Amazon Patent Hints at Self-Driving Car Plans (theguardian.com)
Amazon is working on self-driving cars, according to a new patent that deals with the complex task of navigating reversible lanes. From a report on The Guardian: The patent, filed in November 2015 and granted on Tuesday, covers the problem of how to deal with reversible lanes, which change direction depending on the bulk of the traffic flow. This type of lane is typically used to manage commuter traffic into and out of cities, particularly in the US. Autonomous vehicles, the patent warns, "may not have information about reversible lanes when approaching a portion of a roadway that has reversible lane", leading to a worst-case scenario of them driving headfirst into oncoming traffic. More generally, the inability to plan for reversible lanes means cars and trucks can't optimize their routes by getting into the correct lane well in advance, something that could otherwise prove to be one of the benefits of self-driving cars. Amazon's solution to the problem could have much larger ramifications than simply dealing with highway traffic in large cities. The patent proposes a centralized roadway management system that can communicate with multiple self-driving cars to exchange information and coordinate vehicle movement at a large scale.
centralized roadway management = new car 3-4 years as software updates may stop after 2-3 years and do you really want to pay $2K-5K to update your car's computer at the dealer?
Everybody whose anybody is working on self-driving cars, something that won't pan out for at least a decade due to regulations and liabilities.
The car costs $500, but you'll have to deliver goods for Amazon on your way?
Of course news about a fake are Fake News.
Not actually the cars themselves. Road management system...not car driving system.
Fuck off Amazon.
If all you're doing is replicating the behaviour and mental assessments a human driver would take at an intersection, assessing traffic flow, judging risks etc you don't get to try to patent it and lock others out. 'On a computer via AI speshulsauce' does not render it novel or nonobvious. Humans been doing it since cars were invented.
Amazon sells fucking books. Why is it intefering in the auto industry?
Clearly it hopes to get a royalty check from all the other car manufacturers for its "Amazing Idea". Once again, patents slowing progress to a crawl.
So how are these supposed self driving cars going to handle construction zones where cars get moved into the oncoming lanes?
Central control is the totally wrong way to go, what if communication drops? How will the car move around incidental blockages like blown garbage if it can't figure stuff out for itself.
If you buy an Amazon Car, you get Amazon Prime movies to watch whilst the car drives you around. If you tell it to drive to Walmart, it will take you to one of the new Amazon brick and Mortar stores instead, free of charge.
"That's the way to do it" - Punch
I'd hate to see my Amazon car run into a lorry whilst negotiating a roundabout, or accidentally drive itself off the carriageway. Good heavens.
How about a self-flying car, maybe a cargo-only car? It doesn't even have to operate on roads. Hmm....
Captcha: higher
won't pan out for at least a decade due to regulations and liabilities.
So far the regulations are keeping up with technology. Tesla Autopilot and other driver assistance and self-steering and self-braking is legal in all states. Many states have streamlined approval for SDC testing. By the time SDCs are available to the public, they will be legal in most places.
The liability situation doesn't change much.
Before SDCs: The insurance company pays.
After SDCs: The insurance company pays.
The big difference will be that, instead of the driver buying insurance, the cost of insurance will be built into the price of the car.
Those lanes with the big green arrows and red x's? If you can't read that do how will stoplights work?
How about we just teach the computer to read the sign, all those sensors and we can't OCR one huge sign for: Thru traffic move left
lol, we can't plow the streets or fix potholes here so let's build smart roads and signs cause maintaining those will be different somehow :O Don't worry the lane signals will be fixed later this year, til then watch for oncoming cars aiming for your yellow running lights.
If existing self-driving car technology is unable to cope with the concept of reversible lanes, then it suggests they cannot read road signs, in which case reversible lanes is the least of anyone's worries - road works, temporary diversions, etc, are also going to be a problem.
You are not alone. This is not normal. None of this is normal.
Because someone will insist it gets connected to the public internet to make their personal lives easier.
I wouldn't take this single patent as a sign that Amazon is actively working on self driving cars or anything directly related. Amazon's corporate culture loves patents. They have an internal program to encourage their people to get all their ideas patented, whether it's to do with your current role/project or not. (Amazon gets the patent, you get a cool paperweight for your desk.) Given this fact, combined with the fact that Seattle has reversible lanes on I-5, it would not be surprising that an SDC-enthusiast on their payroll might have come up with a solution to handling reversible lanes. Now, if Amazon has filed dozens of SDC-related patents, that might indicate a active project in this area.
Amazon doesn't want to sell cars to consumers. They do, however, want to make their logistics as efficient as possible. Think autonomous inter and intra city transport.
Autonomous vehicles between logistics hubs; drones for 'last mile' delivery to consumers. No more inconvenient/inefficient drivers.
Another Silicon Valley giant working at self driving cars.
The problem of driving cars is complex, but it is modular (avoid obstacles, visual recognition, ethical decisions) and universal (everybody mostly want the same thing, that is the car not to crash). So, why don't all these companies develop specifics APIs, or modules, and make the core, gluing system open, instead wasting time developing their own distro, *ehm, algorithm. Considering that self-driving cars will likely interact with more and more self-driving cars in the future, this may also improve operability in the future.
PS: I know this is partly what's happening, especially with the visual recognition systems, but I am talking about extending the approach to other functionalities as well.
Over dramatising the worst case.
I doubt any self driving car presented with an oncoming vehicle in what it thinks is its lane would elect to carry on regardless "leading to a worst-case scenario of them driving headfirst into oncoming traffic", rather than switching to another lane or coming to a complete stop. That such a system would be allowed on the road without taking a worst case of only utilising the non-reversible lanes is preposterous.