Comment by ckastner | Hacker News Reader

Comment by ckastner | Hacker News Reader

> Some experts now predict that we may never achieve the original dream of a reliable fully self-driving car. And even if we do, spreading the technology beyond the Arizona suburbs will require costly retrofits to add navigational hardware to existing roads. This is a non-starter in the post-pandemic age of austerity. Most state and local governments now struggle to merely maintain a much simpler, yet more fundamental technology for orienting both human and computer drivers—the painted markings that guide us along the pavement.

I don't buy this argument. Yes, retro-fitting costs money, but even if one were to lay down guide wires next to every major road, the costs wouldn't be as insurmountable as it is being painted here. The cost of digging up a ditch next to the road just isn't that bad.

Furthermore, it's not just costs to bear, but also benefits to weigh. If an AV capable road might get you from city to suburb, the suburb might just pay for the retrofit themselves.

I don't see the point of guidance wires along a road. When you're travelling at speed down a road, you need to know that the surface is flat and clear with achievable turn radius out to your stopping distance.

Guide wires don't tell you whether the surface is flat. The surface could be damaged/potholed, and the guide wires would still be there.

The guide wires don't tell you whether the surface is clear. For that, you need to have vision, to tell you whether cars, pedestrians, obstructions, and toddlers about to run out into the road are there.

The guide wires don't tell you what the turn radius in the road is ahead - it only tells you where the road is right here. Although technically this could be achieved, it would make it much more complicated.

Ultimately, the car needs to build its own picture of where it is safe to drive that is reasonably resilient to adverse conditions and even malicious attack. Trusting road infrastructure for that I think is not resilient enough.

In most population dense parts of the country where traffic is a problem, there is no “side of the road” to dig a ditch into. It’s all built up.

Even in the less dense areas, the side of the road is filled with buried utility lines (it’s cheap).

I’m also curious what the tolerances of these magical “wires alongside the road” have.

I thought of this, too, but believed the solution to be even simpler than rural areas. Dense areas are taken care of even more (snow is plowed, etc.). There's also, much more data being constantly captured.

For example, Google Maps can already hyper-accurately predict traffic down to individual road sections simply by observing the position and speed of Android phones on the road.

Sidenote, I love the idea of the "post-pandemic age of austerity", as if it contrasts with the "pre-pandemic age of austerity".

I doubt retrofitting will be necessary because it isn't necessary for human drivers. I think the author underestimates how good the technology is for LIDAR-aided visual navigation (and how quickly the technology is growing).

Can you expand on the costs? Any numbers to share?

Reply on news.ycombinator.com