GPS Leading Your Phone Astray? We Can Just Fix That In Code, Startup Claims

Interview A startup that's just emerged from stealth claims it's solved smartphone GPS positioning problems by adding software to triangulate a handset with "sub metre absolute accuracy."

Colorado-based Zephr has decloaked to tell the world about its "networked GPS" idea, which it claims in field trials showed sub-60cm GPS accuracy using nothing but existing satellites and mobile devices networked together to share correction data. The startup has also scored $3.5 million in a seed round led by Space Capital and First Spark Ventures.

"Similar to how with more satellite constellations (GPS, Galileo, GLONASS, Beidou) you can connect to the better position you can provide to a phone," Zephr cofounder and CEO Sean Gorman told The Register. "If you also have more measurements from more devices (within 10kms of each other) you have more data to solve the location equation." 

Anyone who's tried navigating an urban environment with smartphone GPS has likely experienced what's known as "multi-path errors," which occur when the signal from a GPS satellite reaches a receiver more than once. Multi-path errors often occur when signals bounce off buildings or other large objects, and can confuse smartphones with a single antenna. 

According to Zephr's marketing spiel, the special sauce is found in its SDK, which can be integrated into apps and, when running, effectively turns smartphones into a network of GPS base stations. 

Base stations are high-accuracy, stationary GPS receivers used in commercial applications to broadcast location correction to roving GPS units. Smartphones running Zephr's code can act as in unison to exchange error data as though they were a base station and use that to all get pinpoint accuracy for their locations.

"All these measurements go into an ensemble optimizer to find the best combination to then create an error correction for all the devices in a region," Gorman told us. That error correction data gets shared between smartphones with Zephr's software installed that are within 10 kilometres of each other, so the more devices the more accurate the whole network should become.

In real-world testing, which Zephr said it did in partnership with research institution SRI International, four smartphones with Zephr software were used and compared to a real-time kinematic (RTK) GPS device. RTK devices use a base station to make real-time corrections to maintain location accuracy to within a single centimetre. According to the company, its tests "generally" kept to within a 1-metre buffer, and "often within 50cm." 

Zephr said it needs around 10 to 15 devices sharing data within 10 kilometres of each other to get a marked improvement in GPS accuracy, meaning multi-path errors could be prevented thanks to multiple sources of data correction. 

Is this a privacy problem?

The thought of sharing location data to every Zephr-capable smartphone within 10 kilometres seems like a privacy nightmare. Even if anonymized, research has shown it's possible to fingerprint a device based on just a few Global Navigation Satellite System data points. 

Gorman tells us that's not an issue, because devices don't broadcast user locations - just error correction data to improve the locations of nearby devices. 

"The same error correction is sent to all devices … so there isn't anything personal about it or that requires an association with the user's ID or phone. We just need to know who is in the region to collect the correct raw measurements and send back the correction," the Zephr cofounder said.

• Telecoms networks could provide next-gen GPS services without the need for satellites
• Starlink decoded for use as GPS alternative – without Elon Musk's help
• Ancient with a dash of modern: We joined the Royal Navy to find there's little new in naval navigation
• NASA's latest AI will navigate the Moon using landmarks

Zephr has a couple pending patents for its tech, and has also licensed a patent from SRI, and plans to make its SDK available in the first half of next year, though Gorman said it could release in Q1 if the US biz meets its stretch goals. 

As for applications Zephr sees itself as useful for augmented reality, autonomous vehicles, delivery and ride-sharing services, search and rescue and other location-centric uses, though Gorman said there weren't any customers that could be shared at this point. It seems to us that Zephr wants to start off with phone apps and progress to other equipment and scenarios in which its SDK can be useful.

For the average person getting directions on their phone, however, the question remains whether sub-60 centimetre GPS accuracy is really worth adding another set of connected services. Anonymous data or not, there's still the security threat posed by yet another SDK to worry about, and those are decidedly harder to lock down, even for bigger firms. ®