools to give you a better understanding of the snowstorm, better understanding of the best way of approaching it,” Nason said. “However, you still own your route.”