Hacker News Reader

Comment by tdicola | Hacker News Reader

I like your optimism, but realistically self driving trucks on American highways are a long way off. How will the trucks refuel? How will they handle loading and unloading? How will they handle inclement weather? What is the insurance liability for putting them on public roads? Who will be sued when a driverless truck crashes into another car (at fault or not)?

Even Mercedes says in that article, "the Highway Pilot system could be launched into production trucks 'as early as 2025 if conditions permit'." Which is basically corporate speak for, "we have no idea if this will ever launch, but it will get investors excited now."

View on HN

>How will the trucks refuel?

Full serve gas stations.

>How will they handle loading and unloading?

Do long haul drivers typically load/unload themselves these days? I would have thought that would have been the task of the warehouse staff on either end.

>How will they handle inclement weather?

Probably better than human driven trucks why would you think they couldn't?

>What is the insurance liability for putting them on public roads?

Probably less than human driven trucks since you don't have to worry about a lot of things that will cause problems for those such as tired/impared drivers.

>Who will be sued when a driverless truck crashes into another car (at fault or not)?

The trucks insurance company. Same as it would currently happen.

Going to disagree with you there on inclement weather. Getting a self-driving system to work on sunny days and clear nights when there's an otherwise clean view of the road and surroundings is one thing, and that's the challenge I see most people working on right now.

Now try the same thing in the middle of winter, the equivalent of visual noise is obstructing both the road and the surroundings. People are good at picking out patterns in this condition, especially if they've been driving for a while, computers notsomuch.

Sure, it's a matter of the tech advancing, but solving the second problem is harder than solving the first.

People are horrible at driving in any kind of bad weather. They speed through foggy conditions where they can't see 50 feet in front of them. They drive too fast or erratically on icy or snowy conditions, or they just flat out don't know how to handle the conditions and swerve/oversteer.

If people were any good at it there would be no increase in crashes during winter, foggy or wet weather conditions.

Self driving vehicles will have the same issues with seeing in the visual spectrum in bad weather as humans but they will also have the advantage of IR for example in knowing if something is in the fog/snow in front of them and how to react to that.

You also remove the emotional component. Humans in really bad weather tend to either ignore it(which is risky) or drive scared(which leads to over reactions which leads to accidents). For the second case things like over correcting, slamming on breaks, swerving, etc all lead to tons of accidents that are easily avoidable.

I said people were good at picking out patterns, not that they were necessarily good drivers.

Yup and I said with technology it doesn't really matter since computers aren't limited to our visual spectrum(they can use thermal imaging when needed for example) so they won't have the same difficulty seeing in those conditions as we do.

This is further compounded by the fact that even if humans can see the patterns they rarely act on them perfectly and often do the exact opposite of what they should do.

Not to mention an automated truck driving a safe 20 MPH over 24 hours is equivalent to a human driver at 60 MPH for 8 hours.

I think it's a false-dilemma to assume it needs to be total automation or nothing at all.

What about "trains" of trucks, where the first truck is piloted and the rest follow? That'd be significantly easier to engineer, and there'd be a person on-site for miscellaneous needs like gassing them up.

One of the major advantages of trucks over trains is precisely that - they're not trains. Making them trains would take away all the flexibility they currently have. If you're taking coal from the mine to the power station then you're probably already using a train. Anything less predictable needs individual trucks.

"Trucks aren't trains" is a tautology.

First, having 2-5 trucks in an automatic convoy is still ENORMOUSLY different than a multi-locomotive train with 10+ or even 100+ wagons on fixed rails.

Second, even if you equalized their "sizes", a three-truck convoy is NOT magically equivalent to a three-wagon train, because they are running on fundamentally different infrastructures.

Automation doesn't mean having to solve every problem simultaneously. You can still pay a driver to stay with the truck, and handle the fueling and loading. Just being able to keep the truck moving while the driver is required by law to rest will be a huge win.

Then the automation doesn't solve the staffing problem, it just perpetuates forced pseudo-downtime that drivers get underpaid for.

Say you have two trucks doing a cross country run(on the I-10 to keep things easy) and a return run. One with an automated system that can drive while the driver sleeps and one without.

Truck A: Can drive 24 hours a day and can cover the distance there and back in about 3 days

Truck B: Can drive 11 hours a day and can cover the distance in ~7 days.

The first Driver can do the trip twice in the same amount of time. Halving the number of drivers required for that route or freeing up drivers on that route for other jobs.

Reply on news.ycombinator.com