Signal Dragging Signal Dragging: Effects of Terminal Movement on War-Driving in CDMA/WCDMA Networks Daehyung Jo MMLab., Seoul National University LNCS

MMLab Table of Contents 1. Introduction 1. Introduction 2. What is Signal Dragging 2. What is Signal Dragging 3. Properties and Implications 3. Properties and Implications 5. Conclusion 5. Conclusion 4. Technical Reasons 4. Technical Reasons 2

MMLab Introduction  Network-based localization  Network-based localization is essential for ubiquitous computing and LBSs Availability of GPS is limited  Resources for network-based localization RSS RSS, TOA, TDOA Cell-ID, antenna orientation and opening  Popular RSS-based localization algorithms Pattern matching (PM) – scene analysis Centroid family - lateration Particle filter – Monte Carlo method  War-driving Practical signal information collection procedure 3

MMLab Signal Pattern Signal Pattern Data  Signal pattern is a series of [RSS, BS ID] pairs  Signal pattern data obtained through war-driving CDMA and WCDMA measurements in Seoul, Korea WCDMA measurements in Seattle, USA Measurement area in Seoul, Korea, 25 km 2 4

MMLab Signal Dragging What is Signal Dragging  A phenomenon A moving mobile terminal tends to retain signal information of old BSs than newly appearing BSs Sum of BS vectors = reverse of moving direction BS vectors stemming from the terminal to BS Sum of all received BS vectors Reverse of sum vector Actual moving direction 5

MMLab Signal Dragging in Real  Signal dragging occurs in cellular networks when the terminal moves fast enough In real circumstances Reverse of sum vector = Estimated direction vector 6

MMLab SDEM Signal Dragging Error Metric (SDEM)  The average of angular difference between the actual user direction vector and the estimated direction vector  If SDEM is less than 90 degrees Signal dragging has occurred mean 31.5 degreesmean 72.6 degrees SDEM distribution 7

MMLab Table of Contents 1. Introduction 1. Introduction 2. What is Signal Dragging 2. What is Signal Dragging 3. Properties and Implications 3. Properties and Implications 5. Conclusion 5. Conclusion 4. Technical Reasons 4. Technical Reasons 8

MMLab Speed Properties: Correlation with Speed  As the terminal moves faster, the signal dragging becomes more notable correlation coefficient speed and (180−SDEM)  Compute the correlation coefficient between the terminal’s moving speed and (180−SDEM) Correlation coefficient is 0.48 in a typical straight road Mean and deviation of the terminal speed in km/h is 31 and 12 each 9

MMLab Direction Change Properties: Direction Change  Estimated direction arrows converge to the changed direction with some delay Increase SDEM in curved areas WCDMA trajectory in Seoul, Korea SDEM distribution 10

MMLab Arrangement of BSs Properties: Arrangement of BSs  The arrangement of BSs and the geographical environment affects the efficacy of signal dragging Fundamental factor to increase SDEM WCDMA trajectory in Seattle, USA SEA 11

MMLab Direction and Arrangement Together  Signal dragging is prevailing in both directions WCDMA data sets in Seoul, Korea Initial direction Reverse direction 12

MMLab Direction and Arrangement Together  Signal dragging hardly occurs due to the uneven arrangement of BSs Signal pattern is different enough to affect the result of localization performance CDMA data sets in Seoul, Korea Initial direction Reverse direction 13

MMLab Implications: PM seed sample  PM system compares pattern database or seed with user’s signal pattern or sample  Signal pattern can be different depending on the movement context in war-driving Different PM results  Potential hint for improvement Construct pattern DB in diverse movement contexts 95 percentile errors (m) Seed by SampleWCDMACDMA Initial by Initial Initial by Reverse Reverse by Reverse Reverse by Initial

MMLab Implications: Centroid Family  Centroid family algorithms do not compare signal patterns Similar results on both directions  Potential hint for improvement Cut out unnecessary BS signals if signal dragging prevails 95 percentile errors (m) DataCentroid Weighted Centroid Cell ID WCDMA initial WCDMA reverse CDMA initial CDMA reverse

MMLab Implications: Direction Estimation  The way we calculate SDEM provides the estimation of moving direction Use only one time signal pattern GT  We have drawn direction arrows based on the BS vectors stemming from GT the resulting position of localization  No big difference if we use the resulting position of localization instead of GT AreaGT basedWC based 72.6 (64.0)73.9 (65.6) SDEM mean (median) 16

MMLab Why Signal Dragging Occurs Technical reasons why signal dragging occurs Synchronization to newly found BS is difficult Pilot signal broadcast every ms Multipath fading becomes severe when moving fast Synchronization Pilot channels are 4 sets managed by 4 sets in CDMA Keep old pilot channel information longer Lower its set priority level when its RSS is weakened Pilot set management Operational characteristic slotted/DRX mode of slotted/DRX mode Terminal in an idle mode wakes up periodically to save its power Updates its signal pattern with relatively long interval Idle mode 17

MMLab Conclusion  Signal Dragging  A phenomenon showing a significant relationship between the signal pattern and the movement context of war- driving Natural phenomenon due to the CDMA/WCDMA mechanism  Understanding Signal Pattern PM results can be different depending on the movement context of war-driving Direction context can be extracted naturally 18

MMLab