Skip to content

Application Case: VLA Pick-and-Carry

Scenario: a tabletop VLA tidying task with the Walker S2 EDU humanoid robot. The robot picks parts from the table, places them into a storage box, and then carries the box to a distant target.

Approach:

  • Train a tabletop-tidying VLA model with the Thinker-IL-LAB toolchain.
  • Deploy with the inference server and hot-load the picking model; call the service with the inference client.
  • Build a task manager for robot navigation, carrying, and VLA picking to switch between multiple tasks and complete the full flow. For robot navigation and carrying, see the Walker S2 high-level motion interface.

Demo video

Further reading