How the heck does Google/Apple maps work?

105 views

I mean, seriously, how do they keep everything so close to perfect and how do they map even the smallest roads ( I imagine, satellites help, but satellites don’t know nor can read the name of some small roads/paths ) Not only that to mix into all that, the almost perfect voice directions.

In: 5

3 Answers

Anonymous 0 Comments

Various mapping services do not in general take satellite photographs in order to map roads and locations. In fact many of the images in Google Maps are actually sourced from aircraft photography and not satellites at all. But the point is that it isn’t the photographs which are used to make the maps, they could map without imagery at all.

Instead they source their information from public records. Roads don’t just appear out of nowhere, someone needs to build them and that someone in most cases is the local government. That local government knows where they built their road, often down to a couple centimeters. That information is public record and accessible by anyone who cares to bother with it; the public paid to make the roads, it is hardly a secret where they are.

Google and similar services then take those records and turn them into usable maps through their proprietary software, generating routes which can be read off by artificial voices. Training those artificial voices would be done from samples provided by a voice actor; the voice of Google Maps and Siri was trained from [Karen Elisabeth Jacobsen](https://upload.wikimedia.org/wikipedia/commons/6/68/Karen_Jacobsen_-_Smaller.jpg). (There is a Karen at your side at all times.)

Anonymous 0 Comments

It’s an accumulation of multiple datasets, waypoint information, roadmaps, sattelite and streetview imagery, 3D maps, height maps – acquired from companies they buy or make deals with. This happens constantly behind the scenes and it’s stuff that we, the user don’t even notice.

It is far from perfect though. Most of the data is the most accurate in bigger urbanized areas. That’s where it is most used of course and where the paying customer resides. If you use it in areas that are lesser inhabited and traveled, you will notice some discrepancies. This already happens in countries outside the US. I was in italy last week and had some minor issues using Apple maps. Seems like north italy uses quite old data. This is probably the case for many other, smaller countries, too.

So, yes, it is a big deal – and there is still work to do because map data constantly changes and companies are competing for the most (accurate) datasets.

Fun fact: there are still around 200 000 GPS related car accidents in the US (as of data from 2021)

Anonymous 0 Comments

The most accurate data is in the most heavily populated and most affluent areas. Here, a combination of vehicles on the street and airplanes and satellites provide the bulk of the data, mixed with high precision GPS, topological maps, building records, and of course public GIS (road and terrain maps).

The street view vehicles actually handle more of the heavy lifting than you might think. All those images are heavily processed with ML-driven computer vision to derive and correct a TON of detail that is unavailable from any other source. Have you wondered how it is that Apple knows exactly how many trees there are in your front yard, along with their size and exact position? Satellite imagery helps but it’s mostly vehicle data.

All of this also gets checked and augmented by people just using Maps. As you drive around, you’re almost always on the road, so if your GPS positions you far away from where the road was supposed to be, or a different speed, or a different direction, etc etc, that becomes data which can be used to surface and correct errors. Isolated data points are just noise, but in aggregate you can derive a lot of accuracy from the masses. This is also how real time traffic works.

So the short answer is really just this: data, and a lot of it, whether it’s images or GPS traces or public maps or anything else. Our tools for processing and understanding incomprehensibly vast quantities of data are so much better than most people realize. This is how Google knows exactly what ads to show you, where and when, and it’s also how they have centimeter precise maps of almost every accessible point on the globe.