Google improves city GPS accuracy with machine learning

Have you ever experienced trying to navigate in a big city, but your map just won’t show you your actual position? An accurate GPS location can be difficult if you’re in a big city with skyscrapers and large buildings, as GPS systems are based on line-of-sight. Skyscrapers reflect GPS signals, causing the distance between your device and the satellites to differ from the intended, direct distance. This causes issues for location accuracy and means that you could end up on the wrong side of the street - or even on a wrong block. For location dependent services such as Uber or food deliveries, it means that a lot of time can be wasted trying to find the right location.

Obviously this can be critical, but it appears now that Google has finally found the answer to this unsolved GPS problem. This new method will increase accuracy by using Google Maps’ 3D building data to measure how buildings interfere with GPS signals. 

Google Maps has 3D models of buildings in more than 3850 big cities all over the world, and these are key in the new GPS accuracy update. While Google doesn’t go into much detail about the technical aspects, it appears that 3D building data along with GPS signals and machine learning can send a more accurate signal to phones to determine a more precise location for the user. 

Users of Google’s own phones, Pixel 5 and 4a, will be the first to experience the update as Google will roll out an update with a 75% accuracy improvement in December. At the same time, Google shares that Android users have already gotten a 50% improvement, and that a bigger update is coming in the first half of 2021. 

It’s also worth noting that while the improvement is already available in many major cities all over the world, the update for now only works while navigation by foot - however, Google claims that it’ll be available on more services in the first half of 2021.

We make sure to keep you updated on all navigation-related news from Google Maps on our blog which you can find in the link below.