Google DeepMind presents Gemini Robotics 2 as a generalist robotics modelA generalist robotics model is trained to support many physical tasks, objects and environments instead of one narrowly programmed behavior. intended to help one robot perform many different tasks in messy human environments. The model acts as the robot's brain, coordinating both broad body movement and precise actions by hands or grippers.
The release focuses on whole-body controlWhole-body robot control coordinates many joints and contact points together so a robot can balance, move and manipulate objects as one system. and dexterous manipulationDexterous robotic manipulation is precise, adaptable control of hands or grippers for objects that require coordinated motion, force and adjustment. beyond simple pick-and-place routines. Demonstrations such as screwing in a light bulb or handling a trash bag illustrate how robots must coordinate many joints and adapt their movements to small objects and changing physical conditions.
Gemini Robotics 2 also lets multiple robots collaborateMulti-robot collaboration is the coordinated use of two or more robots that communicate and divide work toward a shared objective. on one task while each robot runs its own copyDecentralized multi-robot control lets each robot make local decisions while coordinating with peers rather than receiving every action from one controller. of the model. Rather than relying on one network to control both machines, the robots reason independently and orchestrate their actions to collect tools, close a kit and return it to a bin. Background music and general promotional praise are omitted.
Watch on YouTube
