October 19, 2014
When Will Self Driving Cars Become Mandatory?
Self driving cars will eventually be safer on average than human-driven cars. Vivek Wadhwa predicts 15 years until we start debating a ban on human driving of cars. I expect many incremental steps toward autonomous vehicles before that happens.
The regulatory pattern with safety features on cars is that they almost all become mandated as soon as they become affordable. Among the current mandatory car safety features in the United States: Seat belts, air bags, and (started in the 2012 model year) Electronic Stability Control (ESC). The ESC is most interesting because it requires a computer to make decisions to prevent an accident. The ESC computer code makes decisions that effectively override the decision of the driver for how hard to apply the brake. Though ESC actually enhances braking by preventing skidding. So it enhances the ability to carry out the driver's intent.
Before we reach widespread availability of fully autonomous vehicles many more incremental expansions of computer decision making will become first optional and eventually mandatory. For example, lane departure warning systems, adaptive cruise control, forward collision detection warnings, automated braking to prevent collisions, and blind spot warnings could easily all become mandatory before fully autonomous vehicles become the norm.
Early autonomous driving systems will be too expensive for the mass market. But the accident reduction and life saving potential of autonomous vehicles is so large that mandatory installation will come at a higher price point than has been the case with other safety features.
The move to autonomous vehicles will come in at least 5 regulatory steps:
- Allow autonomous vehicles on the road. There may be substeps here for allowing operation on highways versus on surface roads and in cities.
- Require new cars and trucks to come with autonomous driving features.
- Require drivers to activate the autonomous driving features in some road conditions (and the list of circumstances where autonomous operation is required will grow with time and depend on jurisdiction).
- Require higher risk drivers to only operate cars with autonomous features.
- Require non-autonomous vehicles to be upgraded or removed from the road.
When looking at steps 1 thru 5 above it is important to keep in mind that autonomy is not a simple binary attribute of a car. Vehicles will be able to drive themselves on highways before they can do the same on city streets. They'll be able to drive themselves on sunny days before they can in rain storms or snow storms. So step 3 above starts out at a short list of road conditions and gradually expands with later models and software upgrades.
Similarly, the safety gains from autonomous vehicle operation depend on the driver and on the road conditions. Higher risk drivers ought to be shifted to autonomous driving sooner. Also, very old drivers and other drivers with lower performance ought to let computers take over sooner.
Laws and the regulatory environment will slow things down. But some more technophilic legal jurisdictions will first allow computers to take over on highways. Then we will be able to watch what happens to accident rates and death tolls in those jurisdictions. Dropping accident rates for autonomous models and lower death tolls in early adopter jurisdictions will very likely cause big political pressure to allow autonomous vehicles in the remaining jurisdictions.
A lengthening list of jurisdictions reporting death toll drops from non-mandatory use will lead to support for mandate autonomous vehicles for new cars. Mandatory activation of the autonomous features will come soon after. Some jurisdictions will first impose regulatory requirements on higher risk young drivers or very old drivers. In the early stages they'll be required to drive semi-autonomous vehicles with aggressive accident avoidance technology that takes over when danger is detected.
My own prediction: Before 2020 look for early deployment of partial autonomy in luxury vehicles such as the Mercedes S Class, Tesla, BMW, and Cadillac. This is already happening with lane detection, collision avoidance, and self-parking. The luxury cars will take over more of the work on the open highway by 2019 or 2020 model year. Computer operation on city streets will come by the early 2020s (at least in some jurisdictions). In the mid to late 2020s the technology will go mainstream and show up in lower priced vehicles. Various forms of mandatory usage will show up first for dangerous drivers by the mid 2020s.
Dangerous drivers will be told they've got to pay a higher price or not drive at all. The dangerous drivers who can't afford to own autonomous vehicles will have to rent or pay autonomous taxis to move them around. Personal ownership of cars will start to decline.
As costs drop regulatory mandates will encompass a larger fraction of the driving public. By the 2030s we'll see moves in some jurisdictions to force retirement of older cars without autonomous driving tech. Lower insurance premiums for autonomous vehicle operation will accelerate this trend.
Update: But where will autonomy really begin? Long haul trucks which operate without humans on freeways seem to offer the highest ROI in the earlier (and much more expensive) stages of autonomous vehicle employment. The trucks could be driven too and from the freeway by humans. Or the trucks (which will carry expensive sensors and computers) could drop off their trailers in yards along freeways, pick up other trailers, and get back out on the freeways. Then local trucks operated by humans could tow the trailers into towns and cities.
Autonomous trucks will lower shipping costs and also cut shipping time. This will cost railroads some market share.
Randall Parker, 2014 October 19 02:09 PM
The Tesla can respond to a "come-hither" signal: you call, and it leaves it's parking spot and comes to get you! It's not legal yet on public roads, but wow.
"The launch of Dual Motor Model S coincides with the introduction of a standard hardware package that will enable autopilot functionality. Every single Model S now rolling out of the factory includes a forward radar, 12 long range ultrasonic sensors positioned to sense 16 feet around the car in every direction at all speeds, a forward looking camera, and a high precision, digitally controlled electric assist braking system.
Building on this hardware with future software releases, we will deliver a range of active safety features, using digital control of motors, brakes, and steering to avoid collisions from the front, sides, or from leaving the road.
Model S will be able to steer to stay within a lane, change lanes with the simple tap of a turn signal, and manage speed by reading road signs and using active, traffic aware cruise control. It will take several months for all Autopilot features to be completed and uploaded to the cars."
It sounds like fully autonomous highway driving will be here in "several months".
For many people who live in cities, autonomous taxis might be more economical than owning and maintaining a car. The salary of a taxi driver is a significant portion of the cost of taxis. Also, with robotic repair shops and electric cars, the maintenance of vehicles would also be cheaper, making autonomous taxis even more economical.
In fact, even for people who commute a lot, renting autonomous taxis might be economical, since the car company would be more efficient at maintaining cars and constantly using them on time.
Heck, labor is 90%+ of the cost of taxicabs. The industry would be transformed by autonomous vehicles, and personal ownership would have a hard time competing. Personal ownership would become an expensive luxury item.
I am aware of all of Tesla's announced features. But ultrasonic sensors reaching out 16 feet around the car in all directions will not provide enough sensor data. A car coming up fast in another lane will enter that zone too quickly.
Plus, I think they've got tons more software development work to do before they have a mature offering. By "tons" I mean "years".
Taxicab costs: yes, human cab drivers will become unemployed. I think we'll see the rise of very fine grained taxi/bus services where a large number of cars going down the road are really taxis for multiple riders. You'll click on a phone app and the next one going by in your direction will stop and pick you up.The average number of passengers per car will up in more densely populated areas.
In more rural areas car ownership will continue to be compelling. Of course, anyone sees the price of cars as chump change will continue to own one.
hmm. What have you heard specifically about the software?
I have the impression that "tier one" suppliers like TRW are supplying some of these systems. That leverages their R&D over multiple car makers. Heard anything about that?
Something like 95% of all car crashes are due to driver error. 5% are due to mechanical failure. Once autonomous vehicles are demonstrated to be safer than human drivers, expect a political push to remove human drivers from the roads. This kind of push is entirely predictable given the push over the past 30 years to remove drunk drivers from the roads.
"will come in at least 5 regulatory steps"
I don't see #5. What is the 5th step?
If self-driving cars were mandatory, you could expect a problem in places like Detroit and Jerusalem with "vibrant youth" stepping in front of or surrounding cars. For purposes of robbery, terrorism, or even just for kicks. Even today, antics like this are a bit of a problem. At a minimum, I would want a manual override option. Do you see a solution?
Unmentioned & a major deterrent to the universal implementation of self driving cars will be the resistance from local governments & insurance companies to the total elimination of a major income stream...traffic fines.
Traffic fines represent as much as $15BILLION to these groups & many local gov'ts would become insolvent without this income.
Also, considering the unrealistic speed limits which are posted to make virtually all drivers vulnerable, will drivers tolerate the increased travel times that the always "law-abiding" cars cause?
What about the (soon to maligned) car enthusiasts who enjoy the thrill of driving? In the new (dys)topian world of 50mph electric people movers will they just become marginalized like today's heretic climate deniers?
1st they came for the cars.....
I added a 5th step and an additional comment at the end about autonomous trucks.
It really makes sense to do autonomous vehicles on highways first. I'm picturing minibuses that pick up and drop off people in special sidings along highways where people could embark and get off. Way better than big Greyhounds.
I've heard that fuel economy and trash created per mile are dependent on driving style, but I love to roar to a red light and pass some coasting tree-hugger, then step on the gas and see them recede in the rear-view. You can pry my steering wheel from my cold, dead hands!
"you could expect a problem in places like Detroit and Jerusalem with "vibrant youth" stepping in front of or surrounding cars."
Yeah, a cousin of mine got carjacked in Detroit a few years ago, as a result of legally stopping at a stoplight in a bad neighborhood. But I think automated cars are actually ahead of the game there, because they come with dash cams as standard equipment. They could probably also, given the right infrastructure, time their approach to intersections so as to not need to stop, which would be a MAJOR advantage in urban areas.
They could also route around areas known to have high crime rates, but I expect the lawyers would put a stop to that quick. "Racially discriminatory", they'd call it.
I am skeptical. Car manufacturers and Google do not want the legal liability associated with taking full control. The average driver in my metro gets in one accident every 5 years. If Google was liable, and accidents were 1/10 as common, that still would mean every person gets one winning lottery ticket to sue Google in their lifetime. The math on that goes downhill fast.
Even if a law was passed to exempt Google, I still see a problem. Witness how the whole world turned away from nuclear power after an accident in Japan that killed zero people. We tolerate imperfect people much more than we tolerate imperfect technology. A few bad Google accidents seem inevitable because of the law of large numbers, but the PR disaster would be enormous.