Evaluating Ubisense Floor 4 deployment

1
Evaluating Ubisense-Floor 4 deployment

• Kavitha Muthukrishnan
• Pervasive Systems Group

2
Outline

• Presents the deployment at 4th floor
• Evaluate the results for static measurements
• How well can the tag be tracked?
• One-to-one comparison of Inertial data with
  Ubisense
• Conclusions

3
Location of the Ubisensors and the measurement
points ()
Apprx. 9m
14m
4
Static measurements

• Two compact tags at different heights were used
• Approximately 3200 measurements (roughly 3-3.5
  min) at each measurement point

5
Static measurements- Results of tag height 0.75m
6
Static measurements- Results of tag height 1.51 m
7
Mean Error
8
Max error
9
CDF for one of the worst measurement point, for
two different tag heights
60 within 90 cm
10
CDF for one of the best measurement point, for
two different tag heights
80 within 7 cm
11
Tracking tag within the cell with different
update rates
12
Tracking tag within the cell, with different
update rates
13
Dead reckoning with Inertial Sensors

• Inertial navigation self-contained navigation
  technique
• Accelerometers and gyroscopes are used to track
  the position and orientation of an object
  relative to a known starting point, orientation.

XSens MTx sensor
It provides drift-free 3D orientation as well as
kinematic data 3D acceleration 3D rate of turn
(rate gyro) 3D earth-magnetic field

• Integration causes error
• Gyros are major component of INS and gyro
  accuracy limits the overall
• positioning accuracy
• Practically unusable for more than few seconds

14
Pedestrian Dead Reckoning Algorithm

• Exploiting walking dynamics of the user
• Using gyroscope and acceleration data to detect
  stance and swing phase

Stance and Swing phase detection
15
PDR vs Ubisense
16
PDR vs Ubisense
17
PDR vs Ubisense
18
Conclusions

• Good for static measurements
• Tracking the tag can point the person to within
  few cms within the room
• Height and the placement of the tag does matter!
• Ubisense better than PDR in places where the PDR
  heading is impacted
• Compute error for walking logs against PDR (st.
  lines)
• More measurements with lesser number of sensors,
  
• (disabling 2 sensors out of the 6)

19
(No Transcript)
20
(No Transcript)
21
Std deviation
22
PDR vs Ubisense
23
PDR vs Ubisense
24
(No Transcript)
25
CDF of one of the worst point (16) and one of the
best point (2)