Comment by PhilWright | Hacker News Reader

Comment by PhilWright | Hacker News Reader

I think the panic over automation of cars and trucks is a little premature. At the moment you cannot buy a car or truck that will auto drive. I mean auto drive in the sense of not needing a driver at all. Until you can completely remove the driver then there is no danger to the taxi/trucking industry, it will be only a driver aid in the same way cruise control is a driver aid.

There is a huge gap between Google testing a driverless car, with a person in it for emergencies, and actually seeing hordes of driverless taxis of trucks on our streets.

I can imagine a new trend with bored teenagers. Jump in front of the driverless car or truck and watch it emergency stop. Then run away. Or wait for the news report where all the driverless cars in the city have gone to the same car park because a hacker thought it would be hilarious. Just one truck accident that wipes out a few children and the trucks will be taken off the road. After all, if one truck has the bug then they all do. When a person makes a mistake we don't stop all trucks because we blame the person, when a software version crashes then they all could crash.

Worry is not as premature as you think, because replacement of truck drivers does not have to happen all at once. It can start with just replacing the drivers on long haul routes on the interstate highways. So when, say, a container ship arrives in Long Beach from China with goods to be trucked to a Walmart distribution center in Florida instead of needing to dedicate a driver to that truck for something like 80 hours you only need a driver for a few hours at the start and a few hours at the end. That would greatly reduce the number of drivers the long haul trucking industry needs.

When the problem is restricted to self-driving on the interstate highways on fixed routes it is a lot simpler than the problem of general self-driving so we'll get that solved a lot sooner.

Furthermore, there is an even simpler intermediate step, which is to have self-driving trucks that just know how to follow another truck. You then have one leader truck driven by a human, with a line of self-driving trucks follower trucks. This simplifies the problem dramatically, getting rid of almost everything that is hard about self-driving vehicle technology. The limiting factor on when truck trains will be commercially available will be regulatory, not technological.

As cool as the concept is, no, I don't think so. A line of self driving trucks can break up for any reason (cars getting between two trucks, messy junctions, an isolated accident or malfunction). At that point you need again full autonomous driving, as you can't stop a whole column of trucks on the emergency lane, run to the one that got stuck (and the rest of the column behind it) and make it catch up with the rest.

AI-controlled trucks can drive nearly bumper-to-bumper, because there needn't be any reaction-time delay. If the lead truck hits the brakes, every other truck in the column can brake instantly, and with exactly the same pressure. It effectively becomes a single road train with no space in between. Messy junctions can be avoided as long as you stay on the highway proper and don't pass through cities; with this plan, you'd have short-haul truckers at either end for the first-and-last mile work, with the AIs handling the cross-country part.

As for accidents or malfunctions--why can't you stop a column of trucks in the breakdown lane? I'm assuming that the trucks would have at least limited self-parking ability, since that seems to be at least as easy as highway driving for a development standpoint.

> You then have one leader truck driven by a human, with a line of self-driving trucks follower trucks.

We have had these for over a century. They're called trains. :)

And if the rail system extended as widely as the road system, we wouldn't have had semi-trucks in the first place.

The trucker jobs would have never existed then, and the rail could have been already automated by today's date.

I wonder how different the US would have looked if that was the case. It might have also slowed down climate change somewhat, since trains are much more efficient than trucks (about 3x in terms of mile-ton per unit of fuel).

Well, at one time it did[1]. A hundred years ago, I could have walked down to the train station down my street, hopped on a train and gotten off in my home town, 150 miles away, having taken a nap or read a book on the way. Now I've got to drive myself that distance in a hurtling chunk of metal, constantly on the alert for all the other crazy people darting around in their high-velocity chunks of metal, and arrive exhausted.

Ain't that some progress

[1] http://users.humboldt.edu/ogayle/Hist%20111%20Images/RR1890....

Australia already uses trucks as trains.

The majority of the article didn't really focus on the driverless truck issue - it just highlighted that that's the real tipping point.

I think this is a real issue, even if the driverless issue doesn't happen as soon as some think/fear it will. Technology is encroaching on a wide range of jobs, but this time it's to completely automate them, not simply to augment them.

I do agree with most of your analysis of the driverless car/truck issue though - I think we have a hell of a long way before that becomes commonplace.

>I can imagine a new trend with bored teenagers. Jump in front of the driverless car or truck and watch it emergency stop.

I'm not sure why it being automated would change how braking physics work. An emergency stop at 50km/h will still result in a minimum 25m braking distance no matter who is behind the seat.

Having a fairly guaranteed reaction (from software) rather than randomly effective human reflexes turns what was a dangerous game into a prank.

but since the truck can/will have external camera mounts, the interaction gets recorded and the kid gets ARRESTED? :)

25m at 50km/h is an over-conservative approximation peddled by driving test authorities who assume 1.5 seconds of reaction time.

With instantaneous reaction time, the actual braking distance is around 10m unless you jump in front of an escalade or a land rover...

Additionally truck could take an high res photo + location of those teenagers and mail it to concerned authorities with in seconds. I think this will lead to less investigative work and hence less law enforcement employment.

I imagine that we'll get automated convoy AI, with a single driver in the lead truck, well before truly autonomous trucks. That will put a lot of truckers out of work by itself.

We should build dedicated roads for these automated convoys.

Yes, you could give them rails to make them more efficient.

And how exactly does that solve the last mile problem?

We can offload to trucks driven by humans well suited to driving the intricacies of city traffic.

You could just run them on narrow strips of metal rather than bother with a wide stretch of tarmac that the wheels wouldn't be touching most of.

Great idea, then we can put rails on those roads to make them easier to navigate.

Why would we need special roads when we could have truck drones fly over traffic to their destinations?

This is already being tested! https://www.theguardian.com/technology/2016/apr/07/convoy-se...

Reply on news.ycombinator.com