[atroyn]

Not sure why you're getting downvoted, this is a legitimate take.

My thinking is this is two-pronged. It won't make sense to make the infrastructure investment until there are sufficient proof points that this is 1.) something people want and 2.) something that's economically viable.

It's likely that the early geofenced version of passenger autonomy will demonstrate what further infrastructure is needed and how much it would cost.

One interesting side point; dedicated lanes for autonomous traffic are already being proposed on some roadways, particularly interstates in the U.S and highways in Europe. The economic benefits from autonomous logistics (e.g. trucking) are more readily capturable, so the infrastructure investment might make more sense there up front.

[trevyn]

He’s getting downvoted because not many people talk about “smart roadways” since it became clear that sensor-based machine learning autonomous vehicles were commercially feasible, and dare I say inevitable.

Good autonomous vehicles don’t work primarily by sensing and detecting the roadway, they work by sensing the environment and having a complete, detailed 3D map of drivable areas, matching their environment to the map as they go along. This only requires annotation of the terrain, not clear road markings or anything like that.

The 3D detail of the environment is the smart roadway. With always-on GPS and an incredibly detailed map of the environment stored on a honking hard drive in the trunk, there’s never a question where you are in the world.

In fact, now that I think about it, this makes reduced infrastructure possible, since you don’t even need things like signs or streetlights in an all-autonomous world, and can even get by with narrower lanes.

[atroyn]

I work in autonomous vehicle R&D, and if we had an environment filled with reliable sensors it would considerably simplify our problem.

There is no such thing as a _complete, detailed 3D map of driveable areas_. Most current efforts rely heavily on high resolution, large scale, semantically labeled maps, but on any given stretch of road this is only a first approximation of the environment. Live sensors embedded in infrastructure that could pass along real-time information and updates, particularly from directions our on-board sensors can't capture, would be very useful indeed.

And yes, road markings are part of the semantics we use.

[shaki-dora]

I don’t think the two of you are actually disagreeing.

[trevyn]

I also have worked in autonomous vehicle R&D, and you may wish to take my post as a friendly warning, then. :-)

[sundaeofshock]

So how much will it cost to upgrade 4.7 million miles of roadway in the US? How long will it take? What does the technology look like? What impact will it have on the economy to close huge swaths of road?

[atroyn]

These are all questions to be addressed, along with 'how many of those miles should we upgrade', 'what will we get out of it', and 'how do we make a self-driving car'.