Improving urban GPS accuracy for your app Leave a comment


Posted by Frank van Diggelen, Principal Engineer and Jennifer Wang, Product Manager

At Android, we need to make it as straightforward as doable for builders to create probably the most useful apps for their customers. That’s why we goal to offer the very best location expertise with our APIs just like the Fused Location Provider API (FLP). However, we’ve heard from lots of you that the largest location concern is inaccuracy in dense urban areas, corresponding to wrong-side-of-the-street and even wrong-city-block errors.

This is especially essential for probably the most used location apps, corresponding to rideshare and navigation. For occasion, when customers request a rideshare car in a metropolis, apps can not simply find them due to the GPS errors.

The final nice unsolved GPS downside

This wrong-side-of-the-street place error is brought on by mirrored GPS indicators in cities, and we launched into an formidable venture to assist remedy this nice downside in GPS. Our answer makes use of 3D mapping aided corrections, and is just possible to be executed at scale by Google as a result of it includes 3D constructing fashions, uncooked GPS measurements, and machine studying.

The December Pixel Feature Drop provides 3D mapping aided GPS corrections to Pixel 5 and Pixel 4a (5G). With a system API that gives suggestions to the Qualcomm® Snapdragon™ 5G Mobile Platform that powers Pixel, the accuracy in cities (or “urban canyons”) improves spectacularly.

Picture of a pedestrian test, with Pixel 5 phone, walking along one side of the street, then the other. Yellow = Path followed, Red = without 3D mapping aided corrections, Blue = with 3D mapping aided corrections.  The picture shows that without 3D mapping aided corrections, the GPS results frequently wander to the wrong side of the street (or even the wrong city block), whereas, with 3D mapping aided corrections, the position is many times more accurate.

Picture of a pedestrian check, with Pixel 5 telephone, strolling alongside one aspect of the road, then the opposite. Yellow = Path adopted, Red = with out 3D mapping aided corrections, Blue = with 3D mapping aided corrections.

Why hasn’t this been solved earlier than?

The downside is that GPS constructively locates you within the unsuitable place if you find yourself in a metropolis. This is as a result of all GPS techniques are based mostly on line-of-sight operation from satellites. But in massive cities, most or all indicators attain you thru non line-of-sight reflections, as a result of the direct indicators are blocked by the buildings.

Diagram of the 3D mapping aided corrections module in Google Play services, with corrections feeding into the FLP API.   3D mapping aided corrections are also fed into the GNSS chip and software, which in turn provides GNSS measurements, position, and velocity back to the module.

The GPS chip assumes that the sign is line-of-sight and due to this fact introduces error when it calculates the surplus path size that the indicators traveled. The most typical aspect impact is that your place seems on the unsuitable aspect of the road, though your place may also seem on the unsuitable metropolis block, particularly in very giant cities with many skyscrapers.

There have been makes an attempt to handle this downside for greater than a decade. But no answer existed at scale, till 3D mapping aided corrections had been launched on Android.

How 3D mapping aided corrections work

The 3D mapping aided corrections module, in Google Play providers, consists of tiles of 3D constructing fashions that Google has for greater than 3850 cities around the globe. Google Play providers 3D mapping aided corrections at the moment helps pedestrian use-cases solely. When you employ your machine’s GPS whereas strolling, Android’s Activity Recognition API will acknowledge that you’re a pedestrian, and in case you are in one of many 3850+ cities, tiles with 3D fashions will likely be downloaded and cached on the telephone for that metropolis. Cache dimension is roughly 20MB, which is about the identical dimension as 6 pictures.

Inside the module, the 3D mapping aided corrections algorithms remedy the chicken-and-egg downside, which is: if the GPS place will not be in the best place, then how are you aware which buildings are blocking or reflecting the indicators? Having solved this downside, 3D mapping aided corrections present a set of corrected positions to the FLP. A system API then offers this data to the GPS chip to assist the chip enhance the accuracy of the following GPS repair.

With this December Pixel function drop, we’re releasing model 2 of 3D mapping aided corrections on Pixel 5 and Pixel 4a (5G). This reduces wrong-side-of-street occurrences by roughly 75%. Other Android telephones, utilizing Android eight or later, have model 1 carried out within the FLP, which reduces wrong-side-of-street occurrences by roughly 50%. Version 2 will likely be obtainable to the complete Android ecosystem (Android eight or later) in early 2021.

Android’s 3D mapping aided corrections work with indicators from the USA’s Global Positioning System (GPS) in addition to different Global Navigation Satellite Systems (GNSSs): GLONASS, Galileo, BeiDou, and QZSS.

Our GPS chip companions shared the significance of this work for their applied sciences:

“Consumers rely on the accuracy of the positioning and navigation capabilities of their mobile phones. Location technology is at the heart of ensuring you find your favorite restaurant and you get your rideshare service in a timely manner. Qualcomm Technologies is leading the charge to improve consumer experiences with its newest Qualcomm® Location Suite technology featuring integration with Google’s 3D mapping aided corrections. This collaboration with Google is an important milestone toward sidewalk-level location accuracy,” mentioned Francesco Grilli, vp of product administration at Qualcomm Technologies, Inc.

“Broadcom has integrated Google’s 3D mapping aided corrections into the navigation engine of the BCM47765 dual-frequency GNSS chip. The combination of dual frequency L1 and L5 signals plus 3D mapping aided corrections provides unprecedented accuracy in urban canyons. L5 plus Google’s corrections are a game-changer for GNSS use in cities,” mentioned Charles Abraham, Senior Director of Engineering, Broadcom Inc.

“Google’s 3D mapping aided corrections is a major advancement in personal location accuracy for smartphone users when walking in urban environments. MediaTek’s Dimensity 5G family enables 3D mapping aided corrections in addition to its highly accurate dual-band GNSS and industry-leading dead reckoning performance to give the most accurate global positioning ever for 5G smartphone users,” mentioned Dr. Yenchi Lee, Deputy General Manager of MediaTek’s Wireless Communications Business Unit.

How to entry 3D mapping aided corrections

Android’s 3D mapping aided corrections mechanically works when the GPS is being utilized by a pedestrian in any of the 3850+ cities, on any telephone that runs Android eight or later. The greatest manner for builders to make the most of the development is to make use of FLP to get location data. The additional 3D mapping aided corrections within the GPS chip can be found to Pixel 5 and Pixel 4a (5G) immediately, and will likely be rolled out to the remainder of the Android ecosystem (Android eight or later) within the subsequent a number of weeks. We may also quickly assist extra modes together with driving.

Android’s 3D mapping aided corrections cowl greater than 3850 cities, together with:

  • North America: All main cities in USA, Canada, Mexico.
  • Europe: All main cities. (100%, besides Russia & Ukraine)
  • Asia: All main cities in Japan and Taiwan.
  • Rest of the world: All main cities in Brazil, Argentina, Australia, New Zealand, and South Africa.

As our Google Earth 3D fashions develop, so will 3D mapping aided corrections protection.

Google Maps can be getting updates that can present extra avenue stage element for pedestrians in choose cities, corresponding to sidewalks, crosswalks, and pedestrian islands. In 2021, you may get these updates for your app utilizing the Google Maps Platform. Along with the improved location accuracy from 3D mapping aided corrections, we hope we may help builders such as you higher assist use instances for the world’s 2B pedestrians that use Android.

Continuously making location higher

In addition to 3D mapping aided corrections, we proceed to work onerous to make location as correct and helpful as doable. Below are the newest enhancements to the Fused Location Provider API (FLP):

  • Developers wished a neater strategy to retrieve the present location. With the brand new getCurrentLocation() API, builders can get the present location in a single request, slightly than having to subscribe to ongoing location modifications. By permitting builders to request location solely when wanted (and mechanically timing out and shutting open location requests), this new API additionally improves battery life. Check out our newest Kotlin sample.
  • Android 11’s Data Access Auditing API offers extra transparency into how your app and its dependencies entry personal information (like location) from customers. With the brand new assist for the API’s attribution tags within the FusedLocationProviderClient, builders can extra simply audit their apps’ location subscriptions along with common location requests. Check out this Kotlin sample to study extra.

Qualcomm and Snapdragon are emblems or registered emblems of Qualcomm Incorporated.

Qualcomm Snapdragon and Qualcomm Location Suite are merchandise of Qualcomm Technologies, Inc. and/or its subsidiaries.

Leave a Reply

Your email address will not be published. Required fields are marked *