Indoor Location Tracking Using RF Signal Strength for WLAN Networks

Transcription

1 Indr Lcatin Tracking Using RF Signal Strength fr WLAN Netwrks By: Ning Chang Advisr: Dr. M. Ahmadi C-advisr: Dr. R. Rashidzadeh Departmental Reader: Dr. R. Muscedere Departmental Reader: Dr. M. Khalid External Reader: Dr. W. Abdul-Kader

2 Outline f the Presentatin Why Indr Psitining? Existing Slutins Wi-Fi Based Lcatin System The Architecture f Wi-Fi Psitining System The Challenge f Wi-Fi Psitining System State f the Art Micrsft Research s RADAR (INFOCOM 2000) University f Maryland s Hrus (PerCm 2003) Prpsed Methd Future Wrks 2

3 Healthcare Why Indr Psitining? Imprve quality f care fr mentally impaired peple and reduce capital and peratinal expenses and increase safety Asset Visibility fr Manufacturing Facilities Tracking industrial equipment. Real-time visibility f assets thrughut the manufacturing prcess t ptimize peratins, increasing prductin thrughput and cutting csts. Security Access cntrl f wireless devices and detecting device psitins. Entertaining Lcatin based entertainment 3

4 Existing Slutins Transpnder based psitining systems : Radi frequencies [1] Ultrasund [2] Infrared [3] Advantage: Accuracy (1.8meter) Disadvantage: Need heavy infrastructure 4

5 Existing Slutins (Cnt.) Cmputer visin[4] Advantages: N tags required High accuracy Disadvantages : Blind spts Rapid changes in lightning cause errr Targets shadw each ther 5

6 Existing Slutins (Cnt.) Current Wi-Fi Base Lcatin System Advantages: Accurate (1m<errr<4m) Lw cat: uses the exiting Wi-Fi infrastructure Disadvantages: Envirnmental factrs affects the perfrmance and accuracy Needs manual calibratin n regular base r cnsiderable number f reference tags t update radi map 6

7 The Architecture f Wi-Fi indr Psitining System The strength f the Wi-Fi signals received frm different access pints is used as a signature t determine the lcatin f a user. 7

8 Challenges f Wi-Fi Based Psitining Systems Signal prpagatin suffers frm severe multipath fading, and interference in an indr envirnment. Dynamic factrs: peple presence and mvements, humidity. 8

9 Wi-Fi Based Psitining Prcess Tw phases: (a) ffline Training and (b)nline Lcalizatin Offline phase cllect samples frm signal space t build a radi map Online phase Capture the Received Signal Strength (RSS) f access pint s and then find the best match in the radimap t determine the lcatin. 9

10 State f the Art (Academia) Micrsft Research s RADAR [5] K-Nearest Neighbr Methd Offline fr each lcatin, cmpute the signal mean Online estimate lcatin with KNN and triangulatin Strength Small number f samples culd estimate the signal with a reasnable accuracy Weakness Psitining accuracy is relatively lw (errr >3m) 10

11 State f the Art (Academia) University f Maryland s Hrus [6] Maximum Likelihd Estimatin (MLE) Offline fr each lcatin, build the Radi Map f each AP Online apply ML algrithm fr estimatin Strength Gd accuracy (errr <3m) Weakness Needs relatively large number f samples t cnstruct radimap Envirnmental factrs can deterirate the perfrmance 11

12 State f the Art ( Industry) Ekahau Prbabilistic Mdel adpted Gives (x, y, flr) Reference tag needed Accuracy f abut 1-3 meters AerScut Prbabilistic Mdel adpted Gives (x, y, flr) Ultra-Wideband (UWB) needed High accuracy 12

13 Prpsed Methd (Differential Access Pints) The prpsed methd is an analgy t the traditinal differential amplifiers where nise and interference are eliminated thrugh a differential peratin. The methd assume tw access pints are placed at different distances frm a receiver inside a rm, hence path lss between them and the receiver remains equal. 13

14 Mathematical Mdel f the Prpsed Methd The strength f the received signals frm the tw access pints fixed in same rm are given : is a nrmal randm variable with zer mean in db representing the shadwing effects. Therefre the difference between the received signal strengths can be estimated by: S, frm abve equatin, we can see that the effect f shadwing is significantly reduced. 14

15 Search Algrithm Maximum Likelihd (ML) as a mst ppular searching algrithm is adpted t find the best match in the radimap: M 2 1 ( d µ rd) P( ωr ) =.exp{ } m= 1 2π ( m, m) 2 r( m, m) r Where is the Online RSS vectr, represent the O = [ 1, 2... ] ω M r crdinate f the reference lcatin. is a mean vectr and r cvariance th matrix are calculated µ and r stred fr each r during the ff-line stage. ωr Using this methd, each lcatin in the area f interest is represented by a set f PDF Mdels in the radi map. The lcatin f the receiver is where the mean squared errr is reduced t its minimum level. 15

16 Experimental Setup Test-bed: Department f ECE, University f Windsr 60 lcatins (3 by 3 meter) 50 samples per lcatin 65% fr training, 35% testing Repeat each measurement 5 times 16

17 Experimental Results Differential APs methds can reduce the signal strength s deviatin. Errr (meter) Differential APs Single APs lcatin f samples 17

18 Future Wrks Develp necessary sftware tls fr real-time indr lcalizatin using differential access pints Cmpare the perfrmance f the prpsed methd with the state f the art techniques using real-time data. Implement kernel-based smthing methd t eliminate the RSS variatin by small scale fading. 18

19 References [1] Hyunggi Ch, Myungsek Kang, Jnghyuk Park, Byungsung Park and HagbaeKim Perfrmance Analysis f Lcatin Estimatin Algrithm in ZigBeeNetwrks using Received Signal Strength, 21st Internatinal Cnference n Advanced Infrmatin Netwrking and Applicatins Wrkshps [2] Nissanka Bdhi Priyantha. The Cricket Indr Lcatin System, Dctral Thesis. Department f Electrical Engineering and CmputerScience.Massachusetts Institute f Technlgy, June 2 [3] R. Want, A. Hpper, V. Falcã, J. Gibbns. The Active Badge LcatinSystem. ACM Transactins n Infrmatin Systems, vl. 10, Jan pp [4] R. Bdr, B. Jacksn, and N. Papaniklpuls, Visin-based human tracking and activity recgnitin. Prceedings f the 11th Mediterranean Cnf. n Cntrl and Autmatin, Jun [5] P. Bahl and V. N. Padmanabhan, RADAR: An in-building RF-based user lcatin and tracking system, in Prc. INFOCOM, pp , [6] M. Yussef and A. Agrawala, Handling samples crrelatin in the Hrus system, in Prc. INFOCOM, pp , 2004 [7] J. Bach Andersen, T. S. Rappaprt, and S. Yshida, Prpagatin measurements and mdels fr wireless cmmunicatins channels, IEEE Cmmun. Mag., vl. 33, n. 1, pp , Jan