Lag in last_event.frame update after setObjectPoses()?

[anjaliwgoogta]

I have noticed that after an action is taken with setObjectPoses, the object metadata updates immediately and accordingly but the image, as queried through controller.last_event.frame and visualized with plot_frames() in the Ai2-THOR Colab, does not update until after another setObjectPoses action has been taken, like calling a dummy setObjectPoses (not changing any objects' position and rotation).

Am I mistaken about this? Is this a bug that the team is aware of?

Thank you! I include example code and images below.

`def rotate_translate(object_id, translation_vector_xz, degrees_to_rotate):
moveable_pickupable_objects = [obj for obj in controller.last_event.metadata["objects"] if obj["pickupable"] or obj["moveable"]]
object_metadata = controller.last_event.metadata["objects"][object_id]
object_name = object_metadata.get("name")
current_position = object_metadata.get("position")
current_rotation = object_metadata.get("rotation")

if not all([object_name, current_position, current_rotation]):
print(f"Error: Incomplete metadata for object. "
"Requires 'name', 'position', and 'rotation'. Current metadata: {object_metadata}")
else:
new_position = {
"x": current_position["x"] + translation_vector_xz["x"],
"y": current_position["y"],
"z": current_position["z"] + translation_vector_xz["z"]
}

  new_y_rotation = (current_rotation["y"] + degrees_to_rotate) % 360

  new_rotation = {
      "x": current_rotation["x"],
      "y": new_y_rotation,
      "z": current_rotation["z"]
  }

  object_pose_to_set = {
      "objectName": object_name,
      "position": new_position,
      "rotation": new_rotation
  }

  print(f"Attempting to translate '{object_name}' from "
        f"({current_position['x']:.2f}, {current_position['y']:.2f}, {current_position['z']:.2f}) to "
        f"({new_position['x']:.2f}, {new_position['y']:.2f}, {new_position['z']:.2f}).")

  print(f"Attempting to rotate '{object_name}' from "
    f"({current_rotation['x']:.2f}, {current_rotation['y']:.2f}, {current_rotation['z']:.2f}) to "
    f"({new_rotation['x']:.2f}, {new_rotation['y']:.2f}, {new_rotation['z']:.2f}).")

  # these objects will be removed from the scene unless otherwise specified 
  object_poses = [object_pose_to_set]
  for obj in moveable_pickupable_objects:
    if obj["name"] != object_name:
      object_poses.append(
          {"objectName": obj["name"],
          "position": obj["position"],
          "rotation": obj["rotation"]}
      )

  print(object_poses)

  controller.step(
      action='SetObjectPoses',
      objectPoses=object_poses
  )

  if controller.last_event.metadata["lastActionSuccess"]:
      print(f"Translation action for '{object_name}' was SUCCESSFUL.")
  else:
      print(f"Translation action for '{object_name}' FAILED!")
      print(f"Error Message: {controller.last_event.metadata['errorMessage']}")`

original image:

`object_id = 4 # the box to the left of the TV stand
translation_vector_xz = {"x": 0, "z": -0.5}
degrees_to_rotate = 0

rotate_translate(object_id, translation_vector_xz, degrees_to_rotate)
print(controller.last_event.metadata["objects"][4]['position'], controller.last_event.metadata["objects"][4]['rotation'])
plot_frames(controller.last_event.frame)`

Note the difference in the plotted frame before and after calling the additional setobjectposes below.

`# add dummy action so that image frame updates
rotate_translate(object_id, {"x": 0, "z": 0}, 0)

plot_frames(controller.last_event.frame)`