The non-embodied avatar object is structured like this:
Avatar
....SensorContainer (camera)
This is a holdover from earlier versions of TDW that emphasized using embodied avatars. For this reason, the Command API includes commands to rotate the avatar and, separately, to rotate the sensor container. The avatar rotation commands can be considered more or less deprecated; you should translate the avatar with teleport_avatar_to and rotate the camera with sensor container commands.
Option 1: Send the rotate_sensor_container_by command:
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
c = Controller()
object_id = c.get_unique_id()
commands = [TDWUtils.create_empty_room(12, 12),
c.get_add_object(model_name="iron_box",
library="models_core.json",
position={"x": 1, "y": 0, "z": -0.5},
object_id=object_id)]
commands.extend(TDWUtils.create_avatar(position={"x": 2, "y": 1.6, "z": -0.6},
look_at={"x": 1, "y": 0, "z": -0.5},
avatar_id="a"))
c.communicate(commands)
c.communicate({"$type": "rotate_sensor_container_by",
"axis": "pitch",
"angle": -10,
"avatar_id": "a"})
c.communicate([])
c.communicate({"$type": "terminate"})Option 2: With a ThirdPersonCamera, call the rotate() function (which internally creates a rotate_sensor_container_by command):
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
from tdw.add_ons.third_person_camera import ThirdPersonCamera
c = Controller()
object_id = c.get_unique_id()
cam = ThirdPersonCamera(position={"x": 2, "y": 1.6, "z": -0.6},
look_at=object_id)
c.add_ons.append(cam)
c.communicate([TDWUtils.create_empty_room(12, 12),
c.get_add_object(model_name="iron_box",
library="models_core.json",
position={"x": 1, "y": 0, "z": -0.5},
object_id=object_id)])
# Stop following the target object.
cam.look_at_target = None
cam.rotate({"x": -10, "y": 0, "z":0})
c.communicate([])
c.communicate({"$type": "terminate"})Send the rotate_sensor_container_to command or, with a ThirdPersonCamera, call rotate but supply an (x, y, z, w) dictionary instead of an (x, y, z) dictionary.
Quaternions are difficult to use, but they're the only reliable way to repeatedly rotate an object; all other rotation commands such as rotate_sensor_container_by convert parameters specified by the user into quaternions.
It is often use to rotate the camera to a quaternion if you want to store the camera's rotation and apply it later. To get the current quaternion of the camera, send send_image_sensors which will return ImageSensors output data:
from typing import List
import random
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
from tdw.output_data import OutputData, ImageSensors
class RotateSensorContainerToExample(Controller):
def get_random_rotation(self) -> List[dict]:
commands = []
for axis in ["pitch", "yaw", "roll"]:
commands.append({"$type": "rotate_sensor_container_by",
"axis": axis,
"angle": random.uniform(-15, 15),
"avatar_id": "a"})
return commands
def run(self):
commands = [TDWUtils.create_empty_room(12, 12)]
commands.extend(TDWUtils.create_avatar(position={"x": 2, "y": 1.6, "z": -0.6},
avatar_id="a"))
# Set a random initial rotation.
commands.extend(self.get_random_rotation())
# Request image sensors data.
commands.append({"$type": "send_image_sensors",
"frequency": "once"})
resp = self.communicate(commands)
initial_rotation = (0, 0, 0, 0)
for i in range(len(resp) - 1):
r_id = OutputData.get_data_type_id(resp[i])
if r_id == "imse":
imse = ImageSensors(resp[i])
if imse.get_avatar_id() == "a":
initial_rotation = imse.get_sensor_rotation(0)
# Set a random rotation.
self.communicate(self.get_random_rotation())
# Reset to the initial rotation.
self.communicate({"$type": "rotate_sensor_container_to",
"rotation": TDWUtils.array_to_vector4(initial_rotation),
"avatar_id": "a"})
self.communicate({"$type": "terminate"})
if __name__ == "__main__":
c = RotateSensorContainerToExample()
c.run()Reset the camera rotation to its neutral position with the reset_sensor_container_rotation:
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
c = Controller()
commands = [TDWUtils.create_empty_room(12, 12)]
commands.extend(TDWUtils.create_avatar(position={"x": 2, "y": 1.6, "z": -0.6},
avatar_id="a"))
c.communicate(commands)
c.communicate({"$type": "rotate_sensor_container_by",
"axis": "pitch",
"angle": 70,
"avatar_id": "a"})
c.communicate([{"$type": "reset_sensor_container_rotation",
"avatar_id": "a"}])
c.communicate({"$type": "terminate"})Send look_at to rotate the camera such that a target object is at the center of the image (the viewport). Set the "use_centroid" to True to look at the center of the object rather than its bottom-center pivot:
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
c = Controller()
object_id = c.get_unique_id()
commands = [TDWUtils.create_empty_room(12, 12),
c.get_add_object(model_name="iron_box",
library="models_core.json",
position={"x": 1, "y": 0, "z": -0.5},
object_id=object_id)]
commands.extend(TDWUtils.create_avatar(position={"x": 2, "y": 1.6, "z": -0.6},
avatar_id="a"))
commands.append({"$type": "look_at",
"object_id": object_id,
"use_centroid": True,
"avatar_id": "a"})
c.communicate(commands)
c.communicate({"$type": "terminate"})Send look_at_position to rotate the camera such that a target position is at the center of the image (the viewport). This is automatically called whenever we set the look_at parameter of TDWUtils.create_avatar(). This example creates an avatar and points the camera at a position without using the wrapper function:
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
c = Controller()
object_id = c.get_unique_id()
c.communicate([TDWUtils.create_empty_room(12, 12),
{"$type": "create_avatar",
"type": "A_Img_Caps_Kinematic",
"id": "a"},
{"$type": "look_at_position",
"avatar_id": "a",
"position": {"x": 0, "y": 10, "z": 0}}])
c.communicate({"$type": "terminate"})The ThirdPersonCamera wraps both of these commands (look_at and look_at_position in a function: camera.look_at(target). It will continue to look at the target every frame until you call camera.look_at(None).
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
from tdw.add_ons.third_person_camera import ThirdPersonCamera
c = Controller()
object_id = c.get_unique_id()
cam = ThirdPersonCamera(avatar_id="a", position={"x": 2, "y": 1.6, "z": -0.6})
c.add_ons.append(cam)
c.communicate([TDWUtils.create_empty_room(12, 12),
c.get_add_object(model_name="iron_box",
library="models_core.json",
position={"x": 1, "y": 0, "z": -0.5},
object_id=object_id)])
# Look at the object.
cam.look_at(object_id)
c.communicate([])
# Look at a position.
cam.look_at({"x": -2, "y": 0.5, "z": 0})
c.communicate([])
c.communicate({"$type": "terminate"})Next: Follow an object
Command API:
rotate_sensor_container_byrotate_sensor_container_tosend_image_sensorsreset_sensor_container_rotationlook_atlook_at_position
Output Data API:
Python API:
ThirdPersonCameraTDWUtils.array_to_vector4(arr)Convert an array to a vector4 dictionary (in this case, a quaternion).