Getting the bounding box of the grey floor in warehouse USD in Isaac Sim

[mona.jalal]

How can I programmatically get the bounding box of the complete grey floor (not just a couple tiles as shown in selection) using Python code?

I intend to use this bounding box so that when I am spawning random objects they are not spawn outside of the factory.

def main():
    # Open the environment in a new stage
    print(f"Loading Stage {ENV_URL}")
    open_stage(prefix_with_isaac_asset_server(ENV_URL))

    # Create a custom scope for newly added prims
    stage = get_current_stage()
    
     # Get the height of the pallet
    bb_cache = create_bbox_cache()
    curr_spawn_height = bb_cache.ComputeLocalBound(pallet_prim).GetRange().GetSize()[2] * 1.1
    print("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$")
    print('pallet get size: ', bb_cache.ComputeLocalBound(pallet_prim).GetRange().GetSize())

^ also, in above code, does bb_cache.ComputeLocalBound(pallet_prim).GetRange().GetSize() give me the bounding box of the pallet? For example, these are the numbers I got for pallet get size:

pallet get size: (1.213234950604253, 1.0028731312705617, 0.21111953263462624)

I was trying to use same to get the bounding box of the floor in the warehouse but I don’t know how to use a regular expression or better method to find the prim that belongs to the complete floor in the stage.

Please note that in the offline code, warehouse USD is my stage.

[toni.sm]

Hi @mona.jalal

What about using the compute_combined_aabb function for computing a combined AABB for a given a list of prim paths (in your case: floor tiles)

import omni.usd

import omni.isaac.core.utils.bounds as bounds_utils
import omni.isaac.core.utils.prims as prims_utils

stage = omni.usd.get_context().get_stage()
prims = [x.GetPath() for x in stage.Traverse() if "SM_floor" in x.GetName() and prims_utils.get_prim_type_name(x.GetPath()) == "Xform"]

bb_cache = bounds_utils.create_bbox_cache()
combined_range_arr = bounds_utils.compute_combined_aabb(bb_cache, prim_paths=prims)
print(combined_range_arr)

Another possible solution could be to add some objects (small visual spheres, e.g.) to specific positions (opposite corners, e.g.) to determine a working volume from the readings of their positions as show in the image

[mona.jalal]

Hi Toni,

Thanks a lot for your response.

def main():
    # Open the environment in a new stage
    print(f"Loading Stage {ENV_URL}")
    open_stage(prefix_with_isaac_asset_server(ENV_URL))

    # Create a custom scope for newly added prims
    stage = get_current_stage()
    
    floor_prims = [x.GetPath() for x in stage.Traverse() if "SM_floor" in x.GetName() and prims_utils.get_prim_type_name(x.GetPath()) == "Xform"]

    bb_cache = bounds_utils.create_bbox_cache()
    combined_range_arr = bounds_utils.compute_combined_aabb(bb_cache, prim_paths=floor_prims)
    print('combined_range_arr: ', combined_range_arr)

Why is it outputting 6 numbers instead of 4 numbers for the bounding box?

combined_range_arr:  [-2.80000155e+01 -2.34000366e+01 -8.58306866e-08  8.00001977e+00
  3.05999996e+01  1.43051144e-07]

I need this for

    `position=(random.uniform(-20, -2), random.uniform(-1, 3), 0)`

Also is it giving me the AABB for the rectangle or for the cube?

So, I need this information for forklift_prim because sometimes it is spawn outside of the warehouse. So need to figure what numbers to replace -20 and -2 programmatically for a stage where I randomize my objects.

    # Spawn a custom forklift at a random pose
    forklift_prim = prims.create_prim(
        prim_path=f"{SCOPE_NAME}/Custom_Forklift",
        position=(random.uniform(-20, -2), random.uniform(-1, 3), 0),
        orientation=euler_angles_to_quat([0, 0, random.uniform(0, math.pi)]),
        #orientation = np.array(euler_angles_to_quat([45, -60, 180], degrees=True)),
        #orientation=euler_angles_to_quat([random.uniform(0, math.pi), random.uniform(0, math.pi), random.uniform(0, math.pi)]),
        #orientation=euler_angles_to_quat(np.array([0, 90, 0], degrees=True)),
        #orientation=np.array(euler_angles_to_quat([0, 90, 0], degrees=True)),
        usd_path=CUSTOM_FORKLIFT_URL,
        semantic_label="Forklift",
    )

[mona.jalal]

Hi Toni,

based on the numbers I got, does the following sound correct to you?

    # position=(random.uniform(-20, -2), random.uniform(-1, 3), 0),
    position = (random.uniform(-25, 7), random.uniform(-23, 28), 0),

I have:

combined_range_arr:  [-2.80000155e+01 -2.34000366e+01 -8.58306866e-08  8.00001977e+00
  3.05999996e+01  1.43051144e-07]

[mona.jalal]

@toni.sm it seems what I have is wrong since stuff or the camera is being spawn out of the warehouse and shows the black stuff as the photo below

[toni.sm]

Hi @mona.jalal

According to the compute_combined_aabb documentation, the return value is NumPy array with the combined bounding box for input prims: [min x, min y, min z, max x, max y, max z]

---

@mona.jalal

I have:

combined_range_arr:  [-2.80000155e+01 -2.34000366e+01 -8.58306866e-08  8.00001977e+00
  3.05999996e+01  1.43051144e-07]

According to the documentation, then something like the following code should make sense.
use the margin variable to control the distance from bounding box limits

margin = 0.5
min_x, min_y, min_z, max_x, max_y, max_z = combined_range_arr
position = (np.random.uniform(min_x + margin, max_x - margin),  # x
            np.random.uniform(min_y + margin, max_y - margin),  # y
            0)                                                  # z

Keep in mind that the margin should be larger than the radius of the sphere that encapsulates the object, to prevent the object from going out of the box.

[toni.sm]

Hi @mona.jalal

@mona.jalal

@toni.sm it seems what I have is wrong since stuff or the camera is being spawn out of the warehouse and shows the black stuff as the photo below

I notice that some floor tiles in full_warehouse.usd file fall outside the warehouse walls. You need to take this into account for this scenario

d