How does Mapquest DO it???

Me too! I have worked with both of the aforementioned companies, both back in the 90s when they were known as “Navigation Technologies” and “Etak”, respectively. One of my roles in my career in this industry was, in fact, driving around and collecting data. At first, we used paper printouts, and as a pair (one driving, and the other mapping), we would drive every street in an assigned area, and capture this field data directly on the maps, which would then be delivered to our digital mapping department, where they would digitize our scribbles.

The base digital maps were usually standard data from the government acquired very cheaply, but then we would add value by updating with aerial photo data (for positioning), and then the field collected map data that could not be captured via aerial photo (such as turn restrictions, points of interest, traffic barriers, addresses in some cases, best routes, etc).

Back in the day I worked with the prototype GPS data collection units, where we would leave a breadcrumb trail of dots for repositioning and adding geometry (instead of the more expensive and less reliable aerial photos). This is the kind of thing you may see on the Google cars these days that both take photos and assure GPS accuracy.

As far as the routing, each segment of road had attributes such as address ranges, and a network code - the higher the code, the faster it was supposed to be - so you would have something like 1=interstate hwy, 2=US hwy/expressway, 3= arterial, 4=collector, 5= neighborhood street, etc. The programmer wrote code to use these network designators. I am sure it is much different now. And, as suranyi states…

It was a cool, fun, entrepreneurial environment where we felt like we were doing something really different and cool and new.

Current employee of one of the big mapping companies here. I work as a programmer and am quite familiar with the routing algorithms and techniques used, although my specialty (insofar as I have one) is specifically with transit systems, which are a bit different than road networks. I’ll second what suranyi and snowthx said, with some additions.

Additionally, what’s quite vexing is that all this stuff changes. In Chicago, they just finished some big work on Wacker Drive, connecting it differently with the Ike and all kinds of things. First we have to get that new geometry into our database somehow—there’s really no way around field collection for this kind of thing, even though field collection is hideously expensive. Then the changes go into a big bin of staged changes for the next release of our maps. They’re validated using an extensive set of heuristics to try to weed out errors, sent to another group of employees to test and validate further, and then shoved out to customers. What with all the portable and in-dash satnav systems out there, you’re bound to only reach a subset of your customers with the updated maps (even if you give the damn things out for free), which has a tendency to reduce customer satisfaction. Even when the customer is fully up-to-date there’s still some rot just because of the way road networks change. It’s a huge problem.

They haven’t had only mere cameras for a while now. Both Google and NAVTEQ/Nokia/Here also have LiDAR on their vehicles to collect 3D data, which can be used to do all kinds of cool things. (TeleAtlas may or may not have LiDAR; I honestly have no idea and a cursory Googling reveals nothing.)

This concept does exist at my company, although segments have a panoply of different attributes that can be applied. Ranges from the obvious (e.g. one way streets) to the slightly less obvious (e.g. truck restriction) to stuff that you’d have to be a map geek to even consider (e.g. roads with medians: You gonna make that two sets of segments, each one way, or one bidirectional set? How do you decide when to do which?)