Illegal for "Human" to Drive

[Hoosier Daddy]

Even if it doubles every 5 years, averaged out, instead that would still be 64 times the computing power we have today in 30 years. We might be flying cars without drivers by then instead of just having them just drive on the roads for us...

[jvincent]

Assuming Moore's law holds up for the next 30 years, and it may not since semiconductor geometries are approaching the limits of manufacturing, computing power is not the problem.

More compute power does not solve the complexity issue. The ability of the human brain to process the number of inputs required to drive safely is complex on a scale that today's computers can't begin to comprehend.

I'm ignoring here the need for positioning systems that are 99.99999999% reliable. The current GPS system is very far from that.

And don't forget the legal liability issue. More than any technical reason this is going to stop things.

[fang_gt86]

Originally Posted by Tcoat

In case you didn't notice, George is actually flying this thing. He's holding the steering wheel!

[Tcoat]

Originally Posted by Hoosier Daddy

Just compare the electronic systems that drive cars today compared to how cars were driven 30 years ago. In 1985 any type of electronics wasn't widely used.

What would Moore's Law say about this? That computing power will change by the same amount in then next 3 years as all the increase from 1985 up to 2015! By 2045 we'll have 1000 times more computing power than we have today!

1985 cars were loaded with electronics compared to some I had!
Some of my cars still had vacuum tubes in the stock radio so forget about any electronic engine controls.

[Hoosier Daddy]

If Russian engineers could control naturally unstable airframes in the 1980's then commercially viable driving without a driver won't be that much of a stretch for the human brain to reach...

[Tcoat]

Originally Posted by fang_gt86

In case you didn't notice, George is actually flying this thing. He's holding the steering wheel!

Well at least he isn't texting!

[gramicci101]

Originally Posted by Hoosier Daddy

If Russian engineers could control naturally unstable airframes in the 1980's then commercially viable driving without a driver won't be that much of a stretch for the human brain to reach...

It's not a big deal if an aircraft strays 30 feet to the side of its expected path though. It's kind of a problem if the cement truck next to you does the same.


I think instead of each car being completely autonomous in its operation, places would need a grid of traffic management systems, and cars would be managed by the system responsible for the specific area they're in. The car transmits its origin and destination, and the manager slots it in with the other traffic it's managing and routes it either to the next management system or to its destination. Major highways and interstates would have their own dedicated systems.


That way, if there's a single computer controlling everything in a 5x5 mile square and every car in that square is tied in, there would be no accidents and traffic would proceed in an orderly fashion to its various destinations. Rendering manual driving a thing people only do in sanctioned events or the middle of nowhere. Like Wyoming.

[fang_gt86]

Originally Posted by gramicci101

That way, if there's a single computer controlling everything in a 5x5 mile square and every car in that square is tied in, there would be no accidents and traffic would proceed in an orderly fashion to its various destinations. Rendering manual driving a thing people only do in sanctioned events or the middle of nowhere. Like Wyoming.

I can see city like Dubai implement a small scale of this. It'll make all other cities look like they're running DOS.

[Tcoat]

Originally Posted by fang_gt86

I can see city like Dubai implement a small scale of this. It'll make all other cities look like they're running DOS.

Ya! A lot of the vehicle tech already exists and it is the supporting infrastructure that is lacking (and very, very costly) to put a program in place. Someplace like Dubai with a relatively small land mass and relatively large economic base could very well be the ideal location to start the process.

Good thinking!

[Hoosier Daddy]

Originally Posted by gramicci101

It's not a big deal if an aircraft strays 30 feet to the side of its expected path though. It's kind of a problem if the cement truck next to you does the same.

Of course that's true, but if they needed it to be they, in all likelihood, could have done it.

I mean, when I was in University one of the senior projects was an on-going, industry funded, autonomous mining vehicle design. It was fairly sophisticated, lead by a Prof. who was a former NASA controls expert, which started over 15 years ago. Accuracy was within 1 meter, and that's a mile underground! If a pack of engineering kids can do something satisfactory for mining exploration in the early 2000s the future holds little bounds. I have no doubt autonomous cars will happen. But I guess we'll see... Or at least for those that don't kick the bucket first!

Oh, back then my school was working on another well funded project... At that time folks thought it was a real joke. Pilot-less airplanes or Unmanned Aerial Vehicles (UAVs). Or drones. Those won't work they said. Nobody will use those for any reason they said.

They looked suspiciously like the MQ-9 Reaper...

[MokSpeed]

Originally Posted by Eski

however if the world goes the way the movie "idiocracy" portrays it, were all doomed.

"...but it's got electrolytes!"

[rice_classic]

At this point.. I'd settle for proper through lanes.

Like the far left lane that actually has small divider between it and the rest of the freeway so you can't hop in and out at will. Once you're in it.. YOU'RE IN IT. That way someone who lives north of Seattle that wants to drive south of Seattle, isn't impacted by all the on/off ramps and merging that actually creates traffic. The points of entering or exiting the lane don't happen near choke points and only come up about every 20 or so miles. I think a lane like this would do more for reducing traffic than any of the HOV lanes.

"Are you getting on the freeway for the next 50 miles? Well then this lane is for you!" Yeah, we need that.

[stugray]

Originally Posted by jvincent

Assuming Moore's law holds up for the next 30 years, and it may not since semiconductor geometries are approaching the limits of manufacturing, computing power is not the problem.

More compute power does not solve the complexity issue. The ability of the human brain to process the number of inputs required to drive safely is complex on a scale that today's computers can't begin to comprehend.

I'm ignoring here the need for positioning systems that are 99.99999999% reliable. The current GPS system is very far from that.

And don't forget the legal liability issue. More than any technical reason this is going to stop things.

You re 100% correct. We have NOW the computing AND imaging capability to do this right now. It is not a matter of computing power, or software AI, or imaging capability. It is purely a legal & liability issue.

These systems wont be error free until we have years of real-time testing under realistic situations. BUT the car manufacturers wont want to be the first to enable these features because of the risk of litigation.
So it is a chicken & egg issue.

Just a couple of years ago my company flew an imaging system on the last shuttle flight that imaged the ISS with laser Time-of-flight spectrometry that gave a 3D image of the ISS to sub-millimeter resolution as the shuttle docked with the station.

That cost millions of dollars and many years of development to build.

NOW you can go buy one for $250 with the new Kinetic unit for an XBOX One.

Combine a couple of those and a NVIDIA Tegra Drive computing unit with 12X video inputs, and you have an autodrive hardware solution for less than $1000.

So there is no technical reason to not have this technology available now.

In fact, Tesla just announced that they are rolling out an auto-steer function to it's cars as we speak. My guess is that this technology will be rolled out incrementally and more and more features will be available as the car manufacturers get more & more time under realistic driving situations.