Stochastic Radio Interferometric Positioning with Unsynchronized Modulated Signals in Wireless Sensor Networks

Transcription

1 Sochasic Radio Inerferomeric Posiioning wih Unsynchronized Modulaed Signals in Wireless Sensor Neworks Ruiling Gao, Wei-Tse Sun, Alva Couch, Chorng Hwa Chang Deparmen of Elecrical and Compuer Engineering Tufs Universiy Medford, MA, USA Deparmen of Compuer Science Tufs Universiy Medford, MA, USA Absrac Sensor localizaion problem is one of he fundamenal issues in Wireless Sensor Neworks WSN. In his paper, we invesigae his problem and presen an innovaive Radio Inerferomeric Posiioning Sysem RIPS for WSN - Sochasic Radio Inerferomeric Posiioning Sysem wih Unsynchronized Modulaed Signals SRIPS_UMS. To he bes of our knowledge, we are he firs o uilize unsynchronized modulaed signals for radio inerferomeric posiioning and demonsrae is viabiliy in heory. Alhough previous radio inerferomeric posiioning mehods provide cenimeer accuracy, hey are sill no widely adoped due o 1 sric hardware requiremens for Received Signal Srengh Indicaor RSSI circuiry 2 ime synchronizaion problem beween disribued hardware. SRIPS_UMS overcomes hese limiaions by using unsynchronized modulaed signals. Is viabiliy and advanages are demonsraed hrough mahemaical models and simulaions. Exensive simulaions based on radio chip CC2420 demonsrae ha SRIPS_UMS is an efficien posiioning sysem for oudoor WSN. Index Terms sensor localizaion; wireless sensor neworks; radio inerferomeric posiioning sysem; unsynchronized modulaed signals; received signal srengh indicaor; ime synchronizaion; CC2420 I. INTRODUCTION Wireless Sensor Neworks WSN have been aracing explosive ineres from boh academia and indusry 1 2. Sensors have been deployed all over he world o collec and process vas amouns of daa for various of civilian and miliary applicaions, such as smar ciy, weaher sensing, and earhquake monioring. Many wireless nework applicaions depend on locaion informaion o ge he conex of he measured daa or o provide suppor o nework services e.g. rouing. Localizaion can also be a sand-alone applicaion, such as navigaion and arge racking. Mos radiional localizaion sysems, such as Global Posiioning Sysem GPS, are no suiable for WSN because sensors are resource-consrained. Compared wih mos exising localizaion sysems, a quilified localizaion scheme for WSN addiionally needs o have good scalabiliy and low CPU/power/memory consumpion. In his paper, we focus on localizaion in saic Wireless Sensor Neworks in oudoor environmen. Mos curren localizaion soluions are based on Received Signal Srengh RSS 3, 4 using an indicaor of RSS called Received Signal Srengh Indicaor RSSI ha is available on mos sensor radio chips. However is accuracy is no always saisfacory and much calibraion and mainenance work are required. Oher mehods relying on Time Difference of Arrival TDOA, Time of Fligh TOF, and Angle of Arrival AOA are more accurae han he RSS mehods bu require expensive specialized hardware 5, 6. As a resul, sensor localizaion remains a challenging problem. Compared wih he above localizaion mehods, Radio Inerferomeric Posiioning Sysems RIPS offer high accuracy and long range simulaneously, wihou he requiremen of special hardware. RIPS yields an average localizaion error of 3 cm wihin an area wih a 160 m radius in oudoor environmen 7. Many effors have been made o improve RIPS However, i is sill difficul o implemen RIPS algorihms on exising hardware: Many curren radio chips have changed he design of RSSI circuiry. Resoluion requiremen of ime synchronizaion canno be achieved by many curren sensor hardware. Mica2 is proven o be appropriae for RIPS/SRIPS implemenaion, bu i is oo old and is no available in he marke now. Advanced sensors wih direc phase measuremen are no available in producion. ZigBi wih proper micro conroller and radio chip is no suppored in exising producs; some pins may be no conneced and neiher Coniki nor TinyOS suppor i. None of he sensors suppored by he laes version of Coniki and TinyOS has been proved /17/$ IEEE ICIS 2017, May 24-26, 2017, Wuhan, China 57

2 In his paper, we presen SRIPS_UMS 1 a varian of Sochasic RIPS 11 and Sochasic RIPS wih Modulaed Signals 12 ha makes previous RIPS mehods possible on exising hardware. Previous RIPS mehods require sensor hardware o be able o coninuously sample RSS value, such as found on he CC They canno be implemened on curren mainsream hardware such as CC and CC , which insead average he RSS value every 128 us. Anoher problem is ha previous mehods using inerferomeric modulaed signals require microsecond-level ime synchronizaion on senders ha is very hard o be realized on many sensor plaforms. We need an innovaive algorihm ha can operae on he new-generaion RSSI circuiry and also has much lower ime synchronizaion requiremens. Our improved algorihm SRIPS_UMS makes his possible and is viabiliy is demonsraed heoreically hrough mahemaical models and simulaions. Hardware requiremens for differen RIPS mehods are lised in Table I. Direc Sequence Spread Specrum Offse- Quadraure Phase Shif Keying DSSS O-QPSK modulaion is used by many mainsream radio chips This modulaion mehod is also used in mahemaical models and simulaions of his paper. TABLE I HARDWARE REQUIREMENTS FOR DIFFERENT RIPS METHODS Hardware Requiremens RIPS SRIPS SRIPS_MS SRIPS_UMS Average RSS Real-ime RSS Unmodulaed Radios Fine-uning Frequencies Our conribuions in his paper are: We inroduce an innovaive radio inerferomeric posiioning sysem called SRIPS_UMS and heoreically demonsrae ha i can be implemened on curren mainsream radio chips ha do no suppor previous RIPS mehods. We heoreically demonsrae ha i is no necessary for SRIPS_UMS o synchronize senders in ime or o synchronize senders wih receivers in ime. Those ime offses almos have no affec on he posioning accuracy. The remainder of he paper is srucured as follows. In secion II, we inroduce SRIPS_UMS, including is mahemaical models and some relaed simulaion resuls. In secion III, furher simulaion resuls are presened based on CC2420. Finally conclusions are given. previous RIPS mehods are firs inroduced. To solve hese problems, we come up wih SRIPS_UMS. Then we presen he mahemaical models and some preliminary evaluaion resuls of SRIPS_UMS. Mahemaical models are described in deails o demonsrae he viabiliy of SRIPS_UMS in heory, including signal modulaion, signal propagaion, signal inerference, composie signal mix-down process, RSSI sampling, phase offse calculaion, and posiioning algorihm. A se of phase offses are needed so ha he posiioning algorihm can converge o localize he arge. SRIPS_UMS is a combinaion of he SRIPS model and he SRIPS_MS model. SRIPS model uses unmodulaed signals while SRIPS_MS model and SRIPS_UMS model are based on modulaed signals. SRIPS_UMS model improves SRIPS_MS model by geing rid of he need o synchronize senders in ime o microsecond resoluion. A. Hardware Limiaions 1 RSSI Circuiry: One imporan hardware requiremen for previous RIPS mehods, including SRIPS_MS, is ha he RSSI circuiry of receivers should be capable of sampling RSS values coninuously. However, many curren radio chips have abandoned his and insead only expose an average RSS value over 128 us. Alhough an obvious phase offse exiss for acual RSS signals a he wo receivers, he measured RSSI-ime curves are almos he same in phase due o he average process of RSSI circuiry. This will cause huge errors in esimaing he arge s posiion. There is an urgen need o improve radiional RIPS mehods by resolving his problem. 2 Time Synchronizaion : I has been demonsraed in heory ha if senders are synchronized wih receivers in microsecond accuracy, phase offse of RSS signals can be direcly obained from RSS values coninuously measured by receivers in Fig However, when he ime synchronizaion requiremen canno be me, his relaionship breaks up and a new sraegy is needed, which is anoher moivaion for us o come up wih he following mahemaical models using unsynchronized modulaed signal for posiioning a arge. II. STOCHASTIC RADIO INTERFEROMETRIC POSITIONING WITH UNSYNCHRONIZED MODULATED SIGNALS SRIPS_UMS uilizes radio inerferomery of wo unsynchronized modulaed signals wih nearby carrier frequencies o generae a composie signal wih low-frequency envelope for posiion esimaion. In his secion, some problems of 1 Sochasic Radio Inerferomeric Posiioning Sysem wih Unsynchronized Modulaed Signals Fig. 1. Phase Offse Calculaion wih Time Synchronizaion 58

3 B. Mahemaical Models of SRIPS_UMS A leas four sensors are needed o posiion a arge, wih wo as senders and he remaining ones as receivers. Only he arge s posiion is unknown. Denoe senders as A and B and receivers as C and D. D is he arge. Le X be eiher A or B, and Y be eiher C or D. Denoe he clock ime of receivers as and he clock ime of senders as X. f X is he carrier frequency of sender X and a X is he signal ampliude. Assumpions for mahemaical models: Two receivers are synchronized in ime and are lisening simulaneously. Two senders are sending he same symbol sequence coninuously. Symbol sequence is randomly formed from symbol 0~15. Senders have close carrier frequencies f A and f B le f B f A. 1 DSSS O-QPSK Modulaion wih Half-sine Chip Shaping: The modulaion forma in our mahemaical models is DSSS O-QPSK wih half-sine chip shaping, which is equivalen o Minimum-Shif Keying MSK modulaion in mahemaics. As shown in Fig. 2, he DSSS O-QPSK modulaed signal is a summaion of wo signals available on I channel and Q channel respecively. For DSSS O-QPSK, he mahemaical expression for he generaed signal S mod X is: πx S mod X =a I X cos cosf X X πx a Q X sin sinf X X, where X =0means he saring poin when he sender X begins o send symbol sequence. a I X and a Q X are bi-polar square pulses wih ampliude of 1 and pulse edges on X = T,T,3T,... and on X =0,,,... respecively. a I X and a Q X encode even and odd chips of a symbol respecively. In radio chip CC2420, each symbol 4 bis is spread using he IEEE spreading sequence o 32 chips and oupu o he digial-o-analog converers DACs 14. One symbol period is 16 us. This can be rewrien in a form where he phase and frequency modulaion are more obvious: S mod X =cos f X X + b k X π X + ϕ k X, where b k X is +1 when a I X =a Q X and 1 if hey are of opposie signs, and ϕ k X is 0 if a I X is 1, and π oherwise. 2 Signal Propagaion : Denoe he phase inroduced by disance as ϕ d. When he signal propagaes for a wavelengh in space, is phase changes for. The changed phase is in proporion o he ravel disance. The signal received a disance d can be expressed as: 1 2 Fig. 2. DSSS O-QPSK Modulaed Signal S r X =cos f X X + b k X π X + ϕ k X ϕ d =cos b k X π X cosf X X sin + ϕ k X ϕ d b k X π X + ϕ k X ϕ d sinf X X 3 Inerference of Modulaed Signals: Since senders are no required o be synchronized wih each oher and receivers, ime offses may exis and can be expressed as: 3 X = +Δ X 4 In each phase offse measuremen round, Δ X is a consan and X is a variable. In differen phase offse measuremen rounds, Δ X for a cerain sender/receiver pair can be differen and won affec measuremen resuls in heory. The composie signal S Y received by Y is he sum of he wo signals, which can be expressed by Equaion 5: S Y =a A cos f A A + b k A π A + ϕ k A ϕ day + a B cos f B B + b k B π B = a A cos f A + b k A + + ϕ k A ϕ day + a B cos + f B + b k B + ϕ k B ϕ dby f A + b k A f B + b k B Δ B + ϕ k B ϕ dby Δ A 4 Mix Down Process of he Composie Signal: Sensor Y mixes down he composie signal S Y wih a local signal m = cosf LO. m is generaed by a local oscillaor and is used o conver he received signal o a resuling signal S Y wih an inermediae frequency f IF. I can be expressed as: 5 59

4 SY =S Y m = a A + + a A + + a B f A + b k A f A + b k A f A + b k A f B + b k B + a B + f B + b k B f LO Δ A + ϕ k A ϕ day f A + b k A + f LO Δ A + ϕ k A ϕ day f B + b k B f LO + Δ B + ϕ k B ϕ dby f B + b k B + f LO Δ B + ϕ k B ϕ dby Then SY is passed hrough a band pass filer o filer ou he higher frequencies and he wo low-frequency componens are lef. f A f B is very small compared wih b k and f LO. The resuling inermediae signal S IFY only conains wo signal componens wih frequencies f A + b k A f LO and f B + b k B f LO separaely. Se f LO o f A+f B 2 f IF and le δ be f A f B 2,soS IFY can be rewrien as: S IFY = a A + f IF + b k A f A + b k A + a B + f B + b k B + δ Δ A + ϕ k A ϕ day f IF + b k B δ Δ B + ϕ k B ϕ dby 5 RSSI Sampling: The signal power of S IFY is: 6 7 S PY =S 2 IF Y 8 Because RSSI measures RSS in dbm, a non-linear logarihmic funcion is used o read he RSS value. The logarihmic funcion disors he composie signal by adding harmonic frequencies 7. The harmonic frequencies can be filered ou using a properly designed low pass filer, leaving only he fundamenal frequency. Hence we ge he filered RSS signal S RSSY : S RSSY =log { a 2 A + a2 B + a Aa B cos 2δ b k A b k B + + ϕ k A ϕ day } ϕ k B +ϕ dby f A + b k A f B + b k B Δ A Δ B We have heoreically derived he mahemaical expression of he filered RSS signal S RSSY for composie unsynchronized modulaed signals. I is a periodic low-frequency signal wih fundamenal frequency f A f B. IEEE has a ±40 ppm frequency accuracy requiremen, which means ha f A f B generally varies from 0 ~ 200 KHz. The frequency of S RSSY conains a changing componen b k A b k B ha can be 1 1, 0, or over ime. According o CC daa shee, equals o 1 MHz. When ime changes and b k A b k B is 0, he frequency of he signal is f A f B. In oher ime periods when b k A b k B equals o ±1 MHz, he signal s frequency is larger han f A f B and he signal s period becomes shorer. This causes some spikes on he signal curve in he ime zone. As shown in Fig. 3, he curve of he filered RSS signal is no smooh and has many spikes. The phase of S RSSY conains disance informaion beween involved sensors. The absolue phase is difficul o be measured, bu he relaive phase offse beween wo filered RSS signals can be obained and used for posiioning. Fig. 3. Filered RSS Signal in 1 Symbol Period and 32 Symbol Periods 6 Phase Offse Calculaion : According o Equaion 9, we know ha he filered RSS signal of receiver C is periodic wih fundamenal frequency 2δ and absolue phase f A + b k A Δ A + ϕ k A ϕ dac f B + b k B Δ B ϕ k B +ϕ dbc. And for receiver D, 60

5 he filered RSS signal has he same fundamenal frequency and is absolue phase is f A + b k A Δ A + ϕ k A ϕ dad f B + b k B Δ B ϕ k B +ϕ dbd. So he phase offse can be expressed as: Δϕ = f A + b k A f B + b k B f A + b k A f B + b k B Δ A + ϕ k A ϕ dac Δ B ϕ k B +ϕ dbc Δ A + ϕ k A ϕ dad Δ B ϕ k B +ϕ dbd =ϕ dad ϕ dac +ϕ dbc ϕ dbd 10 ϕ dxy is he phase inroduced by he disance beween sender X and receiver Y wih ϕ dxy = d XY c/f. c is he speed of ligh in free space. ϕ dxy has a periodiciy of. SoΔϕ can be furher expressed in he form of disance relaionships as: Δϕ = d AD d AC + d BC d BD mod 11 c/f A c/f B Wih radio chips ha average RSS values over 128 us, he filered RSS signal shown in Fig. 3 will be exposed in a form of averaged RSSI readings, as illusraed in Fig. 4. In Fig. 4, 1600 symbol periods are aken o ge 200 averaged RSSI readings, which equals o 25.6 ms. Phase offse of averaged RSSI readings Δϕ RSSI is observed by comparing he wo averaged RSSI reading curves beween receiver C and receiver D. In secion III, furher simulaion reveals he approximae relaionship beween Δϕ RSSI and Δϕ: Δϕ RSSI =Δϕ 12 Since Δϕ RSSI can be measured, we obain he value of Δϕ. Subsiue he value of Δϕ ino Equaion 11 o ge he disance relaionship. Fig. 4. Phase Offse Refleced by Averaged RSSI Readings a Two Receivers 7 Sochasic Radio Inerferomeric Posiioning Algorihm: One ieraion of he above process is called a phase offse measuremen round. Such measuremen needs o be repeaed under differen carrier frequencies and differen sender/receiver pair combinaions o ge a se of phase offses ha will be inpu ino he posiioning algorihm described in 11. Please refer o 11 for more deails of he algorihm. C. Preliminary Evaluaion of SRIPS_UMS 1 Swiching Time: SRIPS_UMS saves swiching ime. According o he posiioning process described in 7 and 11, sensors use unmodulaed signals for phase offse measuremens and hey also need o swich o modulaion mode for communicaion purpose, such as receiving scheduling commands for localizaion. Denoe he swiching ime as s and he number of swiching is N, hen he oal saved localizaion ime is N s, where s is defined by radio chip daa shee Noise Analysis: In 12, i is poined ou ha RIPS using modulaed signals heoreically performs beer han oher RIPS mehods wih unmodulaed signals in noisy enviromens. We expec SRIPS_UMS o inheri his advanage and experimens have been conducing o conform his. Resuls will be presened in he nex paper. Obviously, noise from oher han he carrier frequencies is rejeced by he ransceiver. So, we re only concerned wih noise ha ges hrough he filer. The noise ha ges hrough is no whie; i is pink in he sense ha i is concenraed in specific frequency bands. 3 Hardware Implemenaion: We are implemening SRIPS_UMS wih five Zoleria Z1 sensors equipped wih radio chips CC2420 using he embedded operaing sysem - TinyOS CC2420 has 16 channels wihin he 2.4 GHz band. The receiving sensors obain 300 averaged RSSI readings per phase offse measuremen in 38.4 ms per frequency over a oal of 16 frequencies in a range of 2405 MHz o 2480 MHz. All RSSI readings will be sen o a compuer and logged for furher processing and daa analysis. III. SIMULATION RESULTS One imporan advanage of SRIPS_UMS is ha i doesn need ime synchronizaion on he sender side, which means ha Δ A, Δ B can be arbirary wihou obvious affecs on posiioning accuracy. In his secion, we demonsrae his advanage hrough exensive simulaions and presen heoreical suppor for Equaion 12. A simple scenario is used in our simulaions. We consider a square field m wih 4 sensors deployed in oudoor environmen. The arge sensor is he only one wih unknown posiion in each phase offse measuremen round. A. Time Offses and Period of Averaged-RSS-Time Curve In Fig. 5 a, alhough Δ A, Δ B are changing, he averaged period of averaged-rss-ime curve remains around 2.5 ms. 2.5 ms is he reciprocal of he frequency difference f A f B. Repea his experimen under a differen frequency 61

6 difference 200 Hz and hese resuls hold, as shown in Fig. 5 b. a frequency difference is 400 Hz b frequency difference is 200 Hz Fig. 5. Averaged Period of Averaged-RSS-Time Curve under Differen Time Offses B. Time Offses and Phase Offse of Averaged-RSS-Time Curves In Fig. 6 a, he phase offse of filered RSS signals Δϕ is 0. and he averaged phase offse of averaged-rss-ime curves Δϕ RSSI remains around 0. under differen ime offses Δ A, Δ B. Repea his experimen under differen phase offses of filered RSS signals 0.4π, 0.6π, 0.8π and Δϕ RSSI sill equals o Δϕ wih differen ime offses, as shown in Fig. 6 b c d. So we can esimae Δϕ from Δϕ RSSI ha can be obained from averaged-rss-ime curves. Time offses Δ A, Δ B almos have no effec on obaining correc Δϕ ha deermines he posiioning accuracy. a phase offse of RSS is 0. b phase offse of RSS is 0.4π c phase offse of RSS is 0.6π d phase offse of RSS is 0.8π Fig. 6. Averaged Phase Offse of Averaged-RSS-Time Curves wih Differen Time Offses IV. CONCLUSIONS SRIPS_UMS heoreically resolves he hardware implemenaion problem and he ime synchronizaion problem of previous RIPS mehods. The key enabling ideas behind his resul are: The mahemaical derivaion of he filered RSS signals. The discovery of he relaionship beween he phase offse of filered RSS signals and he phase offse of averaged- RSS-ime curves. And his relaionship is no effeced by ime offses. Fuure work includes hardware implemenaion of SRIPS_UMS on CC2420, research on SRIPS_UMS uilizing new observaions for possible indoor applicaions. ACKNOWLEDGMENT The auhors wan o hanks Dr. Almir Davis, Zengxu Yang, and oher Tufs Wireless Lab TWL members. REFERENCES 1 L. Zhang, Q. Cheng, Y. Wang, and S. Zeadally, A novel disribued sensor posiioning sysem using he dual of arge racking, IEEE Transacions on Compuers, vol. 57, no. 2, pp , R. Gao, M. Zhao, Z. Qiu, Y. Yu, and C. H. Chang, Web-based moion deecion sysem for healh care, in Compuer and Informaion Science ICIS, 2015 IEEE/ACIS 14h Inernaional Conference on. IEEE, 2015, pp A. Faheem, R. Virrankoski, and M. Elmusrai, Improving rssi based disance esimaion for wireless sensor neworks, in Wireless Informaion Technology and Sysems ICWITS, 2010 IEEE Inernaional Conference on. IEEE, 2010, pp M. A. A. A. Mazlan, M. H. M. Khir, and M. N. M. Saad, Implemenaion of wireless sensor nework in oil and gas specifically for personnel posiioning applicaion, in Compuer, Communicaions, and Conrol Technology I4CT, 2015 Inernaional Conference on. IEEE, 2015, pp R. Dalce, A. Van den Bossche, and T. Val, Indoor self-localizaion in a wsn, based on ime of fligh: Proposiions and demonsraor, in Indoor Posiioning and Indoor Navigaion IPIN, 2013 Inernaional Conference on. IEEE, 2013, pp P. Rong and M. L. Sichiiu, Angle of arrival localizaion for wireless sensor neworks, in rd annual IEEE communicaions sociey on sensor and ad hoc communicaions and neworks, vol. 1. IEEE, 2006, pp M. Marói, P. Völgyesi, S. Dóra, B. Kusỳ, A. Nádas, Á. Lédeczi, G. Balogh, and K. Molnár, Radio inerferomeric geolocaion, in Proceedings of he 3rd inernaional conference on Embedded neworked sensor sysems. ACM, 2005, pp B. Kusy, A. Ledeczi, M. Maroi, and L. Meerens, Node densiy independen localizaion, in Proceedings of he 5h inernaional conference on Informaion processing in sensor neworks. ACM, 2006, pp H.-j. Wu, H.-l. Chang, C.-w. You, H.-h. Chu, and P. Huang, Modeling and opimizing posiional accuracy based on hyperbolic geomery for he adapive radio inerferomeric posiioning sysem, in Inernaional Symposium on Locaion-and Conex-Awareness. Springer, 2007, pp A. Ledeczi, P. Volgyesi, J. Sallai, B. Kusy, X. Kousoukos, and M. Maroi, Towards precise indoor RF localizaion, in proc. of he 5h Workshop on Embedded Neworked Sensors, Charloesville: ACM, B. Dil and P. J. Havinga, Sochasic radio inerferomeric posiioning in he 2.4 GHz range. in SenSys, 2011, pp R. Gao, Z. Yang, W.-T. Sun, C. H. Chang, and A. Couch, Radio inerferomeric posiioning wih modulaed signals in wireless sensor neworks, in 111h Inernaional Conference on Communicaion and Signal Processing, Japan, A. Chipcon, Cc1000 daa shee, T. Insrumens, CC2420: 2.4 ghz ieee /zigbee-ready rf ransceiver, Available a Available a hp://www. i. com/li/gpn/cc2420, vol. 53, , CC2520 daashee: 2.4 ghz ieee /zigbee rf ransceiver, SWRS068, , CC2430 daa shee rev 2.1 sw rs036f,