4.6. Multiple Tasks

4.6.1. Learning Objectives

This tutorial describes how to add multiple tasks in Omniverse Isaac Sim. It explains how to add parameters to Tasks to scale your simulations. After this tutorial, you will have more experience building multi-task simulations in Omniverse Isaac Sim.

15-20 Minute Tutorial

4.6.2. Getting Started

Prerequisites

Begin with the source code open from the previous tutorial, Adding Multiple Robots.

4.6.3. Parameterizing Tasks

Adding parameters to tasks is a useful method to coordinate complex behaviors across multiple tasks. Start by adding an offset parameter to the task. This offset parameter will allow you to translate the assets of all tasks by the offset.

  1from omni.isaac.examples.base_sample import BaseSample
  2from omni.isaac.franka.tasks import PickPlace
  3from omni.isaac.franka.controllers import PickPlaceController
  4from omni.isaac.wheeled_robots.robots import WheeledRobot
  5from omni.isaac.core.utils.nucleus import get_assets_root_path
  6from omni.isaac.motion_generation import WheelBasePoseController
  7from omni.isaac.wheeled_robots.controllers.differential_controller import DifferentialController
  8from omni.isaac.core.tasks import BaseTask
  9from omni.isaac.core.utils.types import ArticulationAction
 10# Find a unique string name, to use it for prim paths and scene names
 11from omni.isaac.core.utils.string import find_unique_string_name        # Creates a unique prim path
 12from omni.isaac.core.utils.prims import is_prim_path_valid              # Checks if a prim path is valid
 13from omni.isaac.core.objects.cuboid import VisualCuboid
 14import numpy as np
 15
 16
 17class RobotsPlaying(BaseTask):
 18    # Adding offset to move the task objects with and the targets..etc
 19    def __init__(self, name, offset=None):
 20        super().__init__(name=name, offset=offset)
 21        self._task_event = 0
 22        self._jetbot_goal_position = np.array([1.3, 0.3, 0]) + self._offset #defined in the base task
 23        self._pick_place_task = PickPlace(cube_initial_position=np.array([0.1, 0.3, 0.05]),
 24                                        target_position=np.array([0.7, -0.3, 0.0515 / 2.0]),
 25                                        offset=offset)
 26        return
 27
 28    def set_up_scene(self, scene):
 29        super().set_up_scene(scene)
 30        #This will already translate the pick and place assets by the offset
 31        self._pick_place_task.set_up_scene(scene)
 32        # Find a unique scene name
 33        jetbot_name = find_unique_string_name(
 34            initial_name="fancy_jetbot", is_unique_fn=lambda x: not self.scene.object_exists(x)
 35        )
 36        # Find a unique prim path
 37        jetbot_prim_path = find_unique_string_name(
 38            initial_name="/World/Fancy_Jetbot", is_unique_fn=lambda x: not is_prim_path_valid(x)
 39        )
 40        assets_root_path = get_assets_root_path()
 41        jetbot_asset_path = assets_root_path + "/Isaac/Robots/Jetbot/jetbot.usd"
 42        self._jetbot = scene.add(
 43            WheeledRobot(
 44                prim_path=jetbot_prim_path,
 45                name=jetbot_name,
 46                wheel_dof_names=["left_wheel_joint", "right_wheel_joint"],
 47                create_robot=True,
 48                usd_path=jetbot_asset_path,
 49                position=np.array([0, 0.3, 0]),
 50            )
 51        )
 52        # Add Jetbot to this task objects
 53        self._task_objects[self._jetbot.name] = self._jetbot
 54        pick_place_params = self._pick_place_task.get_params()
 55        self._franka = scene.get_object(pick_place_params["robot_name"]["value"])
 56        # translate the franka by 100 in the x direction
 57        current_position, _ = self._franka.get_world_pose()
 58        self._franka.set_world_pose(position=current_position + np.array([1.0, 0, 0]))
 59        self._franka.set_default_state(position=current_position + np.array([1.0, 0, 0]))
 60        # This will only translate the task_objects by the offset specified (defined in the BaseTask)
 61        # Note: PickPlace task objects were already translated when setting up its scene
 62        self._move_task_objects_to_their_frame()
 63        return
 64
 65    def get_observations(self):
 66        current_jetbot_position, current_jetbot_orientation = self._jetbot.get_world_pose()
 67        observations= {
 68            "task_event": self._task_event,
 69            self._jetbot.name: {
 70                "position": current_jetbot_position,
 71                "orientation": current_jetbot_orientation,
 72                "goal_position": self._jetbot_goal_position
 73            }
 74        }
 75        observations.update(self._pick_place_task.get_observations())
 76        return observations
 77
 78    def get_params(self):
 79        pick_place_params = self._pick_place_task.get_params()
 80        params_representation = pick_place_params
 81        params_representation["jetbot_name"] = {"value": self._jetbot.name, "modifiable": False}
 82        params_representation["franka_name"] = pick_place_params["robot_name"]
 83        return params_representation
 84
 85    def pre_step(self, control_index, simulation_time):
 86        if self._task_event == 0:
 87            current_jetbot_position, _ = self._jetbot.get_world_pose()
 88            if np.mean(np.abs(current_jetbot_position[:2] - self._jetbot_goal_position[:2])) < 0.04:
 89                self._task_event += 1
 90                self._cube_arrive_step_index = control_index
 91        elif self._task_event == 1:
 92            if control_index - self._cube_arrive_step_index == 200:
 93                self._task_event += 1
 94        return
 95
 96    def post_reset(self):
 97        self._franka.gripper.set_joint_positions(self._franka.gripper.joint_opened_positions)
 98        self._task_event = 0
 99        return
100
101
102class HelloWorld(BaseSample):
103    def __init__(self) -> None:
104        super().__init__()
105        return
106
107    def setup_scene(self):
108        world = self.get_world()
109        world.add_task(RobotsPlaying(name="awesome_task", offset=np.array([0, -1.0, 0])))
110        # For visualization purposes only, so we don't need to add it to the scene
111        # Its there to only see where Franka was supposed to be vs after adding the offset
112        VisualCuboid(prim_path="/new_cube_1",
113                    name="visual_cube",
114                    position=np.array([1.0, 0, 0.05]),
115                    scale=np.array([0.1, 0.1, 0.1]))
116        return
117
118    async def setup_post_load(self):
119        self._world = self.get_world()
120        task_params = self._world.get_task("awesome_task").get_params()
121        self._franka = self._world.scene.get_object(task_params["franka_name"]["value"])
122        self._jetbot = self._world.scene.get_object(task_params["jetbot_name"]["value"])
123        self._cube_name = task_params["cube_name"]["value"]
124        self._franka_controller = PickPlaceController(name="pick_place_controller",
125                                                    gripper=self._franka.gripper,
126                                                    robot_articulation=self._franka)
127        self._jetbot_controller = WheelBasePoseController(name="cool_controller",
128                                                        open_loop_wheel_controller=
129                                                            DifferentialController(name="simple_control",
130                                                                                    wheel_radius=0.03, wheel_base=0.1125))
131        self._world.add_physics_callback("sim_step", callback_fn=self.physics_step)
132        await self._world.play_async()
133        return
134
135    async def setup_post_reset(self):
136        self._franka_controller.reset()
137        self._jetbot_controller.reset()
138        await self._world.play_async()
139        return
140
141    def physics_step(self, step_size):
142        current_observations = self._world.get_observations()
143        if current_observations["task_event"] == 0:
144            self._jetbot.apply_wheel_actions(
145                self._jetbot_controller.forward(
146                    start_position=current_observations[self._jetbot.name]["position"],
147                    start_orientation=current_observations[self._jetbot.name]["orientation"],
148                    goal_position=current_observations[self._jetbot.name]["goal_position"]))
149        elif current_observations["task_event"] == 1:
150            self._jetbot.apply_wheel_actions(ArticulationAction(joint_velocities=[-8.0, -8.0]))
151        elif current_observations["task_event"] == 2:
152            self._jetbot.apply_wheel_actions(ArticulationAction(joint_velocities=[0.0, 0.0]))
153            actions = self._franka_controller.forward(
154                picking_position=current_observations[self._cube_name]["position"],
155                placing_position=current_observations[self._cube_name]["target_position"],
156                current_joint_positions=current_observations[self._franka.name]["joint_positions"])
157            self._franka.apply_action(actions)
158        if self._franka_controller.is_done():
159            self._world.pause()
160        return

4.6.4. Scaling to Many Tasks

To scale to many tasks, you can add the same or different tasks in the world and decide where they operate in the world using the offset parameter. Each task will have its own controller.

  1from omni.isaac.examples.base_sample import BaseSample
  2from omni.isaac.core.utils.nucleus import get_assets_root_path
  3from omni.isaac.franka.tasks import PickPlace
  4from omni.isaac.franka.controllers import PickPlaceController
  5from omni.isaac.wheeled_robots.robots import WheeledRobot
  6from omni.isaac.motion_generation import WheelBasePoseController
  7from omni.isaac.wheeled_robots.controllers.differential_controller import DifferentialController
  8from omni.isaac.core.tasks import BaseTask
  9from omni.isaac.core.utils.types import ArticulationAction
 10from omni.isaac.core.utils.string import find_unique_string_name
 11from omni.isaac.core.utils.prims import is_prim_path_valid
 12import numpy as np
 13
 14
 15class RobotsPlaying(BaseTask):
 16    def __init__(self, name, offset=None):
 17        super().__init__(name=name, offset=offset)
 18        self._task_event = 0
 19        # Randomize the task a bit
 20        self._jetbot_goal_position = np.array([np.random.uniform(1.2, 1.6), 0.3, 0]) + self._offset
 21        self._pick_place_task = PickPlace(cube_initial_position=np.array([0.1, 0.3, 0.05]),
 22                                        target_position=np.array([0.7, -0.3, 0.0515 / 2.0]),
 23                                        offset=offset)
 24        return
 25
 26    def set_up_scene(self, scene):
 27        super().set_up_scene(scene)
 28        self._pick_place_task.set_up_scene(scene)
 29        jetbot_name = find_unique_string_name(
 30            initial_name="fancy_jetbot", is_unique_fn=lambda x: not self.scene.object_exists(x)
 31        )
 32        jetbot_prim_path = find_unique_string_name(
 33            initial_name="/World/Fancy_Jetbot", is_unique_fn=lambda x: not is_prim_path_valid(x)
 34        )
 35        assets_root_path = get_assets_root_path()
 36        jetbot_asset_path = assets_root_path + "/Isaac/Robots/Jetbot/jetbot.usd"
 37        self._jetbot = scene.add(
 38            WheeledRobot(
 39                prim_path=jetbot_prim_path,
 40                name=jetbot_name,
 41                wheel_dof_names=["left_wheel_joint", "right_wheel_joint"],
 42                create_robot=True,
 43                usd_path=jetbot_asset_path,
 44                position=np.array([0, 0.3, 0]),
 45            )
 46        )
 47        self._task_objects[self._jetbot.name] = self._jetbot
 48        pick_place_params = self._pick_place_task.get_params()
 49        self._franka = scene.get_object(pick_place_params["robot_name"]["value"])
 50        current_position, _ = self._franka.get_world_pose()
 51        self._franka.set_world_pose(position=current_position + np.array([1.0, 0, 0]))
 52        self._franka.set_default_state(position=current_position + np.array([1.0, 0, 0]))
 53        self._move_task_objects_to_their_frame()
 54        return
 55
 56    def get_observations(self):
 57        current_jetbot_position, current_jetbot_orientation = self._jetbot.get_world_pose()
 58        observations= {
 59            self.name + "_event": self._task_event, #change task event to make it unique
 60            self._jetbot.name: {
 61                "position": current_jetbot_position,
 62                "orientation": current_jetbot_orientation,
 63                "goal_position": self._jetbot_goal_position
 64            }
 65        }
 66        observations.update(self._pick_place_task.get_observations())
 67        return observations
 68
 69    def get_params(self):
 70        pick_place_params = self._pick_place_task.get_params()
 71        params_representation = pick_place_params
 72        params_representation["jetbot_name"] = {"value": self._jetbot.name, "modifiable": False}
 73        params_representation["franka_name"] = pick_place_params["robot_name"]
 74        return params_representation
 75
 76    def pre_step(self, control_index, simulation_time):
 77        if self._task_event == 0:
 78            current_jetbot_position, _ = self._jetbot.get_world_pose()
 79            if np.mean(np.abs(current_jetbot_position[:2] - self._jetbot_goal_position[:2])) < 0.04:
 80                self._task_event += 1
 81                self._cube_arrive_step_index = control_index
 82        elif self._task_event == 1:
 83            if control_index - self._cube_arrive_step_index == 200:
 84                self._task_event += 1
 85        return
 86
 87    def post_reset(self):
 88        self._franka.gripper.set_joint_positions(self._franka.gripper.joint_opened_positions)
 89        self._task_event = 0
 90        return
 91
 92
 93class HelloWorld(BaseSample):
 94    def __init__(self) -> None:
 95        super().__init__()
 96        # Add lists for tasks,
 97        self._tasks = []
 98        self._num_of_tasks = 3
 99        #  Add lists for controllers
100        self._franka_controllers = []
101        self._jetbot_controllers = []
102        # Add lists for variables needed for control
103        self._jetbots = []
104        self._frankas = []
105        self._cube_names = []
106        return
107
108    def setup_scene(self):
109        world = self.get_world()
110        for i in range(self._num_of_tasks):
111            world.add_task(RobotsPlaying(name="my_awesome_task_" + str(i), offset=np.array([0, (i * 2) - 3, 0])))
112        return
113
114    async def setup_post_load(self):
115        self._world = self.get_world()
116        for i in range(self._num_of_tasks):
117            self._tasks.append(self._world.get_task(name="my_awesome_task_" + str(i)))
118            # Get variables needed for control
119            task_params = self._tasks[i].get_params()
120            self._frankas.append(self._world.scene.get_object(task_params["franka_name"]["value"]))
121            self._jetbots.append(self._world.scene.get_object(task_params["jetbot_name"]["value"]))
122            self._cube_names.append(task_params["cube_name"]["value"])
123            # Define controllers
124            self._franka_controllers.append(PickPlaceController(name="pick_place_controller",
125                                                                gripper=self._frankas[i].gripper,
126                                                                robot_articulation=self._frankas[i],
127                                                                # Change the default events dt of the
128                                                                # pick and place controller to slow down some of the transitions
129                                                                # to pick up further blocks
130                                                                # Note: this is a simple pick and place state machine
131                                                                # based on events dt and not event success
132                                                                # check the different events description in the api
133                                                                # documentation
134                                                                events_dt=[0.008, 0.002, 0.5, 0.1, 0.05, 0.05, 0.0025, 1, 0.008, 0.08]))
135            self._jetbot_controllers.append(WheelBasePoseController(name="cool_controller",
136                                                                    open_loop_wheel_controller=
137                                                                      DifferentialController(name="simple_control",
138                                                                                             wheel_radius=0.03, wheel_base=0.1125)))
139        self._world.add_physics_callback("sim_step", callback_fn=self.physics_step)
140        await self._world.play_async()
141        return
142
143    async def setup_post_reset(self):
144        for i in range(len(self._tasks)):
145            # Reset all controllers
146            self._franka_controllers[i].reset()
147            self._jetbot_controllers[i].reset()
148        await self._world.play_async()
149        return
150
151    def physics_step(self, step_size):
152        current_observations = self._world.get_observations()
153        for i in range(len(self._tasks)):
154            # Apply actions for each task
155            if current_observations[self._tasks[i].name + "_event"] == 0:
156                self._jetbots[i].apply_wheel_actions(
157                    self._jetbot_controllers[i].forward(
158                        start_position=current_observations[self._jetbots[i].name]["position"],
159                        start_orientation=current_observations[self._jetbots[i].name]["orientation"],
160                        goal_position=current_observations[self._jetbots[i].name]["goal_position"]))
161            elif current_observations[self._tasks[i].name + "_event"] == 1:
162                self._jetbots[i].apply_wheel_actions(ArticulationAction(joint_velocities=[-8.0, -8.0]))
163            elif current_observations[self._tasks[i].name + "_event"] == 2:
164                self._jetbots[i].apply_wheel_actions(ArticulationAction(joint_velocities=[0.0, 0.0]))
165                actions = self._franka_controllers[i].forward(
166                    picking_position=current_observations[self._cube_names[i]]["position"],
167                    placing_position=current_observations[self._cube_names[i]]["target_position"],
168                    current_joint_positions=current_observations[self._frankas[i].name]["joint_positions"])
169                self._frankas[i].apply_action(actions)
170        return
171
172    # This function is called after a hot reload or a clear
173    # to delete the variables defined in this extension application
174    def world_cleanup(self):
175        self._tasks = []
176        self._franka_controllers = []
177        self._jetbot_controllers = []
178        self._jetbots = []
179        self._frankas = []
180        self._cube_names = []
181        return

4.6.5. Summary

This tutorial covered the following topics:

  1. Parameterizing Tasks

  2. Changing event velocities of the PickPlace controller

  3. Using world_cleanup to clear the extension application properly while hot reloading and creating a new stage.

  4. Choosing valid prim paths and scene names.

4.6.5.1. Further Learning

  1. See the standalone application at standalone_examples/api/omni.isaac.universal_robots/multiple_tasks.py for adding different tasks.