Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

This Year, Autonomous Trucks Will Take to the Road with No One on Board

spectrum.ieee.org

Post reply on HN
Print view View original

81–90 of 106 posts

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#81

Post by joehartshorn »

ljlolel wrote: post #67
joehartshorn wrote: post #46

Earlier quoted context omitted.

>"Trucks are also a better platform for autonomy, with their large size providing more power for computers" I'm no expert but "power for computers" doesnt strike me as a bottleneck, Geohot/Comma seem to be doing fine, this just sounds like marketing bs

That's right the limitation is Data. You can't effectively use the extra compute if you don't have enough data to train a bigger neural network. https://jperla.medium.com/why-tesla-autopilot-ought-to-be-aw...

Do you actually have data to back that up? I read your article, which (no undue disrespect intended) looked to be full of unsubstantiated qualitative data/graphs and opinion. Are you an "expert"? Or are you trying to get medium clicks on your article/ purchases of your book?

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#82

Post by foobarian »

TameAntelope wrote: post #26
onion2k wrote: post #2

If I block a road an autonomous truck is expecting to drive along, will it automatically reroute to another road? If it does, then these trucks are going to be very easy to rob simply by corralling them in to a quiet location..

But you're still robbing a truck...

Pray tell what are you going to gain by robbing a truck? A bunch of plastic crap you're now going to have to fence? At least in the 80s it would be TVs or cameras you could liquidate on the black market.

A few months ago I listed an old LCD TV on Craigslist. A bunch of offers fussed about coming to get it, and I ended up picking the most polite one and delivered it. I could barely even give it away!

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#83

Post by slfnflctd »

bryanlarsen wrote: post #70
BotJunkie wrote: post #54

Earlier quoted context omitted.

I'm the author of this. It was certainly surprising to learn that a lot of what AVs are doing right now is limited by the amount of power available to the computers. It didn't make it into the article in full, but we were told that in general more computing power means detecting and classifying more object types, tracking those objects over longer periods of time, and doing more detailed motion prediction and plannin…

If so, why not use cars without alternators? 48V mild-hybrids have much higher capacity starter-generators instead of alternators, and electric cars can generally pull power from their batteries at the same rate they can charge it. For example, for a Tesla Model 3 that's 250 kW.

Hybrids are absolutely the most effective way to go here in the near term, there is just no question in my mind. I'm honestly very, very surprised we haven't seen that take off yet-- either it's a lot harder than it looks, or owners of large fleets are irrationally gunshy. I've been expecting it to show up and start kicking ass any minute now for at least the last 7 years.

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#84

Post by bluGill »

jariel wrote: post #9

I cannot believe that this isn't the foundation of autonomous driving: long hauls, highways, no stop signs, semi-controlled circumstances. The economics are more clear and industrialized. Even if a driver had to be present when they got off the highway, I think 'long haul' should be naturally the first thing to be tackled. And I'm surprised Walmart isn't leading the way, they should be.

Walmart doesn't do much long haul. They do the warehouse to store, but that is a lot of local roads so the logistics of getting drivers to/from the truck on highway access is harder than just keeping a driver in the truck.

The big truck companies are interested in this. They mostly highway travel so a team of trucker drivers in each city isn't as logistically difficult. Plus they can promise their drivers home each night instead of twice a month.

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#85

Post by analog31 »

onion2k wrote: post #2

If I block a road an autonomous truck is expecting to drive along, will it automatically reroute to another road? If it does, then these trucks are going to be very easy to rob simply by corralling them in to a quiet location..

They could just hire all of the former drivers as guards. Not just against highway robbers, but even merely Luddites, pranksters, and protestors.

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#86

Post by joehartshorn »

bluGill wrote: post #75
joehartshorn wrote: post #46

Earlier quoted context omitted.

>"Trucks are also a better platform for autonomy, with their large size providing more power for computers" I'm no expert but "power for computers" doesnt strike me as a bottleneck, Geohot/Comma seem to be doing fine, this just sounds like marketing bs

Trucks tend to have bigger alternators, so there is more power available. Even the most basic economy car has more than enough engine power, but the alternator can't deliver that power. The engine power instead mechanically goes through the transmission to the wheels. (in a hybrid car there are two alternators, one for the wheels and when for the rest of the electronics so the same problem exists) the is solvable. yo…

I don't disagree with what you're saying, but my point was that I don't know if power consumption is even relevant, considering what Comma are doing from a phone on a windshield. I don't know (I'm not an expert) but I don't think putting TPUs in trucks is a requirement for solving driving.

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#87

Post by cooljacob204 »

Erlich_Bachman wrote: post #41
pydry wrote: post #35

Earlier quoted context omitted.

It's a bit more difficult with a person on board.

But not really. The person never makes a fuss (nor can realistically do so), and netiher does want to make a fuss for a simple job for one company, driving cargo owned by another company, with overall operation insured by a third company that's just going to reimburse everyone.

But it is, they would need to use violence to compel them. That greatly increases the risk of them getting shot (Many truck drivers carry guns) and it goes from larceny to robbery which greatly increases the criminal penalty.

Re: This Year, Autonomous Trucks Will Take to the Road with No One on Board

#88

Post by jariel »

bluGill wrote: post #84
jariel wrote: post #9

I cannot believe that this isn't the foundation of autonomous driving: long hauls, highways, no stop signs, semi-controlled circumstances. The economics are more clear and industrialized. Even if a driver had to be present when they got off the highway, I think 'long haul' should be naturally the first thing to be tackled. And I'm surprised Walmart isn't leading the way, they should be.

Walmart doesn't do much long haul. They do the warehouse to store, but that is a lot of local roads so the logistics of getting drivers to/from the truck on highway access is harder than just keeping a driver in the truck. The big truck companies are interested in this. They mostly highway travel so a team of trucker drivers in each city isn't as logistically difficult. Plus they can promise their drivers home each n…

Walmart is interested in moving a lot of stuff at scale. Apparently they wanted to help build a massive superhighway from mexico to somewhere in the midwest.

It's a great idea.

Walmart should be taking much more initiative in all of this as one of the only entities that can feasibly take on Amazon.

I wish the President would make a national strategy over this 'Digital Pathways Act 2030' - i.e. make all major highways, railroads, and other 'rights of way' completely autonomous and electric by 2030.

Even for human passenger cargo - long-haul buses may actually compete with trains on price and comfort if they were fully autonomous - especially for shorter routes.

One of the problems with 'trains' in America is that the US is a little spread out, trains can't just go downtown to downtown - but autonomous buses could.

So this could be like quasi-public transport initiative from at very least New England down to DC - that entire corridor.

What we need is someone to 're-brand' bus travel and make it social acceptable for middle/upper class folks to ride instead of the agrarian/lower-class greyhound. Nothing against greyhound but it's just not the experience many people want.

Post reply on HN