An Efficient Alarm Notification Algorithm for Earthquake Early Warning System

Transcription

1 An Effcent Alarm Notfcaton Algorthm for Earthquake Early Warnng System An Effcent Alarm Notfcaton Algorthm for Earthquake Early Warnng System 387

2 388

3 An Effcent Notfcaton Servce Algorthm for Earthquake Early Warnng System Tng-Yun Ch Graduate Insttute of Electrcal Engneerng Natonal Tawan Unversty Tape, Tawan (R.O.C.) Chun-Hao Chen / Han-Cheh Chao Insttute of Computer Scence & Informaton Engneerng Natonal Ilan Unversty, ILan 260 Tawan (R.O.C) Abstract Earthquake Early Warnng Alarm Notfcaton s a servce that addresses devces and users wth messages to be processed at a specfc tme. In ths paper, we propose an effcent notfcaton algorthm for earthquake early warnng system n Tawan. Due to the lack of multcast support n the general IP network, we try to delver notfcaton messages to multple recevers n tme base on locaton nformaton network throughput wth peerng ISPs and prorty wth IoT devces. Wth the proposed algorthm, we can not only reduce the burst message traffc for network but also send the messages n tme. By the SIP protocol, we can ntegrate varous devces nto the same communcaton platform. Keywords Notfcaton; SIP; IoT; Early Earthquake Alert. I. INTRODUCTION The effects of earthquakes nclude: shakng and ground rupture, landsldes, avalanches, fres and Tsunam. In early 20, a major earthquake ht Japan, whch calculated to be at the Mcron Log 9. Accordng to offcal estmaton, there are 6,000 people klled and 0,000 mssed by the tsunam n the nearby area. In 999, the Ch-Ch Earthquake ht Tawan and made the serous damage. By the recent developng of communcaton technology, Earthquake can be estmated the predctve Mcron Log for naton-wde locatons by montor the real- message delvery technology, the earthquake early warnng system s able to announce the alert message before the earthquake arrves. There are three major components for earthquake early warnng system () Real tme earthquake nformaton collecton from the sensor nodes (2) Earthquake estmaton wth predctve result (3) Alarm messages delvery to users. To collect the real tme earthquake nformaton from the sensor nodes, [] [2] use dgtal sesmometers or customzed sensors to collect the data from sensor nodes. In [3],the author proposes an nnovated deas tres to use the three-axs acceleraton nformaton n smart phones or hard dsk as a collaboratve sensor system. The other researches [4][5][6] focus on connecton organzaton mechansms by usng wreless mesh network wreless sensor network P2P technology to collect the data rapdly. There are some studes n [7] [8] [9] [0] try to estmate earthquake to get the predctve result n seconds. In Japan, several studes about earthquake early alert system are carred out. UrEDAS(Urgent Earthquake Detecton and Alarm System) servce [] was launched n 992. The Japan Meteorologcal Agency (JMA) started provdng the Earthquake Early Warnng by several delvery channels such as TV and rado n Another researches n [2][3] tres to provde the alert notfcatons to the moble handsets home automated systems and vehcles. In frst secton, we ntroduce and analyze the status of current earthquake early warnng system. In next secton, we wll ntroduce the status of current earthquake early warnng system n Tawan and analyze the varous alarm notfcaton delvery mechansm. In thrd secton, we modfy exstng system archtecture and propose an nnovatve alarm notfcaton algorthm to reduce system traffc. Our algorthm s verfed by both wth MATLAB smulaton and mplementaton on androd platform n secton 4. Fnally, the concluson and our contrbuton are gven n secton 5. II. RELATED WORK A. Multmeda Broadcast Multcast Servce (MBMS) Messagng Compared to nformaton collecton from hundreds of nodes, delvery the alert to mllons of clents s much dffcult. Alarm emergency message delvery s the most mportant ssue for earthquake early warnng system. Although the japan moble handset alert system proves tself wth successful results n 20, the same archtecture cannot work n Tawan. The telecom servce provders n Tawan do not enable Multmeda Broadcast and Multcast Servces (MBMS) [4] n ther network. MBMS s a broadcastng servce offered va exstng GSM and UMTS cellular networks. The man applcaton s moble TV. MBMS uses multcast dstrbuton n the core network nstead of dedcated uncasts for each end devce. Wth the broadcast capablty, unlmted number of users can be reached wth constant network load. It also enables the possblty to broadcast nformaton smultaneously to many cellular subscrbers, whch s sutable for earthquake early warnng system. 389

4 B. SMS Messagng Usng SMS to delver warnng messages s a bad dea. Even the SMS messagng may beat out other technologes n terms of popularty, t suffers two major dsadvantages. Frst, the cost s relatvely hgh f we would lke to send the message to large number of users. Secondly, although SMS message delvery s usually rapd, the recept tme and relablty can't be guaranteed, whch s the fatal ssue of SMS system. In [5], approxmately 5.% messages were not delvered at all. It s a large amount compared to the end to end message loss for e- mal, whch was only.6%. In general case, each SMS server only can handle two mllon messages per hour (around 500 messages per second). However, t s desgned for general IP network wthout multcast support wth prvate IP ssue. The burst message traffc wll nfluence the low bandwdth network and may be blocked by the frewall. Our prevous work [2][22] already shows how to delver the uncast SIP alert message effcently by locaton nformaton. C. Early earthquake warnng system n Tawan In Tawan, the earthquake wll take 30sec to propagate from Tachung to Tape. CWB (Central Weather Bureau) could collect the data from the sensors and fnsh the estmaton n 0sec. When the earthquake n Tachung, Tape wll receve the warnng message before the earthquake arrves 20 seconds ahead. Fg. shows the system archtecture. Currently, CWB uses propretary Clent-Server TCP communcaton protocol to delver the alert message. TABLE I. Fgure 2. IM Message Flow for RFC3428 (Page Mode) COMPARISON FOR THE DIFFERENT DELIVERED MECHANISMS SMS Propreta ry TCP(cle ntserver) SIP MBMS Relablty Low(wthout Ack) Hgh(wt h Ack) Hgh(wt h Ack) Low(wthou t Ack) Fgure. Earthquake Early Warnng System n Tawan. D. SIP / IMS Messagng The CWB are lookng for a relable and effcent soluton to delver the emergency message to general publc. Due to the prce and performance ssue for SMS system, there are more and more works [6][7][8][9] use SIP or IMS based message delvery mechansm n recent years. Accordng to comparson from Table, SIP wll be the most approprate soluton for NGN (All IP) emergency message delvery system. Compare wth SMS, SIP/IMS clents provde much more multmeda capacty [20]. They usually use Sesson Intaton Protocol for Instant Messagng and Presence Leveragng Extenson (SIMPLE) as the message format. SIP has a number of benefts over SMS, such as explct rendezvous, tghter ntegraton wth other meda-types, drect clent to clent operaton. By RFC3428 page-mode message s enabled n SIP as a much smple way va the SIP MESSAGE method. The message flow shows n Fg. 2. Message delver effcency Advantage Dsadvantag e III. Low Mddle Mddle It can be used wthout IP network Low message delver effcency and expensve The commun caton protocol can be optmze d Hgh mantena nce cost The most popular protocol n the IP network Works wth uncast n the real world Hgh(multc ast) Hgh message delver effcency Only work n partcular cellular network AN EFFICIENT NOTIFICATION SERVICE ALGORITHM FOR EARTHQUAKE EARLY WARNING To ntegrate wth varous smart devces n the same communcaton platform, we adopt SIP page-mode as the next generaton earthquake warnng alert protocol. However, separate uncast messages are stll requred to be sent to users because multcast cannot work n general IP network. It leads to three problems: Frst, t takes tme to send these separate messages sequentally. Second, these messages may block the communcaton channel to peerng ISPs. Thrd, the IoT devces Thanks to MOTC-DPT for fundng the research and NCHC for negotate wth CWB. 390

5 should have the hgh prorty because they usually work wth brake system for elevator and tran. To deal wth these ssues, we desgn an effcent alarm notfcaton algorthm consderng wth both locaton nformaton, prorty for dfferent users and the capacty of communcaton channel to peer ISPs. A. System Archtecture The next generaton earthquake warnng alert system archtecture s llustrated n Fg. 3. When the earthquake occurs, the early earthquake estmaton server wll send the estmated alert to message relay server n Natonal Center for Hgh- Performance Computng (NCHC) after t cal (the tme for collect the data from the sensor node and estmate the result). The message relay server wll then forward the message to message delvery server through JSON nterface. The message delvery server wll delver messages accordng to the proposed alarm notfcaton algorthm consderng wth locaton nformaton, prorty for dfferent users and the capacty of communcaton channel to peer ISPs. Fgure 3. Proposed system archtecture B. User Regstraton Flow When a new user tres to regster wth the servce, the system wll keep the followng nformaton by the steps depcted n Fg. 4: () Network subnet prefx (2) Human or IoT Clent. (3) Locaton nformaton. When a user regsters wth the server, the system wll check user's geographcal locaton. If the user does not set the locaton nformaton n hs regster nformaton, the system wll dentfy hs possble locaton by IP address lookup. Then system wll mark whch peerng ISP the user comes from. Fnally, the system wll check f the user s a real human holdng smart devce or an IoT devce such as elevator brakng system. It wll set the hgher prorty to IoT devce. Fgure 4. User Regstraton Flow C. An Effcent Message Delvery Algorthm The Fg.5 shows the pseudo code for proposed message delvery algorthm. Due to the earthquake estmaton server need a perod of tme (t cal ) to collect the real tme earthquake nformaton from the sensors and estmates the predctve Mcron Log for naton-wde locatons. The clents n the ctes cannot receve the earthquake early warnng n tme that earthquake ht n less than t cal. After t cal, the Earthquake early warnng system starts to delver the messages to clents n each cty. The basc purpose for our delvery algorthm tres to delver messages to clents as many as possble and try to save lfe. If we can send all of the clents n tme, the message delvery server works as normal one. When we cannot send the message due to the capacty for process or network connecton, we set IoT devces have the hgh prorty because they usually much mportant than the human users. The warnng messages wll be sent to the IoT devces frst. The clents are dstrbuted n each ISP and the capacty of communcaton channel to peer ISPs s lmted.we delver the alert messages to each ISP n each cty accordng to the correspondng proporton. When we transmt the burst alert message to clent n each ISP, maybe the capacty of communcaton channels wll be a bottleneck. We actvate the message delvery algorthm when the above-mentoned stuatons happen. The message delvery algorthm can dynamc allocaton message to each ISP based on the bandwdth usage status. Our algorthm makes more effcent use of bandwdth. We set the tme for earthquake estmaton as t cal, the message delvery server can process capablty messages per second. There are ctes n the system and the earthquake wll arrve on t..t. In each cty the IoT and human clents can be marked as ISPIoT + ISPuser. There are j ISPs n each cty and the communcaton channel bandwdth for each ISP s ISPbw j messages per second. If the alert message n the queue can t be sent n tme, t wll be dropped from the delvery queue. The number of messages for each cty that cannot be sent n tme marks n equaton. We wll verfy the 39

6 proposed algorthm by smulaton. The result wll show n the next sesson. LOSSIoT+LOSSUSERS=LOSS Pseudo code of the delvey algorthm and calculate messages loss rato: For( to n) IF( t< tcal) Drop ELSE For( j to n) [ Sum]; IF( ISPbw j > capablty * ISPuser j ISPIoT j / Sum ) IF( capablty * ISPjuser ISPjIoT/ Sum *( t tcal) [( ISPjuser ISPjIoT) - LOSS] Gvng prorty to delver messages to IoT devces than Human users; ISPjuser ISPjIoT Drop [( ISPjuser ISPjIoT) - LOSS] - capablty* ( )* ( t tcal); Sum ELSE // actvate the algorthm Set Capablty* ISPuser j ISPIoT j / Sum ISPbwj; The redundant messages assgn to other ISP; j cal j j j j j cal IF( ISP AdjustMessage *( t t ) [( ISP user ISP IoT ) - LOSS ]) Drop [( ISP user ISP IoT ) - LOSS ]- ISP AdjustMessage *( t t ); Fgure 5. Pseudo code for delvery algorthm IV. SIMULATION A. Smulaton We use the data obtaned Ch-Ch Earthquake n 999. There are 23 major sensor statons n Tawan (N=23) shown n Table. The amount of users n each cty s 0,000. More detal nformaton can be found n Table. The system takes 0 sec to fnsh the smulaton (t cal =0). CWB have one message server that can handle 6,000 messages per second (capacty=6,000). TABLE II. HISTORY DATA FOR 999 CHI-CHI EARTHQUAKE Staton Dstance/ tme Staton Dstance/ tme ID ID Cty0 5.53KM/.38s Cty3 09.6KM/27.4s Cty KM/5.35s Cty4 09.9KM/27.4s Cty KM/6.53s Cty5 23.7KM/30.9s Cty KM/9.58s Cty6 29.KM/32.2s Cty05 4.6KM/0.4s Cty7 32.3KM/33.s Cty KM/3.9s Cty8 37.3KM/34.3s Cty KM/5.6s Cty9 42.5KM/35.6s Cty KM/6.s Cty KM/38.6s Cty KM/6.5s Cty2 56.7KM/39.s Cty KM/9.7s Cty22 7.2KM/42.8s Cty 99.0KM/24.7s Cty KM/53.6s Cty2 03.8KM/25.9s TABLE III. THE NUMBER OF USERS IN EACH CITY Staton ID ISP USER/2/3 Staton ID ISP USER/2/3 Cty0 Cty07 Cty02 Cty03 Cty04 Cty05 Cty06 Cty3 Cty4 Cty5 Cty6 Cty7 Cty8 5000/3000/ /000/ /3000/ /4000/ /3000/ /3000/ /5000/ /2000/ /2000/5000 Cty08 Cty09 Cty0 Cty Cty2 Cty9 Cty20 Cty2 Cty22 Cty /5000/ /3000/ /2000/ /5000/ /2000/ /2000/ /5000/ /6000/ /000/6000 The results of the smulaton show n Fg. 6 and Fg. 7. Fg. 6 show the number of dropped messages wthout consderng the capacty of communcaton channel to peer ISPs. For the frst smulaton, the server delvered the messages wthout usng the algorthm when the earthquake ht. Fg. 6(a) shows the number of loss messages for each ISP. Accordng to smulaton result, the early earthquake estmaton server sends the estmated alert to message relay server on 0 second. It means that we can t send the alert to cty, cty2, cty3, cty4 and even cty5. When the earthquake arrves n the ffth cty, we can see that a consderable number of users and IoT devces cannot receve the earthquake alert message n tme. Fg. 6(b) shows the sum of the number of dropped messages wthout proposed algorthm for IoT devces and Human users. 392

7 Fgure 6. The number of dropped messages wthout proposed algorthm (a)for each ISP. (b)sum for IoT devces and Human users For the second smulaton, we consderng the capacty of communcaton channel to peer ISPs. Fg. 7(a)(b) show the result. By the result, we can fnd that the number of loss message s reduced wth our algorthm that consderng the capacty of communcaton channel to peer ISPs. It shows that our algorthm uses the connecton capacty resource more effcent. Fgure 9. (a) UA settng nterface. (b) UA nterface Fgure 7. The number of dropped messages wth consderng the capacty of communcaton channel to peer ISPs (a)for each ISP. (b)sum for IoT devces and Human users For the fnal smulaton, we try to set the hgher prorty for IoT devces. It make sense that bake system for hgh speed ral carry 989 people s much mportant than a handset. Fg. 8(a) shows the result when the server sets hgher prorty to the IoT devces, we can see the sgnfcant mprovement on the uumber of dropped messages to IoT devces. Fg. 8(b) shows the mprovement n number of dropped messages by campare proposed algorthm(locaton, connecton capacty, prporty to IoT) wth baselne algorthm (only consderng wth locaotn). Fgure 8. The number of dropped messages (a )IoT devces and human users(consderng the hgher prorty to IoT) (b)comparng wth the baselne result(only consderng wth locaotn) B. Implementaton We also mplement a prototype clent on Androd platform. Fg. 9(a)(b) show the user nterface for earthquake early warnng system. V. CONCLUSIONS In ths paper, we propose an nnovatve alarm notfcaton algorthm wth Instant Message (IM) base on Sesson Intaton Protocol (SIP) replacng current propretary Clent- Server protocol. By mprovng the alarm notfcaton algorthm consderng wth locaton nformaton, network throughputs wth peerng ISPs and hgher prorty for IoT devces, we can both reduce the burst message traffc for network and send the message n tme wth fewer servers. It also can be mplemented n real world wth low cost. ACKNOWLEDGMENT We thank the project (MOTC-DPT-00-03) provdes the fundng for the research and Natonal Center for Hgh- Performance Computng (NCHC) makes the great contrbuton to negotate wth CWB for the artcle. REFERENCES [] Jren-Cht Chn; Rautenberg, J.M.; Ma, C.Y.T.; Pujol, S.; Yau, D.K.Y. An Expermental Low-Cost, Low-Data-Rate Rapd Structural Assessment Network. Sensors Journal, IEEE, 2009, vol.9, no., pp [2] Basha, E.; Rus, D. Desgn of early warnng flood detecton systems for developng countres. ICTD Dec. 2007, pp.-0 [3] Hendl, E. Peer-to-peer (P2P) earthquake warnng system based on collaboratve sensng. DEST ' June 2009, pp [4] Nachtgall, J.; Zubow, A.; Sombrutzk, R.; Pcozz, M. The Challenges of Usng Wreless Mesh Networks for Earthquake Early Warnng Systems. MESH 2009,2009, pp [5] Wang Wenjun; Yang Shun. Study of the TDM of the GPS Tmng n Earthquake Early-Warnng Informaton Transfer. ICICTA 200, vol.3, -2 May 200, pp [6] Rajarapollu, P.; Sharma, V.K. Desgn and analyss of satellte based dsaster warnng system II. TENCON Oct. 2007, pp.-4. [7] Gupta, D.; Shahan, D.T.; Khan, S. Fast Fourer Transform based Earthquake Precursor Analyss of Radon n N-W Hmalayas. ICEMI '07. 6 Aug. 2007, pp [8] Amnzadeh, F.; Katz, S.; Ak, K. Adaptve neural nets for generaton of artfcal earthquake precursors. IEEE Transactons on Geoscence and Remote Sensng, Nov 994, vol.32, no.6, pp [9] Ch-Jan Huang; Che-hao Chang; Kuan-Yung Chang. Uncertanty propagaton of Earthquake Loss Estmaton System on the early sesmc 393

8 damage evaluaton. Internatonal Conference on Geonformatcs 7th, 2-4 Aug. 2009, pp.-6. [0] Plag, H.-P.; Blewtt, G.; Hammond, W.; Kreemer, C.; Bar-Sever, Y. Rapd GPS-based determnaton of earthquake dsplacement feld and magntude for tsunam propagaton modelng and warnng, IGARSS 200, July 200, pp [] Gupta, D.; Shahan, D.T.; Khan, S. Fast Fourer Transform based Earthquake Precursor Analyss of Radon n N-W Hmalayas. ICEMI '07. 6 Aug. 2007, pp [2] Nagaosa, T.; Morya, S. An emergency Earthquake warnng system for land moble vehcles usng the earthquake early warnng. ICVES Sept. 2008, pp [3] Koke, N. Catchng the wave: real-tme tsunam warnng systems. IEEE Potentals, October 2009 vol.28, no.5, pp.4-7. [4] Gomez-Barquero, D.; Fernandez-Agulella, A.; Cardona, N. Multcast Delvery of Fle Download Servces n Evolved 3G Moble Networks Wth HSDPA and MBMS. IEEE Transactons on Broadcastng, vol.55, no.4, Dec 2009.pp [5] DEWITT LATIMER. Text Messagng as Emergency Communcaton Superstar? Nt so gr8. EDUCAUSE Revew, June 2008, vol. 43, no. 3, p [6] Sun Ok Park; M-Young Huh; Jae Cheon Han; Shn Gak Kang. Enhanced 9 Mechansm for Internet Telephony Servce. Portland Internatonal Center for Management of Engneerng and Technology, 5-9 Aug. 2007, pp [7] Costa-Requena, J.; Hatao Tang. Enhancng SIP wth spatal locaton for emergency call servces. Tenth Internatonal Conference on Computer Communcatons and Networks Proceedngs.200, pp [8] Maes, S.H. SDP-based IP and Multmeda real tme communcatons ntegraton wth vehcle remote montorng, montorng and emergency systems. ITST Oct. 2008, pp [9] Waraporn, N.; Tryason, T.; Angsuchotmetee, C.; Tlkanont, P. Emergency servce warnng system usng SIP for ntegrated meda. NCM 200, 6-8 Aug. 200, pp [20] Jae Cheon Han; Sun Ok Park; Shn Gak Kang; Hyoung Ho Lee;, "A Study on SIP-based Instant Message and Presence," The 9th Internatonal Conference on Advanced Communcaton Technology, vol.2, 2-4 Feb ,(2007) [2] Tng-Yun Ch, Han-Cheh Chao, Sy-Yen Kuo: An effcent emergency message delvery algorthm for early earthquake warnng system. ICET- 20, 20,Phuket [22] Tng-Yun Ch, Chun-Hao Che, Han-Cheh Chao: An effcent early earthquake alert message delvery algorthm wth mult-isp channels. ITAOI-20,20, Tawan 394