Robotics firm releases open-source AI tools

Silicon Valley startup X Square Robot has released its entire embodied AI stack as open-source software. The company states this decision could speed up development of general-purpose robots by allowing researchers and engineers to build on a shared foundation.

What the stack includes

The codebase, published on GitHub under an Apache 2.0 license, covers perception, motion planning, and low-level control for wheeled and legged robots. It also includes simulation environments built on NVIDIA Isaac Sim and a set of pre-trained vision and language models adapted for physical interaction.

The stack is designed to run on standard hardware, including NVIDIA Jetson Orin modules and Intel RealSense depth cameras. This approach aims to reduce costs for universities and smaller labs that lack resources for custom sensor setups.

One component, X-Motion, manages real-time collision avoidance in crowded areas. While the company hasn’t released full test data, it plans to publish a peer-reviewed paper later this year.

Reasons for the open-source release

The company explained in a press call that the decision responds to fragmentation in robotics. Teams often rebuild the same foundational tools, which slows progress. Standardizing these components could allow more focus on practical applications.

The timing matches a trend in AI research toward open models. Projects like Meta’s Llama and Stability AI’s Stable Diffusion have demonstrated how open-source releases can drive innovation, though they’ve also faced criticism for potential misuse and weak safeguards. Robotics, given its physical risks, may encounter even greater scrutiny.

Some researchers have already expressed concerns about the stack’s lack of built-in safety certifications. While the code includes a basic emergency-stop function, the company acknowledges that meeting industrial safety standards like ISO 13849 would require additional measures from users.

Early adoption and future plans

The project has seen early interest from the research community. The company will maintain the project with quarterly updates. A virtual hackathon next month will award prizes for the most creative use cases. Registration is open, though participants must share modifications under the same open-source license.

The company hasn’t revealed monetization plans but suggested a future enterprise support model similar to Red Hat’s Linux strategy.

The release arrives as embodied AI expands beyond industrial settings. Retailers like Walmart and Amazon are testing robots for inventory tasks, while hospitals explore their use in patient transport. A shared software foundation could simplify adapting robots to these different environments, though it’s unclear whether the open-source approach will work as well for hardware as it has for software.

Performance in dynamic, unstructured spaces will be a key test.

Most of X Square Robot’s demos occurred in controlled lab settings. Real-world deployment, especially in public areas, will likely uncover edge cases the current code doesn’t handle.