scripts/dwm1000.py

   1 #!/usr/bin/env python
   2 # -*- coding: iso-8859-15 -*-
   3
   4 VISULAIZE = False
   5
   6 import threading
   7 import struct
   8 import rospy
   9 import tf
  10 import numpy as np
  11 from math import *
  12 from datetime import datetime
  13 from i2c import i2c
  14 from time import sleep
  15 from std_msgs.msg import Float32
  16 from nav_msgs.msg import Odometry
  17 if VISULAIZE:
  18         import matplotlib.pyplot as plt
  19
  20 class simple_kalman:
  21         def __init__(self, x_est, P_est, Q, R):
  22                 self.x_est = x_est # Systemzustand
  23                 self.P_est = P_est # Fehlerkovarianz
  24                 self.Q = Q # Systemrauschen
  25                 self.R = R # Varianz des Messfehlers
  26
  27         def run(self, y):
  28                 # Korrektur mit der Messung
  29                 # (1) Berechnung der Kalman Verstärkung
  30                 K = self.P_est/(self.R + self.P_est)
  31                 # (2) Korrektur der Schätzung mit der Messung y
  32                 x = self.x_est + K*(y - self.x_est)
  33                 # (3) Korrektur der Fehlerkovarianzmatrix
  34                 P = (1-K)*self.P_est
  35                 #
  36                 # Prädiktion
  37                 # (1) Prädiziere den Systemzustand
  38                 self.x_est = x
  39                 # (2) Präzidiere die Fehlerkovarianzmatrix
  40                 self.P_est = P + self.Q
  41
  42                 return x
  43
  44 class DW1000(threading.Thread):
  45         def __init__(self, name, addr, offset):
  46                 threading.Thread.__init__(self)
  47                 self.setDaemon(1)
  48                 self.dist = 0
  49                 self.offset = offset
  50                 self.addr = addr
  51                 self.name = name
  52                 self.last_update = datetime.min
  53
  54                 self.pub = rospy.Publisher(name, Float32, queue_size=16)
  55
  56                 self.start()
  57
  58         def get_value(self):
  59                 dev = i2c(self.addr)
  60                 ret = struct.unpack("f", dev.read(4))
  61                 dev.close()
  62                 return ret[0]
  63
  64         def distance(self):
  65                 return self.dist
  66
  67         # Returns each distance only if current
  68         def distance_valid(self):
  69                 if (datetime.now() - self.last_update).seconds < 1:
  70                         return self.dist
  71                 return None
  72
  73         def run(self):
  74                 last_val = 10
  75                 while True:
  76                         val = self.get_value()
  77                         if abs(val - last_val)  > 10:
  78                                 print "Ignoring values too far apart %s: %.2f - %.2f" % (self.name, val, last_val)
  79                         elif not isnan(val):
  80                                 self.dist = val + self.offset
  81                                 self.last_update = datetime.now()
  82                                 self.pub.publish(self.distance())
  83                                 last_val = val
  84                         sleep(0.1)
  85
  86 class Position:
  87         def __init__(self):
  88                 # Varianz des Messfehlers
  89                 Rx = 0.2
  90                 Ry = 0.05
  91                 # Fehlerkovarianz
  92                 P_est_x = 0.02
  93                 P_est_y = 0.01
  94                 # Systemrauschen
  95                 Q = 0.002
  96                 self.filter_x = simple_kalman(1.0, P_est_x, Q, Rx)
  97                 self.filter_y = simple_kalman(0.0, P_est_y, Q, Ry)
  98                 self.speed_x = 0
  99                 self.speed_y = 0
 100                 self.speed_z = 0
 101                 self.last_time = rospy.Time.now()
 102                 rospy.Subscriber("/odom_combined", Odometry, self.odomReceived)
 103
 104         def odomReceived(self, msg):
 105                 self.speed_x = msg.twist.twist.linear.x
 106                 self.speed_y = msg.twist.twist.linear.y
 107                 self.speed_z = msg.twist.twist.angular.z
 108
 109         """
 110         TODO:
 111         - variance of kalman should be dependant on distance
 112         """
 113         def filter(self, x, y):
 114                 # Correct estimation with speed
 115                 current_time = rospy.Time.now()
 116                 dt = (current_time - self.last_time).to_sec()
 117                 # Subtract vehicle speed
 118                 pos = np.array([self.filter_x.x_est, self.filter_y.x_est])
 119                 # translation
 120                 pos -= np.array([self.speed_x*dt, self.speed_y*dt])
 121                 # rotation
 122                 rot = np.array([[np.cos(self.speed_z*dt), -np.sin(self.speed_z*dt)],
 123                                 [np.sin(self.speed_z*dt),  np.cos(self.speed_z*dt)]])
 124                 pos = np.dot(pos, rot)
 125                 # copy back
 126                 self.filter_x.x_est = pos[0]
 127                 self.filter_y.x_est = pos[1]
 128
 129                 # run kalman if new measurements are valid
 130                 if x != None and y != None:
 131                         x = self.filter_x.run(x)
 132                         y = self.filter_y.run(y)
 133                 else:
 134                         x = self.filter_x.x_est
 135                         y = self.filter_y.x_est
 136
 137                 self.last_time = current_time
 138                 return x,y
 139
 140
 141 if __name__ == "__main__":
 142         rospy.init_node('DW1000')
 143         dwleft  = DW1000("uwb_dist_left",  0xc2, +0.02)
 144         dwright = DW1000("uwb_dist_right", 0xc0, -0.02)
 145         dist_l_r = 0.285
 146         rate = rospy.Rate(10)
 147         pos = Position()
 148         tf_broadcaster = tf.broadcaster.TransformBroadcaster()
 149
 150         while not rospy.is_shutdown() and dwleft.is_alive() and dwright.is_alive():
 151                 dist_left = dwleft.distance_valid()
 152                 dist_right = dwright.distance_valid()
 153                 if dist_left == None or dist_right == None:
 154                         print "no valid sensor update"
 155                         # run kalman prediction only
 156                         pos.filter(None, None)
 157                 else:
 158                         dir = "left" if (dist_left < dist_right) else "right"
 159
 160                         diff = abs(dist_left - dist_right)
 161                         if diff >= dist_l_r:
 162                                 # difference to high, correct to maximum
 163                                 off = diff - dist_l_r + 0.01
 164                                 if dist_left > dist_right:
 165                                         dist_left -= off/2
 166                                         dist_right += off/2
 167                                 else:
 168                                         dist_left += off/2
 169                                         dist_right -= off/2
 170                         print "%.2f %.2f %.2f %.2f %s" % (dwleft.distance(), dwright.distance(), dist_left, dist_right, dir)
 171
 172                         a_r = (-dist_right**2 + dist_left**2 - dist_l_r**2) / (-2*dist_l_r)
 173                         x = dist_l_r/2 - a_r
 174                         t = dist_right**2 - a_r**2
 175                         if t >= 0:
 176                                 y = sqrt(t)
 177                                 print x,y
 178                                 # Rotate 90 deg
 179                                 x, y = (y, -x)
 180
 181                                 x, y = pos.filter(x, y)
 182                                 tf_broadcaster.sendTransform((x, y, 0.0), (0, 0, 0, 1), rospy.Time.now(), "uwb_beacon", "base_footprint")
 183
 184                                 if VISULAIZE:
 185                                         circle_left = plt.Circle((-dist_l_r/2, 0), dwleft.distance, color='red', fill=False)
 186                                         circle_right = plt.Circle((dist_l_r/2, 0), dwright.distance, color='green', fill=False)
 187                                         plt.gca().add_patch(circle_left)
 188                                         plt.gca().add_patch(circle_right)
 189                                         plt.grid(True)
 190                                         plt.axis('scaled')
 191                                         plt.show()
 192                         else:
 193                                 # No current position, still need up update kalman prediction
 194                                 pos.filter(None, None)
 195
 196                 rate.sleep()