"In one scenario, it appears, a driver is looking for a parking space and drives over a giant QR code on a street or in a parking garage. The QR code might either have exact directions to a specific parking spot, or it might link to a parking garage database of which parking slots are open."
Surely the technical part is no more open to abuse than normal parking signs in cities listing the amount of available spaces in certain car parks.
The QR codes would contain very little information I would imagine but rather a key that is used by the car as an argument for further operations. Leaving all the directional data on the road would be way too easy to abuse.
"it might link to a parking garage database of which parking slots are open"
Imagine creating a QR code, (or copying one that already exists) which tells the car to fetch information from http://your.competitor.example.com/. Then imagine going out late at night and sticking it on a busy motorway.
DOS by traffic.
Also. You better hope that your car doesn't have any vulnerabilities that can be abused by a specially crafted QR code, or by fetching specially crafted data that a QR code prompts it to fetch.
While we're there we better do something about all those people using radio jammers to down auto-landing aircraft.
Also, you better hope their are no vulnerabilities in the automatically driven mono-rail / train system you are riding in.
These type of issues are going to be discussed to the end of the earth but I can't wait to see driverless cars make their way from the lab to the car dealer.
As far as introduction, I would bet these features will be added to high-end luxury models and quite possibly initially limited to main roads (motorways / interstates) where the chances of encountering something unusual are lower. With the addition of a fair splattering of warning labels.
I very, very rarely resort to ad hominem, least of all on HN, but your nickname says it all (if indeed VB refers to Visual Basic). Far too many computer problems of yore were created by VB programmers not being nearly "tin hat" enough. When engineering systems which can damage other systems or cause bodily harm, one's responsibility is to anticipate, mitigate, and prevent the worst case scenario to the maximum extent possible.
Well VBprogrammer is a moniker I've used since I was 14. I now work for a company building websites which have thousands of hits a minute, require PCI compliance and use a variety of tools, none of which are Visual Basic.
Given that you resorted to ad hominem I assume you have no other reasonable arguments against what I said.
It's called speculation. I am fully aware that they will prevent issues like this from happening. It's still interesting to consider them though. In the end, somebody has to take these issues into consideration.
If that's all it takes to DOS your competitor, well, I don't think you have much to worry about from them. Even a busy motorway doesn't have that many cars passing over it per minute.
Also you're going out on a highway, at night, to lay down a big sticker. That's dangerous, suspicious, and illegal (before we even considering what the sticker does). If you want to DOS a competitor there are much more effective and safe ways to do so.
Could you not have the QR codes generated against a private key and the vehicle verify it against the public key to determine whether the instruction came from a trusted source?
Just because it's plain text data in a QR code doesn't mean that it's inherently ripe for abuse.
But if the signs are electronic, as in... providing the location of the nearest empty parking space... then updating them should be effortless.
Lots of assumptions here from everyone about how the signs are implemented and what the QR codes contain, but the general gist is that they could be secured. Assuming otherwise is also an assumption.
Yeah. The obvious first idea is to sign the codes, but then when the private key gets stolen (100% certain to happen sooner or later,) you have to repaint every road.
I suspect this is simply one of many self-driving car patents Google will file, but it's fun to speculate:
Technological hurdles aside, the catch-22 situation is that without much evidence the technology works there will never be public acceptance and legislation for it to be used on the roads. Perhaps Google plan to initially market the system as some sort of beefed-up parking assist - one where you leave your car at the entrance to a car park and collect it from there later. I'm not lawyer, but car parks != public roads, so different laws should apply.
Obviously, nobody would pay loads of extra cash just for that feature, but maybe Google would be willing to subsidise it in the hope that it gets them over the public-acceptance hurdle as people would start getting used to cars driving themselves around without incident and start to trust the technology. It would also build up thousands of hours of evidence of the safety of the system, which the legislators are going to want to see before they okay it for general driving.
Personally, I can't wait until self-driving cars are a reality. It's going to revolutionise transport. :)
Probably autopilots will be sold and marketed as glorified cruise control. To be used only with a driver behind the wheel. And then they will gradually get accepted to the scale that it will be silly not to have a full permit for autonomous vehicles.
Possibly. Though they'd either have to build in something to monitor the driver or accept that they're just going to tune out and read the paper/go to sleep.
Could be a potential way to bring it to market that keeps the insurers happy though. Make the driver 'promise' to pay attention and should an accident happen and the driver didn't take over then the insurer doesn't pay.
This is a patent and not a detailed report on what they intend to use or are implementing.
Strategically speaking the optimal case is a patent to cover the process of sending directions to a car, including everything needed to request them, using any means.
Everyone who produces cars will have to implement some form of that at some point to stay relevant, which makes this a very valuable patent.
Sad that a patent could have so many ways to be implemented. Patents are supposed to be implementable. Google could hide a ton of trade secrets behind this patent.
Can I tell the robot car to run over a pedestrian if he has an ak-47 pointed at me and he is on the curb? A self defense situational awareness algorithm that understands kill or be killed?
This strikes me as ripe for abuse.