def __init__(self):
		self.node_name = rospy.get_name()
		rospy.loginfo("[%s] Initializing " %(self.node_name))
		rospy.Subscriber("/pcl_points", PoseArray, self.call_back, queue_size=1, buff_size = 2**24)
		rospy.Subscriber("/move_base_simple/goal", PoseStamped, self.cb_new_goal, queue_size=1)
		self.pub_map = rospy.Publisher('/local_map', OccupancyGrid, queue_size = 1)
		self.pub_planning_map = rospy.Publisher('/planning_map', OccupancyGrid, queue_size = 1)
		self.pub_rviz = rospy.Publisher("/wp_line", Marker, queue_size = 1)
		self.pub_path = rospy.Publisher("/planning_path", Path, queue_size = 1)
		self.pub_point = rospy.Publisher('/pursue_point', PoseStamped, queue_size=10)
		self.resolution = 0.09
		self.width = 200
		self.height = 200
		self.origin = Pose()
		self.local_map = OccupancyGrid()
		self.dilating_size = 4
		self.wall_width = 3
		self.start_planning = False
		self.goal = []
		self.goal_occupancygrid = []
		self.astar = AStar()
		self.msg_count = 0
		self.border = 50
		self.frame_id = None
		self.map_frame = "X1/base_link"
		self.static_local = False
		self.transpose_matrix = None
		self.old_transform_matrix = None
		self.pre_path = Path()
		self.radius = 2
Example #2
0
 def __init__(self):
     self.node_name = rospy.get_name()
     self.tf = TransformListener()
     self.transformer = TransformerROS()
     rospy.loginfo("[%s] Initializing " % (self.node_name))
     #rospy.Subscriber("/pcl_points", PoseArray, self.call_back, queue_size=1, buff_size = 2**24)
     sub_pcl = message_filters.Subscriber("/pcl_points", PoseArray)
     sub_odom = message_filters.Subscriber("/odometry/ground_truth",
                                           Odometry)
     ats = ApproximateTimeSynchronizer((sub_pcl, sub_odom),
                                       queue_size=1,
                                       slop=0.1)
     ats.registerCallback(self.call_back)
     rospy.Subscriber("/move_base_simple/goal",
                      PoseStamped,
                      self.cb_new_goal,
                      queue_size=1)
     self.pub_map = rospy.Publisher('/local_map',
                                    OccupancyGrid,
                                    queue_size=1)
     self.pub_rviz = rospy.Publisher("/wp_line", Marker, queue_size=1)
     self.pub_poses = rospy.Publisher("/path_points",
                                      PoseArray,
                                      queue_size=1)
     self.resolution = 0.25
     self.robot = Pose()
     self.width = 200
     self.height = 200
     self.origin = Pose()
     self.local_map = OccupancyGrid()
     self.dilating_size = 6
     self.wall_width = 3
     self.start_planning = False
     self.transpose_matrix = None
     self.goal = []
     self.astar = AStar()
     self.msg_count = 0
     self.planning_range = 20
     self.frame_id = "map"
Example #3
0
class mapping():
    def __init__(self):
        self.node_name = rospy.get_name()
        rospy.loginfo("[%s] Initializing " % (self.node_name))
        rospy.Subscriber("/pcl_points",
                         PoseArray,
                         self.call_back,
                         queue_size=1,
                         buff_size=2**24)
        rospy.Subscriber("/move_base_simple/goal",
                         PoseStamped,
                         self.cb_new_goal,
                         queue_size=1)
        self.pub_map = rospy.Publisher('/local_map',
                                       OccupancyGrid,
                                       queue_size=1)
        self.pub_planning_map = rospy.Publisher('/planning_map',
                                                OccupancyGrid,
                                                queue_size=1)
        self.pub_rviz = rospy.Publisher("/wp_line", Marker, queue_size=1)
        self.pub_path = rospy.Publisher("/planning_path", Path, queue_size=1)
        self.pub_point = rospy.Publisher('/pursue_point',
                                         PoseStamped,
                                         queue_size=10)
        self.resolution = 0.25
        self.width = 200
        self.height = 200
        self.origin = Pose()
        self.local_map = OccupancyGrid()
        self.dilating_size = 2
        self.wall_width = 1
        self.start_planning = False
        self.goal = []
        self.goal_occupancygrid = []
        self.astar = AStar()
        self.msg_count = 0
        self.border = 50
        self.frame_id = None
        self.map_frame = "X1/base_link"
        self.static_local = False
        self.transpose_matrix = None
        self.old_transform_matrix = None
        self.pre_path = Path()
        self.radius = 4

        self.start_theta = 20

    def get_pursuit_goal(self):
        path = []
        start = False
        pose = PoseStamped()
        for i in range(len(self.path)):
            p = self.occupancygrid2map([self.path[i][1], self.path[i][0]])
            too_close = self.distanceBtwnPoints(p[0], p[1], 0, 0) < 1.2
            if start or not too_close:
                start = True
                path.append(p)
        pose.header.frame_id = self.map_frame
        pose.pose.position.x = path[0][0]
        pose.pose.position.y = path[0][1]
        self.pub_point.publish(pose)

    def distanceBtwnPoints(self, x1, y1, x2, y2):
        return math.sqrt((x2 - x1)**2 + (y2 - y1)**2)

    def init_param(self):
        self.occupancygrid = np.zeros((self.height, self.width))
        self.local_map = OccupancyGrid()
        self.local_map.info.resolution = self.resolution
        self.local_map.info.width = self.width
        self.local_map.info.height = self.height
        self.origin.position.x = -self.width * self.resolution / 2.
        self.origin.position.y = -self.height * self.resolution / 2.
        self.local_map.info.origin = self.origin

        self.cost_map = np.zeros((self.height, self.width))
        self.planning_map = OccupancyGrid()
        self.planning_map.info.resolution = self.resolution
        self.planning_map.info.width = self.width
        self.planning_map.info.height = self.height
        self.planning_map.info.origin = self.origin

    def cb_rviz(self, msg):
        self.click_pt = [msg.pose.position.x, msg.pose.position.y]
        self.publish_topic()

    def call_back(self, msg):

        self.map_frame = msg.header.frame_id
        self.msg_count = self.msg_count + 1
        self.init_param()
        self.frame_id = msg.header.frame_id
        self.local_map.header = msg.header
        self.planning_map.header = msg.header

        wall_dis = 1.5

        if True or self.goal != []:
            self.start_planning = True
        for i in range(len(msg.poses)):
            p = (msg.poses[i].position.x, msg.poses[i].position.y)
            x, y = self.map2occupancygrid(p)
            width_in_range = (x >= self.width - self.dilating_size
                              or x <= self.dilating_size)
            height_in_range = (y >= self.height - self.dilating_size
                               or y <= self.dilating_size)
            if width_in_range or height_in_range:
                continue  # To prevent point cloud range over occupancy grid range
            self.occupancygrid[y][x] = 100

        for i in range(90 - self.start_theta):
            theta = (90 - i) * math.pi / 180
            r = wall_dis / math.cos(i * math.pi / 180)
            p = (r * math.cos(theta), r * math.sin(theta))
            x, y = self.map2occupancygrid(p)
            self.occupancygrid[y][x] = 100
            p = (r * math.cos(-theta), r * math.sin(-theta))
            x, y = self.map2occupancygrid(p)
            self.occupancygrid[y][x] = 100
            p = (r * math.cos(theta), r * math.sin(-theta))
            x, y = self.map2occupancygrid(p)
            self.occupancygrid[y][x] = 100
            p = (r * math.cos(-theta), r * math.sin(theta))
            x, y = self.map2occupancygrid(p)
            self.occupancygrid[y][x] = 100

        # map dilating
        for i in range(self.height):
            for j in range(self.width):
                if self.occupancygrid[i][j] == 100:
                    for m in range(-self.dilating_size,
                                   self.dilating_size + 1):
                        for n in range(-self.dilating_size,
                                       self.dilating_size + 1):
                            if self.occupancygrid[i + m][j + n] != 100:
                                if m > self.wall_width or m < -self.wall_width or n > self.wall_width or n < -self.wall_width:
                                    if self.occupancygrid[i + m][j + n] != 90:
                                        self.occupancygrid[i + m][j + n] = 50
                                else:
                                    self.occupancygrid[i + m][j + n] = 90

        for i in range(self.height):
            for j in range(self.width):
                self.local_map.data.append(self.occupancygrid[i][j])
        self.local_map.header.frame_id = self.map_frame
        self.planning_map.header.frame_id = self.map_frame
        self.pub_map.publish(self.local_map)

        if self.start_planning:
            self.path_planning()

    def get_goal(self):
        sample = np.linspace(-30, 180, 30)
        goal_list = []
        for degree in sample:
            rad = np.deg2rad(degree)
            x = self.radius * np.cos(rad)
            y = self.radius * np.sin(rad)
            goal = self.map2occupancygrid([x, y])
            if goal not in goal_list:
                goal_list.append(goal)
        return goal_list

    def path_planning(self):
        if self.msg_count % 5 != 0:
            return
        self.msg_count = 0
        #self.cost_map = np.zeros((self.height, self.width))
        for i in range(self.height):
            for j in range(self.width):
                if i > int(self.border) and i < int(self.height - self.border):
                    if j > int(
                            self.border) and j < int(self.width - self.border):
                        self.cost_map[i][j] = self.occupancygrid[i][j]
        start = self.map2occupancygrid([0, 0])
        goal_list = self.get_goal()
        success = False
        for goal in goal_list:
            end = (goal[1], goal[0])
            self.astar.initial(self.cost_map, start, end)
            self.path, success = self.astar.planning()
            if success:
                break
        if success:
            self.old_transform_matrix = self.transpose_matrix
            self.pre_path = self.pub_topic(self.path)
            self.get_pursuit_goal()
        else:
            print("No")
            old_path = []
            for p in self.pre_path.poses:
                p_occupancygrid = self.map2occupancygrid(
                    [p.pose.position.x, p.pose.position.y])
                old_path.append([p_occupancygrid[1], p_occupancygrid[0]])
            _ = self.pub_topic(old_path)

        for i in range(self.height):
            for j in range(self.width):
                self.planning_map.data.append(self.cost_map[i][j])
        self.pub_planning_map.publish(self.planning_map)

    def cb_new_goal(self, p):
        self.goal = [p.pose.position.x, p.pose.position.y]
        # ========== For wall following ==========
        self.dx = self.goal[0]
        self.dy = self.goal[1]

    def occupancygrid2map(self, p):
        x = p[
            0] * self.resolution + self.origin.position.x + self.resolution / 2.
        y = p[
            1] * self.resolution + self.origin.position.y + self.resolution / 2.
        return [x, y]

    def map2occupancygrid(self, p):
        x = int((p[0] - self.origin.position.x) / self.resolution)
        y = int((p[1] - self.origin.position.y) / self.resolution)
        return [x, y]

    def pub_topic(self, path):
        path_msg = Path()
        path_msg.header.frame_id = self.map_frame
        marker = Marker()
        marker.header.frame_id = self.map_frame
        marker.type = marker.LINE_STRIP
        marker.action = marker.ADD
        marker.scale.x = 0.3
        marker.scale.y = 0.3
        marker.scale.z = 0.3
        marker.color.a = 1.0
        marker.color.r = 1.0
        marker.color.g = 0.
        marker.color.b = 0.
        marker.pose.orientation.x = 0.0
        marker.pose.orientation.y = 0.0
        marker.pose.orientation.z = 0.0
        marker.pose.orientation.w = 1.0
        marker.pose.position.x = 0.0
        marker.pose.position.y = 0.0
        marker.pose.position.z = 0.0
        marker.points = []
        for i in range(len(path)):
            p = self.occupancygrid2map([path[i][1], path[i][0]])
            pose = PoseStamped()
            pose.pose.position.x = p[0]
            pose.pose.position.y = p[1]
            path_msg.poses.append(pose)
            point = Point()
            point.x = p[0]
            point.y = p[1]
            point.z = 0
            marker.points.append(point)
        self.pub_rviz.publish(marker)
        self.pub_path.publish(path_msg)
        return path_msg
Example #4
0
class mapping():
    def __init__(self):
        self.node_name = rospy.get_name()
        rospy.loginfo("[%s] Initializing " % (self.node_name))
        rospy.Subscriber("/pcl_points",
                         PoseArray,
                         self.call_back,
                         queue_size=1,
                         buff_size=2**24)
        rospy.Subscriber("/move_base_simple/goal",
                         PoseStamped,
                         self.cb_new_goal,
                         queue_size=1)
        self.pub_map = rospy.Publisher('/local_map',
                                       OccupancyGrid,
                                       queue_size=1)
        self.pub_planning_map = rospy.Publisher('/planning_map',
                                                OccupancyGrid,
                                                queue_size=1)
        self.pub_rviz = rospy.Publisher("/wp_line", Marker, queue_size=1)
        self.pub_path = rospy.Publisher("/planning_path", Path, queue_size=1)
        self.resolution = 0.25
        self.width = 200
        self.height = 200
        self.origin = Pose()
        self.local_map = OccupancyGrid()
        self.dilating_size = 6
        self.wall_width = 3
        self.start_planning = False
        self.goal = []
        self.goal_occupancygrid = []
        self.astar = AStar()
        self.msg_count = 0
        self.border = 50
        self.frame_id = None
        self.map_frame = "/map"
        self.robot_pose = None
        self.static_local = False
        self.transpose_matrix = None
        self.old_transform_matrix = None
        self.pre_path = Path()
        self.radius = 5

    def init_param(self):
        self.occupancygrid = np.zeros((self.height, self.width))
        self.local_map = OccupancyGrid()
        self.local_map.info.resolution = self.resolution
        self.local_map.info.width = self.width
        self.local_map.info.height = self.height
        self.origin.position.x = -self.width * self.resolution / 2.
        self.origin.position.y = -self.height * self.resolution / 2.
        self.local_map.info.origin = self.origin

        self.cost_map = np.zeros((self.height, self.width))
        self.planning_map = OccupancyGrid()
        self.planning_map.info.resolution = self.resolution
        self.planning_map.info.width = self.width
        self.planning_map.info.height = self.height
        self.planning_map.info.origin = self.origin

    def cb_rviz(self, msg):
        self.click_pt = [msg.pose.position.x, msg.pose.position.y]
        self.publish_topic()

    def call_back(self, msg):
        self.msg_count = self.msg_count + 1
        self.init_param()
        self.frame_id = msg.header.frame_id
        self.local_map.header = msg.header
        self.planning_map.header = msg.header
        try:
            position, quaternion = tf_.lookupTransform(self.map_frame,
                                                       msg.header.frame_id,
                                                       rospy.Time(0))
            self.transpose_matrix = transformer.fromTranslationRotation(
                position, quaternion)
            self.robot_pose = np.dot(self.transpose_matrix, [0, 0, 0, 1])
            self.origin.position.x = -self.width * self.resolution / 2. + self.robot_pose[
                0]
            self.origin.position.y = -self.height * self.resolution / 2. + self.robot_pose[
                1]
            self.local_map.info.origin = self.origin
            if True or self.goal != []:
                #print(self.dx, self.dy)# 15, 7
                #goal_temp = np.dot(self.transpose_matrix, [self.dx, self.dy, 0, 1])
                #self.goal_occupancygrid = self.map2occupancygrid(self.goal)
                self.start_planning = True
            for i in range(len(msg.poses)):
                origin_p = np.array([
                    msg.poses[i].position.x, msg.poses[i].position.y,
                    msg.poses[i].position.z, 1
                ])
                new_p = np.dot(self.transpose_matrix, origin_p)
                p = (new_p[0], new_p[1])
                x, y = self.map2occupancygrid(p)
                width_in_range = (x >= self.width - self.dilating_size
                                  or x <= self.dilating_size)
                height_in_range = (y >= self.height - self.dilating_size
                                   or y <= self.dilating_size)
                if width_in_range or height_in_range:
                    continue  # To prevent point cloud range over occupancy grid range
                self.occupancygrid[y][x] = 100
        except (LookupException, ConnectivityException,
                ExtrapolationException):
            return

        # map dilating
        for i in range(self.height):
            for j in range(self.width):
                if self.occupancygrid[i][j] == 100:
                    for m in range(-self.dilating_size,
                                   self.dilating_size + 1):
                        for n in range(-self.dilating_size,
                                       self.dilating_size + 1):
                            if self.occupancygrid[i + m][j + n] != 100:
                                if m > self.wall_width or m < -self.wall_width or n > self.wall_width or n < -self.wall_width:
                                    if self.occupancygrid[i + m][j + n] != 90:
                                        self.occupancygrid[i + m][j + n] = 50
                                else:
                                    self.occupancygrid[i + m][j + n] = 90

        for i in range(self.height):
            for j in range(self.width):
                self.local_map.data.append(self.occupancygrid[i][j])
        self.local_map.header.frame_id = self.map_frame
        self.planning_map.header.frame_id = self.map_frame
        self.pub_map.publish(self.local_map)

        if self.start_planning:
            self.path_planning()

    def get_goal(self):
        sample = np.linspace(-90, 90, 30)
        goal_list = []
        for degree in sample:
            rad = np.deg2rad(degree)
            x = self.radius * np.cos(rad)
            y = self.radius * np.sin(rad)
            p = np.dot(self.transpose_matrix, [x, y, 0, 1])
            goal = self.map2occupancygrid([p[0], p[1]])
            if goal not in goal_list:
                goal_list.append(goal)
        return goal_list

    def path_planning(self):
        if self.msg_count % 5 != 0:
            return
        self.msg_count = 0
        #self.cost_map = np.zeros((self.height, self.width))
        for i in range(self.height):
            for j in range(self.width):
                if i > int(self.border) and i < int(self.height - self.border):
                    if j > int(
                            self.border) and j < int(self.width - self.border):
                        self.cost_map[i][j] = self.occupancygrid[i][j]
        start_point = self.map2occupancygrid(
            (self.robot_pose[0], self.robot_pose[1]))
        start = (start_point[1], start_point[0])
        goal_list = self.get_goal()
        success = False
        for goal in goal_list:
            end = (goal[1], goal[0])
            self.astar.initial(self.cost_map, start, end)
            path, success = self.astar.planning()
            if success:
                break
        if success:
            self.old_transform_matrix = self.transpose_matrix
            self.pre_path = self.pub_topic(path)
        else:
            print("No")
            old_path = []
            for p in self.pre_path.poses:
                p_occupancygrid = self.map2occupancygrid(
                    [p.pose.position.x, p.pose.position.y])
                old_path.append([p_occupancygrid[1], p_occupancygrid[0]])
            _ = self.pub_topic(old_path)

        for i in range(self.height):
            for j in range(self.width):
                self.planning_map.data.append(self.cost_map[i][j])
        self.pub_planning_map.publish(self.planning_map)

    def cb_new_goal(self, p):
        self.goal = [p.pose.position.x, p.pose.position.y]
        # ========== For wall following ==========
        self.dx = self.goal[0] - self.robot_pose[0]
        self.dy = self.goal[1] - self.robot_pose[1]

    def occupancygrid2map(self, p):
        x = p[
            0] * self.resolution + self.origin.position.x + self.resolution / 2.
        y = p[
            1] * self.resolution + self.origin.position.y + self.resolution / 2.
        return [x, y]

    def map2occupancygrid(self, p):
        x = int((p[0] - self.origin.position.x) / self.resolution)
        y = int((p[1] - self.origin.position.y) / self.resolution)
        return [x, y]

    def pub_topic(self, path):
        path_msg = Path()
        path_msg.header.frame_id = self.map_frame
        marker = Marker()
        marker.header.frame_id = self.map_frame
        marker.type = marker.LINE_STRIP
        marker.action = marker.ADD
        marker.scale.x = 0.3
        marker.scale.y = 0.3
        marker.scale.z = 0.3
        marker.color.a = 1.0
        marker.color.r = 1.0
        marker.color.g = 0.
        marker.color.b = 0.
        marker.pose.orientation.x = 0.0
        marker.pose.orientation.y = 0.0
        marker.pose.orientation.z = 0.0
        marker.pose.orientation.w = 1.0
        marker.pose.position.x = 0.0
        marker.pose.position.y = 0.0
        marker.pose.position.z = 0.0
        marker.points = []
        for i in range(len(path)):
            p = self.occupancygrid2map([path[i][1], path[i][0]])
            pose = PoseStamped()
            pose.pose.position.x = p[0]
            pose.pose.position.y = p[1]
            path_msg.poses.append(pose)
            point = Point()
            point.x = p[0]
            point.y = p[1]
            point.z = 0
            marker.points.append(point)
        self.pub_rviz.publish(marker)
        self.pub_path.publish(path_msg)
        return path_msg
Example #5
0
class mapping():
    def __init__(self):
        self.node_name = rospy.get_name()
        self.tf = TransformListener()
        self.transformer = TransformerROS()
        rospy.loginfo("[%s] Initializing " % (self.node_name))
        #rospy.Subscriber("/pcl_points", PoseArray, self.call_back, queue_size=1, buff_size = 2**24)
        sub_pcl = message_filters.Subscriber("/pcl_points", PoseArray)
        sub_odom = message_filters.Subscriber("/odometry/ground_truth",
                                              Odometry)
        ats = ApproximateTimeSynchronizer((sub_pcl, sub_odom),
                                          queue_size=1,
                                          slop=0.1)
        ats.registerCallback(self.call_back)
        rospy.Subscriber("/move_base_simple/goal",
                         PoseStamped,
                         self.cb_new_goal,
                         queue_size=1)
        self.pub_map = rospy.Publisher('/local_map',
                                       OccupancyGrid,
                                       queue_size=1)
        self.pub_rviz = rospy.Publisher("/wp_line", Marker, queue_size=1)
        self.pub_poses = rospy.Publisher("/path_points",
                                         PoseArray,
                                         queue_size=1)
        self.resolution = 0.25
        self.robot = Pose()
        self.width = 200
        self.height = 200
        self.origin = Pose()
        self.local_map = OccupancyGrid()
        self.dilating_size = 6
        self.wall_width = 3
        self.start_planning = False
        self.transpose_matrix = None
        self.goal = []
        self.astar = AStar()
        self.msg_count = 0
        self.planning_range = 20
        self.frame_id = "map"

    def init_param(self):
        self.occupancygrid = np.zeros((self.height, self.width))
        self.local_map = OccupancyGrid()
        self.local_map.info.resolution = self.resolution
        self.local_map.info.width = self.width
        self.local_map.info.height = self.height
        self.origin.position.x = -self.width * self.resolution / 2. + self.robot.position.x
        self.origin.position.y = -self.height * self.resolution / 2. + self.robot.position.y
        self.local_map.info.origin = self.origin

    def cb_rviz(self, msg):
        self.click_pt = [msg.pose.position.x, msg.pose.position.y]
        self.publish_topic()

    def call_back(self, pcl_msg, odom_msg):
        self.robot = odom_msg.pose.pose
        self.msg_count = self.msg_count + 1
        self.init_param()
        self.local_map.header = pcl_msg.header
        self.local_map.header.frame_id = self.frame_id
        self.get_tf()
        if self.transpose_matrix is None:
            return
        for i in range(len(pcl_msg.poses)):
            p = (pcl_msg.poses[i].position.x, pcl_msg.poses[i].position.y,
                 pcl_msg.poses[i].position.z, 1)
            local_p = np.array(p)
            global_p = np.dot(self.transpose_matrix, local_p)
            x, y = self.map2occupancygrid(global_p)
            width_in_range = (x >= self.width - self.dilating_size
                              or x <= self.dilating_size)
            height_in_range = (y >= self.height - self.dilating_size
                               or y <= self.dilating_size)
            if width_in_range or height_in_range:
                continue  # To prevent point cloud range over occupancy grid range
            self.occupancygrid[y][x] = 100

        # map dilating
        for i in range(self.height):
            for j in range(self.width):
                if self.occupancygrid[i][j] == 100:
                    for m in range(-self.dilating_size,
                                   self.dilating_size + 1):
                        for n in range(-self.dilating_size,
                                       self.dilating_size + 1):
                            if self.occupancygrid[i + m][j + n] != 100:
                                if m > self.wall_width or m < -self.wall_width or n > self.wall_width or n < -self.wall_width:
                                    if self.occupancygrid[i + m][j + n] != 90:
                                        self.occupancygrid[i + m][j + n] = 50
                                else:
                                    self.occupancygrid[i + m][j + n] = 90

        for i in range(self.height):
            for j in range(self.width):
                self.local_map.data.append(self.occupancygrid[i][j])
        self.pub_map.publish(self.local_map)

        if self.start_planning:
            self.path_planning()

    def get_tf(self):
        try:
            position, quaternion = self.tf.lookupTransform(
                "/map", "/X1/front_laser", rospy.Time(0))
            self.transpose_matrix = self.transformer.fromTranslationRotation(
                position, quaternion)
        except (LookupException, ConnectivityException,
                ExtrapolationException):
            print("Nothing Happen")

    def path_planning(self):
        if self.msg_count % 5 != 0:
            return
        self.msg_count = 0
        cost_map = np.zeros((self.height, self.width))
        border = self.planning_range / self.resolution
        h_min = int((self.height - border) / 2.)
        h_max = int(self.height - (self.height - border) / 2.)
        w_min = int((self.height - border) / 2.)
        w_max = int(self.width - (self.width - border) / 2.)
        for i in range(self.width):
            for j in range(self.width):
                if i > h_min and i < h_max:
                    if j > w_min and j < w_max:
                        cost_map[i][j] = self.occupancygrid[i][j]
        start_point = self.map2occupancygrid(
            (self.robot.position.x, self.robot.position.y))
        start = (start_point[1], start_point[0])
        goal = self.map2occupancygrid(self.goal)
        end = (goal[1], goal[0])
        self.astar.initial(cost_map, start, end)
        path = self.astar.planning()
        self.pub_path(path)
        self.rviz(path)

    def cb_new_goal(self, p):
        self.goal = [p.pose.position.x, p.pose.position.y]
        self.start_planning = True

    def occupancygrid2map(self, p):
        x = p[
            0] * self.resolution + self.origin.position.x + self.resolution / 2.
        y = p[
            1] * self.resolution + self.origin.position.y + self.resolution / 2.
        return [x, y]

    def map2occupancygrid(self, p):
        x = int((p[0] - self.origin.position.x) / self.resolution)
        y = int((p[1] - self.origin.position.y) / self.resolution)
        return [x, y]

    def pub_path(self, path):
        poses = PoseArray()
        for i in range(len(path)):
            p = self.occupancygrid2map([path[i][1], path[i][0]])
            pose = Pose()
            pose.position.x = p[0]
            pose.position.y = p[1]
            pose.position.z = 0
            poses.poses.append(pose)
        self.pub_poses.publish(poses)

    def rviz(self, path):
        marker = Marker()
        marker.header.frame_id = self.frame_id
        marker.type = marker.LINE_STRIP
        marker.action = marker.ADD
        marker.scale.x = 0.3
        marker.scale.y = 0.3
        marker.scale.z = 0.3
        marker.color.a = 1.0
        marker.color.r = 1.0
        marker.color.g = 0.
        marker.color.b = 0.
        marker.pose.orientation.x = 0.0
        marker.pose.orientation.y = 0.0
        marker.pose.orientation.z = 0.0
        marker.pose.orientation.w = 1.0
        marker.pose.position.x = 0.0
        marker.pose.position.y = 0.0
        marker.pose.position.z = 0.0
        marker.points = []
        for i in range(len(path)):
            p = self.occupancygrid2map([path[i][1], path[i][0]])
            point = Point()
            point.x = p[0]
            point.y = p[1]
            point.z = 0
            marker.points.append(point)
        self.pub_rviz.publish(marker)