Robots are part of the exciting new frontier of technology, but here is the challenge. The robot relies on arrays of sensors, external signals such as GPS and Wi-Fi, and customized software to navigate the environment. Additionally, robotics often include expensive, off-the-shelf hardware solutions, including built-in software and sensors designed for specific tasks, such as relative motion estimation. These products require complex integration and are limited to specific use cases.
As a result, most robots today cannot move in different locations, and only a small percentage of autonomous driving systems use AI for navigation.
However, Tera AI founder and CEO Tony Zhang believes that software known as robotic zero-shot navigation can overcome these obstacles.
At a high level, Tera AI is building a spatial inference AI system to provide affordable visual navigation for autonomous robots. This technology is used in a variety of applications, including robotics, mobile robotics, and autonomous driving.
“We will take a pure software platform-independent approach through over-the-air software updates that run on any robot with existing cameras and GPUs,” Zhang said in an interview with TechCrunch. “The system is cognitively inspired and can be applied to completely new scenarios during inference time, meaning it's a bit similar to a large-scale language model (LLM).”
After leading the machine learning initiative at Google X, Zhang founded San Francisco-based Tera in 2023, where he worked on developing and commercializing geospatial models. He earned his PhD from Caltech under Pietro Perona, a pioneer in computer vision who studied how biological systems solve navigation in a generic way.
The startup team includes AI and simulation researchers from Google AI, Caltech, MIT, and the European Space Agency.
While much of the AI industry focuses on LLMS, Zhang and his team have developed a new approach that allows AI to learn spatial reasoning independently. Spatial Inference AI allows machines to navigate, recognize objects and interact with three-dimensional space. General purpose navigation software that eliminates hardware constraints could further dramatically reduce costs and implementation time, making the robot 1,000 times more valuable, Zhang told TechCrunch.
“It could also enable new capabilities for existing robots in areas where autonomy is simply not possible due to sensor constraints,” he said.
For example, a $250,000 Waymo vehicle can offer a $50,000 localization sensor and a $100,000 LIDAR system. However, according to Tera AI, lightweight robots under $50,000 need a more affordable solution to navigate autonomously. Additionally, a high-precision GPS receiver can cost $10,000, while a top-class IMU (inertia measurement unit) can reach $30,000.
“Our important unique value proposition is that we are completely hardware dependent, meaning we focus on solving generic navigation in the pure software format of robots and new environments without having to readjust every time,” says Zhang. “For the first time in robotics, we can sell software that functions like an operating system, and mobile robotics platforms can make the most of that potential and offer promises to our customers.”
Startups test their products with a variety of US-based players in the robotics industry. The company's clients are robotic manufacturers who primarily have clients but face challenges when expanding their solutions to a variety of autonomous platforms, situations and environments.
This new funding will help Tera deploy its first solution to embedded devices this year and expand its technical team.
“We are looking at a future where software becomes the most valuable asset of robotic platforms. Once we realize that existing cameras already in our robots are sufficient for positioning and navigation, they will be able to deploy cheap robots on a large scale more quickly,” Zhang told TechCrunch. “In the end, we're assuming a future where new features can be installed just by clicking on downloads and booms, like the iOS app store. Robots have a whole new abilities.”
Tera's seed round investors include Felicis, Inovia, Caltech, Wilson Hill, and entrepreneurial investor Naval Ravikant.