Interesting perspective on autonomous freight. Even if self-driving trucks become technologically viable in the coming years, their economic viability may still significantly trail that of personal autonomous vehicles.
Or imagine robotic trucks full of sweet crude cruising down from Alberta.
It's unlikely that robotic trucks can compete with Buffet's railroads. (According to industry definitions, Albert's crude is not "sweet".)
Then again, railroads can't compete with pipelines when it comes to transporting oil.
Good thing we stopped that.
I do think, having a human in the loop who's life is on the line in any accident, will tend to be more creative in avoiding one than software would be.
Training an AI to avoid simulated accidents is not a complete solution. For one thing, there will always be novel situations not trained for, and for another, the AI has no visceral fear of its own death.
Well… yeah, I guess, but it’s not really relevant. We don’t need creativity in avoiding accidents; the solution space is small: accelerate, brake, and/or steer.
The machine knows the velocity vectors, accelerations, and sizes of all other objects on the road. That alone is a huge advantage over a human, which don’t look in all directions at once. You can probably drop the AI part altogether and have an always-running routine that takes over when collision is imminent based on relatively simple vector projection.
Avoiding collision isn’t the hard part here, adhering to the myriad rules (spoken and unspoken) we’ve set up is. I think that’s probably why we all cringe at the Tesla videos of it doing crazy stuff, but somehow it’s not hitting stuff very often.
That is the problem. 99% of the time not hitting stuff is 100% useless
99% is certainly not enough. But at some point Tesla, or another company, will get enough 9’s to surpass humans. It will happen.
Umm, yeah…
Last week the collision avoidance on the truck decided a traffic sign was enough to sound the alarm and hit the brakes while going around a curve on a wet road in sub-freezing weather. Let’s just say that the pucker factor on that one was off the scale.
Actually, the truck was fairly intent on killing me for that leg of the trip since there were a few winding downhill sections, I was at maximum weight and the stupid robo-transmission would randomly downshift instead of just easing down the hills without using the brakes — or randomly upshift into redline which will cause all the important bits inside the engine to come flying through the hood if you don’t catch it immediately.
The thing that really gets my goat is the truck has the ability to be put into manual shift mode (specifically so it doesn’t do something stupid and kills a bus full of kids) but the company has it disabled for fuel economy reasons.
I mean, “move fast and break things” doesn’t really work so well when you’re taking 80,000lbs down an 8% grade and this freightliner is a buggy POS on the best of days.
The thing is, self-driving trucks seem more plausible as something technologically viable compared to self-driving passenger cars. I've heard proposals to have loading docks directly off major highways, so the only driving the things would need to do would on highways (and even in special lanes). At the same time, I wonder what the advantage of this over just trains would be.
It is all the other little hassles and problem solving that you need the driver for.
Proper pay for everybody is the obvious solution. Are markets going to get us there? If so, why not now?