How will a self-driving car negotiate a jam-packed rotary at rush hour?
A startup based in Cambridge and Singapore, Nutonomy, is already pondering that question — and crafting the software that will hopefully avoid making Boston’s rotaries even more harrowing as our cars chauffeur us around. And Toyota is hiring people, also in Cambridge, for a new research lab for self-driving cars that will tackle the rotary and other challenges we’re familiar with, such as “construction projects that change where you’re supposed to drive,” said Gill Pratt, chief executive of the new Toyota Research Institute.
Ever since Google started its self-driving car initiative in 2009, much of the action has been in California. That’s where Google is developing its cutesy coupe, which looks like a partnership with toymaker Fisher-Price, and where Tesla is continually rolling out software updates that enable its electric cars to do things like enter or exit a narrow garage without the driver being inside the vehicle.
California, Nevada, Florida, Michigan, France, and the United Kingdom are racing ahead with so-called autonomous vehicles, making it possible to test or legally drive them on public roads. But here in Massachusetts, we’re still looking for our keys. Regulators have yet to adopt similar rules that would make it possible to take driverless cars for a spin on public roads, and the next generation of autos has yet to capture the imagination of a wide swath researchers.
“Cars haven’t been seen as a sexy research topic at MIT,” says Karl Iagnemma, an MIT research scientist and chief executive of Nutonomy. “You need space to test cars. Universities like Stanford have space — they created a research center and a consortium model with companies. MIT failed to do that.”
Iagnemma’s company is trying to position itself as a supplier of software to carmakers that want to add sophisticated features to their vehicles — like getting into a tight parking spot on their own, or handling dull highway driving safely. Iagnemma says that the toughest thing for self-driving cars will be city driving, where you have to worry about pedestrians, bicyclists, and road closures that likely won’t show up on even the best digital map.
“Google is very advanced,” says Iagnemma, “but we don’t think anyone has a solution that would work everywhere in the world today.”
One of the really cool things the company has been working on is software that tries to predict what pedestrians are going to do next, and be wary around those behaving erratically — like a group of carousers leaving a bar late on Saturday night.
But Massachusetts is still coming from behind. Nutonomy has 22 employees; all but six are in Singapore, where the company has an easier time conducting road tests. Between 2012 and 2015, Apple, reportedly working on its own self-driving automobiles, has hired more than a dozen veterans of A123 Systems, a Waltham supplier of lithium ion batteries to carmakers, to work on unspecified battery-related projects, and shipped most of them out to California.
“Our state is missing the boat,” says Emilio Frazzoli, an MIT professor who took a leave earlier this month to focus on Nutonomy full-time. “It’s crazy for a state that promotes its brand as being technology. We have driven zero miles in Massachusetts so far.” (For comparison, Google said last year that its vehicles had already chalked up 1.8 million miles.)
In October, Toyota announced that it would invest $1 billion in a new initiative led by Pratt, a former MIT and Olin College robotics professor who lives in Lexington. The Toyota Research Institute, with labs in Cambridge and Palo Alto, Calif., will explore topics such as how robots might assist the elderly in moving around their homes, and ways artificial intelligence software might help research scientists to be more productive. But Pratt says it will also explore ways to make cars smarter — and safer for drivers who “may be visually-impaired, old, tired, or young.”
Pratt says it’s hard to overstate “how hard the driverless car challenge is.” How so? “Human beings demand near-perfection from our machines, if we’re trusting them for life and death decisions.” Essentially, while we accept that human drivers are fallible — Pratt says there are about 1.2 million deaths a year because of human error, worldwide — “we insist that machines be far better.” In other words, we likely won’t acknowledge that reducing that 1.2 million figure by 50,000 or 100,000 would be a big deal, if autonomous vehicles are responsible for the remainder of the accidents.
Toyota is funding research at MIT into topics like how self-driving cars will handle traffic congestion, bad weather, and unexpected objects in the road. (Sofa!) The company has also hired as part-time employees two MIT professors who have done work in the field of autonomous vehicles. Over time, says Pratt, the company plans to hire about 50 people for its Cambridge lab.
Several state agencies, including the Massachusetts Department of Transportation, are finally planning a “listening meeting” next month, to discuss potential regulations for self-driving vehicles. In attendance will be representatives from carmakers, startups, academic institutions, and government-affiliated research labs. On the agenda: what other states have done right and wrong in setting up frameworks for operating autonomous vehicles on public roads.
If you’ve ever watched someone video-chatting on a mobile phone as they wait for a light to change, or texting as you pass them on the Pike, it’s not too difficult to believe Iagnemma when he says that self-driving vehicles may prove to be “a massive public health good, a way to reduce accidents and lost lives.” Even better, he says, most of them may be powered by electric motors instead of exhaust-spewing gasoline engines, and smart enough to drive to a charging station after they’ve dropped you off at work.
In the 20th century, the way car makers improved the safety and reliability of their products was all about designing better hardware — crumple zones, anti-lock brakes, airbags. In the 21st century, as Pratt argued in a recent speech, it will be about software and data. Startups, carmakers, and large tech players like Google and Apple will all be vying to develop the best software, and leverage data in the smartest ways.
“This is a frontier market and there is lots of room for innovation,” says Cambridge venture capitalist Eric Paley, an investor in a San Francisco startup called Cruise Automation that is developing an autopilot system for cars already on the road. (One of its vehicles was involved in a minor accident earlier this month.)
A decade from now, when someone rolls down their window to yell at you in a rotary, there may still be profanity involved — but it may be a brusque suggestion to upgrade your software.
0 Comments:
Post a Comment