Skip to main content

We’ve all experienced the frustration of a GPS glitch causing us to miss a highway exit. According to the team at Tern AI, which is developing a low-cost GPS alternative, this is due to the limitations of current technology, which relies on satellite positioning.

Tern AI claims to have found a way to determine a vehicle’s position using only map information and the vehicle’s existing sensor data. The company’s pitch is that its system is cost-effective and doesn’t require any additional expensive sensors.

At SXSW, the Austin-based startup gave TechCrunch an exclusive demonstration of its technology, showing that it could “derive a position from nothing.”

“No triangulation, no satellites, no WiFi, nothing,” said Brett Harrison, co-founder and president of Tern AI, as Cyrus Behroozi, senior software developer, loaded the demo on his iPhone. “That’s really game-changing because, as we move away from triangulation-based technology, we have the ability to be fully off the grid.”

Harrison emphasized that this breakthrough is crucial for several reasons. From a commercial standpoint, companies that rely on GPS, such as ride-hailing apps and delivery companies, lose time, money, and gas when drivers have to double back due to faulty GPS positioning.

More importantly, critical systems like aviation, disaster response, and precision farming rely on GPS. Foreign adversaries have already demonstrated the ability to spoof GPS signals, which could have catastrophic consequences for the economy and national security.

The US has indicated that it wants to prioritize alternatives to GPS. During his first term, President Donald Trump signed an executive order to reduce reliance on a single source of PNT (positioning, navigation, and timing) services, such as GPS. There are also several other initiatives that direct agencies and bodies like the Department of Defense and the National Security Council to ensure resilient PNT by testing and integrating non-GPS technologies.

Harrison compared Tern AI’s achievement to that of DeepSeek, saying, “DeepSeek came out and said it cost them $6 million to do what it took other AI companies billions to do. To get a dot to move across a map in real-time in a vehicle, it took the government billions of dollars and a few decades. We did it with less than $2 million.”

Tern AI emerged from stealth in February 2024 and announced its $4.4 million seed round a few months later. This quick turnaround allowed the company to achieve the level of positioning demonstrated at SXSW.

Testing Tern’s system in Austin

To begin the demonstration, Behroozi connected his 2019 Honda Civic to his phone via Bluetooth, enabling the Tern application to access the vehicle’s existing sensor data. He noted that Tern’s technology can be integrated directly into vehicle models from 2009 and later.

Usually, Tern sets the position manually to speed up the process, but for this demo, the team wanted a “cold start.” Behroozi turned off his phone’s location services, leaving the Tern intelligent system with only a cached map of a 500-square-mile boundary around Austin and vehicle sensors to work with.

As the car drove, the system gathered road data to work towards “convergence.” It took approximately 10 minutes to reach full convergence from a cold start, but Harrison assured me that it usually takes around one to two minutes when a starting point is provided. The traffic we encountered slowed things down, Behroozi added.

Harrison noted that Tern’s system can also localize vehicles in parking garages, tunnels, and on mountains, which GPS struggles to do. However, he wouldn’t explain exactly how, citing proprietary information.

We drove for a few more minutes after the system reached full convergence, and I observed as it steadily tracked our precise movements in a way that appeared as good as, and in some cases better than, GPS. This became more apparent when we drove into downtown Austin, where my Google Maps regularly mislocated me throughout the week due to the towering buildings.

Harrison pointed out that Tern’s system is also safer from a privacy perspective because, with GPS, “if anyone knows your ID, they can find you at any time.”

“Our system is a total closed loop,” he said. “Right now, we’re not emitting anything. It’s independently deriving its own position via edge computing, so there are no external touchpoints.”

Built to scale

“We set up the company and the solution from the start to be scalable,” Harrison said, gesturing towards a Waymo-Uber robotaxi ahead of us. “If you look at that Waymo car and all the hardware embedded in it, we don’t see that going on a Nissan Sentra anytime soon. It’s just too expensive.”

“At the manufacturer level, if Tern is implemented within the infotainment system, it’s just a software download, making it extraordinarily scalable,” Harrison explained. “All new vehicles have the sensor data we need, and the map data already exists with all providers today. So, it’s quite simple.”

Tern’s potential future customers could be anyone from automakers to mobile phone manufacturers, from Google to Uber. Harrison said the startup is open to growing the company or being acquired.

“The primary thing is getting this out into the economy’s hands, with the growing threats and the emergence of tech that’s not realizing its full potential because of the limits of triangulation,” Harrison said.

He noted that Tern is exploring possibilities with the government, having recently received a contract award from the US Department of Transportation after demonstrating its technology alongside nine other companies from around the world.

“We’re hoping we did a good job of showing the government what’s possible now with American innovation,” Harrison said.


Source Link