Impact of automated driving on Trucking industry

Not only that, we're talking take-off and landing as well. These kinds of tasks can't be programmed, and require a human to control the vehicle. I've done real-time simulation systems in the past, and the kind of computational power necessary to do just that requires what's known as a super-mini computer. One stage below the kind of computational power necessary to be called a super-computer. Like a Cray.

Let's talk about computer vision. It's capable now of distinguishing things like bolts on a tray in a production-line environment... but it has to be a controlled environment and expected objects. A few years ago some of my friends were working on autonomous vehicles... the still had enough problems with it that distinguishing between a shadow and something like a step was a 50-50 proposition. That alone will take an order-of magnitude improvement on both the software and hardware side...

This ain't happening any time soon... certainly not on the roads shared with private vehicles.

 

Agree with you. At no point did I say this stuff was currently implementable, there are some serious hurdles, one of which being inclement weather.
Also, i would not start this technology with trucks. that would be insane. proof of concept, car on closed course, then once its bugged out find a small rural area that will allow implementation given success on closed course. next move to a municipality, then maybe a county, then just keep expanding.

just saying its not so far fetched.

EDIT and yes just to add to this I would rather have my family next to a PROVEN out computer controlled vehicle (not until it is proven out though) than an ignorant, arrogant, bored, tired, sick, distracted, impaired, overconfident, underconfident, (pick one or more) driver.

 

the thing doesn't need to read english. ever seen bar codes, or those square dealy things? Those codes can incorporate a good deal of info. add them onto signs with the English that we as humans need to read. The signs are there to tell us information, no reason it cant tell a computer information also.

Also there would have to be many layers of redundancy. signs, RFID, and whatever else could be used. the department of redundancy department would definitely have to be involved.

 

So you're telling me we're going to start implanting RIFD tags in little kids chasing balls out into the middle of the street in low light conditions when they are wearing dark clothes? That's almost invisible to a human eye, and the kind of non-regular object that machine vision not only has to detect, but identify AND determine whether evasive action is required, while determining what kind of evasive action to take. That is difficult for a human, and beyond the capabilities of current computation platforms in real-time systems. Plan on carrying around a trailer-full of computers... not that kind that sits on a desk top, but inhabits a computer room. That is something I know about, because I used to be in the middle of development of that kind of system.

 

of course you dont RFID tag everything. But motion sensors can determine something has entered its path, radar can determine speed and distance. and more quickly than you can determine to brake or swerve. Also the computer wont be fiddling with the radio or its smartphone and just plow over that kid. At a minimum it reduce the severity of the accident. And a good camera which would be needed can have better low light vision than you or i do.

 

You're oversimplifying the problem... it's not just detecting motion. It's detecting the object, AND identifying it. As it stands now, computer vision systems have 50-50 probability of deciding whether an "object" in its path is a shadow or a real object.

You think crash-avoidance systems are a pain now? How about when you go into a full-emergency brake application for the shadow of a cloud. Or worse... decide the kid is a shadow, and plow into him. That is the current state of the art... and from what I see, that won't change any time soon.

 

look, clearly I don't have all the answers, otherwise I wouldn't be driving a truck. I completely understand the complexities of a system like this, all I'm saying is it isn't that far fetched. It's going to take clever software, and fast computers. There will always be bad instances. bad things happen, but clearly we have proven that as a society we don't care that we have millions of motor vehicle accidents and Tens of thousands of motor vehicle fatalities every single year. We just keep on doing what we do. Hell its 2013 and we still cant get people to turn on their #### lights in fog. If we can figure out a way to give computers a chance I'm all for it.

 

Rail is just way too slow, its cheaper but slower,why do you think they still use trucks? do you want your fruit on a train for 2-3 weeks to cross the country

 

Rail delivery windows are +48 hours, -2 weeks. Real timely there.