An Anti-Collision Algorithm for RFID Based on an Array and Encoding Scheme

Transcription

1 information Article An Anti-Collision Algorithm for RFID Based on an Array and Encoding Scheme Baolong Liu * and Xiaohao Su School Computing Science & Engineering, Xi an Technological University, Xi an , China; su-xiaohao@foxmail.com * Correspondence: b.liu@xatu.edu.cn; Tel.: Received: 26 January 2018; Accepted: 7 March 2018; Published: 10 March 2018 Abstract: In order to solve problem tag collision in Radio Frequency Identification (RFID) system, paper proposes a Multi-Bit Identification Collision Tree (MICT) algorithm based on a collision tree. The algorithm uses an array scheme to mark collision bits in identification process, and determines collision information according to first few bits tag, which can effectively reduce number recognitions and amount communication data. The testing results show that proposed algorithm reduces time complexity by about 38% and communication complexity by about 27% compared to existing collision-tree-based algorithms. Through oretical analysis and experimental evaluation, MICT algorithm has obvious advantages in terms time and communication complexity compared to or typical algorithms. The algorithm can be applied to field RFID-related systems to significantly improve system efficiency. Keywords: Radio Frequency Identification; anti-collision algorithm; electronic tags; code identification 1. Introduction RFID technology is a non-contact intelligent identification technology. Compared to traditional automatic identification technology, it has advantages small volume, low cost, large storage, high security, and reusability. As a result, RFID is rapidly replacing outdated barcode systems and various smartcard systems. RFID systems have great application value in fields industrial production, logistics, cargo and health management, safety inspection, and intelligent transportation [1 4]. However, in process RFID identification multiple tags simultaneously, collision between tags is a key issue affecting efficiency RFID identification. Several classical algorithms have been presented, such as ALOHA algorithm [5,6], Query Tree (QT) [7 10] algorithm, Collision Tree (CT) algorithm [11], etc. These anti-collision algorithms can be divided into two categories: probabilistic methods based on ALOHA, and deterministic methods based on tree structure [12 15]. The probabilistic identification algorithms are simple to design and identification time is short, but re will be a tag hunger problem. The deterministic algorithms ensure that all tags in identified range are identified, but ten do not perform well in communication efficiency. Recently, several algorithms such as tag grouping [16,17], bit tracking [18,19], and multi-bit identification have been proposed to improve performance RFID identification. Jung has proposed an Optimized Binary Search with Multiple Collision Bits Resolution Anti-Collision Algorithm (OBS-MCBR) [20]. In this algorithm, collision bits are extracted and coded, and collision location information is restored according to collision situation. The performance algorithm is 30% higher than that CT algorithm. Wang has put forward an anti-collision algorithm based on multi-bit identification (MBI) [21]. The algorithm makes full use collision information a certain length, and uses a distinguishable grouping scheme to group specific Information 2018, 9, 63; doi: /info

2 Information 2018, 9, x FOR PEER REVIEW 2 17 Information 2018, 9, makes full use collision information a certain length, and uses a distinguishable grouping scheme to group specific query prefixes. Thus, several collision bits can be continuously identified query prefixes. when multiple Thus, several collisions collision occur. bits However, can be continuously performance identified proposed when multiple algorithms collisions still has occur. low efficiency. However, performance proposed algorithms still has low efficiency. The The CT CTalgorithm algorithmonly only updates updates query query prefix prefix for for collision collisionlocation, location, which whichreduces reduces unnecessary unnecessary queries queries and and eliminates eliminates idle idle time time slot. slot. In In addition, addition, in in recognition recognition process, process, CT CT algorithm algorithmdoes does not not need need label label memory memory information; information; algorithm algorithm design design is simple is simple and and effective, effective, and and has has good good improvement improvement space. space. This This paper paper proposes proposes a Multi-bit a Multi-bit Identification Identification Collision Collision Tree Tree (MICT) (MICT) algorithm algorithmbased based on on CT CT algorithm. algorithm. The The algorithm algorithm uses uses an an array array scheme scheme to to mark mark collision collision bits, bits, and and adopts adopts coding coding scheme scheme to analyze to analyze collision collision bits, bits, which which can quickly can quickly identify identify collision collision multiple multiple collision collision bits. The bits. experimental The experimental results results show show that that new algorithm new algorithm has a has better a better performance performance in terms in terms time time complexity complexity and and communication communication complexity complexity compared compared to to existing existing algorithms. algorithms. 2. Problem 2. Identification In In RFID system, reader and and tag tag communicate with with each or each through or through radio frequency radio frequency signals. signals. The channel The channel is a signal is a transmission signal transmission medium, medium, and its and roleits is to role transmit is to transmit signal carrying signal carrying information information from from sender to sender receiver. to When receiver. multiple When readers multiple or multiple readers tagsor transmit multiple RF signals tags transmit at same RF signals time, at signals same time, will collide signals withwill eachcollide or with in ach wireless or channel, in wireless causing channel, collision causing problems. collision Theproblems. reader cannot The reader receivecannot correct receive signal correct information signal or information signal fails or tosignal be sent. fails Theto collision be sent. types The collision RFID systems types include RFID tag systems collision include and reader tag collision, and in which reader reader collision, collision in which includes reader reader-tag collision collision includes and reader-tag reader-reader collision collision. and reader-reader collision Tag-Tag Collision Tag-tag collision refers refers to to collision problem between multiple tags, tags, usually usually occurs occurs when when multiple tags tags send send data data signals signals to to reader reader at at same same time, time, as as shown in in Figure The identification range Reader Tag4 Tag2 Reader Tag3 Tag1 Figure Figure 1. Tag-tag 1. Tag-tag collision collision diagram. diagram. Pluralities tags within a reader identification range respond to reader s command after Pluralities tags within a reader identification range respond to reader s command after receiving reader's command. When two or more tags send data signals to reader at same receiving reader's command. When two or more tags send data signals to reader at same time, data signals interfere with each or in wireless channel. At this time, reader time, data signals interfere with each or in wireless channel. At this time, reader cannot cannot correctly receive data signals se tags, causing communication failure. correctly receive data signals se tags, causing communication failure Tag-Reader Tag-Reader Collision Collision Tag-reader Tag-reader collision collision occurs occurs in in a multiple a multiple readers readers identification identification scenario, scenario, as as shown shown in in Figure Figure 2. 2.

3 Information 2018, 2018, 9, x 9, FOR 63 PEER REVIEW Information 2018, 9, x FOR PEER REVIEW 3 17 Figure 2. Tag-reader collision diagram. Figure Figure 2. Tag-reader 2. Tag-reader collision collision diagram. diagram. Tag-reader collision means that when a tag is in identification range two or more readers at Tag-reader same collision means that when a tag is in identification range two or more readers Tag-reader time, collision multiple means readers that will when send a tag a command is in identification to a tag and range tag two fails or to more process readers at at signal same time, multiple readers will send a command to a tag and tag fails to process same after time, it receives multiple signals readers from will multiple send a command readers simultaneously. to a tag and Tag tag 2 fails in Figure to process 2 is in signal signal identification after it receives signals from multiple readers simultaneously. Tag 2 in Figure 2 is in after it receives range signals Reader from multiple 1 and Reader readers2 simultaneously. at same time, Tagand 2 incannot Figure 2respond is in to identification eir m identification range Reader 1 and Reader 2 at same time, and cannot respond to eir range when Reader both signals 1 andare Reader received 2 at at same same time, time. andas cannot a result, respond Readers to 1 eir and 2 cannot m when read both m Tag when both signals are received at same time. As a result, Readers 1 and 2 cannot read signals 2 information. are received at same time. As a result, Readers 1 and 2 cannot read Tag 2 information. Tag 2 information Reader-Reader Collision 2.3. Reader-Reader Collision Readers-reader collision refers refers to to fact fact that that when when one one tag tag is within is recognition range range two two or Readers-reader more more readers collision simultaneously, refers to one one or fact more that more readers when one cannot tag is receive within tag recognition tag signal signal due range due to to frequency two or more interference readers simultaneously, among multiple one readers. or more readers cannot receive tag signal due to frequency In In industrial interference production, among commodity multiple readers. storage, and and or or RFID RFID applications, it is itnecessary is to to fix fix multiple In industrial readers in production, inone onespace commodity to torm a small storage, identification and or RFID area network applications, to ensure it is necessary that objects to in fix in this this multiple space can readers can be be in identified. one space At this At to this form time, time, a small identification recognition areas areas area multiple network multiple readers to ensure readers overlap that overlap each objects or each in to this or form space to interference, form can interference, be identified. as shown as At in shown this Figure time, in 3. Figure Due recognition to3. Due restriction to areas restriction multiple radi frequency readers radio overlap identification frequency each or identification method, to form method, readers interference, thatreaders are as not shown that covered are in Figure not by covered 3. identification Due by to identification area restriction also interfere area also with radio interfere signals frequency with due to identification signals electromagnetic due to method, electromagnetic waves, readers and that waves, more are readers not and covered are arranged, more by readers identification greater are arranged, area probability also interfere greater collisions with signals probability between due readers. to collisions electromagnetic between readers. waves, and more readers are arranged, greater probability collisions between readers. Figure Figure 3. Reader-reader 3. Reader-reader collision collision diagram. diagram. Figure 3. Reader-reader collision diagram. This paper mainly focuses on tag-tag collision problems. Based on deep research into a large number This This paper related paper mainly mainly algorithms, focuses focuses a on new on tag-tag tag-tag multi-tag collision collision collision problems. problems. avoidance Based Based algorithm on on deep deep research is research proposed, into into a which large a large number aims number to provide related related reference algorithms, algorithms, and asupport new new multi-tag for collision collision application avoidance avoidance RFID algorithm algorithm in multi-tag is is proposed, proposed, identification which which aims aims research to provide to through provide reference algorithm reference andory and support support research. for for application application RFID RFID in multi-tag in multi-tag identification identification research research through through algorithm algorithm ory ory research. research.

4 Information 2018, 2018, 9, x 9, FOR 63 PEER REVIEW Array ArrayStorage Scheme Description Description Array Array Storage Storage Scheme Scheme In InQT QTand andct algorithms, a stack stack can can be be adopted adopted to store store prefixes prefixes in in identification identification process process tags. Intags. initial In recognition, initial recognition, reader presses reader an presses emptyan string empty into string stack. into If stack. stackif is not stack empty, is not a binary empty, string a binary is popped string as is popped a query as prefix a query and prefix reader and sends reader prefix sends and waits prefix for and waits response for response tags. tags. In In array arraystorage storagescheme, scheme, tag tagside sideretains no nomemory feature feature CT CTalgorithm and anddoes does not not require require any any setting setting tags. tags. After After receiving receiving reader s reader s query query command, command, tag s IDs tag s are IDs compared compared to received to received prefix. If prefix. front If parts front parts ID numbers ID match numbers match prefix, prefix, tags send tagsremaining send numbers remainingto numbers reader; toif not, reader; tags if do not, nothing. tagsunlike do nothing. stack Unlike structure, stack reader structure, sets two reader arrays, sets as shown two arrays, in Figure as shown 4. One in Figure m, 4. called One Probe_array[], m, called Probe_array[], is used to record is used to collision record information collision information label during label each during round each recognition. round recognition. Anor array, Anor Prefix[], array, is used Prefix[], to store is used query to store prefixes. query If prefixes. ID length If IDtag length to be identified tagis ton, be identified length both is n, arrays length is set to both m, arrays which is set means to m, which re means are two rearrays, are two arrays, Probe_array[0 n-1] Probe_array[0. and.. n-1] Prefix[0 n-1]. and... For n-1]. Probe_array[0 n-1], For. a.. collision n-1], a collision represented is represented with with value 1, and value 0 indicates 1, and 0no indicates collision. nobecause collision. all Because bits array all bits Probe_array Probe_array are 0, all states are 0, all collision states are collision eliminated. are eliminated. In initialization In initialization phase phase identification, identification, all bits all array bitsprobe_array Probe_array are set to are 0. set In to 0. identification In identification phase, phase, assuming assuming that that reader reader sends sends a query a query prefix prefix and tags andreturning tags returning ID message, ID message, collision collision bits are bitsindicated are indicated by c. by The c. reader The reader will will set set Probe_array[c] Probe_array[c] to 1, toand 1, and n n update update prefix prefix by by adding adding one one bit, bit, that that is is Prefix Prefix [0 [0... c-1] c-1] The The reader reader sends sends a prefix a prefix to tags tags to to query query left left branch branchnode; node; if if left left branch branch nodes nodes are aresuccessfully successfullyrecognized, Probe_array[c] Probe_array[c] will will be be set set to to0, 0, which whichmeans means that that collision here herehas has been beeneliminated, eliminated, and andprefix prefixprefix[ c-1] c-1] 1 is 1 is used usedto to query query right right branch branchnodes. nodes. Figure Array Array storage diagram Example Example and and Comparison Comparison Schemes Schemes In Inorder order to to illustrate illustrate implementation implementation steps steps array array storage storage scheme scheme more more intuitively, intuitively, stack stack in inct algorithm CT algorithm is replaced is replaced by two by two arrays. arrays. Assume Assume that that three three labels, labels, A, A, B, B, and and C, C, are are to tobe be identified, identified, with with IDs A: ; B: B: ; ; and and C: C: The The identification identification process process is shown is shown in Table in Table The The symbol x indicates highest highest collision collision bit, bit, and and because because query query prefix prefix CT algorithm CT algorithm is updated is only updated one bit only at aone time, bit at decoding a time, state decoding behind state highest behind collision highest bit iscollision represented bit as is -. represented At beginning as -. At identification, beginning all identification, elements in all Probe_array[] elements in are Probe_array[] set to 0. First, are set reader to 0. First, sends an reader emptysends string an (ε), empty and all string tags (ε), respond and all to tags reader s respond identification to reader s range. identification Secondly, range. reader Secondly, finds highest reader collision finds bithighest first, and collision Probe_array[0] bit first, = and 1. The Probe_array[0] query prefix= is 1. updated The query one prefix bit later, is updated and becomes one bit 0. later, The reader and becomes sends 0. The prefix reader with asends unique tag prefix for B s with response, a unique and tag Tag for BB s is successfully response, and identified. Tag B is Then, successfully prefixidentified. becomes 1Then, and Probe_array prefix becomes [0] = 0. The 1 and reader Probe_array sends [0] prefix = 0. again, The reader and sends decoded prefix state again, is 1. and Thedecoded reader knows state is that highest The reader collision knows bit is that in second highest bit, collision and set bit is in second bit, and set Probe_array [1] = 1. The query prefix becomes 10. The reader sends

5 Information 2018, 9, Probe_array [1] = 1. The query prefix becomes 10. The reader sends prefix, no collision occurs, and tag C is successfully identified. Then, prefix becomes 11, Probe_array [1] = 0 and Tag A is successfully identified. At this time, all elements in Probe_array[] are 0, which indicates that all collisions have been eliminated and all tags are identified. Table 1. The execution array scheme. Query Cycles Prefix 1 ε Probe_Array[0... 7] Reader s Decoding Status Response Tags Identification Status collision Tag B is identified collision Tag C is identified Tag A is identified The CT algorithm requires 16 bits to identify three tags using array, and stack storage takes 22 bits. According to EPC GEN2 standard, length tags ID is set to 96 bits, and IDs are taken in random number mode. By using MATLAB tools, average number bits per cycle required was simulated with tag numbers ranging from 100 to 600 with steps 100 in Table 2. The decimal number in Table 2 represents ratio total number bits produced to identification cycles. It can be seen that advantage array storage scheme in storage space is more obvious in a large-scale tag environment. Table 2. Storage scheme comparison. Number Tags Stack (bits) Array (bits) Tag Encoding Scheme One effective ways improving recognition efficiency tree anti-collision algorithm is using prefix updating collision bit according to collision decoding situation. In process each response tag, selecting first m bits tags as query prefix to encode m-bit IDs into encoded bits M (M = 2 m ) bits is an effective method. Since IDs are also encoded by 0 and 1, and re is only one 1 in each code M bits, position each bit is different. According to such a characteristic, when a collision occurs, a query prefix can be generated based on characteristic encoding to achieve purpose multi-bit identification. Tables 3 and 4 list codes corresponding to m = 2 and m = 3 respectively.

6 For implementation algorithm under different states reader, paper defines reader commands as follows. Broadcast instruction: reader will broadcast prefix to tags. In initial state, prefix is an empty string. Update instruction: updating value Probe_array[] according to value CID. Information 2018, 9, Table 3. m-m code (m = 2). Prefix Bits (m = 2) Encoded Bits (M = 4) Information 2018, 9, x FOR PEER REVIEW Table Table m-m m-m code code (m (m = 3). 3). Prefix Bits (m =3) Encoded Bits (M=8) Prefix Bits (m = 3) 000 Encoded Bits (M = 8) The coding scheme can can be be applied in in CT CT algorithm. In this In this way, way, reader reader updates updates query prefix query according prefix according to previous to previous M-bit collision M-bit collision condition. condition. The query The prefix query canprefix be updated can be from updated 1 bit from to m bits 1 bit in each to m collision bits in cycle, each and collision total cycle, queryand cycles will total be query reduced. cycles The identification will be reduced. process The is identification shown in Figure process 5. is shown in Figure 5. Figure 5. Identification process based on coding. The first bits tag s ID are encoded and sent to reader. After reader receives The first m bits a tag s ID are encoded and sent to reader. After reader receives label code, it decodes collision previous M bits. The M bits binary string is composed label code, it decodes collision previous M bits. The M bits binary string is composed characteristic ID (CID), and reader checks encoding list stored in reader to obtain characteristic ID (CID), and reader checks encoding list stored in reader to obtain corresponding collision bit ID. The reader updates query prefix according to collision bit ID, corresponding collision bit ID. The reader updates query prefix according to collision bit ID, pushes it into stack, and sends query prefix to label. The loop continues until all tags have pushes it into stack, and sends a query prefix to label. The loop continues until all tags have been identified. been identified. 5. An An Anti-Collision Algorithm Based on an Array and Encoding Scheme Based on advantages using array storage to reduce reader s memory space and using coding scheme to to reduce number query, a multi-bit multi-bit identification algorithm is proposed in this section Algorithm Description

7 Information 2018, 9, Algorithm Description For implementation algorithm under different states reader, paper defines reader commands as follows. Broadcast instruction: reader will broadcast prefix to tags. In initial state, prefix is an empty string. Update instruction: updating value Probe_array[] according to value CID. Read instruction: reader reads ID information tags according to instructions. Sleep instruction: after reader sends instructions to a tag, tag does not respond to instructions reader. The encoding scheme is stored using a two-dimensional character array, as shown in Table 5. Since coding scheme converts ID string per m bit into a binary string M bits, reader identifies received tag IDs in an identification interval each length M. Due to characteristics code, reader will generate prefixes based on number collisions in CID. Table 5. The code storage reader-side (m = 2). Str[0] 0 0 Str[1] 0 1 Str[2] 1 0 Str[3] 1 1 Each collision location can be recorded to obtain query prefix based on corresponding instructions for each collision position. If reader decoder is 0, n CID is 1101, and collision occurred at position 0, 2, and 3. The reader sends Request (0, Str), Request (2, Str), and Request (3, Str) to itself to obtain encodings 00, 10, and 11 in 2-d array. Subsequently, reader will update prefixes based on obtained encodings. The algorithm flow chart reader side is described in Figure 6, and both identification steps reader side and tag side are described below. Reader-side identification steps: (1) In initial stage query, reader sends an empty string to tags and all tags within identification range respond to it. (2) The reader takes M bit as identification interval, replies to received M-bits IDs, and decodes remaining ID information. If M-bits IDs do not collide and remaining bits have collisions, query prefix is updated and query continues. If only unique tag replies to its own ID, tag is identified. Reader reads information tag, and n sends Sleep instruction to keep tag in sleep mode and skip to step 5. If re is a collision, it skips to step 3. (3) According to decoding results, bit collision is set to 1, and bit with no collision is marked as 0. The reader get C i (0 I < M) according to position 1 in CID, pushes C i into stack, and sets m-bits value Probe_array [0...n] to a, where a represents number 1 in CID. (4) Extract an element C i from stack, and get Str[C i ] according to two-dimensional character array. Meanwhile, corresponding position 'a' in Probe_array[] is decreased by 1, and let Prefix = Prefix+Str[C i ]. If a in current identification section is changed to 0, it continues to extract an element C i from stack, and let Prefix = Prefix[0,C']+Str[C i ], where, C' is last collision position. The reader broadcasts prefix to tags; skip to step 2. (5) If all elements in Probe_array[] are 0, it means that all tags have been correctly identified and identification process is finished. Tag-side identification steps: After receiving query prefix, first m-bit IDs tags matching query prefix are encoded into M-bit data, and remaining IDs are returned in addition to query prefix.

8 Information 2018, 9, Information 2018, 9, x FOR PEER REVIEW 8 17 Figure Figure 6. Algorithm 6. Algorithm flow flow chart chart reader reader side. side Example Algorithm 5.2. Example Algorithm The execution steps multi-bit collision tree algorithm are described as follows. Here, m = 2 The execution steps multi-bit collision tree algorithm are described as follows. Here, m = 2 in m-m bits scheme, which means first two bits tag response part are encoded. Suppose in m-m bits scheme, which means first two bits tag response part are encoded. Suppose that six tags are A, B, C, D, E, and F, as shown in Table 6. Their IDs are A: ; B: ; that six tags are A, B, C, D, E, and F, as shown in Table 6. Their IDs are A: ; B: ; C: ; D: ; E: ; F: C: ; D: ; E: ; F: It can be seen from Table 6 that reader sends an empty string ε in first step, and all It can be seen from Table 6 that reader sends an empty string ε in first step, and all tags in identification range respond and convert first two ir ID numbers into four-digit tags in identification range respond and convert first two ir ID numbers into four-digit codes and send m to reader. The reader obtains CID value as 1111 after decoding. There codes and send m to reader. The reader obtains CID value as 1111 after decoding. There are are four collision bits, so first two bits in Probe_array[] array are set to 04. At same time, four collision bits, so first two bits in Probe_array[] array are set to 04. At same time, reader gets values four Ci values 0, 1, 2, and 3 according to position 1 in CID reader gets values four C i values 0, 1, 2, and 3 according to position 1 in CID and pushes m into stack. An element 0 is popped from stack, and value 4 in and pushes m into stack. An element 0 is popped from stack, and value 4 in Probe_array[] is decreased by one. The reader searches two-dimensional array, obtains query Probe_array[] is decreased by one. The reader searches two-dimensional array, obtains query prefix 00, and sends it to tags. After obtaining query prefix, tag that matches query prefix 00, and sends it to tags. After obtaining query prefix, tag that matches query prefix prefix encodes first two digits ID except for prefix, and replies to remaining ID. The encodes first two digits ID except for prefix, and replies to remaining ID. The reader reader decodes ID information tags that replied to obtain CID = 1101, which indicates that decodes ID information tags that replied to obtain CID = 1101, which indicates that re are re are three collision bits, and corresponding two bits in Probe_array[] are set to 03. The three collision bits, and corresponding two bits in Probe_array[] are set to 03. The reader obtains reader obtains three values 0, 2, and 3 according to position 1 in CID, and pushes m three values 0, 2, and 3 according to position 1 in CID, and pushes m into stack. into stack. In third query cycle, reader gets query prefix 0000, sends prefix to tags, In third query cycle, reader gets query prefix 0000, sends prefix to tags, and value and value 3 in Probe_array[] is decreased by one. Only a unique tag responds, so tag A is 3 in Probe_array[] is decreased by one. Only a unique tag responds, so tag A is identified. The reader identified. The reader sends a sleep instruction, and tag A is in silence. Continue to pop "2" from sends a sleep instruction, and tag A is in silence. Continue to pop 2 from stack; reader stack; reader accordingly gets query prefix 0010 and broadcasts it. Simultaneously, accordingly gets query prefix 0010 and broadcasts it. Simultaneously, reader changes value reader changes value corresponding position in Probe_array[], by which label B is corresponding position in Probe_array[], by which label B is correctly identified, and correctly identified, and tag C is identified in this step. In this case, value second m bits tag C is identified in this step. In this case, value second m bits in Probe_array[] has been in Probe_array[] has been changed to 00, which indicates that collision at this position has changed to 00, which indicates that collision at this position has been eliminated and returned been eliminated and returned to previous m bits. An element is popped from stack, and to previous m bits. An element is popped from stack, and Probe_array[] will change its value Probe_array[] will change its value 03 to 02. The reader obtains query prefix and sends it to tags. Follow steps described above, tags D, E, and F are identified. Finally, values all positions in Probe_array[] are 0.

9 Information 2018, 9, to 02. The reader obtains query prefix and sends it to tags. Follow steps described above, tags D, E, and F are identified. Finally, values all positions in Probe_array[] are 0. Table 6. MICT algorithm identification table. Query Cycles 1 ε Prefix Probe_Array[0... 7] Decoding State Response Tags Stack ,1,2, ,2,3,1,2,3 Tag A is identified 2,3,1,2,3 Tag B is identified 3,1,2,3 Tag C is identified 1,2,3 Tag D is identified 2,3 Tag E is identified 3 Tag F is identified NULL 5.3. Analysis Algorithm Performance Time Complexity Analysis Time complexity refers to number query response cycles required to complete identification all labels in a collection tags. Suppose that number identified tags is n; re are only collision nodes and identification nodes in process establishing M-tree. The total number nodes is expressed as follows: N = n c + n id (1) where n c is collision nodes and n id indicates identification nodes. Each time a collision occurs, collision nodes generate branch nodes accordingly, and identified tag corresponds to nodes whose degree is 0. We reby obtain following equation: n id = n (2) In worst case, this algorithm implements identification process n tags to form an overriding tree. According to m-m bits scheme, if intermediate nodes collide, number query prefixes can be 2, 3,..., m..., M. The collision tree may be a full 2-ary tree, a full 3-ary tree, a full M-ary tree, or a mixed tree 2 to 2m, so number nodes required for different limits is calculated here. When number query prefixes generated by collision is 2, which means whole collision tree is a full binary tree, time slot derivation process according to CT algorithm can be obtained as follows: N = 2n 1 (3) When number query prefixes generated by collision is 3, that is, when whole collision tree is a full trinary tree, nodes in tree have only nodes with a degree 3 and a degree 0 and a root node. On or hand, suppose that number nodes with degree 0 is N 0

10 Information 2018, 9, and degree 3 is N 3, n number collision nodes n c = N 3, number identified nodes n id = N 0. The total number nodes is as follows: N = N 3 + N 0 (4) Let B be total number branches present in a complete trinary tree; because any node or than root node in tree corresponds to a branch, total number nodes in a complete trinary tree can be represented by total number branches, as follows: N = B + 1 (5) Each intermediate node in tree has three branches, and lowest child node has no branch, so number branches B can be expressed as follows: From Equations (5) and (6), following equation can be obtained: B = 3N 3 (6) N = 3N (7) From Equations (4) and (7), equation can be obtained as follows: N 3 = (N 0 1)/2 (8) Substituting Equation (8) into Equation (4), following equation can be obtained: N = (3N 0 1)/2 (9) Because nodes degree 0 are bottom-most identifying nodes in tree, combined with Equation (2), following equation can be derived: N = (3n 1)/2 (10) According to above derivation, number nodes complete M-ary tree can be obtained as follows: N = (mn 1)/(m 1) (11) where m is number coded bits in m-m coding scheme. Each time a collision occurs, collision node may generate 2 to 2 m prefixes, so average number nodes is used to represent time complexity algorithm. T(n) is time complexity, expressed as follows: T(n) = N = 2 m (mn 1)/(m 1) m=2 2 m 1 (12) Think Equation (11) as a function m, and function can be expressed as follows: f (m) = (mn 1)/(m 1) (13) Derive Equation (13) to get Equation (14): f (m) = (1 n)/(m 1) 2 (14)

11 Information 2018, 9, Since n is greater than 1 and m is greater than or equal to 2, derivative f (m) is always less than 0, which indicates that value f (m) decreases as m increases. When n is equal to 2, N = 2n 1. As n increases, value N is less than 2n-1, and following inequality is established: N 2n 1 (15) Therefore, it is proved that proposed MICT algorithm is superior to CT algorithm in terms time complexity Communication Complexity Analysis Communication complexity refers to amount data transmitted between reader and tag in RFID tag anti-collision algorithm. The communication complexity includes both amount data sent by reader, which is called reader-side communication complexity, and tag side binary number replying, which is called tag-side communication complexity. In RFID system, communication complexity also represents energy consumption. Usually, greater amount data transmitted, higher energy consumption system. Let C(n) be communication complexity MICT algorithm, C R (n) is reader-side communication complexity, C T (n) is tag-side communication complexity, and y satisfy following relationship: C(n) = C T (n) + C R (n) (16) Let l com be length instruction sent by reader. l pre,i is length query prefix in i th cycle, and l rep,i is length binary string sent by tags. Equation (16) can be expressed as follows: C(n) = T(n) (l com + l pre,i ) + i=1 T(n) l pre,i (17) i=1 Since first m bits are encoded each time when tag is returned to ID message. The tag-side communication complexity can be changed as follows: C R (n) = T(n) T(n) l pre,i = i=1 i=1 l rep,i + 2 m m (18) The length l rep,i is equal to length tag s ID minus length l pre,i, and can be expressed by following equation: l ID = l pre,i + l rep,i (19) Therefore, communication complexity MICT algorithm can be obtained via Equations (12) and (17) (19), and is expressed as follows: C(n) = 2 m m= Identification Efficiency Analysis (mn 1)/(m 1) 2 m 1 (l com + l ID + 2 m m) (20) The identification efficiency is also called throughput, which refers to ratio number tags identified to query cycles required to complete identification se tags. Let E MICT (n) be identification efficiency MICT algorithm. The identification efficiency MICT algorithm can be obtained as follows: E MICT (n) = n (21) T(n)

12 Information 2018, 9, Substituting Equation (12) into above equation yields following equation: E MICT (n) = n(2 m 1) 2 m [(mn 1)/(m 1)] m=2 (22) Table 7 lists value identification efficiency under different values m when number tags varies from 100 to Table 7. The oretical value identification efficiency. The Value m Number Tags % 62.11% 62.07% 62.07% % 72.99% 72.99% 72.97% % 81.97% 81.90% 81.89% % 88.50% 88.50% 88.51% % 93.11% 93.02% 93.02% It can be seen from Table 7 that minimum identification efficiency MICT algorithm is maintained at more than 60%, and when value m is greater than 3, identification efficiency is more than 80%. If value m is greater than 6, length after encoding will be greater than length tags. This increases amount data transmitted. Therefore, value m varies from 2 to 6. From oretical analysis in Section 5.3.1, we can see that time complexity MICT algorithm decreases with increase in m. From definition identification efficiency, it can be seen that when number tags is certain, lower time complexity and higher identification efficiency. So, from oretical analysis, we can see that when m = 6, MICT algorithm achieves optimal identification efficiency. As number tags increases, identification efficiency is stable around 93%, even if number tags reaches tens thousands Reader-Side Memory Complexity Analysis Reader-side memory complexity refers to number bits binary data that reader reader needs to store during RFID tag identification. Taking CT algorithm and proposed MICT algorithm as examples, memory complexity reader side is analyzed as follows. Let S(n) be reader-side memory complexity. As described in Sections and 5.3.2, n represents number identified tags, l ID is number identified tags, l com is length instruction sent by reader, and l pre,i is length query prefix in i th cycle. The CT algorithm uses a stack to store newly generated prefix in tag identification process and reader sends prefix and waits for tag s response. Using S CT (n) to represent reader-side memory complexity CT algorithm, it can be expressed as in Equation (23): T(n) S CT (n) = l com + i=1 l pre,i (23) where T(n) is time complexity CT algorithm and value l com is fixed. According to query and response process CT algorithm, it is necessary for reader to update value query prefix according to decoding status in each query cycle. Therefore, in different query-response periods, value l pre,i will constantly change, and its range is 0 l pre,i < m. In MICT algorithm, set up two arrays, one for recording collision, or to store query prefix as long as ID tag to determine length array space is also determined.

13 Information 2018, 9, Using S MICT (n) to represent reader-side complexity CT algorithm, it can be expressed as in Equation (24): S MICT (n) = l com + 2l ID. (24) When number tags is large, value T(n) i=1 l pre,i is much larger than value 2l ID, so in case large-scale tag identification, MICT algorithm is more dominant than CT algorithm in memory complexity reader-side. The Table 8 lists reader-side memory complexity for several existing tree-based anti-collision algorithms. Information 2018, 9, x FOR PEER REVIEW Table 8. Comparison reader-side memory complexity. algorithm in memory complexity reader-side. The Table 8 lists reader-side memory complexity for several Algorithm existing tree-based anti-collision Reader-Side algorithms. Memory Complexity BS l com + [n + log 2 (n!)] l ID [22] Table 8. Comparison reader-side memory complexity. QT l com + n (l ID+2 lgn) Algorithm Reader-Side Memory l Complexity pre,i ( worst situation) [22] BS l i=1 com + [ n + log 2 ( n! )] lid [22] n*( l CT l com + 2n 1 ID + 2 lg n) l QT l + pre,i com l pre, i ( worst situation) [22] i=1 MICT CT i= 1 nl com 1 + 2l ID 2 com + l pre, i i= 1 l It can be seen from MICT Table 8 that reader-side l memory complexity BS, QT, and CT com + 2l ID algorithms is influenced by both number tags and length tag. In MICT algorithm, It can be seen from Table 8 that reader-side memory complexity BS, QT, and CT memory complexity reader is not affected by time complexity. Therefore, compared with algorithms is influenced by both number tags and length tag. In MICT stack storage schemes, an array storage scheme can effectively reduce amount data stored at algorithm, memory complexity reader is not affected by time complexity. Therefore, readercompared side, andwith stack amount storage data schemes, does an not array increase storage scheme with can increase effectively inreduce number amount tags. data stored at reader side, and amount data does not increase with increase in number 6. Algorithm tags. Simulations and Analysis The 6. Algorithm MICT algorithm Simulations and and several Analysis or typical tree anti-collision algorithms are simulated using MATLAB stware. The testing simulates identification one reader and multiple tags. The MICT algorithm and several or typical tree anti-collision algorithms are simulated using The number labels range from 100 to 600. The step length is 100, and tags IDs are 96 bits. Each data MATLAB stware. The testing simulates identification one reader and multiple tags. The point innumber Figures labels 7 12 is range from average 100 to value 600. The obtained step length afteris , rounds and tags experiment. IDs are 96 bits. TheEach paper data obtains m value point in Figures MICT7 12 algorithm is average with value best obtained performance. after 50 rounds Furrmore, experiment. The experimental paper obtains results MICT algorithm m value are compared MICT algorithm to with oretical best performance. value. Finally, Furrmore, proposed experimental algorithmresults is compared to existing classical MICT algorithm tree anti-collision are compared algorithms to oretical with value. time Finally, complexity, proposed communication algorithm complexity, is compared to existing classical tree anti-collision algorithms with time complexity, and recognition efficiency. communication complexity, and recognition efficiency. 3.0 log(number query cycles) m = 2 m = 3 m = 4 m = 5 m = number tags (10 2 ) Figure Comparison query cycles. Figure 7. Comparison query cycles. It can be seen from Figure 7 that maximum number query cycles is required when m is equal to 2, that is, time complexity is largest. With increase in m, time complexity is obviously reduced. When m = 6, number queries required in identification process is lowest. The complexity communication reflects energy consumption system. The less data transferred between reader and tags, lower energy consumption system. The

14 Information 2018, 9, It can be seen from Figure 7 that maximum number query cycles is required when m is equal to 2, that is, time complexity is largest. With increase in m, time complexity is obviously reduced. When m = 6, number queries required in identification process is lowest. The complexity communication reflects energy consumption system. The less data transferred between reader and tags, lower energy consumption system. The MICT Information 2018, 9, x FOR PEER REVIEW algorithm uses m-m encoding scheme and length tag is 96 bits. It can be seen from Figure 8 thatmict communication algorithm uses m-m complexity encoding is scheme largest and when length m = 6. tag When is 96 bits. m = It 4, can be communication seen Information 2018, 9, x FOR PEER REVIEW complexity from is Figure smallest, 8 that andcommunication with increasing complexity is m, largest communication when m = 6. When complexitym = 4, gradually communication complexity is smallest, and with increasing m, communication increases, MICT whichalgorithm is related uses to m-m coding encoding program. scheme and length tag is 96 bits. It can be seen complexity gradually increases, which is related to coding program. from Figure 8 that communication complexity is largest when m = 6. When m = 4, communication complexity 5.2 is smallest, and with increasing m, communication complexity gradually increases, which is related to coding program. log(number log(number transmission transmission bits) bits) m = 2 m = 3 m = 4 m = 5 m = = 6 2 m = 3 m = 4 m = m = 6 6 number tags (10 2 ) Figure 1 8. Comparison 2 3 data 4transmission. 5 6 number tags (10 2 ) Figure 9 shows number query cycles in experimental simulation compared to oretical analysis values. When Figure m 8. = Comparison 2, it can be seen data from transmission. Figure 9 that two lines almost coincide, which indicates that experimental values are highly consistent with oretical values; Figure 9 average shows deviation number is just 0.6%. query Due cycles to in unpredictable experimental specific simulation collision compared different tags, to oretical oretical analysis derivation values. When time complexity m = 2, it can can be only seen be from expressed Figure as an 9 average that value. two In lines almost time coincide, complexity which expression indicates given that in Section experimental 5.3.1, set values values are highly m is C{2,3,... consistent 2 m }; with when oretical value m increases from 2 to 6, number elements in set increases exponentially. At same time, values; average deviation is just 0.6%. Due to unpredictable specific collision different tags, actual collision is affected by m value relatively small, from updating 2-bit prefix to oretical derivation time complexity can only be expressed as an average value. In time 6-bit prefix, so error actual experimental results obtained with oretical analysis complexity expression given in Section 5.3.1, set values m is C{2,3,... 2 value increases with m. It is for this reason that as value m increases, deviation m }; when value between m increases from 2 to 6, number elements in set increases exponentially. At same time, experimental values and oretical values is different. When m = 4, average deviation actual collision is affected by m value relatively small, from updating 2-bit prefix to experimental values and oretical values is about 10%, and when m = 6, average deviation 6-bit prefix, so error actual experimental results obtained with oretical analysis is about 13%. When m increases from 2 to 6, average error between experimental and value oretical increases values with is m. about It is 8%. for this reason that as value m increases, deviation between experimental values and oretical values is different. When m = 4, average deviation experimental values and oretical values is about 10%, and when m = 6, average deviation is about 13%. When m increases 3.0 from 2 to 6, average error between experimental and oretical values is about 8%. log(number log(number query cycles) query cycles) Figure Comparison data transmission. Figure 9 shows number query cycles in experimental simulation compared to oretical analysis values. When m = 2, it can be seen from Figure 9 that two lines almost coincide, which indicates that experimental values are highly consistent with oretical values; average deviation is just 0.6%. Due to unpredictable specific collision different tags, oretical derivation time complexity can only be expressed as an average value. In time complexity expression given in Section 5.3.1, set values m is C{2,3,... 2 m }; when value m increases from 2 to 6, number elements in set increases exponentially. At same time, actual collision is affected by m value relatively small, from updating 2-bit prefix to 6-bit prefix, so error actual experimental results obtained with oretical analysis value increases with m. It is for this reason that as value m increases, deviation between experimental values and oretical values is different. When m = 4, average deviation experimental values and oretical values is about 10%, and when m = 6, average deviation is about 13%. When m increases from 2 to 6, average error between experimental and oretical values is about 8% m = 2 (analysis) m = 2 (simulation) m = 3 (analysis) m = 3 (simulation) m = 4 (analysis) m = 4 (simulation) m = 52 (analysis) m = 52 (simulation) m = 3 (analysis) m = 63 (simulation) m = 4 (analysis) m = 4 (simulation) 2.2 number tags (10 2 ) m = 5 (analysis) m = 5 (simulation) Figure Comparison simulation and analysis m = 6 (analysis) values. m = 6 (simulation) number tags (10 2 ) Figure Figure 9. Comparison 9. simulation and and analysis values. values.

15 Information 2018, 9, Information 2018, 9, x FOR PEER REVIEW Information 2018, 9, x FOR PEER REVIEW Information 2018, 9, x FOR PEER REVIEW log(number log(number query query query cycles) cycles) log(number log(number transmission bits) bits) bits) identification efficiency efficiency (%) (%) (%) BS BS QT QT CT BS CT OBS-MCBR QT OBS-MCBR MICT (m = 4) MICT OBS-MCBR (m = 4) 6) MICT (m = 6) 4) MICT (m = 6) number tags ( ) number tags ( ) number tags (10 2 ) Figure Comparison query cycles. Figure 10. Comparison query cycles. Figure 10. Comparison query cycles BS 4.8 BS QT QT CT BS 4.6 CT OBS-MCBR QT OBS-MCBR MICT (m = 4) 4.4 MICT OBS-MCBR (m = 4) MICT (m = 4) number tags ( ) number tags ( ) Figure 11. Comparison number tags data (10 transmission. 2 ) Figure Figure 11. Comparison data transmission. Figure Comparison Comparison data data transmission. transmission BS 30 BS QT 30 QT CT BS 30 CT OBS-MCBR QT MICT (m = 4) 20 OBS-MCBR MICT OBS-MCBR (m = 4) 20 MICT (m = 4) number tags ( ) number tags ( ) number tags (10 2 ) Figure 12. Comparison identification efficiency. Figure 12. Comparison identification efficiency. Figure Comparison identification efficiency. Figures 10 and 12 demonstrate comparison MICT algorithms with BS, QT, CT, and Figures 10 and 12 demonstrate comparison MICT algorithms with BS, QT, CT, and OBS-MCBR Figures algorithms 10 and 12 demonstrate in terms time complexity, comparison communication MICT algorithms complexity, with and BS, identification QT, CT, and OBS-MCBR algorithms in terms time complexity, communication complexity, and identification efficiency. OBS-MCBR It algorithms can be seen in from terms Figure time complexity, 10 that BS communication algorithm needs complexity, most query and identification cycles when efficiency. It can be seen from Figure 10 that BS algorithm needs most query cycles when efficiency. It can be seen from Figure 10 that BS algorithm needs most query cycles when

16 Information 2018, 9, Figures 10 and 12 demonstrate comparison MICT algorithms with BS, QT, CT, and OBS-MCBR algorithms in terms time complexity, communication complexity, and identification efficiency. It can be seen from Figure 10 that BS algorithm needs most query cycles when number tags increases, and or four algorithms are better than BS algorithm. The MICT algorithm performs best when m = 6. When identifying 500 tags, BS algorithm needs more than 2800 query cycles, QT algorithm needs more than 1200 queries, CT algorithm needs about 1000 times, OBS-MCBR algorithm needs about 700 query cycles, and proposed MICT algorithm needs about 600 query cycles. MICT algorithm s time complexity is reduced by 53%, 38%, and 12%, respectively, compared to QT, CT, and OBS-MCBR algorithms. When m = 4, performance MICT algorithm in terms time complexity is equivalent to that OBS-MCBR algorithm. In terms amount data transmitted, performance BS algorithm is worst, and performance MICT algorithm is best when m = 4. When identifying 500 tags, QT algorithm needs more than 150,000 bits, CT algorithm needs more than 130,000 bits, OBS-MCBR algorithm needs about 110,000 bits, and proposed MICT algorithm needs 9000 bits. Compared to QT, CT, and OBS-MCBR algorithms, communication complexity is reduced by 35%, 27%, and 8%, respectively. Although MICT algorithm has fewest queries when m = 6, amount data transmitted is also largest. The MICT algorithm saves 48 bits on average amount data transmitted by each identification tag when m = 4 than when m = 6. For two aspects (number query times and amount communication data), MICT selects m = 4 when identifying label 96 bits. In terms identification efficiency, MICT algorithm reaches 74% when m = 4, followed by OBS-MCBR algorithm, which reaches 72%, and recognition efficiency CT algorithm is about 51%. The computational burden anti-collision algorithm is related to number queries and amount communication data between readers and tags. When number queries and amount communication data is reduced, computational burden algorithm will decrease accordingly. From above experimental results, it can be seen that, compared to existing algorithms, MICT algorithm reduces number queries and amount communication data. Therefore, computational burden MICT algorithm is lowest among se algorithms. 7. Conclusions Based on CT algorithm, an anti-collision algorithm with multi-bit identification is proposed using array storage scheme and coding scheme. The proposed algorithm extends length prefix one bit in traditional tree-based algorithm to several bits in each query cycle. Through oretical analysis and experimental simulation, it is concluded that MICT algorithm has obvious advantages in terms time and communication complexity compared to several existing algorithms. Acknowledgments: This work was partially supported by Science & Technology Program Shaanxi Province under projects 2017GY-196 and 2015KTCXSF-10-11, and Opening Fund State and Local Engineering Laboratory Advanced Network and Monitoring Control under project GSYSJ Author Contributions: Baolong Liu and Xiaohao Su conceived and designed scheme; Baolong Liu analyzed efficiency; Xiaohao Su simulated system; Baolong Liu and Xiaohao Su wrote paper. Conflicts Interest: The authors declare no conflict interest. References 1. He, M.; Horng, S.-J. A fast RFID tag identification algorithm based on counter and stack. Expert Syst. Appl. 2011, 38, [CrossRef] 2. An, J.; Wu, J.X. Improved RFID binary search anti-collision algorithm. Comput. Eng. Appl. 2009, 45, Tsai, S.-C.; Min, Y. Efficient tag reading protocol for large-scale RFID systems with pre-reading. Comput. Commun. 2016, 88, [CrossRef] 4. Myung, J.; Lee, W. Tag-splitting: Adaptive collision arbitration protocols for RFID tag identification. IEEE Trans. Parallel Distrib. Syst. 2007, 18, [CrossRef]

17 Information 2018, 9, Jung, H. A memory efficient anti-collision protocol to identify memoryless RFID tags. J. Inf. Process. Syst. 2015, 11, Mohammed, U.S.; Salah, M. Tag anti-collision algorithm for RFID systems with minimum overhead information in identification process. Radio Eng. 2011, 20, Shin, J.; Jeon, B. Multiple RFID tags identification with M-ary query tree scheme. IEEE Commun. Lett. 2013, 17, [CrossRef] 8. Agrawal, T.; Biswas, P.K. An optimized query tree algorithm in RFID inventory tracking-a case study evidence. Int. J. Comput. Sci. 2012, 9, Kim, S.; Kim, Y. A tag prediction anti-collision algorithm using extra bits for RFID tag identification. Int. J. Ad Hoc Ubiquitous Comput. 2012, 10, [CrossRef] 10. Jiao, C.-H.; Wang, K.-R. Multi-branch query tree protocol for solving RFID tag collision problem. J. China Univ. Posts Telecommun. 2008, 15, [CrossRef] 11. Jia, X.L.; Feng, Q.Y. An efficient anti-collision protocol for RFID tag identification. IEEE Commun. Lett. 2010, 14, [CrossRef] 12. Safa, H.; El-Hajj, W. A distributed multi-channel reader anti-collision algorithm for RFID environments. Comput. Commun. 2015, 64, [CrossRef] 13. Lai, Y.C.; Hsiao, L.Y. An RFID anti-collision algorithm with dynamic condensation and ordering binary tree. Comput. Commun. 2013, 36, [CrossRef] 14. Klair, K.; Chin, W. A survey and tutorial RFID anti-collision protocols. IEEE Commun. Surv. Tutor. 2010, 12, [CrossRef] 15. Yan, X.Q.; Liu, Y. A memoryless binary query tree based Successive Scheme for passive RFID tag collision resolution. Inf. Fusion 2015, 22, [CrossRef] 16. Guo, H.B.; He, C. PSR: A novel high-efficiency and easy-to-implement parallel algorithm for anti-collision in RFID systems. IEEE Trans. Ind. Inform. 2016, 12, [CrossRef] 17. Zhang, Y.; Yang, F. An anti-collision algorithm for RFID-based robots based on dynamic grouping binary trees. Comput. Electr. Eng. 2018, in press. [CrossRef] 18. Ma, Y.M.; Jin, D. Anti-collision algorithm based on Multi-threaded RFID lock position. Int. J. Hybrid Inf. Technol. 2013, 6, Landaluce, H.; Perallos, A. Managing number tag bits transmitted in a bit-tracking RFID collision resolution protocol. Sensors 2014, 14, [CrossRef] [PubMed] 20. Jung, Y.; Kim, D. Optimized binary search with multiple collision bits resolution anti-collision algorithm for efficient RFID tag identification. IEICE Trans. Fundam. Electron. Commun. Comput. Sci. 2016, 99, [CrossRef] 21. Wang, Y.H.; Liu, Y. A multi-bit identification protocol for RFID tag reading. IEEE Sens. J. 2013, 13, [CrossRef] 22. Su, X.H.; Liu, B.L. An investigation on tree-based tags anti-collision algorithms in RFID. In Proceedings International Conference on Computer Network, Electronic and Automation (ICCNEA), Xi an, China, September 2017; pp by authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under terms and conditions Creative Commons Attribution (CC BY) license (