Early pioneers didn’t realize just how big the Rocky Mountains were until they started climbing up the foothills. In many ways, it’s the same challenge for pioneers in the new world of autonomous vehicles, suggests Gill Pratt.
One of the biggest challenges is trying to learn what you never thought about before – call it “uncertainty on uncertainty,” says the new CEO of the Toyota Research Institute. Funded to the tune of $1 billion, TRI is the think tank Toyota has set up to help it develop self-driving vehicle technology.
That’s one of the most competitive fields in the auto industry today, with some manufacturers hoping to have their first, fully-autonomous vehicles on the road by 2020. But in a keynote speech at the 2016 Consumer Electronics Show in Las Vegas, Pratt throws colder water on those optimistic expectations, warning that, “We’re a long way from the finish line.”
At a show traditionally focused on TVs, digital cameras and smartphones, the auto industry has staked out a major presence at this year’s CES, and the news conference schedule is crowded with shows focusing on autonomous technology. Ford, for example, announced it is tripling its fleet of self-driving prototypes on Tuesday, and Kia revealed its own autonomous vehicle plans.
Until recently, Toyota had been one of the most skeptical of manufacturers, insisting it was limiting the scope of its research to driver assistance systems designed to help a human remain in control. The recent formation of TRI marked a shift in direction, Toyota executives now laying out plans to put driverless vehicles into production – eventually.
For his part, Pratt might best be described as a cautious proponent. He notes that while there are already scores of autonomous prototypes running around on public roads, what has so far been accomplished is “relatively easy because most driving is easy. Where we really need help is when driving isn’t easy.”
(Toyota introducing new autonomous mapping technology. For more, Click Here.)
That can include snowstorms, poorly marked roads, crowded urban centers where drivers and pedestrians don’t obey the rules – or where a critical vehicle sensor suddenly fails.
“We need reliability a million times better than it has been,” he declares.
Pratt is by no means the only one waving the yellow caution flag. Google, considered the leader in autonomous research, recently reported the 16th crash involving one of its self-driving prototypes. In every instance to date the human driver in the other vehicle has taken blame.
At least legally. But Google officials have acknowledged they need to better understand how human drivers behave and that may mean they have to program their cars to occasionally run the yellow light when the car behind isn’t likely to stop.
(Click Here for details about Toyota’s $1 billion artificial intelligence center.)
The closer researchers get to the mountain, Pratt implies, the more challenges they face. Indeed, facial signals are something human drivers rely on – when they’re trying to determine how to negotiate a four-way stop, for example, or are dealing with a pedestrian who may or may not walk out into traffic.
Simply getting an autonomous vehicle to recognize a school crossing guard or a construction zone or accident site not on the digital map poses major challenges.
Where “society tolerates a lot of human error, we expect machines to be perfect,” says Pratt, a former head of autonomous research for the quasi-military Defense Advanced Research Projects Agency – which also helped develop what eventually became the Internet.
Despite his cautions, Pratt believes autonomous vehicles will “eventually” come to market and that they will have a major impact on automotive safety. Toyota’s goal is make its self-driving vehicles unable to crash on their own. If he’s right, the payoff would be tremendous. About 32,000 Americans are killed in traffic accidents each year, with 1.2 million such deaths worldwide.
(To see more Ford’s mobility plans explained at CES, Click Here.)
How soon that will happen is a matter of intense debate. Nissan is shooting for 2020. Kia outlined its own autonomous vehicle goals during a CES presentation, targeting 2030. For his part, Pratt says there are simply too many mountains to climb to come up with a target date he can’t be confident about.