[new299]

The number of negative comments here seems odd to me.

If you actually want practical and safe self driving cars widely deployed it seems obvious that instrumenting roads and making them a better platform for self-driving vehicles is an important part of this process.

To me this work seems like a part of the process of evolving roads from a Ad-Hoc and poorly documented system involving a lot of human guess work into a more robust and reliable platform for self-driving and human driven cars.

[latortuga]

> If you actually want practical and safe self driving cars widely deployed

I can't speak for everyone in this thread but personally this sounds like a nightmare. If we're dreaming about possible future worlds that are better than what we have, I'd rather have less or no cars. Much cheaper to maintain, not hackable.

[schoen]

I agree with that. But almost every cool convenience-enhancing or safety-enhancing "connected" technology has been implemented in a way that makes it easier to track individuals.

If we take aviation as an inspiration, where there are lots of great safety-enhancing uses of radio (for navigation, approach, air traffic control, giving information to autopilots, collision avoidance...), we also end up with "every vehicle can be publicly tracked in real time".

No one seems to have managed to get a "don't facilitate mass surveillance" bullet point into the requirements lists for the majority of transportation innovations. And if you don't have that requirement and you build a system using radio signals, then by default you typically do facilitate mass surveillance.

[philsnow]

> If you actually want practical and safe self driving cars widely deployed

I don't, though.

If we're going to propose a sci-fi future state of the world that will take a mind-boggling amount of investment, not to mention a giant leap of faith that we'll ever actually get there, I would prefer to reclaim all the space that's currently devoted to car infrastructure and be able to walk to everything.

> practical and safe

This isn't even enough; it would need to be cheap and universally accessible as well. I don't want to live in a society where we've agreed that cars are necessary despite a high and growing number of vehicle fatalities per year, and then provide miraculously-effective safety features [0] that only 1% (or 10% or whatever) of people can afford.

[0]

  if about_to_collide()
    dont();

[lm28469]

> The number of negative comments here seems odd to me.

Really ? Individual cars aren't sustainable, you can add more internet of shit in them it doesn't make anything better.

At the end of the day you're still moving 70kg of meat in a 2500kg cage of metal that cost my entire yearly net salary. All we're doing is making them more expensive, more failure prone, harder to repair, &c.

> To me this work seems like a part of the process of evolving roads from a Ad-Hoc and poorly documented system

This is a code monkey take, people in real life do not give a fuck about any of this. It's a road, just be sober, keep your eyes open and drive, it's really not that complex.

That's modern tech doing the only thing it knows, solutions looking for problems nobody has.

[drtgh]

As the article and the linked PDF quickly mention, cybersecurity is a concern, a really big problem difficult to solve.

A cracked traffic or car signal, a spoofed radio signal, or more simply a malfunctioning sensor from both, is something to watch out for. Then, at what point could the data received be trusted without a real trusted source like a visual of what is really happening?

Collapsing a city or causing an accident could be as simple as tricking vehicles into thinking they have another vehicle in front of them by receiving false data with the codes of legitimate vehicles or traffic signals for example.

IMHO vehicles should not react to data from third parties/external, but to a own -and mandatory redundant- sensoring data within the vehicle.

But even nowadays there are problems with this as owners of cars with automatic proximity braking systems could explain. There is also another problem, when the vehicle is connected to a network to receive an OTA or to modify any type of engineering parameter, it already has its own vector of attack, homologous to when one use the remote key to open and start the car, and the signal is captured and cracked by a third party; We didn't saw manufacturers solving this across all this years.

The article concludes like if the problem were political, a sabotage, but without explaining why the cybersecurity is a real problem.

I'm European, so I'm not sure what lobbies are involved there, for sure they exist, but if we ignore it and look at it from a technical point of view, IMHO the cybersecurity problem should be solved -which I'm not sure can be solved- before moving the money.

[adrianN]

We don’t even instrument all the train tracks, a small portion of the network relies on the conductors. I think it’s unlikely that the people commenting today will live to see a sizable portion of the road network instrumented for self driving.

[matsemann]

> If you actually want practical and safe self driving cars widely deployed

That's a big if ;)

Not to be a luddite, but we are many that don't enjoy our cities being designed around car usage. That they take up all space that could have been used for nicer things.

[surfingdino]

Thing is, those who like cars and driving don't want autonomous cars; those who only see cars as a way to transport humans and goods should stop pretending they want cars and simply use Uber or Rent-a-Van. Self-driving cars are a solution to a non-existent problem.

[tbrownaw]

> those who only see cars as a way to transport humans and goods should stop pretending they want cars and simply use Uber or Rent-a-Van

Amazingly enough, working out a cost-benefit calculation between renting on demand vs owning will in fact sometimes turn out in favor of owning.

[surfingdino]

Depends on your location. As the global population relocates to cities the cost-benefit maths work out in favour of not owning. New residential developments in Europe are planned and built without enough space to park at least one vehicle per apartment. Insurance, parking, and extra "emissions" charges (that EVs have to pay too), are all killing car ownership. There are places in London where you will not get a parking permit for your vehicle unless you meet certain criteria. The US was built for life where cars are an essential part of daily routine, Europe wasn't.