Google Using Self-Driving Car Data To Make Cars Smarter

cartechboy (2660665) writes "One thing Google has perfected is using massive data sets generated from users to improve user experience. Google's self-driving cars may be subject to the same cycle of improvement, as they have racked up considerable mileage on public roads, and each mile generates data that Google engineers can use to 'teach' vehicle. Meet Pricilla — a Google test driver on the self-driving car project as she does a video walk through of some of the improvements created so far. Some are fairly simplistic, for example: 'The car does move to avoid large obstacles." That said, the car can also detect a bicyclist signaling and stay clear — oddly, even when that cyclist changes his mind and zig zags a little." Google is now testing cars on the city streets of Mountain View.

Comments

Re:Google claims a lot of things

[cfc-12]

They are around the corner, or at least one of them is. Didn't you watch the end of the video?

Re:Google claims a lot of things

[CRCulver]

Geez, give them a chance. Google didn't really start on self-driving cars until 2010. That's only four years ago. Lots of new technologies have sat around in corporate research labs for much longer before they went on to be a part of our everyday lives.

Still waiting to see 3 things

[NewWorldDan]

1. Cop directing traffic
2. A more complicated construction zone with a badly marked detour
3. A snow storm

Things are coming along nicely, but I still imagine these are a decade away. Still, they should be common and affordable by the time I'm ready to plow through a farmer's market.

Re:Still waiting to see 3 things

[Missing.Matter]

A snow storm is the big one. Also, rain and dust can be a big problem as well. The thing is, when you R&D these systems in sunny California, silly things like "precipitation" seem to get forgotten. I remember seeing a presentation about the Google street view cars, and how when they deployed them to other regions, they had to institute lens cleaning procedures because they had pretty much forgotten it rains in other places in the world.

Re:Still waiting to see 3 things

[khasim]

1. Since they can read a cyclist's hand signals that probably isn't much of a problem.

2. Possibly not on the first attempt. But, ideally, those cars will be sending data back to Google which will then relay the improved instructions to the next cars to attempt it. But hang onto that thought.

3. Probably better than a human would.

Anyway, back to #2. I'd be concerned about the number of idiots around construction zones who are NOT using the autonomous cars. The ones that will change lanes without signalling. The ones that will drive AROUND you if they think you aren't moving fast enough. Given the right set of circumstances an autonomous car could be "stuck" at a construction zone for an hour or more. All the while waiting for the lane in front of them to clear while people in the other lanes move into that opening before it is "safe".

Re:Still waiting to see 3 things

[PvtVoid]

A snow storm is the big one. Also, rain and dust can be a big problem as well. The thing is, when you R&D these systems in sunny California, silly things like "precipitation" seem to get forgotten. I remember seeing a presentation about the Google street view cars, and how when they deployed them to other regions, they had to institute lens cleaning procedures because they had pretty much forgotten it rains in other places in the world.

The cars use a combination of optical, ultrasonic, and radar sensors. I doubt rain presents too much of a problem. Hell, I imagine they could drive pretty well at night with the headlights off.

Re:Still waiting to see 3 things

[CastrTroy]

To add to this, if the car is good enough to drive itself 99.9% of the time, how well will the driver be able to drive when the car fails and they have to take over. All systems I've seen require the user to be paying attention in case something goes wrong with the computer. When the computer is good enough that you haven't had to do any driving in the past 3 months, how much are you really going to be paying attention when something goes wrong?

Re:Still waiting to see 3 things

[mlts]

That is a tough problem, and it affects regular cars as well. If the vehicle is aggressive, someone hits it, and sues Google for metric bum-tons of cash and there goes the self-driving car concept. If the vehicle is too passive, it might not get past a lane that ends, or even getting out on a three way intersection like a highway frontage road.

Of course, a self-driving vehicle has faster "reactions", and if it can get itself into traffic with a fast turn, it can turn a T-bone collision into a rear-ender, shifting fault to the other driver.

Re:Still waiting to see 3 things

[Missing.Matter]

I know what they use. I use the same sensors. Rain is a problem. Dust is a problem. Snow is a HUGE problem. Of the sensors on the car, the Velodyne HDL-64E on top is by far the most important, and provides the most critical data used for localization and detecting dynamic objects. With rain in particular laser beams can get refracted or reflected by the raindrops, which incredibly fuck up your distance measurements. The typical solution is to do a lot of filtering (i.e. take the median of two measurements) but this cuts down your effective frame rate, which already isn't that high to begin with.

Snow pretty much guarantees that "manual operation mode" is going to be a primary interface in autonomous cars for a long time to come, as we wait for not only sensors to get up to speed, but also machine learning in general.

Re:Still waiting to see 3 things

[slashmydots]

Don't forget...
4. moron in the road
5. deer in the road
6. moronic deer in the road

Re:Still waiting to see 3 things

[ColdSam]

Sure, why bother trying to solve the easier problems first? Go straight to solving the most complex cases instead.

Re:Still waiting to see 3 things

[ColdSam]

That 99.9% makes it far safer than a human driver.

Re:Still waiting to see 3 things

[Missing.Matter]

The autonomous miles driven figure is a bit overblown. They've been testing in the same conditions and using the same methods since the 2007 DARPA Urban Challenge. At some point whether you've driven 100 million miles or 100 billion miles it doesn't matter when you haven't driven a single mile in the snow.

Re:Still waiting to see 3 things

[i kan reed]

I think a fallback behavior of stop safely and pull over before [confusion situation] is well within algorithmic acceptability.

Re:Still waiting to see 3 things

[American AC in Paris]

When the computer is good enough that you haven't had to do any driving in the past 3 months, how much are you really going to be paying attention when something goes wrong?

I'd suggest that once this is consumer-ready, the vast majority of "something goes wrong" scenarios where the car doesn't know what to do would fall into one of two categories:

• "I don't know what to do, therefore I will come to a complete stop (and pull over to the side of the road, assuming I can identify a safe path);" or
• "If I can't react adequately to this situation, there's very little chance that you, meatsack, would have done even half as well as I can manage right now."

These things'll never, ever be perfect. They will almost undoubtedly reach a point where they're at least an order of magnitude safer than humans, though. That'll be more than good enough for most people.

Re:Still waiting to see 3 things

[geekoid]

no worse then the driver is now? Most people most of the time do not handle an emergency situation correctly. They may get lucky and avoid it, but is' seldom the best course of action.

You might as well ask the same question about anti-lock brakes.

Re:Still waiting to see 3 things

[geekoid]

1) Around here the cyclist only have one hand signal, and its to tel you were YOU should go.

As for you other points, I look forward to my car sending a video, date, and time of the other vehicle behaving in that manner.

Re:Still waiting to see 3 things

[schlachter]

Maybe in 5 yrs hyper spectral imaging sensors will be cheap enough to be used in these cars...or perhaps inter-car communications will make up for the degraded sensing. For now, I'd think a solution for 95% of the time would work for most people.

Re:Still waiting to see 3 things

[wile_e_wonka]

The summary says that Google is testing cars in Mountain View. According to this website, with respect to Mountain View:

During the cold season, which lasts from November 26 to March 6, there is a 34% average chance that precipitation will be observed at some point during a given day. When precipitation does occur it is most often in the form of light rain (57% of days with precipitation have at worst light rain), moderate rain (31%), and heavy rain (11%).

I think it is safe to say that it rains there, as compared to, for example, Disneyland (18% of cold season days have precipitation) or Las Vegas (12% of cold season days have precipitation).

Re:Still waiting to see 3 things

[lorinc]

I am pretty much waiting for the other way around: when I, as a human, known that I am quite bad at handling the current situation and that the machine will do much better than I'll ever be capable of. In particular, I am thinking of traffic jams where you have to find maximum speed that maintains a smooth global flow, without falling into chaotic start/stop sequences like this: http://www.youtube.com/watch?v...

Most humans typically tend to drive too fast in such situations, leading to an average speed well below what machines could very simply achieve. So basically, I'll drive when it's easy, and let the machine do the complicated stuff, as usual.

Re:Still waiting to see 3 things

[im_thatoneguy]

I'm perfectly OK with the tradeoff though of having the car drive on sunny days and then have a manual operation mode when a blizzard crops up.

Or for that matter, I was driving along the interstate yesterday when a Tornado siren went off. Even I wasn't entirely certain what to do, but it's certainly a problem that could be solved with technology. My phone knew that the tornado siren was activated since it received an emergency text message simultaneously. Unlike me, the car would have instantly known where the nearest off-ramp was and could have taken me to the nearest safe structure assuming it deactivated the LIDAR and relied on stereo optical cameras for "emergency" navigation. Or I could have driven even if I hadn't driven in years at 40mph down a freeway to an offramp and parked. That's an easy enough skill to manage.

But let's look at it from a worst case scenario. Even if I couldn't drive and Google just pulled the car over along the road and said "In the event of debris, exit the vehicle immediately and find a low lying ditch to take cover. Cover your head, and do not hide in an overpass." I would have gotten wetter but would have remained about equally safe. Meanwhile all of the terrible drivers who seemed to want to murder everyone on the interstate would have been far less likely to kill me. I think all things told even in the very worst case scenario I would prefer the automatic car for "net safety". And in the case of snow, it doesn't appear people know how to drive in the snow either, so I can't see performance being *that* much worse than the status quo.

Re:Still waiting to see 3 things

[Comrade Ogilvy]

Please mod parent up! (Had mod points to burn just a couple days ago.)

How confusing the inputs of a system is to the computer is something that the program should be easily able to figure out, except in cases where the human drivers are hosed as well.

Re:Still waiting to see 3 things

[Comrade Ogilvy]

Re #2: People do die from complicated and badly marked construction zones from human error. Happens all the time. So the computer does not have to be 100% to be vastly better than a human. Having the humility to slow down, instead of the human pride that causes us to drive hard forward in the face of uncertainty, will be a huge advantage to the computer program.

Re:Still waiting to see 3 things

[DerekLyons]

They will almost undoubtedly reach a point where they're at least an order of magnitude safer than humans, though.

Considering how good humans actually are (per vehicle mile traveled), that's actually a pretty tall order.

Re:Still waiting to see 3 things

[cheater512]

Many Infrared frequencies will quite happily go through rain and fog. Not sure about snow but don't assume all wavelengths are affected the same way.

Anyway its not like a car would try driving anyway with bad data. It would just pull over to the side of the road and say it can't continue, please drive manually.

Re:Still waiting to see 3 things

[Missing.Matter]

I'm not assuming anything; I've used the actual hardware they are using on the Google car.

Re:Still waiting to see 3 things

[Kjella]

Well, it depends on how many problems people avoid before they become traffic problems. For example imagine that ahead of you there is a truck with apparently a poorly secured load, you can see it moving, ropes fraying so you hit the brakes and give yourself a wide gap in case it spills all over the road. That drunk pedestrian who looks like any moment he'll swerve into the road, maybe we should pass him with extra margin. Those kids that just sent a soccer ball across the road, is anyone running to get it without paying attention. That truck coming down a hill at an awful speed honking his horn, you have the right of way but maaaaaaaybe he's trying to tell you something. It might have 360 degree vision and lightning reflexes, but it might also need them to get out of situations a person would avoid altogether based on a deeper understanding of the world around him.

Re:Still waiting to see 3 things

[Bing Tsher E]

But since you've been using the automatic function for all the sunny days, you're out of practice driving when the manual operation mode is required. In other words, when you need your driving skills most, you're sorta dumped into the role but out of practice.

Re:Still waiting to see 3 things

[mysidia]

At some point whether you've driven 100 million miles or 100 billion miles it doesn't matter when you haven't driven a single mile in the snow.

I'd still be interested in this car, because it never snows down south around here either. If it does snow, everything shuts down, and nobody dare drive, anyways.

Re:Still waiting to see 3 things

[mysidia]

and if it can get itself into traffic with a fast turn, it can turn a T-bone collision into a rear-ender, shifting fault to the other driver.

Except that the incident will be on video/computer record, showing the rapid turn, which will prevent fault from being shifted.

Colliding in the rear does not automatically assign fault to the vehicle in the rear.

Re:Still waiting to see 3 things

[MightyYar]

To be fair, though - the bar for an automated car is pretty low in the snow. So many morons.

Re:Still waiting to see 3 things

[MightyYar]

If that really becomes a problem, we could always make sure the car makes sure you get enough "hours" each year. Then again, we have the technology to measure reaction time at start time and don't do it today. But this isn't a tech problem.

Re:Still waiting to see 3 things

[drinkypoo]

You've gotten a lot of shit for this, but I think you're right. Sadly, I don't think most drivers actually can account for most of those situations much better than a self-driving car would today, especially the oncoming truck which is about to convert their solids directly to vapor. But perhaps I'm wrong. I do notice that most drivers are quite oblivious to what's happening around them most of the time. Perhaps this is an area in which the cars will actually surpass us... but not today.

Re:Still waiting to see 3 things

[Animats]

A better LIDAR sensor is needed. Something more like the Advanced Scientific Concepts flash LIDAR. Right now, it costs too much (about $100K) but that's because they're hand-made in Santa Barbara for DoD and space applications. It has custom ICs made in volumes of tens. Volume production would bring that way down. You don't get the full circle field of view of the Velodyne, so it may take multiple sensors.

To deal with rain and snow, you need "first and last" return data. This is used in air to ground sensing to sense both the top of tree cover and the ground underneath. With that, and a good frame rate, you'll be able to distinguish rain and snow noise from solid objects. You'll lose range in heavy rain and snow, and will have to slow down. That's OK; humans can't see through it either.

Radars can. What's Google doing on the radar front? Off the shelf automotive radars are getting pretty good. Modern millimeter radars can see pedestrians. The older units from the Grand Challenge days could only see car-sized obstacles, maybe motorcycles.

Re:Still waiting to see 3 things

[Gavagai80]

Worst case, it's not hard to make a vehicle pull over and ask a human to drive when the weather gets bad. And we already rely on mechanical windshield wipers for humans to operate cars in weather.

Re:Still waiting to see 3 things

[Gavagai80]

Many people drive rarely already. They cope.

Re:Still waiting to see 3 things

[AmiMoJo]

I have used these sensors too and you vastly over state the problem. In the kind of rain or snow where a human could safely drive they work fine with a bit of filtering, which you need to do anyway to remove passing birds, insects, litter and the like. By the time you start getting unusable data from them a human would have a hard time driving safely at more than 10kph anyway.

Yes, you need to reduce speed in a snow storm (your brakes don't work that well aside from anything else), but it's hardly an insurmountable problem.

Re:Still waiting to see 3 things

[CohibaVancouver]

I'm perfectly OK with the tradeoff though of having the car drive on sunny days and then have a manual operation mode when a blizzard crops up.

How DARE you be OK with this. This is Slashdot, where the edge use-case wins, EVERY time. Now take your ID and go home.

Re:Still waiting to see 3 things

[Quirkz]

All tough scenarios. I've seen a few others myself, including a car coming downhill in the snow that went into a sideways skid and drifted through my lane. The only escape was to move into their oncoming lane since they were occupying mine and there was a ten-foot drop off the side with no appreciable shoulder.

That said, I'll still trade all these outliers for cars that avoid the really stupid stuff that people do on a daily basis. The number of times I've had crazy exceptions like the above is small. The number of times someone has tried to merge or turn into my lane without seeing me, or failed to yield, or yielded when they had right of way, is far, far higher. On the whole I think we'd still benefit, even if all the tough scenarios you outline ended in spectacular crashes.

In the future...

[Ichijo]

...our children and grandchildren would wonder why we ever allowed humans to operate motor vehicles on public roads.

Re:In the future...

[dlt074]

and they will be incapable of doing another simple daily task.

i love technology, but i shudder to think how bad a collapse will be.

Re:In the future...

[FlyHelicopters]

Yes, and this is why the preservation of civilization and knowledge are so important.

The vast majority of people are no longer able to take basic care of themselves without modern society. Or do you know how to grow crops? Because I sure don't, I do however know how to shop at Walmart for food.

Could I figure it out? Yea, sure... in the time required if food left store shelves? No, not really...

A bigger issue? We now have too many people to go back to living off the land, so we need civilization, or we'd have... an interesting day...

Re:In the future...

[HornWumpus]

So grow a vegetable garden and hunt.

It's calming, fresh vegetables and game are good and if the shit hits the fan you are steps up the learning curve for feeding yourself.

It would still suck, hopefully less.

Re:In the future...

[FlyHelicopters]

That is a nice idea, but there are problems with it and it doesn't solve the core problem.

I live in a nice suburb, our population density is much lower than many urban areas, but still WAY too high to live in the 19th century.

We generate a ton of trash, we need a working water and sewer system, and the power needs to be on or we'd all be standing around in the dark and cold (or heat).

Having a small vegetable garden is nice, but wouldn't feed my family and it doesn't scale up to the size that would. First, I don't have the land. Second, I don't have the skills and tools to grow that much food, having a small 10x10 (or even 20x20) vegetable garden in the back yard is not at all the same thing as having an acre of crops to feed the family all year and knowing how and when to harvest them to be a sole food source.

My point was that we have moved beyond that time, we aren't going back there short of disaster and war, and the population would be thinned by quite a bit in the process.

Re:In the future...

[tirerim]

Hell, I wonder that now. As far as I can tell, the vast majority of drivers only barely escape disaster every time they get behind the wheel, mostly because their frequent lapses are adjusted for by other drivers who happen to be paying attention at that time. If two lapses occur simultaneously, *crash*.

Re:In the future...

[HornWumpus]

If you never _need_ the gardening skills, you're still ahead a bunch of tomatoes you can't buy at any price and kids with a clue.

Move to a better neighborhood. Horse properties. With a well. You can afford at least a quarter acre backyard if you can afford helicopter time.

Re:In the future...

[Gavagai80]

If civilization collapses, you don't have cars to drive. So you should only be lamenting how few people know how to ride a horse.

Re:In the future...

[FlyHelicopters]

Then you aren't hearing me...

I don't want everyone else to do anything, I'm just saying that most people in western civilization have moved beyond the time when you could live off the land and be free of modern technology.

To try and go back there would be a nasty mess involving a lot of dead people.

I would prefer that we retain our modern civilization and make sure that we preserve our knowledge so that it may never be lost.

Re:In the future...

[FlyHelicopters]

The challenge is that while I could grow my own tomatoes, my time is not free, they would cost me more to grow than they would cost me from a local farmers market.

As for moving to a better neighborhood, that costs a lot of money, more than I have. Right now I live in one of the best public school districts in Texas. There is another one about 20 miles away I could move to, also one of the best, and all the homes there are on one acre lots, but they also cost a million dollars each, which I do not have. My current home is about $320k and that is the limit of what I can afford. I live here for the schools, I could easily get lots of land for that much money, but it wouldn't be in the best school system.

Re:In the future...

[HornWumpus]

Not even farmers market tomatoes are as good as right off the vine. Lots of farmers market tomatoes are the same as grocery stores.

As I said above, it's calming. It's also amazing how many times I've gone to do a little gardening (to take my mind off a technical problem) and come back in with a solution. Might matter if you're paid for results vs. hourly.

Fucking Cyclists are ruining the future.

[Anonymous Coward]

Fucking cyclists are going to have a free lunch with self driving cars.

Just plain old taking over the streets because they know the computers will give them right of way everytime.

Google needs to program the cars with some assholery in mind where scaring cyclists is as common as checking for the car's fuel.

Re:Fucking Cyclists are ruining the future.

[PvtVoid]

Fucking cyclists are going to have a free lunch with self driving cars.

Just plain old taking over the streets because they know the computers will give them right of way everytime.

Which will be completely fucking awesome.

Re:Fucking Cyclists are ruining the future.

[geekoid]

If grid lock, anger, douch baggery, and killing automated cars is you goal, then sure.
Otherwise, we will just have the self righteous assholes going 15 MPH in a 25 MPH zone becasue they are too precious to use the bike lane.

Re:Fucking Cyclists are ruining the future.

[Ichijo]

What if there's no bike lane, or it's filled with debris, or the bicyclist needs to make a left turn?

Re:Fucking Cyclists are ruining the future.

[PvtVoid]

If grid lock, anger, douch baggery, and killing automated cars is you goal, then sure. Otherwise, we will just have the self righteous assholes going 15 MPH in a 25 MPH zone becasue they are too precious to use the bike lane.

And the car will have the simple common sense to wait ten or fifteen seconds until it's safe to go around the cyclist, who will inevitably catch up at the next traffic light anyway. This behavior seems to elude humans.

Re:Fucking Cyclists are ruining the future.

[slashmydots]

If Jean Claude can do the splits between two volvos, you can certainly have the computer get your Honda 6 inches from a bike rider, lol.

Re:Fucking Cyclists are ruining the future.

[schlachter]

I hear google will program them with a formalized model of aggression which will be required in order to keep up with the flow of manned traffic. :)

Re:Fucking Cyclists are ruining the future.

[seven of five]

An extremely loud air horn, such as those used on locomotives, sounded repeatedly, might get the point across to errant cyclists.

Re:Fucking Cyclists are ruining the future.

[geekoid]

Then it's fine. I wan't talking about those situation. That said, it takes some arragonce to expect everyone to have to be hindered bacause you can't go fast enough.

Everyday I am behind a bicyclist in the middle of the road, when there is a perfectly clear bike lane.
Every fucking day .

Also:
They Run Stop signs, swerve onto and off of the sidewalk at random, stop.

And one time I was rear ended by a bike While I was stopped specifically to let them pass me on the right. What did I get for my troubles? the finger
I used to think it was just a small percentage, but I have gone to too many bike rally and meets. In Portland, well over half are self righteous assholes that want all the rules apply only to others.

Maybe where you live they are actual decent thinking human beings.

It's so bad here I stopped riding because I don't want to be associated with them.

Re:Fucking Cyclists are ruining the future.

[geekoid]

We were talking about cyclist intentional hinders the vehicle and ignoring traffic laws becasue they know the car won't hit them.. you know, abusing the system.

Re:Fucking Cyclists are ruining the future.

[Darinbob]

And while those cases do occur, the vast majority of cyclists outside the bike lanes are just asshats who think the rules of the road don't apply to them, or that hand signals are only for beginner cyclists.

Re:Fucking Cyclists are ruining the future.

[geekoid]

Sorry for the rant. As a former cyclist these people piss me off.

Re:Fucking Cyclists are ruining the future.

[Ichijo]

If you were stopped to let a bicyclist pass on the right, it suggests you were about to turn right, and that you hadn't properly merged into the bike lane as you are legally required in the USA outside of Oregon. A safe bicyclist knows that it's unsafe to pass on the right.

Re:Fucking Cyclists are ruining the future.

[mrsquid0]

You mean like motorists do?

Re:Fucking Cyclists are ruining the future.

[HornWumpus]

More or less all bike lanes are filled with debris. Don't ride a 130psi road bike on the street then complain that debris flattens your tires.

Re:Fucking Cyclists are ruining the future.

[Macman408]

Somebody else already pointed out the link, but you should NEVER put yourself in the situation where you are stopped to let a bicycle pass on the right. Even if I'm 100% absolutely positively certain that a driver has seen me and is waiting for me, I will never pass a car on the right in a situation like that unless there is literally no other option (ie I can't slow down and move behind them or to their left). You should have been paying closer attention, and either known that you could safely turn in front of the bicycle without cutting them off, or slowed down sooner and merged into their lane behind them. The latter is by far the preferable option, since most people tend to overestimate how much time they have to pass a cyclist and make their turn safely - especially if they find their right turn blocked by a pedestrian or another car.

(There are some situations where I will pass a car on the right, like when a long line of traffic is stopped at a traffic light, however once traffic starts moving, I will slow down so that I am in between two cars - so if either of them turns right, I won't be in danger of getting hit.)

This is actually something that the Google car did poorly in the video; when they demonstrate yielding to a bicycle passing on the right, the car should have been on the right in the bike lane, not to the left of the dotted white line. I'm hoping that they did this just to illustrate "hey look, we can see bikes coming from behind the car!", and that the car would normally be in the bike lane/shoulder in order to turn right.

One other thing - often, bicycles need to take the lane even when there is a bike lane. Perhaps the surface in the bike lane is too rough, or there is debris, or a poorly-designed storm sewer grate, or any of a thousand other possibilities. Without knowing the particulars, I can't guess whether one of those is what's happening, or if you just happen to commute along the same route as the world's largest asshole.

That said, yes, there are many crappy bikers out there. I'd like to think that if we started designing our transportation infrastructure more for bicycles (rather than including them as an afterthought), that would go away; more bike paths with fewer grade-level road crossings would go a long way. Also, making some roads (eg residential) nearly dedicated to bicycles can help - for example, put some planter boxes in a few places across the road, so only bicycles can pass through. Car traffic is reduced to only the residents, bicycles get a quiet place to ride, and cars don't have to jockey for position with them on the main thoroughfares. That only works if the bike route is sufficiently straight, goes somewhere interesting, and doesn't have many stop lights and stop signs - otherwise some bikes will prefer riding with more cars in order to have to stop less often.

Enforcement, or even just warnings, would also help. Doubly so if the laws were also adjusted to make sense for cycling; for example, the "Idaho stop", where a cyclist is allowed to treat the stop sign as a yield if there is no traffic approaching. But I'll admit I was amused when I was stopped at a red light, and a cyclist (obviously an inexperienced, low-income, casual biker) biked through, and a cop stopped in the middle of the intersection and yelled at him. ("That other cyclist stopped for the red light. You need to stop too. I don't want to have to fill out the paperwork when you become a road pizza.") The cyclist seemed confused at why he was getting yelled at (or was perhaps just putting on the "no hablo inglés" routine), so I don't know that it made things any better, though.

  Well, one rant begets another, I suppose... ;-)

Re:Fucking Cyclists are ruining the future.

[mysidia]

Google needs to program the cars with some assholery in mind where scaring cyclists is as common as checking for the car's fuel.

Why not just have a system that captures abusive cyclist behavior on video, uses facial/body recognition to identify them, then uploads the video stream to Youtube and submits to law enforcement.

Make abusive cyclist behavior such as entering the roadway/cutting off a vehicle, an offense where the cyclist gets a ticket in the mail and loses their bicycle privileges.

Re:Fucking Cyclists are ruining the future.

[mjwx]

What if there's no bike lane, or it's filled with debris, or the bicyclist needs to make a left turn?

If a cyclist needs to make a turn across traffic (right hand turn here in Australia) they should be doing a hook turn.

http://www.tmr.qld.gov.au/safety/queensland-road-rules/~/media/Travelandtransport/Cycling/Bike%20user%20guide/Road%20rules%20for%20cyclists/Rules_image_six.ashx

A cyclist should never try to merge into traffic for their own safety.

And if you cant ride in the bike lane for any reason or in traffic without obstructing other road users, its time to get off and walk. Unreasonable obstruction of traffic is illegal in my country and should be in any civilised nation.

Re:Fucking Cyclists are ruining the future.

[Ichijo]

That hook turn looks very unsafe.

Is a bicyclist obstructing anyone if cars can change lanes to pass?

In the USA, if five or more vehicles are formed in line behind a slow moving vehicle, that slow moving vehicle is required to pull over when it is safe and allow traffic to pass.

Re:Fucking Cyclists are ruining the future.

[dasunt]

Oh as a cyclist, I'm looking forward to self-driving cars.

Seriously. It's going to be nice. Not because I'm an asshole, but because these things are going to follow the law, which means stuff like 3' passing distance!

Re:Fucking Cyclists are ruining the future.

[mrsquid0]

Wrong. Do you know who were the first to push for paved roads? Hint. It was not motorists.

Re:Fucking Cyclists are ruining the future.

[peter.kingsbury]

The Romans?

Re:Fucking Cyclists are ruining the future.

[Uzuri]

See, this is something I didn't know. I just moved to a place with a lot of bicycles two years ago. I've had to watch myself and relearn how to drive. I would have NEVER guessed that you're supposed to merge into the bike lane. We didn't have them back home; you only cycled on the road if you had a death wish.

Yes dear

[PPH]

It appears that back seat driving is a legitimate profession.

Road to Hana

[portwojc]

Lets see it drive the road to Hana and back. That would be interesting. Bonus points for not getting yelled at.

Re:Road to Hana

[akgooseman]

The only challenge on that road, from either the north or the south, is the rented Mustang convertibles plugging it.

Re:Road to Hana

[Zynder]

Easier! You can do it in way less than 12 parsecs. But if that doesn't convince you enough, most bikes are about 2 meters across. Well, shit, I used to bullseye womprats in my T16 back home. They're not much bigger than 2 meters!

Re:Road to Hana

[Animats]

Solved problem. See 2005 DARPA Grand Challenge. Dirt roads with no guardrails, no problem.

Re:Road to Hana

[SmilingBoy]

Apologies, that was me once in the rented Mustang convertible. And to the grandparent: I did get yelled at - by my wife who was on the passenger seat!

Re:Error in the summary.

[harvestsun]

Yes. That is what companies do, they make money. Astute observation.

Not sure we need it

[nine-times]

These kinds of problems do need to be addressed, but I'm not sure they need to be *fixed* before you turn this into a product. For the snow-storm example, I don't think the car needs to be able to drive in the snow. It's much more important that the car is capable of detecting "this is a situation in which I can't operate safely," and refusing to try. It should be good enough if the car's AI can say, in effect, "Listen, human, I can't take responsibility for driving in this snow storm. If you're comfortable driving in it, go ahead and take manual control. Otherwise, we're staying right here." On the other hand, I could see an interesting application in providing some kind of intelligent 'driver assist' for bad weather conditions that helped the driver maintain traction.

Regarding details, I think the ideal would be for most road conditions, detours, and traffic issues to be kept up-to-date on a database that could allow for dynamic routing instead of the car relying completely on markers. It's not a complete solution, but again, it may be enough to pair a large database with some ability for the car to say, "I don't know what to do here, so I'm going to either give back manual control or pull over and wait."

Re:Not sure we need it

[Kjella]

It should be good enough if the car's AI can say, in effect, "Listen, human, I can't take responsibility for driving in this snow storm. If you're comfortable driving in it, go ahead and take manual control. Otherwise, we're staying right here."

And if you're in the middle of some remote mountain pass, you haven't driven in ages since it handles regular conditions just fine and and the car just says "nuh uh, I'm not moving another inch" then what? It's not like staying to freeze is a real option, it's basically forcing you to take over at the worst possible time. It might be enough, but my guess not until a huge court case decides it's enough.

Re:Not sure we need it

[timeOday]

Meh. It's an edge case that won't even arise for years and years, until driving is mostly automated, such that people start to get out of practice. I would guess that by then roads will be instrumented with special reflectors on the signs, or markers embedded in the surface, so all these fancy sensor systems aren't even needed any more.

Re:Not sure we need it

[geekoid]

maybe the driver needs to be responsible for taking a path he isn't qualified to drive on in the first place?

Anyways, the car will be tied into the weather and it could let you know whats coming up, so you can turn around and go back.
Also, when crossing mountain paths, take extra water and blankets.
I speak form experience. I'm sure my former, and tasty, passenger agrees with me :)

Re:Not sure we need it

[nine-times]

I'm not sure there's a great solution. One way or the other, you could possibly run into a fringe case where the vehicle won't operate safely, so there should be a protocol of "what happens when it can't operate safely?"

It might be that it turns over manual control. Perhaps if you really can't drive, there could be an override that says, "I understand that it's not safe, but I'm overriding the safeguards. I take legal responsibility in case something goes wrong." It would probably be a better answer to have the ability to set of some kind of robust "distress signal" in case of a serious problem, alerting the local search-and-rescue team to come find you.

There are actually a few different options, but my point here is that it's not all about engineering problems, i.e. making the AI "smart enough" to handle every possible eventuality. Some of these things can probably be solved through laws and public policy, and having a sensible protocol for what to do when things aren't working.

Re:Not sure we need it

[HornWumpus]

Look at the accident rate in areas where in snows part of the year.

It appears that people can't remember how to drive in the slick for 6 months. They have to relearn every year.

Re:Not sure we need it

[swillden]

It should be good enough if the car's AI can say, in effect, "Listen, human, I can't take responsibility for driving in this snow storm. If you're comfortable driving in it, go ahead and take manual control. Otherwise, we're staying right here."

And if you're in the middle of some remote mountain pass, you haven't driven in ages since it handles regular conditions just fine and and the car just says "nuh uh, I'm not moving another inch" then what? It's not like staying to freeze is a real option, it's basically forcing you to take over at the worst possible time.

That's easy to solve: The car should force you to make that decision earlier. It knows where it's been told to go. It has access to weather forecasts.

"I'm sorry Dave, but there's a 30% chance of snowfall along that route, and I can't handle driving in a snowstorm. Do you want to chance having to take manual control or should I book you a nearby hotel so you can wait for the weather to clear?"

Re:Not sure we need it

[swillden]

You can also throw in having the car use its Internet access to predict ahead of time when there's a potential for trouble, and ask you to make the decision then. Weather reporting these days is highly accurate when it's only looking a few hours ahead.

Re:Not sure we need it

[bboyjkang]

"I think the ideal would be for most road conditions, detours, and traffic issues to be kept up-to-date on a database that could allow for dynamic routing instead of the car relying completely on markers."

*Crowdsourcing long-term knowledge*
Unlike humans, the cars will already know where the road markings are, even when theyâ(TM)re covered: "Collaborative 3D Scanning with Paracosm and Project Tango" â" âoemultiple entities scan different parts of same the space, and join the data to create a 3-D modelâ: http://i.imgur.com/Y4OOdRe.gif.

*Crowdsourcing real-time knowledge*
Now in terms of real-time conditions, and winter driving, autonomous cars could constantly refresh each other with new information. Thereâ(TM)s a four-way stop thatâ(TM)s about two blocks from my house. When there is a good sheet of black ice, youâ(TM)ll see car after car slip and slide; itâ(TM)s extremely dangerous. As soon as a driverless car detects black ice, itâ(TM)s going to alert every single other autonomous car, and update them with the new info about that location.

Re:Not sure we need it

[drinkypoo]

On the other hand, I could see an interesting application in providing some kind of intelligent 'driver assist' for bad weather conditions that helped the driver maintain traction.

We have that already. It's called yaw control, which also implies traction control. It's mandatory on all passenger cars sold in the US since about 2010, I forget the precise switchover date. Four-wheel ABS and a yaw sensor plus a steering wheel sensor combine to keep the car headed in the direction the driver is requesting by pointing the wheel. Modern ABS can even detect that it's being useless (from accelerometer input) and lock up the brakes for just a moment to build up a pile of snow or sand in front of the wheel so that it can at least attempt to stop on such surfaces.

Regarding details, I think the ideal would be for most road conditions, detours, and traffic issues to be kept up-to-date on a database that could allow for dynamic routing instead of the car relying completely on markers.

The car can only be reasonably ever be considered semi-automatic if it can't sense the road conditions for itself, because the world is a lumpy place. If you're the first person to get to a downed tree, you need to be able to do something intelligent when you get there. Or in this case, the car does.

Re:Not sure we need it

[jabuzz]

While I in general agree with you, it can in practice be absolutely abysmal. Take last weekend, throughout the weekend it kept telling me it was going to rain basically all weekend where I live. It did rain heavily on Friday night into the early hours of Saturday. However after that it stopped, the mist/fog lifted by lunchtime on Saturday and it them remained dry for the rest of the weekend.

Re:Not sure we need it

[swillden]

And on Saturday afternoon what did the four-hour forecast say for Saturday evening?

Re:Not sure we need it

[number17]

You speak like a true southerner. The car needs to drive in the snow. I can't call my boss and say "Im not coming in today, the car doesnt want to drive".

This is a typical highway drive in Toronto. You tell me that your car is going to stop in a live lane because you hit a snow storm? That's a death wish. Road conditions change in an instant just as they do in the 2 minutes of that video.

Re:Not sure we need it

[nine-times]

The car needs to drive in the snow. I can't call my boss and say "Im not coming in today, the car doesnt want to drive".

So then you get in your car and take manual control. Or if the roads are too dangerous for the automatic driving or the manual driving, then they're just too dangerous.

Part of my point here is that the idea of a car-autopilot doesn't have to wait for things to be perfect. We have autopilot for planes, but that doesn't mean that we've taken pilots out of the equation.

Re:Not sure we need it

[nine-times]

If you're the first person to get to a downed tree, you need to be able to do something intelligent when you get there. Or in this case, the car does.

I agree. My argument is essentially that we could still require that a licensed driver is in the car, and that the "something intelligent" that the car does in case of a downed tree could be to pull over, notify the driver that due to a problem it cannot determine a safe action, and wait for the driver to take control.

Now that in itself is a difficult thing for the car to be able to recognize and decide and execute for itself, but it should be a "good enough" solution to start putting these systems into cars.

Re:Not sure we need it

[drinkypoo]

Now that in itself is a difficult thing for the car to be able to recognize and decide and execute for itself, but it should be a "good enough" solution to start putting these systems into cars.

It seems like Google street view provides enough data to do it, now it's just a matter of either putting enough processing power into cars, or abstracting it away to a remote server which is going to require a lot more mobile bandwidth. I wouldn't want that in my own car, but it would be acceptable for a self-driving taxi. If I'm worried about being tattled on, I just don't pick it up from or take it to my actual points of travel, but nearby instead and plan to take a short walk. But Google Street View hasn't done a high-resolution scan up my driveway as it has up some other people's, because I keep a chain up. That's how I want to keep it. I have taken some high-resolution photographs of parts of the property and permitted them to autoupload, but I choose where to point the camera.

Re:Smart cars are an abbomination

[mlts]

That has been tried. People will just drive illegally, using a "friend"'s vehicle, or find other ways to skirt the law.

I rather have self-driving cars as a solution. This means instead of watching for the drunks/texters, I can catch a snooze or read a book on the commute. Highways could be designed without fancy on/off loops, but just as raised, 4-way intersections with each vehicle's computer timing speed so that two highways can cross without stopping.

Are they solving the right problem?

[Minwee]

I guess they couldn't find anything that would make drivers smarter, so designing self-driving cars was the easier choice.

Re:Are they solving the right problem?

[StripedCow]

Also, when I leave work, I have to drive for 30 minutes, then cook 30 minutes.

With a driverless car, I'd still have to spend 30 minutes in the car and cook 30 minutes, a total loss of 1 hour.

If instead, they built me an automatic chef cook, they would save me 30 minutes!

Why only Google?

[StripedCow]

Why is every article about driverless cars about Google's cars?

Car manufacturers, like Mercedes, Audi, VW, are also working on this topic, and are probably even ahead of Google.

Re:Why only Google?

[geekoid]

Because those companies aren't talking about their RnD?

Re:Why only Google?

[kqs]

Those companies don't want to replace drivers. Drivers buy cars, and if you can't "imagine yourself behind the wheel of this car" then it's harder to sell someone an overpriced hunk of metal just because it is styled a bit differently. Those companies just want to assist drivers (help them park or stay in in their lane), so their videos will be much less impressive.

Only google seems to want to replace drivers completely.

Re:Why only Google?

[mjwx]

Why is every article about driverless cars about Google's cars?

Car manufacturers, like Mercedes, Audi, VW, are also working on this topic, and are probably even ahead of Google.

Actually, Mercedes and VAG (Audi and VW are the same company) are way behind Google in the automation race.

Japanese companies like Nissan and Toyota are looking to Google to provide the technology for automated cars (but Japanese auto makers of recent years are extremely risk averse).

Re:Smart cars are an abbomination

[Minwee]

Highways could be designed without fancy on/off loops, but just as raised, 4-way intersections with each vehicle's computer timing speed so that two highways can cross without stopping.

And you thought managing an Active Directory site with time sync problems was tough...

Will there be a human override ?

[raurau]

I'm thinking really rare situations, like post-hurricane Florida where all the traffic lights are out and there are trees on the ground. Or during evacuation when you're supposed to go north even in the southbound lanes.

Re:Will there be a human override ?

[seven of five]

"I don't know how to do this.... grampa, you remember how to drive manual-mode, right? You take over."

Re:Will there be a human override ?

[Ralph Wiggam]

Legally, self driving systems will be classified as elaborate cruise control. They can be disengaged very quickly and easily.

Human controls and overrides will be available for the foreseeable future.

Re:Smart cars are an abbomination

[Arker]

I've driven for multiple decades, across three continents, on both sides of the road. I've drive every day all day long during certain periods and I am sure if I had a mileage metre it would have rolled over a few times by now. I have never earned a single point on my insurance. So yeah, I think it's fair to say I drive better than most of the population.

I do not believe this is the result of any superhuman talent, virtually anyone could do as well if they would take it seriously. Unfortunately people in general are being taught not to take anything seriously and just blame someone else when they screw up.

Hmmm

[zmollusc]

The comments section is filled with people saying how the computer's senses and reflexes are so much better than a puny human, yet in the video clip, whenever the car is faced with anything except an open empty road, the default behaviour seems to be to slow to a crawl. If the other vehicles were also google cars, it would be interesting to see how they reacted to each other's manouvering.

Forget Cars!

[Mullen]

People talk about cars as the next big thing, but in reality, it's trucks. Google 18 wheeler will be able to get on the highway and travel across the country in a couple of days. The big transit points and warehouses are already next to the freeway, just load up Google 18 wheeler, have it take off and two turns later it is on the highway trucking across the country. Just need to make gas stations with attendants that will fill up the trucks gas tanks and check the tires.

Re:Forget Cars!

[eulernet]

Trucks are valuable because of their content.
It won't be easy to avoid robbery, especially if the truck is devised to avoid accidents, since it can be blocked by a single bicycle.

Re:Forget Cars!

[Ralph Wiggam]

Is there an epidemic of truck hijackings? (note that Fast and the Furious is a fictional movie).

Who is going to steal a truck's load not knowing what's inside it? Even if you knew what was in the truck, how does one go about unloading a locked truck in the middle of the road in less time than it takes for the cops to respond to the automated 911 call?

Re:Forget Cars!

[eulernet]

Is there an epidemic of truck hijackings?

This is because hijacking trucks is a little bit complex when humans drive and protect them.
You cannot predict what will happen, so it's dangerous.
With automated trucks, everything is completely predictable.

Who is going to steal a truck's load not knowing what's inside it?

Once the truckers' jobs will disappear, do you believe that more jobs will be created by some miracle ?
This will increase poverty, and poverty encourages stealing (I didn't imply that poverty means stealing), and poor robbers will steal anything that could be sold or consumed.
So yes, I think it's reasonable to imagine that the trucks will be randomly hijacked.

Even if you knew what was in the truck, how does one go about unloading a locked truck in the middle of the road in less time than it takes for the cops to respond to the automated 911 call?

When the automated trucks will be used, it will be for long distances.
Do you think that cops will be placed every 20 miles on the roads ?

Re:Forget Cars!

[Cytotoxic]

It looks like there are roughly 1,000 trucks hijacked each year in the US, representing some half a billion dollars in stolen goods. I don't know if that counts as an epidemic or not. Google tells me that about 100 people die each year from the drug ecstasy. Slate tells me that 297 people have been killed in school shootings since 1980 - or roughly 10 per year. Those topics are all over CNN. State officials are eager to do something! about them. So a thousand could be a pretty big number.

The NHTSA seems to think a couple of million people are injured in car crashes each year - a steadily declining but very large number. There are nearly 4 million home burglaries each year in the US. Which makes 1,000 hijacked trucks sound like small potatoes.

I suppose a truck that doesn't have to stop so the driver can rest is probably less likely to get hijacked, since that seems to be a recurring theme in the hijacking stories.

Re:Smart cars are an abbomination

[Kjella]

oh, right, you drive so much better then everyone else..just like everyone else is.

It's not that hard to beat the average and the ones who drag it down tend to not give a fuck. One is teenagers, I wasn't very good back then but hey the only way to get experience is to drive and you couldn't pry the license from the cold, dead hands of most them. Resulting in several total car wrecks and at least one person in my class who died in a 100 mph crash not that long after graduation. The other is the elderly, my mom finally gave up driving so now I can quite openly say that she should have given it up years ago. Realistically though, what are their options? Bicycle? Not a chance. Public transport? For a few things I guess, but mostly too far to walk and not going when and where it needs to go. Taxi? Possible I guess, but a simple trip to our cabin (1 hour drive x2 for trip/return x2 because they'll charge you both ways since they don't get return passengers) becomes hideously expensive even with the money saved on not having a car.

The same goes for everyone else with some health problem that really suggests they probably shouldn't be operating a motor vehicle and they know it. I think most families with small children would go crazy without cars, back when we didn't have cars we also mostly had stay-at-home moms. Or we're just temporarily impaired, like I've almost fallen asleep at the wheel and mostly likely only the rumble strips on the side kept me on the road. And even then I didn't stop, it was at the end of a five hour drive but I was only half an hour from my bed so I chanced it, loud blaring music, open windows to get fresh air rushing and caused myself a little pain to keep the body in alert mode. Stupid? Dangerous? Sure. But the alternatives sucked donkey balls, which is exactly why it doesn't matter if you're above or below average - you're not going to give it up anyway.

Accidents?

[FlynnMP3]

No matter how many failsafes they put into the engineering and algorithms, there will be accidents. Darn fewer of them since the majority are cause by human error, but they will still happen. I want to know what will happen when the self driving car is in an accident. How will it detect it? How will it determine what works and doesn't work? Will it automatically notify the necessary services (fire, police, ambulance)?

Re:Smart cars are an abbomination

[HornWumpus]

Really: I've got 30 years, no accidents, no tickets. Several autocross event wins. Several non at fault accidents avoided by seeing the fuckwit coming and getting out of his way.

What metric would you use in not accidents/mile driven? Any metric can be gamed, but that's a pretty bottom line one.

Yeah because snow storms can't be predicted

[SmallFurryCreature]

It takes an incredibly narrow minded and anal personality to come up with situations 99% of customers will never encounter and therefor conclude that because 1% might encounter them in a life time, an entire line of products is useless.

Oh no, a product isn't perfect for everybody! USELESS!

Re:Google claims a lot of things

[MightyYar]

You can buy a Q50 right now that - with the right options - will stay in a lane on a highway all by itself, steering as necessary and keeping a safe distance behind the car in front. Yeah, it's not a "self driving car" because it only handles one relatively easy scenario, but the fact that this is in a mass-market vehicle makes me believe that we are not so far as you seem to imply.

Re:Smart cars are an abbomination

[mjwx]

I don't want my car to be smart. Instead, make drivers licenses much more difficult to obtain.

This wont help.

You'll just end up with more people driving without a license.

Beyond that, you've got all the people who cant drive who currently hold license. I know its popular to blame everything on those darn kids, but the fact remains people who got their licenses 20 or more years ago weren't tested to the same standards and almost never kept their skills up to date. The 25+ group will be causing the majority of the accidents on the road.

What you need to do, is make it harder to keep your license. Only proper law enforcement on the road will reduce accident rates in the short to mid term. Proper driving training (note: training, not testing) will reduce it in the long term. Of course this means US drivers will need to give up their entitlement mentality (to be fair, Australian drivers aren't much different) and follow the road rules.

just think...

[slick7]

Smart cars for people with smart phones. How smart is that?

I hope they succeed

[Timmy D Programmer]

Imagine you have reached an age or have a disability where you cannot drive. How amazingly liberating would it be to be able to use a self driving car instead. It's not just a safety thing, think of how profound this would be for the blind community.

fix: Priscilla, not Pricilla

[short]

not a native but Wiktionary says the former

Google Search has been getting worse

[GPS Pilot]

One thing Google has perfected is using massive data sets generated from users to improve user experience.

I have to disagree. Google Search keeps changing the rules and doesn't always respect your query elements.

For example, you can read about how Google replaced the plus-sign operator with quotation marks: http://www.seochat.com/c/a/goo...

But what's worse than that: sometimes Google just plain ignores the quotation marks you put in your query. They're supposed to mean that each search result must contain the search term that you've surrounded with quotes. Nope, lately I've been getting a lot of search results that just don't contain the term in quotes.

Help fight against this trend toward dumbed-down search!

A landscape full of robot-cars

[mpol]

We're all aware that bots are all around us. Anyone with a website has probably more bot-generated traffic than human-generated traffic.
I wonder what will happen with cars. In 10 years we will not just have the occasional Google car filming the neighbourhood.
There will probably a whole industry of robot-cars without humans. Designed to look like cars, like a mini-car, but without seats, without stearingwheel and dashboard, etc.
It will be designed just for the robot. I cannot oversee just for what they will be used, but I reckon it's not just for Google-cars.
You could have a courier-botcar deliver a package, cheaper that a car with a human inside. Or maybe even cheaper than a bike-courier.
There will be new uses invented for cars. 20 years from now the landscape and the roads might look very different from today.

Illegal Right Turn

[Hokan]

In CA (as in most other states) a driver is required to move as far to the right as practicable to prepare for a right turn. This includes merging into a bike lane. The Google car didn't do that. It stayed in the through lane and turned (illegally) across a bike lane.

Seems that the car's programmers need to know a bit more about the driving rules.