You are on page 1of 3

An Adaptive Memoryless Tag Anti-Collision Protocol

for RFID Networks*


Jihoon Myung and Wonjun Lee
Dept of Computer Science and Engineering, Korea University, Seoul, Korea
wlee@korea.ac.kr

Abstract— Tag collision arbitration for passive RFID tags is sig- they have the serious problem that a tag may not be identified
nificant for fast identification. This paper presents an adaptive for a long time so called tag starvation. The tree based tag
memoryless tag anti-collision protocol which is an improvement anti-collision protocols such as the binary tree protocol [3][4]
on the query tree protocol. To identify tags efficiently, we use and the query tree protocol [5] do not cause tag starvation
information about tags the reader already know. We evaluate how though they have relatively long identification delay. They split
many collisions can be reduced by the proposed protocol over the the group of colliding tags into two subgroups until the reader
query tree protocol.
recognizes the IDs of tags without collisions. In the binary tree
Keywords- RFID, tag anti-collision, tag identification. protocol, tags maintain a counter with a random number gen-
erator. The colliding tags are split according to a randomly
selected number. The tags which select 0 transmit their IDs.
I. INTRODUCTION Tags are required to have functionalities of managing the
Radio frequency identification (RFID) system is an auto- counter and the random number generator.
matic identification system. A RFID reader recognizes objects
through wireless communications with tags which have an The query tree protocol is the deterministic tag anti-collision
unique ID and information, and are attached to objects. The protocol. A reader sends a query of a prefix and tags which have
reader must be able to identify tags as quickly as possible. the ID to match the prefix respond. A prefix of the reader has the
However, signals of readers or tags collide because they com- decision power about splitting the colliding tags. Because tags
municate over the shared wireless channel and collisions inter- need not to have additional memory except the ID, the query
fere with fast identification. The reader may not recognize all tree protocol is called the memoryless protocol. It is an advan-
objects because of collisions. Therefore, anti-collision protocols tage to require low-functional tags, which is more cost-effective.
which reduce collisions and identify tags regardless of the oc- However, since it uses prefixes, the performance is sensitive to
currence of collisions are required. the distribution of IDs of tags which a reader is going to recog-
nize. There are variants of the query tree protocol in order to
Collisions are divided into reader collisions and tag colli- reduce the identification delay [5][6], but they simply focus on
sions [1]. Reader collisions occur where neighboring readers shortening time taken for transmitting the tag’s ID. Any re-
interrogate a tag simultaneously and confuse it. Tag collisions searches do not reduce the number of collisions. For this reason,
mean that more than one tags try to respond to a reader at the we propose an adaptive memoryless protocol for RFID tag
same time and make the reader recognize no tags. Reader colli- collision arbitration while drawing from the query tree protocol.
sions can be easily resolved because RFID readers can detect
collisions and communicate with one another [2]. Especially, The essential elements of the adaptive memoryless protocol
since low-functional passive tags cannot figure out neighboring are the query insertion procedure and the query deletion pro-
tags or detect collisions, tag anti-collision protocols are impor- cedure. To identify tags efficiently they generate or do away
tant for identification ability of RFID systems. with prefixes which are obtained from the past processes of tag
identification. A tag does not need additional functions except
Tag anti-collision protocols can be categorized into aloha matching a prefix of a query with its ID. We focus our attention
based protocols and tree based protocols. Aloha based tag on decreasing the number of collisions. A collision makes both
anti-collision protocols such as aloha, slotted aloha, and frame the communication overhead and the transmission delay of a
slotted aloha [3] reduce the probability of the occurrence of tag reader and tags have lost their usefulness. The less signals of
collisions how tags respond at the distinct time. Aloha’s tags tags collide, the faster a reader recognizes tags. The simulation
randomly select their response time and slotted aloha’s tags can result shows that our proposed protocol suppresses the occur-
respond in the beginning of a timeslot which is a certain time rence of collisions and reduces the total delay for identifying all
interval. Frame slotted aloha which shows the best performance the tags while preserves low communication overhead.
of aloha based protocols configures a frame with continual
timeslots. As a tag transmits its ID only at a timeslot every frame, The rest of this paper is organized as follows. Section II re-
the frame slotted aloha reduces collisions. Aloha based proto- views the query tree protocol. Section III describes our adaptive
cols, however, cannot perfectly prevent collisions. In addition, memoryless tag anti-collision protocol. In Section IV, the per-

*
This work was funded by SK Telecom under Contract Number KU-R0405721 to Korea University.
nodes in the query tree are collision nodes and all leaf nodes are
either identified nodes or no-response nodes. The delay of
identification of tags is mostly affected by the collision period
which is the sum of delays of collision nodes. Therefore, re-
ducing the number of collision nodes can improve identification
ability of the reader.

III. ADAPTIVE MEMORYLESS TAG ANTI-COLLISION


PROTOCOL
Since the queue Q is initialized by 0 and 1 in the query tree
protocol, the reader has the long collision period. As the number
of tags is increased, the length of the ID gets longer, and the IDs
of tags are further similar, and as a result, the collision period is
Figure 1. The query tree larger. To shorten the collision period, our proposed adaptive
memoryless tag anti-collision protocol uses information which
formance analysis including the derivations of several major is acquired during the last identification process. The reader
equations referred to in the analysis will be explicated. Finally maintains not only the queue Q but also a candidate queue CQ.
the conclusions of the paper will be drawn in Section V. Before the process of identification, the reader initializes the
queue Q with the candidate queue CQ which stores prefixes
from the last process of identification. The candidate queue CQ
II. QUERY TREE PROTOCOL
manages prefixes with the query insertion procedure and the
The query tree is a data structure for representing prefixes query deletion procedure.
which is sent by the reader in the query tree protocol. A reader
identifies tags through consecutive communications with tags. A. Query Insertion
The query tree protocol consists of rounds. In each round, the
The candidate queue CQ stores prefixes of all leaf nodes of
reader transmits a query and tags respond with their IDs. The
the query tree, that is, identified nodes and no-response nodes.
query has a prefix. Only tags of which IDs match the prefix
The candidate queue CQ makes use of the prefixes of identified
respond. When only a tag answers, the reader successfully
nodes for fast re-recognizing the identified tags as well as pre-
recognizes the tag. When more than one tags answer, responses
fixes of no-response nodes for identifying the new tags. Identi-
collide and the reader cannot get any information about the tags.
fied tags are recognized with prefixes of identified nodes with
The reader, however, can know the existence of tags to have ID
no collision or only a very few collisions. If a new tag which
which match the prefix. For identifying tags which lead to the
matches a prefix of an identified node comes into the identifi-
collision, the reader tries to query with 1-bit longer prefixes in
cation range of the reader, it decreases the number of collision
next rounds. The reader has a queue Q which maintains prefixes.
nodes to use prefixes of identified nodes. A new tag of which ID
The queue Q is initialized by 0 and 1. The reader deques a prefix
does not match a prefix of an identified node is quickly identi-
from the queue Q and transmits a query every round. If the
fied with prefixes of no-response nodes by the reader. Because
prefix p0p1…pi is the collision prefix, the reader enques the
all leaf nodes of the query tree are either identified nodes or
prefix p0p1…pi0 and prefix p0p1…pi1 into the queue Q. The
no-response nodes, the reader can identify all tags quickly.
reader continues the operation until the queue Q is empty. By
extending the prefixes, the reader can recognize all the tags.
B. Query Deletions
A node in the query tree corresponds with a round that the The collision node of which prefix is p0p1…pi has two child
reader transmits a prefix and tags respond. Nodes are catego- nodes as follows: 1) a collision node and a collision node, 2) a
rized as follows: collision node and an identified node, 3) a collision node and a
• Collision node: Responses cause the collision. The no-response node, and 4) an identified node and an identified
reader does not identify any tags, but it detects the col- node. As tags move out of the identification range of the reader,
lision. two child nodes of a collision node change their states as fol-
lows.
• Identified node: Only a tag responds. The reader suc-
cessfully identifies the tag’s ID. • An identified node and a no-response node: A tag which
responds to either the prefix p0p1…pi0 or the prefix
• No-response node: No tags match the prefix. There is p0p1…pi1 is successfully recognized by the reader. Be-
no response. cause there is no tag which has the prefix of the
A collision node has two children nodes as shown in Fig. 1 of no-response node, the reader can also identify the tag
which the query tree is created by identifying three tags which with the prefix p0p1…pi without a collision. Because the
have IDs of 010, 011, and 100. A Left child node has the prefix node of the prefix p0p1…pi is the identified node, there is
which is added 0 to the last of the prefix of the parent node. On no necessity of having child nodes.
the other hand, a right child node has the prefix which is added 1 • A no-response node and a no-response node: There are
to the last of the prefix of the parent node. Thus, all intermediate no tags which respond to the prefix p0p1…pi0 and the
Figure 1. The identification delay Figure 2. The composition of the identification delay

prefix p0p1…pi1. Therefore, tags which have the prefix of the identification range of the reader. All the tags have ran-
p0p1…pi do not reside in the range of the reader. The domly selected 96 bits IDs. Fig. 2 shows the identification delay
node of the prefix p0p1…pi is the no-response node and as the number of tags is increased. The more tags are, the faster
needs not to have child nodes. the adaptive memoryless tag anti-collision protocol. To evaluate
the performance of the query insertion procedure and the query
Due to the query insertion procedure, the size of the queue Q
deletion procedure, we measure the composition of states of
is extending and the length of prefixes in the queue Q are in-
transmitted prefixes in each case as shown in Fig. 3. The query
creasing over time. Unnecessary no-response nodes disturb the
insertion procedure distinguishably reduces collision nodes and
fast identification. To cope with this problem, if the node of the
the query deletion procedure eliminates no-response nodes.
prefix p0p1…pi0 and the node of the prefix p0p1…pi1 are either
the identification node and the no-response node or the
no-response node and the no-response node, the candidate V. CONCLUSION
queue CQ stores the prefix p0p1…pi instead of the prefix In this paper, an adaptive memoryless tag anti-collision
p0p1…pi0 and the prefix p0p1…pi1. The query deletion procedure protocol for passive RFID tags has been proposed and evaluated.
continues until there do not exist two nodes which have the We develop a novel and enhanced query tree protocol to reduce
same parent node, and which are either the identification node the identification delay by using prefixes which do not cause
and the no-response node or the no-response node and the collisions. The simulation results show that the query insertion
no-response node. By the query deletion procedure, the candi- procedure and the query deletion procedure of our tag
date queue CQ can eliminate prefixes of unnecessary anti-collision protocol largely diminish the number of collisions.
no-response nodes.
REFERENCES
IV. IDENTIFICATION DELAY ANALYSIS
[1] S. Sarma, D. Brock, and D. Engels, “Radio Frequency Identification and
We define the performance of the tag anti-collision protocol the Electronic Product Code,” IEEE Micro, v.21 n..6, p. 50-54, November
with the delay of identifying all tags. The total delay dtotal(n) for 2001.
identifying n tags is [2] “Information technology automatic identification and data capture
techniques – Radio frequency identification for item management Air
d total ( n) = ∑ (d p + d response ) ≈ T ( n) ⋅ d round interface - Part 6: Parameters for Air interface communications at
p∈S 860-960 MHZ,” Final Draft International Standard ISO 18000-6.
where S is the set of prefixes for n tags, dp is the delay of sending [3] Don R. Hush and Cliff Wood, “Analysis of Tree Algorithms for RFID
Arbitration,” In IEEE International Symposium on Information Theory,
the query of the prefix p, dresponse is the delay of responding with
1998.
the tag’s ID, dround is the average delay of all rounds, and T(n) is
[4] C. Law, K. Lee, and K.-Y. Siu, “Efficient memoryless protocol for tag
the number of prefixes. The total delay is determined by the
identification.” In Proceedings of the 4th International Workshop on
number of prefixes, T(n). Discrete Algorithms and Methods for Mobile Computing and
To evaluate the total delay of identifying all tags, we meas- Communications, p. 75-84, Boston, MA, August 2000.
ure the number of prefixes, T(n) of the proposed protocol [5] F. Zhou, C. Chen, D. Jin, C. Huang, and H. Min, ”Evaluating and
compared to the query tree protocol when tag has no movement, Optimizing Power Consumption of Anti-Collision Protocols for
when new tags move in, and when some tags move out. There Applications in RFID systems,” ISLPED’04, Newport Beach, California,
USA, August 2004.
have been 350 tags and new 50 tags move in or 50 tags move out

You might also like