Hacker News Reader

Comment by didgeoridoo | Hacker News Reader

The approach that Starsky (http://starsky.io/) is taking seems to be a promising one. They employ truck drivers as remote pilots for the "first and last mile", i.e. the local roads and loading dock work that autonomous vehicles struggle with most. Other advantages include lower fuel costs (driving slower because AVs aren't subject to operator-hour limits) and probably also lower labor costs (you don't need to pay truckers as much if the job is safe and nearby to home).

If the Starsky model ends up dominating, this could all add up to a case where the price of truck transportation drops so much that demand for truck drivers actually increases, possibly even counterbalancing the pressure of individual truckers being able to manage multiple vehicles at once. This would be an example of the "Jevons Effect" https://en.wikipedia.org/wiki/Jevons_paradox

View on HN

The long haul stuff should go on rails anyway. It's very wasteful to run it on the interstates. The majority of the fatigue damage to the highways comes from trucks, then there's all the rubber dust, fuel inefficiency, etc.

The trailers should just be picked up and dropped on flatcars, and trucks used for intracity last mile stuff.

This comment invariably pops up in these threads; If it was going to, it already would have.. It's too late now.

Would it have been a better solution? Probably, but the investment required (laying down the rails, land buyouts, copious use of emanate domain, political greasing of the wheels, etc...) has always been prohibitive. The gains would be shared between many disparate groups that would need to fund this together, taking years to decades for any ROI - certainly a longer time-frame than most publicly traded, quarterly measured, corporations can afford. At the least this would have required some government subsidization.

With driver-less tech around the corner, returns on this investment for any one company become even more dubious, and thus make it less likely to happen. I think the parent comment here will always get some up-votes, but that is as far as I see this going anymore. Maybe it'd go farther if we threw in words like 'Hyperloop' ;)

The rails are already there. When I drive down I5, you can see them.

The reason this sorry state of affairs exists is because truck traffic is heavily subsidized (highways paid for with taxes) whereas rail traffic is heavily taxed (and the rail companies have to pay for the tracks).

It could be improved by increasing weight taxes on trucks, and using the proceeds to subsidize intercity tracks.

Yeah the impact on roads increases with the fourth power of vehicle weight. Why doesn't the tax?

Truck industry lobbying.

There's a lot to this. But it's not just about lobbying, it's also about unions and votes. The American auto dealerships likewise have a lot of what's described as "lobbying power" when really it's more that they have a lot of people that want to see politicians protect them.

Many trucks in the US have a bumper sticker that says, "Without trucks, America stops." Teamster strikes, like the León Vilarín teamster strike in October 1972 in Chile, are a disaster for any country and create national emergencies.

I read years ago that truck loaded to the legal weight limit causes 9,000 times as much fatigue damage as a car. And lots of trucks are overloaded.

At least trucks are carrying something we all need, unlike cars.

Even in France, with paying interstate (called autoroute here) and subsidized rail (it was a state monopole until recently), most of transport of goods is by truck.

I think it's mostly a last mile problem (you'll need big trucks to carry from the station), compounded by the fact that most of the train station of cities and I don't think they is a lot of dedicated cargo train station. Also you do not control part of the schedule: you are dependent on the departure/arrival time of trains. And don't forget the strikes...

I believe SNCF was also know for loosing freight a few years ago. Might have improved these days since I assume most cars must have some sort of GPS. Timing was another problem since freight had to yield priority to passenger traffic, which is pretty dominant in France.

Europe generally prioritises passenger transport over freight transport, but there are rail freight terminals -- probably not anywhere you're likely to visit (example [1])

Rail traffic is allocated on a timetable, for both passenger and goods trains. The time and route is called a "path" in British English, and reserving one costs money.

A power station might have a daily path from a port to their power plant, as they can reliably use all the coal.

Rail freight companies book many paths between freight terminals, container ports and so on and mix together goods from many customers to run on trains on these paths. Therefore, the timings are reliable, potentially more reliable than by road, but the flexibility is less.

If there aren't any goods to move, the company doesn't need to use the path. (I used to live close to a railway line used to transfer trains containing nuclear waste. The reprocessing plant had a path to run a train every day, but only used it every couple of weeks -- presumably, they don't like leaving the waste at the power station any longer than necessary.)

[1] https://en.wikipedia.org/wiki/Rail_freight_in_Great_Britain#...

At least in the United States, this is incorrect. Long haul trucking is expensive compared to shipping things via the rails. The reason things end up being shipped via truck is that its faster, plain and simple. Shipping via rail car is slow and has no guaranteed arrival date. Or should I say, dependable arrival date. And from a logistics stand point you have to engage a last mile trucker at the rail head nearest your true destination to haul your item to the destination. Since you're unlikely to be shipping an entire container/rail car, you'll need to pay extra since the delivery driver will have to drive to your destination then 'dead head' (i.e. return without a load) to the rail yard.

These are just organizational problems that could be worked out if the financial incentives were there, instead of tilted heavily in favor of trucking.

Why aren't warehouses located at or near rail stations?

Because trucking is, or at least was, cheap.

Prior to the rise of the Interstate Highway system, most warehouses were on rail lines, or spurs. You'll still find the trackage, or at least the traces of them, in many urban areas throughout the U.S.

In San Francisco, the region near Potrero Hill by the old Fashion Design Centre. The odd street patterns and building perimeters show the former track rights of way.

https://www.google.com/maps/place/San+Francisco+Design+Cente...

You're arguing past parent. He stated not that trucking is cheaper, but that it is heavily subsidized. In almost all cases, rail would cheaper, and some things will take advantage of that. Likewise, in all situations trucking will be more expensive,yet there are use cases where the reliability and speed are justified (just like couriers are sometimes justified.) The difference made by the subsidies is all in shifting the margin.

Things that would be toss-ups or wins for rail are instead pushed to trucking. This would just be wasteful, and not so pernicious, were it not for the serious negative impacts on the layout of our towns that result from favoring trucks so heavily.

You also have cases like Portland where they shut down the shipping terminals which mean that all the things that would move by rail instead moves by truck now.

The amount of semi traffic from Portland to Seattle on I-5 has exploded, it certainly made driving that stretch a lot less enjoyable.

The current rail network already grows and shrinks with demand so long term you it can continue to expand without massive investments.

IMO, Automated trucking makes trains more not less useful. The core problem with using trains is speed plus how much time and effort loading and unloading takes. With automated trucks you can significantly reduce this friction.

There's very real advantages to long-haul trucking over rail. Shipping time costs logistics operations real money in terms of inventory costs, flexibility, and responsiveness to market conditions. Cargo ships are even cheaper than rail, but shipping costs aren't always the dominant concern. Similarly, some goods get air shipped even with the higher cost of doing so.

It already does - if its longer than 600 route miles, is timed to move slower than 50 mph, its already on the rails, thats what TOFC service is.

What makes you think this isn’t already happening?

Norfolk southern runs behind my house and I see semi trailers on top of rail cars all the time. I see UPS trailers on trains almost every time I watch one.

Refrigerated semi trailers are common on trains too. I’m guessing they have food in them.

> What makes you think this isn’t already happening?

Driving on I5 that seems to be a continuous stream of semi trucks, while I rarely see a freight train next to the freeway.

This is Stefan, from Starsky Robotics.

I hear this all the time, and honestly in many cases a truck is the transportation of last resort. If you need 5 shipping containers of goods going from A->B all arriving at Time X; it's probably going by train.

An issue with rail (as pointed out by others) is how to haul the freight from the rail head (rail yard) to the end warehouses, and the extra time it takes to have goods go on two trucks and one train instead of just one truck. The other issue is that in the US it's really hard to hire and retain long haul truck drivers (it turns out most people want to go home each night after a hard day’s work).

Competition to hire drivers is fierce, and driver turnover can be 100%/yr...which means sometimes your drivers don't show up to pull your freight. Rail is also generally cheaper so if you can put your goods on the train you do.

As far as wear and tear on the highway: that's what weigh stations are for. They not only assign usage taxes, they also make sure weight is evenly distributed across the axles to limit damage to the road.

/SSA

>As far as wear and tear on the highway: that's what weigh stations are for.

That doesn't create a system where trucks pay taxes roughly proportional to the damage they do.

I remember reading the damage a vehicle does is roughly proportional to the fourth power of its weight.

There's a pretty good chart here, near the bottom: http://pedalfortcollins.com/greatest-demand-on-tax-dollars/

Kinda. Has more to do with axle load than total weight.

You could also say that the research is pretty incomplete. All the known studies have been performed at highway speeds...it's entirely plausible that speed exaggerates the damage disparity.

Hey are you all trying to sell your hardware and software or are you trying to create a whole shipping company? Who are your competitors? I've been trying to do some research on companies in this space but it's been a bit hard to find info on anyone except Otto. If you've got the time I'd like to ask you a few questions, my email is in my profile.

This is their competition, and they're much better funded (although funding doesn't necessarily equal success).

http://embarktrucks.com/

Thanks! I really appreciate it!

>As far as wear and tear on the highway: that's what weigh stations are for.

I believe WalterBright meant that even at the legal weight limit, trucks cause 10,000x as much damage as cars. In 2005 this amounting to a $60b taxpayer subsidy.[1]

This isn't to criticize Starsky Robotics (this inefficiency predates it), which is just operating in the environment it finds itself. And in general, inefficiency begets opportunity.

[1] https://truecostblog.com/2009/06/02/the-hidden-trucking-indu...

>As far as wear and tear on the highway...

Right or wrong, I don't think you have to defend that point. The government sets these parameters under which you (or truckers) do business. It's not your responsibility to maintain the highway as long as each truck complies with regulations. And it's not your fault.

American taxpayers subsidise many other things that don't quite make sense (to me), like corn production. You don't see people blame farmers for doing what's best for their farm and family.

Don't forget to take into account there's normally only one or two lanes of train traffic that runs right along side of an 8 lane high way.

Much more chance for congestion. I lived right next to a set of tracks once. Trains would regularly sit for over an hour without moving.

There are a few things I think you've overlooked:

- For one, you're talking about human drivers covering first and last mile, which is a tiny, tiny fraction of the route that a truck covers which is why the Starsky model works well, in theory anyway. That means one given driver can do the work of probably a few hundred trucks in their last mile phase, even if they end up with some downtime in the middle, it's not even close to balancing. You're still talking about a possibly slightly larger market being overseen by a much smaller workforce.

- Secondly, this completely ignores what could be the silver medal effect of truck driving becoming automated; the death of millions of small and large businesses that operate basing heavily on the incomes of truckers, everything from truck stops and diners, all the way to motels, specialty truck accessory companies, on and on and on. These are the ripple effects that many people aren't looking at when they're thinking about this, it isn't just trucking, it's one of the last jobs available to you with naught but high school under your belt that earns around $40k. And that's not even going into how many businesses do a ton of business with truckers that earn that wage, which is usually much higher than their neighbors are earning.

- Thirdly, if all trucking is done from various Starsky-like operations, then truckers in general will be moving into more consolidated living conditions. Again trucking is one of the VERY rare jobs that requires minimal training, pays a comfortable wage, and is available pretty much anywhere in the States where there are roads and people who buy products. If you want to work as a software developer, chances are you'll have to move to Silicon Valley, San Francisco, Austin, etc. but if you're looking to be a trucker, you can go pretty much anywhere and have fairly solid employment opportunities. For now.

> Thirdly, if all trucking is done from various Starsky-like operations, then truckers in general will be moving into more consolidated living conditions.

why? if the controls are network traffic, and one driver being right next to another driver doesn't add value, why not spread them across the country? it'll reduce control latency, increase reliability (can't have all deliveries stopping if the fiber to one building is cut), and make it easier to have drivers awake when the deliveries are tending to take place (certainly some locations might take deliveries in the middle of the night, but mostly they're going to happen when businesses and such have employees available to receive the materials.)

If you want to work as a software developer, chances are you'll have to move to Silicon Valley, San Francisco, Austin, etc

If you're willing to work for truck driver wages ($40K - $70K), you can be a developer in any sizable town in the USA.

If you want to work as a software developer, chances are you'll have to move to Silicon Valley, San Francisco, Austin, etc.

Not that I disagree with your basic point, but that's a pretty narrow view of the software world. There are jobs for developers all over the place, even in podunk towns with a few thousand residents. Granted, that town might have like 3 total jobs for programmers, but a. that's an extreme example and b. "long tail". Consider all the mid-sized cities out there, as well as "non tech hub" big cities and there's plenty of places for programmers to live and work.

Just to illustrate:

https://www.indeed.com/jobs?q=java&l=Buffalo%2C+NY

https://www.indeed.com/jobs?q=java&l=cleveland%2C+oh

https://www.indeed.com/jobs?q=java&l=fargo%2C+nd

https://www.indeed.com/jobs?q=%22visual+basic%22&l=Oklahoma+...

https://www.indeed.com/jobs?q=c%23+&l=Boise%2C+ID

https://www.indeed.com/jobs?q=java&l=chicago%2C+il

https://www.indeed.com/jobs?q=java&l=Cincinnati%2C+OH

https://www.indeed.com/jobs?q=java&l=Phoenix%2C+AZ

https://www.indeed.com/jobs?q=java&l=Tampa%2C+FL

etc., etc. etc.

That's almost certainly a temporary effect though - if the trucks are being outfitted with the hardware necesary for full automation, full automation is only a software update away and the pilots will only have a job for the relatively short period where automation isn't good enough to handle that portion of the journey.

It's going to be a one hell of a software update...

that's how autopilot landed on current Tesla cars.

> this could all add up to a case where the price of truck transportation drops so much that demand for truck drivers actually increases

The problem is that local drivers average about half the pay ($30-40k) vs long-haul drivers ($70-80k). Even if there is more demand for local drivers it likely won't offset the wages lost.

There are a lot of unknowns here. The factors putting downward pressure on trucker salaries are:

- Remote operation of trucks is an easier lifestyle, so more people would be willing to take the job, driving down salaries

- Fewer human truckers are needed per truck on the road, so for a given level of trucking activity, fewer jobs exist

Some potential countervailing factors:

- Price of trucking plummets, potentially jacking up demand (Jevons Paradox) depending on price elasticity of truck transportation

- Managing multiple vehicles may end up being more cognitively challenging than driving a single truck, reducing the pool of available workers and driving up salaries

The values of these parameters (and how they change over time) will be crucial to determining the net impact to trucker salaries and employment numbers. My intuition is that salaries will go down dramatically, employment numbers will go down modestly, and overall trucking activity will expand.

> Managing multiple vehicles may end up being more cognitively challenging than driving a single truck, reducing the pool of available workers and driving up salaries

Tangential to this point: If this mode of truck operation becomes prevalent, I could potentially see a path where warehouses/loading docks get employees certified (a la forklift truck certifications) to operate the truck on premises to handle the "last mile" (i.e. on-site) operations. Now you've completely decimated the number of trucker jobs.

I think you're drastically underestimating the difficulty involved in manuevering an eighteen wheeler around the premises of a typical business. There's no way it would be cost effective to train people for that so that they can spend most of their time doing something else.

The wages aren't "lost", the "saved" money will still be paid to somebody to do something.

Profits kept by shareholders.

and spent by them as well.

Maybe. Depends on tax law at the time. Going to need social safety net money, we might as well tax automation.

Our tax code needs reform, but taxing gains in productivity is a perverse incentive.

> but taxing gains in productivity is a perverse incentive.

Not at all. The end goal should not be productivity onto itself. If those gains are being scooped up by a sliver of the population, we change the rules to benefit all. One person, one vote.

As automation automates away more jobs, the cohort of people without jobs grows, forcing regulations (through democracy) to change to move towards a citizen's dividend funded by a tax on automation. The system will reach equilibrium, or there will be a revolution.

If you don't want your automation lifting the tide of all boats, someone else will come along and build it.

That mindset only works without the reality of globalization.

Events like Brexit and Donald Trump are only the start. If economic conditions gets worse for the disadvantaged, it gets worse for everyone.

Clearly, democracy can override your usual globalization proponents (see: "Okay, The Trans Pacific Partnership Is Dead. What was it?" | https://www.washingtonpost.com/news/monkey-cage/wp/2017/01/2...)

Spent by them in other countries, where they patriotically only spend it locally.

That’s the problem of the modern globalized world, all the profits end up spent in the US.

Rich people spend much less of their money on stuff than poor people do. Another thousand to a millionaire goes on the pile. Another thousand to a single mom goes on kids' shoes and car repairs.

Nobody throws their money into a pile.

Rich people predominantly invest their money which is still paying people to perform work, just like repairing shoes and cars.

Yes. They invest it. Not spend it as you claimed above.

And while both keep money moving, they have very different effects on the spender/investor.

Yes, "spent" was the wrong word.

All I'm saying is that the money will be moving, and still providing employment.

Shareholders aren’t all ‘rich people’ - pension funds, insurance funds, and regular old retiremt funds all hold a lot of stock.

People who have no hope of ever retiring from their low wage job that they live in fear of losing to automation might count anyone with a pension plan/retirement fund access (and even insurance) as rich.

They're lost to those who would have earned them.

Typically the gains would be shared through lower prices to customers and higher profits. Lower costs might also increase demand for services and goods, growing the industry and increasing employment elsewhere in companies, though that latter is likely to be a relatively small effect.

Its not about lower costs or growing the industry.

If we save $1 billion dollars in automating one industry, we are just going to spend that $1 billion dollars elsewhere, growing some other industry.

That's right, it frees up $1bn for other uses. This is why automation, ever since the industrial revolution, hasn't lead to persistent mass unemployment. The capital freed up by more efficient industry was used to create new industries and jobs that didn't even exist before. There have been some painful adjustments, but the more flexible an economy is the more rapidly it adjusts. Historically this has been a US strength.

> They employ truck drivers as remote pilots for the "first and last mile"

It would make even more sense to employ a "local" pilot with local knowledge for the first and last miles, kind of how tug-boats help large container ships in and out of ports.

Harbour pilots actually drive those large ships in and out of port from the bridge.

If Uber is right about it's autonomous taxis, and Tesla's cars really will be able to come when you whistle, then last mile driving and docking are surely solvable problems.

I'm generally an optimistic person, but I think autonomous driving is going to seriously disrupt our societies. I simply cannot see where the huge workforce of unskilled males driving trucks and taxis is going to be soaked up. The industrial revolution did it last time, but anything you can suggest as a potential equivalent (colonising Mars?) will be demanding smart flexible people - and an awful lot of automated systems.

I'm picking the mass unemployment will lead to civil war in the US (where all those unskilled males have semi automatic guns) and universal basic incomes in more civilised countries.

Even so, it will devastate the economies of the places they drive through.

Truck stops, motels, independent semi repair shops, CB radio and accessories, probably more I'm missing. Oh, yeah, the final death knell for DVD porn.

Parallels also to the rise in the demand for horses after the introduction of the railroad.

The availability of long-haul transport increased transportation volume, but the "last mile" connection from origin/destination to the railroad depot couldn't be accomodated as readily. So: horses and carts.

Variously: Robert J. Gordon, The Rise and Fall of American Growth, Vaclav Smil, Energy and World History -- of which there's a heavily revised edition out this year, Energy and Civilization.

I really don't see why autonomous vehicles couldn't handle the first/last miles soon after mastering the open road. I think only regulation would made this a requirement.

For trucks that don't follow the same routes every day (e.g. FedEx/UPS/etc.), even humans frequently have a hard time getting it right. "Oh, no, it's the other side of that bridge that's only for light vehicles. Yes, you'll have to drive around for two miles in heavy traffic to get here. No, it's the entrance on the back facing the tall green building." Etc...

Until AI for vehicles massively increases, humans are going to be better at navigating the truck through smaller roads, abiding by weight restrictions that may not already be known about, having local knowledge of the area that AI might not have. Every delivery location is going to be a little different so you would need a really general AI to understand how to just get the truck up to the loading dock (if you are that lucky).

The long haul portion of the journey is much more suited to the AI of today, cars already have enough tech to get this done with pretty good results.

> Every delivery location is going to be a little different so you would need a really general AI to understand how to just get the truck up to the loading dock (if you are that lucky).

or, as someone with a loading dock, i'd be willing to pay to have someone come out with a bunch of lidar gear and whatever to model up my dock and surrounding area, so i could just provide a big data blob about my dock to the shipping companies, and have their AI trucks pull in correctly every time.

as opposed to worrying about the days when some new guy (with his local knowledge and general intelligence) shows up to deliver and comes in the wrong way and gets stuck for 2 hours.

(there's a new job for you right there, folks, since the model will probably need to be updated every year or two as the shipping companies change their AIs, and changes are made to the surrounding buildings.)

I expect many regular delivery end points will get set up with autonomous vehicle standardised parking bays, with programmed routes in and out added to the AI mapping databases. If you don’t have one of those, you’ll need to pay a premium for driver operated delivery or remote driver guidance.

One word - teamsters.

Corollary - Why no driverless trains yet?

The savings likely aren't worth it. Trains can pull a lot of cargo compared to a truck, so the marginal cost for a driver is a substantially smaller percentage.

Yup, this is often the answer.

Same with "why do we still have humans taking orders at McD?" Then it materializes when the cashier's minimum wage pops up -- in comes the kiosks.

I don't think it has anything to do with cashier wages, and more to do with customer perception.

Kiosks are already cheaper than people.

With minimum wage at $7.25/hour, and a Kiosk being able to replace 2+ shifts (16 hour days 7 days a week), that's $30K/year in savings, which would more than pay for a kiosk and maintenance.

Developing the software to run them isn't free, but spreading $15M in development costs among 14,000 restaurants is only $1K each.

But many customers would balk at having to order at a kiosk, so McDonalds will still hire humans.

> But many customers would balk at having to order at a kiosk, so McDonalds will still hire humans.

A lot of locations around me have both kiosks and staffed registers.

I used the kiosk once and thought it was annoying to work through. If you have a simple order like "A medium #3", then its an order of magnitude faster to order through a person. On the other hand, if you like to customize your burger, or order a several things a la carte, the kiosk might be better depending on the competence of the human at the cash register.

(Of course, whether or not your customized burger or list of a la carte items gets made correctly by the kitchen staff is an entirely different conversation...)

> If you have a simple order like "A medium #3", then its an order of magnitude faster to order through a person.

This is a user interface [fixable] problem, not some inherent defect in kiosks. You should be able to walk up to the kiosk, tap #3, wave your NFC payment at it, and walk away with a hamburger. Today's "Tap. Tap. Scroll, Scroll. Tap. Back. Tap. Tap checkout. Tap Ok. Insert credit card in this other machine. Tap confirm. Sign. Remove card. Tap done." mess is merely incompetent product design.

I think the decision has 3 main factors:

1. Total operational cost,

2. Customer satisfaction,

3. Risk related to change.

You make a good point though -- perhaps it's not primarily driven by the combo of current and projected minimum wage hikes.

Can anyone quantify?

Also usability. Ever watched an older person try to get a drink out of one of those touch-screen drink dispensers with 100 choices of soda+flavors?

Same would happen with a kiosk order taker. Two humans talking can still communicate much more reliably than one human navigating a somewhat unfamiliar UI.

They exist, you can find them on e.g. the London underground.

I'm curious how the minimum wage compares in the various train driver jurisdictions. (I don't know -- possibly not relevant.)

There's an initiative to move to autopilot for trains, or Positive Train Control. The US Gov't started the initiative in 2008 after a particularly gnarly train accident which killed 25 people. It was due to be rolled out in 2015 and we're still waiting...

For long-haul rail freight the economics don't matter as much as they do for trucks. 1-2 people can control a 7 mile train as opposed to trucks which have 1 driver per 50' trailer.

/SSA

There was an ELI5 post the other day with a pretty detailed response[0]

The TL;DR amounts to the braking system and current regulation about when trains are eligible to be automated.

[0]https://www.reddit.com/r/explainlikeimfive/comments/712012/e...

They exist for local rail and subways, and with the unification of the European signaling and braking systems, soon they’ll also dominate long-range rail

http://www.afr.com/business/mining/iron-ore/rio-tinto-makes-...

Reply on news.ycombinator.com