Pick and Place composite task with SimReady assets#
This example uses the agentic environment-generation system to infer an environment graph spec from a prompt for pick and place composite task with SimReady assets on a table-top scene.
Environment Description#
The generated table-top scene: a DROID arm, a beverage can, a tuna can, a mini plastic basket, a hammer and a bean can. The task is to pick up the beverage can and bean can from the maple table and place them into the mini plastic basket.#
The task is composite: instead of a single pick and place, the robot picks up
both the beverage can and the bean can from the maple table and places them into
the mini plastic basket. Two atomic PickAndPlaceTask subtasks are combined
under one root task. A hammer and a tuna can are on the table as distractors.
Two of the assets (i.e. beverage can and plastic basket) are not in the Arena asset catalog. They are found by
asset search through the SimReady service and enter the spec as simready_usd_object entries carrying a usd_path.
The generated spec for this example is available at
isaaclab_arena_environments/maple_table_top/simready_droid_pick_place_cans_hammer_maple_table.yaml.
Workflow#
Prerequisites#
See Prerequisites for the environment and API key setup. The spec shown here depends on the model behind that endpoint — see Inference Model and Spec Quality.
Workflow Steps#
Follow the following steps to complete the workflow: