cis 632 / eec 687 mobile computing...- samsung galaxy s test set - 594 fps 22 1,734 aps 2,028...
TRANSCRIPT
1
CIS 632 / EEC 687Mobile Computing
Indoor Localization
Chansu Yu
Next-Generation Services
2
2
3
Spotlight Mobile, Founded’02, from Cornell
Uses GPS Pseudo-lites
ShopKick - A $100-device at a store sends a special-pitch sound so that the Smartphone app can pick it up to offer discounts, etc.
3
• Exquisite combination and integration of methods, tools, and services Incorporated a variety of COEX services with the indoor
navigation system in myCoex app Support turn-key based proposal for any sites similar to
COEX
WiFi fingerprint
capturing tool
Indoor routing path
drawing and mapping tool
AP environment
analysis tool
4
Integrated First Responder Location and Physiological Monitoring
aa
5
Indoor Localization Techniques
Active Badge (Infrared)
Active Bat, Cricket (Ultrasonic)
UbiSense (UWB)
Smart Floor (Pressure Sensor)
Spot-On (RFID)
RADAR (Wi-Fi)
Wifi-based Indoor Localization or Positioning System (WPS)
Characteristics of Wi-Fi Wi-Fi signal reaches 35m ~ 95m
Each AP has globally unique ID (BSSID)
RSS (Received Signal Strength) in dBm
SSID: XUnivBSSID:00:0e:83:39:a9:80
RSS: -40 (dBm)
SSID: XunivBSSID:00:0e:83:39:a9:80
RSS: -90 (dBm)
No Wi-Fi Signals
in indoors (32m), in outdoors (95m)
6
WPS
Triangulation-Based
Fingerprint-Based
Triangulation-Based WPS
Beacon Database War-Driving/War-Walking
Distance from AP Radio Propagation Model
Location Determination Least Squares Estimation
Davidon Algorithm
AP1
AP2
AP3
d1
d2
d3
7
Fingerprint-Based WPS
Building Radiomap (offline phase) Empirical
Generation
Location Determination (online phase) Nearest Neighbor
SVM
Bayesian
Neural Network
Fingerprint-Based WPS –Offline Phase
* Effective learning data creationSampling-based Signal Propagation ModelInterpolation Model
8
Fingerprint-Based WPS –Online Phase
f1 = {AP1=-44, AP2=-53… APn =-90}
f2 = {AP1=-44, AP2=-58… APn =-93}
……
fk = {AP1=-89, AP2=-97… APn =-52}
Database
Estimation location is the fingerprint which has the
minimum distance among fingerprints in the radiomap
Time complexity = O(|fingerprints| x |APs|)
* Improving accuracyNearest Neighbor Method (kNN)Naïve Bayesian Method Histogram Method
Measured FP fo
kNN (k=3) : obtain the 3 nearest FPs (fx, fy, and fz) => estimated location = average ofloc(fx)+ loc(fy) + loc(fz)
Fingerprint-Based WPS –Clustering (1)
Computational Cost Reducing Methods Reducing search space
Clustering
Decision Tree
Reducing comparison cost AP Select
AP Clustering
9
Fingerprint-Based WPS –Clustering (2)
Fingerprint clustering using the strongest AP(s)
Cluster C1 = set of FPs, the strongest AP of which is AP1Cluster C2 = set of FPs, the strongest AP of which is AP2…
If AP4 is the strongest AP in the measured fo, search C4.
Fingerprint-Based WPS –Clustering (3)
Fingerprint clustering using the strongest AP(s)
Cluster C1 = set of FPs, the strongest AP of which is AP1Cluster C2 = set of FPs, the strongest AP of which is AP2…
If AP5 & AP13 are the strongest APs in the measured fo, search C5 U C13.
10
WiFi indoor localization
High accuracy indoor localization
WiFi enabled smartphone indoor localization
RADAR [INFOCOM’00], Horus [MobiSys’05], Chen et.al[P
ercom’08]
Cricket [Mobicom’00], WALRUS [Mobisys’05], DOLPHIN [Ubicomp’04], Gayathri et.al [SECON’09]
SurroundSense [MobiCom’09], Escort [MobiCom’10], WILL[INF
OCOM’12], Virtual Compass [Pervasive’10]
Academic Research
RADAR [INFOCOM’00]: P. Bahl and V. N. Padmanabhan. RADAR: An In-building RF-based User Location and Tracking System. INFOCOM’00. Cricket [Mobicom’00]: N. Priyantha, A. Chakraborty, and H. Balakrishnan. The Cricket Location-support System. MobiCom’00. DOLPHIN [Ubicomp’04]: M. Minami, Y. Fukuju, K. Hirasawa, and S. Yokoyama. DOLPHIN: A Practical Approach for Implementing A Tully Distributed Indoor Ultrasonic Positioning System. Ubicomp’04.WALRUS [Mobisys’05]: G. Borriello, A. Liu, T. Offer, C. Palistrant, and R. Sharp. WALRUS: Wireless Acoustic Location with Room-level Resolution Using Utrasound. MobiSys’05. Horus [MobiSys’05]: M. Youssef and A. Agrawala. The Horus WLAN Location Determination System. MobiSys’05.Beepbeep [Sensys’07]: C. Peng, G. Shen, Y. Zhang, Y. Li, and K. Tan. Beepbeep: A High Accuracy Acoustic Ranging System Using Cots Mobile Devices. Sensys’07.Chen et.al [Percom’08]: S. Chen, Y. Chen and W. Trappe. Exploiting Environmental Properties for Wireless Localization and Location Aware Applications. PerCom’08.Gayathri et.al [SECON’09]: G. Chandrasekaran, M. A. Ergin, J. Yang, S. Liu, Y. Chen, Marco Gruteser and Rich Martin. Empirical Evaluation of the Limits on Localization Using Signal Strength. SECON’09.SurroundSense [MobiCom’09]: M. Azizyan, I. Constandache, and R. R. Choudhury. Surroundsense: Mobile Phone Localization via Ambience Fingerprinting. MobiCom’09.Escort [MobiCom’10]: I. Constandache, X. Bao, M. Azizyan, and R. R. Choudhury. Did You See Bob? Using Mobile Phones to Locate People. MobiCom’10.Virtual Compass [Pervasive’10]: N. Banerjee, S. Agarwal, P. Bahl, R. Chandra, A. Wolman, and M. Corner. Virtual compass: relative positioning to sense mobile social interactions. Pervasive’10.WILL [INFOCOM’12]: C. Wu, Z. Yang, Y. Liu, and W. Xi. WILL: Wireless Indoor Localization Without Site Survey. INFOCOM’12.
Conferences
IEEE/ACM Mobicom
ACM MobiSys
IEEE Infocom
IEEE Percom
People Tracking conference
The Location Business Summit USA, 2012, San Jose
3rd International Conference on Indoor Positioning and Indoor Navigation (IPIN), 2013, Portugal
The Third International Conference on Mobile Services, Resources, and Users (MOBILITY 2013)
20
11
Large-scale Venue
High localization errors
Generate huge amount of unnecessary traffic
21
Environment- COEX, Seoul- 505 x 237m
APs- 1,734 APs
Data set- Learning Set: 2,028 FPs- Samsung Galaxy S
Test Set- 594 FPs
22
1,734 APs
2,028Locations(FPs)
12
23
AP cardinality (The number of APs observed at a location is an order of magnitude larger at Coex compared to HKUST. Clustering may not work well due to the large feature set per FP. A trend line is added for the distribution of Coex.)
Histogram of RSS (Weaker signals are received better at HKUST partly due to less interference and less movements at an academic building of HKUST. The mode is about 10 dB different.)
Problems in Large-Scale Venues
To obtain FP, a mobile station sends a probe request message (broadcast). It waits for probe response message from APs at Channel 1
for MinChannelTime (1msec).
If none, switch to the next channel.
If yes, wait for MaxChannelTime (10msec).
With many APs in the vicinity, Some APs continue to send probe response message even
though the mobile station switches to the next channel (Probe message explosion)
It leads to incomplete FP, leading to localization accuracy
24
13
Slide 25
Problems in Large-Scale Venues (2)AP 1-1 AP 1-2 AP 1-3 AP 1-… AP 1-nSTA
Probe Request on channel X
Probe Response
AP 1-4 AP 1-5 AP 1-6
Probe Response
Probe Response
Probe Response
Probe Response
Probe Response
Probe Response
Probe Response
Probe Response
Probe Response
Probe ResponseProbe Response
Probe Response
Retransmit Probe Response
MaxChannelTime
STA cannot receive
Probe Response
after MaxChannelTime
Unnecessary
Probe Response
Unnecessary
Retransmission
Probe Response
flooding
Retransmit Probe Response
Retransmit Probe Response
Retransmit Probe Response
Retransmit Probe Response
Retransmit Probe Response
Problems in Large-Scale Venues (3)
26
(Un)detected APs (It shows the number of (un)detected APs. Some APs are not detected via fingerprinting in both Coex and HKUST. While it is mostly due to temporal collisions and random radio propagation in HKUST, it is due to the limited channel time allowed to get responses in comparison to the number of APs in the proximity in Coex.)
14
Remember!It All Depends on RF Propagation!
In free space, receiving power proportional to 1/d² (d = distance between transmitter and receiver)
Suppose transmitted signal is Pt, received signal Pr = h Pt, where h is proportional to 1/d²
2
4
dGG
P
Ptr
t
r
Pr: received power
Pt: transmitted power
Gr, Gt: receiver and transmitter antenna gain
(=c/f): wave length