Comment by Quanttek | Hacker News Reader

Comment by Quanttek | Hacker News Reader

> The cluster of sensors, bolted underneath a chassis like a skid plate, can scan 10 feet beneath the ground to reveal soil, water, roots, and rocks. Once WaveSense has scanned an entire roadway, it creates a map of the subsurface strata that can determine the location of a vehicle within a few centimeters.

nononono! Even assuming you can create sufficient map data as soon as there is any kind of roadworks - or even only the winter cracking the asphahlt - your data will be useless. They don't even need to change the layout of the road. I'll never understand companies that focus so much on maps as priamry source of data. Every day, massive - temporary (construction site) or permanent - changes are occurring to the road network. Also, the centimeter-precise location of the vehicle doesn't matter that much. The actual situation on the road matters much more: You may need to drive a meter more to the left because of a parked truck, a vehicle may obscure signs, a tree branch hangs down into the road, some car is parked too close to intersection, etc.

Any time I drive a new road, I have to drive much more slowly - lanes unexpectedly turn into turning lanes resulting in quick merges, or a manhole has subsided a half foot making for an unpleasantly bumpy ride if I'm not used to it. Or in a rural environment, I have to spend more focus keeping an eye on the road surface for any number of reasons.

If I then drive the road and there's been some changes (traffic cones, lane closures, heavier traffic than normal, dark so I need to watch out for animals more closely....) I adjust my speed for the circumstances

I don't see why gathering surface information would be any different. Gather info, compare to previous fleet-gathered info, adjust driving to compensate. Seems reasonable to me (as ONE source of info for decision-making). And as fleet size increases, info gets gathered more quickly - look at how quickly google maps can inform you of traffic slowdowns in busy urban areas.

> I'll never understand companies that focus so much on maps as priamry source of data.

Maybe they don't really care about the product. Maybe the execs only care about making money, so they'll find funding for anything - whether it will pan out or not.

Self-driving cars are hot and for now there's a lot of money available. I think things will cool down when the hard part starts - like when they really do have to figure out how to make them drive in less than ideal conditions.

Depending on how good that data is, this might be a data play where the data collected in a single repository is extremely valuable. Sell it to the military for operating when GPS is jammed. Sell it to mining companies, insurance companies for a detailed very large map of what lies below.

That's what I was thinking. Worst case, the map data isn't all that critical for self-driving cars, but you have a huge amount of potentially very valuable data that can be used for a lot of other things.

That's an excellent point I had not considered.

It's very useful for navigation not accident avoidance. Cars needs to navigate without GPS in underground parking garages for example. Reading signs and inertial navigation are part of it, but knowing you just cross the same point is really important when trying to find your way out of a maze.

The highway value is when snow covers lane markers. The GPS will still work, but it's only so accurate without those lane markers etc.

> I'll never understand companies that focus so much on maps as primary source of data.

While not sufficient, this does play to one of the fundamental advantages computers have over people: you can copy information from one to the other.

Reply on news.ycombinator.com