2 # -*- coding: iso-8859-15 -*-
12 from datetime import datetime
14 from time import sleep
15 from std_msgs.msg import Float32
16 from nav_msgs.msg import Odometry
17 from wild_thumper.srv import DWM1000Center, DWM1000CenterResponse
19 import matplotlib.pyplot as plt
22 def __init__(self, x_est, P_est, Q, R):
23 self.x_est = x_est # Systemzustand
24 self.P_est = P_est # Fehlerkovarianz
25 self.Q = Q # Systemrauschen
26 self.R = R # Varianz des Messfehlers
29 # Korrektur mit der Messung
30 # (1) Berechnung der Kalman Verstärkung
31 K = self.P_est/(self.R + self.P_est)
32 # (2) Korrektur der Schätzung mit der Messung y
33 x = self.x_est + K*(y - self.x_est)
34 # (3) Korrektur der Fehlerkovarianzmatrix
38 # (1) Prädiziere den Systemzustand
40 # (2) Präzidiere die Fehlerkovarianzmatrix
41 self.P_est = P + self.Q
45 class DW1000(threading.Thread):
46 def __init__(self, name, addr, offset):
47 threading.Thread.__init__(self)
53 self.last_update = datetime.min
55 self.pub = rospy.Publisher(name, Float32, queue_size=16)
61 ret = struct.unpack("f", dev.read(4))
68 # Returns each distance only if current
69 def distance_valid(self):
70 if (datetime.now() - self.last_update).seconds < 1:
77 val = self.get_value()
78 if abs(val - last_val) > 10:
79 print "Ignoring values too far apart %s: %.2f - %.2f" % (self.name, val, last_val)
81 self.dist = val + self.offset
82 self.last_update = datetime.now()
83 self.pub.publish(self.distance())
89 # Varianz des Messfehlers
97 self.filter_x = simple_kalman(1.0, P_est_x, Q, Rx)
98 self.filter_y = simple_kalman(0.0, P_est_y, Q, Ry)
102 self.last_time = rospy.Time.now()
103 rospy.Subscriber("/odom_combined", Odometry, self.odomReceived)
105 def odomReceived(self, msg):
106 self.speed_x = msg.twist.twist.linear.x
107 self.speed_y = msg.twist.twist.linear.y
108 self.speed_z = msg.twist.twist.angular.z
112 - variance of kalman should be dependant on distance
114 def filter(self, x, y):
115 # Correct estimation with speed
116 current_time = rospy.Time.now()
117 dt = (current_time - self.last_time).to_sec()
118 # Subtract vehicle speed
119 pos = np.array([self.filter_x.x_est, self.filter_y.x_est])
121 pos -= np.array([self.speed_x*dt, self.speed_y*dt])
123 rot = np.array([[np.cos(self.speed_z*dt), -np.sin(self.speed_z*dt)],
124 [np.sin(self.speed_z*dt), np.cos(self.speed_z*dt)]])
125 pos = np.dot(pos, rot)
127 self.filter_x.x_est = pos[0]
128 self.filter_y.x_est = pos[1]
130 # run kalman if new measurements are valid
131 if x != None and y != None:
132 x = self.filter_x.run(x)
133 y = self.filter_y.run(y)
135 x = self.filter_x.x_est
136 y = self.filter_y.x_est
138 self.last_time = current_time
142 def handle_center_call(req):
143 diff = dwleft.distance_valid() - dwright.distance_valid()
144 dwleft.offset -= diff/2
145 dwright.offset += diff/2
146 print "Centering to %.2f %.2f" % (dwleft.offset, dwright.offset)
147 return DWM1000CenterResponse()
149 if __name__ == "__main__":
150 rospy.init_node('DWM1000')
151 dwleft = DW1000("uwb_dist_left", 0xc2, +0.02)
152 dwright = DW1000("uwb_dist_right", 0xc0, -0.02)
154 rate = rospy.Rate(10)
156 tf_broadcaster = tf.broadcaster.TransformBroadcaster()
157 rospy.Service('/DWM1000/center', DWM1000Center, handle_center_call)
159 while not rospy.is_shutdown() and dwleft.is_alive() and dwright.is_alive():
160 dist_left = dwleft.distance_valid()
161 dist_right = dwright.distance_valid()
162 if dist_left == None or dist_right == None:
163 print "no valid sensor update"
164 # run kalman prediction only
165 pos.filter(None, None)
167 dir = "left" if (dist_left < dist_right) else "right"
169 diff = abs(dist_left - dist_right)
171 # difference to high, correct to maximum
172 off = diff - dist_l_r + 0.01
173 if dist_left > dist_right:
179 print "%.2f %.2f %.2f %.2f %s" % (dwleft.distance(), dwright.distance(), dist_left, dist_right, dir)
181 a_r = (-dist_right**2 + dist_left**2 - dist_l_r**2) / (-2*dist_l_r)
183 t = dist_right**2 - a_r**2
190 x, y = pos.filter(x, y)
191 tf_broadcaster.sendTransform((x, y, 0.0), (0, 0, 0, 1), rospy.Time.now(), "uwb_beacon", "base_footprint")
194 circle_left = plt.Circle((-dist_l_r/2, 0), dwleft.distance, color='red', fill=False)
195 circle_right = plt.Circle((dist_l_r/2, 0), dwright.distance, color='green', fill=False)
196 plt.gca().add_patch(circle_left)
197 plt.gca().add_patch(circle_right)
202 # No current position, still need up update kalman prediction
203 pos.filter(None, None)