Research Article QSA: Query Splitting-Based …downloads.hindawi.com/journals/ijdsn/2013/674698.pdf4. QSA: Query Splitting-Based Anticollision for Mobile IoT.. Overview of the Anticollision

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2013 Article ID 674698 9 pageshttpdxdoiorg1011552013674698

Research ArticleQSA Query Splitting-Based Anticollision forMobile RFID-Based Internet-of-Things

Jianliang Gao Jianxin Wang Jianbiao He and Weiping Wang

School of Information Science and Engineering Central South University Changsha 410083 China

Correspondence should be addressed to Weiping Wang wpwangcsueducn

Received 28 March 2013 Accepted 2 July 2013

Academic Editor Lu Liu

Copyright copy 2013 Jianliang Gao et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Mobility is a common character of the emerging RFID-based internet-of-things However most of prior RFID anticollision algo-rithms ignore the movement of tags which can degrade the identification performance seriously and even result in tag starvationproblem This paper presents a novel anticollision algorithm named Query Splitting-based Anticollision (QSA) for mobile RFID-based internet-of-things By designing adaptive query QSA reduces the number of collisions efficiently and makes it possible toidentify multiple mobile tags without rollback In QSA we propose a query stack technology to avoid the rollback operation causedby new arriving tags which solves the tag starvation problem under mobile environments The performance evaluation shows thatthe proposed algorithm takes fewer timeslots and has better performance in identifying mobile tags

1 Introduction

Radio frequency identification (RFID) is a contactless wire-less communication technology The scope for using thistechnology boosts as RFID tag becomes a low-cost device dueto mass production [1] As the rapid proliferation of RFIDtags it has given rise to various concepts that integrate thephysical world with the virtual one One of the most popularconcepts is the Internet-of-Things (IoT) a vision in which theInternet extends into physical entities In the IoT RFID is thefoundation to connect the things together [2]

RFID systems consist of a reading device called readerand multiple tags which are attached to physical entities inthe IoT The reader is typically a powerful device with amplememory and computational resources On the other handtags vary significantly in their computational capabilitiesAmong tag types passive ones are emerging to be a popularchoice for large scale deployments due to their low cost [3]For passive tags they respond only at reader commands withthe energy provided by the reader [2 4]

In RFID systems the reader usually needs to communi-cate with multiple tags If there are multiple tags in a readerrsquosinterrogation zone the reader receives the responses fromthese tags simultaneously For a reader it is not able to distin-guish exact information from the interfered wireless signals

which is called collision Collision is a serious problem inRFID systems since the reader will not receive the messagesrightly once collision occurs [5] An example is shown inFigure 1 Simultaneous responses transmitted bymultiple tagscollide resulting in an increase of identification delay evenfailure of reading the tagsTherefore an efficient anticollisionalgorithm is required to reduce collisions and to achieve fastidentification

The tag collision problem becomes more serious inmobile RFID-based IoT The physical entities with tags areoftenmobile which facilitates competitive advantage throughbenefits such as improved efficiency increased visibilityreduced cost and many others [6] The mobile devices canbe part of numerous products gadgets and vehicle parts [7]These devices close the gap between the real word and its vir-tual representation via for example seamless identificationand integration with other wirelessly-embedded devices andtheir surroundings

However the mobile devices result in new challenges foranticollision problem One of the challenges is tag starvationFor example as shown in Figure 1 the reader is processing thecollision caused by Tag 1 to Tag 119873 Assuming the reader haspartitioned the tags for several rounds and will identify onetag out right now If Tag119873+1 enters the readerrsquos interrogationzone at this time the new arrived tagmight cause the rollback

2 International Journal of Distributed Sensor Networks

Reader

Tag NTag 1 Tag 2

Readerrsquosinterrogation zone

Antenna

middot middot middot

Tag N + 1

Collision

Figure 1 The tag collision with mobile tag

of the partition operations Therefore the done operationshave to be repeated and the delay of identifying the tagsincreases If the new tags enter the readerrsquos interrogation zoneat an interval that causes rollback repeatedly some tagsmightdepart the readerrsquos interrogation zone before being identifiedwhich is called ldquotag starvationrdquo in mobile environmentsTherefore it is necessary to provide efficient anticollision formobile RFID-based IoT

In this paper we will deal with the ldquotag starvationrdquo andperformance problem for anticollision in mobile IoT By thedesign of query stack and query rules the proposed anticol-lision algorithm avoids repetition operations and decreasesthe collisions consequently In summary our contributionsinclude the following

(i) We characterize the problem of anticollision inmobile environment which is more practical for theemerging IoT technology

(ii) We propose a novel anticollision algorithm namedQuery Splitting-basedAnticollision (QSA) formobileRFID-based IoT which solves the problem of tag star-vation and performance degradation resulted fromthe tag mobility

(iii) Simulation and analysis evaluate the efficiency ofthe proposed algorithm The results show the betterperformance of the proposed algorithm to the priormethods

The rest of this paper is organized as follows Section 2outlines preliminary includes background related workSection 3 presents the problems for mobile tags Section 4gives the detailed design of our algorithm The simulationresults are provided in Section 5 Finally Section 6 concludesthe paper

2 Preliminary

21 Collision Detection Code technologies are widely usedfor collision detection Manchester code is one of the mostpopular technologies for RFID systems [5] In Manchestercode the value of one bit is defined as the voltage transitionwithin a bit windowA bit ldquo0rdquo is coded by a positive transitionwhile a bit ldquo1rdquo is coded by a negative transition In RFID

systems if two (ormore) tags transmit different values simul-taneously the positive and negative transitions of the receivedbits cancel each other out This state is not permissible inManchester code during data transmission and is recognizedas an error Therefore Manchester code makes it possible toldquotrace a collision to an individual bitrdquo and ldquofind where thecollided bit isrdquo [8]

Figure 2 shows an example of Manchester code forcollision detection The IDs of tag 1 and tag 2 are ldquo10101100rdquoand ldquo10001001rdquo respectively When tags 1 and 2 send theirIDs simultaneously using Manchester code the decodeddata from the interfered signal received by the reader isldquo10119909011199090119909rdquo where ldquo119909rdquo represents a collided bit In thisexample the locations of the collided bits are the 3rd 6thand 8th bits This information helps the reader separate thecollided tags into subsets more smartly and identify the tagsmore quickly

Manchester code can be utilized to detect the collisionbits but the tags should be strictly synchronized Fortunatelythe tags in passive RFID systems are all driven by the readerwith both energy and the same clock frequency

22 Anticollision Algorithms Many researches have focusedon the issue of anticollision including tag-driven and readerdriven procedures [9] Tag-driven anticollision protocolsfunction asynchronously [5] For example in Aloha-basedprotocols [10] time is divided into slots and each tag ran-domly transmits its ID in each timeslot The tags contin-uously retransmit their IDs until the reader acknowledgestheir transmission However the Aloha-based protocols haveseveral serious problems For example a specific tag maynot be identified for a long time leading to the so-calledldquotag starvationrdquo problem The performance of Aloha-basedprotocols is sensitive to the number of tags Furthermore itis very difficult to predict the number of tags in mobile envi-ronments if not impossible

For reader-driven anticollision protocols they functionsynchronously since all tags are controlled and checked bythe reader simultaneouslyTherefore the reader can avoid tagstarvation under static environments Furthermore they canbe categorized into Binary Tree algorithm (BT) [11ndash13] andQuery Tree algorithm (QT) [14ndash16]

221 Binary Tree Algorithm (BT) BT performs collision res-olution by splitting collided tags into disjoint subsets Thesesubsets become increasingly smaller until they contain onetag Each tag has a random binary number generator Forexample in Figure 3 tags with a counter value of zero areconsidered to be in the transmit state otherwise tags are inthe sleep state After each timeslot the reader informs tagswhether there is a collision or not If therewas a collision eachtag in the transmit state generates a random binary numberTags will become in sleep state after being identified

As can be seen that the average number of timeslots toidentify the first tag is

119879 (119873) = log2119873 (1)

where119873 is the number of tags Note that the reader needs tobroadcast the universal condition (all bits are ldquo1rdquo) before the

International Journal of Distributed Sensor Networks 3

1 0 11 0 1 0 0

1 0 0 0 1 0 0 1

1 0 10 0

Tag 1

Tag 2

Decoded datax x x

Figure 2 Collision detection with Manchester code

0 1

0 1

0 10 1

0 1

0 10 1

CollisionTag

The first timeslot

The second timeslot

The third timeslot

Figure 3 Binary tree based anticollision procedure

first timeslots to collect all tags lying in its interrogation zeroFurthermore the total number of timeslots for identifying all119873 tags is

119879 (119873) =

119873

sum

119896=1

log2119896 (2)

The time consumption is the most serious defect since ithas not recorded the history information which results in alarge amount of repeat operations

222 Query Tree (QT) Query tree (QT) algorithms storetree construction at the reader and tags only need to have aprefix matching circuit The reader transmits a serial numberto tags which they then compare against their IDs The tagswhose IDs equal to or lower than the serial number respondto the readerrsquos commandThe reader thenmonitors tags replybit by bit usingManchester code and once a collision occursthe reader splits tags into subsets based on collided bits Thereader then transmits another query by replacing the mostsignificant collided bit with ldquo0rdquo and sets the other bits to ldquo1rdquoThis procedure stops until a single tag has been selected out

Rollback Query Tree (RQT) is proposed to reduce theaverage number of timeslots for identifying the tags Duringthe partition operation these records are saved at the readerThus the anticollision can avoid the repeating operationsbased on the saved informationwhen the reader tries to selectout the next tag The timeslot number of RQT is

119879 (119873) = log2119873 +119873 minus 1 (3)

where119873 is the number of tags

3 Problem for Identifying Mobile Tags

In static environments Rollback Query Tree reduces thenumber of timeslots to deal with collision in RFID systemsHowever both tag starvation and delay problems occurred inmobile environments If one tag enters the readerrsquos interro-gation zone when the reader is processing the entered tagssome obtained results might be destroyed by the new arrivedtag The total number of identifying tags is

119879 (119873 +119872) = 119879 (119873) +

119872

sum

119894=1

(119879119894+ 119877119894) (4)

where119873119872 are the numbers of the tags in the readerrsquos inter-rogation zone already and the new arrived tags respectively119879(119873) is the timeslots number as is defined in formula (3)119879119894 119877119894are the timeslots for processing the new arrived tag 119894

and the timeslots that caused by the repeat operations Theincreased timeslot for a new arrived tag 119894 is 119879

119894+ 119877119894

If the increased time is greater than the interval timebetween the arriving tags the existing tags might not beprocessed in a long time since the reader has to reexecute therollback operations for the new arrived tags Therefore tagsstarvation problem will happen if the following condition ismet

119879119894rarr 119894+1le 119879119894+ 119877119894 (5)

where 119879119894rarr 119894+1

is the interval time between the new arrivedtag 119894 to 119894 + 1 If each tag 119894 meets the requirement in formula(5) no tags will be selected out successfullyThis requirementis considerably strong but to a moving tag it will fail to beidentified only if the reader has not processed before it leavesthe readerrsquos interrogation zone


In the following we take an example to illustrate therollback caused by a new arriving tag As shown in Figure 4there are already four tags (IDs 10100011 11100010 11100011and 11110010) in the readerrsquos interrogation zone After tworounds of query (named 119902

1and 119902

2) the tag 10100011 is the

first one to be identified Then the reader should broadcastthe queries 119902

3 1199024 and 119902

5to select the next smallest tag

11100010 in the readerrsquos interrogation zone If a new tag01100011 enters the interrogation zone between broadcasting1199024and 119902

5 it will respond to the query 119902

5ldquole11100010rdquo Then

the reader receives the response ldquo119909110001119909rdquo (combined by11100010 and 01100011) instead of selecting out tag 11100010Therefore the reader should broadcast the new request com-mand ldquole11111111rdquo according to the highest uncertain ldquo119909rdquo bitUnfortunately this query command includes the responsesfrom all tags and the queries ldquo119902

3rdquo and ldquo119902

4rdquo have to be

executed again in the following In an extreme situation theoperation might be repeated many times for the consecutivearriving tags as shown in formula (5) Thus tag starvation ispossible for query tree schemes in mobile environments

4 QSA Query Splitting-BasedAnticollision for Mobile IoT

41 Overview of the Anticollision Algorithm Firstly a set ofqueries 119876 = (119902

1 1199022 119902

119897) is defined for the reader where

1199021is initialized as ldquole1119899rdquo The reader executes the following

steps to identify tags

(1) Broadcast the current query in 119876 to all tags(2) When receiving the responses from tags

(21) if the reply is string 119908 without ldquo119909rdquo bits thenselect and process the tag with ID 119908

(22) if a collision is detected that is the reply is string119908 with ldquo119909rdquo bits then set the next query in 119876

(23) if there is no reply from tags do nothing

(3) Update 119876 according to the received responses

Repeat the above procedure until 119876 is empty In thisprocedure four commands (REQUEST SELECT READ-DATA and DESELECT) are adopted as defined in Table 1

For tags let119908 = 11990811199082 119908

119899be the tagrsquos IDThe query is

defined as follows if119908 le 119902 then the tag sends string119908 to thereader where 119902 is the query string received from the reader

As can be seen from the above procedure that the keyproblem is how to design the query scheme our main ideais to keep the history records which can avoid rollback of thedone operations

42 Design of Adaptive Query

421 Query Stack Design for Rollback Operation As shownin Table 1 REQUEST is the query command for the readerto split tags into subsets To reduce the total timeslots foridentifying all tags we design a query stack structure toimplement the rollback operations during query processldquoPush Queryrdquo and ldquoPop Queryrdquo are the two basic operations

Table 1 Definition of commands

REQUEST(ID Condition)

Reader sends REQUEST command withparameter ID Condition to tags Tags comparetheir IDs against the received ID Conditionand reply their IDs to the reader if ID leID Condition

SELECT (ID) Reader selects the tag IDREAD-DATA Reader reads data from the selected tag

DESELECT Reader cancels the selected tag and this tagthus enters silent state

of query stack Each query response is pushed into querystack if and only if the response includes ldquo119909rdquo

We also take the example in Figure 4 to explain thestack operation The response is ldquo111119909001119909rdquo for the query1199023 ldquole11111111rdquo The response ldquo111119909001119909rdquo includes ldquo119909rdquo thus

it is pushed into the query stack as shown in Figure 5 In thesame way the response ldquo1110001119909rdquo is also pushed into thequery stack with the query 119902

4 ldquole11101111rdquo When the reader

broadcasts the query 1199025 ldquole11100010rdquo a new tag ldquo01100011rdquo

enters the readerrsquos interrogation zone Thus the response willbe ldquo119909110001119909rdquo and it is also pushed into the stack as shownin Figure 5

Once a response includes no ldquo119909rdquo the reader will select thistag (using the command SELECT as shown in Table 1) andread the selected tag (using the command READ-DATA asshown in Table 1) After that the reader will pop one elementfrom the query stack For example when the new arrived tagldquo01100011rdquo has been processed the element ldquo119909110001rdquo willthen be popped out The following design is how to set theldquo119909rdquo bits and broadcast new query commands

422 Rule Design for Query Condition The rule for querycommand is one of the most important designs in QSA Asshown in Table 2 the rules can be presented as two kindsOne is for the obtained response and the other is for the popstack

(1) Rule for Obtained Response When the reader receivesresponse including ldquo119909rdquo bits the next query condition shouldbe set according to the collision bits in the response Thesecond line in Table 2 shows the rule for this kind of obtainedresponse

In the first iteration all bits are set as ldquo1rdquo that isMax(ID) to collect the responses from all tags For examplethe parameter is ldquo11111111rdquo for eight bits ID In the followingiteration the highest bit ldquo119909rdquo is set as ldquo0rdquo the bits which arelower than the highest ldquo119909rdquo are all set as ldquo1rdquo According to thesetting of ldquo119909rdquo bits we can draw the following theorem

Theorem 1 Rule for obtained response implements binarysplitting for multiple tags

Proof The response is a string 119878 = 0 1 119909119899 where 119899 is thebit number of tag ID The highest ldquo119909rdquo bit in 119878 is denoted as 119896and it must meet the following requirement

119878 [119896] = 119909 119878 [119899 sdot sdot sdot 119896 + 1] = 0 1119899minus119896 (6)


A

B

11110010C

01100011

Rollback forthe new arrived tagq3 le 11111111

q1 le 11111111

q2 le 10111111

q7 le 11111111q4 le 11101111

q5 le 11100010

q6 le 11100011

10100011

Processed tag

To be processed

New arrived

1110001111100010

Collision

Figure 4 Rollback problem for the new arrived tag

Stack 3

Stack 2

Stack 1

Push query Pop query

x110001x

1110001x

111x001x

Figure 5 Query stack for new arrived tags

Table 2 Rule design for query condition

The firstiteration The nth iteration

Rule forobtainedresponse

Max(ID)bit(119896) = 0 bit(119896 minus 1 sdot sdot sdot 1) = 1119896minus1(where k is the highest ldquoxrdquo bit in theresponse)

Rule for popstack

Cannothappen

if there is new arrived tagbit(119896) = 1bit(119896 minus 1 sdot sdot sdot 1)= 0119896minus1

elsebit(119896 sdot sdot sdot 1) = 1119896

end

Then there is at least one tag whose 119896th bit is ldquo1rdquo and atleast one tagwhose 119896th bit is ldquo0rdquoOtherwise the 119896th bit cannotbe ldquo0rdquoTherefore REQUEST ldquole bit(119896) = 0 bit(119896minus1 sdot sdot sdot 0) = 1rdquo

splits the tags into two catalogs as binary division one catalogis with 119878[119896] = 0 and the other is with 119878[119896] = 1

This rule can reduce the query scope by dividing theresponding tags into two catalogs until only one tag respond

(2) Rule for Pop Stack After one tag is selected and processedthe top element is popped out This element includes ldquo119909rdquo bitThe rule for pop stack deals with the problem of setting thesebits which is shown in the third line of Table 2 Note that itcannot be the first iteration in anticollision process (ldquocannothappenrdquo in Table 2) since the stack is empty at the beginning

To the popped element there are two situations withoutnew arrived tag and with arrived tag when it pushed intothe stack For the situation without new arrived tag all bitsare set as ldquo1rdquo in order to include all unprocessed tags thatis bit(119896 sdot sdot sdot 1) = 1119896 where 119896 is the highest ldquo119909rdquo bit On theother hand if there are new arrived tags the highest ldquo119909rdquobit is set as ldquo1rdquo and the lower bits are all set as ldquo0rdquo that isbit(119896 minus 1 sdot sdot sdot 1) = 0119896minus1

Whether there is new arrived tag or not it can be judgedaccording to the following formula

Highest119909 (Top) gt Highest119909 (Top-1) (7)

whereHighest119909 is the function to obtain the highest ldquo119909rdquo bit inits parameter Top means the first element in the query stackand Top-1 is the second one

Theorem 2 If the highest ldquo119909rdquo bit of the first stack elementis higher than that of the second stack element that is thecondition of formula (7) is met there must be new arrived tag


Begin

Request (max ID)

Wait for a given period

No

No

No

Yes

Received theresponse

ldquoXrdquo bit in V

Yes

Read-data

Deselect

Read one tag

Stack is empty

End

Process right subtree

Process left subtree pop stackrdquo and

Set V as ldquorule forobtained valuerdquo and

request (V)

Yes

Decode the response V

Push V onto stack

Select (V)

V998400 = pop stackSet V998400 as ldquorule for

request (V998400)

Figure 6 Flow chart of the proposed QSA algorithm

Proof According to rule for obtained response the readerfirst processes the tags whose highest ldquo119909rdquo bit is ldquo0rdquoThereforeonly when the higher ldquo119909rdquo bit is processed it is possiblefor lower ldquo119909rdquo to be processed Thus the highest ldquo119909rdquo alwaysdecreases as the distance to the top of query stack Howeverit is possible that the ldquo119909rdquo can be at any bit if new tags arriveThat is to say formula (7) is a result caused by new arrivedtags

For example as shown in Figure 5 the highest ldquo119909rdquo bitin the top stack element Stack 3 is 8 and that in Stack 2 is1 Formula (7) meets and there is new arrived tag In theexample the query for Stack 2 element ldquo1110001119909rdquo is 119902

5

ldquole11100010rdquo in the example of Figure 4While the highest ldquo119909rdquobit in top element Stack 1 is the 8th bit There is new arrivedtag whose 8th bit is ldquo0rdquo Otherwise the new arrived tag isimpossible to be included by the query ldquole11100010rdquo and thetop element cannot be ldquo119909rdquo at the 8th bit In fact there is anew arrived tag whose ID is 01100011 in the example

43 The Procedure of the Proposed QSA Based on the designof query stack and query rule we describe the procedureof the proposed QSA algorithm in the following In QSA

the reader first broadcasts the query with the parameter(max ID) and receives the replies from all tags in its inter-rogation zone If the received response of this query includesldquo119909rdquo the tree can be divided into the left subtree and the rightsubtree according to the highest bit ldquo119909rdquo In the left subtreethe highest ldquo119909rdquo bit is ldquo0rdquo and that of the right tree is ldquo1rdquo Theprocedure of the anticollision algorithm is shown in Figure 6Itmainly includes three parts process the left subtree processthe right subtree and read one tag

431 Process the Left Subtree If there is ldquo119909rdquo in the receivedresponse the reader begins to process the left subtree Firstlythe reader pushes the response which has ldquo119909rdquo bits into thequery stack We design this query stack which can record theentrance to the right subtreeThen it sets the query conditionaccording to the rule for obtained response in Table 2 Finallythe reader broadcasts the query with the set condition whichenables a new round communication

432 Read One Tag If the received response has no ldquo119909rdquo bitthere is no collision and a single tag is identified The readerwill select this tag and read data from it After the tag is


0 100 200 3000

500

1000

1500Th

e num

ber o

f col

lisio

ns (t

imes

)

The number of tags

QT RQT Proposed QSA

(a)

0 100 200 3000

400

800

1200

The n

umbe

r of c

ollis

ions

(tim

es)

The number of tags

QT RQT Proposed QSA

(b)

Figure 7 The number of collisions versus various numbers of tags (a) Tag ID is 64 bits (b) tag ID is 128 bits

0 100 200 300

45

48

51

The number of tags

RQT Proposed QSA

Iden

tifica

tion

effici

ency

()

Figure 8 Identification efficiency

processed command ldquoDESELECTrdquo will make this tag keepsilent state in the following

433 Process the Right Subtree After one tag is read thealgorithm will begin to process the right subtree Firstly thetop element of the query stack is popped out and it is setaccording to the rule for pop stack As described in previousSection there are two possibilities that is with new arrivedtag and without new arrived tag when setting the querycondition If there is new tag which causes higher ldquo119909rdquo bitthe query will enter the left tree process after receiving theresponse Otherwise the process will further deal with theright subtree Note that the process will end until the query

200

400

600

800

The n

umbe

r of c

ollis

ions

(tim

es)

QT RQT Proposed QSA

0 2 4 6The interval of entering tags (timeslots)

Figure 9The number of collisions versus various mobile velocities

stack is empty which means that all tags have been identifiedalready

5 Simulation Results

To evaluate anticollision performance we compare the timeconsumption and identification efficiency of the proposedQSA algorithm with that of two representative schemesQuery tree (QT) and Rollback Query Tree (RQT) Undermobile environments we measure the number of timeslotsused to read out all tags Firstly we fix mobile velocity of thetags and evaluate the resultsThemobile velocity of tags is set


as the following rule a group of five tags enter the readerrsquosinterrogation zero at the interval of five timeslots

Figure 7 shows the results of collision numbers undervarious numbers of tags As can be seen that QT meets themost number of collisions since it has not recorded thequery results during the procedure of finding the least tagID RQT always takesmore collisions than the proposedQSAalgorithm which is consistent with formula (4) Both 64 bitsand 128 bits tag ID in Figures 7(a) and 7(b) the proposedQSA takes the least number of collisions which illustratesthe efficiency of the QSA since it overcomes the rollbackoperations when new tags arrive

To identify the tags as soon as possible the algorithmshould keep high identification efficiency Identification effi-ciency is defined as the ratio of the number of successtimeslots to the total timeslot number Higher identificationefficiency means less wasted timeslots for collisions Figure 8describes the comparison of identification efficiency betweenRQT and the proposed QSA The identification efficiency ofQSA is higher than that of RQT It validates the analysis ofnumber of consumed timeslots in Figure 7(a) The identifi-cation efficiency of the proposed QSA especially decreasesslowly as the tag number increasesWhile the RQT is of loweridentification efficiency when the tag number increases Forexample the identification efficiency of the RQT is below 45when the tag number is 300

From the theoretical analysis it can be known that themobile velocity will affect the results of collisions Finally weevaluate the impact of tagmobile velocity In this experimenttotal 200 tags enter the readerrsquos interrogation zone at differentintervals which varies from 1 to 5 timeslots The samewith the above experiments five tags come into the readerrsquosinterrogation zone at a pointed timeslot As can be seen fromFigure 9 the proposed QSA achieves the least number ofcollisions under all intervals of entering tags An interest-ing change happens in QT scheme the collision numberincreases firstly and then decreases It is because the tagnumber increases quickly at the smaller interval of enteringtags which causes more collisions The gap between RQTand the proposed QSA increases as the interval increases Itindicates that the velocity of mobile tags will cause differentrollback operations in RQT while it is eliminated in our QSAalgorithm

6 Conclusions

Collision is considered as one of the most important issuesthat affect the identification process in RFID systems Mobiletags of IoT especially induce new problems such as perfor-mance degrade and tag starvation In this paper we proposea novel anticollision algorithm named QSA to solve theseproblems The proposed QSA can overcome the problemsresulted from mobile tags efficiently Compared to priorschemes the QSA presents a better performance at differentmobile velocity and various tag numbers

Acknowledgments

This work is supported in part by the National NaturalScience Foundation of China under Grant nos 61106036

61173169 and 61272147 in part by the science and technologyprojects of Hunan province and in part by State Key Lab-oratory of Computer Architecture Institute of ComputingTechnology Chinese Academy of Sciences

References

[1] A-H Mohsenian-Rad V Shah-Mansouri V W S Wong andR Schober ldquoDistributed channel selection and randomizedinterrogation algorithms for large-scale and dense RFID sys-temsrdquo IEEE Transactions on Wireless Communications vol 9no 4 pp 1402ndash1413 2010

[2] F Villanueva V David and M Francisco ldquoInternet of Thingsarchitecture for a RFID-based product tracking businessmodelrdquo in Proceedings of the 6th International Conference onInnovative Mobile and Internet Services in Ubiquitous Comput-ing pp 811ndash816 2012

[3] Y-H Chen S-J Horng R-S Run et al ldquoA novel anti-collisionalgorithm in RFID systems for identifying passive tagsrdquo IEEETransactions on Industrial Informatics vol 6 no 1 pp 105ndash1212010

[4] J Panneerselvam L Liu R Hill Y Zhan and W Liu ldquoAninvestigation of the effect of cloud computing on networkmanagementrdquo in International Conference on High PerformanceComputing and Communications pp 1794ndash1799 2012

[5] D K Klair K-W Chin and R Raad ldquoA survey and tutorial ofRFID anti-collision protocolsrdquo IEEE Communications Surveysand Tutorials vol 12 no 3 pp 400ndash421 2010

[6] Y Jiang and R Zhang ldquoAn adaptive combination query treeprotocol for tag identification in RFID systemsrdquo IEEE Commu-nications Letters vol 16 no 8 pp 1192ndash1194 2012

[7] S Jiang J Miao and L Wang ldquoMobile node authenticationprotocol for crossing cluster in heterogeneous wireless sensornetworkrdquo in Proceedings of the 3rd International Conference onCommunication Software and Networks (ICCSN rsquo11) pp 205ndash209 May 2011

[8] X Jia Q Feng and C Ma ldquoAn efficient anti-collision protocolfor RFID tag identificationrdquo IEEE Communications Letters vol14 no 11 pp 1014ndash1016 2010

[9] L Kang J Zhang K Wu D Zhang and L Ni ldquoRCSMAreceiver-based carrier sense multiple access in UHF RFIDsystemsrdquo IEEE Transactions on Parallel and Distributed Systemsvol 23 no 4 pp 735ndash743 2012

[10] M Al-Medhwahi A Alkholidi and H Hamam ldquoA new hybridframe ALOHA and binary splitting algorithm for anti-collisionin RFID systemsrdquo in Proceedings of the 18th InternationalConference on Software Telecommunications and ComputerNetworks (SoftCOM rsquo10) pp 219ndash224 September 2010

[11] Y-C Lai and C-C Lin ldquoTwo blocking algorithms on adaptivebinary splitting single and pair resolutions for RFID tagidentificationrdquo IEEEACM Transactions on Networking vol 17no 3 pp 962ndash975 2009

[12] Y-C Lai and C-C Lin ldquoA pair-resolution blocking algorithmon adaptive binary splitting for RFID tag identificationrdquo IEEECommunications Letters vol 12 no 6 pp 432ndash434 2008

[13] Y-C Lai and L-Y Hsiao ldquoGeneral binary tree protocol forcoping with the capture effect in RFID tag identificationrdquo IEEECommunications Letters vol 14 no 3 pp 208ndash210 2010

[14] Y Lai L Hsiao H Chen C Lai and J Lin ldquoA novel querytree protocol with bit tracking in RFID tag identificationrdquo IEEETransactions on Mobile Computing pp 1ndash13 2012


[15] J H Choi D Lee andH Lee ldquoQuery tree-based reservation forefficient RFID tag anti-collisionrdquo IEEE Communications Lettersvol 11 no 1 pp 85ndash87 2007

[16] J Shin B Jeon and D Yang ldquoMultiple RFID tags identificationwith M-ary query tree schemerdquo IEEE Communications Lettersvol 17 no 3 pp 604ndash607 2013

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



Reader

Tag NTag 1 Tag 2

Readerrsquosinterrogation zone

Antenna

middot middot middot

Tag N + 1

Collision

Figure 1 The tag collision with mobile tag

of the partition operations Therefore the done operationshave to be repeated and the delay of identifying the tagsincreases If the new tags enter the readerrsquos interrogation zoneat an interval that causes rollback repeatedly some tagsmightdepart the readerrsquos interrogation zone before being identifiedwhich is called ldquotag starvationrdquo in mobile environmentsTherefore it is necessary to provide efficient anticollision formobile RFID-based IoT

In this paper we will deal with the ldquotag starvationrdquo andperformance problem for anticollision in mobile IoT By thedesign of query stack and query rules the proposed anticol-lision algorithm avoids repetition operations and decreasesthe collisions consequently In summary our contributionsinclude the following

(i) We characterize the problem of anticollision inmobile environment which is more practical for theemerging IoT technology

(ii) We propose a novel anticollision algorithm namedQuery Splitting-basedAnticollision (QSA) formobileRFID-based IoT which solves the problem of tag star-vation and performance degradation resulted fromthe tag mobility

(iii) Simulation and analysis evaluate the efficiency ofthe proposed algorithm The results show the betterperformance of the proposed algorithm to the priormethods

The rest of this paper is organized as follows Section 2outlines preliminary includes background related workSection 3 presents the problems for mobile tags Section 4gives the detailed design of our algorithm The simulationresults are provided in Section 5 Finally Section 6 concludesthe paper

2 Preliminary

21 Collision Detection Code technologies are widely usedfor collision detection Manchester code is one of the mostpopular technologies for RFID systems [5] In Manchestercode the value of one bit is defined as the voltage transitionwithin a bit windowA bit ldquo0rdquo is coded by a positive transitionwhile a bit ldquo1rdquo is coded by a negative transition In RFID

systems if two (ormore) tags transmit different values simul-taneously the positive and negative transitions of the receivedbits cancel each other out This state is not permissible inManchester code during data transmission and is recognizedas an error Therefore Manchester code makes it possible toldquotrace a collision to an individual bitrdquo and ldquofind where thecollided bit isrdquo [8]

Figure 2 shows an example of Manchester code forcollision detection The IDs of tag 1 and tag 2 are ldquo10101100rdquoand ldquo10001001rdquo respectively When tags 1 and 2 send theirIDs simultaneously using Manchester code the decodeddata from the interfered signal received by the reader isldquo10119909011199090119909rdquo where ldquo119909rdquo represents a collided bit In thisexample the locations of the collided bits are the 3rd 6thand 8th bits This information helps the reader separate thecollided tags into subsets more smartly and identify the tagsmore quickly

Manchester code can be utilized to detect the collisionbits but the tags should be strictly synchronized Fortunatelythe tags in passive RFID systems are all driven by the readerwith both energy and the same clock frequency

22 Anticollision Algorithms Many researches have focusedon the issue of anticollision including tag-driven and readerdriven procedures [9] Tag-driven anticollision protocolsfunction asynchronously [5] For example in Aloha-basedprotocols [10] time is divided into slots and each tag ran-domly transmits its ID in each timeslot The tags contin-uously retransmit their IDs until the reader acknowledgestheir transmission However the Aloha-based protocols haveseveral serious problems For example a specific tag maynot be identified for a long time leading to the so-calledldquotag starvationrdquo problem The performance of Aloha-basedprotocols is sensitive to the number of tags Furthermore itis very difficult to predict the number of tags in mobile envi-ronments if not impossible

For reader-driven anticollision protocols they functionsynchronously since all tags are controlled and checked bythe reader simultaneouslyTherefore the reader can avoid tagstarvation under static environments Furthermore they canbe categorized into Binary Tree algorithm (BT) [11ndash13] andQuery Tree algorithm (QT) [14ndash16]

221 Binary Tree Algorithm (BT) BT performs collision res-olution by splitting collided tags into disjoint subsets Thesesubsets become increasingly smaller until they contain onetag Each tag has a random binary number generator Forexample in Figure 3 tags with a counter value of zero areconsidered to be in the transmit state otherwise tags are inthe sleep state After each timeslot the reader informs tagswhether there is a collision or not If therewas a collision eachtag in the transmit state generates a random binary numberTags will become in sleep state after being identified

As can be seen that the average number of timeslots toidentify the first tag is

119879 (119873) = log2119873 (1)

where119873 is the number of tags Note that the reader needs tobroadcast the universal condition (all bits are ldquo1rdquo) before the


1 0 11 0 1 0 0

1 0 0 0 1 0 0 1

1 0 10 0

Tag 1

Tag 2

Decoded datax x x


0 1

0 1

0 10 1

0 1

0 10 1

CollisionTag

The first timeslot

The second timeslot

The third timeslot



119879 (119873) =

119873

sum

119896=1

log2119896 (2)




119879 (119873) = log2119873 +119873 minus 1 (3)




119879 (119873 +119872) = 119879 (119873) +

119872

sum

119894=1

(119879119894+ 119877119894) (4)



119894+ 119877119894


119879119894rarr 119894+1le 119879119894+ 119877119894 (5)

where 119879119894rarr 119894+1




1and 119902



3 1199024 and 119902







4rdquo have to be




1 1199022 119902










For tags let119908 = 11990811199082 119908

























A

B

11110010C

01100011


q1 le 11111111

q2 le 10111111

q7 le 11111111q4 le 11101111

q5 le 11100010

q6 le 11100011

10100011

Processed tag

To be processed

New arrived

1110001111100010

Collision


Stack 3

Stack 2

Stack 1


x110001x

1110001x

111x001x






Rule for popstack

Cannothappen



end











Begin

Request (max ID)


No

No

No

Yes



Yes

Read-data

Deselect

Read one tag

Stack is empty

End




request (V)

Yes


Push V onto stack

Select (V)


request (V998400)




5







0 100 200 3000

500

1000

1500Th

e num

ber o

f col

lisio

ns (t

imes

)

The number of tags

QT RQT Proposed QSA

(a)

0 100 200 3000

400

800

1200

The n

umbe

r of c

ollis

ions

(tim

es)

The number of tags

QT RQT Proposed QSA

(b)


0 100 200 300

45

48

51

The number of tags

RQT Proposed QSA

Iden

tifica

tion

effici

ency

()




200

400

600

800

The n

umbe

r of c

ollis

ions

(tim

es)

QT RQT Proposed QSA











6 Conclusions


Acknowledgments



References




















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










1 0 11 0 1 0 0

1 0 0 0 1 0 0 1

1 0 10 0

Tag 1

Tag 2

Decoded datax x x


0 1

0 1

0 10 1

0 1

0 10 1

CollisionTag

The first timeslot

The second timeslot

The third timeslot



119879 (119873) =

119873

sum

119896=1

log2119896 (2)




119879 (119873) = log2119873 +119873 minus 1 (3)




119879 (119873 +119872) = 119879 (119873) +

119872

sum

119894=1

(119879119894+ 119877119894) (4)



119894+ 119877119894


119879119894rarr 119894+1le 119879119894+ 119877119894 (5)

where 119879119894rarr 119894+1




1and 119902



3 1199024 and 119902







4rdquo have to be




1 1199022 119902










For tags let119908 = 11990811199082 119908

























A

B

11110010C

01100011


q1 le 11111111

q2 le 10111111

q7 le 11111111q4 le 11101111

q5 le 11100010

q6 le 11100011

10100011

Processed tag

To be processed

New arrived

1110001111100010

Collision


Stack 3

Stack 2

Stack 1


x110001x

1110001x

111x001x






Rule for popstack

Cannothappen



end











Begin

Request (max ID)


No

No

No

Yes



Yes

Read-data

Deselect

Read one tag

Stack is empty

End




request (V)

Yes


Push V onto stack

Select (V)


request (V998400)




5







0 100 200 3000

500

1000

1500Th

e num

ber o

f col

lisio

ns (t

imes

)

The number of tags

QT RQT Proposed QSA

(a)

0 100 200 3000

400

800

1200

The n

umbe

r of c

ollis

ions

(tim

es)

The number of tags

QT RQT Proposed QSA

(b)


0 100 200 300

45

48

51

The number of tags

RQT Proposed QSA

Iden

tifica

tion

effici

ency

()




200

400

600

800

The n

umbe

r of c

ollis

ions

(tim

es)

QT RQT Proposed QSA











6 Conclusions


Acknowledgments



References




















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











1and 119902



3 1199024 and 119902







4rdquo have to be




1 1199022 119902










For tags let119908 = 11990811199082 119908

























A

B

11110010C

01100011


q1 le 11111111

q2 le 10111111

q7 le 11111111q4 le 11101111

q5 le 11100010

q6 le 11100011

10100011

Processed tag

To be processed

New arrived

1110001111100010

Collision


Stack 3

Stack 2

Stack 1


x110001x

1110001x

111x001x






Rule for popstack

Cannothappen



end











Begin

Request (max ID)


No

No

No

Yes



Yes

Read-data

Deselect

Read one tag

Stack is empty

End




request (V)

Yes


Push V onto stack

Select (V)


request (V998400)




5







0 100 200 3000

500

1000

1500Th

e num

ber o

f col

lisio

ns (t

imes

)

The number of tags

QT RQT Proposed QSA

(a)

0 100 200 3000

400

800

1200

The n

umbe

r of c

ollis

ions

(tim

es)

The number of tags

QT RQT Proposed QSA

(b)


0 100 200 300

45

48

51

The number of tags

RQT Proposed QSA

Iden

tifica

tion

effici

ency

()




200

400

600

800

The n

umbe

r of c

ollis

ions

(tim

es)

QT RQT Proposed QSA











6 Conclusions


Acknowledgments



References




















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










A

B

11110010C

01100011


q1 le 11111111

q2 le 10111111

q7 le 11111111q4 le 11101111

q5 le 11100010

q6 le 11100011

10100011

Processed tag

To be processed

New arrived

1110001111100010

Collision


Stack 3

Stack 2

Stack 1


x110001x

1110001x

111x001x






Rule for popstack

Cannothappen



end











Begin

Request (max ID)


No

No

No

Yes



Yes

Read-data

Deselect

Read one tag

Stack is empty

End




request (V)

Yes


Push V onto stack

Select (V)


request (V998400)




5







0 100 200 3000

500

1000

1500Th

e num

ber o

f col

lisio

ns (t

imes

)

The number of tags

QT RQT Proposed QSA

(a)

0 100 200 3000

400

800

1200

The n

umbe

r of c

ollis

ions

(tim

es)

The number of tags

QT RQT Proposed QSA

(b)


0 100 200 300

45

48

51

The number of tags

RQT Proposed QSA

Iden

tifica

tion

effici

ency

()




200

400

600

800

The n

umbe

r of c

ollis

ions

(tim

es)

QT RQT Proposed QSA











6 Conclusions


Acknowledgments



References




















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Begin

Request (max ID)


No

No

No

Yes



Yes

Read-data

Deselect

Read one tag

Stack is empty

End




request (V)

Yes


Push V onto stack

Select (V)


request (V998400)




5







0 100 200 3000

500

1000

1500Th

e num

ber o

f col

lisio

ns (t

imes

)

The number of tags

QT RQT Proposed QSA

(a)

0 100 200 3000

400

800

1200

The n

umbe

r of c

ollis

ions

(tim

es)

The number of tags

QT RQT Proposed QSA

(b)


0 100 200 300

45

48

51

The number of tags

RQT Proposed QSA

Iden

tifica

tion

effici

ency

()




200

400

600

800

The n

umbe

r of c

ollis

ions

(tim

es)

QT RQT Proposed QSA











6 Conclusions


Acknowledgments



References




















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 100 200 3000

500

1000

1500Th

e num

ber o

f col

lisio

ns (t

imes

)

The number of tags

QT RQT Proposed QSA

(a)

0 100 200 3000

400

800

1200

The n

umbe

r of c

ollis

ions

(tim

es)

The number of tags

QT RQT Proposed QSA

(b)


0 100 200 300

45

48

51

The number of tags

RQT Proposed QSA

Iden

tifica

tion

effici

ency

()




200

400

600

800

The n

umbe

r of c

ollis

ions

(tim

es)

QT RQT Proposed QSA











6 Conclusions


Acknowledgments



References




















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














6 Conclusions


Acknowledgments



References




















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Documents

Research Article QSA: Query Splitting-Based …downloads.hindawi.com/journals/ijdsn/2013/674698.pdf4. QSA: Query Splitting-Based Anticollision for Mobile IoT.. Overview of the Anticollision