Hacker News Reader

Comment by bluehawk | Hacker News Reader

Agreed. When theres 3 inches of snow on the road (before the plows have hit your neighborhood) and/or snow blowing there's zero chance of an autonomous car being able to decide where the lanes are.

View on HN

It is something that puzzles me that there is no-one talking about making lane markers reflective on a radio frequency that passes through water (liquid and frozen) without too much trouble and put a radar on those frquencies on the car. Is it because there is no such frequency (I doubt)? Or would that be for whatever reason prohitively expensive? Or something else, like difficulties in getting good enough resolution from radar or that there exist no such material that reflects those frequencies well enough?

Who is gonna pay for that?

We are, through the government, as usual.

Laying down some RF reflective paint/markers on the road wouldn't be nearly as expensive as other proposals (like embedding electronics or special wire underneath the road surface), it would be similar in expense to pavement markers[0] that are applied to some roads today, plus with a HUD it could even help human drivers in snow environments. Lots of places make special modifications to the road specifically for snow, and especially for snowplows.

[0]: http://www.highway-markers.com/pavement-reflectors/PM290.htm...

Many localities can't even be bothered to maintain their paint. It's been speculated that this is one reason why Tesla's Autopilot is so much better than Mercedes' supposedly equivalent system: Mercedes is testing their system on German roads with perfect markings everywhere, Tesla is testing theirs on California roads that haven't been painted since the Reagan administration.

I can give you a better test environment, Come here and test in India.

If you can get your car to self drive here, I assure you, your car will run anywhere on earth.

Now that you mention it, it will be really interesting to see what happens with self-driving cars in places like India.

I believe that should be solved with improvements in computer vision. We are probably decade or two away from parsing realtime 2x100 megapixel streams.

In winter driving, where the paint actually is matters very little.

> there's zero chance of an autonomous car being able to decide where the lanes are

Let alone the seemingly indefinite construction zones that have become commonplace across America - often with the lane markers missing or going off in directions that do not exist anymore (old lanes now blocked, etc...).

Is that right, have construction zones become more commonplace in highways and/or roads across America recently?

I had not heard about that but that sounds interesting and I would love to learn more. Any sources you can point me at?

I wouldn't say recently... it's a symptom of how the roads and highways were built in the US. They are essentially in a constant state of construction due to repaving needs, growing traffic congestion, etc.

What sort of "source" do you expect would exist for this? It's just the day-to-day job of DOT.

I guess I misunderstood your usage of "have become" to mean that this is a new state of affairs, and it didn't use to be this way in the past.

In the midwest, we have two season, Winter, and road construction. As soon as winter ends, road construction begins. Neverending road construction.

http://www.ll.mit.edu/news/Highly-accurate-vehicle-localizat...

The drawback with that approach is that it requires the road to have been previously scanned. So all roads would have to be scanned and added to a database, much like Streetview.

So it just needs to be driven in the summer first, then. Seriously, how do people navigate in whiteout conditions if they don't already know the terrain?

Reply on news.ycombinator.com