This post is also available in:
עברית (Hebrew)
US-based Skild AI has introduced the Skild Brain, an artificial intelligence system designed to operate across a broad spectrum of robotic platforms. Unlike traditional robotic AI solutions that are tailored for specific machines or tasks, this new technology is built to support many types of robotic systems, offering a flexible foundation for various applications such as defense and tactical operations.
According to the company, the Skild Brain employs a hierarchical control architecture consisting of two levels. A low-frequency controller is responsible for higher-level functions like navigation and path planning. Meanwhile, a high-frequency controller converts these strategic decisions into precise motor commands, such as joint movements and torque control. This design enables robots to autonomously interpret and respond to their surroundings with improved spatial awareness and adaptability.
A notable safety feature integrated into the system limits the force exerted by robots, ensuring they can interact safely around humans and manage unexpected physical encounters, according to NextGenDefense. This allows Skild-powered robots to operate reliably in dynamic, real-world environments rather than controlled demonstration settings.
One of the challenges in robotics AI development has been the scarcity of large-scale, diverse action data needed to build comprehensive models. Many current approaches rely heavily on vision-and-language models (VLMs) with minimal real-world robotic data, limiting their ability to generalize beyond simple tasks. Skild AI addresses this by training its brain on extensive simulation data and real-world examples, which enhances its robustness and versatility.
The Skild Brain’s ability to generalize across tasks and hardware is strengthened by incorporating data from various robot forms, including human movement data, effectively broadening the training set. This omni-bodied intelligence approach supports greater resilience to hardware changes or failures.
With ongoing development and increased data exposure, the Skild Brain continues to evolve, improving its adaptability and operational reliability. The technology reflects a significant step towards creating a unified AI system capable of powering multiple robot types in real-world applications, moving closer to the goal of artificial general intelligence in robotics.
