GM has begun road testing prototype Cadillacs capable of semi-autonomous driving. GM calls the system ‘Super Cruise’. just-auto spoke with Dr Nady Boules, director of GM’s Electrical and Control Integration Research Lab. He outlined how GM sees the pathway, timeframe and costs in moving through semi- to fully autonomous driving.
In GM’s vision, fully autonomous driving is the only way to get cars that don’t crash. “We didn’t just wake up one morning and say ‘Let’s do cars that drive themselves,’” Boules says. “It was a result of developments for safety.”
And once crashes stop happening, not only are lives saved, but cars can be built that are far lighter and more efficient, he points out.
“Super Cruise” allows the vehicle to do its own steering, lane centring and speed control, from highway speed down to a stop. It could come to the market “by mid-decade” said GM in a statement.
The sensor technology and individual systems are already available on the new Cadillacs as well as competitor vehicles, using 360-degree crash sensing via radar, ultrasonics, visual-spectrum cameras and GPS data.
Despite the fact that many of these individual elements are on the market now, the effect of combining them is striking, Boules says, “It’s a very discernible step-change when you go semi-autonomous, hands-free and foot-free at the same time. The management, when they try it, always come away wowed. Letting go of the steering wheel really is a big deal.
“We have collision-imminent braking now. We should also have collision-imminent steering, where the car steers itself around an obstacle by using an adjacent lane [that it has sensed is clear].”
Boules says semi-autonomous driving will initially be applicable on highways, where the traffic is one-way and entirely vehicular. And the systems will detect driver alertness, shutting down the car if the driver can’t take over immediately.
The cost of such packages to the customer would be in the sub-$5000 range, Boules thinks. There are no legislative changes needed for semi-autonomous driving.
It is GM’s corporate goal to get cars capable of fully autonomous driving by 2020, although it would not be deployed until afterward because it needs legislative changes.
Fully autonomous driving requires that the vehicle has sufficient information about the surrounding traffic and environment – roads, junctions, topography, speed limits, etc – to make all the decisions it needs. Boules says one way to acquire all this information is to use high-resolution sensors with advanced processing.
On the other hand, he says, “When vehicle to vehicle [V2V] communication is universal, autonomous driving becomes more affordable because you need fewer sensors.” V2V allows vehicles to inform each other of their position and trajectory, so it would not need to be sensed.
V2V via DSRC at 5.9GHz is now standardized and a number of automotive OEMs have proved inter-operability, to the extent that Boules claims there is no issue with its adoption as a global standard.
“The V2V technology is pretty much ready,” he says. “There are embedded systems, but GM is taking the lead on portable devices which use the vehicle’s input/outputs wirelessly. These will give faster market penetration.”
A penetration of at least 15 percent of the fleet is needed before the major safety gains can be attained, for instance in warning of a highway obstruction out of a driver’s sight. But Boules is aware of the challenge of reaching that penetration for a safety system. “People think safety should be for free,” he acknowledges. So early penetration would be accelerated by enabling customers to use the devices for downloading entertainment and social networking.
Fitment of V2V could also be encouraged by lower insurance premiums, as the insurance companies would pass on a portion of their lowered accident costs, Boules suggests.
Accurate 3d mapping is vital to autonomous driving, Boules says, which he suggests partly explains why Google is researching the subject so intensively. It’s a way to leverage its mapping. Also if a driver isn’t driving, he or she can use Google’s other services on his journey.
Despite cars broadcasting their positions and speeds, privacy issues don’t bother Boules. He points out that the DSRC data is anonymised. Besides. “We consider privacy as one of the most import areas we have to protect. People trust GM with their data. We have has OnStar since 1996 and we have kept the data safe.” He says no speeding ticket has ever used OnStar data.
Some manufacturers have expressed reservations about autonomous driving because it would take away their marketing advantage of producers of cars drivers actually want to drive. Says Boules, “We recognise some drivers say ‘You’ll take away the steering wheel over my dead body.’ I personally believe there will be a time when people will feel insecure actually driving on the road and will go to special ‘driving ranges’ to practice driving themselves for fun. But if you have your family with you, you’ll say ‘I’ll let the car take care of this.’”
So where’s GM’s competitive advantage in all this? “All these areas are being worked on by our competitors,” Boules says. “They’re industry challenges. But we at GM have a depth and a legacy, and since the restructuring we have the means to invest in technology and products. We’re looking for a leading position in all these fronts. Our Cadillacs need to be the standard of the world again.”