Google DeepMind Expands Gemini Robotics AI to Control Full Humanoid Bodies
Google DeepMind has announced a significant update to its Gemini Robotics AI model, enabling it to control full humanoid robot bodies rather than just upper bodies as before.
Gemini Robotics 2 now supports whole-body motions spanning from feet to fingertips. This includes walking, crouching, stretching, and object manipulation—capabilities previously limited to specialized robotics systems.
In demonstrations, Apptronik's Apollo 2 humanoid robot performed tasks such as bending over to pick up a watering can and locating specific items on a shelf. These actions require coordinated movement across the robot's entire frame, something the expanded model now supports.
The advancement marks a step toward more versatile humanoid robots capable of performing a broader range of tasks in real-world environments.