California Says Autonomous Cars Don't Need Human Drivers (bloomberg.com)

Currently, California law requires that all self-driving cars used for testing purposes be done with a human behind the wheel, so that they can take control if necessary. While California has been fairly strict on how self-driving cars are to be used in the state, they appear to be relaxing several of the rules. "The state's Department of Motor Vehicles released proposed regulations Friday for autonomous vehicles, dropping an earlier requirement that a human driver had to be present while testing on public roads," reports Bloomberg. "The DMV also backed down on a previous rule that vehicles needed a steering wheel and pedals for the operator to take back control." From the report: "When we think of driverless vehicles they can either have conventional controls, which are steering wheels, pedals, things like that, or they cannot," said California DMV Chief Counsel Brian Soublet during a conference call with reporters. If companies test vehicles without conventional controls, they have to show the California DMV that they have approval from the National Highway Traffic Safety Administration, he added. NHTSA said in early 2016 that self-driving software systems, not just humans, can be considered drivers. "If California was going to keep that level of development activity in the state, what they did was necessary and timely," said Eric Noble, president of The CarLab, an automotive consulting firm. "They kind of had to do it because at some point manufacturers can't move autonomous vehicles forward without getting controls out of cars." The proposed regulations have a 45-day public comment period that ends April 24. That will be followed by a public hearing. During Friday's conference call, the California DMV said the rules should be completed by the end of the year.

Comments

Good

[kqs]

This seems like an important next step. Expecting even a trained human to take over with only a few seconds (or less) leeway is crazy and cannot work.

I expect that these regulations will evolve a bit as we see which self-driving car developers can handle this and which ones cannot. There will likely be a few accidents, hopefully none serious. But since these cars have no egos and no temper, they're likely to drive far safer than the average human.

Re:Good

[elrous0]

Yeah, what could possibly go wrong, right?

Re:Good

[religionofpeas]

Expecting even a trained human to take over with only a few seconds (or less) leeway is crazy and cannot work,

True, but there are plenty of circumstances where there is more time for the human to intervene. Suppose the road is blocked somewhere, and there's somebody directing traffic and letting people drive over the sidewalk, or on the wrong lane, or explaining how to make a detour, or tell the driver to wait for the pilot car. Plenty of situations are too difficult for an autonomous car to handle, but not imminently dangerous, assuming that the self driving car is smart enough to stop when it notices the road is blocked.

Re:Good

[Mr D from 63]

The regulations would be smart to impose some restrictions for vehicles operating with no driver. One of them could a an successful operating history with a driver. Speed limits should also be considered. We may not want driver-less cars on the road doing, say, 40 mph + when there are zero demonstrations of technology being able to do that safely.

Even Tesla, the self proclaimed leader in this technology, is struggling to get the simplest things to work reliably; http://bgr.com/2017/03/02/tesl...

Re:Good

[amiga3D]

Most? I think it's not most are terrible. Some are terrible, some are bad, some are so-so, some are pretty fair, some are good and some are excellent. Having a self driving car suddenly lose it's capability to judge where the center of the road is would be equivalent to a stone drunk human for example. As long as it happens less often then you get a wasted driver on a per mile basis I'd guess it's safer. Still, we're talking about a "testing" program here. If I'm a company that wants to develop and build self-driving cars I'd be a little cautious about bad press from an accident involving a car with no operator. One single death due to a bad autonomous mistake will over reach the nearly 10K dead in 2014 from drunk drivers.

Re:Good

[arth1]

Fortunately, while the new regulation is going to say that there doesn't have to be a human driver in the car, it doesn't actually prevent him from being actually present.

When opening for the autonomous car not having controls for the human, it doesn't much matter whether the human isn't prevented from being there.
The first time a cop tries to direct traffic around an accident, and there either isn't a driver in the car, or there aren't controls in the car for the driver to operate, this will be challenged. If it led to severe delays for important people, I expect it to not survive.

Re:Good

[ColdSam]

If you are just labels to a Bell curve then, sure, that works for everything, but it is meaningless. You've not contradicted the idea that most human drivers, by the standards we would like for operating as 2 or 3 ton vehicle, are terrible. Most will get into several significant accidents in their lifetime.

You are also completely wrong about a self driving car not finding the center line being like a stone cold drunk. That car may have little clue where the line is, but it will STILL not just run over a pedestrian or crash into another car or a fast moving tree. It also will be able to safely just stop, if it gets too confused, unlike those thousands of drunks.

Where you are correct is that one fatal accident for an autonomous car will be magnified by the press and the luddites. People are idiots. That is a reason to educate them, not to fear technology or needlessly delay its lifesaving potential.

Re:Good

[coastwalker]

A machine is likely to be safer at driving humans in very little time. I have been driving for more than 30 years and can attest that for at least the first 5 years I took too many risks and for the last five I probably have not been quick thinking enough for dynamic traffic situations. Society does not give a toss how safe you think you can drive, all society cares about is the total cost of road traffic accidents. Automatic driving is so close you can smell it.

Re:Good

[dgatwood]

Most? I think it's not most are terrible. Some are terrible, some are bad, some are so-so, some are pretty fair, some are good and some are excellent.

... at any given moment. The problem is that every driver is terrible at least some of the time. People get distracted (both externally and internally), fatigued, can look in only one direction at a time, can't see in certain directions from the driver's seat, etc., all of which can effectively mimic impaired reaction time just as easily as drunk driving, albeit for a shorter period of time. Fortunately, most of the time, you don't need fast reaction time to drive safely.

Its too early IMO

[NotInHere]

I think its too early for autonomous cars to drive around without drivers. Imagine what happens when an accident occurs. Then the technology will be demonized. That would be horrible. Only allow autonomous cars to drive around without drivers once you are certain they are not just better than the average driver, but than 95% of all human drivers.

But I guess its like with most people who have a risky driving style: they say "who cares", until something horrible happens due to that carelessness, and then they are either unable to say anything any more, or are terribly sad.

Re:Its too early IMO

[elrous0]

I have a friend who is a truck driver who worries about driverless trucks one day putting him out of work. I laughed and told him that the first time some driverless 80,000-pound semi has a software glitch and piles full speed into a busload of kids, his future employment will be secured forever.
 

Re:Its too early IMO

[Junta]

I wonder though how much that is currently done with big trucks will become smaller vehicles, but more numerous. One of the big motivators for piling tons of stuff onto one truck is because each vehicle needs an expensive driver.

Now there are other motivators, but in scenarios where the big truck is used because only because you need to amortize the large expense of a human driver, you'll probably see smaller things on the road when/if autonomous cargo transport happens.

Re:Its too early IMO

[ranton]

I have a friend who is a truck driver who worries about driverless trucks one day putting him out of work. I laughed and told him that the first time some driverless 80,000-pound semi has a software glitch and piles full speed into a busload of kids, his future employment will be secured forever.

If that was true human drivers would already be out of a job, since about 5000 people are killed every year in trucking accidents. While it's certainly likely for a new crop of Luddites to fly off the rails at every autonomous accident, I don't share your pessimism about what the outcome would be of their crackpot protests.

"Great idea!" said the CIA

[elrous0]

In today's news, a North Korean businessman who was visiting the U.S. on a tourist visa was killed in a freak accident involving a driverless car in Los Angeles.

It depends...

[Junta]

If it is a 4,000 lb. passenger vehicle with human occupants going at 70 MPH, I agree. For a vehicle that will be carrying humans *anyway*, I don't see any need to remove the controls in the near future, even if they are not going to be used much. Maybe one day if you have unaccompanied humans who cannot be trusted with that option, but I think it's too early for that.

If vehicle + payload is less than a couple hundred pounds, about the size of a scooter, and doesn't go more than 35 MPH or so, then I think it could be reasonable to not mandate human controls. Think smallish delivery vehicle. No space or crash readiness for human passengers, low weight, size, and speed significantly mitigates the risk of an 'oops'. Vehicle wouldn't need the rigid passenger compartment and could take on more energy of a crash impact. Lower inertia means there's less damage for it to do.

I personally think that long term, delivery will move from passenger cars (food delivery) and big box trucks carrying a bunch of packages (really trying to optimize for number of drivers) to many small vehicles carrying a handful of deliveries. Also, away from flying drones to road vehicles. Flying drones are neat and all, but it's far easier and less energy to roll around.

Re:It depends...

[vtcodger]

"So if a vehicle carrying packages runs over a pedestrian that's ok?"

Hey, this is America. If we allow people to take precedence over commerce, we will lose the freedoms our forefathers fought for. Right?

The irony of safety concerns.

[geekmux]

"I don't trust these newfangled vehicles! Those computers will crash and cause accidents!"

...says the human ranting on social media, texting behind the wheel while driving on the freeway...

Re:Reason I want an automated driving vehicle

[K. S. Kyosuke]

You're not being lazy, you're being practical. Parking in general is a problem in many cities, and autonomous vehicles could ameliorate city planning processes.

Re:Do they need driving tests?

[Junta]

Think about commercial fleets/motor pools/rentals. Replaced every 2-3 years, in some scenarios many cars for relatively few drivers. Generally speaking, every driver gets *one* road test for their entire lifetime. Saying that they will operate 10 vehicles in their lifetime is probably *conservative*.

Re: for whom?

[whopis]

Sure because once you have told the automated car where to go there is no possible way of stopping it along the way or changing its destination.

Also I would venture to guess that the majority of driving is commuting to work for a lot of people. And many of them would welcome this.

Premature given their disengagement rates

[wired_parrot]

It seems premature. According to Google's last disengagement report, humans had to take control of the wheel at a rate of 0.2 per 1,000 miles, or 1 per 5,000 miles. While this is significant improvement from their previous report, which showed human intervention once every 1,000 miles, it would not give me confidence that the cars are ready to be in public streets without a driver present. They should be aiming for a rate of human intervention of no more than once per the lifetime of the vehicle (1 per 200,000 miles) before allowing the cars without a human driver.

Re:Correct

[ColdSam]

Being forced to follow nominal traffic laws is a tradeoff most people would be willing to make for the safety and convenience of self driving cars. People who want better can certainly lobby for relaxed traffic laws (or driving algorithms) for self-driving cars. I'd have no issue with software that let the self-driving car speed up to 10% over the nominal limit while coasting down a hill (but then I'd also let them go 100 mph if I get to make all the rules). Of course, there are other ways to go: e.g. regenerative braking.