Google demos driverless cars | Hacker News Reader

Google demos driverless cars

engadget.com

169 pointsmartythemaniak114 comments

If Google plays their cards right over the next several years, this could easily produce more revenue for them than ads do.

There are quite a few teams working on driveless cars. for example six won the last DARPA challenge , a european funded vehicle won did a 8000 miles trip from italy to china , israel is using similar technologies for military purposes, and GM and folksvagen is working on systems.

from the other side , google's total investment in the project is a yearly $15 million ,which is not unsourmountable.

I don't have a very deep knowledge about this , but my guess is that the big competitive barriers here are gaining enough real life experience , legal and regulatory barriers , and branding.

The companies that would make most the money from this technology would have those strengths. this looks more like a big car company than google.

Google hired many of those involved in the DARPA Urban challenge, and they are leading the effort. In particular, Sebastian Thrun, who appears to be in charge of Google's program, was the head of the winning Stanford team.

Additionally, Google's cars are already far more sophisticated than those that won the Challenge, which involved navigating a relatively simplistic simulated urban environment. According to their blog post on the subject [0] Google has been running their autonomous cars on real roads and highways in real traffic, a dramatically harder problem.

[0] http://googleblog.blogspot.com/2010/10/what-were-driving-at....

You can buy cars today that can almost drive by itself. Adaptive cruise control. Automated parking. Lane assist. Automatic braking to avoid accidents. BMW has demonstrated a driverless racing lap on Top Gear before can be seen in the link I posted in another comment.

I was downvoted for asking what this Google car did that other companies haven't already demonstrated. No answer though.

it doesn't 'almost drive by itself'. it drives by itself.

Car companies have made cars that do that too.

No completely autonomous car is on the market though. Not from Google nor anyone else.

This.

This is why it's smart to be a technology company first. It worked like crazy for Bell Labs for generations.

Agreed. Amazon's done amazing things where their competitors have... well, filed for bankruptcy.

Also a good example. 37Signals has great advice here: sell your byproducts. Amazon was forced to invest heavily in scaling dynamic websites as a byproduct of running a high-volume retail store, it invented a lot of internal tools to support that some of which they've adapted and spun-off into standalone products like AWS, EC2, S3, etc. In the process they've helped leverage themselves into more of a technology company than merely a web-store.

Did Google demo something that car companies haven't been working on for many years, and demonstrated years ago?

To the downvoters: Please answer the question. BMW has made a driverless car too. How is it different?

The iPhone has been able to drive cars since version 3.0.1. Google should really be focusing on making lists more springy at their ends instead of this sort of trivial thing.

Is this a serious post ? How does an iphone "drive" a car ? If you think GPS and turn by turn instructions is the same thing has actually controlling the car on a road, you couldn't possibily be more wrong ...

Your second point is strange too, do you actually think that a big tech company like google should not try to research new stuff outside of its comfort zone ? Or because one of their product is imperfect, they shouldn't work on another even if they have the ressources ?

I hope this was a joke that I didn't get ...

It was sarcasm. The point was that the iOS vs. Android wars pale in comparison to potentially world-changing research like this.

The ubiquitous spread of smartphones throughout the world is not world changing?

We're talking about platforms that struggle for mastery over an emerging market that will be bigger than the PC revolution.

"Android vs. iOS wars" refers to all the senseless flame wars that often get more attention here at HN. That's not to imply smartphones aren't world-changing devices.

Maybe this

http://www.youtube.com/watch?v=oHDwKT564Kk&feature=playe...

Tight turns look cool, but they're not especially useful. When they can demonstrate robust resilience to pedestrians, consistently don't think a bird or pebble flying in front of the car is another car, etc., then we'll have something to celebrate.

I'm not saying they're not a lot of the way there already - they probably are - I'm just saying that that's the demo video that would be really exciting.

So, I took a test drive in one of these today.

They can do the hard stuff but the tight turns make for a more interesting demo.

I will try to post a video later.

Awesome, thanks, please do.

Is it feasible to see free google taxi cabs in major cities that display ads give you internet and keep street view up to date in the next few years?

I would really like this on top gear and compare it to the other drivers on their closed course. I wonder how well it would do in their celebrity challenge that they do every episode?

It's worth noting that Stanford and Audi (an extension of the VW/Stanford work with DARPA) - will be competing in the upcoming Pikes Peak challenge. For those unaware, it's a balls out race up a mountain road. Asphalt, dirt, weather changes, and sheer dropoffs.

So far they've demo'd the car drifting and powersliding into parking spots:

Stanfords Audi (3min): http://www.youtube.com/watch?v=u2UaYIHiiOA

As mentioned above, BMW did just that a while ago (2min): http://www.youtube.com/watch?v=50rbUfTYykc

Lastly (I can't view the video), Google mentioned a while back they've already put over 50,000 miles on their fleet.

They already did it a few years ago in a self-driving BMW. Enjoy.

http://www.youtube.com/watch?v=DRF_KaWzxq4

I don't know if Jeremy Clarkson would let a Prius back on the show...

Why not? He has had them on the show before. There would have to be a reason for it to be on again though. The driverless lap has already been done. The Google Prius would just be slower.

I believe there are other models involved in the program; not just Prius-es.

Looked up the NYTimes article... it's also installed on an Audi TT.

If tuned to a single track it should be completely unbeatable, right?

If that were the case, then Shelly, the car racing up Pikes Peak would have been able to beat the human time of 11:21, instead of it's recorded time of 27 minutes.

Sounds like the Watson for cars. It'll happen.

Not necessarily there might be latency concerns or even other unknown factors that would hinder it. Also what if its wet or other weather concerns? I don't work on the project so I wouldn't know what exactly would be involved.

I do! Oh wait, and then I woke up.

That Prius corners pretty mean ...

Ahh so the all the recalled priuses were actually alpha versions of google's self driving cars. I see.

Ok, I get it there is no place for humor here. Xkcd can do it, I can't. Well I have a great respect for google as a company, that's the company I have always dreamt of working for, and my real feelings about this are that I feel great respect and a tad bit of jealousy for our real life iron men larry and sergey who backed by their enormously successful businesses can work on their world changing toys.

This is unrelated to the parent and I don't know where else to ask...

What is the karma threshold on down-voting?

If anyone can make this work out to be used by the public, Google can. They seem to be fairly good at pissing off established business. (Though they're still learning as Android shows.) Demos are great, just reinforces that Google is really an AI Company.

Where my pessimism lies for this reaching the public even if it's technically sound is that there are a lot of established multi-billion dollar businesses to contend with. Insurance is a big one, traffic tickets are another.

I don't think insurance and traffic tickets are nearly as big of a roadblock as safety regulations for selling a computer that creates 2,000lb projectiles with humans inside.

Wouldn't it be fairly simple to run tests for awhile and give the thumbs up if the average, shall we say, carnage rate is lower than with human drivers? Seems like it would be a no-brainer.

And when there are inevitably problems (or even deaths) caused by the AI systems, it's not like car companies don't currently have liability for malfunctions in their designs. Seems it would be the same...

You obviously don't work in a highly regulated industry :-)

Things that seem like "no brainers" to software developers are absolute show stoppers to regulatory bureaucracies like the Department of Transportation.

It's not impossible and it will certainly happen eventually, but it will take years of proposals and studies for a new set of regulations to evolve that these cars will have to follow. Liability is a separate issue that the car companies are used to dealing with. Changing minds at the US DOT (for example) will be a lot harder.

Insurance is easy: Google provides the insurance. Looking at the market share and market caps of Progressive and Allstate, which together have an 18% market share, Google is larger than the entire US auto insurance industry.

If applied correctly this could solved a lot of traffic problem. I still don't understand why cars don't have a retractable bar under them. So you normally drive everywhere, but big roads like highways are in fact rails where your car gets on.

Possibly because the necessary switching apparatus would be a nightmare with modern technology, and would have been entirely unworkable even 20 years ago. That said, it's a really good idea, and one which I would like to see attempted - or at least a good explanation of why it would be completely crazy on a technological level.

That idea sounds too simple to be true, but I'd file that under "crazy enough to work"

The craziest ideas are the best ;) and what's crazier: a car on rails or a self driving Prius from a search and advertising company ?

After looking at google maps/earth, especially street view, I would say they used driving data from that to feed the car driving learning mechanism.

Yeah, you blend your maps data with a bunch of sensors together and you have a self driving car. It's just a matter of speed and the other drivers + knowing where your going.

Do you mean something like this? http://www.ruf.dk/

It hasn't caught on. And I don't see the advantages that might be there outweighing the disadvantages.

I can see why cutting a car in half never caught on...

Wow! Nice idea. You could even skip the retractable when the rails fits in between your wheels.

This is amazing. I remember an old CMU paper from 1996 about AI driven cars being one of the reasons I decided to go to grad school. It's a classic IMO, and a very accessible introduction to the whole problem: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.15.4... (click the cached pdf icon in the top right corner). I imagine the solutions have moved on since then though.

Anyone care to hack up a little frontside car camera and steering position sensors?

Looks like what BMW had on Top Gear some a few years ago. Just faster. http://www.youtube.com/watch?v=0AYCUSAMF9U

A bonus video about what is more environmentally friendly and economical - a Prius or a BMW M3: http://www.youtube.com/watch?v=dKTOyiKLARk

Note that, witty and knowledgeable as he is, Clarkson isn't exactly the most reliable and unbiased source on the merits of cars he dislikes.

Disclaimer: I have an M3, not a Prius. But I average 27mpg.

It's entertainment. Do you think he said anything factually wrong in the clip, though?

"It's entertainment"? Shades of Michael Moore fans.

Unless all cars are driven by machine, the driverless car won't make a huge difference in our lives. You can already experience the magic of not having to drive by paying someone else to do it for you.

The real gains of automation occur when a machine doing the job can do it faster than a human. Computers calculate and communicate faster than people, and factories/robots build stuff faster. A machine driving a car has to obey the rules of the road and react to other human drivers. This puts a hard limit on their potential.

Now once all cars are driven by machine, we will see amazing changes. No traffic lights for starters.

This view seems deeply flawed. One of the upsides of public transportation is that I can read / sleep / work / talk with people with 0 distractions. The downside is that the commute is 3-5 times longer.

If I can get the shorter commute time of my car AND have the benefits of not having to pay attention, that's a huge gain even if my travel time isn't any faster.

Additionally, there are plenty of use cases that don't rely on interaction. Ideally, my car can drop me off at the door to the grocery store and go park itself and then come pick me up when I text it that I'm ready to go home.

The comparison to current public transportation is unfair. First, imagine what public transportation would be like today if as much money had been invested as has been invested in cars. There are more licensed vehicles in the U.S. than there are licensed drivers.

Cars simply cannot beat public transportation in raw material use, operating cost or road space utilization. My car occupies 70 sq ft of road space. At most, it can fit 5 people. That's 14 sq ft per person. A typical bus uses half that (calculated from an MCI D series). Since raw material use is more dependent on the volume, the savings will be even greater there.

My car gets 30 miles per gallon. Fully loaded, that's 150 person-miles per gallon. A fully loaded bus beats that even if it only get 3 miles per gallon. However, since buses run standard routes, they can be wired directly to the power grid, eliminating the need for separate fuel and eliminating the need for the batteries than an electric car needs.

And this is with current technology, developed despite substantially more money being spent on cars and highways.

The reason your commute is faster with a car is that there aren't enough buses/trains. Vehicles with dedicated routes (such as trains and buses driving on bus-only roads) don't get caught in traffic. If they were comparably common, there's no way you could beat them on time either.

First, imagine what public transportation would be like today if as much money had been invested as has been invested in cars.

Honestly, I do this all too often. What if the American passenger railroad system had never been left to die? What if the Big Three hadn't bought up streetcar services all over the country and run them into the ground? What if we'd been investing money for the past half century in making advances in new rail technology instead of trying to make dangerous cars slightly less dangerous?

The intermittent dialog about high-speed rail here brings to mind an obese diabetic who has resolved that he definitely needs to start going to the gym one of these days.

Ok, I think I'm done ranting.

Imagine cars on demand. You don't own a car, you text a car to get you to work at 8:30, the car picks you up at home, drops you off (no parking space required) and goes on to the next ride.

I can't wait to see driverless cars.

Almost like a taxi?

I guess with driverless it's more like cloud carpooling.

It'd be cheaper than a taxi since there's no driver to pay.

Yeah. Just buy the car first. And automated car insurance. And the monthly fee for internet connectivity so the car can get data from google.

That's an ingenious concept! I will start a company like that right now, I will call them... Taxis!

Yes almost like a taxi, what I meant to say is that today you don't drive to work in a taxi because they are so expensive most of the time or inconvenient because few are available.

If driverless cars became maintream I'm pretty sure it would drastically decrease the price of such commutes thus changing the way we own a car today.

It's called PRT (personal rapid transit) and in my view, it's the one glimmer of hope in turning a century's worth of investment in automobile infrastructure from a liability into an asset. It's a fantastic idea that's gaining some limited traction in places like London Heathrow (http://www.ultraprt.com/prt/), where hopefully it can gain enough of a foothold to expand. Right now, PRT systems tend to use their own rights-of-way, but with the advent of true driverless cars, all that could change.

This is wrong of course. Imagine you could hire a driver, now imagine he worked for free and was always at his best driving ability.

The things that self driving cars will do first, really really well, are: 1) Drive people around who currently use chauffers because they can use the driving time productively. 2) Lower the barrier to attending events, restaurants, etc in urban areas where parking is an issue. (if you car can go park itself a couple of miles away and you can phone it up when you want to go home, win) 3) Reduce the number of DUI accidents as people go out, have a good time, and their car drives them home.

Self driving cars that can drive amongst human driven cars will be a huge win. Now why a search company would do with that, who knows.

Now why a search company would do with that, who knows.

Google is an AI company. Search and ads just happen to be the currently most profitable application.

> 2) Lower the barrier to attending events, restaurants, etc in urban areas where parking is an issue. (if you car can go park itself a couple of miles away and you can phone it up when you want to go home, win)

It lowers the barrier on travel in another way too. Put a comfortable bed in the self driving car and anywhere within 8 hours drive is now a night's sleep away. Weekend tourism would be a lot more fun that way.

Aggregate data about trips and make statistics and whatnot.

Rush hour in major cities would be an entirely different experience with automated cars.

In my observation, the vast majority of congestion is caused by drivers either overcompensating for changes in conditions (ie slowing down even more than the car in front of you slowed down) or otherwise reacting slowly.

I believe highways could handle considerably more traffic if all the traffic were automated.

I think it's possible to see significant gains if only some of the cars were automated, actually.

Left lane for automated cars, right lane for manual prudes.

In mine and a lot of others' observation the main cause of traffic is people's inability to merge. People suck at it. And that in turn creates traffic waves... http://trafficwaves.org/

Increased density of cars on the road has the potential for a huge economic wins, in terms of getting governments out of widening highways to increase throughput.

Automated cars would be be able to (one day) receive information about down-the-road conditions and congestion which would enable the automation of William Beaty's "jam busting" techniques at a more expansive scale: http://amasci.com/amateur/traffic/trafexp.html (For example, slowing down gradually, rather than screeching to a stop as you pile into the back of a jam.)

I think it's possible to see significant gains if only some of the cars were automated, actually.

I've thought about this as well many times.

Is it that an automated car could break a chain of overcompensation?

This sounds really interesting could someone knowledgeable elaborate?

It would be sweet to see a time where only a majority of the cars on the road are driven by machine. I imagine them driving in packs, and leaving lots of buffer space between them and the lone human driver.

Yes, these secondary affects are interesting to think about. Also, if people don't have to drive 95% of the time, people will forget how to drive, and will probably be unable to instantly switch into 'operator' mode when/if the car decides to transfer back into manual mode. For this reason, I don't see the technology being successful until it can drive reliably enough that it never has to switch back into manual mode.

Can you imagine reading a book, watching a movie, eating (or whatever) in your car and suddenly the situation gets so hairy that the car switches back into manual? I don't think that would work out, especially if the last time you had taken the wheel was 5 years ago.

Not exactly real-world conditions, but a nice demo nonetheless. Maybe this is how they've been able to Street View so much of the world?

I wonder how long until we see the iDrive. I'd love to not-drive myself everywhere, it's a massive time-sink IMO, I could be doing better things in the same time period. I understand some people enjoy it, but I don't.

AFAIK, these cars are still being manned by the engineers, they're not used in any official capacity.

I have seen one on the 101. They are very weird looking things when you look in and see that the "driver" is on a laptop. It's pretty freaky.

If you don't enjoy driving, there are alternatives such as public transportation for short surfaces, and trains/airplanes for longer. And I don't mean this in a bad way, I'm just saying.

It would take me several hours to get to or from work on the bus, and trains around me are both far away and quite expensive. Heck, it sometimes took the buses two hours to get me to school 3 miles away, when they're supposed to be coming by every 30 minutes.

I would if I could. Honestly. I want to get rid of my car. It's just not possible here, currently, while still being able to do anything necessary.

It seems like folks in cities, and some foreign countries, don't understand how much different the transportation situation is out in rural America. Getting to mass transit is frequently no easier than getting to your actual destination directly, even for non-trivial trips.

i wonder what proportion of the labor market this will affect. basically anyone using cars for delivery?

no more taxi drivers, take-out delivery, postal delivery, etc.

There's still a problem to solve though:

who's going to accept payments, process credit cards, verify signatures etc.

Just like there needs to be people who do that at Amazon, Zappos, etc... </sarcasm>

Honestly, a taxi with only a Google maps interface and a credit card swiper would be interesting.

No more tipping!

That is currently the most annoying part of taking a taxi when travelling. Why taxi companies are still using technology from 15 years ago for processing credit cards (assuming they even accept them) is beyond me. If dumping the driver makes it easier for me to swipe, the sooner the better.

How do vending machines work currently? :)

Does anyone know about the algorithms behind this? Are they still using Neural networks?

I just can't fathom how a computer can know how to drive.

"Remember the times where we used to drive our cars?"

That looks really cool although I'm also interested in how it works. Any infos there?

I, for one, want to stop and thank google for making html5 videos from youtube embeddable.

I want to see how it handles a blowout...in the rain.

I would imagine way better than a human. Computers don't get scared.

Instead of scared, they crash :)

The biggest danger is an overreaction, which a computer is much less likely to do (if programmed correctly of course). Newer luxury cars already save you from yourself when you try and overcorrect, so I suspect this is already a solved problem.

So what OS does it run?

Driverless cars are not going anywhere until they can prove themselves 20% superior than the average driver in ALL the following conditions:

Dealing with Rain (visibility and deep water on the road),

Hard to see tire eating potholes causing damage,

Ice and snow (on road and caking on instruments),

Snow accumulation ruts, causing car to move unexpectedly,

Evasive driving - cutting across a lawn, (avoiding the pool and coming to rest over the garden gnomes),

hurricane winds, hailstorms,

ambiguous and contradictory road signs,

unmarked roads,

interstate highway pileups,

white-out snow conditions.

I'm pretty sure your going to need pretty strong AI for all this to happen. By then we shouldn't be needing to go to work because everyone on earth can go on vacation and the machines will do everything that needs to be done. AND all this equipment has to be less expensive then hiring someone to drive around on your behalf.

Naysaying aside, self driving cars will cause a revolution in how cars are used, the concept of "owning a car" will disappear and then single passenger cars will fill the roads instead of 4 door seudans/SUV's.

Interestingly enough, a car which simply refuses to drive in hurricanes, hailstorms, white-out snow, and so forth would be rather more than 20% superior to the average driver. So that's easily dealt with - if insufficient sensory data is available, either hand control over to the driver or, if possible, pull to the side of the road. The average driver likely does really horrible in large crashes and pileups, so merely the lowered reaction time might allow a huge improvement there (humans take a really long time to react, compared to computers). The average driver doesn't need to do evasive driving (as well as likely being horrible at it), so that's an unnecessary criteria. Similarly with tire eating potholes; even a tiny improvement would meet the 20% requirement you propose.

Ambiguous/contradictory road signs and unmarked roads are a navigation problem; given a sufficiently accurate database, as is likely to be present in most cities, they would be irrelevant (I expect that initial roll-out of driverless cars would occur in cities, where most traffic would be low-speed and highly structured. I may be wrong).

Unexpected car movements (due to snow, mud, or whatever else) I would agree to be fair things that need to be targeted, but I very much disagree with strong AI being needed for all of this. However, it's probably going to be quite a long while until the equipment is less expensive than hiring someone else to drive (which is a barrier for some group behaviors and sharing of data between cars). On the other hand, that's not necessarily a consideration which people factor in when buying extremely expensive cars.

One of the nice things about software drivers is that you can spend as much time as you want training a module for driving with, say, a given tire blown out and then swap that module in when actual conditions match the conditions that its trained for.

It's more of a sensory problem than a logic problem.

Problem is, in twenty years, the car would be handing over to a really amateurish driver.

If it allows the car to be driven. I'm afraid I can't let you do that Dave.

Much as it would today.

car which simply refuses to drive in hurricanes, hailstorms, white-out snow...

While admitted these conditions can sometime be anticipated, I think the point is that sometimes these conditions cannot be anticipated. You experience a sudden white-out a minute after a snow storm starts. What do you do? Breaking immediately is not the answer! (thus, the cars will need to drive...)

White-out conditions are much easier to see through with radar than with visible light. I think the computer will easily top us on that one.

How Ironic. My very first job out of college, in 1995, was working for an avionics firm that was developing a system called "Tundra Tracker" - that would allow vehicles to navigate in whiteout conditions in the far north. I had the amazingly easy portion of taking the DGPS coordinates of maps/previous tracker runs and turning them into navigation maps. Challenges with the (dynamic) appearance of people, other vehicles, animals (Moose will ruin your day) crevasses, and other elements on the path were left to others to solve. :-)

I think your 20% figure might be low, but that doesn't mean that we won't see specific applications for autonomous cars earlier.

Autonomy is going to creep into autos slowly, things like: anti-lock brakes, adaptive cruise control and collision avoidance are all small pieces of an "autonomous" automobile.

A car dealer at the Ford dealership told me Ford is demoing cars that parallel park themselves.

Lexus has had automatic parallel parking for a few years now. http://en.wikipedia.org/wiki/Intelligent_Parking_Assist_Syst...

Not really; to be useful it doesn't have to do everything all at once.

I'm not in the market for a new car, but if I were, and there was one available that could drive itself in good weather only, on an expressway in reasonably good condition, then it would be worth a small premium.

Most of my driving is on the interstate or other expressway in relatively little traffic. A car that could handle this would make the daily commute, or weekend visits to friends that live 3 hours away a lot more pleasant.

On the other hand, I could care less about automatically driving in the city, since I don't live there and go very rarely.

For most of those conditions, safely pulling off the road and telling the human to take over would be satisfactory.

An automated driving system also has a lot more potential sensory inputs than eyes and ears. 360-by-360-degree cameras, radar, IR, etc, which might be able to see through rain, snow, hail, etc. just fine.

Unmarked roads, bad traffic patterns, and routing problems are things we already rely on maps/GPS data for anyway. This system can look at these and get real-time updates without taking its 'eyes' off the road.

There will doubtless be lots of peripheral benefits from this piece of research. But private vehicles being automatically driven on the road ? Is Ford/Audi/Chrysler going to swallow the liability implications ? Look how much they get burnt by questionable brake cables....

Reply on news.ycombinator.com