News - General News - Technology
Tech guru dismisses Google car as 'a joke'
Google driverless car highlights autonomous limitations, says tech leader
20 May 2015
By IAN PORTER
A LEADING software analyst has described Google's driverless car as “a joke”, criticising its inability to perform simple tasks such as changing lanes on a freeway.
Google's autonomous car has clocked up more than 1.6 million kilometres of testing on private and public roads, and the tech giant has announced that it will start trialing a few of its bubble-shaped cars on the roads of Mountain View, the Californian city where Google is headquartered.
But Russell Shields, chairman of leading software company Ygomi – which is working with several car-makers on automation, telematics, in-car applications and services for electric vehicles – says despite raising the profile of autonomous cars, the Google car is not achieving a lot.
“The Google thing is a joke,” he said at the Australian Intelligent Transport Systems Summit in Melbourne last week. “It’s very good in that it has got people aware of automated driving, but they are so far away from being practical.”
Google’s hardware and software systems have now logged the equivalent of 75 years of typical American adult driving, and the fleet is currently covering 16,000km a week.
The cars will be limited to 40km/h and the company will be assessing how other drivers perceive and interact with them. The test team will monitor how the cars manage real-life situations such as inability to park at a desired destination.
Mr Shields said that, while the hardware needed for autonomous driving was getting better and cheaper, progress with the software was more difficult to achieve.
“But the software is much, much, much harder. Just in general, that’s becoming the biggest challenge for the whole vehicle industry.”
And the challenge is not just getting the software right. Already, more than half the recalls of new cars are caused by software issues, Mr Shields said, and the software needed for Level 3 and Level 4 automation will be much more complex than what Level 2 cars have today (see levels explanation below).
Mr Shields said that, while there are Level 2 vehicles now available and on the roads, the sticking points that are holding up the move to Level 3 are the task of changing lanes on a multi-lane expressway, and the issue of handing back control to the driver.
Cars already offer systems such as cruise control lane-keep assist, but Mr Shields said that is basic compared with some of the tougher challenges offered by driving.
“Doing that on the expressways is pretty much doable.
“But we have some challenges in some cases like on a complicated expressway – four lanes, lots of exits and entrances with cars on it – changing lanes is not easy.
“Google is playing around with that, but they have no way of changing lanes.
“How we humans do it is we look at the guy in the other car, we try to get his eye and we stare at him and we sort of creep into his lane and will he let us go?“The automated car doesn’t know how to do those sorts of things. That’s a significant issue still, probably one of the hardest software issues still, how do you, on a complicated expressway, deal with lane changing and traffic.”
Mr Shields also highlighted situations where a driver needs to be in the correct lane well in advance of the required off-ramp.
“We are trying to figure out the way can get CAD-type vector databases of the road network so you can figure out that, four kilometres from here you need to be two lanes over, and start trying to get there ahead of time.”
Mr Shields also touched on the issue of the automated vehicle handing control back to the driver.
“How can you guarantee a driver can take control? No-one I know actually knows how you can guarantee the driver will take control of the vehicle,” he said.
“We don’t know how to test for that.”
Mr Shields said the car companies can use simulators to test how long it takes a driver to get their hands back on the wheel when the car hands back control, but added that it was a more complicated issue.
“Why is the car giving back control to you? Well, it’s because it doesn’t know what to do. Something is complicated, dangerous or different.
“What are you going to do? Suddenly holding the steering wheel is the easy part.
“You have to get the whole situational awareness around you of what’s going on and not jerk the steering wheel and do something dumb because you see a car over here but there is something on the road over there.
“Simulators aren’t going to test that kind of thing. That’s real serious stuff, figuring out what to do in a unique situation.
“That’s a tough problem and it’s only going to be resolved by real testing in the real world, on regular roads.”
And, according to Mr Shields, that means accidents are likely.
“We’re going to have some car company engineers injured or killed, or what have you, and we are trying to figure that out.”
Mr Shields also pointed out that drivers are going to have to regain control of the car in what are essentially no-win situations.
“I’m following the speed limit at 85mph (136km/h) and suddenly I have to focus.
I don’t know what the conditions are: a hole in the road, or somebody has dropped a refrigerator off their truck in front of me and there are cars on both sides of me.
“So what happens? The automated car turns control back to the driver and lets him make a mess out of it. We don’t know.”
He said car manufacturers will likely be keen to prevent image-damaging collisions by augmenting the sensing technology outside the car with sensing technology inside the car.
“Most of the car companies I know, that we are working with, are going to have a camera up by the mirror watching the driver and it’s going to make sure he is not playing with his girlfriend or his iPad.
“I think they will allow you to watch the football game on the car display, but they are not going to let you play with your smartphone (in Level 3 autonomous mode),” Mr Shields said.
The National Highway Traffic Safety Administration in the United States deems there to be four levels of automation in vehicles.
Level 1 has one or more control functions, such as cruise control, allowing the driver to cede limited authority.
Level 2 has at least two automated functions that work in combination, such as range sensing, cruise, brake or throttle control, which relieve the driver of those functions.
Level 3 allows the driver to cede control of all safety functions, including steering, brake and throttle control and lane navigation, but the driver can regain control with sufficient transition time.
Level 4 sees the vehicle perform all safety functions and monitor road conditions for the full trip. Human presence is optional.
The Road to Recovery podcast series
15th of May 2015
Automated cars to require expensive infrastructure
Driverless cars will change some of the basic concepts of motoring
15th of May 2015
US trial to test how ITS and automation combine
Autonomous transportation system could be operating in Michigan early next decade
24th of December 2014
Google polishes self-driving car
First complete prototype of Google autonomous car ready for testing
7th of November 2014
Society must keep up with driverless car tech: Google
Google says society must accept pace of tech change as driverless cars near
Click to share
General News articles
Research General News
Motor industry news