A Highly Efficient Predetection-Based Anticollision Mechanism for Radio-Frequency Identification

Transcription

1 Journal of Sensor and Actuator Networks Article A Highly Efficient Predetection-Based Anticollision Mechanism for Radio-Frequency Identification Yu-Hsiung Lin 1 and Chiu-Kuo Liang 2, * 1 Department of Electrical Engineering, Chung Hua University, Hsinchu 30012, Taiwan; yuhsiunglin@chu.edu.tw 2 Department of Computer Science and Information Engineering, Chung Hua University, Hsinchu 30012, Taiwan * Correspondence: ckliang@chu.edu.tw; Tel.: Received: 9 January 2018; Accepted: 6 March 2018; Published: 8 March 2018 Abstract: One of research areas in radio-frequency identification (RFID) systems is reduction of identification processing time for a number of tags within an RFID reader recognition region. In last decade, many research results regarding anticollision algorithms have been presented in literature. Most of m are tree-based protocols. However, it is important for tree-based protocols to enhance stability and system throughput, since y may face long identification delays when network density is high. In this study, we present a highly efficient predetection tree-based algorithm to achieve more efficient tag identification performance. Our proposed mechanism can effectively reduce both collisions and idle cycles by exploiting predetection technique and adjustable slot size mechanism. The simulation results show that proposed mechanism can effectively improve tag identification time performance by around 29.9% to 64.8% over previous techniques. Furr, number of query cycles, number of collisions, and total number of slots are reduced compared to previous predetection-based protocols. It is also observed that proposed scheme can have good performance in large-scale RFID systems. Keywords: RFID systems; tag identification; anticollision scheme; query tree; predetection scheme 1. Introduction Radio-frequency identification (RFID) is an automatic technology that is widely used in modern industrial practical applications, such as inventory management [1], object identification and tracking, supply-chain management, and wireless sensor networks [2,3]. It also plays an important role in realizing Internet of Things concept [4]. Traditional identification systems, such as barcodes and smart-card systems, are inefficient at automatic identification and data collection due to ir read rate, visibility, and contact limitations. RFID systems, on or hand, provide fast and reliable communication without requirement of physical visibility or contact between readers and tags. Based on se advantageous features, today s RFID technology goes beyond identification of objects and is being used for localization [5,6] and sensing applications [7]. In addition, more and more battery-less tags can be augmented with sensors, so that tags can send not only ir unique IDs, but also sensed data to readers [8]. One of areas of research in this field is reduction of identification processing time for a given number of tags within an RFID reader recognition zone. To achieve goal of fast tag identification, anticollision protocols are required. An anticollision protocol aims to reduce collisions during tag identification process. The collisions can be categorized into two types: reader collisions and tag collisions. When two or more neighboring readers inquire a tag simultaneously, reader collisions occur. As a result, tag cannot respond with its ID to appropriate inquiring readers. The reader collision problem can be solved easily by detecting collisions and communicating with J. Sens. Actuator Netw. 2018, 7, 13; doi: /jsan

2 J. Sens. Actuator Netw. 2018, 7, 13 2 of 18 or readers. On or hand, tag collisions occur when more than one tag tries to respond to a reader concurrently, which causes reader to identify no tag. In RFID systems with low-cost passive tags, only thing for a tag to do is to respond to reader inquiry. Therefore, tag anticollision protocols are necessary to achieve efficiency of identifying tag IDs in RFID systems. Many research results for reducing identification collisions have been presented in literature. These tag identification mechanisms can be classified into two main categories, Aloha-based anticollision scheme [9 11] and tree-based scheme [12 23]. In Aloha-based scheme, RFID reader creates a frame with several time slots, and n adds frame length to inquiry message sent to tags in its recognition region. Tags respond to reader s inquiry by choosing a random time slot. Tags that respond simultaneously in same time slot cannot be recognized by reader because of collisions. Thus, reader needs to send inquiries repeatedly until all tags are identified, which makes Aloha-based scheme suffer long processing times in identifying large-scale RFID systems [20]. In a tree-based scheme, such as query tree (QT) algorithm [12,13], Improved Bit-by-bit Binary-Tree (IBBT) [14], binary tree splitting protocol [15], tree working algorithm [16,17], and Intelligent Query Tree (IQT) [18], RFID readers use a scheme similar to a binary search algorithm to recognize tags. Once tags collide, readers split collided tags into two subgroups and repeat process until y recognize IDs of tags without collisions. Therefore, readers using tree-based scheme are able to identify all tags. Choi et al. [14] proposed a fast anticollision algorithm called Improved Bit-by-bit Binary-Tree algorithm (IBBT) for ubiquitous identification systems and evaluated its performance along with three older schemes. The reader in IBBT algorithm requests all bits of tag IDs. After reader receives responses from tags, it saves results of each receiving bit. Therefore, reader knows which bits have collided, and it sequentially sends request only for collided bits to tags in a bit-by-bit fashion. Myung et al. [13] proposed adaptive memoryless tag anticollision protocol, which is an extended scheme based on query tree protocol. The reader in this proposed approach uses not only a queue to maintain prefixes, but also an additional candidate queue for maintaining both prefixes of identified nodes and no-response nodes of last identification process. As a result, when number of tags is increased, collision period can be shortened. Sahoo et al. [18] proposed an intelligent query tree algorithm (IQT), which modified query tree protocol for scenarios where tags have some common prefix. In [19], Zhou et al. considered scenarios where multiple readers can be deployed in a region to improve coverage. In such environments, multiple readers can perform tag reading concurrently, and collisions may occur when multiple readers are used within a close vicinity. They discussed problem of slotted scheduled access of RFID tags and developed centralized algorithms in a slotted time model to read all tags. In addition to proof of NP-hardness, y also proposed approximation algorithms for single-channel cases and heuristic algorithms for multiple-channel cases. Unlike Aloha-based protocols that cause tag starvation problems, tree-based protocols deliver 100% guaranteed read rates, but identification delay is relatively long due to large amounts of collisions and idle time. Therefore, collision resolution and idle time elimination become major issues in tree-based anticollision protocols. Jia et al. [24] proposed a collision tree (CT) algorithm, which is based on QT, aiming to eliminate collisions and idle time. The proposed algorithm both generates prefixes and splits tags according to first collide bit, which eliminates idle slots effectively. Lai et al. [25] proposed an optimal binary tracking tree protocol that employs a bit estimation algorithm to split tags into small sets and n uses a binary tracking tree method to quickly identify tags. Yan et al. [26] proposed a hybrid anticollision algorithm, called anticollision protocol based on improved collision detection, to reduce tag collisions. Their approach employs idea of bit-tracking technology and dual-prefix matching into a collision arbitration mechanism in RFID system. Landaluce et al. [27] presented a bit window procedure to manage length of tags bit streams in order to reduce energy consumption in collisions. They modified both QT and CT protocols by adding bit window procedure to produce query window tree protocol and

3 J. Sens. Actuator Netw. 2018, 7, 13 3 of 18 collision window tree protocol. However, se algorithms have complicated implementation and increase identification delay. Ryu et al. [20] proposed a hyper-hybrid query tree protocol (HQT), aiming to reduce identification delays. The proposed protocol reduces identification delay by eliminating collisions and idle time. To reduce collisions, a quarternary query tree is used instead of a binary tree. In quaternary query tree mechanism, prefix string of a collided query will be expanded by two bits instead of one bit in QT protocol. As a result, collision cycles will be reduced substantially, but idle time will increase as a side effect. To eliminate idle time, y introduced a slotted back-off mechanism to reduce unnecessary query commands. When a tag responds to a reader, its back-off time is set from two bits following prefix of tag ID to query string sent by reader. When a collision occurs, reader can partially detect to which subtrees tags belong, and unnecessary queries can potentially be reduced. However, slotted back-off mechanism in HQT protocol cannot check idle cycles between busy slots. To solve problem, a hybrid hyper query tree algorithm (H 2 QT) was proposed by Kim et al. [21]. The protocol aims to reduce idle cycle by using a different slotted back-off mechanism. When a tag responds to a reader, its back-off time is determined from three bits that follow prefix of tag ID to query string. Unlike HQT protocol, back-off mechanism in H 2 QT protocol counts number of 1 s in three bits and uses this number to select a response slot in tag response window. As a result, idle cycles can be reduced substantially. However, collision cycles cannot be reduced by using back-off mechanism in H 2 QT protocol and identification delay is still long. In a preliminary work [22], we addressed key design aspect of using a predetection-based technique called Pre-Detection Broadcasting Query Tree () protocol, to eliminate those unnecessary idle cycles and avoid collision cycles. The reader in protocol allocates some tiny predetection slots for tags to respond in order to reveal distribution of tags. Then, by knowing distribution of tags, only existing tags will respond in corresponding response time slots of following tag response cycle. As a result, no idle slots are wasted and collision slots can be reduced. In this paper, we extend previous work by investigating optimal assignment of tag response time slots in predetection scheme after collecting tag distribution information. The proposed identification scheme will (1) reduce communication overhead between reader and tags during predetection process, (2) reduce tag response time as much as possible, and (3) complete tag identification process as quickly as possible. Our contributions may be summarized as follows: This paper introduces a tag identification problem for RFID systems, that readers collect tags IDs with goal of minimizing transmission completion time. We exploit predetection scheme to reveal distribution of tags IDs, and no idle time slots are wasted. Then, by knowing distribution of tags IDs, reader can allocate appropriate time slots and broadcast arranged information to tags with minimum communication overhead. Simulation results validate that its performance is better than previously proposed predetection solutions. The rest of this paper is organized as follows: in Section 2, we describe idea and function of predetection scheme as preliminary work of this study. Section 3 introduces our proposed predetection-based tag identification technique. In Section 4, we evaluate performance of our proposed scheme, Efficient Pre-Detection based Query Tree (EPDQT) algorithm. Performance comparisons and analyses are also given. Finally, we conclude paper in Section Predetection-Based Scheme The main idea of our anticollision algorithm is to understand distribution of tag IDs as much as possible. When distribution is realized, reader will send this information to tags so that each tag is able to realize appropriate time slot to respond in response cycle. As a result, different

4 J. Sens. Actuator Netw. 2018, 7, 13 4 of 18 tags will n respond to reader in different time slots, which can reduce collisions substantially. J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 4 of 18 In addition, no empty slot occurs, since reader allocates optimal time slots for tags to respond. different To do so, tags wewill propose n arespond predetection-based to reader scheme in different to reveal time slots, distributions which can of reduce tag IDs. collisions Recall that all passive substantially. tags follow In addition, request-respond no empty slot occurs, mode, since y reader cannot allocates initiate optimal communication time slots for between tags mselves to respond. and readers. Therefore, tag identification process can be viewed as several iterations of two operations: To do so, we reader propose request a predetection-based and tag response. scheme to Inreveal our predetection distributions scheme, of tag after IDs. Recall reader request that operation all passive (i.e., tags follow reader sends request-respond request mode, command since y or prefix cannot string initiate to tags), communication operations between mselves and readers. Therefore, tag identification process can be viewed as several in tag response cycle are composed of three phases: (1) predetection phase, (2) broadcasting iterations of two operations: reader request and tag response. In our predetection scheme, phase, and (3) tag response phase, as shown in Figure 1. The purposes or functions of se three after reader request operation (i.e., reader sends request command or prefix string to phases can be described as follows: tags), operations in tag response cycle are composed of three phases: (1) predetection phase, Predetection (2) phase: broadcasting The purpose phase, of and (3) predetection tag response phase phase, is to collect as shown information Figure about 1. The tags to purposes understand or functions overall of se distribution three phases of can tags. be To described do so, as follows: reader allocates many tiny time slots, andpredetection each tiny slot phase: will The bepurpose responsible of for predetection collectingphase tag information is to collect information a particular about group. tags to These tinyunderstand slots are allocated overall fordistribution collecting tag s of tags. information, To do so, instead reader of allocates tag s ID. many Once tiny time slots, has information and each tiny available, slot will reader be responsible can understand for collecting distribution tag information of tagsfor fora that particular particular group. group, andthese this distribution tiny slots are information allocated will collecting help tag s reader information, determining instead of if tag s ID. time Once slot is needed time to receive slot has tag s information ID duringavailable, tag response reader can phase. understand There are 2distribution n tiny timeof slots tags numbered for that particular from 0 to 2 n 1 group, and each and time this distribution slot can be information represented will as an help n-bit binary reader string. for determining Once a tag if matches time slot is prefix needed to receive tag s ID during tag response phase. There are 2 string, it will respond a message to tiny slot that has same label as n tiny time slots next n-bit of tag s ID. numbered from 0 to 2 The message of tags to respond n 1 and each time slot can be represented as an n-bit binary string. Once depends on width of each tiny time slot. In general, width a tag matches prefix string, it will respond a message to tiny slot that has same label of se as next slots n-bit is of smaller tag s ID. than The message length of of tags to tag s respond ID, in depends order to on reduce width communication of each tiny overhead. time slot. After In general, receiving width responses of se from slots tags, is smaller reader than can length recognize of tag s status ID, in oforder each tiny slot: to idle, reduce success, communication or collision. When overhead. no tag After responds, receiving it responses leads to an from idletags, state. When reader only can one tagrecognize responds, it status is recognized of each tiny as aslot: success idle, success, state. When or collision. morewhen than one no tag tagresponds, tries to respond it leads to to reader s an idle query, state. When reader only can one recognize tag responds, it as it ais collision recognized state. as a success state. When more than Broadcasting one tag tries phase: to respond After to predetection reader s query, phase, reader reader can can recognize encode it as a state collision of each state. time into bit Broadcasting 0 or 1. The success phase: After state ispredetection encoded into phase, bit 1, and reader both can idle encode and collision state of states each time are encoded into into bit bit 0 or Then, success in broadcasting state is encoded phase, into bit 1, reader and both sends idle and binary collision string states representing are encoded into bit 0. Then, in broadcasting phase, reader sends binary string representing status of each time slot to tags, so that those tags with IDs that match prefixes can realize how status of each time slot to tags, so that those tags with IDs that match prefixes can realize how many time slots are allocated and which are available to respond with ir IDs. many time slots are allocated and which are available to respond with ir IDs. Tag Tag response response phase: phase: After After receiving receiving broadcasted broadcasted binary binarystring string representing representing status status of each of each tiny tiny slot slot in in predetection predetection phase, phase, each each tag tag is able is able to realize to realize appropriate appropriate time time slot slot to respond to with respond its ID. with Note its that ID. Note eachthat bit each in bit received in received 2 n -bit 2binary n -bit binary string string represents represents status status of corresponding of corresponding tiny slot intiny slot predetection in predetection phase. If itphase. is bit 1, If n it is only bit 1, one n tagonly is detected one tag and is its ID detected can be successfully and its ID can identified be successfully by reader. identified For by this purpose, reader. For this reader purpose, will allocate reader a long time will slot allocate to receive a long tags time IDs. slot On to receive or tags hand, IDs. ifon it is bitor 0, n hand, idle if it cycle is bit or 0, n collision an idle cycle is detected. cycle or The collision readercycle will is not detected. allocatethe anyreader time slot will tonot receive allocate IDsany fortime bothslot cases. to receive As a result, IDs for tag responses both cases. can As be arranged a result, intotag a sequence responses of can time be slots arranged and collisions into a sequence or idleof slots time can slots be and avoided. collisions or idle slots can be avoided. Figure 1. The tag response cycle in predetection scheme. Figure 1. The tag response cycle in predetection scheme.

5 J. Sens. Actuator Netw. 2018, 7, 13 5 of 18 J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 5 of 18 In In our our preliminary preliminary work work [22], [22], we we proposed proposed an an anticollision anticollision scheme scheme called called Pre-Detection Pre-Detection Broadcasting Broadcasting Query Query Tree Tree () () protocol, protocol, which which is is based based on on aforementioned aforementioned predetection predetection scheme. scheme. In In protocol, reader reader will will allocate allocate four four tiny tiny slots, slots, say say 00, 00, 01, 01, 10, 10, and and 11, 11, in in predetection predetection phase phase for for tags tags to to respond. respond. Then, Then, by by using using quaternary quaternary query query tree tree mechanism, mechanism, each each tag tag that that matches matches reader s query query prefix prefix string string responds on on time time slot slot that that is is determined determined from from next next two two bits bits of of its its tag tag ID. ID. In In order order to to reduce reduce communication communication overhead, overhead, only only a four-bit four-bit random random number number is is responded responded to to by by aa tag tag instead instead of of whole whole tag tag ID. ID. If re If re is no is tag no response tag response in a time in a time slot, slot, reader reader can realize can realize that itthat is anit idle is an slot, idle and slot, noand time no slot time will slot bewill allocated be allocated in tag in response tag response phase. If phase. receiving If receiving four bitsfour can bits be identified can be identified correctly, correctly, reader can reader realize can that realize time that slot is time a success slot is slot, a success and slot, corresponding and corresponding time slot will time beslot allocated will be inallocated tag response in tag phase response to receive phase to receive whole tag ID. whole If more tag than ID. If one more tag responds, than one since tag responds, each tag responds since each to atag four-bit responds random to number, a four-bit a collision random slot number, can bea identified collision slot if different can be identified random numbers if different arerandom transmitted numbers to are reader. transmitted Since to probability reader. of Since transmitting probability same random of transmitting numbersame by different random tags number is low, by different reader can tags identify is low, a collision reader with can high identify accuracy a collision when with multiple high tags accuracy respond. when As amultiple result, re tags isrespond. a high possibility As a result, thatre reader is a high can realize possibility exact that distribution reader can ofrealize tag IDs. exact distribution of tag IDs. As As mentioned, mentioned, in in predetection predetection phase, phase, those those tags tags that that match match prefixes prefixes respond respond on time slot time that slot isthat determined is determined from from next two next bits, two which bits, which follow follow prefixes prefixes of irof IDs ir identical IDs identical to query to string. query The string. reader The uses reader a four-bit uses a binary four-bit string binary representation string representation to indicateto indicate status of time status slots of and time broadcasts slots and broadcasts status to tags. status After to broadcasting, tags. After broadcasting, those tags with those IDs tags identical with IDs to identical query string to tag query canstring realizetag can exact realize time slotexact to respond time slot by checking to respond by corresponding checking bit corresponding of receivedbit four-bit of binary received string. four-bit Thebinary tag canstring. respond The with tag its can IDrespond in upcoming with its ID tagin response upcoming phase iftag response corresponding phase bit if in corresponding broadcastingbit four-bit in string broadcasting is 1. Orwise, four-bit string no tagis response 1. Orwise, occurs. no For tag those response tags that occurs. can respond For those intags that tag-response can respond phase, in tag-response exact time slot phase, to respond exact can time alsoslot beto obtained respond bycan counting also be obtained number by of counting 1 s broadcasting number of 1 s four-bit in string broadcasting from four-bit start bitstring to from corresponding start bit bit to that represents corresponding next bit that tworepresents bits to prefixes next two of ir bits to IDs. Consider prefixes of ir example IDs. Consider in Figure 2. example The tag responding Figure 2. on The tag 11responding time slot inon predetection 11 time slot phase in is predetection aware that its phase ID can is aware be responded that its ID tocan by be responded reader onlyto onby third reader timeonly slot, on since third received time slot, four-bit since broadcasted received string four-bit is 1101, broadcasted and it is string third is 1101, 1 fromand it beginning is third bit 1 to from corresponding beginning bit bit. to corresponding bit. Figure 2. The tag response cycle of our preliminary work. Figure 2. The tag response cycle of our preliminary work. 3. Proposed Anticollision Scheme 3. Proposed Anticollision Scheme Recall that previous proposed protocol can eliminate all idle cycles. However, it is Recall that previous proposed protocol can eliminate all idle cycles. However, it is possible for different tags to respond to same four-bit random numbers in same time slot. In possible for different tags to respond to same four-bit random numbers in same time slot. such a situation, collision will be identified in tag response phase instead of predetection In such a situation, collision will be identified in tag response phase instead of predetection phase. As a result, more time slots will be needed to identify those collision tags and time phase. As a result, more time slots will be needed to identify those collision tags and time needed needed to identify all tags will be prolonged. In this paper, we propose a highly efficient algorithm, to identify all tags will be prolonged. In this paper, we propose a highly efficient algorithm, called called Efficient Pre-Detection based Query Tree (EPDQT) algorithm, to improve Efficient Pre-Detection based Query Tree (EPDQT) algorithm, to improve performance in terms of performance in terms of increasing system throughput and minimizing identification delay. increasing system throughput and minimizing identification delay System Model of EDPQT Scheme In this section, we will describe system model of our proposed EDPQT scheme. Suppose re are t tags in reader field and each tag has same length of m-bit ID, which can be

6 J. Sens. Actuator Netw. 2018, 7, 13 6 of System Model of EDPQT Scheme J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 6 of 18 In this section, we will describe system model of our proposed EDPQT scheme. Suppose re denoted are t tagsas ina binary reader string fieldb1b2...bm, and eachwhere tag has each bi same is eir length 0 or of1. m-bit Since ID, our which scheme can be is based denoted on as a predetection binary stringmechanism, b 1 b 2...b m, where tag eachresponse b i is eir cycle 0 oris 1. composed Since ourof scheme three is phases: based predetection predetection phase, broadcasting mechanism, phase, tag response and tag response cycle is composed phase, as of mentioned three phases: above. predetection In predetection phase, broadcasting phase, number phase, and of time tag response slots depends phase, on as a mentioned parameter, above. say n, Inwhich predetection is varied and phase, can be defined number by ofusers time when slots depends using this onalgorithm. a parameter, After saya n, reader which sends is varied an inquiring and canp-bit be defined string (prefix), by users0 when p m, using to tags, this algorithm. allocates After 2 n time a reader slots for sends tags an to inquiring respond. p-bit Then, string tags located (prefix), within 0 p m, reader s to tags, it close allocates vicinity 2 n respond time slotsto for tags reader to respond. if inquiring Then, tags bits located are within same as reader s first p close bits of vicinity tag IDs. respond When to a tag responds, reader if it selects inquiring one bits of 2are n time slots sameto asrespond, firstdepending p bits of tagon IDs. When n bits afollowing tag responds, first it selects p bits of one its of tag 2 n ID, time that slots is, bp+1bp+2...bp+n. to respond, Thus, depending time on slot nrepresents bits following value first of n pbits. After of itsdetermining tag ID, that is, btime p+1 b p+2 slot...bto p+n respond,. Thus, time tag sends slot represents (n + 1)th bit value following of n bits. After first p determining bits of its tag ID, time that slot is, bp+n+1, to respond, to reader. tag sends For example, (n + 1)th if n bit = 2 following and tag ID first is p010110, bits of its n tagafter ID, that is, reader b p+n+1 sends, to a two-bit reader. For prefix example, string, ifsay n = 01, 2 and tag tag will IDrespond is , with nits after fifth bit, reader which sends is 1, on a two-bit 01 prefix time slot string, saypredetection 01, tag will phase, respond since with its following fifth bit, two which bits isafter 1, on first 01 time two slot bits inis 01. predetection This means that phase, in our since algorithm, following re two is only bitsone after bit for first tag twoto bits respond is 01. in This means predetection that in phase. our algorithm, Furrmore, rein is onlytag oneresponse bit for phase, tag tothose respond successfully in predetection recognized phase. tags Furrmore, will send inremaining tag response IDs, that phase, is, bp+n+2...bm, those successfully to reader. recognized As a result, tags will communication send remaining overhead IDs, can that be is, reduced b p+n+2...b substantially. m, to reader. Figure As 3 a shows result, communication system model overhead of can interaction be reduced between substantially. reader Figure and 3 tags shows in our system proposed model EPDQT of scheme. interaction between reader and tags in our proposed EPDQT scheme. Figure 3. The interaction between reader and tag in Efficient Pre-Detection based Query Tree Figure 3. The interaction between reader and tag in Efficient Pre-Detection based Query Tree (EPDQT) scheme. (EPDQT) scheme. After predetection phase is completed, distribution of tags can be realized by collecting After predetection phase is completed, distribution of tags can be realized by collecting status of each time slot in predetection phase. The status of each time slot can be categorized status of each time slot in predetection phase. The status of each time slot can be categorized as as one of four states: idle, collision, 0-success, or 1-success. The idle state indicates that no tag matches one of four states: idle, collision, 0-success, or 1-success. The idle state indicates that no tag matches with with prefixes. The collision state occurs when more than one tag matches with prefixes. prefixes. The collision state occurs when more than one tag matches with prefixes. Then it is Then it is necessary to resolve collisions in upcoming query cycles. The 0-success and necessary to resolve collisions in upcoming query cycles. The 0-success and 1-success states 1-success states indicate that reader receives only 0 or 1, respectively, in predetection phase. It indicate should that be noted reader that receives 0-success only 0 or state 1, respectively, or 1-success in state predetection may occur phase. when It only should one be tag noted is matched that 0-success or more than state one or 1-success tag is matched state may but returns occur when same only bit one string. tag is The matched latter or case more will than cause one a collision, tag is matched as many but tags returns return ir same IDs bit on string. same The time latter slot case in will tag cause response a collision, phase. as As many a result, tags return reader ir needs IDs on more same cycles time to identify slot in those tag collided response tags. phase. It also As a should result, be noted reader that needs once more collision cycles to state identify has those been detected, collided tags. eir It in also should predetection be noted phase that or once in tag collision response state phase, has been prefix detected, string, eir along in with predetection corresponding phase or time in slot tag string, response will phase, be added prefix to a waiting string, along queue with for furr corresponding inquiry by time reader. slot string, This will means be added that to whole a waiting tag queue identification for furr process inquiry will by be repeated reader. until This means waiting that queue whole is empty. tag identification process will be repeated until waiting queue is empty. As states of all time slots in predetection phase are collected, each state is encoded into a two-bit pattern. The idle, 1-success, 0-success, and collision states are represented as 00, 01, 10, and 11 bit patterns, respectively. Then, in broadcasting phase, reader broadcasts whole bit patterns of all time slots in predetection phase. Since re are 2 n slots allocated in

7 J. Sens. Actuator Netw. 2018, 7, 13 7 of 18 As states of all time slots in predetection phase are collected, each state is encoded into a two-bit pattern. The idle, 1-success, 0-success, and collision states are represented as 00, 01, 10, and 11 bit patterns, respectively. Then, in broadcasting phase, reader broadcasts whole bit patterns of all time slots in predetection phase. Since re are 2 n slots allocated in predetection phase and each slot can be encoded into a two-bit pattern, re will be a total 2 n+1 -bit binary string broadcasted by reader. Each bit in broadcasting binary string indicates wher a time slot is needed in tag response phase. The time slots needed in tag response phase can also be represented as an (n + 1)-bit binary string. Thus, for n = 1, re may be four time slots, which can be represented as 00, 01, 10, and 11, in tag response phase, and for n = 2, we may have eight time slots, denoted as 000, 001,..., 111. However, not every time slot will exist in tag response phase. In our scheme, reader will allocate time slots for those bits in broadcasting binary string that have 1 s. No time slot is needed for those 0 s. For example, reader will allocate three time slots during tag response phase, namely 010, 100, and 111, if broadcasting binary string is As tags receive broadcasting 2 n+1 -bit string, each tag will check next (n + 1)-bit after first p-bit of its ID to see if corresponding bit in broadcasted binary string is 1 or 0. The tag will respond during tag response phase only if corresponding bit is 1. Furrmore, tag can also be aware of exact time slot to respond by counting number of 1 s in received broadcasting string from start bit to corresponding bit. Again, if n = 2, tag ID is , and prefix string is 01, n after receiving broadcasted binary string , tag will check next three-bit, which is 011, after first 01 bits of its ID to see if corresponding bit in broadcasted binary string is 1 or 0. Furrmore, to accelerate identification, we propose different procedures for some special cases where length of a prefix query string approaches length of tag IDs. These cases will be explained in detail in following section Special Cases of EPDQT Scheme In our scheme, as number of iterations grows, length of prefix query string also grows. This means that value of p will increase as identification process keeps running. In such scenarios, special procedures are needed to speed up identification process when p is approaching m. Based on our observations, re are three possible special cases that can occur during identification process, which can be described as follows: p + n = m 1: In this case, after predetection phase, re is no bit left for each tag to respond in tag response phase. Thus, reader can immediately identify tag IDs after predetection phase by realizing status of each time slot in predetection phase. Assume that prefix string is b 1 b 2...b p and binary string of corresponding time slot in predetection phase is b p+1 b p+2...b p+n. Then, when reader realizes status of a time slot as a 0-success or 1-success state, it can immediately identify a tag with ID as b 1...b p b p+1...b p+n followed by bit 0 or bit 1, respectively. For collision state, reader can also identify two tags with IDs b 1...b p b p+1...b p+n 0 and b 1...b p b p+1...b p+n 1. p + n m: In this case, reader will change value of n to m p 1 before sending prefix string. Then, reader will follow procedure described in previous case, since now p + n = m 1. p + n = m 2: In this case, re is only one bit left for each tag to respond in tag response phase. The reader can identify tag IDs by verifying results of each slot in tag response phase. Note that re are three possible results: idle, success, and collision. For collision case, reader can immediately identify two tags with IDs b 1...b p b p+1...b p+n b p+n+1 0 and b 1...b p b p+1...b p+n b p+n+1 1.

8 J. Sens. Actuator Netw. 2018, 7, 13 8 of An Example of EPDQT Scheme To facilitate understanding of EPDQT algorithm, we show below an example of identification procedure of our proposed protocol. We assume that re are eight tags with eight-bit unique IDs: tag A (with ID ), tag B (with ID ), tag C (with ID ), tag D (with ID ), tag E (with ID ), tag F (with ID ), tag G (with ID ), and tagj. H Sens. (with Actuator IDNetw ). 2018, 7, x FOR The PEER whole REVIEW process is also shown in Figure 4. The iterations 8 of 18 of identification process using EPDQT protocol with n = 2 are described as follows: Iteration iterations 1: First of ofidentification all, waiting process queue using is empty EPDQT andprotocol reader with sends n = 2 are empty-prefix described as to follows: tags and allocates four time slots for predetection phase. All tags respond to this empty-prefix request command. Iteration 1: InFirst this of case, all, tags waiting A, B, queue andis Cempty respond and onreader 00sends time slot empty-prefix and sendto bit 1 to reader. Tag tags Dand responds allocates four on time 10slots timefor slot with predetection bit 0. Tags phase. E, All F, G, tags and respond H respond to this on empty-prefix 11 time slot, request command. In this case, tags A, B, and C respond on 00 time slot and send bit 1 to both with bit 0, to reader. After receiving responses from tags, reader can realize status reader. Tag D responds on 10 time slot with bit 0. Tags E, F, G, and H respond on 11 time of four time slot, both slotswith in bit predetection 0, to reader. phase After receiving as 1-success responses in 00from slot, tags, idle in reader 01 slot, can realize 0-success in 10 slot, and status 0-success of four time in slots 11in slot. predetection Thus, broadcasting phase as 1-success string representing 00 slot, idle in status 01 slot, of time slots in 0-success predetection in 10 phase slot, and is 0-success encodedin as slot. Thus, This indicates broadcasting that re string will representing be three time slots, namely status 001, 100, of and time 110, slots allocated in predetection in tag phase response is encoded phase. as After receiving This indicates broadcasting that re string, will be three time slots, namely 001, 100, and 110, allocated in tag response phase. After tags A, B, and C will respond with 01010, 10101, and 11011, respectively, to reader on time slot 001. receiving broadcasting string, tags A, B, and C will respond with 01010, 10101, and 11011, As a result, reader detects a collision and 001 will be added to waiting queue for furr query. respectively, to reader on time slot 001. As a result, reader detects a collision and 001 will be Furrmore, added tag to D waiting responds queue with for furr onquery. slot 100, Furrmore, which can tag be D responds successfully with identified on slot by100, reader. Tags E, F, which G, and can Hbe will successfully all respond identified on time by slot reader. 110 Tags withe, 10001, F, G, and 10101, H will 11001, all respond and 11010, on time respectively. slot Then a collision 110 with 10001, is detected 10101, and 11001, 110 and will 11010, be added respectively. to Then waiting a collision queue. is Figure detected 4and shows 110 will be situation that occurs added into this iteration. waiting queue. Figure 4a shows situation that occurs in this iteration. (a) Response of tags in iteration 1 (b) Response of tags in iteration 2 (c) Response of tags in iteration 3 Figure 4. Cont.

9 J. Sens. Actuator Netw. 2018, 7, 13 9 of 18 J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 9 of 18 (d) Response of tags in final iteration Figure 4. An example of identification process of our scheme with n = 2. Figure 4. An example of identification process of our scheme with n = 2. Iteration 2: The reader takes prefix string 001 out of queue and sends it to tags. This time, only tags A, B, and C will respond in predetection phase, since tag D is identified and tags E, F, Iteration G, and 2: H The cannot reader match takes with prefix prefix string, string as shown 001 out in of Figure queue 4b. In this and case, sends tag A itresponds to tags. on This time, only tags A, 01 B, time and slot Cwith willbit respond 0, tag B responds in predetection 10 time phase, slot with since bit 1, tag and Dtag is identified C responds and on tags E, F, G, and H11 cannot time slot match with bit with 0. Then prefix reader string, realizes as shown status in of slots Figure in 4b. predetection In this case, phase tagas Aidle responds in on 01 time 00 slot slot, with 0-success bit 0, in tag B01 responds slot, 1-success on in time slot, slot and with 0-success bit 1, in and 11 tag slot. CThe responds reader on n broadcasts status string to tags, which indicates that re are three slots, namely 11 time slot with bit 0. Then reader realizes status of slots in predetection phase as idle in 010, 101, and 110, allocated in tag response phase. After receiving broadcasting string, tags 00 slot, A, B, 0-success and C will in respond 01 slot, with 1-success 10, 10, and in 1110 to slot, reader and 0-success on time in slots 010, 11 slot. 101, The and reader 110, n broadcasts respectively. status As a string result, reader can to tags, successfully whichidentify indicates all of that m. re are three slots, namely 010, 101, and 110, Iteration allocated 3: The inreader tag takes response prefix phase. string 110 After out of receiving queue and broadcasting sends it to tags. string, This time, tags A, B, and C will tags respond E and F will with respond 10, 10, in and 11 predetection to reader phase on on time 10 slots time 010, slot 101, with and 0 and 110, 1, respectively, respectively. As a and tags G and H will respond on 11 time slot with bit 0, as shown in Figure 4c. In this case, result, reader can successfully identify all of m. reader realizes status of slots as idle state in 00 slot, 0-success in 01 slot, 1-success in Iteration 3: The reader takes prefix string 110 out of queue and sends it to tags. This time, 10 slot, and 0-success in 11 slot. The reader n broadcasts status string to tags, tags E and which F will indicates respond that re in are predetection three slots, namely phase 010, on 101, 10and time 110, slot allocated with 0in and 1, tag respectively, response and tags G and phase. H will After respond receiving on broadcasting 11 time slot string, withtag bite 0, will as respond shown in 01 Figure on time 4c. slot In100, thistag case, F will reader realizesrespond status 00 on oftime slotsslot as 101, idleand state tags in G and 00H slot, will 0-success respond 01 inand 10, 01respectively, slot, 1-success on time in slot slot, and 0-success As ina result, 11 slot. reader The can reader successfully n broadcasts identify tags E status and F. string However, tags G toand tags, H cannot whichbe indicates identified, since a collision occurred on time slot 110. Therefore, reader will add binary that re are three slots, namely 010, 101, and 110, allocated in tag response phase. After receiving string to waiting queue and execute next iteration. broadcasting string, tag E will respond 01 on time slot 100, tag F will respond 00 on time slot Iteration 4: The reader takes prefix string out of queue for furr query. At this 101, andmoment, tags G andlength H will p of respond prefix 01string and 10, is 6, respectively, which satisfies on time condition slot 110. of Asspecial a result, case as reader can successfully mentioned identify above; that tags is, p E+ andn = m. F. However, Therefore, tags reader G and will Hchange cannot be value identified, of n to 1 since and send a collision occurred onprefix timestring slot 110. and n Therefore, to tags. This means reader that, will in add predetection binaryphase, stringonly two to time slots waiting will queue and execute be allocated next for iteration. tags to respond. In this case, tag G will respond on time slot 0 with bit 1 and tag H will respond on time slot 1 with bit 0. Since tags G and H are responding with ir last bit of ID to Iteration 4: The reader takes prefix string out of queue for furr query. At this reader in predetection phase, reader can identify ir IDs according to status of ir moment, corresponding lengthtime p ofslot. In prefix this case, string both is time 6, which slots for satisfies tags G and H condition are recognized of as special success case as mentioned states. above; Thus, both that tags is, p can + n be = identified m. Therefore, immediately reader and no will tag response change phase value is needed. of n to Figure 1 and send prefix 4d shows string and situation. n to tags. This means that, in predetection phase, only two time slots will be allocated for tags to respond. In this case, tag G will respond on time slot 0 with bit 1 and tag H 4. Performance Evaluation will respond on time slot 1 with bit 0. Since tags G and H are responding with ir last bit of ID to reader into evaluate predetection performance phase, of reader our proposed can identify mechanism, ir IDs we according implemented to EPDQT status of ir scheme with three different settings, n = 2, n = 3, and n = 4, which are indicated as, corresponding time slot. In this case, both time slots for tags G and H are recognized as success states., and, respectively, along with H 2 QT and algorithms. We conducted Thus, both tags can be identified immediately and no tag response phase is needed. Figure 4d shows a set of simulation experiments for proposed algorithms. All experiments were performed on a situation. computer equipped with a 3.0 GHz central processing unit and 4 GB memory in C# on.net platform. Every experiment was repeated 100 times, and recorded data was averaged over those 4. Performance runs into Evaluation final results. According to EPCglobal C1 G2 standard [28], simulation environment is as follows: we To evaluate performance of our proposed mechanism, we implemented EPDQT scheme consider an RFID system that has one reader and t tags within a reading range where t = 100, 500,..., with three different settings, n = 2, n = 3, and n = 4, which are indicated as,, and, respectively, along with H 2 QT and algorithms. We conducted a set of simulation experiments for proposed algorithms. All experiments were performed on a computer equipped with a 3.0 GHz central processing unit and 4 GB memory in C# on.net platform. Every experiment was repeated 100 times, and recorded data was averaged over those runs into final results.

10 J. Sens. Actuator Netw. 2018, 7, of 18 According to EPCglobal C1 G2 standard [28], simulation environment is as follows: we consider an RFID system that has one reader and t tags within a reading range where t = 100, 500,..., The IDs of all tags are 96 bits long. We also consider two different distributions of tag IDs, uniform random distribution and sequential distribution. The tag IDs in sequential distribution are grouped and consecutive. The data rate of channels is set to be 128 kbps. For convenience, we consider a noise-free channel between reader and tags and ignore propagation delay of signal, since all aforementioned algorithms would be impacted equally. The parameter settings in our simulations are listed in Table 1. More practically, we formulate time needed for reader query command, predetection phase, broadcasting phase, and tag response phase in different protocols, as shown in Table 2. Table 1. Parameter settings in simulations. Parameter Value Meaning t 100, 500, 1000,..., 4000 Number of tags L uid 96 bits Length of tag ID L q 128 bits Length of query command L b 22 bits Length of broadcast command L res Max. 96 bits Length of returned ID from tags R com 128 kbps Data rate between reader and tags T bit µs 1-bit transmission time, i.e., 1/R com T rt 20 µs Waiting time from reader transmission to tag response T tr 20 µs Waiting time from tag response to reader transmission Table 2. Parameters used in simulations. Protocol Reader Query Predetection Phase Broadcasting Phase Tag Response Phase H 2 QT L q T bit - - T rt + (L uid T bit ) 4 L q T bit T rt + 16 T bit T tr + (L b + 4) T bit T rt + (L res T bit ) N 1 1 L q T bit T rt + 4 T bit T tr + (L b + 8) T bit T rt + (L res T bit ) N 1 2 L q T bit T rt + 8 T bit T tr + (L b + 16) T bit T rt + (L res T bit ) N 1 3 L q T bit T rt + 16 T bit T tr + (L b + 32) T bit T rt + (L res T bit ) N N 1, N 2, N 3, and N 4 indicate numbers of slots in tag response phase in protocols,,, and, respectively. Our simulations focus on determining performance for impact of number of tags in terms of number of queries, number of total slots, delay time, and system efficiency. System efficiency is measured by two different perspectives: in terms of slots and time. In terms of slots, system efficiency is measured as SE s = S id /S tot, where S id is amount of identification slots, which is equal to number of tags, and S tot is total number of slots. In terms of time, system efficiency is measured as SE t = T id /T tot, where T id is time taken by identification slots and T tot is total execution time Impact of : Uniform Distribution Number of Queries vs. This experiment evaluates effect that number of tags has on performance of number of queries needed by reader to complete tag identification of H 2 QT,,,, and approaches. The results are shown in Figure 5. In this graph, we can see that as number of tags increases, number of queries of each algorithm increases linearly. This is obvious, because number of queries is proportional to number of tags. Furrmore, number of queries is also dependent on number of collisions that occurred during identification process. Therefore, as number of tags increases, number of collisions increases proportionally, which implies that number of queries increases as well. However, our proposed EPDQT schemes

11 J. Sens. Actuator Netw. 2018, 7, of 18 require fewer queries compared with or schemes. For example, when t = 4000, number of J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW queries needed by reader in H 2 11 of 18 QT,,,, and protocols is 3539, J. Sens. 2883, Actuator 2540, Netw. 2145, 2018, and 7, x FOR 2018, PEER respectively. REVIEW Thus, our proposed EPDQT schemes perform11 around of 18 schemes perform around 29% to 43% better than H 2 29% to 43% better than H QT scheme. This is because 2 QT scheme. This is because EPDQT EPDQT protocols generate fewer collision slots. schemes protocols It can perform generate also be seen around fewer collision that 29% number to slots. 43% better It can also of queries than be in seen H 2 QT that scheme. number is around This of 19% is queries because in less than in EPDQT is around 19% less than in. This shows that for EPDQT schemes, number of. protocols generate fewer collision slots. It can also be seen that number of queries in queries is This decreases shows that as forparameter EPDQT increases, schemes, since more number time ofslots queries are decreases allocated in as predetection parameter increases, around 19% less than in. This shows that for EPDQT schemes, number of queries phase since to identify more time more slots tags. are allocated in predetection phase to identify more tags. decreases as parameter increases, since more time slots are allocated in predetection phase to identify more tags Number Number of Queries of Queries Figure 5. Number of queries required to complete identification. Figure 5. Number of queries required to complete identification. Figure 5. Number of queries required to complete identification Number of Total Slots vs Number of Total Slots vs This Number experiment of Total examines Slots vs. Number effect of that Tags number of has on number of total slots generated This experiment by reader examines to complete effect tag identification that number of of Htags 2 QT, has, on, number of, total slots This experiment examines effect that number of tags generated and by approaches. reader to complete In this experiment, tag identification we ignore of tiny H 2 has on number of total slots QT, slots, in predetection, generated by reader to complete tag identification of H phase and and take only approaches. number of Inslots this that experiment, occurred we in ignore tag response tiny 2 QT,,,, phase slots into account, predetection since it will phase and approaches. In this experiment, we ignore tiny slots in predetection phase andtake take more onlytime to number finish of slots process that compared occurredto in time tag spent response in phase predetection into account, phase. since The it slot will and take only number of slots that occurred in tag response phase into account, since it will take take comparison more more time time results to to finish finish are given process process in Figure compared compared 6. to to time time spent spent in in predetection predetection phase. phase. The The slot slot comparison comparison results results are are given given in in Figure Figure ,000 Number Number of Slots of Slots 16,000 14,000 14,000 12,000 12,000 10,000 10, Number 2000 of 2500 Tags Figure 6. Number of slots generated by each protocol. Figure 6. Number of slots generated by each protocol. In general, Figure 6 Figure shows 6. that Number H 2 QT ofhas slots generated most total by each slots. protocol. That is because H 2 QT allocates four In slots general, to split Figure tags in 6 each shows query that cycle. H 2 QT Many has idle most or collision total slots. slots That will is not because reduced. H 2 QT However, allocates four in In our slots general, proposed to split Figure EPDQT tags 6in shows each and query that Hcycle. protocols, 2 QTMany has idle most reader or collision total uses slots. a predetection That will not is because mechanism reduced. H 2 QT However, to realize allocates four in slots our distribution proposed to split tags of EPDQT tag inids, each and which query decreases cycle. protocols, Many most idle of reader or collision uses a slots, predetection slotsand willonly notmechanism be a few reduced. collisions to However, realize may in occur ourdistribution proposed in tag EPDQT of response tag IDs, and phase. which As decreases a protocols, result, most number of reader collision of total uses slots aslots, predetection decreases and only as a mechanism few well. collisions For example, to may realize occur when in t = 4000, tag response number phase. of total As a slots result, generated number by of total reader slots in decreases H 2 QT, as, well. For, example, when, t = 4000, and number protocols of total slots is 14,150, generated 4096, by 6539, 6144, reader and in 6017, H 2 QT, respectively.,, Thus, our, proposed EPDQT and schemes protocols perform is 91.9% 14,150, to 93.8% 4096, 6539, better 6144, than and H6017, 2 QT respectively. scheme. Our Thus, proposed our proposed EPDQT schemes perform 91.9% to 93.8% better than H 2 QT scheme. Our proposed

12 J. Sens. Actuator Netw. 2018, 7, of 18 distribution of tag IDs, which decreases most of collision slots, and only a few collisions may occur in tag response phase. As a result, number of total slots decreases as well. For example, when t = 4000, number of total slots generated by reader in H 2 QT,,,, and protocols is 14,150, 4096, 6539, 6144, and 6017, respectively. Thus, our proposed EPDQT schemes perform 91.9% to 93.8% better than H 2 J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW QT scheme. Our proposed EPDQT protocols cost fewer slots than H 2 12 of 18 QT. However, EPDQT protocols cost more slots than. EPDQT Theprotocols main reason cost fewer is that slots than Huses 2 QT. ahowever, random number EPDQT forprotocols each tagcost to respond more slots during than predetection. The phase, main and reason no slot is that will be allocated uses a inrandom tagnumber response for phase each once tag to respond collision during is detected. On predetection or hand, phase, and EPDQT no slot protocols will be allocated will allocate in slots tag to response split phase collided once tags. collision As a result, is many detected. collisions On detected or in hand, predetection EPDQT protocols phase will will still allocate collide slots into split tag response collided phase. tags. As Wea also result, many collisions detected in predetection phase will still collide in tag response phase. note that number of total slots in is 7.98% less than in. This is because as We also note that number of total slots in is 7.98% less than in. This is parameters in EPDQT protocols increase, both queries and collisions decrease, which decreases because as parameters in EPDQT protocols increase, both queries and collisions decrease, total slots needed to identify tags. which decreases total slots needed to identify tags Delay Time vs Delay Time vs. This This experiment experiment examined examined effect that number of tags has on total time required to complete tag identification of 2 effect that number of tags has on total time required to complete tag identification of H 2 QT,,,,, and and approaches. approaches. The The time time comparison results results are are shown in in Figure 7. In Figure In Figure 7, we 7, we can can see see that that as as number of tags increases, total total time time required for for each each algorithm to to complete identification increases. However, our proposed EPDQT schemes require require less less time time compared compared to or to or schemes. schemes. For For example, when when t = t 4000, = 4000, total total time time required in in H 2 QT,, H 2 QT,,,,,, and and protocols protocols is 16.07, 8.07, is 16.07, 7.11, 8.07, 6.01, 7.11, and6.01, 5.65 and s, respectively s, Thus, respectively. our proposed Thus, EPDQT our proposed schemes EPDQT perform schemes 54.5% perform to 64.8% 54.5% better to than 64.8% better H 2 QT than scheme. H 2 QT This is scheme. This is because EPDQT protocols can reduce idle and collision slots substantially. because EPDQT protocols can reduce idle and collision slots substantially. We also note that We also note that total time required in is 20.5% better than in PDQT-2. This indicates total time required in is 20.5% better than in PDQT-2. This indicates that for EPDQT that for EPDQT schemes, total time required decreases as parameter value increases, since schemes, total time required decreases as parameter value increases, since more time slots are more time slots are allocated in predetection phase and refore more tags can be identified allocated successfully. in predetection phase and refore more tags can be identified successfully. Delay Time (s) Figure 7. Time required to complete tag identification. Figure 7. Time required to complete tag identification System Efficiency vs System Efficiency vs. In following, we examine effect that number of tags has on system efficiency in terms In of following, slots and we time examine for H 2 QT, effect, that, number of, tags hasand on system efficiency approaches. in terms of slots Results and for time slot forsystem H 2 efficiency QT,, and, time system, efficiency andare presented approaches. in Figures Results 8 and 9. for In slot system terms efficiency of slot and system timefficiency, system efficiency each of are compared presented approaches in Figures 8experiences and 9. In terms similar of slot system system efficiency, efficiency each as of number compared of tags approaches increases. experiences For example, similar slot system efficiency as of number H 2 QT, of,,, and protocols is around 28.51%, 97.68%, 62.04%, 65.16%, and 66.58%, respectively, all of which are independent of number of tags. The rationale behind se results is quite clear. Since numbers of identification slots and total slots both increase as number of tags increases, slot system efficiency almost flattens. Furrmore, slot system efficiency of,,, and protocols significantly outperforms

13 J. Sens. Actuator Netw. 2018, 7, of 18 tags increases. For example, slot system efficiency of H 2 QT,,,, and protocols is around 28.51%, 97.68%, 62.04%, 65.16%, and 66.58%, respectively, all of which are independent of number of tags. The rationale behind se results is quite clear. Since numbers of identification slots and total slots both increase as number of tags increases, slot system efficiency almost J. Sens. flattens. Actuator Furrmore, Netw. 2018, 7, x FOR slot PEER system REVIEW efficiency of,,, and 13 of 18 protocols J. Sens. Actuator significantly Netw. 2018, outperforms 7, x FOR PEER REVIEW H 2 QT protocol. The rationale behind this result is 13 also of 18 clear. In H 2 2 QT protocol, protocol. each The rationale query cycle behind allocates this result fouris time also slots clear. for In tags Hto 2 QT respond. protocol, Some each ofquery slots maycycle H receive allocates 2 QT protocol. one tag s four response time The rationale slots and for n tags behind to identify respond. this result Some is also tag s of ID, clear. but slots In may H rest of receive 2 QT protocol, mone maytag s each receive response query multiple cycle tags and responses n allocates identify four or no response, time tag s slots ID, for and but tags as a to respond. result rest of m Some slot may may of collide receive slots or multiple may receive be idle. Therefore, tags one responses tag s response utilization or no and response, n of slots in H 2 and identify as a result tag s ID, slot but may collide rest or of be m idle. may Therefore, receive multiple utilization tags of responses slots in H QT is poor. On or hand, in predetection-based protocols, no idle slot is 2 or QT no is allocated response, poor. On and or as a result hand, in slot predetection-based may collide or be protocols, idle. Therefore, no idle slot utilization is allocated of and slots most in H 2 of QT is and poor. most of collisions are detected during predetection phase. As a result, allocated slots collisions On are detected or hand, during predetection-based phase. protocols, As a result, no idle slot allocated is allocated slots and are almost of used are almost collisions used for identifying tags. However, protocol outperforms EPDQT protocols, for identifying are detected tags. However, during predetection protocol phase. As outperforms a result, allocated EPDQT slots protocols, are almost since used it sincefor reduces it reduces identifying most most collision tags. collision However, slots. slots. Furrmore, Furrmore, for protocol for EPDQT outperforms EPDQT protocols, protocols, as EPDQT parameter as protocols, parameter increases, since increases, slot it slot reduces efficiency most increases. collision Figure Figure slots. 8 Furrmore, 8shows that that when for when t EPDQT = t = 4000, 4000, protocols, as protocol parameter protocol performs performs increases, around around slot 5.3% efficiency 5.3% better better than increases. than Figure 8 shows protocol. that when t = 4000, protocol performs around 5.3% better than protocol. Slot Slot System System Efficiency Number 2000 of 2500 Tags Figure 8. Slot system efficiency of compared protocols. Figure Figure Slot Slot system system efficiency efficiency of of compared compared protocols. protocols Figure 9. Time system efficiency of compared protocols. Figure 9. Time system efficiency of compared protocols. Figure 9. Time system efficiency of compared protocols. In terms of time system efficiency, as shown in Figure 9, each of compared approaches experiences In terms similar of time time system system efficiency, efficiency as shown as in number Figure of 9, tags each increases. of compared For example, approaches time experiences system In terms efficiency ofsimilar time of system time Hsystem 2 QT, efficiency,, efficiency as, shown as in number, Figureof 9, and tags each increases. of compared protocols For example, is approaches around time experiences system 18.66%, efficiency 38.05%, similar time 68.95%, of system H76.71%, 2 QT, efficiency, and 79.85%, as, respectively, number, of tags all and increases. of which are Forprotocols independent example, is time around of system efficiency 18.66%, number of38.05%, tags. H %, The QT, reason, 76.71%, behind, and this 79.85%, result, respectively, is similar and to all of reason which protocols for are slot independent system is around efficiency. of 18.66%, 38.05%, number Furrmore, 68.95%, of tags %, time The system reason andefficiency 79.85%, behind respectively, this of result, is similar all, of which to, reason are independent for and slot system of efficiency. protocols number of tags. Furrmore, also Thesignificantly reasontime behind outperforms system thisefficiency result ish similar 2 of QT protocol., The reason, rationale for, slot behind system this and efficiency. result is quite Furrmore, protocols clear. also Since significantly re are many outperforms more queries Hin 2 QT protocol. H 2 QT protocol The rationale than in behind predetection-based this result is quite protocols, clear. Since by adding re are time many for more reader queries query in commands H 2 QT protocol and than waiting in time predetection-based from reader query protocols, tag by response, adding time system time for efficiency reader query of H 2 QT commands is lower than and or waiting protocols. time Furrmore, from reader query proposed to tag response, EPDQT protocols time system outperform efficiency of H 2 is scheme lower than by or around protocols. 30% 41%. Furrmore, The rationale proposed behind EPDQT protocols outperform scheme by around 30% to 41%. The rationale behind Time Time Systen Systen Efficiency

14 J. Sens. Actuator Netw. 2018, 7, of 18 time system efficiency of,,, and protocols also significantly outperforms H 2 QT protocol. The rationale behind this result is quite clear. Since re are many more queries in H 2 QT protocol than in predetection-based protocols, by adding time for reader query commands and waiting time from reader query to tag response, time system efficiency of H 2 QT is lower than or protocols. Furrmore, proposed EPDQT protocols outperform scheme by around 30% to 41%. The rationale behind se results is quite clear. Since communication J. Sens. Actuator Netw. overhead 2018, 7, x infor PEER predetection REVIEW phase of proposed EPDQT protocols is much 14 of 18 less than that in protocol, time for each query cycle in EPDQT is much less than in. As ase result, results proposed is quite clear. EDPQT Since protocols communication can achieve better overhead timein system predetection efficiency phase than of ors. proposed EPDQT protocols is much less than that in protocol, time for each query 4.2. Impact cycle in ofepdqt Number is much of Tags: less Sequential than in. Distribution As a result, proposed EDPQT protocols can achieve better time system efficiency than ors Number of Queries vs Impact of : Sequential Distribution In this simulation, we compare number of query cycles of test protocols as value of t increases Number from 100 of Queries to 4000vs. innumber sequential of Tags distribution fashion. The query cycle comparison results are given in Figure 10. In general, Figure 10 shows that each protocol generates fewer query cycles In this simulation, we compare number of query cycles of test protocols as value of t in sequential distribution than in uniform distribution. This is because tag IDs are consecutive increases from 100 to 4000 in sequential distribution fashion. The query cycle comparison results are in sequential given in Figure distribution. 10. In general, Thus, Figure a sequence 10 shows of tags that may each be protocol evenly generates distributed fewer into query cycles query in slots. As asequential result, more distribution tags canthan be identified in uniform in distribution. a query cycle, This which is because costs fewer tag query IDs are cycles consecutive than uniform in distribution. sequential Figure distribution. 10 also Thus, shows a sequence that Hof 2 QT tags has may be most evenly query distributed cycles. into The reason query slots. is quite As a clear. Despite result, more different tags can distributions be identified of tag in a IDs, query H 2 QT cycle, spends which more costs query fewer cycles query than cycles than ors, uniform since it cannot distribution. reduce Figure collisions 10 also shows regardless that H 2 QT of tag has ID distribution. most query cycles. However, The reason in is our quite proposed clear. Despite EPDQT protocols, different each query distributions cycle can of tag identify IDs, Hmore 2 QT spends tags than more Hquery 2 QT. The cycles rationale than behind ors, this since result it cannot is clear. Duereduce to predetection collisions regardless mechanism, of tag ID distribution. tiny slots are However, allocated in our sequentially. proposed EPDQT As a result, protocols, a sequence each of tags query cancycle be identified can identify in more a query tags cycle, than which H 2 QT. The decreases rationale behind number this of result query is clear. cyclesdue dramatically. to predetection mechanism, tiny slots are allocated sequentially. As a result, a sequence of tags can be Furrmore, protocol outperforms and protocols. This is because identified in a query cycle, which decreases number of query cycles dramatically. Furrmore, prefix string in is extended by five bits instead of three bits and four bits in protocol outperforms and protocols. This is because prefix string and protocols, respectively, which costs fewer query cycles for to identify in is extended by five bits instead of three bits and four bits in and a tag s protocols, ID. respectively, which costs fewer query cycles for to identify a tag s ID. Number of Queries Figure 10. Number of queries required to complete identification in sequential distribution. Figure 10. Number of queries required to complete identification in sequential distribution Number of Total Slots vs Number of Total Slots vs. In this simulation, we compare number of total slots of test protocols. The slot comparison results In thisare simulation, given in Figure we compare 11. Figure number 11 shows ofthat total each slots protocol of test generates protocols. fewer The slot total comparison slots in results sequential are given distribution in Figure than 11. uniform Figure distribution 11 shows that due each to protocol same reason generates given fewer in total previous slots in sequential subsection. distribution Figure 11 also thanshows uniform that distribution EPDQT protocols due tosignificantly same outperform reason given H 2 QT, in since previous re are much fewer query cycles in EPDQT than in. However, performs a little better than EPDQT. This is because generates much fewer collisions than EPDQT protocols, which costs fewer total slots.

15 J. Sens. Actuator Netw. 2018, 7, of 18 subsection. Figure 11 also shows that EPDQT protocols significantly outperform H 2 QT, since re are much fewer query cycles in EPDQT than in. However, performs a little better than EPDQT. This is because generates much fewer collisions than EPDQT protocols, which costs fewer total slots. J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 15 of 18 J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 15 of Number Number of Slots of Slots Figure 11. Number of slots generated by each protocol in sequential distribution. Figure 11. Number of slots generated by each protocol in sequential distribution. Figure 11. Number of slots generated by each protocol in sequential distribution Delay Time vs Delay Time vs This Delay simulation Time vs. Number compares of Tags time needed to complete tag identification in sequential distribution This simulation for test compares protocols. The time time needed comparison to complete results tag are identification shown in Figure in sequential 12. Figure 12 distribution shows This simulation compares time needed to complete tag identification in sequential for test that protocols. each protocol Thecosts timeless comparison time than uniform results are distribution. shown in Figure also Figure shows 12 that shows EPDQT that each distribution for test protocols. The time comparison results are shown in Figure 12. Figure 12 shows protocol protocols costsignificantly less time than outperform uniform distribution. H 2 QT and Figure protocols. 12 also shows that EPDQT protocols that each protocol costs less significantly outperform H 2 time than uniform distribution. Figure 12 also shows that EPDQT protocols significantly outperform QT and H protocols. 2 QT and protocols. 10 Delay Delay Time Time (s) (s) Number 2000 of 2500 Tags Figure 12. Time required to complete tag identification in sequential distribution. Figure 12. Time required to complete tag identification in sequential distribution System Figure Efficiency 12. Time vs. required Number toof complete Tags tag identification in sequential distribution The System comparison Efficiency results vs. Number for system of Tags efficiency in terms of slots and time for test protocols are given System in Figures Efficiency 13 vs. and Number 14. Each of of Tags compared approaches experiences similar slot system The comparison results for system efficiency in terms of slots and time for test protocols are efficiency as number of tags increases, as shown in Figure 13. Figure 13 also shows that each given The comparison in Figures 13 results and for 14. system Each of efficiency compared in terms approaches of slots and experiences time for test similar protocols slot system are given protocol achieves better performance under sequential distribution than uniform distribution. This in Figures efficiency 13 as and 14. number Each of tags compared increases, as approaches shown in Figure experiences 13. Figure similar 13 also slot shows system that efficiency each is because of large amount of tag identifications in query cycles of each protocol. Specifically, as protocol number achieves of tags better increases, performance as shown under insequential Figure 13. distribution Figure 13 than also uniform showsdistribution. that each protocol This EPDQT protocols achieve very high slot system efficiency (around 85.4% to 98.0%). This is achieves is because because better of performance large amount EPDQT protocols under of tag generate sequential identifications much distribution in query fewer collisions than uniform cycles of each protocol. Specifically, under sequential distribution. distribution This is than because of EPDQT protocols achieve very high slot system efficiency (around 85.4% to 98.0%). This is uniform large amount distribution. of tag However, identifications time system in efficiency query cycles of ofepdqt each protocol. protocols Specifically, is much worse than EPDQT protocols because EPDQT protocols generate much fewer collisions under sequential distribution than H 2 QT achieve and. very high The rationale slot system behind efficiency this result (around is also 85.4% clear. tosince 98.0%). This EPDQT is because protocols cost EPDQT a uniform distribution. However, time system efficiency of EPDQT protocols is much worse than protocols few collisions generate in much sequential fewer distribution, collisions under time sequential spent in tag distribution identification than is much uniform less distribution. than H communication 2 QT and. The rationale behind this result is also clear. Since EPDQT protocols cost a overhead for reader query commands. As a result, time system efficiency is low few collisions in sequential distribution, time spent in tag identification is much less than compared to H 2 QT and protocols. communication overhead for reader query commands. As a result, time system efficiency is low compared to H 2 QT and protocols.

16 J. Sens. Actuator Netw. 2018, 7, of 18 However, time system efficiency of EPDQT protocols is much worse than H 2 QT and. The rationale behind this result is also clear. Since EPDQT protocols cost a few collisions in sequential distribution, time spent in tag identification is much less than communication overhead for reader query commands. As a result, time system efficiency is low compared to H 2 QT and protocols. J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 16 of 18 J. Sens. Actuator Netw. 2018, 7, x FOR PEER REVIEW 16 of 18 Slot Slot System System Efficiency Number 2000 of Tags Figure 13. Slot system efficiency of compared protocols in sequential distribution. Figure 13. Slot system efficiency of compared protocols in sequential distribution. Figure 13. Slot system efficiency of compared protocols in sequential distribution. Time Time Systen Systen Efficiency Number 2000 of 2500 Tags Figure 14. Time system efficiency of compared protocols in sequential distribution. Figure 14. Time system efficiency of compared protocols in sequential distribution. Figure 14. Time system efficiency of compared protocols in sequential distribution. 5. Conclusions 5. Conclusions 5. Conclusions Developing a highly efficient tag identification process in a large-scale RFID system is a crucial and very Developing challenging a highly task. efficient Many collisions tag identification may occur process during a large-scale tag identification RFID system process is a crucial due to and Developing nature very challenging of large-scale a highly task. efficient RFID Many systems. tag collisions identification On may or occur process hand, during many in a large-scale idle tag identification cycles RFID may system also process occur, is due a since crucial to and tag very nature IDs in, challenging of say, large-scale a big task. shopping RFID Many systems. mall collisions may On have may some or occur common hand, during many prefix idle (e.g., tag cycles common identification may also product occur, process code since or due to tag vendor IDs nature in, ID). say, of Identification large-scale a big shopping protocols RFIDmall systems. may such have as On Hsome 2 QT can common or reduce hand, prefix idle many cycles, (e.g., idle common but cycles re product may are still also code many occur, or since vendor collisions. tag IDs ID). Therefore, in, Identification say, acollision big shopping protocols resolution such mallbecomes as may H 2 QT have a can major some reduce issue common idle in improving cycles, prefix but (e.g., performance. re common are still In product many this code collisions. paper, or vendor we Therefore, proposed ID). Identification collision a nearly resolution collision-free protocols becomes such tag identification as a Hmajor 2 QTissue can algorithm reduce in improving idle to cycles, reduce performance. but collisions rein are and this still many paper, improve collisions. we identification proposed Therefore, a performance nearly collision collision-free substantially. resolution tag becomes identification We used a major algorithm predetection issue into improving technique reduce collisions to performance. detect and all In improve idle this cycles paper, identification and we proposed many performance possible a nearly collided collision-free substantially. inquiries. We tag As used identification a result, predetection we were algorithm able technique to tonot reduce only to detect collisions reduce all andidle unnecessary improve cycles identification and collided many inquiries, possible performance but collided also substantially. eliminate inquiries. idle As We time a result, used slots. we Therefore, predetection were able to performance technique not only reduce tof detect tag allunnecessary identification idle cycles and collided can many be significantly inquiries, possiblebut collided improved. also eliminate inquiries. To evaluate idle As time a result, slots. efficiency Therefore, weof were our able proposed performance to not protocol, onlyof reduce tag we unnecessary identification implemented collided can our be proposed inquiries, significantly EPDQT butimproved. also protocol eliminate To along evaluate idle with time previous slots. efficiency Therefore, H 2 QT of our and proposed performance protocol, protocols of we in tag identification implemented two tag ID distributions: can our beproposed significantly uniform EPDQT improved. distribution protocol along Toand evaluate sequential with previous efficiency distribution. H 2 QT and ofthe our simulation proposed protocols results protocol, in we two show implemented tag that ID our distributions: proposed our proposed protocol uniform EPDQT can distribution protocol provide and along considerable sequential with previous improvement distribution. H 2 QTin and The terms simulation of number protocols results of in show queries, that number our proposed of total protocol slots, system can provide efficiency, considerable and total improvement time required in for terms tag identification. of number of queries, Simulation number results of also total show slots, system that efficiency, proposed and EPDQT total time protocol required can for provide tag identification. substantial Simulation improvement results in identification also show time that for a large-scale proposed RFID EPDQT system. protocol can provide substantial improvement in identification time for a large-scale RFID system. Author Contributions: Yu-Hsiung Lin and Chiu-Kuo Liang conceived and designed experiments; Author Yu-Hsiung Contributions: Lin performed Yu-Hsiung experiments; Lin and Yu-Hsiung Chiu-Kuo Lin Liang and Chiu-Kuo conceived Liang and analyzed designed data; experiments; Yu-Hsiung

17 J. Sens. Actuator Netw. 2018, 7, of 18 two tag ID distributions: uniform distribution and sequential distribution. The simulation results show that our proposed protocol can provide considerable improvement in terms of number of queries, number of total slots, system efficiency, and total time required for tag identification. Simulation results also show that proposed EPDQT protocol can provide substantial improvement in identification time for a large-scale RFID system. Author Contributions: Yu-Hsiung Lin and Chiu-Kuo Liang conceived and designed experiments; Yu-Hsiung Lin performed experiments; Yu-Hsiung Lin and Chiu-Kuo Liang analyzed data; Yu-Hsiung Lin contributed reagents/materials/analysis tools; Chiu-Kuo Liang wrote paper. Conflicts of Interest: The authors declare no conflict of interest. References 1. Vogt, H. Efficient Object Identification with Passive RFID Tags. Lect. Notes Comput. Sci. 2002, 2414, Li, Y.; Ding, X. Protecting RFID communications in supply chains. In Proceedings of 2nd ACM Symposium on Information, Computer and Communications Security, Singapore, March 2007; ACM: New York, NY, USA, 2007; pp Shirehjini, A.; Yassine, A.; Shirmohammadi, S. Equipment location in hospitals using RFID-based positioning system. IEEE Trans. Inf. Technol. Biomed. 2012, 16, [CrossRef] [PubMed] 4. Lee, S.R.; Joo, S.D.; Lee, C.W. An enhanced dynamic framed slotted ALOHA algorithm for RFID tag identification. In Proceedings of Second Annual International Conference on Mobile and Ubiquitous Systems: Networking and Services, San Diego, CA, USA, July 2005; pp Dardari, D.; Decarli, N.; Guerra, A.; Guidi, F. The future of ultra-wideband localization in RFID. In Proceedings of 2016 IEEE International Conference on RFID (RFID), Orlando, FL, USA, 3 5 May Qiu, L.; Huang, Z.; Wirström, N.; Voigt, T. 3DinSAR: Object 3D localization for indoor RFID applications. In Proceedings of 2016 IEEE International Conference on RFID (RFID), Orlando, FL, USA, 3 5 May Naderiparizi, S.; Parks, A.N.; Kapetanovic, Z.; Ransford, B.; Smith, J.R. WISPCam: A battery-free RFID camera. In Proceedings of 2015 IEEE International Conference on RFID (RFID), San Diego, CA, USA, April 2015; pp Philipose, M.; Smith, J.R.; Jiang, B.; Mamishev, A.; Roy, S.; Sundara-Rajan, K. Battery-free wireless identification and sensing. IEEE Pervasive Comput. 2005, 4, [CrossRef] 9. Park, J.; Chung, M.; Lee, T.J. Identification of RFID Tags in Framed-Slotted ALOHA with Robust Estimation and Binary Selection. IEEE Commun. Lett. 2007, 11, [CrossRef] 10. Klair, D.K.; Chin, K.W.; Raad, R. An investigation into energy efficiency of pure and slotted aloha based RFID anticollision protocols. In Proceedings of IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks (WoWMoM), Espoo, Finland, June 2007; pp Zhen, B.; Kobayashi, M.; Shimizui, M. Framed aloha for multiple RFID objects Identification. IEICE Trans. Commun. 2005, E88-B, [CrossRef] 12. Law, C.; Lee, K.; Siu, K.Y. Efficient Memoryless Protocol for Tag Identification. In Proceedings of 4th International Workshop on Discrete Algorithms and Methods for Mobile Computing and Communications, Boston, MA, USA, 11 August Myung, J.; Lee, W. An adaptive memoryless tag anticollision protocol for RFID networks. In Proceedings of 24th Annual IEEE Conference on Computer Communications (INFOCOM 05), Poster Session, Miami, FL, USA, March Choi, H.S.; Cha, J.R.; Kim, J.H. Improved Bit-by-bit Binary Tree Algorithm in Ubiquitous ID System. In Proceedings of 5th Pacific Rim Conference on Multimedia, Tokyo, Japan, 30 November 3 December 2004; pp Myung, J.; Lee, W.; Srivastava, J. Adaptive binary splitting for efficient RFID tag anti-collision. IEEE Commun. Lett. 2006, 10, [CrossRef] 16. Capetanakis, J.I. Tree algorithms for packet broadcast channels. IEEE Trans. Inf. Theory 1979, 25, [CrossRef]

18 J. Sens. Actuator Netw. 2018, 7, of Feng, B.; Tao, L.J.; Bo, G.J.; Hua, D.Z. ID-Binary tree stack anti-collision algorithm for RFID. In Proceedings of 11th IEEE Symposium on Computers and Communications (ISCC 06), Cagliari, Italy, June 2006; pp Sahoo, A.; Iyer, S.; Bhandari, N. Improving RFID System to Read Tags Efficiently; KRSIT Technical Report; IIT Bombay: Mumbai, India, June Zhou, Z.; Gupta, H.; Das, S.R.; Zhu, X. Slotted Scheduled Tag Access in Multi-Reader RFID Systems. In Proceedings of IEEE International Conference on Networks Protocols (ICNP), Beijing, China, October 2007; pp Ryu, J.; Lee, H.; Seok, Y.; Kwon, T.; Choi, Y. A Hybrid Query Tree Protocol for Tag Collision Arbitration in RFID systems. In Proceedings of IEEE International Conference on Communications (ICC-07), Glasgow, UK, June 2007; pp Kim, T.H.; Lee, S.J. A Hybrid Hyper Tag Anti-Collision Algorithm in RFID System. In Proceedings of 11th International Conference on Advanced Communication Technology (ICACT 2009), Phoenix Park, Korea, February 2009; Volume 2, pp Liang, C.K.; Chien, Y.C.; Tsai, C.H. A Pre-Detection Query Tree Tag Anti-Collision Scheme in RFID Systems. In Proceedings of Seventh International Conference on Sensor Technologies and Applications (SENSORCOMM 13), Barcelona, Spain, August 2013; pp Zhou, F.; Jin, D.; Huang, C.; Hao, M. Optimize Power Consumption of Passive Electronic Tags for Anti-collision Schemes. In Proceedings of 5th International Conference on ASIC, Beijing, China, October 2003; Volume 2, pp Jia, X.; Feng, Q.; Yu, L. Stability analysis of an efficient anti-collision protocol for RFID tag identification. IEEE Trans. Commun. 2012, 60, [CrossRef] 25. Lai, Y.C.; Hsiao, L.Y.; Lin, B.S. Optimal slot assignment for binary tracking tree protocol in RFID tag identification. IEEE/ACM Trans. Netw. 2015, 23, [CrossRef] 26. Yan, Y.N.; Xiong, J. An Efficient Tag Identification Algorithm Based on Improved Collision Detection. IEICE Trans. Commun. 2016, E99-B, Landaluce, H.; Perallos, A.; Onieva, E.; Arjona, L.; Bengtsson, L. An Energy and Identificaion Time Decreasing Procedure for Memoryless RFID Tag Anticollision Protocols. IEEE Trans. Wirel. Commun. 2016, 15, [CrossRef] 28. EPCglobal. EPC Radio-Frequency Identity Protocols Generation-2 UHF RFID. November Available online: (accessed on 12 December 2017) by authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under terms and conditions of Creative Commons Attribution (CC BY) license (