Skip to main content

VLA Examples

This document demonstrates practical examples of Vision-Language-Action (VLA) applications.

Example 1: Object Recognition to Action​

  • Input: Image of a table with objects and a text prompt: "Pick up the red cup."
  • Process:
    1. The VLA model identifies objects in the image.
    2. Maps the instruction to the robot's action plan.
    3. Generates ROS 2 commands to execute the pick-and-place action.
  • Output: Robot picks up the red cup successfully.

Example 2: Scene Navigation​

  • Input: Image of a room and prompt: "Move to the door."
  • Process:
    1. Scene is parsed for obstacles and target location.
    2. VLA generates a path and controls the robot movement.
  • Output: Robot navigates safely to the door.

Note: These examples assume ROS 2 nodes are running and the humanoid robot is connected to the simulation or real hardware.