[yourapostasy]

> ...filled to the brim with sensors...

Has someone actually worked out that tons of sensors will cost far less than people-driven trucks? As it is, fuel is the big cost, followed by driver salary [1]. L5 autonomous driving is not going to come cheap, that gear is going to price as close to 3X driver salary as they can get away with, on the assumption they can run close to around the clock. Whose margin is getting compressed for the additional sensor gear?

This doesn't even touch upon that as soon as L5 is available and if 24x7 L5 operations approved, you suddenly just increased industry transport capacity 3X, leading to a sudden oversupply in certain segments and scenarios, while still requiring a certain baseline to handle peak load demands. That chaos will cause a lot of margin compression, and lots of rosy profit projections from L5 autonomous driving without drivers will turn into a race for finding more customer demand.

I can see some modest sensor gear, but nothing fancy, and not a lot of them. Perhaps high resolution visual and night vision cameras coupled with lots of street camera access, with lots of back-end software processing will deter most theft attempts?

We might ironically get to L5, only to stick lower-paid security guards on a random number of trucks.

[1] https://www.thetruckersreport.com/infographics/cost-of-truck...

[elsonrodriguez]

The average truck driver earns about $70k a year. Even assuming that the sensor suite costs as much as an entire Tesla Model S(which in addition to a sensor suite includes an actual car), the system will pay itself back in a year.

Also once automated, the trucks can engage in all sorts of hyper-miling shenanigans since they don't have to worry much about traffic during a significant part of their 24/7 operation, especially on more remote roads. That's additional fuel savings.

[jandrese]

Sensor packages and near-AI compute clusters constantly get cheaper as the technology improves.

Today it doesn't make sense, but in 10-15 years when a full self-driving solution costs maybe $1k? It's a no brainer, especially for long haul trucking.

I'd guess that putting a security guard on a truck will be an exceptional occurrence, probably only used when the truck is hauling an especially valuable cargo or going through a known trouble spot.

[yourapostasy]

> ...but in 10-15 years when a full self-driving solution costs maybe $1k? It's a no brainer, especially for long haul trucking.

I once spoke with a highly-paid driver (used to be a programmer, using the gigs in truck driving to decompress because our industry generally has worse work-life balance than truck driving...chew on that for a bit) about the US logistics industry.

The driver was in high demand because they consistently tested drug-free, carried various kinds of specialized certs, was always on-schedule or always in communication about problems, and fixed many problems on their own. As I remember the explanation, there is some kind of trucking industry-wide database that contains every driver's trucking records, and it shows every ping of their record to everyone. Might have the details wrong, but the gist was every single time a competing trucking company pinged their record, they got a raise to stay without even asking. So they were in a good position to watch from the best of what the trucking industry could offer. Their contention after observing from inside the industry for a number of years is that the bulk of the US logistics industry is the train companies' to lose.

Placing enough sensors along the tracks and looking outward to the sides of tracks to detect conditions requiring trains to slow down way ahead of time, but otherwise clearing trains to run at much higher speeds than they are allowed to now, would go a long ways to fixing many of the train industry's delivery speed. Upgrade the tracks themselves and the rolling stock to boost the speed even more to match trucking's coast-to-coast delivery time, and there isn't much incentive to use trucks for those corridors rail serves.

[ethbro]

Makes sense. Logstics is fundamentally a network problem optimizing for latency and throughput.

Jitter is toxic to both, and road had a huge amount.

Rail networks are at least in control of most of their own variables!

[ethbro]

Given: (1) sensors necessary to autonomously navigate, (2) large stretches of rural America (out west, maybe 100+ miles to the nearest police station), (3) adaptable human adversay, (4) no humans to injure on the vehicle

... I just don't see how you economically protect a vehicle (vs cargo value).

And more sensors simply mean more things to steal. The minimum law enforcement response time along your entire route is the real issue, and there's no way you decrease that short of drastically increasing police staffing.

[jandrese]

These thieves have to steal from a moving vehicle? Sure it's not impossible, but it's more Hollywood heist than something you'd see in real life. It will almost certainly happen a few times, but for practical purposes the amount of stuff stolen before the guys are caught will be less than what it costs to hire tens of thousands of rent-a-cops to sit in the backs of trucks.

Plus, these are thieves we are talking about. Pointing a gun at a human driver and telling them to pull over so they can rob the truck is something that could easily happen today but is extremely rare. The minimum police response time is something that's hard to measure. It might be many minutes 90% of the time, but if you do 10 or 20 of these heists eventually you're going to get unlucky and the cop will happen to be sitting at a speed trap right there are you're busting into the truck.

This is the fundamental problem with crime. You will get away with it most of the time, but when you don't you're fucked. It's a lousy career choice because the upsides are modest relatively speaking, and the downsides are huge. If you're going to be a criminal the trick is to steal enough to retire on and then immediately retire. Knocking over one random truck is not going to do it.

[ethbro]

You can get a 50 BMG rifle for a few thousand dollars and put a hole right through an engine block.

Unload truck at your leisure.

That's not even getting into ways to make an automated system stop by putting up an emergency / stop sign in the middle of the road.

The risk to stealing is directly proportional to the chance of getting caught, which broadly correlates into something unexpected happening, which is drastically diminished by not having a human driver.

[smileysteve]

> You can get a 50 BMG rifle for a few thousand dollars and put a hole right through an engine block.

So, how does a single, or even a 2 driver truck prevent this TODAY? The only reason I can think is conspiracy felony murder of 1 or 2 people w/ clubs, guns, cell phones has a higher risk than conspiracy felony larceny?

You've concocted a great hollywood heist scene; but if you're Fast and the Furious driving through South America, it's The Rock and the Swat team hanging out in cargo hold you have to worry about; not whether the truck has an autonomous driver or not.

[ethbro]

> So, how does a single, or even a 2 driver truck prevent this TODAY

See my last paragraph.

Defending an unmanned vehicle is the walls vs guards argument: it's far harder to build an impassible, unguarded wall than it is to build a difficult, guarded one.

[jandrese]

This assumes the truck has no way to call for help when something goes wrong, or that people driving on the same road won't call the cops on you. Long haul trucking doesn't usually happen on deserted backroads.

I'm just saying that the worry about theft is probably overblown. It's almost certainly going to be rare enough that a regular insurance policy will be sufficient protection. People can be hired for exceptional cargo, but that's true today too.

[ethbro]

Have you ever driven interstate east of the Rockies and west of the Mississippi?

If not, you'd be surprised how bare it is. Especially if automation kills the truck stops.

We'll see. The means are trivial for any enterprising farmer, and I think the moral calculus drastically changes once you remove risk to a human driver from the equation.

[james_s_tayler]

As my IT teacher always used to say "don't go to jail for 50 bucks, go for 50 million"