Tera AI Emerges from Stealth with $7.8M for Robot Visual Navigation

# robotics # Tera AI

Robots are pushing the boundaries of technology, but they face some big hurdles. They depend heavily on a mix of sensors, external signals like GPS and Wi-Fi, and specialized software just to get around. Plus, many robots use pricey, off-the-shelf hardware that comes with its own software and sensors tailored for specific jobs, like figuring out how they're moving. This setup needs a lot of tweaking to work right and often only fits certain situations. Because of all this, most robots today can't switch between different places easily, and only a tiny fraction of self-driving systems use AI to find their way. But Tony Zhang, the founder and CEO of Tera AI, believes he's got a solution. His company is working on something called zero-shot navigation for robots, and investors are backing him up with a hefty $7.8 million in seed funding. In simple terms, Tera AI is developing a spatial reasoning AI system that can help robots navigate on the cheap. This tech can be used for all sorts of things, like moving objects, mobile robots, and even self-driving cars. "We're all about software that can be updated over the air and works with any robot that's already got a camera and a GPU," Zhang told TechCrunch. "Our system is inspired by how the brain works and can adapt to new situations on the fly, kind of like a big language model." Zhang started Tera in San Francisco in 2023 after working on machine learning at Google X. He got his PhD at Caltech under Pietro Perona, a big name in computer vision who looked at how living things navigate in a flexible way. Tera's team is made up of AI and simulation experts from Google AI, Caltech, MIT, and even the European Space Agency. While everyone else is chasing after language models, Zhang and his team are focusing on teaching AI to understand space. This kind of spatial reasoning lets machines move around, spot objects, and interact with the world in 3D. By getting rid of hardware limits, their general-purpose navigation software could cut costs and speed up deployment by a huge amount, making robots way more useful, Zhang says. "It could also unlock new tricks for robots that were stuck because of sensor limits," he added. For example, a Waymo car might splash out $250,000, which includes a $50,000 sensor for pinpointing its location and a $100,000 lidar system. But smaller robots that cost less than $50,000 need cheaper ways to navigate on their own, according to Tera AI. A fancy GPS receiver can set you back $10,000, and a top-notch IMU (Inertial Measurement Unit) might hit $30,000 — prices that make autonomous navigation a no-go for many smaller robots. "Our big selling point is that our software works with any hardware out there," Zhang said. "We're solving navigation in a way that works for any robot, in any new place, without needing to adjust it every time. It's like an operating system for robots, letting them do what they're meant to do and keep their customers happy." The startup has been testing its product with major U.S. robotics companies. Their main customers are robot makers who already have a customer base but struggle to adapt their solutions to different autonomous platforms, situations, and environments. With the new funding, Tera plans to roll out its first solution on small devices this year and beef up its tech team. "We see a future where software is the star of the show for robots," Zhang told TechCrunch. "Once people realize that the cameras already on robots are enough for navigation, they'll be able to put more robots to work faster and cheaper. We're dreaming of a day where, just like with an iOS app store, you can add new skills to your robot with a simple download." Tera's seed round was backed by investors like Felicis, Inovia, Caltech, Wilson Hill, and Naval Ravikant. > SoftBank to invest $500M in robotics startup Skild AI

Related article

Comments (35)