Dont worry Raza, we have a long ways to go before full automation

[12,000 RPM]

http://jalopnik.com/this-is-how-bad-self-driving-cars-suck-in-the-rain-1666268433

Video is too long and not worth watching but the long and short of it is fully automated cars + rain = no bueno

Also shit like construction etc flusters it

Driving automation will continue to come in the form of added convenience. Stop n go traffic cruise control and shit like that.

[68_427]

There was also a Korean competition where ten feet of fog like conditions sent a few cars into a ditch.

[Raza]

[12,000 RPM]

There was also a Korean competition where ten feet of fog like conditions sent a few cars into a ditch.

I think this was the same competition. Link is about some competition in Korea.

Few more decades to enjoy the thrill of full control stop n go traffic in the rain.

[Galaxy]

A couple of years old but...

https://www.youtube.com/watch?v=BYY7OfQ4-5A

[GoCougs]

Self-driving cars will never happen. To again whip out my big fat engineering experience which I know you guys just love, I've designed and programmed plenty of machine vision and sensing systems and robots in general (and what I've be doing for the last ~6 months).  It's not a matter of processing power or resolution or imaging or software - it's a matter of what to look for and on your average roadway the process window is simply far too large - there's no conceivable way to capture all (or even most) scenarios in order to model/program the system such as to make it reliable. The ONLY way this ever becomes remotely possible is on a highly controlled roadway - in effect a tunnel with specific features, but then you're not far from some sort electrified roadway, which I've been saying all along is the only solution to this and transit in general. Sadly, this was solved ~100 years ago with electric buses, trollies and trains but that's too boring to ever get any press, let alone government $$$.

[AutobahnSHO]

I don't agree with your assessment.

Engineers in the 1960s could never have fathomed walking around with a portable 10MB/sec window into the network of all human knowledge.

[SVT666]

Cougs, you're wrong.  It will happen.  Technology is advancing at such a rapid rate that I think there will be self driving cars available to purchase in less than 20 years.

[Soup DeVille]

Cougs is right about the current obstacles. He's probably right about the nature of the first consumer available automaton cars.

I would not even hazard a guess as to how soon the technical problems will be solved; but it will be at least 20 years before the legal and public perception problems are.

[Morris Minor]

Automated cars will not have much smarts on board - they'll have the packet headers with routing info for the road systems that control them.

[AutobahnSHO]

Automated cars will not have much smarts on board - they'll have the packet headers with routing info for the road systems that control them.

aha!  See, great idea there.

[GoCougs]

The challenge isn't "smarts" or communicating information. The challenge is in real-time processing of a nearly infinitely variable environment. Validating a process is the cycle of testing it to failure, incorporating fixes, and wash/rinse/repeat until the # of failures reaches a predetermined minimum. If the minimum can't be reached a common approach is to revisit the premise by limiting the variation. Thing is with driving a car there is near infinite variation in the environment, esp. that is is in 3D, ergo, the environment has to be limited if the process is to be validated (to be usable and safe).

Again, automated vehicles much like many other tech fads (3D printing, mag lev, robots, etc.) was solved by the manufacturing industry many years ago. Note how the environment is controlled - there are tracks in the floor, every vehicle has the same OS and real-time detection/info of what the others are doing, and there are no people around (and when anyone/things gets remotely close it brings the entire system to a halt). Wash/rinse/repeat for self-driving cars - they'd need the same highly controlled (closed) environment to be even remotely feasible. Since this is prohibitively expensive it will never happen.

https://www.youtube.com/watch?v=QKPIOny1AL4

[Morris Minor]

The challenge isn't "smarts" or communicating information. The challenge is in real-time processing of a nearly infinitely variable environment. Validating a process is the cycle of testing it to failure, incorporating fixes, and wash/rinse/repeat until the # of failures reaches a predetermined minimum. If the minimum can't be reached a common approach is to revisit the premise by limiting the variation. Thing is with driving a car there is near infinite variation in the environment, esp. that is is in 3D, ergo, the environment has to be limited if the process is to be validated (to be usable and safe).

Again, automated vehicles much like many other tech fads (3D printing, mag lev, robots, etc.) was solved by the manufacturing industry many years ago. Note how the environment is controlled - there are tracks in the floor, every vehicle has the same OS and real-time detection/info of what the others are doing, and there are no people around (and when anyone/things gets remotely close it brings the entire system to a halt). Wash/rinse/repeat for self-driving cars - they'd need the same highly controlled (closed) environment to be even remotely feasible. Since this is prohibitively expensive it will never happen.

https://www.youtube.com/watch?v=QKPIOny1AL4

I can see it happening on controlled access highways - at least to start with. So you would drive to the freeway & the car would latch itself to a virtual train to be taken to its destination.

Long term I think the whole Google-type project, with self-driving "smart" cars festooned with sensors trying to avoid random collisions in uncontrolled environments, will be a dead-end approach...  chimpanzees navigating minefields.

[AltinD]

Automakers say: "It's not the technical aspect, it's the legal aspect that at the moment keeps automated driving off the road"

[12,000 RPM]

I can see it happening on controlled access highways - at least to start with. So you would drive to the freeway & the car would latch itself to a virtual train to be taken to its destination.

Long term I think the whole Google-type project, with self-driving "smart" cars festooned with sensors trying to avoid random collisions in uncontrolled environments, will be a dead-end approach...  chimpanzees navigating minefields.

We barely maintain highways as is.... where would the money for highway tracks come from

Naw, the future in driving "automation" is definitely in increased driver assists. But the driver will remain part of the equation over our lifetimes. I mean even in aviation, there is a LOT of automation, to the point that some planes are inherently unstable without computer assistance. But at the end of the day, even with autopilot, the pilot is still necessary and in control. Driving will become easier, but not unnecessary.

[565]

Sadly still better than 90% of Boston drivers in the rain. 

[MexicoCityM3]

Automated multi-storey parking coming soon from your favourite automaker:

http://f30.bimmerpost.com/forums/showthread.php?t=1069610

"Fully automated parking in multi-storey car parks – dynamic and safe even without the driver. The fully automated Remote Valet Parking Assistant in the BMW i3 research vehicle combines information from the laser scanners with the digital site plan of a building, for example a multi-storey car park. If the driver uses the Smartwatch to activate the fully-automated Remote Valet Parking Assistant, the system will steer the vehicle independently through the levels, while the driver has already got out of the car and is on his way to a business appointment. The fully automated Remote Valet Parking Assistant recognises the structural features of the car park and equally reliably steers round any obstacles that appear unexpectedly – such as incorrectly parked vehicles. Once the BMW i3 has arrived at the parking space, the vehicle locks itself and waits to be called by Smartwatch and voice command. The fully automated Remote Valet Parking Assistant then calculates the exact time until the driver arrives at the car park and starts up the BMW i3 so that it arrives at the car park exit at exactly the right time.
Navigation without GPS signals.

BMW has succeeded in achieving fully automated control of the vehicle by connecting up vehicle sensor systems and a digital site plan. This avoids dependence on the GPS signal, which is not at all precise in multi-storey car parks. Alongside the laser sensors, the research vehicle also has the processing units and necessary algorithms on board and this means it can determine its exact position in the car park, monitor the environment perfectly, and carry out independent and fully automated navigation. It is not necessary to provide car parks, for example, with complex infrastructure in order to allow cars to orientate and navigate around the area safely."

[Soup DeVille]

Automakers say: "It's not the technical aspect, it's the legal aspect that at the moment keeps automated driving off the road"

I think that's partially true: So far as we basically know how to work towardsa the solutions to the technical problems; but nobody can lay out a roadmap to solving the legal issues.

Even when automated driving becomes safer than manual driving, the liability issue is huge. People may make more mistakes than machines: but when they do, its pretty easy to say "the driver is at fault." Who do we blame when an automated car crashes?

[MexicoCityM3]

I think that's partially true: So far as we basically know how to work towardsa the solutions to the technical problems; but nobody can lay out a roadmap to solving the legal issues.

Even when automated driving becomes safer than manual driving, the liability issue is huge. People may make more mistakes than machines: but when they do, its pretty easy to say "the driver is at fault." Who do we blame when an automated car crashes?

Laconian and his Google overlords.

[Raza]

I think that's partially true: So far as we basically know how to work towardsa the solutions to the technical problems; but nobody can lay out a roadmap to solving the legal issues.

Even when automated driving becomes safer than manual driving, the liability issue is huge. People may make more mistakes than machines: but when they do, its pretty easy to say "the driver is at fault." Who do we blame when an automated car crashes?

Oh, the driver will still be at fault, I'm sure.  Everyone will likely have to sign contracts that disclaim liability in case of an accident that is not covered under an express warranty and it'll be upheld in court, since corporations win everything all the time.

[Tave]

Pfffft! If Google's car wasn't involved, and it doesn't appear to be, then why should any of us care?

That a bunch of undercapitalized, start-up, R&D units have put out a product with a bunch of glitches should come as a surprise to nobody.

*Gasp!* You mean to tell me that KAIST Unmanned System Research Group's entry wasn't up to snuff?!?! The horror! The horror!

[AltinD]

Pfffft! If Google's car wasn't involved, and it doesn't appear to be, then why should any of us care?

That a bunch of undercapitalized, start-up, R&D units have put out a product with a bunch of glitches should come as a surprise to nobody.

*Gasp!* You mean to tell me that KAIST Unmanned System Research Group's entry wasn't up to snuff?!?! The horror! The horror!

I don't really get the rant, but the pioneers of automation driving are not start-ups or even Google, rather the automakers themselves

[Tave]

I don't really get the rant, but the pioneers of automation driving are not start-ups or even Google, rather the automakers themselves

Nah Google is head and shoulders above the industry at the moment and leading the charge quite well.

Notice Google was nowhere to be seen at this South Korean test as it's been busy expanding into additional States, driving on actual roads, and releasing its newest prototype.

[Morris Minor]

Germany to test self-driving cars on digitized autobahn, 'won't rely on Google'

The German government wants to convert part of the A9 Autobahn in Bavaria into a test-field for advanced car technology. The project is key to ensuring the country's 'digital sovereignty,' according to its transport minister.

The track, part of the "Digitales Testfeld Autobahn" project, would be launched this year, Alexander Dobrindt said on Monday in an interview with the Frankfurter Allgemeine Zeitung newspaper. The plan involves equipping the road with infrastructure to allow cars to communicate with each other and the road's own sensors to provide necessary data on traffic.

More:
http://rt.com/news/226279-robotic-car-autobahn-germany/