Vector人工智慧機器人SDK使用筆記
Cozmo是2016年推出的，2兩年後的2018年Vector上市，具備語音助手和更多功能，元件數由300+升級到700+。
Vector的SDK具體說明在：developer.anki.com/vector/docs/。目前是測試版本
Vector人工智慧機器人SDK使用筆記
首先下載Vector的SDK（github）：
docs是文件，examples是示例，還有一些說明文件和安裝指令碼等。
SDK支援Windows、Linux和MacOS，具體安裝和使用流程參考官網。
由於是測試版本，示例還比較少，不如Cozmo豐富：
face_images放置可以在Vector面部顯示的圖案jpg、png，可以自定義。
apps是一些綜合應用程式；tutorials是教程。
1. tutorials教程
教程並沒有分章節，一共 13個：
1.1 hello world
"""Hello World
Make Vector say 'Hello World' in this simple Vector SDK example program.
"""
import anki_vector
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
print("Say 'Hello World'...")
robot.say_text("Hello World")
if __name__ == "__main__":
main()
對比Cozmo的hello world
'''Hello World
Make Cozmo say 'Hello World' in this simple Cozmo SDK example program.
'''
import cozmo
def cozmo_program(robot: cozmo.robot.Robot):
robot.say_text("Hello World").wait_for_completed()
cozmo.run_program(cozmo_program)
有些區別，需要注意哦～
1.2 drive square
"""Make Vector drive in a square.
Make Vector drive in a square by going forward and turning left 4 times in a row.
"""
import anki_vector
from anki_vector.util import degrees, distance_mm, speed_mmps
def main():
args = anki_vector.util.parse_command_args()
# The robot drives straight, stops and then turns around
with anki_vector.Robot(args.serial) as robot:
robot.behavior.drive_off_charger()
# Use a "for loop" to repeat the indented code 4 times
# Note: the _ variable name can be used when you don't need the value
for _ in range(4):
print("Drive Vector straight...")
robot.behavior.drive_straight(distance_mm(200), speed_mmps(50))
print("Turn Vector in place...")
robot.behavior.turn_in_place(degrees(90))
if __name__ == "__main__":
main()
1.3 motors
"""Drive Vector's wheels, lift and head motors directly
This is an example of how you can also have low-level control of Vector's motors
(wheels, lift and head) for fine-grained control and ease of controlling
multiple things at once.
"""
import time
import anki_vector
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
robot.behavior.drive_off_charger()
# Tell the head motor to start lowering the head (at 5 radians per second)
print("Lower Vector's head...")
robot.motors.set_head_motor(-5.0)
# Tell the lift motor to start lowering the lift (at 5 radians per second)
print("Lower Vector's lift...")
robot.motors.set_lift_motor(-5.0)
# Tell Vector to drive the left wheel at 25 mmps (millimeters per second),
# and the right wheel at 50 mmps (so Vector will drive Forwards while also
# turning to the left
print("Set Vector's wheel motors...")
robot.motors.set_wheel_motors(25, 50)
# wait for 3 seconds (the head, lift and wheels will move while we wait)
time.sleep(3)
# Tell the head motor to start raising the head (at 5 radians per second)
print("Raise Vector's head...")
robot.motors.set_head_motor(5)
# Tell the lift motor to start raising the lift (at 5 radians per second)
print("Raise Vector's lift...")
robot.motors.set_lift_motor(5)
# Tell Vector to drive the left wheel at 50 mmps (millimeters per second),
# and the right wheel at -50 mmps (so Vector will turn in-place to the right)
print("Set Vector's wheel motors...")
robot.motors.set_wheel_motors(50, -50)
# Wait for 3 seconds (the head, lift and wheels will move while we wait)
time.sleep(3)
# Stop the motors, which unlocks the tracks
robot.motors.set_wheel_motors(0, 0)
robot.motors.set_lift_motor(0)
robot.motors.set_head_motor(0)
if __name__ == "__main__":
main()
1.4 animation
"""Play an animation on Vector
"""
import anki_vector
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
robot.behavior.drive_off_charger()
# Play an animation via its name.
#
# Warning: Future versions of the app might change these, so for future-proofing
# we recommend using play_animation_trigger when it becomes available.
#
# See the remote_control.py example in apps for an easy way to see
# the available animations.
animation = 'anim_pounce_success_02'
print("Playing animation by name: " + animation)
robot.anim.play_animation(animation)
if __name__ == "__main__":
main()
1.5 play behaviors
"""Tell Vector to drive on and off the charger.
"""
import anki_vector
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
print("Drive Vector onto charger...")
robot.behavior.drive_on_charger()
print("Drive Vector off of charger...")
robot.behavior.drive_off_charger()
if __name__ == '__main__':
main()
1.6 face image
png--jpg
import os
import sys
import time
try:
from PIL import Image
except ImportError:
sys.exit("Cannot import from PIL: Do `pip3 install --user Pillow` to install")
import anki_vector
from anki_vector.util import degrees
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
# If necessary, move Vector's Head and Lift to make it easy to see his face
robot.behavior.set_head_angle(degrees(45.0))
robot.behavior.set_lift_height(0.0)
current_directory = os.path.dirname(os.path.realpath(__file__))
image_path = os.path.join(current_directory, "..", "face_images", "cozmo_image.jpg")
# Load an image
image_file = Image.open(image_path)
# Convert the image to the format used by the Screen
print("Display image on Vector's face...")
screen_data = anki_vector.screen.convert_image_to_screen_data(image_file)
robot.screen.set_screen_with_image_data(screen_data, 4.0)
time.sleep(5)
if __name__ == "__main__":
main()
1.7 dock with cube
"""Tell Vector to drive up to a seen cube.
This example demonstrates Vector driving to and docking with a cube, without
picking it up. Vector will line his arm hooks up with the cube so that they are
inserted into the cube's corners.
You must place a cube in front of Vector so that he can see it.
"""
import anki_vector
from anki_vector.util import degrees
def main():
args = anki_vector.util.parse_command_args()
docking_result = None
with anki_vector.Robot(args.serial) as robot:
robot.behavior.drive_off_charger()
# If necessary, move Vector's Head and Lift down
robot.behavior.set_head_angle(degrees(-5.0))
robot.behavior.set_lift_height(0.0)
print("Connecting to a cube...")
robot.world.connect_cube()
if robot.world.connected_light_cube:
print("Begin cube docking...")
dock_response = robot.behavior.dock_with_cube(
robot.world.connected_light_cube,
num_retries=3)
if dock_response:
docking_result = dock_response.result
robot.world.disconnect_cube()
if docking_result:
if docking_result.code != anki_vector.messaging.protocol.ActionResult.ACTION_RESULT_SUCCESS:
print("Cube docking failed with code {0} ({1})".format(str(docking_result).rstrip('\n\r'), docking_result.code))
else:
print("Cube docking failed.")
if __name__ == "__main__":
main()
1.8 drive to cliff and back up
"""Make Vector drive to a cliff and back up.
Place the robot about a foot from a "cliff" (such as a tabletop edge),
then run this script.
This tutorial is an advanced example that shows the SDK's integration
with the Vector behavior system.
The Vector behavior system uses an order of prioritizations to determine
what the robot will do next. The highest priorities in the behavior
system including the following:
* When Vector reaches a cliff, he will back up to avoid falling.
* When Vector is low on battery, he will start searching for his charger
and self-dock.
When the SDK is running at a lower priority level than high priorities
like cliff and low battery, an SDK program can lose its ability to
control the robot when a cliff if reached or when battery is low.
This example shows how, after reaching a cliff, the SDK program can
re-request control so it can continue controlling the robot after
reaching the cliff.
"""
import anki_vector
from anki_vector.util import distance_mm, speed_mmps
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
print("Vector SDK has behavior control...")
robot.behavior.drive_off_charger()
print("Drive Vector straight until he reaches cliff...")
# Once robot reaches cliff, he will play his typical cliff reactions.
robot.behavior.drive_straight(distance_mm(5000), speed_mmps(100))
robot.conn.run_coroutine(robot.conn.control_lost_event.wait()).result()
print("Lost SDK behavior control. Request SDK behavior control again...")
robot.conn.request_control()
print("Drive Vector backward away from the cliff...")
robot.behavior.drive_straight(distance_mm(-300), speed_mmps(100))
if __name__ == "__main__":
main()
1.9 show photo
"""Show a photo taken by Vector.
Grabs the pictures off of Vector and open them via PIL.
Before running this script, please make sure you have successfully
had Vector take a photo by saying, "Hey Vector! Take a photo."
"""
import io
import sys
try:
from PIL import Image
except ImportError:
sys.exit("Cannot import from PIL: Do `pip3 install --user Pillow` to install")
import anki_vector
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
if len(robot.photos.photo_info) == 0:
print('\n\nNo photos found on Vector. Ask him to take a photo first by saying, "Hey Vector! Take a photo."\n\n')
return
for photo in robot.photos.photo_info:
print(f"Opening photo {photo.photo_id}")
val = robot.photos.get_photo(photo.photo_id)
image = Image.open(io.BytesIO(val.image))
image.show()
if __name__ == "__main__":
main()
1.10 eye color
"""Set Vector's eye color.
"""
import time
import anki_vector
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial) as robot:
print("Set Vector's eye color to purple...")
robot.behavior.set_eye_color(hue=0.83, saturation=0.76)
print("Sleep 5 seconds...")
time.sleep(5)
if __name__ == '__main__':
main()
1.11 face event subscription
"""Wait for Vector to see a face, and then print output to the console.
This script demonstrates how to set up a listener for an event. It
subscribes to event 'robot_observed_face'. When that event is dispatched,
method 'on_robot_observed_face' is called, which prints text to the console.
Vector will also say "I see a face" one time, and the program will exit when
he finishes speaking.
"""
import functools
import threading
import anki_vector
from anki_vector.events import Events
from anki_vector.util import degrees
said_text = False
def main():
evt = threading.Event()
def on_robot_observed_face(robot, event_type, event):
print("Vector sees a face")
global said_text
if not said_text:
said_text = True
robot.say_text("I see a face!")
evt.set()
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial, enable_face_detection=True) as robot:
# If necessary, move Vector's Head and Lift to make it easy to see his face
robot.behavior.set_head_angle(degrees(45.0))
robot.behavior.set_lift_height(0.0)
on_robot_observed_face = functools.partial(on_robot_observed_face, robot)
robot.events.subscribe(on_robot_observed_face, Events.robot_observed_face)
print("------ waiting for face events, press ctrl+c to exit early ------")
try:
if not evt.wait(timeout=5):
print("------ Vector never saw your face! ------")
except KeyboardInterrupt:
pass
robot.events.unsubscribe(on_robot_observed_face, Events.robot_observed_face)
if __name__ == '__main__':
main()
1.12 wake word subscription
"""Wait for Vector to hear "Hey Vector!" and then play an animation.
The wake_word event only is dispatched when the SDK program has
not requested behavior control. After the robot hears "Hey Vector!"
and the event is received, you can then request behavior control
and control the robot. See the 'requires_behavior_control' method in
connection.py for more information.
"""
import functools
import threading
import anki_vector
from anki_vector.events import Events
wake_word_heard = False
def main():
evt = threading.Event()
def on_wake_word(robot, event_type, event):
robot.conn.request_control()
global wake_word_heard
if not wake_word_heard:
wake_word_heard = True
robot.say_text("Hello")
evt.set()
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial, requires_behavior_control=False, cache_animation_list=False) as robot:
on_wake_word = functools.partial(on_wake_word, robot)
robot.events.subscribe(on_wake_word, Events.wake_word)
print('------ Vector is waiting to hear "Hey Vector!" Press ctrl+c to exit early ------')
try:
if not evt.wait(timeout=10):
print('------ Vector never heard "Hey Vector!" ------')
except KeyboardInterrupt:
pass
if __name__ == '__main__':
main()
1.13 custom objects
"""This example demonstrates how you can define custom objects.
The example defines several custom objects (2 cubes, a wall and a box). When
Vector sees the markers for those objects he will report that he observed an
object of that size and shape there.
You can adjust the markers, marker sizes, and object sizes to fit whatever
object you have and the exact size of the markers that you print out.
"""
import time
import anki_vector
from anki_vector.objects import CustomObjectMarkers, CustomObjectTypes
def handle_object_appeared(event_type, event):
# This will be called whenever an EvtObjectAppeared is dispatched -
# whenever an Object comes into view.
print(f"--------- Vector started seeing an object --------- \n{event.obj}")
def handle_object_disappeared(event_type, event):
# This will be called whenever an EvtObjectDisappeared is dispatched -
# whenever an Object goes out of view.
print(f"--------- Vector stopped seeing an object --------- \n{event.obj}")
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial,
default_logging=False,
show_viewer=True,
show_3d_viewer=True,
enable_camera_feed=True,
enable_custom_object_detection=True,
enable_nav_map_feed=True) as robot:
# Add event handlers for whenever Vector sees a new object
robot.events.subscribe(handle_object_appeared, anki_vector.events.Events.object_appeared)
robot.events.subscribe(handle_object_disappeared, anki_vector.events.Events.object_disappeared)
# define a unique cube (44mm x 44mm x 44mm) (approximately the same size as Vector's light cube)
# with a 50mm x 50mm Circles2 image on every face. Note that marker_width_mm and marker_height_mm
# parameter values must match the dimensions of the printed marker.
cube_obj = robot.world.define_custom_cube(custom_object_type=CustomObjectTypes.CustomType00,
marker=CustomObjectMarkers.Circles2,
size_mm=44.0,
marker_width_mm=50.0,
marker_height_mm=50.0,
is_unique=True)
# define a unique cube (88mm x 88mm x 88mm) (approximately 2x the size of Vector's light cube)
# with a 50mm x 50mm Circles3 image on every face.
big_cube_obj = robot.world.define_custom_cube(custom_object_type=CustomObjectTypes.CustomType01,
marker=CustomObjectMarkers.Circles3,
size_mm=88.0,
marker_width_mm=50.0,
marker_height_mm=50.0,
is_unique=True)
# define a unique wall (150mm x 120mm (x10mm thick for all walls)
# with a 50mm x 30mm Triangles2 image on front and back
wall_obj = robot.world.define_custom_wall(custom_object_type=CustomObjectTypes.CustomType02,
marker=CustomObjectMarkers.Triangles2,
width_mm=150,
height_mm=120,
marker_width_mm=50,
marker_height_mm=30,
is_unique=True)
# define a unique box (20mm deep x 20mm width x20mm tall)
# with a different 50mm x 50mm image on each of the 6 faces
box_obj = robot.world.define_custom_box(custom_object_type=CustomObjectTypes.CustomType03,
marker_front=CustomObjectMarkers.Diamonds2, # front
marker_back=CustomObjectMarkers.Hexagons2, # back
marker_top=CustomObjectMarkers.Hexagons3, # top
marker_bottom=CustomObjectMarkers.Hexagons4, # bottom
marker_left=CustomObjectMarkers.Triangles3, # left
marker_right=CustomObjectMarkers.Triangles4, # right
depth_mm=20.0,
width_mm=20.0,
height_mm=20.0,
marker_width_mm=50.0,
marker_height_mm=50.0,
is_unique=True)
if ((cube_obj is not None) and (big_cube_obj is not None) and
(wall_obj is not None) and (box_obj is not None)):
print("All objects defined successfully!")
else:
print("One or more object definitions failed!")
return
print("\n\nShow a marker specified in the Python script to Vector and you will see the related 3d objects\n"
"display in Vector's 3d_viewer window. You will also see messages print every time a custom object\n"
"enters or exits Vector's view. Markers can be found from the docs under CustomObjectMarkers.\n\n")
try:
while True:
time.sleep(0.5)
except KeyboardInterrupt:
pass
if __name__ == "__main__":
main()
2. apps
分為四個部分，每個部分只有一個示例：
2.1 3d viewer
"""3d Viewer example, with remote control.
This is an example of how you can use the 3D viewer with a program, and the
3D Viewer and controls will work automatically.
"""
import time
import anki_vector
def main():
args = anki_vector.util.parse_command_args()
with anki_vector.Robot(args.serial,
show_viewer=True,
enable_camera_feed=True,
show_3d_viewer=True,
enable_face_detection=True,
enable_custom_object_detection=True,
enable_nav_map_feed=True):
print("Starting 3D Viewer. Use Ctrl+C to quit.")
try:
while True:
time.sleep(0.5)
except KeyboardInterrupt:
pass
if __name__ == "__main__":
main()
2.2 interactive shell
"""Command Line Interface for Vector
This is an example of integrating Vector with an ipython-based command line interface.
"""
import sys
try:
from IPython.terminal.embed import InteractiveShellEmbed
except ImportError:
sys.exit('Cannot import from ipython: Do `pip3 install ipython` to install')
import anki_vector
usage = """Use the [tab] key to auto-complete commands, and see all available methods and properties.
For example, type 'robot.' then press the [tab] key and you'll see all the robot functions.
Keep pressing tab to cycle through all of the available options.
All IPython commands work as usual.
Here's some useful syntax:
robot? -> Details about 'robot'.
robot?? -> More detailed information including code for 'robot'.
These commands will work on all objects inside of the shell.
You can even call the functions that send messages to Vector, and he'll respond just like he would in a script.
Try it out! Type:
robot.anim.play_animation('anim_pounce_success_02')
"""
args = anki_vector.util.parse_command_args()
ipyshell = InteractiveShellEmbed(banner1='\nWelcome to the Vector Interactive Shell!',
exit_msg='Goodbye\n')
if __name__ == "__main__":
with anki_vector.Robot(args.serial,
enable_camera_feed=True,
show_viewer=True) as robot:
# Invoke the ipython shell while connected to Vector
ipyshell(usage)
2.3 proximity mapper
"""Maps a region around Vector using the proximity sensor.
Vector will turn in place and use his sensor to detect walls in his
local environment. These walls are displayed in the 3D Viewer. The
visualizer does not effect the robot's internal state or behavior.
Vector expects this environment to be static - if objects are moved
he will have no knowledge of them.
"""
import asyncio
import concurrent
from math import cos, sin, inf, acos
import os
import sys
sys.path.append(os.path.join(os.path.dirname(__file__), 'lib'))
from proximity_mapper_state import ClearedTerritory, MapState, Wall, WallSegment # pylint: disable=wrong-import-position
import anki_vector # pylint: disable=wrong-import-position
from anki_vector.util import parse_command_args, radians, degrees, distance_mm, speed_mmps, Vector3 # pylint: disable=wrong-import-position
# Constants
#: The maximum distance (in millimeters) the scan considers valid for a proximity respons.
#: Wall detection past this threshold will be disregarded, and an 'open' node will
#: be created at this distance instead. Increasing this value may degrade the
#: reliability of this program, see note below:
#:
#: NOTE: The proximity sensor works by sending a light pulse, and seeing how long that pulse takes
#: to reflect and return to the sensor. The proximity sensor does not specifically have a maximum
#: range, but will get unreliable results below a certain return signal strength. This return signal
#: is impacted by environmental conditions (such as the orientation and material of the detected obstacle)
#: as well as the distance. Additionally, increasing this radius will reduce the resolution of contact
#: points, necessitating changes to PROXIMITY_SCAN_SAMPLE_FREQUENCY_HZ and PROXIMITY_SCAN_BIND_THRESHOLD_MM
#: to maintain effective wall prediction.
PROXIMITY_SCAN_DISTANCE_THRESHOLD_MM = 300
#: The distance (in millimeters) to place an open node if no proximity results are detected along
#: a given line of sight. This should be smaller than the distance threshold, since these nodes
#: indicate safe points for the robot to drive to, and the robot's size should be taken into account
#: when estimating a maximum safe driving distance
PROXIMITY_SCAN_OPEN_NODE_DISTANCE_MM = 230
#: How frequently (in hertz) the robot checks proximity data while doing a scan.
PROXIMITY_SCAN_SAMPLE_FREQUENCY_HZ = 15.0
#: How long (in seconds) the robot spends doing it's 360 degree scan.
PROXIMITY_SCAN_TURN_DURATION_S = 10.0
#: How close (in millimeters) together two detected contact points need to be for the robot to
#: consider them part of a continuous wall.
PROXIMITY_SCAN_BIND_THRESHOLD_MM = 30.0
#: A delay (in seconds) the program waits after the scan finishes before shutting down.
#: This allows the user time to explore the mapped 3d environment in the viewer and can be
#: Tuned to any desired length. A value of 0.0 will prevent the viewer from closing.
PROXIMITY_EXPLORATION_SHUTDOWN_DELAY_S = 8.0
# @TODO enable when testing shows it is ready to go
#: ACTIVELY_EXPLORE_SPACE can be activated to allow the robot to move
#: into an open space after scanning, and continue the process until all open
#: spaces are explored.
ACTIVELY_EXPLORE_SPACE = True
#: The speed (in millimeters/second) the robot drives while exploring.
EXPLORE_DRIVE_SPEED_MMPS = 40.0
#: The speed (in degrees/second) the robot turns while exploring.
EXPLORE_TURN_SPEED_DPS = 90.0
#: Takes a position in 3d space where a collection was detected, and adds it to the map state
#: by either creating a wall, adding to wall or storing a loose contact point.
async def add_proximity_contact_to_state(node_position: Vector3, state: MapState):
# Comparison function for sorting points by distance.
def compare_distance(elem):
return (elem - node_position).magnitude_squared
# Comparison function for sorting walls by distance using their head as a reference point.
def compare_head_distance(elem):
return (elem.vertices[0] - node_position).magnitude_squared
# Comparison function for sorting walls by distance using their tail as a reference point.
def compare_tail_distance(elem):
return (elem.vertices[-1] - node_position).magnitude_squared
# Sort all the loose contact nodes not yet incorporated into walls by
# their distance to our reading position. If the nearest one is within
# our binding threshold - store it as a viable wall creation partner.
# (infinity is used as a standin for 'nothing')
closest_contact_distance = inf
if state.contact_nodes:
state.contact_nodes.sort(key=compare_distance)
closest_contact_distance = (state.contact_nodes[0] - node_position).magnitude
if closest_contact_distance > PROXIMITY_SCAN_BIND_THRESHOLD_MM:
closest_contact_distance = inf
# Sort all the walls both by head and tail distance from our sample
# if either of the results are within our binding threshold, store them
# as potential wall extension candidates for our sample.
# (infinity is used as a standin for 'nothing')
closest_head_distance = inf
closest_tail_distance = inf
if state.walls:
state.walls.sort(key=compare_tail_distance)
closest_tail_distance = (state.walls[0].vertices[-1] - node_position).magnitude
if closest_tail_distance > PROXIMITY_SCAN_BIND_THRESHOLD_MM:
closest_tail_distance = inf
state.walls.sort(key=compare_head_distance)
closest_head_distance = (state.walls[0].vertices[0] - node_position).magnitude
if closest_head_distance > PROXIMITY_SCAN_BIND_THRESHOLD_MM:
closest_head_distance = inf
# Create a new wall if a loose contact node is in bind range and
# is closer than any existing wall. The contact node will be removed.
if closest_contact_distance <= PROXIMITY_SCAN_BIND_THRESHOLD_MM and closest_contact_distance < closest_head_distance and closest_contact_distance < closest_tail_distance:
state.walls.append(Wall(WallSegment(state.contact_nodes[0], node_position)))
state.contact_nodes.pop(0)
# Extend a wall if it's head is within bind range and is closer than
# any loose contacts or wall tails.
elif closest_head_distance <= PROXIMITY_SCAN_BIND_THRESHOLD_MM and closest_head_distance < closest_contact_distance and closest_head_distance < closest_tail_distance:
state.walls[0].insert_head(node_position)
# Extend a wall if it's tail is within bind range and is closer than
# any loose contacts or wall heads.
elif closest_tail_distance <= PROXIMITY_SCAN_BIND_THRESHOLD_MM and closest_tail_distance < closest_contact_distance and closest_tail_distance < closest_head_distance:
state.walls.sort(key=compare_tail_distance)
state.walls[0].insert_tail(node_position)
# If nothing was found to bind with, store the sample as a loose contact node.
else:
state.contact_nodes.append(node_position)
#: Takes a position in 3d space and adds it to the map state as an open node
async def add_proximity_non_contact_to_state(node_position: Vector3, state: MapState):
# Check to see if the uncontacted sample is inside of any area considered already explored.
is_open_unexplored = True
for ct in state.cleared_territories:
if (node_position - ct.center).magnitude < ct.radius:
is_open_unexplored = False
# If the uncontacted sample is in unfamiliar ground, store it as an open node.
if is_open_unexplored:
state.open_nodes.append(node_position)
#: Modifies the map state with the details of a proximity reading
async def analyze_proximity_sample(reading: anki_vector.proximity.ProximitySensorData, robot: anki_vector.robot.Robot, state: MapState):
# Check if the reading meets the engine's metrics for valid, and that its within our specified distance threshold.
reading_contacted = reading.is_valid and reading.distance.distance_mm < PROXIMITY_SCAN_DISTANCE_THRESHOLD_MM
if reading_contacted:
# The distance will either be the reading data, or our threshold distance if the reading is considered uncontacted.
reading_distance = reading.distance.distance_mm if reading_contacted else PROXIMITY_SCAN_DISTANCE_THRESHOLD_MM
# Convert the distance to a 3d position in worldspace.
reading_position = Vector3(
robot.pose.position.x + cos(robot.pose_angle_rad) * reading_distance,
robot.pose.position.y + sin(robot.pose_angle_rad) * reading_distance,
robot.pose.position.z)
await add_proximity_contact_to_state(reading_position, state)
else:
# Convert the distance to a 3d position in worldspace.
safe_driving_position = Vector3(
robot.pose.position.x + cos(robot.pose_angle_rad) * PROXIMITY_SCAN_OPEN_NODE_DISTANCE_MM,
robot.pose.position.y + sin(robot.pose_angle_rad) * PROXIMITY_SCAN_OPEN_NODE_DISTANCE_MM,
robot.pose.position.z)
await add_proximity_non_contact_to_state(safe_driving_position, state)
#: repeatedly collects proximity data sample and converts them to nodes and walls for the map state
async def collect_proximity_data_loop(robot: anki_vector.robot.Robot, future: concurrent.futures.Future, state: MapState):
try:
scan_interval = 1.0 / PROXIMITY_SCAN_SAMPLE_FREQUENCY_HZ
# Runs until the collection_active flag is cleared.
# This flag is cleared external to this function.
while state.collection_active:
# Collect proximity data from the sensor.
reading = robot.proximity.last_sensor_reading
if reading is not None:
await analyze_proximity_sample(reading, robot, state)
robot.viewer_3d.user_data_queue.put(state)
await asyncio.sleep(scan_interval)
# Exceptions raised in this process are ignored, unless we set them on the future, and then run future.result() at a later time
except Exception as e: # pylint: disable=broad-except
future.set_exception(e)
finally:
future.set_result(state)
#: Updates the map state by rotating 360 degrees and collecting/applying proximity data samples.
async def scan_area(robot: anki_vector.robot.Robot, state: MapState):
collect_future = concurrent.futures.Future()
# The collect_proximity_data task relies on this external trigger to know when its finished.
state.collection_active = True
# Activate the collection task while the robot turns in place.
collect_task = robot.conn.loop.create_task(collect_proximity_data_loop(robot, collect_future, state))
# Turn around in place, then send the signal to kill the collection task.
robot.behavior.turn_in_place(angle=degrees(360.0), speed=degrees(360.0 / PROXIMITY_SCAN_TURN_DURATION_S))
state.collection_active = False
# Wait for the collection task to finish.
robot.conn.run_coroutine(collect_task)
# While the result of the task is not used, this call will propagate any exceptions that
# occured in the task, allowing for debug visibility.
collect_future.result()
#: Top level call to perform exploration and environment mapping
async def map_explorer(robot: anki_vector.robot.Robot):
# Drop the lift, so that it does not block the proximity sensor
robot.behavior.set_lift_height(0.0)
# Create the map state, and add it's rendering function to the viewer's render pipeline
state = MapState()
robot.viewer_3d.add_render_call(state.render)
# Comparison function used for sorting which open nodes are the furthest from all existing
# walls and loose contacts.
# (Using 1/r^2 to respond strongly to small numbers of close contact and weaking to many distant contacts)
def open_point_sort_func(position: Vector3):
proximity_sum = 0
for p in state.contact_nodes:
proximity_sum = proximity_sum + 1 / (p - position).magnitude_squared
for c in state.walls:
for p in c.vertices:
proximity_sum = proximity_sum + 1 / (p - position).magnitude_squared
return proximity_sum
# Loop until running out of open samples to navigate to,
# or if the process has yet to start (indicated by a lack of cleared_territories).
while (state.open_nodes and ACTIVELY_EXPLORE_SPACE) or not state.cleared_territories:
if robot.pose:
# Delete any open samples range of the robot.
state.open_nodes = [position for position in state.open_nodes if (position - robot.pose.position).magnitude > PROXIMITY_SCAN_DISTANCE_THRESHOLD_MM]
# Collect map data for the robot's current location.
await scan_area(robot, state)
# Add where the robot is to the map's cleared territories.
state.cleared_territories.append(ClearedTerritory(robot.pose.position, PROXIMITY_SCAN_DISTANCE_THRESHOLD_MM))
# @TODO: This whole block should ideally be replaced with the go_to_pose actions when that is ready to go.
# Alternatively, the turn&drive commands can be modified to respond to collisions by cancelling. After
# either change, ACTIVELY_EXPLORE_SPACE should be defaulted True
if ACTIVELY_EXPLORE_SPACE and state.open_nodes:
# Sort the open nodes and find our next navigation point.
state.open_nodes.sort(key=open_point_sort_func)
nav_point = state.open_nodes[0]
# Calculate the distance and direction of this next navigation point.
nav_point_delta = Vector3(
nav_point.x - robot.pose.position.x,
nav_point.y - robot.pose.position.y,
0)
nav_distance = nav_point_delta.magnitude
nav_direction = nav_point_delta.normalized
# Convert the nav_direction into a turn angle relative to the robot's current facing.
robot_forward = Vector3(*robot.pose.rotation.to_matrix().forward_xyz).normalized
turn_angle = acos(nav_direction.dot(robot_forward))
if nav_direction.cross(robot_forward).z > 0:
turn_angle = -turn_angle
# Turn toward the nav point, and drive to it.
robot.behavior.turn_in_place(angle=radians(turn_angle), speed=degrees(EXPLORE_TURN_SPEED_DPS))
robot.behavior.drive_straight(distance=distance_mm(nav_distance), speed=speed_mmps(EXPLORE_DRIVE_SPEED_MMPS))
if PROXIMITY_EXPLORATION_SHUTDOWN_DELAY_S == 0.0:
while True:
await asyncio.sleep(1.0)
else:
print('finished exploring - waiting an additional {0} seconds, then shutting down'.format(PROXIMITY_EXPLORATION_SHUTDOWN_DELAY_S))
await asyncio.sleep(PROXIMITY_EXPLORATION_SHUTDOWN_DELAY_S)
if __name__ == '__main__':
# Connect to the robot
args = parse_command_args()
with anki_vector.Robot(args.serial,
enable_camera_feed=True,
show_viewer=True,
enable_nav_map_feed=False,
show_3d_viewer=True) as robotInstance:
robotInstance.behavior.drive_off_charger()
loop = asyncio.get_event_loop()
loop.run_until_complete(map_explorer(robotInstance))
2.4 remote control
"""Control Vector using a webpage on your computer.
This example lets you control Vector by Remote Control, using a webpage served by Flask.
"""
import io
import json
import sys
import time
from lib import flask_helpers
import anki_vector
from anki_vector import util
try:
from flask import Flask, request
except ImportError:
sys.exit("Cannot import from flask: Do `pip3 install --user flask` to install")
try:
from PIL import Image
except ImportError:
sys.exit("Cannot import from PIL: Do `pip3 install --user Pillow` to install")
def create_default_image(image_width, image_height, do_gradient=False):
"""Create a place-holder PIL image to use until we have a live feed from Vector"""
image_bytes = bytearray([0x70, 0x70, 0x70]) * image_width * image_height
if do_gradient:
i = 0
for y in range(image_height):
for x in range(image_width):
image_bytes[i] = int(255.0 * (x / image_width)) # R
image_bytes[i + 1] = int(255.0 * (y / image_height)) # G
image_bytes[i + 2] = 0 # B
i += 3
image = Image.frombytes('RGB', (image_width, image_height), bytes(image_bytes))
return image
flask_app = Flask(__name__)
_default_camera_image = create_default_image(320, 240)
_is_mouse_look_enabled_by_default = False
def remap_to_range(x, x_min, x_max, out_min, out_max):
"""convert x (in x_min..x_max range) to out_min..out_max range"""
if x < x_min:
return out_min
if x > x_max:
return out_max
ratio = (x - x_min) / (x_max - x_min)
return out_min + ratio * (out_max - out_min)
class RemoteControlVector:
def __init__(self, robot):
self.vector = robot
self.drive_forwards = 0
self.drive_back = 0
self.turn_left = 0
self.turn_right = 0
self.lift_up = 0
self.lift_down = 0
self.head_up = 0
self.head_down = 0
self.go_fast = 0
self.go_slow = 0
self.is_mouse_look_enabled = _is_mouse_look_enabled_by_default
self.mouse_dir = 0
all_anim_names = self.vector.anim.anim_list
all_anim_names.sort()
self.anim_names = []
# Hide a few specific test animations that don't behave well
bad_anim_names = [
"ANIMATION_TEST",
"soundTestAnim"]
for anim_name in all_anim_names:
if anim_name not in bad_anim_names:
self.anim_names.append(anim_name)
default_anims_for_keys = ["anim_turn_left_01", # 0
"anim_blackjack_victorwin_01", # 1
"anim_pounce_success_02", # 2
"anim_feedback_shutup_01", # 3
"anim_knowledgegraph_success_01", # 4
"anim_wakeword_groggyeyes_listenloop_01", # 5
"anim_fistbump_success_01", # 6
"anim_reacttoface_unidentified_01", # 7
"anim_rtpickup_loop_10", # 8
"anim_volume_stage_05"] # 9
self.anim_index_for_key = [0] * 10
kI = 0
for default_key in default_anims_for_keys:
try:
anim_idx = self.anim_names.index(default_key)
except ValueError:
print("Error: default_anim %s is not in the list of animations" % default_key)
anim_idx = kI
self.anim_index_for_key[kI] = anim_idx
kI += 1
self.action_queue = []
self.text_to_say = "Hi I'm Vector"
def set_anim(self, key_index, anim_index):
self.anim_index_for_key[key_index] = anim_index
def handle_mouse(self, mouse_x, mouse_y):
"""Called whenever mouse moves
mouse_x, mouse_y are in in 0..1 range (0,0 = top left, 1,1 = bottom right of window)
"""
if self.is_mouse_look_enabled:
mouse_sensitivity = 1.5 # higher = more twitchy
self.mouse_dir = remap_to_range(mouse_x, 0.0, 1.0, -mouse_sensitivity, mouse_sensitivity)
self.update_mouse_driving()
desired_head_angle = remap_to_range(mouse_y, 0.0, 1.0, 45, -25)
head_angle_delta = desired_head_angle - util.radians(self.vector.head_angle_rad).degrees
head_vel = head_angle_delta * 0.03
self.vector.motors.set_head_motor(head_vel)
def set_mouse_look_enabled(self, is_mouse_look_enabled):
was_mouse_look_enabled = self.is_mouse_look_enabled
self.is_mouse_look_enabled = is_mouse_look_enabled
if not is_mouse_look_enabled:
# cancel any current mouse-look turning
self.mouse_dir = 0
if was_mouse_look_enabled:
self.update_mouse_driving()
self.update_head()
def update_drive_state(self, key_code, is_key_down, speed_changed):
"""Update state of driving intent from keyboard, and if anything changed then call update_driving"""
update_driving = True
if key_code == ord('W'):
self.drive_forwards = is_key_down
elif key_code == ord('S'):
self.drive_back = is_key_down
elif key_code == ord('A'):
self.turn_left = is_key_down
elif key_code == ord('D'):
self.turn_right = is_key_down
else:
if not speed_changed:
update_driving = False
return update_driving
def update_lift_state(self, key_code, is_key_down, speed_changed):
"""Update state of lift move intent from keyboard, and if anything changed then call update_lift"""
update_lift = True
if key_code == ord('R'):
self.lift_up = is_key_down
elif key_code == ord('F'):
self.lift_down = is_key_down
else:
if not speed_changed:
update_lift = False
return update_lift
def update_head_state(self, key_code, is_key_down, speed_changed):
"""Update state of head move intent from keyboard, and if anything changed then call update_head"""
update_head = True
if key_code == ord('T'):
self.head_up = is_key_down
elif key_code == ord('G'):
self.head_down = is_key_down
else:
if not speed_changed:
update_head = False
return update_head
def handle_key(self, key_code, is_shift_down, is_alt_down, is_key_down):
"""Called on any key press or release
Holding a key down may result in repeated handle_key calls with is_key_down==True
"""
# Update desired speed / fidelity of actions based on shift/alt being held
was_go_fast = self.go_fast
was_go_slow = self.go_slow
self.go_fast = is_shift_down
self.go_slow = is_alt_down
speed_changed = (was_go_fast != self.go_fast) or (was_go_slow != self.go_slow)
update_driving = self.update_drive_state(key_code, is_key_down, speed_changed)
update_lift = self.update_lift_state(key_code, is_key_down, speed_changed)
update_head = self.update_head_state(key_code, is_key_down, speed_changed)
# Update driving, head and lift as appropriate
if update_driving:
self.update_mouse_driving()
if update_head:
self.update_head()
if update_lift:
self.update_lift()
# Handle any keys being released (e.g. the end of a key-click)
if not is_key_down:
if ord('9') >= key_code >= ord('0'):
anim_name = self.key_code_to_anim_name(key_code)
self.queue_action((self.vector.anim.play_animation, anim_name))
elif key_code == ord(' '):
self.queue_action((self.vector.say_text, self.text_to_say))
def key_code_to_anim_name(self, key_code):
key_num = key_code - ord('0')
anim_num = self.anim_index_for_key[key_num]
anim_name = self.anim_names[anim_num]
return anim_name
def func_to_name(self, func):
if func == self.vector.say_text:
return "say_text"
if func == self.vector.anim.play_animation:
return "play_anim"
return "UNKNOWN"
def action_to_text(self, action):
func, args = action
return self.func_to_name(func) + "( " + str(args) + " )"
def action_queue_to_text(self, action_queue):
out_text = ""
i = 0
for action in action_queue:
out_text += "[" + str(i) + "] " + self.action_to_text(action)
i += 1
return out_text
def queue_action(self, new_action):
if len(self.action_queue) > 10:
self.action_queue.pop(0)
self.action_queue.append(new_action)
def update(self):
"""Try and execute the next queued action"""
if self.action_queue:
queued_action, action_args = self.action_queue[0]
if queued_action(action_args):
self.action_queue.pop(0)
def pick_speed(self, fast_speed, mid_speed, slow_speed):
if self.go_fast:
if not self.go_slow:
return fast_speed
elif self.go_slow:
return slow_speed
return mid_speed
def update_lift(self):
lift_speed = self.pick_speed(8, 4, 2)
lift_vel = (self.lift_up - self.lift_down) * lift_speed
self.vector.motors.set_lift_motor(lift_vel)
def update_head(self):
if not self.is_mouse_look_enabled:
head_speed = self.pick_speed(2, 1, 0.5)
head_vel = (self.head_up - self.head_down) * head_speed
self.vector.motors.set_head_motor(head_vel)
def update_mouse_driving(self):
drive_dir = (self.drive_forwards - self.drive_back)
turn_dir = (self.turn_right - self.turn_left) + self.mouse_dir
if drive_dir < 0:
# It feels more natural to turn the opposite way when reversing
turn_dir = -turn_dir
forward_speed = self.pick_speed(150, 75, 50)
turn_speed = self.pick_speed(100, 50, 30)
l_wheel_speed = (drive_dir * forward_speed) + (turn_speed * turn_dir)
r_wheel_speed = (drive_dir * forward_speed) - (turn_speed * turn_dir)
self.vector.motors.set_wheel_motors(l_wheel_speed, r_wheel_speed, l_wheel_speed * 4, r_wheel_speed * 4)
def get_anim_sel_drop_down(selectorIndex):
html_text = """<select onchange="handleDropDownSelect(this)" name="animSelector""" + str(selectorIndex) + """">"""
i = 0
for anim_name in flask_app.remote_control_vector.anim_names:
is_selected_item = (i == flask_app.remote_control_vector.anim_index_for_key[selectorIndex])
selected_text = ''' selected="selected"''' if is_selected_item else ""
html_text += """<option value=""" + str(i) + selected_text + """>""" + anim_name + """</option>"""
i += 1
html_text += """</select>"""
return html_text
def get_anim_sel_drop_downs():
html_text = ""
for i in range(10):
# list keys 1..9,0 as that's the layout on the keyboard
key = i + 1 if (i < 9) else 0
html_text += str(key) + """: """ + get_anim_sel_drop_down(key) + """<br>"""
return html_text
def to_js_bool_string(bool_value):
return "true" if bool_value else "false"
@flask_app.route("/")
def handle_index_page():
return """
<html>
<head>
<title>remote_control_vector.py display</title>
</head>
<body>
<h1>Remote Control Vector</h1>
<table>
<tr>
<td valign = top>
<div id="vectorImageMicrosoftWarning" style="display: none;color: #ff9900; text-align: center;">Video feed performance is better in Chrome or Firefox due to mjpeg limitations in this browser</div>
<img src="vectorImage" id="vectorImageId" width=640 height=480>
<div id="DebugInfoId"></div>
</td>
<td width=30></td>
<td valign=top>
<h2>Controls:</h2>
<h3>Driving:</h3>
<b>W A S D</b> : Drive Forwards / Left / Back / Right<br><br>
<b>Q</b> : Toggle Mouse Look: <button id="mouseLookId" onClick=onMouseLookButtonClicked(this) style="font-size: 14px">Default</button><br>
<b>Mouse</b> : Move in browser window to aim<br>
(steer and head angle)<br>
(similar to an FPS game)<br>
<h3>Head:</h3>
<b>T</b> : Move Head Up<br>
<b>G</b> : Move Head Down<br>
<h3>Lift:</h3>
<b>R</b> : Move Lift Up<br>
<b>F</b>: Move Lift Down<br>
<h3>General:</h3>
<b>Shift</b> : Hold to Move Faster (Driving, Head and Lift)<br>
<b>Alt</b> : Hold to Move Slower (Driving, Head and Lift)<br>
<b>P</b> : Toggle Free Play mode: <button id="freeplayId" onClick=onFreeplayButtonClicked(this) style="font-size: 14px">Default</button><br>
<h3>Play Animations</h3>
<b>0 .. 9</b> : Play Animation mapped to that key<br>
<h3>Talk</h3>
<b>Space</b> : Say <input type="text" name="sayText" id="sayTextId" value="""" + flask_app.remote_control_vector.text_to_say + """" onchange=handleTextInput(this)>
</td>
<td width=30></td>
<td valign=top>
<h2>Animation key mappings:</h2>
""" + get_anim_sel_drop_downs() + """<br>
</td>
</tr>
</table>
<script type="text/javascript">
var gLastClientX = -1
var gLastClientY = -1
var gIsMouseLookEnabled = """ + to_js_bool_string(_is_mouse_look_enabled_by_default) + """
var gIsFreeplayEnabled = false
var gUserAgent = window.navigator.userAgent;
var gIsMicrosoftBrowser = gUserAgent.indexOf('MSIE ') > 0 || gUserAgent.indexOf('Trident/') > 0 || gUserAgent.indexOf('Edge/') > 0;
var gSkipFrame = false;
if (gIsMicrosoftBrowser) {
document.getElementById("vectorImageMicrosoftWarning").style.display = "block";
}
function postHttpRequest(url, dataSet)
{
var xhr = new XMLHttpRequest();
xhr.open("POST", url, true);
xhr.send( JSON.stringify( dataSet ) );
}
function updateVector()
{
console.log("Updating log")
if (gIsMicrosoftBrowser && !gSkipFrame) {
// IE doesn't support MJPEG, so we need to ping the server for more images.
// Though, if this happens too frequently, the controls will be unresponsive.
gSkipFrame = true;
document.getElementById("vectorImageId").src="vectorImage?" + (new Date()).getTime();
} else if (gSkipFrame) {
gSkipFrame = false;
}
var xhr = new XMLHttpRequest();
xhr.onreadystatechange = function() {
if (xhr.readyState == XMLHttpRequest.DONE) {
document.getElementById("DebugInfoId").innerHTML = xhr.responseText
}
}
xhr.open("POST", "updateVector", true);
xhr.send( null );
}
setInterval(updateVector , 60);
function updateButtonEnabledText(button, isEnabled)
{
button.firstChild.data = isEnabled ? "Enabled" : "Disabled";
}
function onMouseLookButtonClicked(button)
{
gIsMouseLookEnabled = !gIsMouseLookEnabled;
updateButtonEnabledText(button, gIsMouseLookEnabled);
isMouseLookEnabled = gIsMouseLookEnabled
postHttpRequest("setMouseLookEnabled", {isMouseLookEnabled})
}
function onFreeplayButtonClicked(button)
{
gIsFreeplayEnabled = !gIsFreeplayEnabled;
updateButtonEnabledText(button, gIsFreeplayEnabled);
isFreeplayEnabled = gIsFreeplayEnabled
postHttpRequest("setFreeplayEnabled", {isFreeplayEnabled})
}
updateButtonEnabledText(document.getElementById("mouseLookId"), gIsMouseLookEnabled);
updateButtonEnabledText(document.getElementById("freeplayId"), gIsFreeplayEnabled);
function handleDropDownSelect(selectObject)
{
selectedIndex = selectObject.selectedIndex
itemName = selectObject.name
postHttpRequest("dropDownSelect", {selectedIndex, itemName});
}
function handleKeyActivity (e, actionType)
{
var keyCode = (e.keyCode ? e.keyCode : e.which);
var hasShift = (e.shiftKey ? 1 : 0)
var hasCtrl = (e.ctrlKey ? 1 : 0)
var hasAlt = (e.altKey ? 1 : 0)
if (actionType=="keyup")
{
if (keyCode == 80) // 'P'
{
// Simulate a click of the freeplay button
onFreeplayButtonClicked(document.getElementById("freeplayId"))
}
else if (keyCode == 81) // 'Q'
{
// Simulate a click of the mouse look button
onMouseLookButtonClicked(document.getElementById("mouseLookId"))
}
}
postHttpRequest(actionType, {keyCode, hasShift, hasCtrl, hasAlt})
}
function handleMouseActivity (e, actionType)
{
var clientX = e.clientX / document.body.clientWidth // 0..1 (left..right)
var clientY = e.clientY / document.body.clientHeight // 0..1 (top..bottom)
var isButtonDown = e.which && (e.which != 0) ? 1 : 0
var deltaX = (gLastClientX >= 0) ? (clientX - gLastClientX) : 0.0
var deltaY = (gLastClientY >= 0) ? (clientY - gLastClientY) : 0.0
gLastClientX = clientX
gLastClientY = clientY
postHttpRequest(actionType, {clientX, clientY, isButtonDown, deltaX, deltaY})
}
function handleTextInput(textField)
{
textEntered = textField.value
postHttpRequest("sayText", {textEntered})
}
document.addEventListener("keydown", function(e) { handleKeyActivity(e, "keydown") } );
document.addEventListener("keyup", function(e) { handleKeyActivity(e, "keyup") } );
document.addEventListener("mousemove", function(e) { handleMouseActivity(e, "mousemove") } );
function stopEventPropagation(event)
{
if (event.stopPropagation)
{
event.stopPropagation();
}
else
{
event.cancelBubble = true
}
}
document.getElementById("sayTextId").addEventListener("keydown", function(event) {
stopEventPropagation(event);
} );
document.getElementById("sayTextId").addEventListener("keyup", function(event) {
stopEventPropagation(event);
} );
</script>
</body>
</html>
"""
def get_annotated_image():
# TODO: Update to use annotated image (add annotate module)
image = flask_app.remote_control_vector.vector.camera.latest_image
if image is None:
return _default_camera_image
return image
def streaming_video():
"""Video streaming generator function"""
while True:
if flask_app.remote_control_vector:
image = get_annotated_image()
img_io = io.BytesIO()
image.save(img_io, 'PNG')
img_io.seek(0)
yield (b'--frame\r\n'
b'Content-Type: image/png\r\n\r\n' + img_io.getvalue() + b'\r\n')
else:
time.sleep(.1)
def serve_single_image():
if flask_app.remote_control_vector:
image = get_annotated_image()
if image:
return flask_helpers.serve_pil_image(image)
return flask_helpers.serve_pil_image(_default_camera_image)
def is_microsoft_browser(req):
agent = req.user_agent.string
return 'Edge/' in agent or 'MSIE ' in agent or 'Trident/' in agent
@flask_app.route("/vectorImage")
def handle_vectorImage():
if is_microsoft_browser(request):
return serve_single_image()
return flask_helpers.stream_video(streaming_video)
def handle_key_event(key_request, is_key_down):
message = json.loads(key_request.data.decode("utf-8"))
if flask_app.remote_control_vector:
flask_app.remote_control_vector.handle_key(key_code=(message['keyCode']), is_shift_down=message['hasShift'],
is_alt_down=message['hasAlt'], is_key_down=is_key_down)
return ""
@flask_app.route('/mousemove', methods=['POST'])
def handle_mousemove():
"""Called from Javascript whenever mouse moves"""
message = json.loads(request.data.decode("utf-8"))
if flask_app.remote_control_vector:
flask_app.remote_control_vector.handle_mouse(mouse_x=(message['clientX']), mouse_y=message['clientY'])
return ""
@flask_app.route('/setMouseLookEnabled', methods=['POST'])
def handle_setMouseLookEnabled():
"""Called from Javascript whenever mouse-look mode is toggled"""
message = json.loads(request.data.decode("utf-8"))
if flask_app.remote_control_vector:
flask_app.remote_control_vector.set_mouse_look_enabled(is_mouse_look_enabled=message['isMouseLookEnabled'])
return ""
@flask_app.route('/setFreeplayEnabled', methods=['POST'])
def handle_setFreeplayEnabled():
"""Called from Javascript whenever freeplay mode is toggled on/off"""
message = json.loads(request.data.decode("utf-8"))
if flask_app.remote_control_vector:
isFreeplayEnabled = message['isFreeplayEnabled']
connection = flask_app.remote_control_vector.vector.conn
connection.request_control(enable=(not isFreeplayEnabled))
return ""
@flask_app.route('/keydown', methods=['POST'])
def handle_keydown():
"""Called from Javascript whenever a key is down (note: can generate repeat calls if held down)"""
return handle_key_event(request, is_key_down=True)
@flask_app.route('/keyup', methods=['POST'])
def handle_keyup():
"""Called from Javascript whenever a key is released"""
return handle_key_event(request, is_key_down=False)
@flask_app.route('/dropDownSelect', methods=['POST'])
def handle_dropDownSelect():
"""Called from Javascript whenever an animSelector dropdown menu is selected (i.e. modified)"""
message = json.loads(request.data.decode("utf-8"))
item_name_prefix = "animSelector"
item_name = message['itemName']
if flask_app.remote_control_vector and item_name.startswith(item_name_prefix):
item_name_index = int(item_name[len(item_name_prefix):])
flask_app.remote_control_vector.set_anim(item_name_index, message['selectedIndex'])
return ""
@flask_app.route('/sayText', methods=['POST'])
def handle_sayText():
"""Called from Javascript whenever the saytext text field is modified"""
message = json.loads(request.data.decode("utf-8"))
if flask_app.remote_control_vector:
flask_app.remote_control_vector.text_to_say = message['textEntered']
return ""
@flask_app.route('/updateVector', methods=['POST'])
def handle_updateVector():
if flask_app.remote_control_vector:
flask_app.remote_control_vector.update()
action_queue_text = ""
i = 1
for action in flask_app.remote_control_vector.action_queue:
action_queue_text += str(i) + ": " + flask_app.remote_control_vector.action_to_text(action) + "<br>"
i += 1
return "Action Queue:<br>" + action_queue_text + "\n"
return ""
def run():
args = util.parse_command_args()
with anki_vector.AsyncRobot(args.serial, enable_camera_feed=True) as robot:
flask_app.remote_control_vector = RemoteControlVector(robot)
robot.behavior.drive_off_charger()
flask_helpers.run_flask(flask_app)
if __name__ == '__main__':
try:
run()
except KeyboardInterrupt as e:
pass
except anki_vector.exceptions.VectorConnectionException as e:
sys.exit("A connection error occurred: %s" % e)
相關文章
- Cozmo人工智慧機器人SDK使用筆記（1）-基礎部分basics人工智慧機器人筆記
- Cozmo人工智慧機器人SDK使用筆記（3）-視覺部分vision人工智慧機器人筆記視覺
- Cozmo人工智慧機器人SDK使用筆記（2）-顯示部分face人工智慧機器人筆記
- Cozmo人工智慧機器人SDK使用筆記（9）-判斷部分if_this_then_that人工智慧機器人筆記
- Cozmo人工智慧機器人SDK使用筆記（8）-應用部分apps人工智慧機器人筆記APP
- Cozmo人工智慧機器人SDK使用筆記（7）-補充說明人工智慧機器人筆記
- Cozmo人工智慧機器人SDK使用筆記（4）-任務部分cubes_and_objects人工智慧機器人筆記Object
- Cozmo人工智慧機器人SDK使用筆記（6）-並行部分Parallel_Action人工智慧機器人筆記並行Parallel
- Cozmo人工智慧機器人SDK使用筆記（5）-時序部分async_sync人工智慧機器人筆記
- Vector人工智慧情感機器人SDK釋出和說明（ROS2GO 2.0 + Vector 2.0）人工智慧機器人ROSGo
- Cozmo人工智慧機器人SDK使用筆記（X）-總結- |人工智慧基礎（中小學版）實踐平臺|人工智慧機器人筆記
- Anki Cozmo（Vector）人工智慧機器人玩具部分文件人工智慧機器人
- Tello無人機的使用筆記之dji-sdk/Tello-Python無人機筆記Python
- 【機器學習】支援向量機（個人筆記）機器學習筆記
- C++ ——vector陣列筆記C++陣列筆記
- 軒田機器學習技法課程學習筆記1 — Linear Support Vector Machine機器學習筆記Mac
- 林軒田機器學習技法課程學習筆記6 — Support Vector Regression機器學習筆記
- 關於聊天機器人的閱讀筆記機器人筆記
- 讀人工智慧時代與人類未來筆記06_機器學習的力量人工智慧筆記機器學習
- 林軒田機器學習技法課程學習筆記3 — Kernel Support Vector Machine機器學習筆記Mac
- 林軒田機器學習技法課程學習筆記2 — Dual Support Vector Machine機器學習筆記Mac
- 《深入理解Java虛擬機器》個人筆記Java虛擬機筆記
- 林軒田機器學習技法課程學習筆記4 — Soft-Margin Support Vector Machine機器學習筆記Mac
- 機器學習筆記機器學習筆記
- Windows sdk程式設計筆記Windows程式設計筆記
- 使用人工智慧機器人提高農業效率 | 資料標註人工智慧機器人
- RK3568開發筆記（五）：在虛擬機器上使用SDK編譯製作uboot、kernel和ubuntu映象筆記虛擬機編譯bootUbuntu
- RK3568開發筆記（四）：在虛擬機器上使用SDK編譯製作uboot、kernel和buildroot映象筆記虛擬機編譯bootUI
- 讀AI新生：破解人機共存密碼筆記12人工智慧辯論AI密碼筆記人工智慧
- 《機器學習初步》筆記機器學習筆記
- 機器人搭建記錄 HoshinoBot機器人
- 讀寫給大家的AI極簡史筆記05機器人AI筆記機器人
- 機器學習個人筆記（三）之無監督學習機器學習筆記
- 人工智慧NLP專案_QA機器人（7）人工智慧機器人
- 計算機網路個人筆記計算機網路筆記
- V-Rep虛擬機器器人實驗平臺在Ubuntu18.04LTS安裝與使用筆記虛擬機Ubuntu筆記
- Webots R2019和ROS使用筆記（機器人模擬軟體與作業系統）WebROS筆記機器人作業系統
- C++ 學習筆記（1）：STL、Vector 與 SetC++筆記