Tommoro Robotics has introduced HABILIS Brain 0, its proprietary robot foundation model (RFM). HABILIS Brain 0 is a Vision-Language-Action (VLA) model designed to understand task environments through visual and language inputs and generate robot actions.

HABILIS Brain 0 combines Geometry-Change Vision-Language-Action (GC-VLA), which learns geometric changes between robots and objects, with Guarded Corrective Residual Flow (GCRF), a reinforcement learning-based action correction method. Together, these technologies improve the accuracy and reliability of robotic manipulation.

On LIBERO, a simulation benchmark for robotic manipulation, GC-VLA achieved a 95.20% success rate, while the final model incorporating GCRF reached 99.55%. The research has also been released as a technical report, detailing the model architecture, training methodology, and evaluation results.

The company has also validated manipulation capabilities on physical robotic platforms. Rainbow Robotics’ RB-Y1 completed 400 consecutive logistics picking operations, while Universal Robots’ UR5e and ROBOTIS’ AI Worker demonstrated precision manipulation, repetitive task execution, and recovery from unexpected changes during operation.

Through the HABILIS Brain series, Tommoro Robotics is developing general-purpose robot intelligence that is not tied to a specific robot manufacturer or hardware platform. The company plans to expand deployment across diverse robotic systems and accelerate the commercialization of Physical AI through further validation in manufacturing and logistics environments.