A vision-language-action model is a robot policy that maps camera images and a natural-language instruction directly to motor actions in embodied AI.
Visual odometry is a robot perception method that estimates how a camera moves through space by tracking changes between consecutive camera images.