FREQUENCY HOPPING FOR ROBUST INDOOR LOCALIZATION IN DAILY ENVIRONMENTS

1、Sensor Network for Localization and Its Application in Healthcare,Polly Huang Electrical Engineering National Taiwan University .tw/,1,Road Map,Sensor network? Indoor localization? The system & testbed Technical challenges Causality Study Frequency Hopping Hospital Installation

2、,Sensor Network,Sensors?,Also the biomedical sensors: EMG, EKG, pulses, emotions, etc,camera,mic,accelerometer,gyro,pressure,thermal,GPS,Sensor Nodes Today,MICA, 2001-2002 5.7cm X 3.18cm 4 MHz CPU 128K ROM 512K RAM 40kbps Radio range x00 feet Sensors, battery not included,Embedded Sensor Node,Intel

3、Xscale CPU Analog and digital radio Flash and SRAM memory Sensors,Embedded Sensor Network,Indoor Localization,Localization Systems,Outdoor GPS Indoor UWB, ultrasound, infrared, etc RF-based system,9,Radio as Location Sensor,802.15.4 (Zigbee) radio RF radio Active RFID,Photo courtesy: Crossbow,Beacon

4、,Tag,Signal Strength Propagation Time, Distance,Classic Techniques,Ranging & Triangulation Each radio source to a receiver RSSI distance Radio Fingerprinting A bunch of radio sources to a receiver RSSIs(position_X) RSSIs(position_Y),11,Localization - Triangulation,Infrastructural Beacons,Mobile User

5、s,(x, y, z),(x1, y1, z1),(x2, y2, z2),(x3, y3, z3),(x4, y4, z4),(x5, y5, z5),Requirement,Important property of the electromagnetic wave used Radio time of flight transmitter receiver Proportional to distance transmitter receiver High frequency EM waves, e.g., Infrared Ultrasound Problem Line of sigh

6、t Placement of tags a challenge,Localization Radio Fingerprint,Beacon 4,Beacon 1,Beacon 3,Beacon 2,(B1,B2,B3,B4),Look up the table,Road Map,Sensor network? Indoor localization? The system & testbed Technical challenges Causality Study Frequency Hopping Hospital Installation,1. Survey Phase,Radio Fin

7、gerprints,2. Tracking Phase,RSSI,Radio Fingerprints,Location Estimation,K-Nearest Neighbor(KNN) K=3,18,K closest fingerprints (with minimum Euclidean distance),Output Position,Radio Fingerprints,Tracking : Beacon 1=-70 Beacon 2=-77 ,Weighted Average,Modulator Demodulator,Testbed Deployment,19,PC gat

8、eway + Power supply,USB Hub,Beacon nodes,Training area,USB Extender,Maximum Transmission Distance: Issue I,Power Supply: Issue II,Power Confusion: Issue III,Localization Error,Road Map,Sensor network? Indoor localization? The system & testbed Technical challenges Causality Study Frequency Hopping Ho

9、spital Installation,Antenna Orientation: Experiment I,Average location error: 0.88 m,Average location error: 2.05 m,Average location error: 3.17 m,Antenna Radiation Pattern,On-board antenna,Radiation Pattern,Foreground Obstacle: Experiment II,Tag on chair,Tag in hand,Human wandering around,Human sta

10、nding still,Foreground Obstacle Experimental Result,RSSI of tag on chair: More stable RSSI of tag held in hand: High in variance,Foreground Obstacle Experimental Result,RSSI of tag on chair: More stable RSSI of tag held in hand: High in variance,(a),(b),Foreground Obstacle Experimental Result,RSSI o

11、f tag on chair: More stable RSSI of tag held in hand: High in variance,(a),(b),Average localization error: 12:00AM 7:00AM: 1.58m 9:00 AM 12:00AM: 1.52m,Daytime Midnight,Background Noises: Experiment III,Beacon Density: Experiment IV,1/3 beacons (I),1/3 beacons (II),1/3 beacons (lll),Take Aways,Major

12、 factors: Antenna orientation Foreground obstacles Suggestions: Use of uniform radiation pattern antenna Same antenna orientation in training and tracking The method in training and tracking should be consistent. Denser beacon deployment higher accuracy Not an issue: Background noise,Road Map,Sensor

13、 network? Indoor localization? The system & testbed Technical challenges Causality Study Frequency Hopping Hospital Installation,Radio InterferenceA Common Problem,35,Location estimation,(beacon1, RSSI1) (beacon2, RSSI2) (beacon3, RSSI3) (beacon4, RSSI4) .,Interference,Beacon node 1,Beacon node 2,Be

14、acon node 3,Beacon node4,Interference Localization Error,Implementation Localization Zigbee RSSI-signature Data WiFi Experiments WiFi data rate 68 2835 kbps,36,1.60m,3.85m,Cause: Beacon Packet Loss,37,Problem Statement,Frequency hopping to mitigate the impact of interference on localization accuracy

15、!,38,Longer beacon packet collection time? 1. Location sampling delay 2. Data traffic drop due to interference,Contributions,Quantify the impact of interference Devise a frequency hopping mechanism Evaluate localization stability under realistic influences,39,Done!,Next!,Mechanism Components,How to

16、hop In a relay fashion When to hop A diagnostic test decide whether to hop,40,Hopping Notification,Diagnostic Test,Hop,Timeout,41,Diagnostic Test 3 Parts,Easy part Not-so-easy part Difficult part,42,The Easy Part,Strong evidence Beacon loss Location error Beacon packet loss A good criteria Packet Re

17、ception Rate (PRR) PRR # packets received in T / # packets sent in T 5 packets/sec,43,The Semi-Easy Part,Beacon loss Location error Observed at the mobile tag Need to know PRR from a beacon when theres no interference So mobile tag can tell if PRR is relatively PRR from a beacon Differ depending on

18、Location of mobile tag,44,The Solution,Every beacon node listens to other beacon nodes! At system setup (no interference) Recording PRR from all other beacons Select good neighbor beacons During operation Record PRR from good neighbor beacons Use this PRR to determine whether theres interference,45,

19、The Difficult Part,How bad PRR interference Too low = beacon packet losses Too high = hop unnecessarily,46,Proposed Solutions,1. Thresholding Measure the system for a while Find a proper threshold that minimize error Brute-force search (nave ”) 2. Hidden Markov Model(HMM) Measure the system for a wh

20、ile Train a model based on measured relationship between PRR vs. localization errors More scalable (scientifically sound),47,Why HMM?,Time dependency in background WiFi traffic Quiet Quiet & Busy Busy (Often) Quiet Busy & Busy Quiet (Seldom) Time dependency in PRR Background WiFi Traffic Beacon PRR

21、HMM models temporal pattern in PRR Transition probability,48,transit ,remain ,Quiet,Busy,Data Sets: Real WiFi Usage,49,Localization Errors,1.24m,2.74m,1.32m,1.82m,a slight penalty of receiving no beacon packets during transition of hopping,50,Hopping delay,Loss pattern of hopping message beacon mess

22、age Time for all beacons to hop to the next channel is simulated by beacon PRR,51,No Timeout,Take Aways,RSSI-based localization system sensitive to interference Localization error under interference Degrade by 141% Frequency hopping effective Localization error with frequency hopping Reduced to 45%,52,Road Map,Sensor network? Indoor localization? The system & testbed Technical challenges Causality Study Frequency Hopping Hospital Installation,台大醫院北護分院室內定位系統,54,室內定位系統,甚麼是定位系統? 找出人或物品目前所在的位置 市面上有沒有商品化的定位系統 GPS衛星定位 可是，衛星定位