Can it understand the scene and plan the task?

The robot's eyes and planner: vision-language models that look at a scene, work out where things are and what can be done, and plan the steps.