← Glossary Term
Vision-language-action model
A robot control model that takes in camera images and spoken instructions and outputs the actual movements.
A vision-language-action model, usually shortened to VLA, does three jobs in one network: it sees through the robot’s cameras, understands an instruction in ordinary language, and produces the motor commands that move joints and grippers. The name is basically the explanation.
Before VLAs, those three parts were separate systems bolted together, and the joints between them were where most robots failed. Combining them is what makes it possible to say “put the mug in the sink” instead of scripting every centimetre of the motion. Google DeepMind’s Gemini Robotics and Nvidia’s Isaac GR00T are current examples.
Mentioned in