Quora Question: How Will the Transition to Self-Driving Cars Work?

The Android Auto interface is seen in an Audi S3 at the Google I/O developers conference in San Francisco on June 26, 2014. Elijah Nouvelage/Reuters

Quora Questions are part of a partnership between Newsweek and Quora, through which we'll be posting relevant and interesting answers from Quora contributors throughout the week. Read more about the partnership here.

Answer from Jason Lancaster, editor of AccurateAutoAdvice.com.

The transition to fully autonomous cars should happen in three stages.

Stage 1 - Limited autonomy that doesn't depend on data sharing.
Stage 2 - Semi-autonomy that utilizes shared data.
Stage 3 - Full autonomy with shared data, advanced sensors and (perhaps) high-resolution mapping.

Stage 1 is where we are today. Vehicles have various safety systems built with the basic technologies behind autonomous cars, like 3-D camera systems and millimeter wave radar. Examples of "Stage 1" technologies:

  • Lane keeping/lane warning systems, which help you stay in your line by controlling steering or warn you as soon as you leave (data that's determined by a camera system).
  • Cruise control radar systems that let you set your cruise control to follow the vehicle ahead, slowing or braking automatically to avoid a collision.
  • Cross traffic/blind spot detection systems, which use radar and/or cameras and/or ultrasonic sensors to "see" around corners, etc.
  • The so-called autopilot system from Tesla (which is basically just a combination of lane keeping and cruise control radar), Distronic Plus from Mercedes-Benz, Hyundai's Auto Braking, etc. are all systems that combine sensors and programming to offer excellent safety and convenience features.

Stage 1 technologies are affordable and powerful, and I look forward to the day when all new vehicles come with these features as standard equipment (they're likely to become federally mandated safety features this decade).

Stage 2 is beginning, but we're 5-10 years away from meaningful market penetration.

Essentially, Stage 2 is V2X communications, which stands for "vehicle to x." "X" can be other vehicles as well as infrastructure. V2X will allow vehicles to share their position and course information with all the surrounding vehicles, as well as data from their onboard sensing systems.

V2X illustrated

This, in turn, will allow vehicle software to build "models" of the world around them, filling in gaps with information shared by other vehicles as well as the roadway. V2X offers tremendous opportunities for improving vehicle safety, and it may be sufficient (when combined with Stage 1 technologies) to offer nearly autonomous driving.

In terms of market penetration, something like 50% of vehicles will need to have both sensing systems and V2X communication for the driving public to see a major shift. But even at low market penetration percentages, V2X and advanced sensing will save lives.

Sidebar: V2X communications will need to be federally mandated to make any sort of inroads into the marketplace, and that's years away. Likewise, infrastructure needs to be upgraded to support V2X, and that's probably decades away.

Stage 3 is a combination of V2X, advanced sensors and (potentially) high-resolution map data.

Google, Audi, M-B, Toyota, Honda, Ford, etc., etc.—all are experimenting with Stage 3 right now. All automakers (or almost all of them) are deploying high-priced LIDAR (laser radar) pods on their test vehicles, and then combining these high-powered LIDAR systems with radar, 3-D cameras and (in most cases) high-resolution map data. The vehicles are designed to be 100% autonomous without any sort of communication (e.g., no V2X), which would make them able to drive themselves down a rural road without any driver input or any shared data. These vehicles are also able to handle all weather conditions, something that 3-D camera systems can't manage (snow and heavy rain cripple the effectiveness of 3-D camera systems).

Check out this older lidar pod on this Google Prius, which sort of looks like a single emergency light but is actually a $20k housing for a bunch of lasers.

The technology for stage three vehicles exists today. The challenge isn't technological—it's cost. High-resolution LIDAR pods are tens of thousands of dollars (some cost upward of $80k). Lower-resolution pods are "only" $8,000 a piece, but they're not really sufficiently precise enough for complete autonomy.

...Which brings us to map data. If you combine highly precise map data with good (but not great) LIDAR pods, radar and 3-D cameras, you get a highly autonomous car that's almost affordable enough for mass production. That's the technology that Google is pursuing—they're going to put "cheap" LIDAR pods on slow-moving cars, give these cars hyper-accurate maps and then let people use them like taxis.

The new Google car, with a "cheap" LIDAR pod, a lot of map data and a 30 mph top-speed. Google thinks they've got the future of the auto industry in their hands here, but I think the windows need to be tinted much darker. No one is going to want to be seen in this vehicle as-is.

Audi, M-B, Tesla, etc., are all kicking around systems like I've described, which may or may not rely upon hyper-accurate mapping data. Of course, as V2X systems become more commonplace, the need for hyper-accurate map data decreases...vehicles can build and share their own real-time maps with V2X. There would be no need to store or download map data.

Thus, it's possible that Stage 2 and Stage 3 will arrive at the same time...that we'll figure out affordable, full autonomous vehicles right about the time V2X becomes commonplace.

"But what about all those cars without self-driving systems? Are they going to ruin it for the rest of us because they're too technologically primitive?"

Probably not. Autonomous cars will be able to measure what's going on in the world around them, so if the driver of the old-fashioned car decides to do something crazy and cut you off, your autonomous car will detect it and adjust. With V2X, your vehicle could identify vehicles that are on autopilot and vehicles that aren't, and then devote extra resources to checking the activity of the human-piloted vehicle just in case.

"What about bugs/mistakes in these systems that cause accidents (or worse)?"

This is the question that makes me think autonomous cars might actually be 20 years away. Currently, if an automaker makes a safety mistake, they're severely penalized by the NHTSA and they're subjected to class-action lawsuits.

If something isn't done to protect automakers from liability, I worry that anyone driving an "old fashioned" car can expect a massive settlement should they get into even a minor fender-bender with an autonomous car. A few billion-dollar class action lawsuit settlements could put a big damper on autonomous car sales.

"Will government embrace fully autonomous cars?"

Not right away. We can expect the NHTSA to demand drivers sit at the steering wheel, hands at the ready to "take over" whenever they release their first set of standards. We can also expect NHTSA to demand that drivers are in the driver's seats of fully autonomous vehicles, especially commercial vehicles that do nothing more than haul goods on the highway (e.g., commercial trucks).

Sidebar: Commercial trucks are ripe for automation. Truck operators can afford to invest in expensive autonomous systems, and the savings from eliminating a truck driver is substantial. However, I don't see NHTSA allowing trucking companies to send their 80k lbs big rigs down the road without a human at the wheel.

Check it out—a self-driving Mercedes-Benz tractor trailer. You can bet that these vehicles hit the road sooner rather than later, as trucking companies will be happy to replace human drivers with computers (assuming the authorities and politicians will let them, of course).

Summing Up

A lot of very useful autonomous car technology is here today, and most of it is backed by radar and 3-D camera systems. Automakers can do a LOT with what they have now, and I fully expect all vehicles will have things like lane keeping assistance, automatic braking, etc. in the next five years.

But full autonomy, LIDAR systems and hyper-accurate maps are at least a decade away from production. And cars that drive us everywhere we want to go no matter what, with no help or concern from us humans? Fifteen to 20 years seems reasonable to me.

"How will the transition to self-driving cars work?" originally appeared on Quora: The best answer to any question. Ask a question, get a great answer. Learn from experts and access insider knowledge. You can follow Quora on Twitter, Facebook and Google+.