Hacker News Reader

Comment by jonplackett | Hacker News Reader

Yet again this article mentions 5G being "a must for automated-driving technology"

I don't see why this is necessary? A car is going to need to make decisions without a connection so why does everyone go on about this all the time? Genuine question.

View on HN

For context, I work in the industry. There are effectively two major use cases for which 5G makes sense: teleoperator backup, and V2X.

The former is for Level 4 vehicles that might get into situations requiring a human to take over (ex. traffic is being directed by a human at a broken light). If you don't have a safety driver, the vehicle needs to be able to call home and ask for help with video feeds and no wi-fi connection.

The latter is for a bunch of concepts that are mostly drawing board-only right now, but could maybe become real in time. Ex., platooning, bypassing traffic signals entirely, automated redirection for traffic decongestion. The issue with this is that OEMs really don't want to pay for the data cost, so there's at least as much effort going into non-cellular alternatives as there is into 5G.

But in general, yes, articles stating 5G is necessary to autonomous driving are pretty far off the mark.

> OEMs really don't want to pay for the data cost

Isn’t a large part of the point of 5G the enablement of direct Device-to-Device (D2D) connections as a transparent optimization of device-to-tower-to-device links? That’s what is supposed to enable the platooning thing, IIRC.

I assumed that there would be no cost for 5G D2D data, since the carriers aren’t a part of the D2D conversation. Just like you’re not travelling over a particular 802.11 AP if you’re temporarily talking to a peer via an ad-hoc connection (e.g. AirDrop.)

Re: V2X. I believe that it can do a lot to increase safety and systems margins ultimately as autonomous and semi-autonomous vehicles become more pervasive.

Letting vehicles announce hazard plus their location and intentions can let other vehicles benefit.

I don't see how it can materially increase safety, as the vehicles will have to be designed to operate safely when 5G/V2X is unavailable.

Seems like a solution in search of a problem.

You don't think that e.g. alerting as to a wrong-way driver or that panic braking is starting before the car behind you could detect it from relative motion has any ability to add margin?

V2V is being discussed to augment human drivers and prevent them from rear-ending / improve collision mitigation systems. Why can't it help machines too.

> A car is going to need to make decisions without a connection so why does everyone go on about this all the time?

Because they don't know it?

I'd say even in such reputable establishment as Nikkei, airhead "tech journalists" are running the show.

People are afraid to show their incompetence, and switch to technical gobbledygook.

It comes up in virtually every article about automated cars, with zero explanation. If it really is just that journalists are too afraid to ask / don't know better, that's pretty damning for the entire tech journalist industry!

Sure, they can operate without 5G, but I can see strong benefits to having connectivity. Traffic, construction, road changes happen and fully automated needs to know where it’s going. It seems like a must to me, and definitely assumed for any future car platforms.

I'm not saying they don't need the internet, just that 5G is specifically for super high speed internet. Why can't 4G do all that? Or your home wifi connection at night?

First, all of that can be done with 4g.

Second, supposing you did need connectivity to have a self driving car, the challenge is having a good signal everywhere, not having 5g. You can't even get 4g everywhere.

Third I would never purchase a self driving car that requires any connectivity in order to work well.

> Third I would never purchase a self driving car that requires any connectivity in order to work well.

Communication is quite a key to solve the task, but I believe the entire framing of the problem is wrong: you don't need a self-driving car, you need an automatic road.

Why wont we write some feedback to Nikkei?

It comes up in virtually every article about automated cars, with zero explanation.

I've noticed that.

If you were just doing car to car communication for automatic driving and didn't need to send ads, track, or remotely update the car, the way to do it would be through the vehicle radars. That's how aircraft send altitude, ID, and anti-collision data back to the ground and to each other. Since you know the range and direction of the signal source, totally bogus signals are readily detected. Doesn't even need a permanent ID; choose a big random number at each start.

But no. The anal probe approach is essential to so many business models.

> If you were just doing car to car communication for automatic driving and didn't need to send ads, track, or remotely update the car, the way to do it would be through the vehicle radars.

In a vehicle application wouldn't that suffer heavily from line-of-sight problems?

That's the whole point. You only need info about things you can run into.

Likely due to an intentional misdirection campaign by the carriers poisoning the mindspace of journalists: https://pxlnv.com/linklog/5g-pixie-dust/

Automated driving will make collective decision making more plausible. For example, individual cars and particularly trucks outfitted with a coupling mechanism to form chains in order to reduce drag when driving on the highway.

Wouldn't it make more sense for them to connection locally somehow? Via wifi or something, rather than that go through a 5G tower, which might not have signal at that point anyway?

Still, you'll need network for authentication and receive push notification in case of rerouting, maybe public key crypto can help in some cases but network is always useful.

Also to stream Netflix to bored passengers or zoom while zooming about.

Netflix+Zoom for all the bored passengers is the single best answer I've heard to date about why this is needed! You really might need 5G (rather than 4G) for 5X Netflix streams.

You don't need collective decision making if you can just run something like an air trafic control for cars, and do it using a computer.

If you do so, the ennormous technical challenge of SLAM, and route planning becomes redundant, and obviated, completely removing the need in some magic AI.

But you can do so with a cheap, and reliable UHF link, and not a fancy 5G modem.

Imagine weaponizing the platooning feature using remote C&C mesh-swarms. It will have to be limited in features with regulation to prevent that from happening.

what do you mean, like linking many trucks as a type of kinetic bomb. Haven't thought about security implications, but that might be one. Safety is defiantly a reason to have some oversight over what the swarm is doing.

AI 8k+5G, duh.

Reply on news.ycombinator.com