Google DeepMind has unveiled Gemini Robotics On-Device, a new artificial intelligence model designed specifically for robots. A key feature of this model is its ability to function without an internet connection, eliminating the need for robots to rely on the cloud or Wi-Fi for operation. This AI model empowers bi-arm robots with capabilities such as speech processing, language understanding, and action execution. It operates with low latency, making it suitable for real-time systems. In internal tests, the AI model demonstrated proficiency in tasks like folding clothes, opening and closing bags, identifying and picking up objects, and performing precision tasks such as industrial belt assembly. The model can understand human instructions in natural language and execute them accordingly. The system has been tested on various robotic platforms, including ALOHA, Franka Emika FR3, and Apptronik Apollo humanoid robots, showcasing its versatility across different robot types, even when handling novel or complex tasks. Currently, the system is accessible only to selected developers through Google’s Trusted Tester program, who can experiment with the Gemini Robotics SDK and MuJoCo physics simulator.
