Futuristic driverless personal cars that can operate safely in rain, hail or shine remain a distant reality, says Toyota.
However, autonomous ride-share cars restricted to covering certain areas in certain conditions are a very different story, given they're being prepared for commercial sale inside a decade, and could technically operate today as technology demonstrators with some clever “hacking”.
CarAdvice recently spoke with ex-Google engineering director James Kuffner, who’s now the CEO of Toyota’s Advanced Research Institute (TRI) focusing on car-to-car communication and self-driving evolution through exponential ‘deep learning’, and other trick algorithms.
We asked Dr Kuffner to give us a clearer picture of how the world's biggest car company is tackling the issue, and whether the wider industry’s notorious over-promising around when and where we will experience these driverless cars has been an impediment to acceptance.
For background: the common ‘scoring card’ for autonomously driving cars contains five levels.
Most advanced cars of today are at level 2, able to briefly steer between lanes using cameras and radars, to mirror the speed of a car ahead, and to automatically emergency brake. Toyota/Lexus itself will deliver automated driving “on highways” to cars next year, as part of a technology suite called ‘Teammate’.
But what everyone thinks of when they think about driverless cars is Level 5, which denotes a true autonomous vehicle enabling complete reliance on the AI “equivalent to a human driver under any weather condition, any traffic, with a map, anywhere in the world”.
“Let me just be really clear: nobody is even close at that level,” Kuffner said.
“If you say level 4, now you're adding restrictions. You’re saying it’s geofenced, it’s lighter traffic, lower speeds, good weather. Then the answer is today, we could do that if we restricted things. Even dedicated lanes, good weather, light traffic, good maps, we could deploy that today.
“So the real question is, how much are you restricting it and whether or not that's providing value.”
“And the other thing is, unfortunately, it isn't that hard to make a demonstration car. Colleagues and I could spend a weekend and hack a Prius, and say ‘hey we just made a self-driving Prius! But it would be nowhere near a product,” he added.
What this means is, a product sold publicly needs to be tested far more exhaustively, and given the massive cost of roof-mounted LIDARs and other necessary technologies such as rapid wireless communication streams, they’d be hugely expensive. Not that this is necessarily an issue in all cases.
Kuffner continued: "[But] there are two products often conflated in the media. One is a car you sell to a customer, the other is a fleet-owned and operated, everyday-maintained driverless taxi. That’s a completely different product because its higher utilisation amortises the higher costs.
“It is true that… overpromising by other people does sometimes hurt us, but we do believe this technology is going to have a huge impact. One million people dying a year because of traffic accidents is a travesty. We’ll try to accelerate it ourselves and through partnerships,” he added.
As if to emphasise that final point, TRI last week announced today that its Platform 4 (level 4) automated driving test Lexus LS with its ‘Lexus Team Mate’ suite of tech including roof-mounted lidars, will be available for public demonstration rides next summer in Tokyo, from July to September.
“The P4 experience will take place in Tokyo's Odaiba district, a busy and often congested waterfront sub-centre. Odaiba's complex environment of pedestrians, vehicle traffic, diverse road infrastructure and tall glass buildings provide a challenging setting in which to demonstrate the capabilities of Toyota's automated driving technology,” the company says.
The public will be invited to register for the experience, and individuals will be selected to participate. In accordance with Japanese law, a Safety Driver will be present during the experience.
Lexus’s publicised rollout plans show it will embark on larger-scale field tests of this very technology in Beijing in the "early 2020s" and offer service deployment of full highway automation and automated driving on surface roads (urban roads) in the late 2020s.
Much of this echoes what another leader in the field, Mercedes-Benz’s global board member for R&D Markus Schäfer, told this writer recently, while discussing thorny issues such as sensor fusion, lidar cost, AI development, and cloud-based exponential learning.
“It’s very hard to predict, obviously the challenges are way bigger than anybody thought... from my point of view it’s going to take a couple more years, then you might have some usage in some restricted areas, maybe on a highway which is an easier case and will definitely come earlier.
“If it’s an urban environment it would be limited routes [only]. Free-floating routes in all areas, that’s not very near,” he said.
Consensus on timings from Daimler AG and Toyota is significant.