#! /usr/bin/env python
# See run_fifo_standalone.py for documentation. The only difference is that
# a LIFO planner is created instead of a FIFO planner.
from comp313p_reactive_planner_controller.occupancy_grid import OccupancyGrid
from comp313p_reactive_planner_controller.lifo_planner import LIFOPlanner
occupancyGrid = OccupancyGrid(21, 21, 0.5)
for y in xrange(1, 19):
occupancyGrid.setCell(11, y, 1)
start = (3, 18)
goal = (20, 0)
planner = LIFOPlanner('Depth First Search', occupancyGrid)
planner.setRunInteractively(True)
planner.setWindowHeightInPixels(400)
goalReached = planner.search(start, goal)
path = planner.extractPathToGoal()
class MapperNode(object):
def __init__(self):
# Get the ground truth map from stdr; we use this to figure out the dimensions of the map
rospy.init_node('mapper_node', anonymous=True)
self.mapServer = rospy.ServiceProxy('static_map', GetMap)
resp = self.mapServer()
# Drawing options
self.showOccupancyGrid = rospy.get_param('show_mapper_occupancy_grid',
True)
self.showDeltaOccupancyGrid = rospy.get_param(
'show_mapper_delta_occupancy_grid', True)
# Create the publisher for the regular map messages
self.mapUpdatePublisher = rospy.Publisher('updated_map',
MapUpdate,
queue_size=1)
# Get the map scale
mapScale = rospy.get_param('plan_scale', 5)
# Create the occupancy grid map. This is the "raw" one from the sensor.
self.occupancyGrid = OccupancyGrid(resp.map.info.width,
resp.map.info.height,
resp.map.info.resolution, 0.5)
self.occupancyGrid.setScale(mapScale)
self.occupancyGrid.scaleEmptyMap()
# Create the delta occupancy grid map. This stores the difference since the last time the map was sent out
self.deltaOccupancyGrid = OccupancyGrid(resp.map.info.width,
resp.map.info.height,
resp.map.info.resolution, 0)
self.deltaOccupancyGrid.setScale(mapScale)
self.deltaOccupancyGrid.scaleEmptyMap()
windowHeight = rospy.get_param('maximum_window_height_in_pixels', 600)
if self.showOccupancyGrid is True:
self.occupancyGridDrawer = OccupancyGridDrawer('Mapper Node Occupancy Grid',\
self.occupancyGrid, windowHeight)
self.occupancyGridDrawer.open()
if self.showDeltaOccupancyGrid is True:
self.deltaOccupancyGridForShow = OccupancyGrid(
resp.map.info.width, resp.map.info.height,
resp.map.info.resolution, 0)
self.deltaOccupancyGridForShow.setScale(mapScale)
self.deltaOccupancyGridForShow.scaleEmptyMap()
self.deltaOccupancyGridDrawer = OccupancyGridDrawer('Mapper Node Delta Occupancy Grid',\
self.deltaOccupancyGridForShow, windowHeight)
self.deltaOccupancyGridDrawer.open()
# Flag set to true if graphics can be updated
self.visualisationUpdateRequired = False
# Set up the lock to ensure thread safety
self.dataCopyLock = Lock()
self.noOdometryReceived = True
self.noTwistReceived = True
self.noLaserScanReceived = True
self.enableMapping = True
# Register the supported services
self.changeMapperStateService = rospy.Service('change_mapper_state',
ChangeMapperState,
self.mappingStateService)
self.requestMapUpdateService = rospy.Service(
'request_map_update', RequestMapUpdate,
self.requestMapUpdateService)
# Set up the subscribers. These track the robot position, speed and laser scans.
self.mostRecentOdometry = Odometry()
self.odometrySubscriber = rospy.Subscriber("robot0/odom",
Odometry,
self.odometryCallback,
queue_size=1)
self.mostRecentTwist = Twist()
self.twistSubscriber = rospy.Subscriber('/robot0/cmd_vel',
Twist,
self.twistCallback,
queue_size=1)
self.laserSubscriber = rospy.Subscriber("robot0/laser_0",
LaserScan,
self.laserScanCallback,
queue_size=1)
def odometryCallback(self, msg):
self.dataCopyLock.acquire()
self.mostRecentOdometry = msg
self.noOdometryReceived = False
self.dataCopyLock.release()
def twistCallback(self, msg):
self.dataCopyLock.acquire()
self.mostRecentVelocity = msg
self.noTwistReceived = False
self.dataCopyLock.release()
def mappingStateService(self, changeMapperState):
self.enableMapping = changeMapperState.enableMapping
# rospy.loginfo('Changing the enableMapping state to %d', self.enableMapping)
return ChangeMapperStateResponse()
def requestMapUpdateService(self, request):
rospy.loginfo(
'requestMapUpdateService with deltaOccupancyGridRequired %d',
request.deltaOccupancyGridRequired)
mapUpdateMessage = self.constructMapUpdateMessage(
request.deltaOccupancyGridRequired)
return RequestMapUpdateResponse(mapUpdateMessage)
# Handle the laser scan callback. First process the scans and update the various maps
def laserScanCallback(self, msg):
# Can't process anything until stuff is enabled
if self.enableMapping is False:
return
# Can't process anything until we have the first scan
if (self.noOdometryReceived is True) or (self.noTwistReceived is True):
return
# Process the scan
gridHasChanged = self.parseScan(msg)
# If nothing has changed, return
if gridHasChanged is False:
return
# Mark that there is a pending update to the visualisation
self.visualisationUpdateRequired = True
# Mark that a laser scan has been received
self.noLaserScanReceived = False
# Construct the map update message and send it out
mapUpdateMessage = self.constructMapUpdateMessage(True)
self.mapUpdatePublisher.publish(mapUpdateMessage)
# Predict the pose of the robot to the current time. This is to
# hopefully make the pose of the robot a bit more accurate. The
# equation is: currentPose = lastPose + dT * lastTwist. Note this
# isn't quite right. e.g. a more proper implementation would store
# a history of velocities and interpolate over them.
def predictPose(self, predictTime):
# Copy the last odometry and velocity
self.dataCopyLock.acquire()
currentPose = copy.deepcopy(self.mostRecentOdometry.pose.pose)
currentPoseTime = self.mostRecentOdometry.header.stamp.to_sec()
currentTwist = copy.deepcopy(self.mostRecentTwist)
self.dataCopyLock.release()
dT = 0 #predictTime - currentPoseTime
quaternion = (currentPose.orientation.x, currentPose.orientation.y,
currentPose.orientation.z, currentPose.orientation.w)
euler = tf.transformations.euler_from_quaternion(quaternion)
theta = euler[2]
# These are the "ideal motion model" prediction equations from
# stdr which attempt to accurately describe the trajectory of
# the robot if it turns as it moves. The equations are precise
# if the angular and linear velocity is constant over the
# prediction interval.
if (abs(currentTwist.angular.z) < 1e-6):
x = currentPose.position.x + dT * currentTwist.linear.x * math.cos(
theta)
y = currentPose.position.y + dT * currentTwist.linear.x * math.sin(
theta)
else:
x = currentPose.position.x - currentTwist.linear.x / currentTwist.angular.z * sin(theta) + \
currentTwist.linear.x / currentTwist.angular.z * sin(theta + dT * currentTwist.angular.z)
y = currentPose.position.y - currentTwist.linear.x / currentTwist.angular.z * cos(theta) + \
currentTwist.linear.x / currentTwist.angular.z * cos(theta + dT * currentTwist.angular.z)
theta = theta + currentTwist.angular.z * dT
return x, y, theta
def parseScan(self, msg):
# Predict the robot pose to the time the scan was taken
x, y, theta = self.predictPose(msg.header.stamp.to_sec())
# Clear the flag which shows that the map has changed
gridHasChanged = False
# Clear the delta map, to imply that no changes have been detected
self.deltaOccupancyGrid.clearMap(0)
# For each ray, check the range is good. If so, check all the
# cells along the ray and mark cells as either open or
# blocked. To get around numerical issues, we trace along each
# ray in turn and terminate when we hit the first obstacle or at the end of the ray.
for ii in range(
int(
math.floor((msg.angle_max - msg.angle_min) /
msg.angle_increment))):
# rospy.loginfo("{} {} {}".format(msg.ranges[ii],msg.angle_min,msg.angle_max))
detectedRange = msg.ranges[ii]
# If the detection is below the minimum range, assume this ray is busted and continue
if (detectedRange <= msg.range_min):
continue
rayEndsOnObject = True
# If the detection range is beyond the end of the sensor,
# this is the mark which says that nothing was detected
if detectedRange >= msg.range_max:
rayEndsOnObject = False
detectedRange = msg.range_max
# Get the angle of this ray
angle = msg.angle_min + msg.angle_increment * ii + theta
# Get the list of cells which sit on this ray. The ray is
# scaled so that the last sell is at the detected range
# from the sensor.
between = self.ray_trace(detectedRange, x, y, angle, msg)
# If between is empty, something went wrong with the ray cast, so skip
if len(between) == 0:
continue
# Traverse along the ray and set cells. We can only change
# cells from unknown (0.5) to free. If we encounter a
# blocked cell, terminate. Sometimes the ray can slightly
# extend through a blocked cell due to numerical rounding
# issues.
traversedToEnd = True
for point in between:
try:
if self.occupancyGrid.getCell(point[0], point[1]) > 0.5:
traversedToEnd = False
break
if self.occupancyGrid.getCell(point[0], point[1]) == 0.5:
self.occupancyGrid.setCell(point[0], point[1], 0)
self.deltaOccupancyGrid.setCell(
point[0], point[1], 1.0)
if self.showDeltaOccupancyGrid is True:
self.deltaOccupancyGridForShow.setCell(
point[0], point[1], 1.0)
gridHasChanged = True
except IndexError as e:
print(e)
print "between: " + str(point[0]) + ", " + str(point[1])
print "extent: " + str(self.occupancyGrid.getExtent())
# If we got to the end okay, see if we have to mark the
# state of the end cell to occupied or not. To do this, we
# Note that we can change a cell
# from unknown and free to occupied, but we cannot change
# the state from occupied back to anything else. This gets
# around the issue that there can be "blinking" between
# whether a cell is occupied or not.
if (traversedToEnd is True) & (rayEndsOnObject is True):
lastPoint = between[-1]
if self.occupancyGrid.getCell(lastPoint[0],
lastPoint[1]) < 1.0:
self.occupancyGrid.setCell(lastPoint[0], lastPoint[1], 1)
self.deltaOccupancyGrid.setCell(lastPoint[0], lastPoint[1],
1.0)
if self.showDeltaOccupancyGrid is True:
self.deltaOccupancyGridForShow.setCell(
lastPoint[0], lastPoint[1], 1.0)
gridHasChanged = True
return gridHasChanged
def ray_trace(self, dist, x, y, angle, scanmsg):
"""
Function to get a list of points between two points
:param origin: position of the origin in world coordinates
:param dist: distance to end point
:param angle: angle from robot
:param scanmsg: Laser Scan message
:return: list of points in between the origin and end point
"""
startPoint = self.occupancyGrid.getCellCoordinatesFromWorldCoordinates([math.cos(angle) * scanmsg.range_min + x, \
math.sin(angle) * scanmsg.range_min + y])
endPoint = self.occupancyGrid.getCellCoordinatesFromWorldCoordinates([math.cos(angle) * dist + x, \
math.sin(angle) * dist + y])
points = bresenham(endPoint, startPoint)
return points.path
def updateVisualisation(self):
# If anything has changed the state of the occupancy grids,
# visualisationUpdateRequired is set to true. Therefore, the
# graphics are updated. If set to false, we flush anyway to
# make sure that the windows are properly (re)drawn in VNC.
if self.visualisationUpdateRequired is True:
if self.showOccupancyGrid is True:
self.occupancyGridDrawer.update()
if self.showDeltaOccupancyGrid is True:
self.deltaOccupancyGridDrawer.update()
self.deltaOccupancyGridForShow.clearMap(0)
self.visualisationUpdateRequired = False
else:
if self.showOccupancyGrid is True:
self.occupancyGridDrawer.flushAndUpdateWindow()
if self.showDeltaOccupancyGrid is True:
self.deltaOccupancyGridDrawer.flushAndUpdateWindow()
def constructMapUpdateMessage(self, deltaMapRequired):
# Construct the map update message
mapUpdateMessage = MapUpdate()
mapUpdateMessage.header.stamp = rospy.Time().now()
mapUpdateMessage.isPriorMap = self.noLaserScanReceived
mapUpdateMessage.scale = self.occupancyGrid.getScale()
mapUpdateMessage.resolution = self.occupancyGrid.getResolution()
mapUpdateMessage.extentInCells = self.occupancyGrid.getExtentInCells()
mapUpdateMessage.occupancyGrid = self.occupancyGrid.getGridAsVector()
if deltaMapRequired is True:
mapUpdateMessage.deltaOccupancyGrid = self.deltaOccupancyGrid.getGridAsVector(
)
return mapUpdateMessage
def run(self):
while not rospy.is_shutdown():
self.updateVisualisation()
rospy.sleep(0.1)