Human roadblock for Japanese firms developing autonomous cars

November 15, 2015 by Anne Beade
The Toyota logo on a Sienta car displayed at its headquarters in Tokyo
The Toyota logo on a Sienta car displayed at its headquarters in Tokyo

Japanese car manufacturers will have to convince the public that letting go of the wheel in a self-driving car is safe, while also dealing with the biggest threat to the cars' security: the humans using them.

Toyota, Nissan and Honda are intent on putting autonomous cars on highways—and also city roads for Nissan—by 2020, and the triumvirate of Japan's auto industry were keen to stress the advances made so far at the recent Tokyo Motor Show.

Their stated goal—preventing deaths on the road—is laudable, but the technological arms race is also highly lucrative: consultancy firm AT Kearney has estimated the market for the self-driving car could be worth more than $566 billion by 2035.

Nissan chief executive Carlos Ghosn told reporters at the Tokyo show the company has high hopes the technology will save lives while altering car journeys forever.

"It compensates for human error, which causes more than 90 percent of all car accidents.

"As a result, time spent behind the wheel is safer, more efficient and more fun," he said.

But Ghosn's comments belie the work still to be done, as its engineers edge forward in steps rather than leaps.

Google offers promises of a fully autonomous car, but these automakers are taking a more gradual approach, focusing on aspects such as self-parking and crash avoidance technology.

Functions such as emergency braking and speed-limiting devices that track the distance between vehicles already exist, but getting drivers to abandon the steering wheel completely is a harder sell.

"We must make sure our clients understand how the machine works," said Nissan's chief planning officer, Philippe Klein.

To instil confidence, the artificial intelligence that will power Nissan's autonomous cars will mirror the driver's driving style as closely as possible, while "ironing out any bad habits", the automaker said.

A driver tries out a Lexus GS450h on the Tokyo metropolitan highway during Toyota's advanced technology presentation
A driver tries out a Lexus GS450h on the Tokyo metropolitan highway during Toyota's advanced technology presentation

Intersection conundrum

Obtaining the trust of drivers is crucial, as without it "we cannot move forward", said Moritaka Yoshida, a Toyota executive.

And even if the user of a self-driving car is convinced of its superior safety, other road users need to feel secure sharing the tarmac.

Manufacturers are experimenting with icons or written messages appearing on wind-shields, warning sounds, and in one case a light-strip along the length of the car whose colour and intensity would alter in different situations.

Intersections present a particular challenge, said Melissa Cefkin, who is based at Nissan's Silicon Valley research centre.

"Sometimes drivers communicate between themselves and with pedestrians or cyclists directly, by swapping looks, with a hand gesture, or even verbally," she said.

"Sometimes it's interpretative: we look for signals while judging the vehicle's speed and movements."

The tiny pointers that motorists pick up from one another are not yet within the reach of the technology.

"Currently, the machine isn't capable of grasping all the subtlety of these clues," Cefkin said.

To better understand them, Nissan is undertaking the immense task of studying thousands of intersection scenarios in an attempt to identify cultural patterns by country or context.

U.S. Transportation Secretary Anthony Foxx (R) and Google Chairman Eric Schmidt (L) get out of a Google self-driving car at the
U.S. Transportation Secretary Anthony Foxx (R) and Google Chairman Eric Schmidt (L) get out of a Google self-driving car at the Google headquarters on February 2, 2015

Regulatory roadmap

Besides, once everyone on the road feels comfortable with in their midst, auto firms must still convince regulators of their safety before they can hit the streets.

"Today, you have to drive with your eyes on the road and your hands on the wheel. If the regulation doesn't change, having a self-driving car will be totally useless," said Nissan's Ghosn.

"Everything depends on public-private cooperation," he added.

Car companies are already conducting intense exchanges with such authorities across the globe, said Toyota's Yoshida, adding he hoped that Europe, Japan and the United States would succeed in shaping "global norms" for the industry.

This month, Toyota unveiled a vehicle that can drive itself along a highway, but CEO Akio Toyoda is acutely aware of the legal minefield his firm must navigate before putting it on commercial sale.

"Imagine if a major accident occurred that implicated a self-driving car. We have to go step by step," Toyoda said.

And while humans may be the architects of their own downfall on the roads, self-driving technology cannot yet match our level of sensory perception, emphasised Honda's chairman Fumihiko Ike, urging caution over the limits of the technology in the short term.

"Human intelligence has no equal for working out what is happening on the road, so I think fundamentally it won't be easy to leave it to the machine except in very restricted conditions such as motorways or specific routes," Ike said.

Explore further: CEO: Nissan will be ready with autonomous driving by 2020

Related Stories

Recommended for you

Microsoft aims at Apple with high-end PCs, 3D software

October 26, 2016

Microsoft launched a new consumer offensive Wednesday, unveiling a high-end computer that challenges the Apple iMac along with an updated Windows operating system that showcases three-dimensional content and "mixed reality."

Making it easier to collaborate on code

October 26, 2016

Git is an open-source system with a polarizing reputation among programmers. It's a powerful tool to help developers track changes to code, but many view it as prohibitively difficult to use.

Dutch unveil giant vacuum to clean outside air

October 25, 2016

Dutch inventors Tuesday unveiled what they called the world's first giant outside air vacuum cleaner—a large purifying system intended to filter out toxic tiny particles from the atmosphere surrounding the machine.


Adjust slider to filter visible comments by rank

Display comments: newest first

5 / 5 (3) Nov 15, 2015
How can AT Kearney be so accurate to estimate the market for the self-driving car could be worth more than $566 billion by 2035. Surely its something like $500-$600 billion--and even that is probably stating with too much precision given known unknowns and unknown unknowns.

3.4 / 5 (5) Nov 15, 2015
Sometimes drivers communicate...with a hand gesture, or even verbally,
I'm shocked - we would never do such a thing here! Maybe the cars could be fitted with a middle finger, and vocal cords.

2.3 / 5 (3) Nov 15, 2015
It all started with algorithm based (individual wheel) anti-lock braking, for regaining control in difficult situations. (feedback systems)

This progressed to driving aids for Lamborghini and Ferrari, give the driver the illusion of being superior at the wheel, while not killing the millionaire weekend warrior. A bit of hard stroking for the rich clientele, sold to them as a benefit ----not a nanny.

Driving the car will become pure illusion.
3.4 / 5 (5) Nov 15, 2015
"Currently, the machine isn't capable of grasping all the subtlety of these clues," Cefkin said.

That's a HUGE understatement. The current level of technology isn't even capable of grasping what an intersection is without GPS and a pre-recorded route map that tells it where to drive down to the centimeter.

The way e.g. the google cars work, you drive them through the route once, they 3D-scan the whole way to record how it looks like to make a map for themselves, align it with the GPS record of the route, and the human trainer marks where the intersections are, what lines to take, where to look for the traffic light signals etc. and stores it in an online database for later use.

It doesn't even understand what a road is. The computer itself just sees "flat area here, not so flat there", so it can't take any turn independently. It's basically an elaborate line following robot with limited capability to not collide with temporary obstacles.

3 / 5 (2) Nov 15, 2015
humans are model recognition based.

the clue is that model recognition systems might work for this application, IF it does as humans do, which is constantly rejigger the models in relation and realization, in situ.

At least, smart drivers are always in the act of re-compensating and re-realization in situational awareness and analysis. These aspects shift within the given situation, ie highway vs city, speed, etc.

The problem comes in building the model recognition database.

For the larger part, humans do it via SIGHT ALONE, so it is possible to do the same, with stereoscopic image capture, as this is how humans do it.

2.6 / 5 (5) Nov 15, 2015
The most difficult thing that Google does with their autonomous cars is the combination of GPS data with local 3D ranging and scanning to keep an accurate and reliable positioning data. Once you got that, all you need is for someone somehow to draw a line in your virtual map and tell the computer "follow this".

The cars are much dumber than the media hype would lead you to believe. They only do the absolute minimum amount of "thinking" necessary not to crash into things.

The idea is that eventually Google would have a huge central online 3D-mapped database of roads that is continuously updated by the cars, so they can externalize the thinking from the cars to the central database where people with supercomputers can clean it up and plan the routes that the cars should take.

The cars themselves are simply like trolley busses under virtual overhead lines. You can't arbitrarily send them from any point A to another B unless there exists a "wire" for that route.

2.6 / 5 (5) Nov 15, 2015
For the larger part, humans do it via SIGHT ALONE, so it is possible to do the same, with stereoscopic image capture, as this is how humans do it.

It is possible, but it also happens to be very hard. The whole rear lobes of your brain, about a quarter of the mass, is devoted to visual processing. Then you also need the cognitive ability to take the model you have of the world and understand what is a road, what is an intersection, what is a traffic sign - for example so that you don't speed up to 100 mph just because someone put up a round piece of cardboard with the number on it.

The current crop of autonomous vehicles do absolutely none of that.

They don't see, they use lidar or radar, they don't hear because they don't listen - they couldn't make any sense of audio in the first place because they don't think; all the thinking has been outsourced to the human trainers. The cars are literally deaf, dumb and blind.
2.4 / 5 (5) Nov 15, 2015
That's actually a good test, a kind of reverse-turing for autonomous vehicles.

If I paint a fake road sign on a piece of plywood and stick it up next to the road, a human driver will instantly recognize it's wrong. They won't follow it.

The autonomous vehicle - because the programmers are trying to get around the problem the easy way - will not recognize a fake roadsign from a real one because it looks close enough that simple statistical match algorithms would break down.
1 / 5 (2) Nov 15, 2015
How many billions in revenue do these companies risk if they continue to interfere with a driver's ability to control the car?

I don't intend to buy a self braking car. I will switch to a different brand without the bells and whistles or I will buy an older car in good condition which does not endanger my life and the lives of my passengers.
Lex Talonis
5 / 5 (1) Nov 15, 2015
Autonomous vehicles... SAFER - yes, if programmed to the Nth degree with a very large dataset, and all of the information is aggregated, up dated and there is wireless connectivity etc.,

But like this is like having super computers, and failing to include the pencil and paper option.

Malware, hackers, power outages, storms, floods, Evil Government Controls, land slides, washING out roads and bridges and things that deviate from the norm... like there is an aircraft falling out of the sky and it's going to land on us, unless we turn left - and the idiot car has no steering wheel.... (Google Morons) and it keeps right on going.......

Then there is the statistical overall saving the masses from crashes, and the minority of unaccounted for issues that do cause crashes, injuries and deaths.

It's the TOTAL lack of any control over any thing, to do with being in a moving vehicle, that disturbs me...
5 / 5 (2) Nov 15, 2015
Well, it sounds most of you are saying that these self-driving cars are lacking in one essential element, awareness. Of course, if they had that then they would eventually realize that, as the author stated, "the biggest threat to the cars' security: (is) the humans using them", and do something about it.
5 / 5 (1) Nov 16, 2015
Once you ask the driver to let go you have to have FULL ability to drive that car without human intervention. You cannot expect the driver to do nothing and still be ready to jump in if something goes wrong. I guarantee you 99% of drivers will fail at that.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.