1. Implementation and Evaluation
of Indoor Localization System
using WiFi Channel State
Information
Chang-Ning Tsai
Prof. Hsin-Mu Tsai
1
2. Outline
• Introduction
o Background & Motivation
o Challenge & Related work
• System Model
o Signal Preprocessing
o Doughnut
o System Architecture
o Trilateration
• Methodology
• Evaluation
• Conclusion & Future Work
2
4. Background & Motivation
4
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
5. Background & Motivation
• Offline Phase
• Online Phase
5
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
6. Background & Motivation
• Offline Phase
6
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
7. Background & Motivation
• Offline Phase
o Collect training data for model building
7
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
8. Background & Motivation
• Offline Phase
o Collect training data for model building
o Build a system model
• Propagation model: path loss exponent
• Fingerprinting: radio map
8
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
9. Background & Motivation
• Online Phase
9
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
10. Background & Motivation
• Online Phase
o Using the positioning model to estimate the location
10
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
11. Background & Motivation
• Online Phase
o Using positioning model to estimate location
o Positioning Model:
• Propagation model: trilateration
• Fingerprinting: compare RSS with radio map
11
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
12. Background & Motivation
12
• Introduction
o Background &
Motivation
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
13. Challenge & Related work
• Reduce the deployment cost
• Reduce the impact of
environmental change
• Provide sufficient accuracy
13
• Introduction
o Challenge & Related
work
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
14. Challenge & Related work
• Ultrasound – additional cost
• Infrared – additional cost
• Visible Light – additional cost
• G-sensor + Accelerometer - landmark
• Magnetic – high deployment cost
• Radio Frequency
14
• Introduction
o Challenge & Related
work
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
16. Signal Preprocessing
16
• Introduction
• System Model
o Signal Preprocessing
• Methodology
• Evaluation
• Conclusion & Future Work
• Measurement:
o 802.11n
• OFDM-multiple subcarriers
• MIMO-multiple antennas
• Channel State Information(CSI)
17. Signal Preprocessing
17
• Introduction
• System Model
o Signal Preprocessing
• Methodology
• Evaluation
• Conclusion & Future Work
• Other CSI based positioning system
• Our positioning system
18. Signal Preprocessing
• Remove outliers
18
• Introduction
• System Model
o Signal Preprocessing
• Methodology
• Evaluation
• Conclusion & Future Work
Why?
1. Human movement
2. Shadowing
3. Small-scale fading
23. Signal Preprocessing
23
• Introduction
• System Model
o Signal Preprocessing
• Methodology
• Evaluation
• Conclusion & Future Work
• IFFT
• Choose LOS component
24. Doughnut
24
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
• Estimate the most
probable position.
• Remove all unlikely
positions
25. Doughnut
• Propagation Model
o Whether Regression is piecewise or not
25
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
29. Doughnut
29
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
0 5 10 15
0
5
10
15
AP1
AP
2
AP3
AP4
30. Doughnut
30
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
0 5 10 15
0
5
10
15
AP1
AP
2
AP3
AP4
31. Doughnut
31
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
0 5 10 15
0
5
10
15
AP1
AP
2
AP3
AP4
32. Doughnut
32
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
0 5 10 15
0
5
10
15
AP1
AP
2
AP3
AP4
33. Doughnut
33
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
0 5 10 15
0
5
10
15
AP1
AP
2
AP3
AP4
34. Doughnut
34
• Introduction
• System Model
o Doughnut
• Methodology
• Evaluation
• Conclusion & Future Work
0 5 10 15
0
5
10
15
AP1
AP
2
AP3
AP4
Estimated location
Ground truth
35. System Architecture
• Offline phase
35
• Introduction
• System Model
o System Architecture
• Methodology
• Evaluation
• Conclusion & Future Work
Outlier
Removal
Multipath
mitigation
Build RSS and
distance table
Online Phase
36. System Architecture
• Online phase
36
• Introduction
• System Model
o System Architecture
• Methodology
• Evaluation
• Conclusion & Future Work
Online Phase
Determine possible
distance range for a
particular AP
37. 37
Find intersection of
possible locations of all
APs
Has
intersection ?
Remove
one of the
APs
Find intersection centroid Estimated location
Yes
No
38. Trilateration
• Trilateration
o Propagation model based
o Calculate distance to a particular AP
o Estimate the user’s location with the coordinates
of and the distances to the APs
38
• Introduction
• System Model
o Trilateration
• Methodology
• Evaluation
• Conclusion & Future Work
40. Methodology
• CSI tool
o 30 subcarriers
o CSI is measured from a received packet under the
following conditions:
• The packet is received correctly
• The packet is sent to a hardcoded, fixed mac
address
40
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
Intel-5300 NIC
49. Evaluation
• Multi-path Error Analysis
49
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
Average Error without mitigation: 1.93 m
Average Error with mitigation:1.76 m
Average Error withuot mitigation: 2.14 m
Average Error with mitigation:1.56 m
Trilateration Doughnut
0 2 4 6 8
0
0.2
0.4
0.6
0.8
1
Distance error (m)
Probability
Without multipath mitigation
With multipath mitigation
0 2 4 6 8
0
0.2
0.4
0.6
0.8
1
Distance error (m)
Probability
With multipath mitigation
Without multipath mitigation
50. Evaluation
• Trilateration
o Cost Analysis
50
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
0 2 4 6 8 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability 165 data
110 data
55 data
35 data
25 data
51. Evaluation
• Cost Analysis
51
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
0 2 4 6 8 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability
165 data
110 data
55 data
35 data
25 data
0 1 2 3 4 5 6 7 8 9 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability 165 data
110 data
55 data
35 data
25 data
Trilateration Doughnut
52. Evaluation
• Cost Analysis
52
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
25 45 65 85 105 125 145 165 185
Number of training data
Averageerror
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
25 45 65 85 105 125 145 165 185
Number of training data
Averageerror
Trilateration Doughnut
53. Evaluation
• Accuracy of Trilateration and Doughnut
53
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
4AP
0 1 2 3 4 5 6 7
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability
Doughnut
Trilateration
Trilateration with diff exponent
Average error:
Doughnut: 1.56 m
Trilateration: 1.76 m
Tri piecewise: 1.64 m
54. Evaluation
• Accuracy of Trilateration and Doughnut
54
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
Average error:
Doughnut: 2.08 m
Trilateration: 3.01 m
Tri piecewise: 3.22 m
0 2 4 6 8 10 12 14
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability
Doughnut
Trilateration
Trilateration with diff exponent
3AP
55. Evaluation
• Accuracy of Trilateration and Doughnut
55
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
0 1 2 3 4 5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability
Doughnut
Trilateration
Trilateration with diff exponent
Average error:
Doughnut: 1.74 m
Trilateration: 1.65 m
Tri piecewise: 1.56 m
3AP
56. Evaluation
• Accuracy of Trilateration and Doughnut
56
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
Average error:
Doughnut: 1.76 m
Trilateration: 2.27 m
Tri piecewise: 2.28 m
0 1 2 3 4 5 6 7
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability
Doughnut
Trilateration
Trilateration with diff exponent
3AP
57. Evaluation
• Accuracy of Trilateration and Doughnut
57
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
Average error:
Doughnut: 1.90 m
Trilateration: 1.75 m
Tri piecewise: 2.04 m
0 2 4 6 8 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance error (m)
Probability
Doughnut
Trilateration
Trilateration with diff exponent
3AP
59. Conclusion & Future Work
• Doughnut
o Using CSI rather than RSSI
o Remove unlikely locations rather than estimate
the most probable location
o Based on our data, the accuracy is improved by
11.6%
• Trilateration: average error =1.764 m
• Doughnut: average error = 1.560 m
o Reduce the impact of environmental change
59
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
60. Conclusion & Future Work
• Develop a mobile navigation system
o Include time series filter (Kalman or particle filter)
• Large room
o Parking garage
o Supermarket
60
• Introduction
• System Model
• Methodology
• Evaluation
• Conclusion & Future Work
各位口委以及他在場的聽眾大家好!我是蔡欣穆老師的學生:蔡彰寧。今天我要講的題目是Implementation and Evaluation of Indoor Localization system using WiFi Channel State Information!
(使用無線區域網路頻道狀態資訊的室內定位系統實作與評估)