Positioning in Wireless Sensor Network for Human ... - Science Direct

Available online at www.sciencedirect.com

ScienceDirect ScienceDirect

Available online at www.sciencedirect.com Transportation Research Procedia 00 (2016) 000–000 Available online at www.sciencedirect.com Transportation Research Procedia 00 (2016) 000–000 Available online at www.sciencedirect.com

ScienceDirect ScienceDirect Transportation Research Procedia 00 (2016) 000–000 ScienceDirect Transportation Research Procedia 00 (2016) 000–000 Transportation Research Procedia 21 (2017) 56–64

www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia

2016 International Symposium of Transport Simulation (ISTS’16 Conference), June 23~25, 2016 2016 International Symposium of Transport Simulation (ISTS’16 Conference), June 23~25, 2016

Positioning Wireless Network(ISTS’16 for Human Sensing Problem 2016 InternationalinSymposium of Sensor Transport Simulation Conference), June 23~25, 2016 2016 International Symposium of Transport Simulation (ISTS’16 Conference), June 23~25, 2016 Positioning in Wireless Sensor Network for Human Sensing Problem Stanislav Lykova*, Yasuo Asakuraa, Shinya Hanaokab a a b Lykov *, Yasuo Asakura ,Environmental Shinya Hanaoka aWireless inStanislav Sensor Network Human Sensing Tokyo Institute of Technology, Department of Civil andfor Engineering, 2-12-1

Positioning Problem Ookayama Meguro-ku 152-8550 for Tokyo Japan Positioning inTokyo Wireless Sensor Network Human Sensing Problem Institute of Technology, Engineering, 2-12-1 a Department of Civil and Environmental a b a b Tokyo

Institute of Technology, of Transdisciplinary ScienceHanaoka and Engineering, 2 -12-1 Stanislav Lykov *,Department Yasuo Asakura , Shinya Ookayama Meguro-ku 152-8550 Tokyo Japan Stanislav Lykov *, Yasuo Asakura , Shinya Hanaoka

Ookayama Meguro-ku 152-8550a Tokyo Japan aDepartment b 2 -12-1 b Tokyo Institute of Technology, of Transdisciplinary Science and Engineering, a Tokyo Institute of Technology, Department of Civil and Environmental Engineering, 2-12-1

Ookayama Meguro-ku 152-8550 Tokyo Japan Ookayama Meguro-kuof152-8550 Japan Engineering, 2-12-1 Institute of Technology, Department Civil andTokyo Environmental b Tokyo Institute of Technology, Department of Transdisciplinary Science and Engineering, 2 -12-1 Ookayama Meguro-ku 152-8550 Tokyo Japan Ookayama Meguro-ku 152-8550 TokyoScience Japan and Engineering, 2 -12-1 b Tokyo Institute of Technology, Department of Transdisciplinary Ookayama Meguro-ku 152-8550 Tokyo Japan a Tokyo

Abstract

Abstract In this paper the possibility of utilization of wireless sensor network in application to human sensing problem is examined. With the emphasis on crowd dynamics monitoring, the most critical issuesinregarding thistoapplication are analyzed. on limitations In this paper the possibility of utilization of wireless sensor network application human sensing problem Based is examined. With Abstract shortcomings of contemporary crowd monitoring approaches, well as this the specificity problem,Based modified positioning theand emphasis on crowd dynamics monitoring, the most critical issues as regarding applicationof arethe analyzed. on limitations Abstract algorithm, combined with noisy data preprocessing technique is as proposed. Moreover, physical of interaction between and contemporary crowd monitoring approaches, well as the specificity of sensing the principles problem, modified positioning Inshortcomings this paperand theof possibility of utilization of wireless sensor network in application totohuman problem is of examined. With pedestrians sensor network are analyzed. An experiment had been conducted demonstrate the concept data collection algorithm, combined with noisy data preprocessing technique is proposed. Moreover, physical are principles of Based interaction between the emphasis on crowd dynamics monitoring, the most critical issues regarding this application analyzed. on limitations In this paper the possibility of utilization of wireless sensor network in application to human sensing problem is examined. With system aimed obtain relevantare information regarding pedestrian movements. pedestrians and to sensor network analyzed. An experiment had been conducted to demonstrate the concept of data collection and shortcomings of contemporary crowd monitoring as regarding well as thethis specificity of the the emphasis on crowd dynamics monitoring, the mostapproaches, critical issues application are problem, analyzed. modified Based on positioning limitations system aimed to obtainwith relevant regardingtechnique pedestrian movements. algorithm, combined noisyinformation data preprocessing is proposed. physical principles interaction between and of contemporary monitoring as well asMoreover, the specificity of the problem,ofmodified positioning © shortcomings 2016 The Authors. Published bycrowd Elsevier B. V. approaches, pedestrians and sensor network are analyzed. An experiment had been conducted to demonstrate the concept of data collection algorithm, combined with noisy data preprocessing technique is proposed. Moreover, physical principles of interaction between Selection Peer-review under of Transportation Engineering, University of Seoul. © 2016 aimed Theand Authors. Published by responsibility Elsevier B. V.of Dept. system to sensor obtain relevant regarding pedestrian pedestrians and network information are analyzed. An experiment had movements. been conducted to demonstrate the concept of data collection Selection and Peer-review under responsibility of Dept. of Transportation Engineering, University of Seoul. system aimed to obtain relevant information regarding pedestrian movements. sensor network; disasters; ©Keywords: 2016 Thewireless Authors. Published bycrowd Elsevier B. V.human sensing; smart dust Copyright © 2017 The Authors. Published by Elsevier Keywords: wireless sensor network; disasters; sensing; smart dust Selection andAuthors. Peer-review undercrowd responsibility ofB.V. Dept. of Transportation Engineering, University of Seoul. © 2016 The Published by Elsevier B. human V. Selection and Peer-review under responsibility of Dept. of Transportation Engineering, University of Seoul. Selection and Peer-review under responsibility of Dept. of Transportation Engineering, University of Seoul. Keywords: wireless sensor network; crowd disasters; human sensing; smart dust

1. Introduction Keywords: wireless sensor network; crowd disasters; human sensing; smart dust 1. Introduction One of the major trends of 21st century is intensive urbanization process. This process is leading to increasing urban population overtrends the World. a result are urbanization becoming more congested, therefore the probability of dangerous One of theall major of 21stAs century is cities intensive process. This process is leading to increasing urban 1. Introduction situations is increasing. Especially, public venues such as train stations, stadiums, airports etc. are highly vulnerable to population all over the World. As a result cities are becoming more congested, therefore the probability of dangerous 1. Introduction so-called crowd disasters, such as stampede during mass events or emergency situations during rush hours. As an situations is increasing. Especially, public venues such as train stations, stadiums, airports etc. are highly vulnerable to One of the major trends of 21st century is intensive urbanization process. This process is leading to increasing urban and example, one can consider deadly stampede during the Hajj in Saudi Arabia 2015, where over to 700 people died so-called disasters, such ascentury during mass events or emergency situations during rush hours. As an population allmajor over the World. As astampede result cities are becoming more congested, probability of dangerous One ofcrowd the trends ofa21st is intensive urbanization process. Thisintherefore process isthe leading increasing urban nearly 900 were injured (Holly 2015). Notable that this tragedy occurred despite the fact that modern surveillance and example, one can consider a deadly stampede during the Hajj in Saudi Arabia in 2015, where over 700 people died situations increasing. Especially, such as train more stations, stadiums, airportsthe etc.probability are highly vulnerable to populationisall over the World. As apublic result venues cities are becoming congested, therefore of dangerous systems being used and2015). a lot ofNotable preparation work had been done in order provide safety. AmongAsmany nearly 900crowd were injured (Holly that this occurred despite thetofact that surveillance so-called disasters, such as stampede during mass events or emergency situations during rush hours. an situations iswere increasing. Especially, public venues such as tragedy train stations, stadiums, airports etc. aremodern highly vulnerable to other reasons, the key complexity of Hajj the problem. Crowd behavior is difficult to people predict, describe systems were being usedchallenge and a as lotis ofthe preparation had in in order to where provide safety. Among many example, one can consider asuch deadly stampede during the inbeen Saudi Arabia 2015, over 700 died so-called crowd disasters, duringwork mass events ordone emergency situations during rush hours. Asand an and reasons, analyze, since (especially during emergency dodespite not follow strict movement rules. other the keypeople challenge is stampede the complexity of the problem. Crowd behavior isfact difficult to predict, describe nearly 900 were injured (Holly 2015). Notable that this tragedy occurred thewhere that modern surveillance and example, one can consider a deadly during the Hajj insituations) Saudi Arabia in 2015, over 700 people died More precisely, the milestone in the process of analyzing the crowd dynamics is complicated process of human sensing. and analyze, since people (especially during emergency situations) do not follow strict movement rules. More systems900 were being used(Holly and a2015). lot of Notable preparation had been done indespite order the to provide many nearly were injured thatwork this tragedy occurred fact thatsafety. modernAmong surveillance Under the we assume the of process of been extracting information regarding people’s movement precisely, theterm milestone in sensing” the process analyzing thethe crowd dynamics complicated process of human sensing. other reasons, the“human keyused challenge is the complexity problem. Crowd behavior difficult to predict, describe systems were being and a lot of of preparation work had done isin order to is provide safety. Among many in some environment. Indeed, understanding of people’s movement and behavioral pattern play crucial role in Under the term “human sensing” we assume the process of extracting information regarding people’s movement and analyze, people (especially during emergency situations) do not followisstrict movement rules. More other reasons,since the key challenge is the complexity of the problem. Crowd behavior difficult to predict, describe many applications. in some environment. Indeed, understanding of people’s movement and behavioral pattern play crucial role in precisely, the since milestone in the process of analyzing the crowd dynamics complicated process of human sensing. and analyze, people (especially during emergency situations) doisnot follow strict movement rules. More many applications. Under the the term “human in sensing” we assume the process of extracting regarding movement precisely, milestone the process of analyzing the crowd dynamicsinformation is complicated processpeople’s of human sensing. in sometheenvironment. Indeed, understanding of process people’sofmovement and behavioral pattern people’s play crucial role in Under term “human sensing” we assume the extracting information regarding movement * Corresponding author. Tel.: 03-5734-2575. many applications. in some environment. Indeed, understanding of people’s movement and behavioral pattern play crucial role in E-mail address: [email protected] * Corresponding author. Tel.: 03-5734-2575. many applications. E-mail address: [email protected] Copyright © 2017 The Authors. Published by Elsevier B.V. 2214-241X 2016 The authors. Published byofElsevier V. Selection and © Peer-review under responsibility Dept. ofB.Transportation Engineering, University of Seoul. * Corresponding Tel.: 03-5734-2575. Selection andauthor. Peer-review under responsibility of Dept. of Transportation Engineering, University of Seoul. 10.1016/j.trpro.2017.03.077 2214-241X © 2016 The authors. Published by Elsevier B. V. E-mail address: [email protected] * Selection Corresponding author. Tel.:under 03-5734-2575. and Peer-review responsibility of Dept. of Transportation Engineering, University of Seoul. E-mail address: [email protected] 2214-241X © 2016 The authors. Published by Elsevier B. V. Selection and under responsibility of Dept. B. of V. Transportation Engineering, University of Seoul. 2214-241X © Peer-review 2016 The authors. Published by Elsevier

2

Stanislav Lykov / Transportation Research Procedia 00 (2016) 000–

2

Stanislav Lykov / Transportation Research Procedia 00 (2016) 000–

For instance, this information is essential for crowd management systems aimed to monitor crowd dynamics, detect Stanislav Lykov / Transportation Research Procedia 00 (2016) 000– For instance, Lykov et al. / Transportation Research Procedia 21to (2017) 56–64 57 dangerous situations and insure public safety et al. 2010). Another application is urban planning. Here the this information is Stanislav essential for(Jacques crowd management systems aimed monitor crowd dynamics, detect analysis of situations movement pattern is used forsafety provisioning for the design of and publicHere venues. dangerous and insure public etguidelines al. 2010).systems Another application is urban planning. the For instance, this information is essential for(Jacques crowdthe management aimed to infrastructure monitor crowd dynamics, detect However, despite the fact that modern surveillance systems are highly several majorplanning. shortcomings are analysis of situations movement pattern is used forsafety provisioning for the advanced, design of infrastructure and publicHere venues. dangerous and insure public (Jacquesthe etguidelines al. 2010). Another application is urban the stillHowever, exist. instance, video systems have limited capabilities andmajor experience difficulties despite the fact that modern surveillance systems areinstallation highly several shortcomings are analysis of For movement pattern issurveillance used for provisioning the guidelines for the advanced, design of infrastructure and public venues. during operating in crowded places. Moresurveillance precisely, it’s very are difficult observe and identify individuals dueare to still exist. Fordespite instance, surveillance systems have limited installation capabilities andmajor experience difficulties However, the video fact that modern systems highlytoadvanced, several shortcomings frequent occlusions. Another example is utilization ofit’s mobile Recent advances in mobile computing during operating in crowded More precisely, very devices. difficult to observe and identify individuals dueand to still exist. For instance, videoplaces. surveillance systems have limited installation capabilities and experience difficulties sensingoperating provide a in good opportunity toMore collect a wide range of new and rich that can toindividuals detect pedestrian frequent occlusions. Another example is utilization ofit’s mobile devices. Recent advances inused mobile computing during crowded places. precisely, very difficult to data observe and be identify dueand to movement. Examples include GPS location information participatory et al.be2012), Bluetooth identifier sensing provide a good opportunity to collect a wide range of new and sensing rich data(Wirz that can to detect pedestrian frequent occlusions. Another example is utilization of inmobile devices. Recent advances inused mobile computing and collectedprovide from sensors integrated smartphones (Weppner et new al. 2013), and data social media data (Haghighi al. 2013). movement. Examples GPSonlocation information in participatory (Wirz et al.be 2012), Bluetooth identifier sensing a goodinclude opportunity to collect a wide range of and sensing rich that can used to detectetpedestrian The mainfrom disadvantage of these approaches is (Weppner the availability of devices. These coupled with the collected sensorsinclude integrated smartphones et al. 2013), and social media (Haghighi et al. 2013). movement. Examples GPSonlocation information in participatory sensing (Wirz etdrawbacks, al.data 2012), Bluetooth identifier complexity of the task itself, mayon lead to failureisof(Weppner these systems. The mainfrom disadvantage of these approaches the availability of devices. These drawbacks, coupled with the collected sensors integrated smartphones et al. 2013), and social media data (Haghighi et al. 2013). Inmain order disadvantage to some of these limitations study propose the system based oncoupled Wireless Sensor complexity of overcome the task itself, may lead to failureisofthis these systems. The of these approaches the availability ofconcept devices.of These drawbacks, with the Network (WSN) a core component human system, with the application the problem of crowd dynamics In order to some of these limitations study propose concept of the to system based on Wireless Sensor complexity of overcome theastask itself, may leadofto failuresensing ofthis these systems. monitoring. term WSN we of assume distributed autonomous capable to monitor physical or Network (WSN) as the a core component humanspatially sensing system, with the application the problem of crowd dynamics In order toUnder overcome some of these limitations this study propose concept ofsensors the to system based on Wireless Sensor environmental conditions, and cooperatively pass this data through the Need to problem mention, that a special case monitoring. Under term WSN we of assume spatially distributed autonomous sensors capable to monitor physical or Network (WSN) as the a core component human sensing system, with the network. application to the of crowd dynamics of sensor network istheconsidered inwe thisassume study,pass inspired by “smart dust” project,sensors proposed by thetoresearchers fromcase the environmental conditions, cooperatively this data through the network. Need capable to mention, that a special monitoring. Under termand WSN spatially distributed autonomous monitor physical or UCsensor Berkeley. According toand (Warneke al. 2001), thisdata is a“smart “…system many proposed tiny microelectromechanical systems of network is considered in thisetstudy, inspired by dust” project, by the researchers from the environmental conditions, cooperatively pass this through the of network. Need to mention, that a special case (MEMS) such as sensors, robots, or this other devices, that can for example, light, magnetism, UCsensor Berkeley. According to (Warneke etstudy, al. 2001), this is detect, a“smart “…system of many tinytemperature, microelectromechanical systems of network is considered in inspired by dust” project, proposed by the vibration, researchers from the or The intent is that devices will eventually become size oftemperature, amicroelectromechanical grain of sand or a dust particle, (MEMS) such According as sensors, robots, or these other that can for example, light, vibration, magnetism, UCchemicals…”. Berkeley. to (Warneke et devices, al. 2001), this is detect, a “…system ofthe many tiny systems hence thesuch nameassmart dust. or chemicals…”. The intent is that devices will become the size oftemperature, a grain of sand or a dust particle, (MEMS) sensors, robots, or these other devices, that eventually can detect, for example, light, vibration, magnetism, Thethe keyname difference from issimilar researches thateventually the concept of "smart dust" considered study. This hence smart dust. or chemicals…”. The intent that these devicesiswill become the size of aisgrain of sandinorthis a dust particle, approach gives more in terms of monitoring pedestrians through network, since denser network Thethe key difference from similar researches is that the concept of "smartthe dust" is considered in this study. setup This hence name smart flexibility dust. allows obtain positioning information. nowadays devices are smaller and cheaper, approach gives more precise flexibility in terms of monitoring pedestrians through the network, since denser network setup The tokey difference from similar researches is thatMoreover, the concept of "smart dust" is becoming considered in this study. This therefore there will possible realize yet theoretical concept in devices near future. Tosince meetdenser this goal, allows to obtain precise positioning information. Moreover, nowadays are becoming smaller andmodified cheaper, approach gives morebe flexibility intoterms of this monitoring pedestrians through the network, network setup positioning algorithm for sensor nodes localization had been developed, ofmodified “human therefore there will possible to realize this yet theoretical concept inand nearseveral future. To meetprinciples this goal, allows to obtain morebe precise positioning information. Moreover, nowadays devices are physical becoming smaller and cheaper, sensing” were examined. Insensor orderto to realize test and investigate applicability proposed solution, anprinciples experiment had been positioning algorithm nodes localization hadthe been developed, physical “human therefore there will befor possible this yet theoretical concept of inand nearseveral future. To meet this goal,ofmodified conducted. sensing” were examined. order nodes to test and investigate applicability ofand proposed solution, experiment been positioning algorithm forInsensor localization hadthe been developed, several physicalanprinciples ofhad “human conducted. sensing” were examined. In order to test and investigate the applicability of proposed solution, an experiment had been 2. Human sensing techniques conducted. 2. Human sensing techniques The needsensing of advanced tools for understanding human behavior from sensory data is indeed emerging recently, since 2. Human techniques such applications as health monitoring, computerhuman graphics, machine vision and data human-machine interface are in asince high The need of advanced tools for understanding behavior from sensory is indeed emerging recently, need forneed this information. According to the proposed solution, WSNfrom should be used foristhe purpose of crowd monitoring. The of advanced tools for understanding human behavior sensory data indeed emerging recently, such applications as health monitoring, computer graphics, machine vision and human-machine interface are in asince high However, sensors (ref. asAccording MOTEs) have toproposed interact with pedestrians in order tohuman-machine collect information. Therefore, in such applications as health monitoring, computer graphics, machine vision and interface are in a high need for this information. to the solution, WSN should be used for thethe purpose of crowd monitoring. this chapter the description and analysis human-sensing techniques provided. other words, the description of However, sensors (ref. asAccording MOTEs) have toproposed interact with pedestrians in isorder to collect the information. Therefore, in need for this information. to theof solution, WSN should be used forInthe purpose of crowd monitoring. different physical principles of analysis interaction betweenwith MOTEs and pedestrians. According to researchers from MIT However, sensors (ref. as MOTEs) have of to interact pedestrians in isorder to collect the information. Therefore, in this chapter the description and human-sensing techniques provided. In other words, the description of (Teixeira et al. 2011), different types of interactions can be caught by different types of sensors. On the picture this chapter the description of human-sensing techniques is provided. In other to words, description of different physical principlesand of analysis interaction between MOTEs and pedestrians. According researchers fromabove, MIT taxonomyphysical measurable human traits is provided. Static traits stem from the physiological properties, andpicture arefrom produced different principles oftypes interaction between MOTEs and by pedestrians. According to researchers MIT (Teixeira etofal. 2011), different of interactions can be caught different types of sensors. On the above, whenever a person is present, irrespective of what he or she is doing. Common static traits are weight and shape. On (Teixeira etofal. 2011), different interactions can traits be caught different types of sensors. On the above, taxonomy measurable humantypes traits of is provided. Static stem by from the physiological properties, andpicture are produced the other hand, dynamic traits are those that arise from human activity. They are only present when people move, and taxonomy aofperson measurable humanirrespective traits is provided. fromCommon the physiological properties, andand areshape. produced whenever is present, of whatStatic he ortraits she isstem doing. static traits are weight On are other not detectable for reasonably stationary restisof this chapter is devoted to the description wireless the hand, dynamic traits are those thatofpersons. arise activity. They are only present people move, and whenever a person is present, irrespective whatfrom heThe orhuman she doing. Common static traits arewhen weight andof shape. On sensor methods of traits human the other hand, dynamic aresensing. those that persons. arise from human activity. They are only present people of move, and are notbased detectable for reasonably stationary The rest of this chapter is devoted to thewhen description wireless are notbased detectable for reasonably stationary persons. The rest of this chapter is devoted to the description of wireless sensor methods of human sensing. sensor based methods of human sensing. 2

Fig. 1. Texonomy of measurable human traits (Teixeira al. 2001) Fig. 1. Texonomy of measurable human traits (Teixeira al. 2001)

Author name / Transportation Research Procedia 00 (2016) 000–000

3

2.1. Wireless sensor based methods Author name / Transportation Research Procedia 00 (2016) 000–000 3 Author name / Transportation Research Procedia 00 (2016) 000–000 3 2.1. Wireless sensor based methods 58 Stanislav Lykov et al. / Transportation Research00 Procedia 21 (2017) 56–64 Author name / Transportation Research Procedia (2016) 000–000 3 2.1. Wireless sensor based methods Several researches had been done in order to investigate the applicability of wireless sensor network to the problems 2.1. Wireless sensor based methods of human andhad crowd monitoring. For instance, (Hao et al. 2006) presented a wireless pyroelectric Severalsensing researches beendynamics done in order to investigate the applicability of wireless sensor network to the problems Several researches been done inmodules order to(slaves), investigate the applicability of wireless sensor network to the problems sensor system, composed of sensing synchronization error rejection module (master), and a of human sensing andhad crowd dynamics monitoring. Forainstance, (Hao et and al. 2006) presented a wireless pyroelectric Several researches had been done in order to investigate the applicability of wireless sensor network to the problems of human sensing and crowd dynamics monitoring. For instance, (Hao et al. 2006) presented a wireless pyroelectric data fusion module (host),of to sensing performmodules human tracking. precisely, researchers an implementation of a sensor system, composed (slaves), More a synchronization and error presented rejection module (master), and of human sensing and crowd dynamics monitoring. ForMore (Hao researchers et and al. 2006) presented a wireless pyroelectric sensor system, composed of modules (slaves), ainstance, synchronization error rejection module (master), and wireless distributed pyroelectric sensor system for human motion tracking based on presented TI’s microcontroller MSP430149 data fusion module (host), to sensing perform human tracking. precisely, an implementation of aa sensor system, composed of sensing modules (slaves), a synchronization and error rejection module (master), and data fusion module (host), to perform human tracking. More precisely, researchers presented an implementation of and RF transceiver for the reported results, motion WSN appeared be suitable human sensing, howeveraa wireless distributed TRF6901. pyroelectricAssensor system for human tracking to based on TI’s for microcontroller MSP430149 data fusion module (host), to perform human tracking. More precisely, researchers presented an implementation of a wireless distributed pyroelectric sensor system for human motion tracking based on TI’s microcontroller MSP430149 this study was limited to only small interest.results, WSN appeared to be suitable for human sensing, however and RF transceiver TRF6901. As forarea theofreported wireless distributed pyroelectric sensor system for human motion tracking based on TI’s microcontroller MSP430149 and RF transceiver TRF6901. As for the reported results, WSN appeared to be suitable for human sensing, however collaborative research by interest. the researchers from Ohio State University, The University of Texas, thisInstudy was limited to only conducted small area of and RF transceiver TRF6901. As for theof reported results, WSN to beUniversity, human sensing, this was limited to only conducted small area interest. University of Iowa, Michigan State University andresearchers Kent State University etsuitable al. 2004)for the applicability ofofhowever wireless Instudy collaborative research by the fromappeared Ohio(Arora State The University Texas, this study was limited to only small area of interest. In collaborative research conducted by the researchers from Ohio State University, The University of sensor network to the problem of human sensing had been investigated. According to this comprehensive study, aTexas, novel University of Iowa, Michigan State University and Kent State University (Arora et al. 2004) the applicability of wireless In collaborative research conducted by the researchers from Ohio State University, The University of Texas, University of Iowa, Michigan State University and Kent State University (Arora et al. 2004) the applicability of wireless spatial statistic useful for classification is defined, and a distributed estimator for it is developed and tested. Moreover, sensor network to the problem of human sensing had been investigated. According to this comprehensive study, a novel University of Iowa, Michigan State University and Kent University (Arorafor etital. the of wireless sensor network to observations the problem of had been investigated. According tois2004) this comprehensive study, a novel the experimental arehuman based on more than adistributed thousand empirical tracks tested at applicability a and dozen sites including spatial statistic useful for classification issensing defined, and aState estimator developed tested. Moreover, sensor network to the problem of human sensing had been investigated. According to this comprehensive study, a novel spatial statistic useful for classification is defined, and a distributed estimator for it is developed and tested. Moreover, multiple deployments at MacDill Air Force Base in Tampa, Florida. Based on this nearly year-long study, researchers the experimental observations are based on more than a thousand empirical tracks tested at a dozen sites including spatial statistic useful for classification is defined, and a distributed estimator for it is developed and tested. Moreover, the experimental observations are based on more than a thousand empirical tracks tested at a dozen sites including identified several lessons that have practical value to sensor network designers. However, despite the fact that the multiple deployments at MacDill Air Force Base in Tampa, Florida. Based on this nearly year-long study, researchers the experimental observations are based on more than a thousand empirical tracks tested at a dozen sites including multiple deployments at MacDill Air Force Base in Tampa, Florida. Based on this nearly year-long study, researchers concept considered in this study was similar to the one, considered in current paper (network consist from a huge identified several lessons that have practical value to sensor network designers. However, despite the fact that the multipleof deployments MacDill Air Force in Tampa, Florida. Based on thispaper nearly year-long study, researchers identified several that have value toone, sensor network designers. However, despite the from fact that the number nodes), lessons the authors we focused onBase intrusion detection, rather crowd dynamics monitoring. Researchers concept considered inat this study waspractical similar to the considered inthan current (network consist a huge identified several lessons that have practical value to sensor network designers. However, despite the fact that the concept considered in this study was similar to the one, considered in current paper (network consist from a huge form the Beijing University of Science and Technology (Zhang et al. 2012) proposed a human motion tracking approach number of nodes), the authors we focused on intrusion detection, rather than crowd dynamics monitoring. Researchers concept considered in this study was similar to the one, considered in current paper (network consist from a huge number of nodes), the authors we focused on intrusion detection, rather than crowd dynamics monitoring. Researchers for daily life surveillance in a distributed wireless sensor network using ultrasonic range sensors. In this study the form the Beijing University of Science and Technology (Zhang et al. 2012) proposed a human motion tracking approach number of nodes), the authors we focused on intrusion detection, rather than crowd dynamics monitoring. Researchers form the Beijing University of Science and Technology (Zhang et al. 2012) proposed a human motion tracking approach authors claimed that instead of the centralized processing tracking system based on vision, a promising alternative for daily life surveillance in a distributed wireless sensor network using ultrasonic range sensors. In this study the form thenamed Beijing University ofof and Technology (Zhang et al. 2012) proposed a human motion tracking approach for daily life surveillance in aScience distributed wireless sensor network using ultrasonic range sensors. In this study the system distributed wireless network (WSN) has been quickly developed recently. The authors also authors claimed that instead the sensor centralized processing tracking system based on vision, a promising alternative for daily life surveillance in a distributed wireless sensor network using ultrasonic range sensors. In this study the authors claimed that instead of the centralized processing tracking system based on vision, a promising alternative emphasized that despite the fact that there are many applications on WSN on target tracking problems, only a few system named distributed wireless sensor network (WSN) has been quickly developed recently. The authors also authors claimed that instead of the centralized processing tracking system based on vision, a promising alternative system named distributed network has been quickly developed recently. authors papers can bethat found on human motion tracking real(WSN) time systems. Therefore developed such aThe system by WSN. emphasized despite thewireless fact thatsensor there areinmany applications on WSN onthey target tracking problems, only a also few system named distributed sensor network (WSN) has been quickly developed authors emphasized that despite thewireless fact there arebased applications on WSN onthey target tracking problems, only a also few As a result, this paper anthat UKF filter adaptive sensor scheduling scheme for recently. human tracking inby wireless papers can be found onpresented human motion tracking inmany real time systems. Therefore developed such aThe system WSN. emphasized that despite the fact there are applications on WSN onthey target tracking problems, only a few papers can be found onpresented human motion tracking inmany real time systems. Therefore such a system WSN. sensor networks. It uses cheap range sensor nodes in wireless sensor networks bydeveloped jointly selecting the next tasking As a result, this paper anthat UKF filter based adaptive sensor scheduling scheme for human tracking inby wireless papers can be found on human motion tracking in real time systems. Therefore they developed such a system by WSN. As a result, this paper presented an UKF filter nodes based adaptive sensor scheduling scheme human tracking inthe wireless and determining the sampling interval based in onwireless predicted tracking accuracy andfor tracking cost under UKF sensor networks. It uses cheap range sensor sensor networks by jointly selecting the next tasking As a result, this paper presented an UKF filter nodes based adaptive sensor scheduling scheme human tracking inthe wireless sensor networks. It uses cheap range sensor sensor networks by jointly selecting the next tasking filter frame. sensor and determining the sampling interval based in onwireless predicted tracking accuracy andfor tracking cost under UKF sensor networks. It uses cheap rangesensor sensor nodes wireless sensor networks by jointly selecting nextthe tasking sensor and determining the sampling interval based in on predicted tracking accuracy andup tracking costthe under UKF Several concepts of tiny wireless network, such as “smart dust” project, made of microelectromechanical filter frame. sensor and determining thereferred sampling interval based predicted tracking accuracy andup tracking cost under the UKF filter frame. sensors (MEMS), usually tosensor as MOTEs, thaton have sensing,made computation, communication and Several concepts of tiny wireless network, such as self-contained “smart dust” project, of microelectromechanical filter frame. Several concepts ofintiny wireless network, such as self-contained “smartobjects. dust” project, made up of microelectromechanical power were proposed order to gather about different For example, according to the “smart dust” sensors (MEMS), usually referred tosensor asinformation MOTEs, that have sensing, computation, communication and Several concepts ofintiny wireless network, such as self-contained “smart dust” made of microelectromechanical sensors (MEMS), tosensor asinformation MOTEs, that have sensing, communication and project created by ausually team ofreferred UC researchers, within each of theseproject, MOTEs iscomputation, anup integrated package of sensor, power were proposed order to Berkeley gather about different objects. For example, according to the “smart dust” sensors (MEMS), to asinformation MOTEs, that have self-contained sensing, and power were proposed in order to Berkeley gather about different objects. For example, accordingcommunication to the “smart dust” a semiconductor diode and passive optical transmission, an optical receiver, signal and control circuit, project created bylaser ausually team ofreferred UC researchers, within each of these MOTEs iscomputation, anprocessing integrated package of sensor, power were proposed in order to gather information about different objects. For example, according to the “smart dust” project created bylaser abased team of Berkeley researchers, within each of these is anprocessing integrated package ofinclude: sensor, a power source on UC thick-film and solar cells. toMOTEs the authors, possible applications aand semiconductor diode and passivebatteries optical transmission, anAccording optical receiver, signal and control circuit, created by abased team of Berkeley researchers, within each of these is anprocessing integrated package ofinclude: sensor, aproject semiconductor laser diode and passive optical transmission, anvibration, optical receiver, signal and control circuit, monitoring and measurement things like temperature, sound, pressure, and environmental pollutants. For and a power source on UC thick-film batteries and solar cells. According toMOTEs the authors, possible applications amonitoring semiconductor laser diode and passive optical transmission, optical signal processing and pollutants. control circuit, and a power source based thick-film batteries anddust solar cells.anvibration, According to the authors, possible applications include: civilian applications, a on wide variety of smart applications arereceiver, possible, such as agricultural, industrial, and measurement things like temperature, sound, pressure, and environmental For and a power source based thick-film batteries solar cells. According to the authors, possible applications include: monitoring and measurement things like temperature, sound, vibration, pressure, and environmental pollutants. For environmental, or healthcare monitoring, welland asdust traffic control uses. the biggest challenge in industrial, achieving civilian applications, a on wide variety ofas smart applications areHowever, possible, such as agricultural, monitoring and measurement things likeofas temperature, sound, vibration, pressure, and environmental pollutants. For civilian applications, a wide variety smart applications possible, such as agricultural, actual smart dust are the power consumption issues and making thearebatteries small enough. Additionally, several environmental, or healthcare monitoring, well asdust traffic control uses. However, the biggest challenge in industrial, achieving civilian applications, a wide variety of smart dust applications are possible, such as agricultural, industrial, environmental, or already healthcare monitoring, as well as traffic control However, the biggest in achieving experiments in order to demonstrate and testuses. the batteries sensor network system.challenge For example, research actual smart were dust are the conducted power consumption issues and making the small enough. Additionally, several environmental, or already healthcare monitoring, well as traffic control However, the biggest challenge in achieving actual smartUC dust are the(UC power consumption issues and making the enough. Additionally, several group from Berkeley Berkeley andasMLB Co.) conducted anuses. experiment insmall order to test the feasibility ofresearch “smart experiments were conducted in order to demonstrate and test the batteries sensor network system. For example, actual smartUC dust aredeployed the(UC power consumption issues and making the batteries Additionally, several experiments were already conducted in order to demonstrate and test theunmanned sensor network system. For example, dust” network. They a sensor network onto conducted a road from aerial vehicle (UAV), established group from Berkeley Berkeley and MLB Co.) anan experiment insmall orderenough. to test the feasibility ofresearch “smarta experiments were already conducted in order to demonstrate test theunmanned sensor network system. For example, group from UC Berkeley (UC Berkeley and MLB Co.) an experiment in aerial orderand to test the feasibility ofresearch “smarta time-synchronized multi-hop communication network among the nodes on the ground detected, as well as tracked dust” network. They deployed a sensor network onto conducted a roadand from an vehicle (UAV), established group from UC They Berkeley (UC Berkeley and MLB conducted an experiment in aerial order to test the feasibility of “smart dust” network. deployed a sensorAs network onto a road from an unmanned vehicle (UAV), established vehicles passing through the network. for the Co.) reported results, according the researcher they managed totracked detecta time-synchronized multi-hop communication network among the nodes on thetoground and detected, as well as dust” network. They deployed a sensor network onto a road from an unmanned aerial vehicle (UAV), established time-synchronized multi-hop communication among the nodes on thetoground and detected, as well astotracked and trackpassing vehiclesthrough successfully. vehicles the network. As for network the reported results, according the researcher they managed detecta time-synchronized multi-hop communication network among the nodes on the ground and detected, as well as tracked vehicles passing the network. As researchers for the reported according the researcher they follow thethrough concept proposed by the fromresults, The University of to Berkeley, however themanaged emphasistoisdetect made andWe track vehicles successfully. vehicles passing the network. As for the reported according the researcher they toisdetect and track vehicles successfully. on We human sensing and crowd dynamics monitoring problems. key issues regarding this application are described follow thethrough concept proposed by the researchers fromresults, TheThe University of to Berkeley, however themanaged emphasis made and track vehicles successfully. follow the Experiment concept proposed by the researchers fromtoThe University of Berkeley, the emphasis is made andWe investigated. had been conducted inproblems. order validate solution.however on human sensing and crowd dynamics monitoring The keyproposed issues regarding this application are described follow the Experiment concept proposed by the researchers fromtoThe University of Berkeley, the emphasis is made on human sensing and crowd dynamics monitoring The keyproposed issues regarding this application are described andWe investigated. had been conducted inproblems. order validate solution.however on human sensing and crowd dynamics monitoring problems. The key issues regarding this application are described and investigated. Experiment had been conducted in order to validate proposed solution. 3. Proposed solution and investigated. Experiment had been conducted in order to validate proposed solution. 3. Proposed solution 3. Proposed solution sensor network consists of spatially distributed autonomous sensors capable to monitor physical A typical wireless 3. Proposed solution properties of wireless the environment and objects. Theofnumber ofdistributed sensor nodes could varysensors from acapable few to several hundreds or A typical sensor network consists spatially autonomous to monitor physical A typical wireless network consists ofnumber spatially autonomous sensors to monitor physical thousands, where eachsensor node is connected toThe others, and capable perform some computation, information and properties of the environment and objects. ofdistributed sensortonodes could vary from acapable few gather to several hundreds or A typical wireless network consists of spatially autonomous sensors to physical properties of the environment and number ofdistributed sensor vary acapable fewcase to several hundreds or transmit this information over theobjects. network. As mentioned earlier, wecould consider afrom special ofmonitor sensor network thousands, where eachsensor node is connected toThe others, and capable tonodes perform some computation, gather information and properties of the environment The number of sensor could vary a fewcase to several or thousands, where each node is and connected others, and project. capable tonodes some computation, gather information and considered. The solution isover inspired by theto“smart dust” It isperform anwe emerging technology, which a hundreds potential to transmit this information theobjects. network. As mentioned earlier, consider afrom special ofhas sensor network thousands, where each node is connected to others, and capable to perform some computation, gather information and transmit this information over the network. As mentioned earlier, we consider a special case of sensor network become commercially successful. The concept of this of network is still a theoretical of tiny considered. The solution is inspired by fundamental the “smart dust” project. It istype an emerging technology, which hasconcept a potential to transmit thisThe information network. As dust” mentioned earlier, we a special case ofhas sensor network considered. solution isover inspired by the “smart project. It istype anrobots emerging technology, which a potential to wireless sensor network, made upthe of microelectromechanical or devices, usually referred to as MOTEs, become commercially successful. The fundamental concept ofsensors, this ofconsider network is still a theoretical concept of tiny considered. Thenetwork, solution is inspired by the “smart dust” project. It istype anrobots emerging technology, which hasconcept a potential to become successful. The fundamental concept ofsensors, this of network is still a theoretical tiny that havecommercially self-contained sensing, communication and power. According tousually the design created aofteam wireless sensor made upcomputation, of microelectromechanical or devices, referred to asby MOTEs, become commercially successful. The fundamental concept of this type of network is still a theoretical concept of tiny wireless sensor network, made up of microelectromechanical sensors, robots or devices, usually referred to as MOTEs, from UCself-contained Berkeley researchers, within eachcommunication of this MOTE an integrated of different that have sensing, computation, and is power. According package to the design created bysensors. a team wireless network, made upcomputation, ofwithin microelectromechanical sensors, robots or devices,package referred to asby MOTEs, that self-contained sensing, and is power. According tousually the design created a team fromhave UCsensor Berkeley researchers, eachcommunication of this MOTE an integrated of different sensors.

4

Stanislav Lykov / Transportation Research Procedia 00 (2016) 000–

motivation behind utilization ofStanislav Lykov etfollowing. al. / Transportation Research Procedia (2017) 56–64 The WSN is Nowadays sensors are21 becoming smaller and cheaper. 59 In 4 Stanislav Lykov / the Transportation Research Procedia 00 (2016) 000– future there will be possible to assemble tiny devices with great sensing capabilities, which could be literally The motivation behind utilization of WSN is the following. Nowadays sensors are becoming smaller and cheaper. In everywhere future there around will beus.possible to assemble tiny devices with great sensing capabilities, which could be literally

everywhere around us. 3.1. Wireless sensor based methods

3.1. Wireless sensor based methods One of the most important challenges in sensor networks is positioning. Positioning has to be done in order the network to the function intended.challenges Due to specificity the solution, sensor’s observations be be greatly One of most as important in sensorofnetworks is positioning. Positioningcould has to doneaffected in orderwith the noise, which may result in decreasing of applicability. Therefore, noisy data preprocessing has to be considered. network to function as intended. Due to specificity of the solution, sensor’s observations could be greatly affected with Moreover, since “smart concept assumes a huge number of devices, use commonly approaches noise, which may resultdust” in decreasing of applicability. Therefore, noisy we datacannot preprocessing has toused be considered. such as GPS, simply because it would be costly. More precisely, in our model we assume that following entities exist: Moreover, since “smart dust” concept assumes a huge number of devices, we cannot use commonly used approaches tiny devices (ref. as MOTEs) equipped with several sensors, and user-related devices such as smartphones or wearable such as GPS, simply because it would be costly. More precisely, in our model we assume that following entities exist: devices (ref. (ref. as BEACONs). section is several dedicated to the and description of modified algorithm,orcombined tiny devices as MOTEs) This equipped with sensors, user-related devices positioning such as smartphones wearable with noisy data preprocessing. performance of a network. Moreover, in our model we assume a huge numbercombined of sensor devices (ref. as BEACONs). This section is dedicated to the description of modified positioning algorithm, nodes, therefore cooperative positioning should also be considered. In this section, the description of proposed with noisy data preprocessing. performance of a network. Moreover, in our model we assume a huge number of sensor positioning algorithm is provided. nodes, therefore cooperative positioning should also be considered. In this section, the description of proposed positioning algorithm is provided. 3.2. Positioning algorithm description

3.2. Positioning algorithm description The need of modified positioning algorithm is mainly due to the following factors. First of all, positioning algorithms which suitable for convenient sensor networks may not in ourfactors. case, since have smaller size, as Theare need of modified positioning algorithm is mainly duebe to suitable the following FirstMOTEs of all, positioning algorithms awhich resultare smaller signal range and transmission power. Therefore, the noise could affect greatly the performance ofasa suitable for convenient sensor networks may not be suitable in our case, since MOTEs have smaller size, network. Moreover, in our model we assume a huge number of sensor nodes, therefore cooperative positioning should a result smaller signal range and transmission power. Therefore, the noise could affect greatly the performance of a also be considered. theassume description of number proposed algorithm is cooperative provided. positioning should network. Moreover,In inthis our section, model we a huge ofpositioning sensor nodes, therefore also be considered. In this section, the description of proposed positioning algorithm is provided. Phase 0. Assumptions

Phase 0. Assumptions In our model we assume that entities of two types exist. Small number of fast moving BEACONs, user-related devices, GPS,that andentities large number of slowly MOTEsofequipped withBEACONs, several sensors. While In ourequipped model wewith assume of two types exist.moving Small number fast moving user-related BEACONs are moving in vicinity of MOTE’s sensing radius, MOTE remain motionless. However, MOTEs also could devices, equipped with GPS, and large number of slowly moving MOTEs equipped with several sensors. While change their position in time. Moreover, BEACONs periodically broadcast the following information: ID and current BEACONs are moving in vicinity of MOTE’s sensing radius, MOTE remain motionless. However, MOTEs also could GPS coordinates. MOTE receive this information andperiodically use it for thebroadcast purpose the to determine position. ID and current change their position in time. Moreover, BEACONs followingits information: GPS coordinates. MOTE receive this information and use it for the purpose to determine its position. Phase 1. Data smoothing

Phase 1. Data smoothing The key point of proposed positioning algorithm is noisy data preprocessing. Since “smart dust” concept assume utilization tinyofdevices, their signal transmission be highly Since vulnerable the noise. Therefore, The key of point proposed positioning algorithm iscapabilities noisy datacould preprocessing. “smarttodust” concept assume received data has to be processed before using. As an approach for data preprocessing, utilization of smoothing cubic utilization of tiny devices, their signal transmission capabilities could be highly vulnerable to the noise. Therefore, ����� ,� ) ∀�= �� � , modeled by the relation ��= �(� spline is proposed. Here we assume that we have a set of observations (� � preprocessing, � �). received data has to be processed before using. As an approach for data utilization of smoothing cubic The smoothing spline estimates �̂ ofthat the we function is of defined to be the(�minimizer: �� �, modeled by the relation ��= �(��). spline is proposed. Here we assume have a�set observations �,��) ∀�= ����� The smoothing spline estimates �̂ of�the function � is defined to be the minimizer:

∑������� � �̂ ��� � � � � �� � �̂ �� ��� � �� �� ∑������� � �̂ ��� � � � � �� � �̂ �� ��� � �� �

(1) (1)

Phase2. MOTE’s position estimation

Phase2. MOTE’s position estimation After smoothing noisy observations from BEACONs, we need to estimate the position of MOTEs. For this purpose, theAfter utilization of least-squares circle filfrom algorithm is proposed. BEACONs are moving inside MOTE’s sensing smoothing noisy observations BEACONs, we needWhile to estimate the position of MOTEs. For this purpose, radius, they broadcast their position information. A fraction among these observations, near the edge of MOTE’s sensing the utilization of least-squares circle fil algorithm is proposed. While BEACONs are moving inside MOTE’s sensing ����� �� �. We want find a circle that “best” (in radius isthey used. Assumetheir thatposition this fraction is a finite of points (��,� �) ∀�= radius, broadcast information. A set fraction among these observations, neartothe edge of MOTE’s sensing

least-squares fits that the points. Defineis a finite set of points (��,��) ∀�= ����� �� �. We want to find a circle that “best” (in radius is used.sense) Assume this fraction least-squares sense) fits the points. Define � �

�̅ � ∑� �� � �� � ∑� �� � �� � �� � �̅ � �� � �� � �� � � � � �̅ � ∑� �� � �� � ∑� �� � �� � �� � �̅ � �� � �� � ��

� Further�we define the circle with coordinates of center (��,). We want to minimize

Further we define the circle with coordinates of center (��,). We want to minimize

(2) (2)

60

Author name / Transportation Research Procedia 00 (2016) 000–000

5

Author name / Transportation Research Procedia 00 (2016) 000–000

5

� Procedia 21 (2017) � 56–64 Stanislav Lykov et al. / Transportation Research � � ∑������ � �� ��� � � ����� ��� � � �� � � � � � �� � � � � � �� � � � � � ∑������ � �� ��� � � ����� ���� � �� � � � � � �� � � � � � � �� � � � �

To do so, we differentiate S(� � �� � �� � : To do so, we differentiate S(� � �� � �� � : ��

��

� � ∑� �� �� � �� � ��� � �� � � �� ∑� �� �� � �� � �� �� � � ∑� �� �� � �� � ��� � �� � � �� ∑� �� �� � �� � �� �� �� �� �� Thus, � �, iff ∑� �� �� � �� � � �, and similar ∑� �� � � �� � �� � � � for � �� �� �� �� � �� �� ��� Thus, � �, iff ∑� �� �� � �� � � �, and similar ∑� �� � � �� � �� � � � for � �� �� ��� ��� � Futher, defining �� � ∑� �� and ��� � ∑� �� we obtain a system of equations Futher, defining �� � ∑� �� and ��� � ∑� �� � we obtain a system of equations �� ��

�� ��� � �� ��� �� � � � � �� ��� � ��� ��� �� � � �� �� ��� � �� ���

�

� ����� � ���� � � � � ��� � ���� � � � � �� ����� � ���� � � � � ����� � ���� �

(3) (3)

(4) (4) (5) (5)

(6) (6)

�

Solving this system gives the coordinates of center of circle (the position of MOTE). Solving this system gives the coordinates of center of circle (the position of MOTE). Phase3. Cooperative positioning Phase3. Cooperative positioning An important aspect in proposed positioning scheme is cooperative positioning technique. Due to spatial isolation, and limited sensing radius, several MOTEs could not receive necessary information for localization from BEACONs. An important aspect in proposed positioning scheme is cooperative positioning technique. Due to spatial isolation, Therefore, nearby MOTEs with known positions are used to help them to identify the positions. Geometric location and limited sensing radius, several MOTEs could not receive necessary information for localization from BEACONs. based on time-difference-of-arrival (TDOA) is proposed to use. TDOA measures the distance difference between an Therefore, nearby MOTEs with known positions are used to help them to identify the positions. Geometric location unknown node (target MOTE) and two synchronized reference nodes (reference MOTEs): based on time-difference-of-arrival (TDOA) is proposed to use. TDOA measures the distance difference between an unknown node (target MOTE) and two synchronized reference nodes (reference MOTEs): ‖�� ��� ‖

‖� �� ‖

�

�

� ��� � ��� � ��� � � � � � ��� � ��� � ‖ � ‖� ‖� �� � ��� ‖ � ��� � ��� � ��� � � � � � ��� � ���

where �� – position of target MOTE and ��,�� – positions of reference MOTEs.

where �� – position of target MOTE and ��,�� – positions of reference MOTEs. Phase4. Dynamic position update

(7) (7)

Phase4. Dynamic position update According to the specificity of proposed solution, MOTEs positions could change over time due to the influence of external factors. To account this, in current positioning scheme we assume that newly arrived BEACONs broadcast their According to the specificity of proposed solution, MOTEs positions could change over time due to the influence of position information to MOTEs, so that adjust their position. This process is repeated continuously. external factors. To account this, in current positioning scheme we assume that newly arrived BEACONs broadcast their position information to MOTEs, so that adjust their position. This process is repeated continuously.

Fig 2 a. Positioning scheme description (Dong et al.). Modified. b. Cooperative positioning scheme (Gholami). Modified. Fig 2 a. Positioning scheme description (Dong et al.). Modified. b. Cooperative positioning scheme (Gholami). Modified.

6

Stanislav Lykov / Transportation Research Procedia 00 (2016) 000–

6 Stanislav Lykov / Transportation Research Procedia 00 (2016) 000– Positioning algorithm simulation results 3.3. 6

Stanislav Lykov / Transportation Research Procedia 00 (2016) 000–

Positioning algorithm simulation results 3.3. Simulation Stanislav Lykov et al. /the Transportation Research Procedia 21algorithm. (2017) 56–64 61 had been done in order to investigate performance of proposed According to the setup, MOTE’s sensing radius was represented by a circle with defined radius. Observations from BEACONs were 3.3.Simulation Positioning algorithm simulation results been done in order to investigate the performance of proposed According toradius the setup, represented byhad randomly distributed entities within MOTE’s sensing radius. Duringalgorithm. the simulation, sensing was MOTE’s sensing radius was represented by a circle with defined radius. Observations from BEACONs were set,Simulation and the performance of utilization of least-squares circle fit algorithm combined with smoothing cubic spline for had been done in order to investigate the performance of proposed algorithm. According to the setup, represented by randomly distributed entities within MOTE’s radius.were During the simulation, sensing radius was noisy datasensing pre-processing, according to different of sensing observations investigated. According to the second MOTE’s radius was represented by anumbers circle with defined radius. Observations from BEACONs were set, and thesetup, performance ofassumed utilization of least-squares circletargeted fit algorithm combined with smoothing spline for simulation MOTEsdistributed to be distributed inside area, which was acubic gridradius 100x100. represented by randomly entities within MOTE’s sensing radius. During therepresented simulation,by sensing was noisy data pre-processing, according to different numbers of observations were investigated. According to the second MOTEs wereperformance represented of byutilization randomly distributed entities withfitvariable sensing radius.with The smoothing objective ofcubic this simulation set, and the of least-squares circle algorithm combined spline for simulation setup, MOTEs assumed of to be distributed inside targeted area, which was represented by aiterational/nongrid 100x100. was todata investigate the performance algorithm to MOTEs sensing radius and noisy pre-processing, according topositioning different numbers of according observations were investigated. According to the second MOTEs were represented by randomly distributed entities with variable sensing radius. The objective of this simulation iterational techniques. simulation positioning setup, MOTEs assumed to be distributed inside targeted area, which was represented by a grid 100x100. wasFirst, to investigate the performance positioning algorithm100 according toBEACON MOTEs sensing radius and iterational/nonthe proposed algorithm hadofbeen tested using and 150 observations. Inofcase uniform MOTEs were represented by randomly distributed entities50, with variable sensing radius. The objective thisof simulation iterational positioning techniques. distributed observations, proposed algorithm, as intended, MOTE's location close radius to its true (which was to investigate the performance of positioning algorithmshowed according to MOTEs sensing and position iterational/nonproposed algorithm tested usingobservations 50, 100 and (which 150 BEACON caseinofreal uniform wasFirst, set athe priori at (0,0)techniques. point). Inhad casebeen of non-uniform emulatedobservations. existence of In noise case), iterational positioning distributed observations, proposed algorithm, as intended, showed MOTE's location close to its true(150 position (which theFirst, proposed algorithmalgorithm demonstrated (50 observations); observations); 24.26% the proposed had 29.92% been tested using 50, 100 25.62% and 150(100 BEACON observations. In caseobservations) of uniform was a priori at (0,0) point). case of non-uniform (which emulated existence of noise in real case), bettersetresults comparing to the In one without noisy dataobservations pre-processing. Scatter plots close in case are distributed observations, proposed algorithm, as intended, showed MOTE's location to of its150 trueobservations position (which the proposed algorithm demonstrated 29.92% (50 observations); 25.62% (100 observations); 24.26% (150 observations) shown performance measurement in case of observations non-uniform (which distribution is theexistence differenceofofnoise distances between was setbelow. a prioriThe at (0,0) point). In case of non-uniform emulated in real case), better results and comparing to mean the one without noisyapproaches data pre-processing. Scatter plots in case of 150 observations are trueproposed position calculated values for both (with/without pre-processing). the algorithm demonstrated 29.92% (50 observations); 25.62% (100data observations); 24.26% (150 observations) shown below. The performance measurement in case of non-uniform distribution is the difference of distances between better results comparing to the one without noisy data pre-processing. Scatter plots in case of 150 observations are true position and calculated mean values for both approaches (with/without data pre-processing). shown below. The performance measurement in case of non-uniform distribution is the difference of distances between true position and calculated mean values for both approaches (with/without data pre-processing).

Fig 3. Scatter plots for 150 observations. Proposed algorithm is shown using orange colour, while the algorithm without data pre-processing is shown using blue colour. Fig 3. Scatter plots for 150 observations. Proposed algorithm is shown using orange colour, while the algorithm without data pre-processing is shown using blue colour. Fig 3. Scatter plots for 150 observations. Proposed algorithm is shown using orange colour, the number observations which MOTEs receive couldblue be colour. adjusted via setting while theof algorithm without data pre-processing is shown using

In real life scenario, the frequency of transmitted signals from beacons. Therefore, the obtained results could be used if one wants to meet the certain level In real life scenario, the number of observations which MOTEs receive could be adjusted via setting the frequency of accuracy. of transmitted signals from beacons. Therefore, the obtained results could be used if one wants to meet the certain level In real life scenario, the number of observations which MOTEs receive could be adjusted via setting the frequency of accuracy. of transmitted signals from beacons. Therefore, the obtained results could be used if one wants to meet the certain level 4. Experiment of accuracy. 4. Experiment An important part in conceptual study is validation part, since applicability and reasonability of proposed approach

need to be shown. Therefore, an experiment had been conducted in order to demonstrate the concept of data collection 4. Experiment An important part be in used conceptual studysensing is validation since applicability and reasonability of proposed approach system, which could for human and forpart, crowd dynamics monitoring. need to be shown. Therefore, an experiment had been conducted in order to demonstrate the concept had of data collection Notable that this hadstudy beenisdone under part, the assumption that theand position of MOTEs been already An important partexperiment in conceptual validation since applicability reasonability of proposed approach system, which could be used for human sensing and for crowd dynamics monitoring. determined, using modified positioning algorithm described in previous section. More precisely, there are two phases need to be shown. Therefore, an experiment had been conducted in order to demonstrate the concept of data collection Notable that this experiment had been done that under the assumption that the position of MOTEs had been already in proposed solution. The first phase assumes we distribute MOTEs inside targeted area, and determine their system, which could be used for human sensing and for crowd dynamics monitoring. determined, using modified devices positioning algorithm described in previous section.using Moreproposed precisely,positioning there are two phases positions using user-related (such as smartphones or wearable devices), algorithm. Notable that this experiment had been done under the assumption that the position of MOTEs had been already in proposed we solution. first phase assumes that to wetrack distribute MOTEs targeted area, and under determine their Afterwards, usemodified thisThe network of MOTEs in order pedestrians. Theinside laterMore phase is considered thephases scope determined, using positioning algorithm described in previous section. precisely, there are two positions using user-related devices (such as smartphones or wearable devices), using proposed positioning algorithm. of the experiment. In this the description of experiment setup, roadmap and targeted results isarea, provided. in proposed solution. Thesection first phase assumes that we distribute MOTEs inside and determine their Afterwards, we use this network of MOTEs in order to track pedestrians. The later phase is considered under the scope positions using user-related devices (such as smartphones or wearable devices), using proposed positioning algorithm. of the experiment. In this section the description of experiment setup, roadmap and results is provided. Positioning algorithm simulation resultsin order 4.1. Afterwards, we use this network of MOTEs to track pedestrians. The later phase is considered under the scope of the experiment. In this section the description of experiment setup, roadmap and results is provided. Positioning simulationthe results 4.1.An important algorithm aspect is choosing communication protocol for data transmission. Since many constraints exist, protocol has to be simple, fast and allow to process data from several sources. For this purpose, MQTT (Message Queue Positioning algorithm simulationthe results 4.1.An important aspectconnection is choosing communication protocol for datatotransmission. Sincethis many constraints exist, Telemetry Transport) protocol had been chosen. According the specification, is publish/subscribe, protocol has to be simple, fast and allow to process data from several sources. For this purpose, MQTT (Message Queue very and lightweight messaging designed for constrained and low-bandwidth, high-latency or Ansimple important aspect is choosing theprotocol, communication protocol for data devices transmission. Since many constraints exist, Telemetry Transport) connection protocol had been chosen. According to the specification, this is publish/subscribe, unreliable networks. protocol has to be simple, fast and allow to process data from several sources. For this purpose, MQTT (Message Queue very simple andprinciples lightweight protocol, designed forand constrained devices requirements and low-bandwidth, also high-latency or The design are messaging to minimize bandwidth device to resource attempting Telemetry Transport) connection protocolnetwork had been chosen. According the specification, thiswhilst is publish/subscribe, unreliable networks. very simple and lightweight messaging protocol, designed for constrained devices and low-bandwidth, high-latency or The design principles are to minimize network bandwidth and device resource requirements whilst also attempting

62

Stanislav et al. / Transportation Research00 Procedia 21 (2017) 56–64 Author nameLykov / Transportation Research Procedia (2016) 000–000

7

to ensure reliability and some degree of assurance of delivery. The key idea of the experiment is utilization of WiFi received signal strength indicator (RSSI) as a metric for determining the distance between MOTEs and client. In more complicated scenarios, the way of interaction between sensors and pedestrians could be different, which had been described above in section devoted to human sensing techniques.

Fig 4. Experiment setup

4.2. Experiment results The experiment had been divided into several scenarios. The purpose of the first scenario was to investigate if the proposed system could be used for the purpose to determine crowd behavior pattern. More precisely, an artificial bottleneck was created, and the target was to detect this bottleneck. The second scenario was targeted to investigation of influence of MOTEs density onto positioning accuracy. Bottleneck emulation After the calibration, observations using four RPIs installed inside targeted area had been done. The main objective of this phase was to test that received signal strengths indicators, as a functions of time have similar shape and peaks which corresponds to the situation when pedestrian is in vicinity of RPIs. The result is shown on the figure below. According to the results, RSSIs from all the RPIs have similar pattern, therefore the proposed system could be used in order to obtain signal strength indicators from all devices simultaneously. Moreover, during the experiment, true location of all RPIs had been noted. When pedestrian was moving in vicinity of RPI, this time slot was measured. All the peaks shown of the picture below correspond to this situation, which means that the system works as intended. After this step, an artificial bottleneck had been created inside the targeted area. Five people were moving towards this bottleneck, and then waited for each other, while one of them was passing through. The objective of this scenario was to detect this bottleneck. The results of the observations of RSSI are shown on the figure below.

8

Stanislav Lykov et al. / Transportation (2017) 56–64 Stanislav Lykov / Transportation ResearchResearch Procedia Procedia 00 (2016)21 000–

63

Fig 5 a. Observations of received signal strength indicator using four RPIs. b. Observations of received signal strength indicator using four RPIs for bottleneck scenario.

As clearly seen from the figure above, there is a flat region from 4th till 12th seconds, when the pedestrian had to wait before enter the bottleneck. This waiting near the entrance also widens the peaks (marked red on figure), so that the signal level remains the same during the waiting time. Of course, more precise detection and analysis is essential to use for the purpose of more accurate detection, however, the objective of this scenario was to show the applicability of proposed technique, so obtaining the precise results was not the goal of the study. Investigation of impact of MOTEs density on positioning accuracy Another purpose of this experiment was to investigate the influence of MOTEs density on positioning accuracy. The experiment setup was the following. Four RPIs were installed inside targeted area, and were displaced from the center of the area every 0.25 meters. The initial displacement was 0.5 meters. Moreover, five control points were tracked in order to investigate the accuracy of proposed solution. The error measure was assumed to be the mean area of uncertainty in positioning. The results are shown on the figure below. The horizontal axis represents the total area where RPIs were installed. The vertical axis represents the uncertainty in positioning, calculated for five control points and averaged. The result is shown on the figure above.

Fig 6. Positioning accuracy according to MOTEs density..

According to the results, the error in positioning follows exponential law as intended, since RSSI also follows exponential law. However, due to the limitations of used equipment, the accuracy appeared to be lower than expected. There are several reasons for this. First of all, the results are affected by the position of smartphone in space. Another factor which affected the accuracy is interference between signals from different RPIs, therefore more careful consideration regarding the channels is intended to be done in future. Moreover, since the antennas had different angles during the observation for different control points, received signals were also affected.

64

Stanislav Lykov et al. / Transportation Research Procedia 21 (2017) 56–64 Author name / Transportation Research Procedia 00 (2016) 000–000

9

5. Conclusions This study had an attempt to incorporate the idea of “smart dust” into the concept of the system suitable for human sensing, with the emphasis on crowd dynamics monitoring. For this purpose, the solution based on wireless sensor network as a core component was proposed. Relevant challenges regarding this application of wireless sensor network had been investigated and analyzed. More precisely, modified positioning algorithm, combined with noisy data preprocessing technique was presented. Comprehensive analysis of physical principles for human sensing problem was conducted. Additionally, experiment had been conducted for the purpose to test and demonstrate the concept of data collection system aimed to obtain relevant information regarding pedestrian movements. Need to mention that the key novelty of this study is the attempt to merge several promising technologies, and use their synergy for better performance. However, this study has several limitations. The results of this study only provides the concept of the system which could be used for human sensing and crowd dynamics monitoring, without considering engineering aspects. Therefore, only some of the engineering challenges had been described. As one of the possible extensions, deeper investigation of equipment which could be used, need to be done. Another future objective could be devoted to more comprehensive investigation of physical principles of human sensing. Acknowledgements This work is supported by JSPS Kakenhi Grant Numbers 25249070 and 26220906. References Holly Y., 2015. 717 people dead: what caused a stampede? CNN official report. Soomaroo L., Murray V., 2012. Disasters at mass gatherings: Lessons from history. Public Library of Science. Jacques J. Jr. et al., 2010. Crowd analysis using computer vision techniques. IEEE Signal Processing Magazine. Dublon G., Savvides A., 2011. A survey of human-sensing: Methods for detecting presence, count, location, track, and identity. ACM Computing Surveys. Saad M. Khan, Murbarak S., 2006. A multiview approach to tracking people in crowded scenes using a planar homography constraint. Springer., 3954:133-146 Yusuke K., Yoshikatsu K., Takashi N., 2011. Human skin detection by visible and near-infrared imaging. Proceedings of the 12th IAPR Conference on Machine Vision Applications. Hamid S., Mahmoud M., Luis F., 2013. Atlas project developing a mobile based travel survey. Australasian Transport Research Forum Proceedings. Alexander E., James M. S., 2008. Human detection range by active doppler and passive ultrasonic methods. Proceedings of SPIE - The International Society for Optical Engineering. Mathias V., Tijs N., Matthias D., Nico Van de W., 2012. The use of Bluetooth for analyzing spatiotemporal dynamics of human movement at mass events: A case study of the Ghent festivities. Applied Geography, pp. 208-220. Warneke B., Last M., Liebowitz B., Pister K.S.J., 2001. Smart dust: communicating with a cubic-millimeter computer. IEEE Computer, pp. 44-51. UC Berkeley and MLB Co., 2014. Tracking vehicles with a uav-delivered sensor network. Rpi2 model b hardware general specifications, 2015. Official Raspberry Pi Page. HiveMQ, 2015. Official Documentation. MQTT 101 how to get started with the lightweight IOT protocol. Hao Q., Brady D., 2006. Human tracking with wireless distributed pyroelectric sensors. IEEE sensors journal, vol. 6, no. 6. Arora A., Dutta P., Bapat S., Kulathumani V., Zhang H., Naik V., Mittal V., Cao H., Demirbas M., Gouda M., Choi Y., Herman T., Kulkarni S., Arumugam U., Nesterenko M., Vora A., Miyashita M, 2004. A line in the sand: a wireless sensor network for target detection, classification, and tracking. Computer Networks: The International Journal of Computer and Telecommunications Networking - Special issue: Military communications systems and technologies. Vol. 46, pp. 605-634. Zhang S., Xiao W., 2012. Human tracking for daily life surveillance based on a wireless sensor network. Proceedings of 7th International Conference, WASA 2012, Yellow Mountains, China. Vol. 7405, pp. 677-684.