Yeah, it'd be nice if more people read patent claims before passing judgement.
I took the obvious practical implementation of this to be a method to guide autonomous vehicles into a parking lot/structure. The car rolls up to the landing strip, the driver gets out, the car switches to autonomous mode, gets the connection information to the parking management server, asks for and receives a spot assignment and instructions on getting there, etc.
GPS and map data are simply insufficient to handle autonomous parking. There are precious few universals in parking laws, customs, lot-flow, etc. And automatic parking is one of the unavoidable expectations of self-driving cars.
> it'd be nice if more people read patent claims before passing judgement.
It'd also be nice if patents were written in a language that made them readily understandable. Then more people might read them. I'm sure I'm not the only person who finds them obscure bordering on incomprehensible.
If you want more people to read patents, you need to make them more understandable (as you say), and get rid of treble damages for potentially accidental infringement of a patent you "should have" been aware of because you were reading patents.
In many cases, it is exactly their intention to write the patent in that obscure way, in order to have more wiggle room, and in most cases it's lawyers who write those patents, not engineers.
And automatic parking is one of the unavoidable expectations of self-driving cars.
Actually, I would be perfectly content if I only had to park my car, but it drove itself from just outside the source parking lot (or street) and the destination one.
And just not having to drive on freeways (a much more controlled environment than usual roads) would be incredibly useful :)
Presumably based on the wording of the patent, the QR code just contains a URL from which to fetch the actually instructions for the car. Thus I would assume there are security measures built into the instruction retrieval, SSL or some other authentication/encryption scheme. Thus simply changing the QR code wouldn't allow forged instructions to be sent to the car, you'd have to break the security on the data link.
There are laws, and social mores - but those certainly don't stop people from hacking road signs or removing traffic signals.(1,2) Moreover, unlike spikes on the road, a tampered-with indicator might not be immediately obvious, even in its effects.
Anyway, I finally had a chance to read through the patent; it's enough of a concern that they specifically address it:
"In some embodiments, the data provided by the reference indicator 250 may be encoded or cryptographically signed for security purposes. By providing encoded or cryptographically signed data, the ability for a third party to disrupt the autonomous vehicles may be mitigated."
Symbols on the ground are infrastructure, you should not be able to patent a thing like that. Imagine every company that makes vehicles having to license someone else's tech in order to be able to use public roads, or a huge variety (one for every brand) of competing symbols polluting the roads.
Signs on roads == traffic signs, open formats and readable by everybody.
The person who invented the original traffic signal patented it in 1923 (http://www.google.com/patents?id=EHJoAAAAEBAJ&printsec=f...). The idea was unique, novel, and non-obvious at the time. It's easy to see an invention now that is ubiquitous and say that it's obvious.
Stuff like this makes me twitch. Why are researchers so keen to have my car controlled by radio tags? The potential for abuse is seemingly never discussed because I feel that admitting it will kill this idea stone dead, as it should.
Are you implying that researchers want to take control of your autonomous vehicle on behalf of the government or some overreaching corporation? The marker/URL scheme is just a solution to the problem of delivering instructions when needed, I don't think it's malicious.
I think the implication is that researchers aren't discussing the possibility of someone else using their system to take control of your car. If you are security conscious, and your car can be controlled externally, you might want to assume that it will be controlled externally until you can prove otherwise.
Look at it this way: if driverless cars are widely adopted in Country A, and the new Country A government decides to go rogue and oppressive on its citizens, the rogue government could prevent people from getting to protests by shutting their cars down remotely.
Most protests take place in cities, where one can walk. What you describe is as 'first world problem.' In repressive countries, the government just sends out some troops to establish a checkpoint and stop cars by pointing guns at the drivers. No special technology required.
Let's suppose "Country A" is a country that is subject to first world problems, and that many of Country A's angry citizens are in suburbs and semi-rural areas within driving distance of ideal protest zones, but they want to express their anger in person.
There's no need for fancy engineering to prevent driving to a protest. Most cities already have traffic control mechanisms in place - bridges, toll booths, tunnels - that are far simpler and harder to work around. I believe New York City has shut down various passages into Manhattan in response to threats or protests over the past few years.
Sure, but it's significantly easier for a small group of people (a government, hackers, whoever) to remotely disable someone's car than it is to mobilize enough police to have the same effect. Believe me, I want self-driving cars. But I want my car under my control, even if I'm not literally at the wheel.
I'd take issue with your definition of "easy". Even if we had self-driving cars today, it would still be a matter of a phone call to get major traffic arteries closed, whereas there's no guarantee that any group of individuals could work out a way to turn off the cars. You're comparing a known, implemented process with a hypothetical possibility. Similarly, terrorists _could_ conceivably hack their way into a traffic control system, but it's a lot more likely they'd simply try to take over a plane with brute force.
This is similar to "tin hat" theories: they favor the dramatic and insidious, but unlikely, fear over the much more imminent and mundane problem.
I'm more than willing to give up the right to absolute control over my vehicle in return for better safety and faster commutes. The lives saved in both time and accidents would far outweigh the remote possibility that it would be used as an instrument of oppression. Maybe folks can control their own vehicle when off of major arteries, but be required to give up control on heavily shared routes (the same as they have to when taking public transportation.)
"Disclosed are methods and devices for transitioning a mixed-mode autonomous vehicle from a human driven mode to an autonomously driven mode. Transitioning may include stopping a vehicle on a predefined landing strip and detecting a reference indicator. Based on the reference indicator, the vehicle may be able to know its exact position. Additionally, the vehicle may use the reference indictor to obtain an autonomous vehicle instruction via a URL. After the vehicle knows its precise location and has an autonomous vehicle instruction, it can operate in autonomous mode."
So it does not patent "driverless vehicle", but a particular method of transition from mixed-mode to driverless mode.
"Driverless Vechicle" is a compound adjective modifying patent. It's not a patent for (all) driverless vechicles, it's a patent regarding driverless vehicles.
Car models aren't like software—people aren't constantly producing new, accidentally-infringing works. I'm pretty sure this is one of those patents that will get licensed by all the other big companies, rather than sued over.
Well, there are clear prior art examples for driverless cars, but that's beside the point. The real issue here is that a driverless car manufacturer is going to absorb all liability for all accidents while with a driver, the driver absorbs that liability.
I seriously doubt Google will monetize this patent before it expires.
In Germany you are always liable for all damage your car does, no matter whether it is caused by your own mistakes or because of a technical failure. This might not seem fair but the second piece to this puzzle is that you are also forced to have insurance.
Making it clear and unmistakable who is liable makes sure that it is always clear who has to pay for the damage done by cars. The insurance makes sure that whoever is liable can also actually pay. That way the victims don’t get left with the costs (no matter how much money the liable person has) and for the liable person accidents don’t turn into life-destroying events.
This is a working system and I see no reason why it wouldn’t work with driverless cars.
(Even if you want to change it and switch liability from the owner to the manufacturer, that’s not really an insurmountable problem. The solution is – as always – called insurance. Google would just have to insure every car they sell. That’s no big deal, seeing as cars are already pretty much all insured. The cars would get more expensive but the cost of ownership would decrease since it wouldn’t be necessary any longer to pay for insurance. Such a change would be pretty pointless, though. It would only serve to switch around who pays. In the end, though, the owner still has to bear the costs.)
I like the clarity you describe in the German system, but the introduction of autonomous-mode-enabling infrastructure adds a point of "technical failure" that is way outside the vehicle owner's maintenance responsibility.
The concept of "damage my car does" becomes murky if an infrastructure-level failure causes my vehicle to come in contact with another, especially if both vehicles are provably-in-autonomous-mode at the time of the accident.
Which isn't to say that owner-pays wouldn't work, but we do want the system provider to have skin in the game, and such a policy could dramatically slow adoption of the technology by anyone who doesn't also want to assume liability for something they probably just massively funded through taxes (for example).
A solvable problem, in any case, and it's great that the tech is getting to the point where details like insurance effects are worth working on.
I think with autonomous cars it’s even easier to just split liability 50:50 (or any other way) when there is doubt about the cause. That already happens now, and if there are no egos but only machines involved I’m guessing it’s even easier.
I do agree that manufacturers have to have skin in the game, but that doesn’t necessarily have to happen through liability (though that’s certainly a possibility). I would imagine that there will, for example, be very tight and thorough certification of any new autonomous car.
This is in any way a very interesting topic to think about and I think that all the legal hurdles are very solvable once the technology is solid.
(By the way, I only added the remark about the German system because it’s the only one I know. I was under the impression that there are similar systems in place elsewhere, I’m just not sure about that.)
I agree the manufacturer would not necessarily take on the bulk of the risk. However, there is supposed to be less risk once this technology is perfected. If the technology is indeed superior to human drivers, auto insurers will offer discounts to self-driving car owners. Furthermore, autonomous cars make more sense economically if they are part of a fleet service like a robotic taxi service. The future subscription-based Zipcar or ad-based Google Cars service would carry the insurance. The human is just a passenger no more responsible for an accident than a human-driven-taxi passenger.
But in the US, one is not held to be responsible just because something bad happens. It has to be negligence at least.
So the idea here is that if the car crashes, people find out that an algorithm didn't handle a situation very well, and they sue the manufacturer over negligence regarding product safety.
I don't think that was mis-parsed - I think that JabavuAdams is (rightfully ?) cynical that the insurance companies will lower prices.
Rather, they will make the current price the standard for (self-driving cars), and increase the price for (human-driven cars), thereby ensuring more money for themselves.
Do you really think there will be commercial driverless cars on the road in a volume that is financially meaningful to Google before this patent expires?
A lot of things have to happen first - and the liability issue is just one of them. The auto insurance markets and laws have to adapt and change, and who knows who will object to this. Consumers have rethink the very meaning of car. Safety needs to be worked out. Someone has to pay for all this. This is decades away at best.
If/when this happens, it's more likely that Google doesn't exist than Google profiting from driverless cars.
I think that there will be a significant number of driverless cars operating on the road by 2020. I would be very surprised if they're banned, but only mildly surprised if by 2020 there's a mandate or serious talk of a mandate for driverless technology on all cars sold, like emissions legislation now.
Is it just a hunch you have? Why specifically 2020? No snark, just wondering if you've read something that indicates they're coming sooner than later. I mean, I want my driverless car in my lifetime :)
The technology is already available, and merely allowing it would have enormous savings in time and money for virtually everyone who isn't being paid to drive. Further, the savings can be captured by the person making the decision to put it on, so as soon as it's allowed, I expect to see widespread adoption, retrofitting, etc.
Again, as soon as the statistics come in about accidents with driverless systems, I expect a push for mandating that all ordinary vehicles have them, because putting them on every car would likely save almost 30K lives a year in the US alone: http://www-fars.nhtsa.dot.gov/Main/index.aspx . Interesting sidenote: why did vehicular deaths decline by 25% between 2007 and 2009, without any real decrease in miles traveled? Odd.
Interesting sidenote: why did vehicular deaths decline by 25% between 2007 and 2009, without any real decrease in miles traveled? Odd.
My hypotheses:
Older, less-safe cars are breaking down and getting off the road. Newer cars are better able to handle higher speeds and collision avoidance maneuvers. Safety features previously found only in luxury cars are becoming available in lower-end models.
Many roads were widened and/or improved as part of the economic stimulus. More lanes and wider shoulders means more room to dodge a potential accident.
Maybe the weather has been more mild on particularly dangerous roads.
I think it is almost entirely that cars are more safe nowadays than they used to be. And I expect it has less to do with "collision avoidance maneuvers", and more to do with better design and crumple zones, and a significant focus on crash test ratings. There are probably as many accidents as ever, it's just that less of them are fatal.
But my original question was (expanded for clarity): "Why was there a sharp drop from 2007 to 2009 after many years of very gradual decrease or even flatness? Your answer could explain a steady drop as safety features become more widespread, but that didn't happen from 1995 to 2005, even though presumably people bought new cars during this time.
consumer cars being sold today already partially drive themselves (can stop for you, swerve to avoid debris, parallel park). Autonomous cars have been under test for decades and google has already solved the problem of creating a car that can drive itself safely in traffic.
Everything I've been hearing indicates they are coming sooner than later, and that it's a revolution not many people are that aware of.
The Wright Brothers first powered flight was in 1903. The first commercial flight according to Wikipedia was in 1914 and the first Commercial airline company starting in 1919. I would argue that the rate of technical change has quickened since then. It is true that the rate of legislative change seems slower today, but I think that technical triumphs have a way of creating massive displacements almost overnight. Btw the internet only started seeing serious use in 1995. There were questions about anonymity, privacy, fair use, but somehow they have sorted themselves out.
1. Google is NEVER going into the auto business, but they'll licence or even donate their software to run on any auto manufacturers car.
2. The auto manufacturers will only tepidly roll this out very with tons of regulatory and legal tiptoeing.
3. The end user will have to sign away any legal rights to recourse and/or the government(s) will have to amend statutory rights accordingly before the first cars hit the road.
4. It will happen, eventually. It will save hundreds of thousands of lives/year.
they'll licence or even donate their software to run on any auto manufacturers car
Of course, as you're riding down the road in your Google-powered driverless car, there will be ads on the dashboard monitor, paid for by the businesses you pass by on your route.
3. The end user will have to sign away any legal rights to recourse and/or the government(s) will have to amend statutory rights accordingly before the first cars hit the road.
Since the end user will probably not be the plaintiff "The care you made owned by my neighbor ran over my son!" it will have to be a statutory fix IMO.
Question: How do you get "hundreds of thousands of lives per year?" How much safer do you expect these to be and where do you expect them to be driven?
"The real issue here is that a driverless car manufacturer is going to absorb all liability for all accidents while with a driver, the driver absorbs that liability."
True. I think the point, however, is that the enormous amounts of money at stake will ensure that this decision process will take quite some time to work out.
Not all driverless vehicles will carry passengers. e.g., I can see this patent being applied to automated forklifts in a warehouse, robotic supply carriers in hospitals (a la HelpMate robots), etc.
If they don't patent it, somebody else would and would drag them to court. (gosh, I'm defending a multinational - I really hope they don't misbehave with this)
This may indicate the Google has found 100% autonomous operation is not practical yet and so they spent time working on how to easily transition in/out of that mode. I could see a situation where you have a toll-both type area before entering highways that would trigger your vehicle's autonomous mode and another one to switch back to manual as you exit.
Practical or not, I think the first jurisdictions to permit robot drivers will limit where they are permitted[1]. This would work well for that, though realistically, I think an online map of allowed regions and a GPS will be the real solution unless regular drivers want to be informed that others are robot driven, then it will be road side signs.
EOM
[1] I drove south through Nevada at 70+mph on a ruler straight road for 90 minutes without encountering another vehicle, human, or cow. Start there, expand as experience with the systems is gained.
While I am super excited for the future of auto-driving auto-motives, does this patent really prove anything novel and non obvious? QR codes are already used for orientation by many systems, and it seems quite obvious that an autonomous car would need orientation sensors and a computer with at least one line of code relating the operation of an autonomous vehicle. I think google has made some real patent-worthy innovations with their driver-less vehicles but this seems to be of the overly broad progress stifling type of patent.
I noticed that too. I was actually patenting after I published my post on this earlier today because I thought such a quick turnaround was impossible and the patent was not in fact granted. But it was! Really mind-blowing considering we've seen some take four or five years.
I was surprised that Sebastian Thrun is not on the inventors' list. At least two CMU robotics types are among those on the list. Perhaps it speaks to the depth of Google's robotics expertise?
When those cars make into our roads, I expect them to have way less accidents than us, humans. Therefore, more and more the human driving insurance will start to spike to levels that make it not cheap to drive your own car.
I'm not sure about the Grand Challenge, but a fair amount of work for DARPA involves the performers (companies/organizations who did the work) walking away from the program with full IP rights to the result with the ability to patent it or do whatever ... as long as the US government retains I think free license to at least that product/whatever and also maybe free license to anything produced by it? I think that might vary program by program.
DARPA is a "stimulating" organization. giving it patents would not fit into its mission.
When the issue is music or software people pirate, patents are scorned and declared the enemy of progress. When the issue is driverless cars, the attitudes seem much more relaxed.
Agreed. The entire patent departments of all world governments should be abolished. There is no reason the human race should be putting these restrictions on ourselves. It's completely absurd.
Seriously, though, this is a complete waste of time. Mass transit has been solving the problem of getting people from point A to point B without them having to do the piloting for a hundred years. What we really need to be spending our money on is rebuilding the streetcar and intercity rail systems that used to connect nearly every town in the country and electrifying them. This will drastically reduce our reliance on fossil fuels (because oil is the only portable energy source cheap and powerful enough for automobiles) and reducing greenhouse gasses. Instead, we're talking about an overly complex solution involving creation of an automated driving system on all our highways and compelling thousands of users to upgrade to driverless cars, which may or may not be electric.
In other words, Google's driverless car is to transportation as Dart is to DOM scripting.
So, I've been a life-long rider of mass-transit in Toronto, and Southern Ontario. I think I was 30 before I actually had to drive to a job.
You know what -- I hate transit during rush hour. It's uncomfortable, you don't get a seat. Ugh. During the off-times, it's great.
Traffic sucks, but at least you've got a little bubble of space over which you have control. This is psychologically significant.
Think of the difference between having a desk whose layout (or lack thereof) you control, versus working elbow-to-elbow at a long table with other people's jostling you and their crap falling in your space.
People need a certain amount of control over their space.
I get the impression you think that electricity is automatically clean energy. Have you ever thought about how that electricity is generated? Ever heard about coal or nuclear power plants etc.? Do you think they're are environmentally friendly?
Yes, I realize that most of our electricity comes from coal, natural gas, and nuclear. However, that fact doesn't preclude the idea of using solar, wind, tidal, and geothermal energy to generate all electricity. It's much more difficult to use these energy sources to power automobiles than it is for trains.
Actually, the QR-code (or equivalent) is used to fetch "instructions for performing the autonomous vehicle instruction" (which could be anything).
And the QR-code can be replaced by a lot of things (radio signal, etc).