Volvo’s first self-driving cars now being tested live on public roads in Sweden
kurzweilai.net
kurzweilai.net
Building high quality safety software and sharing it with all car manufacturers wins way more brownie points in my book than using it as a selling point.
The latest models also detect pedestrians, bikes, and even animals -- and the car sets off an alarm and brakes so that we don't run over them. High beams are politely turned off when meeting another car, as well as when passing near pedestrians. The software also performs lane detection, shows you where you're heading when moving in reverse, signals to the driver when another vehicle is close by to the left or right, even if you can't seen them in the rearview mirrors.
Volvo has a goal of preventing or mitigating accidents to such an extent that by 2020 nobody will die in a Volvo when there is a car accident. Their commitment to the safety of car passengers was such that they let everyone freely use their patented seat-belt system when they improved it. They've been running a team in Sweden for decades that travels to every single car accident involving a Volvo, so that they can understand the causes and consequences of the accident, sometimes even repeat the event in lab conditions -- and they can improve the cars accordingly.
Yes, we own a (new) Volvo -- we picked it up at the factory in Sweden, drove around Europe for a few weeks, and returned it for them to ship it over to our US address.
There are a lot of other variables to optimize for, both on individual basis (fuel, travel time, battery life, ...) and collective basis (congestion locallity, total time lost in congestion, emissions, ...)
I guess legislatures - over time - will try to optimize some variables for 'the greater good' instead of minimizing your travel time :-)
I'm not sure what you mean with the other option ?
It's sad that asking questions is considered bad.
This functionality is demonstrated in one of the more recent ridealongs of Google's car.
[1] http://commons.wikimedia.org/wiki/File:VBA_showing_100_kilom...
It also covers the (admittedly limited) situations where no GPS is available (tunnels, skyscrapers). If there are temporary changes in speed limits due to construction or other events, OCR could immediately identify those with no need to download new information.
* Severe congestion pricing
* Expensive due to severe taxes (generally over 100k USD)
* Licensing for a car is time-limited. I think this is about 10 years or so.
The last point in particular makes it feasible to cycle out the entire country's cars in a relatively short time.
EDIT: Add more newlines between bullets
In US it is 4 years. I was surprised to see that low time frame as in India it is at least 20 years. And, in US, you are also required to renew your registration plates every year
Also, your US numbers of course vary by state.
For example, Arizona driver's licenses don't expire until the person's 65th birthday (and every five years thereafter).
(source: Volvo Cars press release: https://www.media.volvocars.com/global/en-gb/media/pressrele... )
I have a bad feeling that Google is going to keep everything sealed in Google Maps. The industry would benefit from sharing all of their data with OpenStreetMap.
Yes: http://www.reuters.com/article/2014/02/03/us-autos-technolog...
But I don't think it's for mapping, just safety.
And yes the safety applications like emergency brake, intersection collision warning, or curve speed warning etc., have high priority but other Applications like mapping are also possible. I think current mapping applications use road side infrastructures (also through DSRC) but I guess you could share traffic information (Waze?) using inter-vehicle communication.
Here's an technical paper if anybody is interested: http://secs.oakland.edu/~gpcorser/vanet-kenny-DSRC.pdf
What is the barrier to entry for an open platform for people to submit this same data to OpenStreetMap. Precise Positioning + GoPros + data processing on the backend.
Major projects like this have a ton of hardware limitations.
- Why Sweden has so few road deaths [1]
- Vision Zero (fatalities and serious injuries on highways by 2020) [2]
[1] http://www.economist.com/blogs/economist-explains/2014/02/ec...
Google car drives only in good weather in California for a reason. In Sweden they could drive maybe a week in a year.
... Sommaren är kort
de mesta regnar bort ..Cracking the heavy rain/snow problem would require more than just a new design though - you'd need a totally different sensor modality, a mix of sensors that can extract enough data, or a system that's far more robust to noise.
I have no doubt that smarter roads are in our future, but they pose too high of a price for wide spread adoption of autonomous vehicles at the outset.
I really don't think it would be that expensive, at least along Interstates and major highways, to embed a trace wire to help guide self driving cars. (At least, during regular maintenance and new construction- tearing up roads to add it would obviously be expensive).
http://youtu.be/EkvazIZx-F0?t=1m56s
Maybe it works only because they use multiple cameras and they don't need too high resolution.
Regardless I wasn't aware of this before. Why would Google chose this technology ? It's bizarre.
Google is not planning to monetize this technology anytime soon, despite the hype.
The difference between crude human/animal intelligence and top notch AI-research is still huge. If people would need the accuracy that Google's car needs to move reliably and do split second decisions, we could never leave our house. We operate using just two cameras and accelerometers. The clear picture and spatial recognition is done using top notch heuristics in the unconscious. With self driving car it's the opposite. They need millions of very accurate distance measurements per second to drive. Driving like Google car does with cameras only is not happening yet.
Mercedes Benz from Germany is doing active research in dynamic computer vision of driverless cars since the 1980s.
"1758 km trip in the fall of 1995 from Munich in Bavaria to Odense in Denmark to a project meeting and back. Both longitudinal and lateral guidance were performed autonomously by vision. On highways, the robot achieved speeds exceeding 175 km/h" ... "This is particularly impressive considering that the system used black-and-white video-cameras"
-- http://en.wikipedia.org/wiki/Ernst_Dickmanns
"In August 2013, Daimler R&D with Karlsruhe Institute of Technology/FZI, made a Mercedes-Benz S-class vehicle with close-to-production stereo cameras and radars drive completely autonomously for about 100 km from Mannheim to Pforzheim, Germany, following the historic Bertha Benz Memorial Route."
I was with you up to there.
'A million measurements' sounds really impressive but it does not have much to do with anything. What's a measurement? A single distance measurement in front of the car? Ok, at what opening angle, how many returns, how many pulses / second and so on.
As it stands that's just a 'big number' but those are not impressive at all without context.
now you put second camera near-by and run stereo analysis algorithm to build 3D scene. 10+ years ago (DARPA Grand Challenge - where roots of Google self-driving car architecture comes from) with 1M cameras and the available hardware you'd get lucky to get 1 scene/sec and very crude one at that as 1M is much lower resolution than our eyes, and resolution is the key to stereovision. With LIDAR you just get 3D point for each measurement, no processing (beside regular filtering and coordinate transformation)
I wonder (haven't touched it myself for years nor checked the literature) what stereoprocessing one gets today on 10M-20M cameras on Intel CPUs of today plus GPU. It should be pretty close to what our eyes do, and what is most important - using several 20M cameras you can probably do _better_ than our eyes.
That said, stereo runs pretty damn fast these days. On ASICs. TYZX, who was bought by Intel, sold a stereo camera about 3 years ago that ran ~52 fps with full point cloud returns. I think those were running 2+ Mpx.
this is one of the reasons why i said about several cameras - each camera, pair of them, can cover different [overlapping] subranges of light sensitivity and each do it better than eyes in each respective subrange, and thus the integrated image may be better than eyes'
(how many times per second is the same point revisited)
(1) Delivery-bot. A car that drives itself to drop off a package and only delivers on non-rainy days. (if you need delivery on a rainy day, you pay extra for a person-driven delivery service. Otherwise the package waits at the warehouse)
(2) Transport option for people who can't or shouldn't drive themselves - too old, too blind, too young, or physically impaired. The self-driving car takes them where they need to go when weather permits, otherwise they have to call a cab or van service as a backup option.
(Much of California only has a couple weeks of rain per year.)
On the other hand, I imagine that auto manufacturers are much more interested in getting to viable product--whether it's improved assistive driving features (collision avoidance, speed matching, etc.) or, in the somewhat longer term, autonomy for some limited range of conditions. Hence, for example, Volvo's involvement of government as well.
https://plus.google.com/105073266122249925354/posts/a7XDM2xg...
There's SOME kind of thingy on the roof. Not sure what it is though. Any ideas?
Didn't want to go too close though: https://www.youtube.com/watch?v=A_83efj9Xdg :)
Why do you say that? Somebody is still going to need to be awake at the controls, at least for the first few decades of self driving tech. There's a reason pilots on planes don't go to sleep when the auto-pilot kicks on.
Good point on the pilots, though I always felt that it's unnecessary for them to stay awake (especially on cargo planes). And it looks like many do fall asleep inflight: http://uk.reuters.com/article/2013/10/08/oukoe-uk-britain-pi...
Truckers, on the other hand, need to be alert all the time, falling asleep for even a few minutes can be catastrophic.
I'm mostly hoping for fewer human drivers - two drivers/technicians should be enough to drive a whole convoy of trucks between cities/countries.
Maybe autonomous vehicles will make smaller trucks cost-effective, and big trucks will then be charged their true cost in road damage and traffic risks.
That would be a much more visible change in the traffic landscape.
This is an important step, although I must say they appear to be far behind Google. My money is on Google getting to an acceptable deployment phase far earlier.
Google appears to be making an advanced and expensive solution to "full" autonomous driving. The limitation will not be in where it can go, but it is initially limited in weather/environmental conditions. The target audience is likely not production cars in the short term, but a system like theirs can for example make their street view photography cars autonomous: they can work in summer, stop if it rains, and it doesn't matter too much if they cost twice as much as a regular car. It will of course be a regulation nightmare to allow the driverless cars out in the street though.
Volvo on the other hand will probably try to make a simpler and much less expensive solution that will have to work in rain/darkness/snow, and not add a significant amount to the cost of the car. Clearly that means having other restrictions, presumably in the overall intelligence or in which roads it works on. Volvo's system is likely intended for making day-to-day travel more secure at the lowest cost possible, not for creating completely autonomous navigation and problem solving (e.g. 4 way stop turn-taking). It must work in almost any environmental conditions, but it will probably require a driver at all times, and will almost certainly not work everywhere.
Street View is hopelessly out of date in many places (4 years in some parts of Sydney for example) and it has to be super expensive to keep thousands of people continuously driving around the world. And Street View will likely be a vital and profitable data source if VR/AR (in all its forms) is ever used en masse.
Sounds like too much expense for just that. If the problem of street view was it's cost you could just pay people to add a StreetView system to the roof of their cars and get the data much cheaper than having your own car. They may very well want to start by using Street View to validate the technology but the end-game must be larger.
I think you're underestimating the number of people you'd have to pay to get full coverage of cities. Most people do not drive around on every street in a systematic way; they have a small number of destinations and they usually take the same routes between them every time.
That doesn't mean that once you do have self-driving cars you wouldn't use them for that. Or even that StreetView wouldn't be the first thing you'd use them for, since that's a good way to get more training data for your algorithms. But to suggest they'd go to all the expense of creating self-driving cars to then use them just as a cost saving measure for StreetView sounds strange.
Not unlike the experience for drivers on Lyft and UberX it would be trivial for Google to create a rig and an app. You set the magnetic camera rig on your roof and sign in. It shows you available routes and how much they'll pay for them. The offer increases every so often if nobody selects the route. Done.
> Vissa korsningar har en stoppskylt vid varje väg, ett så kallat flervägsstopp. Det innebär att alla som kommer till korsningen måste stanna, oavsett vilken väg man kommer på. Tanken är att man vid flervägsstopp ska ta ögonkontakt med medtrafikanter för att på så sätt komma överens om vem som ska köra först.
Or my translation "Certain intersections have a stop sign at every way, a so-called multi-way stop. This means that everyone who comes to the intersection must stop, despite which way they come from. The thought is that people at the multi-way stop will make eye contact with their fellow drivers in order to come to an agreement on who shall drive first.
(Apparently the actual law says "ömsesidig hänsyn" - "mutual consideration.")
That said, they aren't anywhere near as common as the US.
If you're more curious, see http://www.trafikverket.se/PageFiles/3038/komplett_slutrappo... , among others. That one talks about how they were tested out during the 1980s, and by 2007 there were a few hundred in the country.
Unless I misunderstood the problem.
I've driven a bit around Europe, but I have never seen a single such intersection in my life - if they exist, they must be rare.
Of course, "if you have an intersection of non-priority roads and a car from each direction" then it's the same, but in practice it seems that 'they' make sure that such intersections are only in extremely low-traffic places where you'll very rarely see another car at the same time.
In the countryside you'll often see this warning sign to remind you that "just because your road is bigger and faster than the one you're about to cross doesn't mean you have priority" http://www.transportstyrelsen.se/Vagmarken/Varningsmarken/A2...
In Europe you can have ambiguous situations with the priority to the right, where you are supposed to be courteous and wave the driver if both (or more) have the right of way.
So basically they're testing under the 'worst' circumstances, making it likely that it will perform well on 'easy' routes as well. This approach makes sense - what good is it to clock 1000's of kms of test drives on straight roads with no traffic and in broad daylight?
However, one major traffic simplification in Europe is that there is no such insanity as "go ahead and turn right even when the light is red -- and run over crossing pedestrians unless they back off."
I'm excited to see more "in the wild" units - more competition and innovation is always a good thing.
"The self-driving algorithms do not work because there has been some breakthrough in artificial intelligence; they run on maps. Every road that Google’s robo-cars drive on was first surveyed by a human-driven pilot car outfitted with sensors accurate enough to measure the thickness of the painted lines in the middle of the road. Every detail of the road has been mapped beforehand. According to Peter Norvig, Google’s head of research, it’s a hard problem for computer vision and artificial intelligence to pick a traffic light out of a scene and determine if it is red, yellow or green. But it is trivially easy to recognize the color of a traffic light that you already know is there."
http://www.nytimes.com/2013/12/15/magazine/googles-plan-for-...
Plus, every time the automatic car goes over the road, it's presumably checking its prior maps and flagging any anomalies from its preset mission. It still needs to recognize a new stop sign, but if a stop sign disappears, it will stop anyway while waiting for an answer on "was it was taken out on purpose, or did some kids steal it?".
I may be reading too much into this, but when I read "test route" in the article, it gave me the impression that the Volvo approach is sort of like a train without the tracks -- only able to follow a pre-programmed route from point A to point B.
The quote ghaff posted https://news.ycombinator.com/item?id=7698610 it seems like you are describing Google more than Volvo.
That's not an explanation, it's speculation just like my speculation is.
note the:
Our aim is for the car to be able to handle all possible traffic scenarios by itself, including leaving the traffic flow and finding a safe ‘harbour’ if the driver for any reason is unable to regain control,” explains Erik Coelingh, Technical Specialist at Volvo Car Group.
Plus the goals listed.
Volvo did do a land-train style project called SARTRE http://www.motorauthority.com/news/1072088_sartre-self-drivi...
Given the court system, I don't think you could do that in the US.
It's a science experiment.
Well, that's better than most human Volvo drivers can do, by my experience. Whenever I see a car make a bone-headed move in traffic, chances are it's a Volvo.
SO if we end up with self driving taxis then this form of transport will become ridiculously cheap as the driver's salary is now eliminated. The rationale to own a car will become harder to justify.
Also I read recently that our actual usage of cars is very low over a year. This means huge inefficiency. There is plenty of spare capacity per car to do alot more journeys. If we all start taking self driving taxis that run 24x7 in a few years time where does that leave the car companies. One assumes with far fewer car sales. Are they shooting themselves in the foot?
> We have a big emotional attachment to cars evidenced by
> the fact that the annual cost of taking taxis everywhere
> is probably cheaper than the annual costs of owning a
> car if you live in a city
Note: I know EyeSight can't drive the car itself.
https://www.youtube.com/watch?v=lcJ9xgBZJdI
The explanation was "The failure of the test was due to the dummy not being set up properly", which I find rather less than reassuring.
http://www.drive.com.au/motor-news/volvo-crash-test-goes-wro...
On the 'up side'; if the pedestrian safety air bag fails, it'll be just like any other car.
They are pretty useless if they only fire when the car is already turning. The whole point of them is to give a warning prior to turning.
But I like manual transmissions as they force me to pay attention to the act of driving quite frequently. Without that constant reminder, I have a tendency to zone out and go on autopilot myself, which is of course dangerous.
Manual transmissions are my hack to attack this issue - but I’d love to have a self driving car where I was reasonably confident that my zoning out would not be a safety hazard. (As a manual transmission isn’t a great hack, and I live in an urban area, I mainly take the bus, which solves the "zoned out problem" completely.)
You can't have a system that doesn't need a person to do anything 99% of the time but requires that same person to take over control at a moment's notice to deal with some situation that the computer doesn't know how to handle or, indeed, didn't even recognize.
[1] http://en.wikipedia.org/wiki/CAFE#Agreed_standards_by_model_...
[2] I'm not saying the days of the V8 are over (like in Mad Max 2 -- "The last of the V8 Interceptors. A piece of history.") .. but if you want one, you might want to start thinking about buying one to keep.
http://consumerreports.org/cro/2012/01/save-gas-and-money-wi...
If you choose the CVT in the Nissan Versa, you'll get 5mpg better economy over either the manual or the automatic choices.
http://www.edmunds.com/fuel-economy/five-myths-about-stick-s...
That said, trucks and passenger vehicles with rubbery shifter bushings might be a different story. There's a lot of shifter and clutch variation out there.