The Documents From Google's First DMV Test In Nevada

An anonymous reader writes "IEEE Spectrum contributor Mark Harris obtained a copy of the DMV test Google's autonomous car passed in Nevada in 2012 and associated documents. What has not been revealed until now, is that Google chose the test route; that it set limits on the road and weather conditions that the vehicle could encounter; and that its engineers had to take control of the car twice during the drive.

Comments

Who would have thought

[penguinoid]

I'm sure the editors are shocked and amazed that "Google chose the test route; that it set limits on the road and weather conditions that the vehicle could encounter; and that its engineers had to take control of the car twice during the drive."

Re:Who would have thought

[i kan reed]

Yeah, and it went into manual controlled mode when it recognized obstacles it couldn't handle:

A. A railroad crossing without signals
B. A roundabout
C. Construction work
D. "Some specific turns"

Obviously not ready for the real world yet.

Re:Who would have thought

[Anonymous Coward]

Ofcourse it is not 100% ready for the real world. It does not mean it should not be deployed though.

Not every technology and device that made it to market first worked with 100% accuracy the first time.

It just means that drivers should still be paying attention to the road at all times instead of being distracted behind the wheel. The faster it comes to market, the more quickly the improvements can be made to advance the technology.

- SK

Re:Who would have thought

[OzPeter]

Yeah, and it went into manual controlled mode when it recognized obstacles it couldn't handle:

B. A roundabout

Yeah .. but I found the reasoning of the roundabout issue to be particularly lame:

It also noted: “[Roundabouts are] particularly challenging, where many drivers don’t know the proper rules in the first place.” In an e-mail to colleagues at the DMV, Breslow wrote, “We can’t fail an applicant for not being able to navigate a traffic circle if they say that there [sic] vehicle can’t yet do it.”

Really? Whats so hard about a roundabout that Google's famed engineers couldn't program the requirements into the car? Isn't the point of these autonomous cars that they drive better than people and can deal with people driving around with them?

Re:Who would have thought

[i kan reed]

Why does version 2.0 of your application still not have all its features?

Because code takes time. And you can't just manpower your way through it.

Re:Who would have thought

[bobbied]

Ofcourse it is not 100% ready for the real world. It does not mean it should not be deployed though.

We need the power they said, it will be fine they said, don't worry they said.

The citizens of Chernobyl

Re:Who would have thought

[WrongMonkey]

The roundabout is the only one that's a show stopper. The other three are exactly the kind of circumstances where it would be reasonable for the car to come to a stop and switch to manual control.

Re:Who would have thought

[Mr D from 63]

Part of driving is dealing with a certain percentage of drivers who either don't know or simply fail to follow the 'rules'. But, I can see why roundabouts, particularly multi-lane ones, would be difficult to program the algorithms for. Part of effective roundabout driving is a bit more anticipatory than many other driving tasks.

Re:Who would have thought

[MyFirstNameIsPaul]

I agree. This is far more dangerous than cruise control and hand-free talking on the cell phone, dealing with kids in the back, arguing with a passenger, fumbling with other car controls, etc., and should not be allowed in vehicles, because the drivers wouldn't be able to handle it.

Re:Who would have thought

[OzPeter]

Why does version 2.0 of your application still not have all its features?

Because code takes time. And you can't just manpower your way through it.

A roundabout is simply nothing more than a right hand turn at a yield sign, followed by an exit ramp. If your application can't handle that, then why the hell is it on the road in the first place? And what connection does this have to what *other* drivers do?

Re:Who would have thought

[tibit]

The only slight problem with that is that in order to react at all in time, you must be paying the same amount of attention as you would if there was no autonomous drive system at all. This is otherwise known as the human being in the loop. Removing the human from the loop in aircraft automation has been a source of unending problems, and only recently one could say that it's a reasonably well understood problem - if not quite solved just yet. Don't forget we're talking about trained professional pilots here.

So, when faced with a self-driving car, the relatively untrained non-professional driver will always be so far out of the loop, that there's no way for him to overtake control safely in real time.

Of course, the solution for that is simple: the car's control transfer must, by default, happen in a fail-safe state - with the car stopped, with emergency blinkers on, etc. Only if the control transfer is explicitly acknowledged in a preset time, would the fail-safe be bypassed.

Re:Who would have thought

[tibit]

The hard part is that the car must absolutely be able to read the horizontal markings on the pavement. This is not a trivial problem at all, since those markings are often of poor legibility in ideal circumstances even to a human, never mind a machine. I'm talking about the U.S., Western Europe is probably much better in that respect.

Re:Who would have thought

[OzPeter]

Part of driving is dealing with a certain percentage of drivers who either don't know or simply fail to follow the 'rules'.

That comment applies to all driving and not just roundabouts. The car successfully dealt with a cyclist who wandered into the cars path, so it should be able to deal with other cars that do wrong things - if it can't then get it the hell off the road.

But, I can see why roundabouts, particularly multi-lane ones, would be difficult to program the algorithms for. Part of effective roundabout driving is a bit more anticipatory than many other driving tasks.

All I see at a roundabout is turn right on yield and merge right to exit ramp. There is nothing particularly hard about that, even with multiple lanes. I'm sure that fundamental to the google car is an algorithm that says (very simplified) "Analyze the objects around you, predict where they are going and don't generate a path of your own that intersects with those predicted paths". That works just as well on any right turn as well as a roundabout.

Even navigating the famed magic roundabout is just an extension of these principles (albeit in mirror image to the US)

Re:Who would have thought

[shadowrat]

I don't expect fully autonomous cars on day one. I'd expect to have to pay attention and take control from time to time, and I can live with all of that. Having to upload my planned routes to google for approval a month ahead of time doesn't really live up to my expectations though.

Re:Who would have thought

[Archangel Michael]

"Whats so hard about a roundabout that Google's famed engineers couldn't program the requirements into the car?"

Nothing. It isn't the roundabout that is likely the issue, it is the other drivers doing stupid shit because they don't know how to handle the roundabout. There is no instructions on roundabout protocols, people are just expected to "know" them. AND they don't and guess wrong all the time. In our city, they are mostly wrong all the time. You can't program a car to do what is right, when everyone else is doing it 100% wrong.

Re:Who would have thought

[shadowrat]

I think the part about other drivers was the dmv claiming that failure to navigate a roundabout is not grounds for failing a human applicant. This report does more to illuminate how low the bar is set for any driver, machine or human.

Re:Who would have thought

[i kan reed]

Oh, if it's so simple, why didn't you just write the code?

(Because the systems involved in the car's core design are complex? oooooooooooh)

Re:Who would have thought

[timeOday]

Removing the human from the loop in aircraft automation has been a source of unending problems

Commercial aviation is now safer than it ever was in the past.

Fully autonomous driving is doable IF it is only along routes that have been verified and to some extent instrumented. I predict we'll see a few Approved Routes initially, such as stretches of Interstate. Fairly soon, the approved routes will account for the majority of vehicle miles driven. Then there will be a long tail of routes and conditions that won't be automated anytime soon. Basically, just like cellphone coverage.

Re:Who would have thought

[i kan reed]

I dunno.

Train tracks seems like a normalish navigation hazard. "Look and listen for train" should be as doable as "look for pedestrians".

Just less common.

Re:Who would have thought

[OzPeter]

Oh, if it's so simple, why didn't you just write the code?

Why should I? Google already has written the right turn and merge right code.

Re:Who would have thought

[bws111]

Huh? Here are the actual rules for roundabouts in NYS:

If you are going one exit (ie a right turn), you start in the right lane of the approach road, go in the outer circle of the roundabout, and exit in the right lane of the exit road.

If you are going two exits (ie straight thru), you start in either lane of the approach road. If you start in the right lane you stay in the outer ring and exit right. If you start in the left lane you cross the outer ring, travel in the inner ring, and cross the outer ring at your exit, exitting in the left lane.

If you are going more than two exists (a left turn), you start in the left lane of the approach road, cross the outer ring and travel in the inner ring, cross the inner ring at your exit, and exit in the left lane.

Not quite what you said. When they first started using roundabouts several years ago many people had no idea what to do. Some idiots still think the outer ring is some kind of racetrack or something.

Re:Who would have thought

[Moof123]

If you have to be at full driver attentiveness at all times, what the heck is the point? A self-driving car should let you take a nap, read a book, or otherwise tune out from the road. Having to be a secondary driver ready to jump in could be even more stressful than just driving by yourself. And what if you have gone for a few months without touching the controls, can you be reasonable expected to be a competent driver if you only drive for 10 minutes a handful of times a year, and under conditions that are atypical?

I am all for giving Google and other access to the real road for continued development under very strict supervision, but there is a chasm between that and what the headlines say about the capabilities of our new robot overlords. Google is pushing for cars without steering wheels long before the technology is close to ready. It is that apparent dissonance that really bugs me. Little things like rain or a snow flurry blinding the thing, or being unable to recognize a traffic cop are pretty major obstacles to overcome, but their marketing department does not appear to be too keen on waiting for those to be solved.

Re:Who would have thought

[Anonymous Coward]

Really? Whats so hard about a roundabout that Google's famed engineers couldn't program the requirements into the car? Isn't the point of these autonomous cars that they drive better than people and can deal with people driving around with them?

Many locales have different roundabout rules. Calabasas, for instance, has specific rules about when to use which turn signal - how long your left turn signal is supposed to be on, when to switch it to your right one. Nevada requires you to use different lanes of the roundabout depending on the angle between your origin and destination exit. Another common rule is not to enter a multi-lane roundabout beside an already-existing car - a rule that doesn't mesh well with traffic law in many states where it's okay to turn right onto a multi-lane road next to another car, provided proper caution is exercised. It is easy to say 'well that seems obvious' when you're reading it on the internet, but when you just happen to be exploring a new area and run into a multi-lane roundabout, you might not realize that in order to travel to the exact opposite side you need to change lanes TWICE before you get there - you may not just use the right-most lane the whole time. And, in Calabasas, you gotta do it without signals (though to be fair, they don't have any multi-lane roundabouts that I know of).

On the off chance that you REALLY ARE an expert on roundabouts and you already know to call up a completely different traffic paradigm than for straight-road driving when approaching one (ie, all your rules about how to signal, what lane to use when completely change), how many of the general public do you think are? Good practice for avoiding collisions involves "assume the other guy is an idiot." Roundabout laws aren't complex, but they ARE significantly different, and they ARE rare enough that most people don't have to deal with them in their daily lives. I may be king shit of turd boulevard when it comes to roundabouts, but my task is more complex when the twelve other people on the roundabout are not familiar with what's going on. Multi-lane roundabouts are a choreography of lane changes that gets messed up when even one guy doesn't know it.

So yes, I imagine that dealing with a higher rate of unexpected driver behavior is a bit of a hurdle for engineers designing, building, programming, and testing a device that is supposed to replace the reactions and decision making faculties of humans piloting a two-ton weapon through all manner of the quirks of the american transportation system.

Jeez, man, this is a system of physics married with social anticipation, all done with very few giants on whose shoulders one can stand. So they said "nope! We're not designing our car to handle roundabouts, you gotta take over for that part." So what? Cut 'em some slack.

Re:Who would have thought

[bws111]

How many places, other than roundabouts, is it legal (or wise) to start in the left lane, and CROSS (not merge into) the right lane to get off at your exit? But that is exactly what you are supposed to do in a roundabout.

Re:Who would have thought

[dywolf]

no one knows how to handle roundabouts.
and the general rule for construction work always seems to be "ignore the signs, and hope you dont hit anyone".

Re:Who would have thought

[i kan reed]

Oh, believe me, I'm not an anti-automated vehicle person yet.

But at some point, we're going to have to acknowledge that people will want use them as completely automated.

Self park? Driving drunk people home? Picking your kids(or more reasonably, teens) up? Dropping you off at the airport? Actually sleeping on long drives? Replacing expensive semi-truck drivers?
There's so many applications that totally automated vehicles. And like it or not, some people are going to assume they work for these.

Re:Who would have thought

[lister king of smeg]

Why does version 2.0 of your application still not have all its features?

Because code takes time. And you can't just manpower your way through it.

A roundabout is simply nothing more than a right hand turn at a yield sign, followed by an exit ramp. If your application can't handle that, then why the hell is it on the road in the first place? And what connection does this have to what *other* drivers do?

I have seen perfectly sober drivers go the wrong way on a roundabout. Self driving cars in a beta if not alpha program having to contend with asinine drivers like that would understandably need human intervention remember this is not a finished product, not even a release candidate, it is a limited alpha release it will be improved by the time it gets to market.

Re:Who would have thought

[HornWumpus]

'Look for pedestrians' is going to be the last, hardest part of automating a car. Because you have to infer their presence. Toolbox next to a car? What constitutes a tool vs road debris? Toys on the street? Same question.

They will be automatically rolling down highways for decades before they can handle a 25mph residential street.

Then the kids will start fucking with car AIs. It will be good for laughs all around when teenagers learn that 'pulling an invisible rope' on the shoulder will cause panic stops at highway speed. When they fix that, they start tossing kickballs off of overpasses. Yes, I was a rotten kid.

Re:Who would have thought

[OzPeter]

How many places, other than roundabouts, is it legal (or wise) to start in the left lane, and CROSS (not merge into) the right lane to get off at your exit? But that is exactly what you are supposed to do in a roundabout.

Do you mean like traveling in the left hand lane of a freeway and wanting to get off on an exit ramp on the right hand side without colliding with anyone? I'd say that that is a pretty basic function to master. It may seem like its different, but it is effectively the same topology, just with different scale factors - and to a computer, once you have mastered the topology, the scale should be irrelevant.

Re:Who would have thought

[rockmuelle]

And let's just remember that planes don't actually fly themselves:

http://www.askthepilot.com/que...

-Chris

Re:Who would have thought

[blackraven14250]

I see that as the main use for automatic driving in the near future anyway. Going around the corner to the grocery store? No big deal, just drive it. Taking a road trip from NYC to Florida, or a commute from the suburbs into your local major city? Set it to autopilot and go, take control once you're in the ballpark of your destination. You're still going to need a license to use one of these for a long while, so that kind of limitation isn't crippling.

Re:Who would have thought

[viperidaenz]

On my drive to work this morning my car had to be put in manual control as well.

Obviously traditional cars are not ready for the real world either.

Re:Who would have thought

[Zalbik]

A roundabout is simply nothing more than a right hand turn at a yield sign

Not if it's a two-lane roundabout.

Re:Who would have thought

[PRMan]

You have a train on the map and you have radar. How hard is it?

Re:Who would have thought

[viperidaenz]

So why do so many drivers get confused with the give way rules at roundabouts? especially when they have multiple lanes.
You just need to follow the same rules you apply to other situations.

Re:Who would have thought

[viperidaenz]

Here's the first result to a .gov page on "how to use a roundabout"
http://www.wsdot.wa.gov/safety...
Looks like instructions to me.

Re:Who would have thought

[WrongMonkey]

I don't know how hard it is. But many types of vehicles are required to stop at all train crossings, so having the first generation of autonomous cars do the same isn't a big deal.

Re:Who would have thought

[bws111]

Where do you drive that is acceptable to do that (drive in the left lane and cross the right lane to exit)? I sure hope Googles cars are smarter than that.

If you are in the left lane on a freeway, and your exit is coming up on the right, you are supposed to merge into the right lane, then onto the slowdown lane and onto the ramp. You sure as hell aren't supposed to shoot directly across the right lane at a 45 degree angle and out the exit, bypassing the slowdown lane altogether, which is what you do in a roundabout.

Re:Who would have thought

[mythosaz]

The only slight problem with that is that in order to react at all in time, you must be paying the same amount of attention as you would if there was no autonomous drive system at all.

I snipped the quote, but to the post as a whole...

Why? Based on other technologies we have, the autonomous system seems capable of stopping the car when it's confused. It doesn't just keep rolling at the unknown at full speed hoping you take over at the split-second that it leaves it's mapped world and enters the unknown.

Along the same lines, it doesn't seem difficult to take control of the system while it's actively driving. It's not hard to disengage cruise controls or stop a car using Park Assist or Lane Assist from turning into something not seen by the sensor system. Why is it hard for me to grab the wheel from the "hands" of the auto-pilot in the Google car?

...just my thoughts.

Re:Who would have thought

[HornWumpus]

No. The racetrack is all lanes. First and last apex's are on the outside, middle apex halfway through (assuming symmetrical) on the inside. Duh.

Re:Who would have thought

[mythosaz]

A roundabout is simply nothing more than a right hand turn at a yield sign, followed by an exit ramp. If your application can't handle that, then why the hell is it on the road in the first place?

Oh, yeah, sure.

http://transportblog.co.nz/wp-...

Re:Who would have thought

[OzPeter]

I have seen perfectly sober drivers go the wrong way on a roundabout.

So have I, and in splitting my time between countries that drive on the left and on the right I had to seriously question if the car that was coming at me was because it was my fault or his fault - which is not a good feeling when you just rented a fully restored '67 mustang convertible for the weekend

Self driving cars in a beta if not alpha program having to contend with asinine drivers like that would understandably need human intervention remember this is not a finished product, not even a release candidate, it is a limited alpha release it will be improved by the time it gets to market.

However, the cause of how people came to drive at an autonomous car is irrelevant, whether it be in a roundabout or just driving down the wrong side of the street - only the response is relevant and that should be the same in both of those situations. So I contend that if the car is safe to drive autonomously on a straight street (within googles limitations), it must also be safe to drive in a roundabout.

Re:Who would have thought

[HornWumpus]

He's likely from Oregon. That's SOP up there. Along with Left turns from the right lane.

Re:Who would have thought

[OzPeter]

So why do so many drivers get confused with the give way rules at roundabouts? especially when they have multiple lanes.

Because they're idiots and can't think outside of their small world view?

But those drivers don't just simply appear when in roundabouts, they are idiots all of the time, on all of the roads. How is the poor google car going to deal with people who have the right of way, yet stop and wave you through, even though you are the person facing the sign and they are facing nothing but open road?

Re:Who would have thought

[OzPeter]

A roundabout is simply nothing more than a right hand turn at a yield sign

Not if it's a two-lane roundabout.

Which just means an extra lane change left and right. Your google car can change lanes can't it?

Re:Who would have thought

[OzPeter]

A roundabout is simply nothing more than a right hand turn at a yield sign, followed by an exit ramp. If your application can't handle that, then why the hell is it on the road in the first place?

Oh, yeah, sure.

http://transportblog.co.nz/wp-...

What is the point of this? Is it that its a left turn and left merge? Or that there are multiple lanes?

Re:Who would have thought

[bws111]

And just how to you propose to 'predict' where the other cars are going? That is the whole problem. If you are 'turning left' you are SUPPOSED to be in the left lane the whole time, but many people don't do that. They are driving in the right lane, and stay in the right (outer) lane until they exit, no matter how far they go around. If you are turning right, you are SUPPOSED to start in the right lane and stay there, but many people use the inner ring as just another opportunity to pass, especially if the first exit is more than 1/4 way around.

Re:Who would have thought

[OzPeter]

And just how to you propose to 'predict' where the other cars are going?

Are you saying that spinning laser pointer thingy on the roof of the google car is just there for show? That one that tracks the objects in the vicinity of the car, that is held up high so it can get a good view of its surroundings?

Re:Who would have thought

[bws111]

Do actually know how to use a roundabout? There are NO 'lane changes' involved. You should not 'change lanes' in a roundabout, ever. There are lane CROSSINGS in a roundabout, essentially a right turn from the left lane while the right lane continues straight. Not something sane people have a lot of experience doing.

Re:Who would have thought

[Mycroft-X]

Along the same lines, it doesn't seem difficult to take control of the system while it's actively driving. It's not hard to disengage cruise controls or stop a car using Park Assist or Lane Assist from turning into something not seen by the sensor system. Why is it hard for me to grab the wheel from the "hands" of the auto-pilot in the Google car?

...just my thoughts.

Here's the best example I can think of -- let's say you are the understudy for a radio actor with narcolepsy. You both have the script, you the understudy are following along word for word as the actor is performing. Suddenly the actor falls asleep and the words stop. How many seconds pass before you pick up where he left off? You are as aware and able as you can possibly be without actually anticipating something you can't anticipate, and I believe it would still take me a few seconds to switch myself from simply paying attention to audibly reading words.

Second scenario is the same except that you are both in sound booths reading the words and the actor is the only one with a hot mic. I believe it would be faster for me to be reading aloud with the actor and trigger my mic to go live at the necessary time -- however in doing so I am saving no effort over doing all the reading myself in the first place, so the application to automated vehicles is somewhat limited.

Re:Who would have thought

[OzPeter]

Where do you drive that is acceptable to do that (drive in the left lane and cross the right lane to exit)?

As I said .. the topology (and I mean topology in its purest sense) is *exactly* the same for both roundabouts and freeway exits, the only difference is the scales of elements *within* that topology, such as how far from the exit you should consider moving from left to right hand lanes, and perhaps the "attraction" factor you should have for a particular lane. The fact that one object curves and that one is straight is irrelevant.

Re:Who would have thought

[viperidaenz]

Facial recognition.
If the other driver looks like a retard, follow a different set of rules.

Re:Who would have thought

[OzPeter]

Many locales have different roundabout rules.

Many locales have different rules in for driving in general. Such as passing on the right. A fully fledged autonomous car will have to take all of those differences into account. But this was a limited test in a limited locale for which google pre-mapped the route, and probably tuned the car so that it would behave accordingly to the situations it encountered.

Re: Who would have thought

[Redbehrend]

Planes flying themselves is easier and already done. They are even building a new system to revamp it for commercial. The only reason it's not mainstream is because the cost and airlines are cheap, if it wasn't for the regulations they wouldn't even maintain their own planes. A lot of plane's still use tube tv's lol

Re:Who would have thought

[Noah Haders]

Ofcourse it is not 100% ready for the real world. It does not mean it should not be deployed though

lol lol lol lol lol when an autonomous SUV hits you at 60 MPG it hits you with more energy than a stick of dynamite. Let's make sure it's ready before deploying it.

Re:Who would have thought

[OzPeter]

Do actually know how to use a roundabout?

Yes ..and on both left and right side driving countries.

There are NO 'lane changes' involved.

The lane crossing in a roundabout is just a special case of lane changing in which time spent within the lane is minimized as you merge into or out of the roundabout. The fundamentals and topology are still the same.

Re:Who would have thought

[bws111]

No, they are not the same at all, and your insistance that they are says to me you don't know how to use a roundabout.

The only thing you have right is that curved or straight does not matter.

On a freeway, your major direction of travel is ALWAYS parallel with the travel lanes, even when getting ready to exit. In a roundabout, it is not. You do not 'change lanes' in a roundabout, even temporarily. You cross lanes.

Re:Who would have thought

[OzPeter]

The hard part is that the car must absolutely be able to read the horizontal markings on the pavement.

And how is that different from any other form of driving? Why is it only in a roundabout that you are concerned about the car needing to read markings on the road?

Re:Who would have thought

[Noah Haders]

wow those are tortured examples. I don't even think you need examples. cruise control, park assist and lane assist all help you drive. they don't do it for you. you can't take a nap, or text with your bestie, or fap. if autonomous cars were on the road I guarantee you that 50% of guys would fap. if you're fapping then yes, it will be hard to react in time to an emergency.

Re:Who would have thought

[dnavid]

Ofcourse it is not 100% ready for the real world. It does not mean it should not be deployed though.

We need the power they said, it will be fine they said, don't worry they said.

The citizens of Chernobyl

Interesting analogy, since the Chernobyl accident was not caused by the power plant's automated systems, but by human beings that overrode the safety systems designed to prevent just such an accident. Interestingly, the Three Mile Island accident occurred for essentially the same reasons: humans prevented the automatic systems from functioning correctly to prevent an accident.

Re:Who would have thought

[Mr D from 63]

I would think it would be a lot easier to track the motion of cars going in generally the same direction and limit the decision based on a few than it would be to track several cars traveling in a circle in front of you, with cars entering from multiple directions. It may seem simple from a rule perspective, but from an actual real world processing standpoint, it is a quite complex task. It seems that the google staff has run into this challenge, and indicator that it may be more complex than you are assuming.

Re:Who would have thought

[bws111]

If you are at a normal intersection (not a roundabout), and you cross the intersection, have you 'changed lanes'? Any sane person would say no.

Even the markings in the roundabout show you are wrong. There is not just 'left lane' and 'right lane', there are 'crossing lanes'. Yes, when you are in the inner ring and need to exit you make a lane change, into the crossing lane, not the outer ring. The major difference being a portion of the crossing lane is shared with the outer ring, but not going in the same direction, something completely unique to roundabouts.

Re:Who would have thought

[sexconker]

Ofcourse it is not 100% ready for the real world. It does not mean it should not be deployed though.

Not every technology and device that made it to market first worked with 100% accuracy the first time.

It just means that drivers should still be paying attention to the road at all times instead of being distracted behind the wheel. The faster it comes to market, the more quickly the improvements can be made to advance the technology.

- SK

But it does mean that it should not have fucking passed.
I don't know how the heathens in Nevada do it, but in any sane state if another person has to take control of the vehicle during the test, you automatically fail.
If the driver is the autopilot, then the engineer taking control should cause the autopilot to automatically fail.

Re:Who would have thought

[OzPeter]

No, they are not the same at all, and your insistance that they are says to me you don't know how to use a roundabout.

The only thing you have right is that curved or straight does not matter.

On a freeway, your major direction of travel is ALWAYS parallel with the travel lanes, even when getting ready to exit. In a roundabout, it is not. You do not 'change lanes' in a roundabout, even temporarily. You cross lanes.

When you change lanes on a freeway you are driving at an angle to the direction of travel - albeit very small.

When you cross lanes in a roundabout you are driving at an angle to the direction of travel - but at a much larger angle than in the freeway case.

In both case you are driving at an angle to the direction of travel. The topology is the same .. only the size of the angle differs.

Re:Who would have thought

[CODiNE]

I also wonder if this test included turning from and into one way streets.

Recently I actually turned into the wrong way while making a left from a one way to a 2 way street.

O_O
Not cool. That's very easy to do when you're circling a block looking for parking with one ways and two ways all mixed up.

Re:Who would have thought

[sexconker]

Ofcourse it is not 100% ready for the real world. It does not mean it should not be deployed though

lol lol lol lol lol when an autonomous SUV hits you at 60 MPG it hits you with more energy than a stick of dynamite. Let's make sure it's ready before deploying it.

When an SUV hits 60 MPG I'll still not buy one of those ridiculous ego mobiles.
SUVs only exist because American males felt emasculated when they had families and had to drive around in minivans, branding them as lifeless slaves to their spawn and spouse. So fuck safety, fuck fuel economy, fuck design, hell, fuck every color except black. If we pretend these things are cool people will pay $$$$$$ for them despite the fact that they're worse than the minivans they're replacing in nearly every way.

Re:Who would have thought

[jfengel]

Interstates are the perfect place for it: relatively few surprises and extremely boring for drivers. They're all "limited access highways", so you don't have to worry about pedestrian crossings or children running into the street.

If they just left it at that, I'd consider it an enormous advance over the present state of things.

Re:Who would have thought

[OzPeter]

If you are at a normal intersection (not a roundabout), and you cross the intersection, have you 'changed lanes'? Any sane person would say no.

I was equating a roundabout to turning right, not crossing an intersection.

Regardless what you call them a roundabout have multiple lanes of traffic. I have driven on roundabouts in the UK that are a good 1/4 mile in diameter with significant distances between on/off ramps. Do you consider that those roundabouts do not have a left and right lane?

Re:Who would have thought

[Noah Haders]

don't be purposely obtuse MPG -> MPH. to the rest of your comment, there's a guy in my neighborhood who has a black hummer with all-chrome trim. license plate "MTV STAR". I thought he was the biggest douche until I realized that he's actually living in the car. I felt bad for him.

Re:Who would have thought

[sexconker]

Why does version 2.0 of your application still not have all its features?

Because code takes time. And you can't just manpower your way through it.

Version X of my program has all of its features. Version X.Y may have more features, but version X has precisely every feature version X has. Version X.Y is a different program.

If you'd like to buy an upgrade license to version X.Y, please click below.

Note: Starting 3 months after you order your licenses, we will be rebranding Product Z to better serve our customers. Product Z will be known as Product W. Your Product Z licenses will carry over to Product W as follows:

Product Z - Product W Home
Product Z Pro - Product W Standard
Product Z Enterprise - Product W Server

Product W Home does not include Feature A or Feature B from Product Z. Upgrade to Product W Standard if you use Feature A or Feature B in Product Z.
Product W Standard includes Feature C from Product Z, but use is limited to 1 Device or 1 User. Upgrade to Product W Server if you use Feature C on more than 1 device or with more than 1 User.
Product W Server includes all of the features of Product Z Enterprise, as well as 1 license for Product W Server, 3 client device licenses, and 3 user licenses. Note that product Z Enterprise had no per-user or per-device licenses. If you require additional licenses, click below for a quote from one of our partners. We won't tell you the price, what each license enables, how many you need, whether you can reuse them, whether they expire, whether our software phones home, the actual terms, how to apply your license, or even the actual name of the license you're ordering. For installations on multiple servers, please consider Product W Datacenter Edition, which contains all the features of Product W Server with none* of the licensing restrictions. *Product W Datacenter requires 1 license per physical CPU, and if you are running on VMs the software will do its best to bitch endlessly.

Re:Who would have thought

[Archangel Michael]

I know the rules for a roundabout. That is not the problem. The problem is that no driver is required to know the rules of the roundabout. It is NOT required. Additionally you have people like my wife who thinks they know the rules, but give wrong instructions ... AND won't believe me when I correct her .. who won't read the rules for herself.

Re:Who would have thought

[Xylantiel]

Yep, at least in the US a roundabout is all about guessing, based on their approach, whether the other approaching driver has any clue how a roundabout is supposed to work. Just because you have the right of way doesn't mean it won't be a total mess if you hit somebody in the driver's door because they pulled out in front of you instead of yielding.

Re:Who would have thought

[tlhIngan]

And let's just remember that planes don't actually fly themselves:

A growing problem with a LOT of airline pilots is they lack the basic skills to fly. And this is, unfortunate a product of the environment in which they fly - most countries either prohibit or massively regulate general aviation flight, resulting in the only flying most airline pilots get are from the simulators and the jetliner cockpits.

It's better in North America as many jet pilots do actually fly little light aviation aircraft (like ye olde Cessna 172s and 182s and Pipers, etc). These pilots usually exercise basic flying skills because the most automation they have is a single axis autopilot typically. And are subsequently able to land in fairly normal conditions.

A lot of airline pilots get horrendously scared if the ILS is down, or they have to circle and land, or the VASI is off or a bunch of other things that most commercial airports have. It causes undue stress and if you really challenge them, give them perfect VFR conditions and they still can't land without automation's help. Yes, your basic VFR landing - something every student pilot has to do - can give the willies the most seasoned jetliner pilot simply because the skill has rotted away.

Oh yeah, there are also a few fundamentals to every plane that are always true for that plane. Stuff like Attitude + Power = Performance. Give it a set attitude, a set power level, and you know what to expect - climb rate, airspeed, descent rate, etc. It stays the same for the plane, and is true from your tiniest light sports to your Airbus A380s. It would've saved many lives when the airspeed indicators are off (mean flying instructors love to cover up instruments during landing - but it teaches the student that if they can hold a certain attitude and power setting, the airplane's performance is predictable and yes, you can even land without stalling. And landings can be oddly far better this way when the student is concentrating on the outside rather than inside on the instruments).

Re:Who would have thought

[sl149q]

Or a roundabout that has two lanes in one direction and one in the other. E.g. two lanes east/west and one lane north/south.

Yes we have some of those in BC.

Re:Who would have thought

[Zynder]

Can't fap with cruise control? The hell you say!

Re:Who would have thought

[sl149q]

With feedback (there is this Internet thing...) and / or official documentation from the owners of the particular road you are driving, eventually your model of the driving environment will be reasonably accurate.

So mostly you are looking to see if there are any changes today (like a car stopped, or cones and a flagger out, or a mattress or other obstacle.) Although again if feedback is active and the obstacle has been "seen" by another vehicle, your car may actually know about it before it arrives.

Re:Who would have thought

[sl149q]

And this is how different from now?

Other than many human drivers will be distracted and might not notice... but if they do notice you get the same panic effect.

Which is also why (at least in BC) most over passes (especially pedestrian ones) now have fences so that idiots cannot throw things off them.

In other words this has nothing much to do with autonomous vehicles.

Re:Who would have thought

[mysidia]

Yes, your basic VFR landing - something every student pilot has to do - can give the willies the most seasoned jetliner pilot simply because the skill has rotted away.

I would suggest an annual recertification requirement for commercial pilots with a minimum of 100 hours flying with no automation.

Re:Who would have thought

[suutar]

This requires giving the pilot an opportunity to spend that time, and according to GP a lot of places outside the US make that very difficult.

I agree with your idea, but implementing it seems to hit some roadblocks.

Re:Who would have thought

[mysidia]

If you have to be at full driver attentiveness at all times, what the heck is the point? A self-driving car should let you take a nap, read a book, or otherwise tune out from the road.

I don't think you have to be at full attentiveness at all times, but you won't be able to take a nap either. On the other hand, you would be safely able to study a paper map a bit.

So the car sounds an alarm and stops due to this obstacle it doesn't know how to handle, and you then need to pick up the wheel, look around, and navigate the situation safely.... if you're sleeping or stuck in a book, you will be stopped in the middle of the road and inconveniencing other drivers.

If LaneAssist is a useful feature, then this should still be a useful luxury feature as well..... just not as useful as a fully independent self-driving car would be.

Likely tangible safety improvements will still be a good reason for this feature.

Re:Who would have thought

[mysidia]

If the driver is the autopilot, then the engineer taking control should cause the autopilot to automatically fail.

Apparently it was agreed upon before the test:

In communication with the Nevada DMV before the test, Google said its policy was to prohibit autonomous operation at railroad crossings that lack signals and for human drivers to take over. It also noted: “[Roundabouts are] particularly challenging, where many drivers don’t know the proper rules in the first place.” In an e-mail to colleagues at the DMV, Breslow wrote, “We can’t fail an applicant for not being able to navigate a traffic circle if they say that there [sic] vehicle can’t yet do it.”

Re:Who would have thought

[viperidaenz]

I don't need your lecture, I was just pointing out there are instructions, since you said:

There is no instructions on roundabout protocols, people are just expected to "know" them.

Re:Who would have thought

[amRadioHed]

This test wasn't for a drivers license, it was for a ”self-driving vehicle testing license”. Sounds like a learners permit more than anything else, and as I recall learners permits for people are pretty simple to acquire too. There isn't even a road test.

Re: Who would have thought

[tapspace]

I love minivans. I spent the best years of my life in a minivan, as a teen and young 20-something borrowing the family minivan. I loved taking a road trip with 5 friends. I loved the comfort. But minivans are not superior to crossovers (which have largely replaced SUVs in the market) in design, fuel economy, safety or color choice (!?). I've been looking to buy a people hauler, and I've looked at crossovers and SUVs. I like the Ford Flex and the Chrysler Town and Country (the classic minivan!). They are similarly specced in terms of fuel economy, price and safety. In fact, the Ecoboost engine option in the Flex gives it superior fuel economy to most minivans. Design? Not even a contest. The Flex wins. More cubic feet, more fit and finish (barely when compared to a T&C) and better powertrain options. The Flex is not an SUV, but crossovers have largely taken the place of SUVs.

I really don't see the value in minivans anymore. Crossovers are better, and wagons and hatchbacks are a solid option if hauling stuff is your goal. Even SUVs get similar fuel economy to minivans these days, sacrificing people hauling prowess per MPG and ease of drivability for superior design and bad condition drivability. You had some axe to grind, but I think you're like the person who rags in American automobile reliability: stuck in the past.

Re:Who would have thought

[Richy_T]

Yup, and if you are sane, you change lane well ahead of your exit rather than try and cut across a lane of moving traffic.

Re:Who would have thought

[Richy_T]

I don't understand this "but not going in the same direction" thing. A post below indicates that BC has some with lanes going in different directions but on the ones I have ever used, all lanes go in the same direction.

Of course, a nice innovation in the UK is spiral laned roundabouts. As long as you get in the correct lane for your exit, getting off is very easy.

Re:Who would have thought

[Richy_T]

Nevermind, I misread the BC post. I still don't understand what you're getting at though.

Re:Who would have thought

[Richy_T]

There's nothing unusual about that.

Re: Who would have thought

[tapspace]

I never knew that. Are you American? I was absolutely never taught that, and I've never seen any sign which implies that is the correct action. I hate roundabouts actually, as they instill anxiety in me.

Re:Who would have thought

[Richy_T]

And when you took control, did you start your engine, check your mirrors, signal and pull smoothly away or were you bundling along at 50mph and your car suddenly shouted "Think quick, Bub, I got nothing"

Re:Who would have thought

[gman003]

Or let's see if it could handle the commute I had this morning: There was a power outage. Seems to have affected at least two blocks, including the traffic lights. I was trying to make a left turn, onto Main Street. For those whose traffic laws may be different, in this jurisdiction, a downed traffic signal is treated as an all-way stop sign. Or at least, it's supposed to be. Traffic coming from the left refused to stop. They just blew right through it, most didn't even slow down. Traffic from the right stopped occasionally (they had passed through other downed lights to get there, so I guess they had a few seconds to think), but that didn't help me. I eventually had to turn around and find a different route, specifically one that would not hit any traffic lights. I wonder if automated cars would have done better. Would they have stopped, as they were supposed to? Would they recognize it as a dangerous situation, stop and hand control back to the driver? Would they just plow right through it like all the dumb humans? The first is obviously correct behavior, the second would be tolerable, and the third would be merely no better than humans were doing. What if my car was automated? Would it have stopped? I sure hope so. Would it have realized that humans are morons and that it would never be able to make that turn? Would it plan a backup route that avoided all traffic lights, or would it continue to be surprised every time humans failed to remember an obscure driving law? What if a policeman had been directing traffic? Do Google's cars know how to obey hand signals? People can usually figure them out even if they don't remember them, since they're fairly intuitive to us, but that has no bearing on whether it makes sense to a machine.

Re:Who would have thought

[gman003]

You're making the assumption that any situation the car cannot handle is both an immediate danger, and a situation that can be handled perfectly by a human.

When I try to think of situations where an automated car would fail, most tend to be ones where a response of "come to a full stop, don't do anything until the human orients himself and takes charge" is a perfectly valid one. Traffic lights not working? Let the human figure it out. Bridge out ahead? Let the human figure it out. Conditions so bad you can't see the road markings? Let the human try to do better, and if he wants to sit on his ass until it clears, that's probably a good idea anyways.

Sure, there are situations where an AI might not be able to avoid an accident an alert human would. Let's say a trailer detaches from a truck in front of you, but not in your lane. As it skids, it suddenly tumbles into your lane. An automated car might have ignored it until it was too late, while an alert human would have slammed on the brakes as soon as they saw it.

But how many humans would have been that alert? Even if they weren't on a phone, or sipping their coffee, or fiddling with the radio, most drivers end up in a sort of trance, doing things automatically. I've seen people crash just because they weren't paying attention - not distracted by anything, just driving without conscious thought. Automated cars won't have that problem - they don't *get* bored. Even if they can't dodge a freak accident, they'll be avoiding plenty of routine accidents. Net gain for people who don't like car wrecks.

Re:Who would have thought

[mythosaz]

Here's the best example I can think of...

A radio actor with narcolepsy? Seriously?

Re:Who would have thought

[itsenrique]

He has a point though, about prevailing behavior. It's pretty easy to a license in my state, you never have to prove that you remember anything you once read. Honestly people DO mess up roundabouts, hell they even manage to mess up 4 way stops (the person who waves everyone else to "just go" because they don't even know when its their turn). I'm not sure its an excuse for driverless cars, but I believe he is right.

Re:Who would have thought

[MMC Monster]

Because many human drivers totally lose it in a roundabout. Especially here in the U.S. If you ever drive in one that just pops up in an area without them, you'll know what I'm talking about.

Some drivers have no idea who has right of way and will come to a dead stop in the round. Others (I swear I've seen this) try to make a U-turn when they miss their exits and get confused at all the oncoming traffic, and others try to zip through, (probably) thinking that you're supposed to maintain speed when entering the circle.

I cringe whenever I see one in PA (I've only encountered a couple so far). I don't even want to think about dealing with Jersey drivers in one.

Re:Who would have thought

[viperidaenz]

I wasn't travelling at 50mph. My speedo is in kph.

Re:Who would have thought

[mjwx]

Removing the human from the loop in aircraft automation has been a source of unending problems

Commercial aviation is now safer than it ever was in the past.

Fully autonomous driving is doable IF it is only along routes that have been verified and to some extent instrumented. I predict we'll see a few Approved Routes initially, such as stretches of Interstate. Fairly soon, the approved routes will account for the majority of vehicle miles driven. Then there will be a long tail of routes and conditions that won't be automated anytime soon. Basically, just like cellphone coverage.

I think you're right, we're looking at a slow, long term roll out. Initially it will start out as road designed as or retrofitted to specifically carry automated cars, similar to an O-Bahn except instead of being a mechanical design, it will be electronic.

The anti-government and revenue conspiracy nuts will hate it as it will mean every vehicle on this road will be tracked in real time (as automated cars will be sharing the road with human controlled cars).

Re:Who would have thought

[mjwx]

I have seen perfectly sober drivers go the wrong way on a roundabout.

I have never seen this. Ever, not even with blind drunk drivers. Not even with the ditziest of SUV drivers and Australia has a hell of a lot more roundabouts than the US.

Re:Who would have thought

[mjwx]

I think the part about other drivers was the dmv claiming that failure to navigate a roundabout is not grounds for failing a human applicant. This report does more to illuminate how low the bar is set for any driver, machine or human.

I think its saying that the DMV has been paid for a specific result, regardless of the actual events.

In order to pass a driving test, an autonomous car must not only be able to complete a random course without assistance from the assessor (in Australia interference or assistance from an assessor is an instant fail) it must be able to take natural language instructions from the assessor.

In a (Western) Australian driving test, you do not know the route in advance, the assessor will literally tell you where to go as you go (I.E. "Please take a left turn at roundabout" or "Take a left turn at the lights and move into the right hand lane"). Right now natural language interpretation requires it to be trained to your voice and is still very inaccurate. Basically:
You: "Car, dont take the parkway."
Car: "Calling your mother now. Beep"

We can make the other systems good, but also need the interface to be usable and accurate (I.E. dont let Apple anywhere near it).

Re: Who would have thought

[bws111]

Yes, I am American. Here is the New York DOT information.

Re:Who would have thought

[Anonymous Coward]

Ofcourse it is not 100% ready for the real world. It does not mean it should not be deployed though

lol lol lol lol lol when an autonomous SUV hits you at 60 MPG it hits you with more energy than a stick of dynamite. Let's make sure it's ready before deploying it.

Guess what. When a car driven by a human hits you at 60 MPG (or MPH, your choice), it has the exact same energy as the autonomous car. Guess we shouldn't let humans drive until we can ensure they drive perfectly.

Re:Who would have thought

[0123456]

Right. So we'll have people driving along the highway in their self-driving cars, and then suddenly the car will see something it can't handle and... stop in the middle of the highway, where you'll then have to put down your smartphone and figure out where you are and what's going on and drive through it. All before someone slams into the back of you, or the other dozen 'self-driving' cars now stopped in the road.

That'll work.

Re:Who would have thought

[0123456]

I've seen it in the UK. The first time appeared to be a couple of Americans in a rental car going one way around the roundabout while about twenty other cars were trying to go the opposite direction.

Re:Who would have thought

[Theaetetus]

The only slight problem with that is that in order to react at all in time, you must be paying the same amount of attention as you would if there was no autonomous drive system at all. This is otherwise known as the human being in the loop. Removing the human from the loop in aircraft automation has been a source of unending problems, and only recently one could say that it's a reasonably well understood problem - if not quite solved just yet. Don't forget we're talking about trained professional pilots here.

So, when faced with a self-driving car, the relatively untrained non-professional driver will always be so far out of the loop, that there's no way for him to overtake control safely in real time.

If you read the article, in the instances where the automation didn't know what to do, it pulled over and stopped:

Construction work, however, proved trickier. When faced with a partially blocked-off road, the car switched between autonomous and manual modes and then braked to a halt, requiring Urmson, the safety driver, to take control.

The driver doesn't need to react in time - the car does that. The driver merely needs to make the next decision to start moving again and guide the car to where it needs to go.

Re:Who would have thought

[lister king of smeg]

I have seen perfectly sober drivers go the wrong way on a roundabout.

I have never seen this. Ever, not even with blind drunk drivers. Not even with the ditziest of SUV drivers and Australia has a hell of a lot more roundabouts than the US.

and that is exactly why you have never seen it because people encounter them more and thus learn to use them in US they are rare enough that people don't learn to use them properly.

Re:Who would have thought

[mjwx]

I've seen it in the UK. The first time appeared to be a couple of Americans in a rental car going one way around the roundabout while about twenty other cars were trying to go the opposite direction.

Fair enough.

We get fewer Americans in Oz and they generally dont drive very far because of the drop bears, hoop snakes and land sharks.

Re:Who would have thought

[AmiMoJo]

Firstly they have fixed all the issues they encountered. If you have a look at their YouTube videos you can see that they handle road works and unexpected stuff like that fine now. This was two years ago, after all.

Secondly the car didn't suddenly switch to manual. There was no need for the driver to suddenly grab the wheel or anything like that. It just recognised a situation it couldn't handle and politely asked the human to take over.

Re:Who would have thought

[Brianwa]

I do it all the time, certain parts of America have roundabouts that are impossible to navigate without stopping to back up multiple times in a vehicle that's slightly above average length. If you have to go left it's safer and easier (and, oddly enough, often legal) to just go left.

We also have roundabouts with stop signs at some of the entrances, "roundabouts" that are legally regular intersections where oncoming traffic does NOT yield, and roundabouts shaped such that people going one direction can cruise through at very high speeds and others can't see far enough to know if they should yield or not until it's too late.

Re:Who would have thought

[macson_g]

I don't think anyone was consulting the construction of the plant with local population.

Re:Who would have thought

[macson_g]

LOT pilots don't lack any skills: https://www.youtube.com/watch?...

Re:Who would have thought

[Noah Haders]

dude, I thought our comments were really funny, but nobody else engaged. i guess that's the danger of being so far down the list of comments. safest bet is always to make an offtopic comment to the first thread.

on the other hand, since nobody is reading anyway, we can say whatever we want. monkey butts!

Re:Who would have thought

[RivenAleem]

But for your 4 examples, the car would have plenty of advanced warning and if no response came from the driver, the car would be capable of stopping in time. The tests they were passing with the DMV was to allow them to do further testing in more conditions. They were not trying to get ubiquitous permission for autonomous driving, just an assurance that it would be safe for other road users for Google to perform more testing.

Re:Who would have thought

[buck-yar]

Not quite. Chernobyl was caused by operators not understanding the reactor's function in low power situations which happened to be during a test to see if the cooling system would work in the time after the core shut down but before the diesel generators were back up. They brought the power levels too low to where the reactor was nearly shut down, then to raise it back up they brought the control rods all the way out creating hot spots. Then when the power came on too strong, the response was to lower the control rods, which in a very hot reactor actually does the opposite- causes a huge reaction.

Re:Who would have thought

[cyn1c77]

Yeah, and it went into manual controlled mode when it recognized obstacles it couldn't handle:

A. A railroad crossing without signals
B. A roundabout
C. Construction work
D. "Some specific turns"

Obviously not ready for the real world yet.

Ummm, have you driven lately? Most people with licenses also can't handle those obstacles!

Re:Who would have thought

[cyn1c77]

And let's just remember that planes don't actually fly themselves:

http://www.askthepilot.com/que...

-Chris

Your link is misleading. Autopilot can control aircraft at all times (climb, cruise , descent, approach, and landing phases) except during the taxi and landing phases.

Thus, with modern avionics, autopilot can control can be engaged immediately after takeoff (once flaps are off and gear is up) and kept on through landing.

A separate feature, called autothrottle, can be used to automatically control the plane's engines through the entire flight.

Of course, pilots are essential for programming these systems, monitoring them during flight, and stepping in during anomalous situations, but the planes really can fly themselves.

Re:Who would have thought

[JimFive]

The thing that bugs me the most is when the driver in front of me stops at the yield before the roundabout because someone is approaching (not in) the roundabout from the left or just got in the roundabout from the right. The whole point is that you can get on the roundabout AND the person to the left can get on the roundabout at the same time. It's not a one car at a time thing.
--
JimFive

Re:Who would have thought

[tibit]

And that's what's needed. My point still stands in all its irrelevant glory :)

Re:Who would have thought

[dnavid]

Not quite. Chernobyl was caused by operators not understanding the reactor's function in low power situations which happened to be during a test to see if the cooling system would work in the time after the core shut down but before the diesel generators were back up. They brought the power levels too low to where the reactor was nearly shut down, then to raise it back up they brought the control rods all the way out creating hot spots. Then when the power came on too strong, the response was to lower the control rods, which in a very hot reactor actually does the opposite- causes a huge reaction.

That's true in broad strokes, but you're overlooking key details. First, the automatic cooling systems were shut off. Second, when the power dropped to very low levels the operators instead of allowing the reactor to bring power up normally disabled the control rod systems designed to ensure the reactor couldn't "bounce critical" when they removed control rods to bring up power. And finally, when the test did not go as planned, and after several previous failures, the operators of the reactor went off-script and began overriding other safety features such as the coolant override systems in an attempt to drive the reactor to the desired test parameters.

You're not correct about the behavior of the control rods. The problem was a design issue of the control rods which caused them to displace coolant and neutron moderators as they are inserted, which means there's a short lag between when the control rods begin moderating the reaction and when they ironically increase the reaction due to reducing the effects of coolant moderation. This always happens, not just in "hot" reactors, but in a normally functioning reactor this is not a problem because the momentary increase in reaction is not significant. But all of the operators' previous actions caused the reactor to be placed into an extremely unstable situation. In particular, had they not disabled the automatic control rod safety systems the reactor would have automatically moderated them out of the unstable situation. But because they had essentially taken control away from the control rod systems for most of the control rods, the reactor could not effectively stop the problem the operators were introducing.

System logs show the operators did not simply try to lower the control rods back into the reactor to slow it down, they had attempted an emergency scram which drops all the control rods into the reactor at once. They would have only done that in response to an emergency condition. Which means when they did what you are pointing to as the cause of the accident, they were already in the middle of an accident. What they did not know was that they had already put the reactor into a condition beyond the ability of a scram to fix.

Re:Who would have thought

[mysidia]

and then suddenly the car will see something it can't handle and... stop in the middle of the highway

The issues were in city driving, and they are situations Google knows the car can't handle, so they agreed in advance with the state that a human would pilot the vehicle through these obstacles.

Highway driving tends to be "easier" in terms of fewer obstacles --- but yeah, if there's an unmapped stoplight or road construction site, the vehicle might have difficulties ----- it turns out, that in such areas, usually a lower speed limit is posted and normal highway speeds are not advisable or legal.

Re:Who would have thought

[HornWumpus]

The AI won't have the judgement to just keep driving when the kickball (or whatever exploit they find) is bouncing around the interstate. It will be much more rewarding to the kids.

I dont know why this is a bad thing

[Anonymous Coward]

All of the recent articles about autonomous cars seem to be trying to make people think they're terrible will never work and are a disaster waiting to happen.

All of these tests and such aren't being done so they can release an autonomous car tomorrow, its an ongoing process and will take time. I don't blame google for not wanting to publish all the details about it, its a research project and the media seems to have an agenda to make autonomous cars into the boogeyman

Re:I dont know why this is a bad thing

[Russ1642]

Somehow people think that if there's one flaw then the whole project should be scrapped. They also think that it should be at an insanely advanced stage before it's ready for use. They want to skip the decades of development that happen from real world experience and go straight to some futuristic magical car-pod.

Re:I dont know why this is a bad thing

[SJHillman]

No, they want to go way past 2014 to a magical time in which no smartphone every experiences glitches, bugs, and has infinite battery life thanks to magic fairy dust.

Re:I dont know why this is a bad thing

[maliqua]

Exactly, to a scientist / researcher / engineer etc a failed test is not doomsday, as long as it gives them more information to continue solving the problem its just a lesser form of success

Re:I dont know why this is a bad thing

[nleven]

Exactly. These limitations are well known in the autonomous car community. These are not much harder to solve than what's been already solved. It makes sense to solve the majority of the cases and then start working on the edge cases.

Re:I dont know why this is a bad thing

[bws111]

They are just a counter balance to the excessive hype that comes from Google and its fanboys.

Re:I dont know why this is a bad thing

[DerekLyons]

All of the recent articles about autonomous cars seem to be trying to make people think they're terrible will never work and are a disaster waiting to happen.

I don't blame google for not wanting to publish all the details about it, its a research project and the media seems to have an agenda to make autonomous cars into the boogeyman.

No, not so much. The recent articles are more in response to the numerous [Google press release based] articles with headlines like "Autonomous car drives 10,000 miles safely!". So, yeah, I do blame Google for not publishing the full truth, choosing instead in favor of spin and hype. It's not a media agenda, or a conspiracy, it's called balance and investigative journalism. (Something everyone here routinely calls for - right until the spotlight hits their fandom. Then it's an "agenda" and a "conspiracy".)

Re:I dont know why this is a bad thing

[vux984]

All of these tests and such aren't being done so they can release an autonomous car tomorrow, its an ongoing process and will take time.

That's understood.

IEEE Spectrum contributor Mark Harris obtained a copy of the DMV test Google's autonomous car passed in Nevada in 2012

It "passed"?

Would your or I pass a drivers test if we couldn't handle weather, road construction, roundabouts, "specific turns", and had our mom in the back seat reaching over to take the controls whenever we weren't sure of ourselves? Of course not.

When it passes a drivers test that'll be a big day. But that test in 2012 was not that day. There's much to celebrate, and as you said its an ongoing research project. But it didn't pass a Nevada drivers test. So why say it did?

It did very well, impressively well, for an autonomous vehicle even... but its still well short of being issued a drivers license, which is what an actual pass of an actual test would imply.

its a research project and the media seems to have an agenda to make autonomous cars into the boogeyman

Because the hype machine says it passed a drivers test, leading to the inference that its ready to be let loose on the streets, when CLEARLY they are no where near ready in actuality.

Re:I dont know why this is a bad thing

[0123456]

No, we're just sick of the hype that 'OMG! Self-driving cars will be here tomorrow! Google cars have driven themselves a billion miles without a crash!'

But, yes, they do need to be at 'an insanely advanced stage' before they're ready for use. Automated cruise control is likely to be available in the mass market within ten years, but it's a huge step from there to a car that can drive from A to B through town without human intervention, and anyone who says 'oh, if the car doesn't know what to do, it will just hand control back to the driver' is a moron.

Too chaotic of a system

[blueshift_1]

I'm really not that suprised. The sheer amount of decisions that we make while driving (which is a pretty complex and chaotic system) is pretty staggering. Not to mention you have local/state/national laws that vary - just to make it a bit more challenging. I really don't forsee driverless cars becoming a reality until local communication between all cars can resolve a lot of problems.

Re:Too chaotic of a system

[war4peace]

Agreed, nut I'd really see an advantage of having a "hybrid" driving car. Imagine a long trip which starts in a city and then you have to drive through a long straight highway stretch which involves little to do. You could let your car drive in autonomous mode while you relax and look around, and take control back towards the end of the trip.

Re:Too chaotic of a system

[war4peace]

doh, typo. nut = but

Re:Too chaotic of a system

[maz2331]

A hybrid could sill be useful as a cruise control that handles steering.

It's called testing

[jovius]

The conditions are controlled and constant in order to test the system's response, to ready the system for variable conditions with solid behavior data.

Re:It's called testing

[SternisheFan]

The conditions are controlled and constant in order to test the system's response, to ready the system for variable conditions with solid behavior data.

Yes, and the GoogleCar is many years away from ever being an 'alpha' release. Right now some known difficulties for it to handle are rain/snow conditions and active road construction zones. A human will need to be always aware and vigilant for the entire drive, ready to take over in a moments notice, or be held liable in case of an accident. It's going to be a long time before we have a car that a drunk can just get into and say, "Take me home", then wake up at his destination. An awful lot of time still needs to be invested in full development.

In Google's Defense...

[sandytaru]

From TFA: 'In communication with the Nevada DMV before the test, Google said its policy was to prohibit autonomous operation at railroad crossings that lack signals and for human drivers to take over. It also noted: “[Roundabouts are] particularly challenging, where many drivers don’t know the proper rules in the first place.” In an e-mail to colleagues at the DMV, Breslow wrote, “We can’t fail an applicant for not being able to navigate a traffic circle if they say that there [sic] vehicle can’t yet do it.” '

So the two times that the Google engineer took over were for the two things that Google said they felt it was unsafe for the car to handle - a railroad track without a signal, and a roundabout.

Re:In Google's Defense...

[twistedsymphony]

It's also important to note that this test took place 2 years ago, a lot of development and improvements can happen in 2 years.

Re:In Google's Defense...

[umafuckit]

"Roundabouts are] particularly challenging, where many drivers don't know the proper rules in the first place."

Only in the US. Elsewhere drivers have figured this out.

Re:In Google's Defense...

[viperidaenz]

Excuse me? I see drivers failing to correctly navigate a roundabout all the time.
I don't live in the US.

Re:In Google's Defense...

[Kjella]

I also think you have to consider Google's risk/reward here. They wanted to pass the test and most of all they did not want to get involved in any accidents, even one where the car was driving technically correct but "unhuman" as that would be used in no small about of FUD. They passed, they got their license to drive, they got the PR and the news that they couldn't drive a roundabout two years ago is nothing compared to the bad PR they'd get for crashing in a roundabout.

And the railroad thing was just policy, it can't be that much harder than crossing a priority road. Or maybe they just hadn't bothered to implement that logic, since almost all crossings in central areas have signals/gates. I just checked here in Norway and all public roads now have that, there are ~3100 unsecured crossings but all on private roads.

Re:In Google's Defense...

[the phantom]

The dinky little roundabouts that you see in the US don't do much about that level of congestion. They are just as bad (if not worse) than a four-way stop at intersections that get that kind of traffic. If an intersection routinely has half-mile backups, then it might be a good time to install a stoplight.

Re:In Google's Defense...

[Richy_T]

It depends on the type of congestion. If the traffic is mostly north & south for example, it can work out a treat.

Re:In Google's Defense...

[onezeno]

They're cropping up in Missouri and they always seem to operate smoothly to me.

Re:In Google's Defense...

[the phantom]

That is probably true, though I have to wonder why either a 4-way stop or roundabout would be needed at such an intersection---put a stop sign on the low-traffic street and don't stop the high-traffic street, or put a signal there and only stop the high-traffic street when there are cars waiting to cross. I get the impression that the major goal of those small roundabouts is traffic calming---traffic is forced to slow down for the roundabout, but not forced to stop. Of course, I am not a traffic engineer and am largely talking out of my ass, so I will readily concede that I am likely wrong. ;)

Re:In Google's Defense...

[Richy_T]

Roundabouts lead to improved traffic flow over a four way stop and often can be navigated with little to no slowing down. They are not (usually) for traffic calming.

Re:In Google's Defense...

[Obfuscant]

The dinky little roundabouts that you see in the US don't do much about that level of congestion.

That's because they aren't true roundabouts and aren't treated as such legally.

We had one in our town. I treated it as a roundabout. I got a ticket for failing to yield to someone who was waiting to enter from a street to my right. That's right -- traffic in the circle was expected to yield to traffic waiting to enter the circle.

Meh

[LookIntoTheFuture]

Let me know when these things can drive themselves in the rain and snow. Until then, yawn.

Prior art

[jpellino]

"engineers had to take control of the car twice during the drive." You mean like many of our parents did reflexively when we were learning how to drive? Not to mention the dent in the passenger side floor where the extra brake pedal wasn't.

Re:Prior art

[KingOfBLASH]

At some point I moved to Europe.

This exchange really took place:

Mom: (looking at my speedometer) "Stop going so fast!"
Me: "Mom the speed limit is 130"
Mom: "That's unsafe! Go 60"
Me: "It's KPH not MPH. 100 KPH == 60 MPH"
Mom: "Well it doesn't matter. 120 just sounds too fast. Go 60"

Re:and they picked a lemon state

[WrongMonkey]

Are we to seriously believe Nevada thought it was entirely acceptable for the vehicle to come to a complete standstill in construction traffic?

Why wouldn't that be entirely acceptable? A construction zone means that traffic can come to complete standstill at any time, even on the freeway. Having autonomous cars momentarily stop and switch to manual control shouldn't be any more of a problem than other impediments due to construction.

Re:and they picked a lemon state

[Wraithlyn]

> google got to cherrypick a desolate strip of highway

They drove down the Las Vegas strip. In May.

Pretty much the opposite of "desolate strip of highway".

hype origin

[Livius]

This shows that the claims of Google have been overhyped, but while I have no reason to trust Google, I have to wonder if the fault is Google or the people who pretend to be journalists.

Re:Why didn't it fail?

[PRMan]

Maybe if they say the car won't handle it and the car correctly recognizes the situation and shifts into manual mode, that's a pass. Makes sense.

Re:Why didn't it fail?

[MobyDisk]

I guess so.

Car: "Warning: 500-foot cliff detected. Switching to manual mode in 3...2...1..."
Human: "AAAAaaaaaaaaaaaaaaaa......"

TEST PASSED!

What Revision?

[Moof123]

I am curious as to how the license works. Is it for a specific revision of software, or does Google have free reign to make large changes to the software without re-testing?

It seems to me that as more of these vehicles hit the road, even in test mode, there should be tight control over how big of a change can be Beta tested on live streets without a re-qualification.

It's not a license to deploy

[mysidia]

It's a license to test AND apparently Google allowed it to expire, the spectrum page states:

A DMV committee then met and approved the country’s first self-driving vehicle testing license. On 4 May 2012, the Prius was issued special gold-on-red autonomous license plates, number AU-001. However, Google seems to have done little with its historic license plate. Most of the company’s testing has since been focused on California, and when its Nevada license expired in May 2013, Google failed to renew it for at least eight months.

Re:and they picked a lemon state

[Richy_T]

Hah, you seriously think that's a sane and safe mode of operation. Why do you think construction zones are so dangerous in the first place? Because of stationary objects (including people) where you wouldn't normally expect them. And you're just going to arbitrarily add to that? In an active lane?

Re:Not even close

[Richy_T]

Though with a bit of luck, that could actually lead to people not driving in conditions where they would be much better off staying put. There are some people who need to travel in snow and ice but mostly not.

Re:Not even close

[vux984]

Though with a bit of luck, that could actually lead to people not driving in conditions where they would be much better off staying put

And if luck goes against us, the inclement weather lands mid-day, and everyone has to get home from work, or is halfway home from work when the car throws up its hands, pulls over, and gives up.

Me, personally, I can generally avoid driving on snow days, but a lot of people simply don't have that luxury. Walmart and McDonalds and the rest of the service and retail industry don't close at the first sign of snow and they tend to be less than generous when staff claim difficulty getting to work.

First step towards flying cars

[Undead Waffle]

Everyone criticizing them seems to be missing the important part: We need self-driving cars so we can have self-driving flying cars. I don't trust most people to handle a flying car but self-driving actually becomes easier when crossroads can be handled on different positions on the Z axis. Don't ruin it, jerks!

The biggest failing

[ThatsNotPudding]

The biggest failing autonomous cars have is assuming they are *not* surrounded by malicious, aggressive, incompetent, brain-dead animals.

Re:Who would have thunk

[TheRealHocusLocus]

A. A railroad crossing without signals
B. A roundabout
C. Construction work
D. "Some specific turns"

As it approached the railroad crossing the Google Car coasted almost to a halt, at ~1% of full power. This was due to an excess 'poisoning' of Xenon-135, a persistent condition that was generally understood but specifics directly relating to operational safety at low power had not been addressed. At this point conditions for the driving test were inherently unfavorable and dangerous. Tuptunov sensed this, but he lacked the knowledge, vocabulary and resolve to communicate this to Dyatlov -- who ordered the car to be driven manually over the crossing.

This led them into the roundabout, where a single path to the destination exists but the Google Car was not configured to confidently know how and when to exit. The automated systems drove the car in circles for several minutes in low gear at high RPM. The car was still in a state of equilibrium and would eventually have allowed the excess Xenon to absorb neutrons and decay to Xenon-136, which has a much smaller cross-section. But again Dyatlov was impatient for the test to complete and he was getting dizzy, so he ordered to withdraw all but six of the control rods and manually lurch the car into the turnoff.

At this point the car was screaming at full RPM in low gear as it approached the Construction Zone, lurching and swaying. The operators knew they were in some sort of trouble, but the Google Car jerked forward automatically until it spotted the red cones and barricades. It disengaged to manual control and Tuptunov slapped the vehicle into its lowest possible gear. Under normal conditions this engine-assisted braking procedure was the best possible course of action. But the pistons and rods were tipped with graphite which causes a temporary neutron flux when inserted, which escalated power and deformed the rods.

At this point things in general took "some specific turns" for the worse.

Re:Not even close

[Richy_T]

True. But that is the wrong attitude and if it an be forced to adjust, that may be a good thing.