Google Launches Gemini Robotics OnDevice for Offline, RealTime Robot Control
Gemini Robotics OnDevice, a small voicelanguageaction VLA AI model that operates solely on robots without internet access, was introduced by Google. I
Gemini Robotics OnDevice, a small voicelanguageaction VLA AI model that operates solely on robots without internet access, was introduced by Google. It was created by DeepMind and allows biarm robots, such as ALOHA, Franka FR3, and Apollo humanoids, to use natural language commands to carry out complicated tasks, such as making belts or folding clothes. Designed for offline, lowlatency settings, the model requires less computing power and can learn new jobs with only 50–100 demos. Even with invisible objects or multistep instructions, it competes with cloudbased models in generalisation and dexterity despite its small size. Developers can test it using Googles MuJoCo simulator and access it through a Gemini Robotics SDK. This is a significant step towards autonomous, realtime robotics in both home and industrial environments.
Read More..
B2B Tech News | 3 months ago