Copyright The Boston Globe
Rideshare drivers said they feared sudden and devastating job loss should autonomous taxis flood the city. Skeptics questioned whether the robotaxis are clever enough to navigate Boston’s compact and confounding streets. Others fretted that self-driving cars could worsen safety on roads already plagued with aggressive and lawless human drivers. But perhaps one of the biggest, if less politically charged, hurdles mentioned was the weather: can driverless cars handle Boston’s snowy winters? “We all know that sometimes Boston can have very terrible winter temperatures, with snowfalls and black ice and unsafe roads,” Boston city councilor Enrique Pepén said during the hearing. “Winter weather capability... [has] been a primary area of focus for development,” replied Matthew Walsh, a representative for Waymo, the driverless car giant that completed its mapping project in Boston earlier this year. “We’re continuing to get more and more confident and more and more assertive in operating in those conditions.” Snowy environments present a number of unsolved technical hurdles to robot drivers, according to experts. The software controlling a Waymo car, anthropomorphized by the company as the “Waymo Driver,” relies on a system of finely-tuned sensors and cameras to register its surroundings, according to the firm. The car’s on-board computer ingests information about the speed and position of nearby pedestrians, cyclists, vehicles, and other items in the road, dictating the car’s movement accordingly. But snow can blur cameras and muddle the car’s sensors. Snowfall and its slushy residue can also obscure road markings, helpful cues for direction and spacing. And autonomous car software also has to adjust its driving dynamics when traveling slippery streets. “Snow is just unique in its blockage,” said Jonathan How, an engineering professor at the Massachusetts Institute of Technology. “It’s hard to see through it, and when it falls on the ground and sticks, it tends to cover things.” Robocars, much like tourists or transplants from warmer climates, must also acclimate to the strange and daunting features of harsh winters. “We just don’t have enough data in those environments,” said John Dolan, a professor at Carnegie Mellon University. “The primary deployments [of autonomous vehicles] have been in nicer weather places.” Waymo’s autonomous taxi fleets currently offer rides in Los Angeles, Phoenix, San Francisco, Austin, and Atlanta, according to the company’s website. A subsidiary of Google’s parent conglomerate Alphabet, Waymo is teeing up deployments in Miami and Washington, D.C. next. To a driverless car trained exclusively in the sun belt, snowbanks and snow-coated cars are foreign sights. “Everything starts looking different,” said How. “Intersections just have things in them that they didn’t have before.” Training the software to navigate in winter climates will require extensive testing in winter conditions, experts said. “You need to collect the data,” said Hayder Radha, an electrical and computer engineering professor at Michigan State University. “You’re going to have to do the butt miles of driving around here and collecting the data that helps you understand what the differences are really like in those types of environments,” said How. Technological breakthroughs — more robust sensors or sharper cameras with tough, miniature wipers — could also help edge progress along, experts added. Waymo, for its part, said it is up for the challenge and working its way toward a solution. “For years, we’ve been advancing our system in some of the snowiest conditions across the country — regularly driving in Upstate New York, Michigan’s Upper Peninsula, and the Sierra,“ the company wrote in a blog post published on Monday. ”We’ve amassed tens of thousands of miles in diverse, snowy conditions. This has allowed the Waymo Driver’s AI to learn from real driving experience and train to navigate a wide range of winter weather.“ Waymo could theoretically begin deploying its fleet in Boston only in warmer months as its software trains to conquer the snow. Even human drivers tend to avoid roads when the worst snowstorms hit. But such a strategy might sow doubts about the robotaxis’ reliability, How said. “If you want to be a system that is deployed and is robust, in the sense that people use you, then you can’t have out days,“ said How. ”You can’t have days where you phone in sick and say, ‘We as a company can’t operate on this day because it’s too much slush.’“
