RFID Transponder Collision Control Algorithm

Transcription

1 Wireless Pers Commun (2011) 59: DOI /s RFID Transponder Collision Control Algorithm Ahmed Wasif Reza Tan Kim Geok Kiew Joh Chia Kaharudin Dimyati Published online: 21 February 2010 Springer Science+Business Media, LLC Abstract Transponder collision problem can be significant when a large number of RFID (radio frequency identification) transponders exist in field. Most existing anti-collision algorithms can solve this problem. However, problem arises when all or part of these transponders are having identical UID (unique identification). This paper proposes a new transponder collision control algorithm to overcome overlapping that occurs among transponders with identical UID in RFID large scale deployment (e.g., in a large warehouse), so that the RFID reader can successfully identify the quantity of transponders for each particular UID with high identification accuracy. The proposed anti-collision algorithm adopts a modified version of frequency domain method by adding stochastic delays in time domain. The obtained results show that the proposed method can achieve optimum frequency bandwidth utilization and at the same time poses high identification accuracy (almost 100%) with low identification delay. Keywords Collision Overlapping I.Code Transponder Tag identification RFID 1 Introduction RFID (radio frequency identification) technology has been an active area of research since the last few years as it can be implemented in broad range of applications, ranging from manufacturing, retailing, security, traffic control, and military to disease control [1]. The main feature that makes RFID technology superior than other automatic data capture (ADC) technologies is its contactless high-speed identification ability without line of sight (LOS) in A. W. Reza (B) K. Dimyati Faculty of Engineering, Department of Electrical Engineering, University of Malaya, Kuala Lumpur 50603, Malaysia awreza98@yahoo.com T. K. Geok K. J. Chia Faculty of Engineering and Technology, Multimedia University, Jalan Ayer Keroh Lama, Bukit Beruang, Melaka 75450, Malaysia

2 690 A. W. Reza et al. the shared medium. Generally, RFID system consists of two main components with antennas embedded on both of them: transponder (or generally known as RFID tag) and interrogator (or known as RFID reader). Transponder carries an unique identification (UID) number to identify individual type of item and respond to interrogation signal only when adjacent interrogator activates it passively or actively [2]. Passive transponder relies on power coupled from the interrogator by reflecting or backscatters the transmission from interrogator. Active transponder is battery powered or assisted to perform all operation functions and power up its internal circuit. The interrogator is responsible to communicate with transponders in field by generating radio frequency field and detecting radio wave from the transponder. However, this feature is overwhelmed by crucial RFID collision problem, which is one of the problems that prevent the widespread implementation of advanced RFID technology. The collision control protocol for multiple channel access communication system which is widely used in wired or wireless network cannot be applied directly to passive RFID transponder anti-collision due to limitation in physical size, memory, and power [3]. Therefore, this unintelligent transponder requires collision control to be done in interrogator. The passive transponder anti-collision method should be designed with minimal delay, minimal power consumption, reliability and completeness, LOS independence, robustness to environment, and scalability. Generally, transponder collision can be resolved either by using space domain, frequency domain or time domain methods. By deploying space domain distribution, the interrogator will distribute its power over several transponder zones in the predetermined or manually set sequence to achieve the goal of anti-collision. This method requires directional antenna rather than isotropic antenna. Therefore, it requires proper geometry arrangement of hardware i.e., interrogator or transponder. In time domain anti-collision method, interrogator-transponder communication is carried out in different time slots. It is also known as time division multiple access (TDMA) scheme. Generally, this method is divided into two schemes: deterministic synchronous scheme where interrogator polls through a list of transponders UID and probabilistic asynchronous scheme where transponders respond to interrogator s request at random generated time period. In synchronous scheme, tree-based anti-collision protocol, such as binary tree, memoryless query tree, adaptive memoryless query tree, and bit arbitration may be implemented. For asynchronous ALOHA-based protocol scheme, there have several approaches, such as pure ALOHA, slotted ALOHA, frame slotted ALOHA, and adaptive frame slotted ALOHA. In frequency domain method, total available bandwidth is divided into several channels and interrogatortransponders communication will be carried out by particular dedicated channel. This method is widely known as frequency division multiple access (FDMA). Hence, this method requires a larger bandwidth and accurate frequency source, which will lead to a costly system [4]. In the scenario where multiple readers or/and transponders co-exist, collision between RFID signals will occur. There are two types of collision: reader collision [5 7] and transponder collision [8 10]. Reader collision problem arises when multiple interrogators are used, normally with the purpose to improve the batch reading. This can be seen as frequency or channel assignment problems in mobile cellular communication. In recent years, several major supply chain companies, such as Wal-Mart and Tesco have deployed RFID systems in some of their supply chains to increase the visibility of goods and assets [11]. In these deployments, tens to hundreds of reader antennas will be in operation within close proximity to each other. When two such readers are activated at the same time, the tags in the overlapped region cannot differentiate between the two signals. Reader collision may lead to misreading and thus can depress the improvement of read rate and correctness. There exist various multi-reader anti-collision algorithms to solve this interference problem. However, using a master interrogator for timing control to set the particular client interrogators, so that

3 RFID Transponder Collision Control Algorithm 691 Fig. 1 Possibility of overlapping and collision among non-uniform placement of tags in 15 m 15 m area no more than one interrogator can communicate with the transponders in field within the same time slot, thus can easily solve this problem [12]. During simultaneous transponders identification, modulated data from the transponders may collide and cancel out each other or to be identified only as single data from one transponder. This will lead to the loss of information and can be significant when a large number of transponders exist in field. For grid RFID reader antenna deployment in a large warehouse as reported in our paper [13,14], where several reader antennas are positioned in a building to locate a number of distributed tags, the interference between all these transponders must be studied carefully to avoid transponder collision. Refer to our paper [13], 64 reader antennas are installed at grid points in an area of 15 m 15 m as shown in Fig. 1. Inthis setup, 45 tags are distributed non-uniformly in the grid antenna network area. It is observed that, for example, the tags 11, 12, 13, and 14 positioned at pair zone of grid reader antenna network, attempt to communicate with the reader antenna at the same time. The possibility of collision or overlapping among transponders in the read zone is indicated by the red circle as shown in the figure. Currently, a fundamental problem in RFID large scale deployment, such as in warehouse RFID deployment, is overlapping or collision among the tagged objects distributed at a reader environment. In this paper, a new transponder collision control algorithm is proposed which focuses only on the transponder collision problem. The goal of this study is to optimize the Philips I.Code performance so that it is more pervasive in most system. The main contributions of the proposed method can be summarized as follows: (1) it proposes tag identification (TID) anti-collision algorithm based on Philips I.Code, such as I.Code SLI and I.Code 1 [15 17]to overcome overlapping or collision occurs among transponders with identical UID (2) it can achieve optimum frequency bandwidth utilization and at the same time has low identification

4 692 A. W. Reza et al. delay with almost negligible identification error (identification accuracy of about 100% in some cases) (3) it can be compatible for RFID system from any manufacturer and (4) it may be appropriate to avoid collision or overlapping among tags in a large warehouse or retail industry where tagged objects are heavily populated in an RFID reader deployment. The organization of the remaining part of this paper is as follows: in Sect. 2, TID anticollision algorithm based on Philips I.Code is proposed. Section 3 covers the proposed TID protocol and the proposed RFID system modification in order to implement the TID anti-collision algorithm. In Sect. 4, performance of TID anti-collision algorithm is evaluated in terms of transponder identification accuracy and identification duration in multiple transponders scenarios. Lastly, Sect. 5 presents the conclusions of the proposed work. 2 The Proposed TID Anti-Collision Algorithm The transponder collision problem is easier when all transponders have unique UID. Most existing anti-collision algorithms are able to handle this problem. However, the transponder collision becomes complicated when all or some of the transponders have identical UID. For example, when both unintelligent transponders, namely, Tag A and Tag B (shown in Fig. 2) transmit their data simultaneously once they are energized by the interrogator, the collision of data may occur. During simultaneous transponders identification, collision may occur between transponders with different UID that will lead to incorrect receiving of modulated data, or among transponders with same UID that will cause the modulated data from transponders cancel out each other or to be identified only as data from one transponder as illustrated in Fig. 2. Due to overlapping, most existing anti-collision algorithms cannot identify the quantity of transponders for each particular UID. The proposed anti-collision algorithm adopts a modified version of frequency domain method or FDMA in combining with time domain method or TDMA, and thus can avoid any interference among the transponders. This anti-collision algorithm has the capability to overcome the problems of both the transponder collision and overlapping. In this study, the proposed TID anti-collision algorithm is developed based on the passive RFID which is operating at 13.56MHz frequency, such as I.Code standard. Modulation Tag A Modulation Tag B Modulation Tag A+Tag B Identical UID: Overlapping occurs, collision not detected Fig. 2 Overlapping and collision of transponders Unique UID: Collision detected

5 RFID Transponder Collision Control Algorithm 693 f c+4f s Tag 2 Tag 6 Tag 4 f c+3f s Tag 2 f c+2f s Tag 7 f c+f s Tag 3 Tag 1 f c-f s Tag 6 Tag 4 f c-2f s Tag 1 Tag 8 f c-3f s Tag 8 f c-4f s Tag 8 Tag 5 Tag 7 t s Fig. 3 Channel assignment of RFID system by adding stochastic delays to avoid collision among transponders with identical UID Figure 3 shows the entire two dimensional scenario of the proposed TID anti-overlapping protocol, where Tag represents transponders response with delay time at particular channel, f c is I.Code downlink carrier s frequency of MHz, f s is frequency deviation used for subcarriers assignment spacing, and t s is duration of each delay. The proposed TID anticollision algorithm is resolved around the fundamental concept of random number to delay the response of particular transponder in particular random backward link. The algorithm is based on the method of channel assignment of cellular phone system by adding some stochastic delays to avoid collision among transponders with identical UID as shown in Fig. 3. This feature allows automatic computing of total amount of identical transponders in an RF (radio frequency) field for a particular acceptable identification time. The whole transponder identification process is controlled by the proposed TID protocol, which is discussed in the next section. This algorithm requires no additional circuit to be added to transponders as the processing load of anti-collision algorithm is placed on the reader side rather than the transponders in order to reduce the cost and the size of the transponder. The proposed anti-collision algorithm consists of two random number generators: timer random generator and frequency random generator that work in parallel. The timer random generator is used to generate random delay for each addressed group of transponders instead of individual transponder as proposed in [18]. This delay method allows reader to determine the quantity of transponders with identical UID. To obtain higher guarantee of TID anti-collision, the second random generator is needed. The second random generator is used to assign a random frequency shift keying (FSK) signal to an addressed group of transponders with identical UID. This method does not require any extra circuit to be added into transponder as FSK signal is generated by the oscillator and divider that already exist inside I.Code transponder. The identification concept is similar to channel assignment in cellular communication system. Due to channel bandwidth limitation, frequency assignment in the proposed anti-collision algorithm is only used to determine the amount of transponders rather than used to distinguish between the transponders with unique UID. Thus, this proposed algorithm works well with a small bandwidth requirement only. However, since several frequencies are used to modulate the response data, therefore, to implement the proposed algorithm, the reader as in Fig. 4 must be designed to include several demodulators (e.g., Demodulator 1...m ) to demodulate each FSK signals. This small modification on reader s circuit does not add the complexity and the cost of the reader. In general, the objective of random signals generation is to allow the reader to assure the total amount of transponders in field. Thereby, it is important to have random number generating mechanism. To reduce the duplication of unpredictable random number to the

6 694 A. W. Reza et al MHz RF-amplifier 50 ohms Host Microcontroller Demodulator 1 Demodulator 2 Antenna Impedance Matching Antenna I.Code1 Tag I.Code SLI Tag Demodulator m-1 Demodulator m Reader Fig. 4 Reader design to support TID anti-collision algorithm lowest minimum level, the generation of random numbers is realized dynamically by these two generators with several iterations by processor. The implementation of linear feedback shift register cannot be applied here; as seed value obtained from a reader will cause existing loop repeats again and same random number will be created. To obtain higher guarantee of TID anti-collision with higher channel occupancy, timer random number must be integer multiple of the response data length in order to avoid partially collision among transponders response. The three bits delay data may be used to assign for return response delay. Additionally, to further suppress the chances of overlapping or collision, TID protocol employs mathematical preliminary (M-ary) FSK frequency assignment, where eight FSK signals of each encoded with three bits data is proposed for allowing the transponders response at different frequencies. Since I.Code has carrier frequency of MHz with tolerant around 70 khz, therefore, the proposed FSK signal assignment will have a frequency deviation of more than 70 khz in order to avoid the interference among these frequencies. For more secure, the proposed frequency deviation, f s is set to 0.1 MHz. 3 TID Protocol The proposed TID protocol in this study is an extension of UID protocol of I.Code. In this protocol, three commands are proposed: SelTag, RFSK, and WriteOK commands [15 17] as illustrated in Fig. 5. A SelTag command is used to awake the TID processor that is initially in sleep mode. Its frame consists of five blocks with a constant length of 88 bits. Refer to Fig. 5, the instruction frame begins with the most significant bit (MSB) of start of frame (SOF) bits and ends with the least significant bit (LSB) of end of frame (EOF) bits to allow synchronization between new command frames. After SOF bits and 8 bits SelTag command code, 64 bits UID field holds the serial number of the transponder that will be awoke to perform the TID protocol. Also, 16 bits cyclic redundancy check (CRC) is used for error control. When the addressed group of transponders is activated by the SelTag command, RdEn bit that is used to identify transponders activation state is transitioned from default value of zero to one. If this RdEn bit is reset to zero, the TID protocol processor is in inactive state again. Otherwise, the active

7 RFID Transponder Collision Control Algorithm 695 Reader SOF SelTag 8 bits UID 64 bits CRC 16 bits EOF Tag RdEn nrden 2 bits CRC 16 bits RdEn nrden = 1 0 : TID protocol processor of specified transponder is awaked RdEn nrden = 0 1 : TID protocol processor of specified transponder is in sleep mode Reader SOF RFSK 8 bits Mask FSK 4 bits Mask Delay 4 bits CRC 16 bits EOF Tag SOF FSK Data 4 bits Delay Data 4 bits CRC 16 bits EOF Reader SOF WriteOK 8 bits CRC 16 bits EOF Tag SOF Wok nwok 2 bits Wok nwok = 1 0 : Mark operation had been successfully executed Wok nwok = 0 1 : No mark operation CRC 16 bits EOF Fig. 5 Command and response structure: SelTag, RFSK, and WriteOK transponders return the changes value of RdEn and nrden bits to indicate that they are currently in awake state and can proceed to answer the interrogator requests. In SelTag s response frame, 2 bits of RdEn and nrden are proposed to indicate the TID protocol processor status as if only one bit is used. The RFSK command as illustrated in Fig. 5 consists of six blocks with 32 bits. Similar with SelTag command structure, RFSK command frame includes SOF, command code, CRC, and EOF fields. For proper operation of TID algorithm, mask FSK and mask delay fields in RFSK command frame need to be emphasized. With RFSK command, interrogator desires to obtain the stated three bits FSK signals and three bits delay signals that are generated by the built-in random generators for the specified transponders. Upon receiving the RFSK instruction, selected transponders respond their FSK data and delay data in respective channels with corresponding time delays. This command can be sent more than once for another iteration of TID protocol to further suppress the collision chances. The WriteOK command is a simple command of four blocks with 24 bits. It is proposed to indicate the successfully identified items by marking the WriteOK field in response data. Once WriteOK command is executed, RdEn will be reset, so that the TID anti-collision protocol processor will be in sleep mode again. Simultaneously, Wok and nwok bits change their values. In general, TID protocol consists of three steps: activating selected group of transponders, obtaining random signals, and identifying successfully recognized transponders. To improve the anti-collision performance, the number of TID cycles can be adapted according to the transponder population. Typically, TID cycles increase with the transponders population to reduce the probability of collisions. On contrary, if transponders population is small, the number of cycles should be limited so that the average identification delay is minimized. Figure 6 demonstrates an entire TID protocol to overcome the transponder collision problem. Prior to TID anti-overlapping protocol algorithm, an individual UID must be identified so that interrogator only transacts with the identified UID transponders in TID protocol. After the completion of UID protocol, an interrogator will initiate TID protocol by transmitting a SelTag command in order to activate the identified transponders with ID number matches the number in SelTag command s UID fields. All transponders with the selected UID of 25B9 Hex (example of TID protocol is explained in Fig. 7) will later be awaked to perform TID algorithm. The awaked transponders acknowledge interrogator about their current

8 696 A. W. Reza et al. Fig. 6 The proposed TID protocol

9 RFID Transponder Collision Control Algorithm SelTag (UID = 25B9) 25B9 is awaked 25B9 is awaked 25B9 is awaked 25B9 is awaked 1A08 in silence 2 RFSK (Mask FSK = 0x00, Mask delay = 0x00) 25B9 (3, 2) 25B9 (3 2) 25B9 (3, 4) 25B9 (2, 4) RFSK (M ask FSK= 0x03, M ask delay = 0x02) RFSK (Mask FSK = 0x02, M ask delay = 0x04) 25B9 (3,1) WriteOK 25B9 (5, 2) WriteOK 3 25B9 (5, 2) WriteOK 5 RFSK (M ask FSK = 0x03, M ask delay = 0x04) 25B9 (2, 2) WriteOK 4 Fig. 7 Example of TID protocol status by returning a set RdEn bit s SelTag response. However, due to absence of the specified transponders, the interrogator also may not receive any response signal (power can be saved due to postpone of transmission). This condition less happens, unless the transponders are unpowered or left the interrogator s field. If there is no response signal detected within a short period, interrogator may reconfirm the absent of transponders by re-instruct the transponders. In other words, if the interrogator does not receive response after a request is sent and after a pre-determined threshold period, the request is discarded. As stated, this denotes that the addressed transponders do not exist in field. This may speed up the process. Same procedures will be taken for all other interrogator commands. Also, the interrogator purposely sends commands to the transponders in order to deactivate its TID activation state by making RdEn bit of zero value. Besides, WriteOK command is instructed to acknowledge for completeness of particular TID cycle. However, on expiry of time out, interrogator will proceed to send another SelTag command with another UID (i.e., correspondent channel is switched for another transponder). Once TID protocol processors of particular group of transponders are activated, other unselected transponders or UIDs are all kept in sleep mode during this TID protocol round. After interrogator detects the existence of the selected transponders, it then transmits a RFSK command to the addressed transponders for amount computation with initial zero value of mask FSK and mask delay. Once addressed transponders receive RFSK, it will generate a random FSK signal and a random delay time. These random data (m, n) with m denotes mask FSK hexadecimal value data and n denotes mask delay hexadecimal value data, will be put into mask FSK and mask delay fields, respectively in response packet to be reflected back after wait for a certain amount of time corresponding to delay data at assigned channel

10 698 A. W. Reza et al. corresponding to FSK data. Whilst transponders are waiting for response, their random counters are decremented until they reached to zero. However, this may lead to amount estimation error as fully collided of these responses may occur. To obtain further higher accuracy, the proposed TID anti-collision algorithm sends second RFSK command to transponders with respective (m, n). With same procedures as in the first RFSK iteration that are described previously, transponders respond with their respective data of (i, j) in mask FSK and mask delay fields, where i denotes mask FSK hexadecimal value data and j denotes mask delay hexadecimal value data for the second iteration with respect to each (m, n). After the desired c TID iterations, the transponders quantity with same UID that is specified in SelTag command s UID field can then be determined by summing up the number of transponders with respective to random data, where (m, n, i, j)is shown for two iterations TID protocol in Fig. 6. Once the quantity had been determined, the transponders are switched into sleep mode by WriteOK command from interrogator. This is indicated by a response with set WOK bit and in the meantime, RdEn bit in transponders memories are preset to zero. With repetition of procedures for each identified transponder by UID protocol, in field transponders with their respective data and quantity can be obtained. Figure 7 presents an example of TID overlapping anti-collision procedures. In this example, four transponders that are labeled with A own the identical UID of 25B9 Hex and a single transponder that is labeled with B owns the identical UID of 1A08 Hex present in RF field. To begin with TID protocol, SelTag command is issued with a specified UID together with this command, saying that transponders A with data 25B9 Hex to awake these specified transponders for TID algorithm operation. Upon activated by SelTag command, transponders A hence respond with RdEn bit of one to acknowledge the interrogator. However, the interrogator does not know how many of transponders A are present in an RF field in this state. Therefore, it sends a RFSK command to the addressed transponders A for amount computation with initial zero value of mask FSK and mask delay. As shown in Fig. 7, the interrogator will able to recognize the three transponders with (m, n)of (3, 2), (2, 4),and (3, 4)from the transponders A response. This leads to amount estimation error as the actual four transponders A exist in field rather than three. Therefore, TID anti-collision algorithm sends second RFSK command to the transponders with respective (m, n) of mask FSK and mask delay. With same procedures as discussed before, transponders A respond with their respective data of (i, j) in mask FSK and mask delay fields, where (3, 1) and (5, 2) respond for (m, n) of(3,2) instruction, (2, 2) for (m, n) of (2, 4) instruction, and (5, 2) for (m, n) of (3, 4) instruction. The transponders are quantified with UID that is specified in SelTag command s UID field can then be determined by summing up the transponders with respective (m, n, i, j), saying (3, 1, 3, 2), (5, 2, 3, 2), (2, 2, 2, 4), and (5, 2, 3, 4) in the described example. With the explained procedures, four transponders A with data 25B9 Hex will be identified successfully at the end of TID protocol. After performing the TID anti-overlapping protocol for UID of 25B9 Hex, 1A08 Hex will later perform same procedures to determine its quantity. 4 Results and Analysis Before carrying out some fundamental analysis on the performance of the TID protocol, it is appropriate to define several symbols to be used for overall analysis and discussion, as in Table 1. The TID anti-collision algorithm performance can be analyzed by the identification accuracy, the identification duration, and the bandwidth requirement as explained in the following subsections.

11 RFID Transponder Collision Control Algorithm 699 Table 1 TID protocol analysis symbols Symbols N n r k m r S C Definition Number of identified transponders (different UID) in UID protocol cycle Number of transponders (same UID) to perform single TID protocol cycle Number of collided transponders (same random mask FSK and delay) Number of total random mask FSK and delay combinations Number of slots with r transponders Number of total identified transponders in TID protocol loop Number of total iterations Table 2 Frequency assignment FSK signal f c = MHz Digital signal (3 bits) Frequency spacing f s = 0.1MHz f 1 = f c 7 f s = MHz 000 f 2 = f c 5 f s = MHz 001 f 3 = f c 3 f s = MHz 010 f 4 = f c f s = MHz 011 f 5 = f c + f s = MHz 100 f 6 = f c + 3 f s = MHz 101 f 7 = f c + 5 f s = MHz 110 f 8 = f c + 7 f s = MHz M-ary FSK Frequency Assignment and Bandwidth The following are frequency parameters used in performance analysis of TID anti-collision algorithm. f i = f c + (2i 1 M) f s, where i is positive integer f c = carrier frequency = MHz f s = frequency spacing = 0.1 MHz M = number of different signal elements = 8 i = 1, 2,...,M Table 2 shows the details of M-ary FSK frequency assignment. As mentioned earlier, three bits of FSK digital signal is proposed for TID anti-collision algorithm in order to keep the bandwidth at minimum level. This three bits FSK signal with frequency spacing of 0.1 MHz requires a total maximum bandwidth of 1.6 MHz (total bandwidth required = 2 M f s = MHz= 1.6 MHz).

12 700 A. W. Reza et al. 4.2 Identification Accuracy Results Philips I.Code slotted anti-collision algorithm implements Binomial distribution in Eq. 1 to determine the occupancy of the slots, where the number of transponders in one particular slot is binomially distributed with parameter of total number of slots in a time frame and total number of transponders [19]. The Eq. 1 applies to all slots, so that the expected number of slots that are occupied by the particular number of transponders can be determined by using Eq. 2. Occupancy of slots, B n,1/n (r) = C(n, r)(1/n) r (1 1/N) n r (1) where, n: number of in field transponders N: frame size (number of slots in time frame) r: number of transponders in one slot Expected number of slots with r, ar N,n = NB n,1/n(r) (2) In Eq. 1, total random data combinations that is denoted as k is analogous to total number of slots that is denoted as N, number of transponders with same UID to perform single TID iteration is analogous to total in field transponders with both of them are denoted as n, and amount of collided transponders are analogous to number of transponders in a single slot where both of them are denoted as r. With same rationale behind, Eqs. 1 and 2 with some modifications and extensions can be applied in TID protocol analysis in order to determine the error chances in identified total items for particular UID. In this study, the proposed TID can adopt the Binomial distribution concept for error analysis as TID condition follows all stipulations of it [20]. Before discussing the proposed equations to analyze the error characteristics of the TID protocol, the underlying idea of TID protocol accuracy rate is captured in Fig. 8. InFig. 8, each matrix s row corresponds to each transponder and each matrix s column corresponds to each slot of possible random data. Suppose, one interrogator starts to identify a fixed set of transponders with identical UIDs, saying four transponders with five random combinations in example of Fig. 8, one that does not cancel out each other due to UID overlapping able to be identified without error. However, some transponders are unable to be recognized and this happens when they fall in same slot. This is indicated as dark gray slot where two transponders, respectively at row one and at row two occupy the same column, saying slot one of first matrix. The similar situation occurs for other combination of the two collided transponders out of four in field transponders. It can be visualized that there are six possible arrangements for the overlapping or collision that happens in slot one. Therefore, theorem of combination [20] can be applied to determine the total possible of arrangements of r collided transponders out of n total in field identical transponders, as it will yield to an equivalent value. The probability that the particular transponder falls into a slot, saying slot one in the described example is inversely proportional to total k number of random mask FSK and delay data or slots. The constant probability is held over all in field transponders and the same rationale is applied to all slots. Hence, the TID protocol identification error depends on the parameter of number of transponders in single TID protocol iteration, total number of random data, and total number of TID protocol iterations. With the established Eqs. (3) to(7), the TID protocol identification error level may be determined. Since those transponders are binomially distributed with probability of p over k slots, Eq. 3 can be adopted to determine the distribution of n identified transponders with

13 RFID Transponder Collision Control Algorithm 701 Let say: k = 5 random combinations, r = 2 transponders, n = 4 transponders k k k n n Total possible of r arrangement for each m r, C(n,r) = C(4,2) = 6 Note: n = number of transponders (same UID) to perform single TID protocol cycle r = number of collided transponders (same random mask FSK and delay) Fig. 8 Total possible arrangement of r collided transponders from n identical UIDs identical UID over k random mask FSK and delay data. In the stated equation, i defines the sequence of iteration number with total of C TID protocol iterations to be performed. Therefore, first sub-equation of Eq. 3 will be picked up for the first iteration to determine the distribution of r collided transponders. For the second and subsequent iterations up to C iterations that have non-zero number of slots for particular number of collided transponders, the second sub-equation will be employed. The method to determine the expected number of slots that owns the particular number of collided transponders will be discussed later in Eq. 4. However, Eq. 3 sub-equation depends upon the results that are achieved from the previous iteration, where the distribution is determined over number of collided transponders in each particular collided slot rather than over total number of transponders that is performing TID protocol cycle. Therefore, the sample size for distribution decreases with the increase of iteration number. Optimally, the unit distribution is achieved if there are less than two collided transponders in the first iteration as a confidence identification level is certainly obtained in single transponder slot or empty slot. For the rest of the conditions, the zero distribution will be achieved. After the Binomial distribution of n transponders is determined, the number of slots that is denoted as m r with r overlapped transponders may be calculated by using Eq. 4,wherethe obtained results are rounded to the nearest integer number. With this determined m r, the total identified transponders by TID protocol will be finally obtained. As shown in Eq. 5, S quantity of identified transponders is the summation of all m r that is obtained for non-empty slots, if there is only a single iteration be adopted by TID protocol. However, if more than single iteration is implemented to resolve the UID overlapping or collision problem, the product of the last iteration non-empty slots amount with its respective collided transponders quantity in previous iterations, adding to each of previous iterations of single transponder slots quantity contributes to total recognized transponders by TID protocol. After these steps, probability of identified transponders and probability of error can be easily determined with Eqs. (6) and (7), respectively. For example, the probability of two iterations TID protocol identification

14 702 A. W. Reza et al. Table 3 TID protocol probability of identification error n = 10, c = 2 k = 16 k = 64 i = 1 i = 2 i = 1 i = 2 r f(r 1) m r,1 f(r 1) m r, r f(r 2) m r,2 r f(r 2) m r, S S P identified 1 or 100% 1 or 100% P e 0 0 error is computed by the proposed equations as presented in Table 3,where f (r 1 ) and f (r 2 ) in the table is the Binomial distribution of transponders in the first and the second iteration, respectively. The mathematical computation shows that only 8 transponders are recognized in the first iteration of two bits TID protocol with 16 random combinations. Therefore, the second iteration is performed to obtain higher identification accuracy, which is almost 100% in this example. This investigation emphasizes the importance of TID protocol to be performed for more than one iteration. The 100% of read accuracy can be even achieved in the first iteration with higher random data combinations. However, it may introduce penalty due to bandwidth and delay consumption. For the above reason, appropriate value of parameter should be chosen depending upon the implemented applications. TID protocol Binomial distribution of transponders in particular slot, C(n, r i )p r i q n r i C(r f (r i ) = i 1, r i )p r i q (r i 1) r i 1 0 ; 0 r i n, i = 1 ; 0 r i r i 1, 2 i c, m r,i = 0, r i 1 2 ; r i 1 1, 2 i c ; else where, p = 1/k and q = 1 p Here, n denotes the number of in field transponders, r is the number of transponders in one slot, k represents total number of random combinations, p is the probability that occurs for each trial, q is the probability that does not occur for each trial, and c is the number of iterations. Number of slots with r i, m r,i = round {k f (r i )} ; 0 r i n, 1 i c (4) Total identified transponders, { 1 ri S = n m r,i ; c = 1 1 i<c ri=1 m r,i +{ 1 ri r i 1,i=cm r,i } m r,i 1 ; c > 1 (5) Probability of TID identified transponders, P identified = S/n (6) Probability of error, P e = 1 P identified (7) (3)

15 RFID Transponder Collision Control Algorithm 703 Table 4 Accuracy results of TID anti-collision algorithm FSK bits Delay bits Iteration Error (%) Accuracy (%) 50 tags 1 iteration 2 2 1st st tags 2 iterations 2 2 1st nd tags 2 iterations 3 3 1st nd st nd st nd st nd st nd tags 2 iterations 3 3 1st nd st nd tags 2 iterations 3 3 1st nd Nonetheless, the identification accuracy results are computed based on the predetermined parameters, such as FSK data bits, delay data bits, number of iterations, and total number of transponders with identical UID exist in field. To maximize the read accuracy, more random data combinations of FSK signal and delay sessions may be deployed (can be observed from Table 3), where the large number of random combinations contributes to higher identification accuracy. From the computed results shown in Table 4, three bits FSK and delay data can generate sufficiently low error in identification of transponders with identical UID. Therefore, six bits random data with three bits for FSK signal and three bits for delay data in double iterations TID protocol is optimal to provide better identification accuracy with the tolerance to identification delay and bandwidth requirement. However, a safety margin may be added with implementation of additional fourth FSK bit or fourth delay bit to practically avoid missing of some transponders in wireless medium. It is also observed that the identification error percentage is higher for one iteration TID algorithm. Therefore, in order to minimize the error to an assurance level, at least two iterations are required. Thus, the obtained identification accuracy results improved to about % with two iterations. Furthermore, the identification delay analysis yields that increase in number of random data bits causes only a minor increase in time-to-last transponder (will be explained in Sect. 4.3). The obtained results also indicate that the TID anti-collision algorithm is more accurate with less in field transponders with identical UID. More specifically, the identification error percentage increases with the quantity of identical transponders in awake state. For 20 identical UID transponders (shown in Table 4), the proposed algorithm can achieve the

16 704 A. W. Reza et al. identification accuracy as high as 80% even with single iteration. For practical case implementation, such as in warehouse management, more iterations together with more reserved data bits can be chosen to further ensure 100% identification accuracy at the price of identification delay. 4.3 Tag Identification Duration Figure 9 shows the time period of each command involves in each cycle of TID protocol based on Philips I.Code SLI and Philips I.Code 1 [15,16]. The detailed time period information is includedintable 5. Table 5 shows that Philips I.Code 1 IC does not embed SOF or EOF into either of its command or response structures. Besides, there is no difference in communication format for each kind of Philips I.Code in either mode. The difference only relies on the instruction and response timing periods, which are stated in Table 5 above. According to Table 5 and Fig. 9, the duration period of each command and response structure for TID protocol, such as SelTag, RFSK, and WriteOK are determined by the following parameters using Eqs. (8) to 11. SelTag 11 Bytes RFSK 4 Bytes RFSK 4 Bytes WOK 3 Bytes SelTag Response 18 bits RFSK Response 24 bits RFSK Response 24 bits WOK Response 18 bits t cd t rd t sc t sr t rc t rr t rc t rr t wc t wr t rd t cd t rd t cd t rd t cd Note: Timing format is same for both standard and fast mode. t sc is SelTag time, t rc is RFSK time, t wc is WriteOK time, t sr is SelTag response time, t rr is RFSK response time, t wr is WriteOK response time, t rd is response delay and t cd is command delay. Fig. 9 Timing characteristic of Philips I.Code with TID protocol Table 5 Timing parameters of I.Code SLI and I.Code 1 Period I.Code SLI I.Code 1 Fast mode Standard mode Fast mode Standard mode Downlink SOF 75.52µs µs 9.44µs (start pulse) 18.8 µs (start pulse) Downlink EOF µs µs Uplink SOF µs µs Uplink EOF µs µs Total bit position Command bit position µs 18.8 µs µs 18.8 µs Nominal response bit µs µs µs µs Nominal command delay, t cd µs µs µs µs Nominal response delay, t rd µs µs µs µs Note: There are no SOF and EOF in I.Code 1 command and response structure. Only start pulse presents to indicate start of the command

17 RFID Transponder Collision Control Algorithm 705 Command time, t c = Start of Frame (SOF) or start pulse period + total command bytes total bit position for each bytes period for each bit position + End of Frame (EOF) or end pulse period (8) Response time, t r = SOF period + total response bits response bit period + EOF period (9) Total command time, t tc = t sc + t rc + t wc (10) Total response time, t tr = t sr + t rr + t wr (11) Let x represents the number of TID iterations and y represents the number of delay sessions. In this case, 3 bits delay data is used; therefore, y = 8. The tag identification time, T TID can be calculated by applying Eq. 12 as follows. T TID = 2 (t cd + t rd ) + t sc + t wc + t sr + t wr + x(t rc + t cd + y (t rr + t rd )) (12) To determine the total identification time and the overall anti-collision speed, Eqs. (13)and (14) are established as below. Total identification time, T total = N (T TID + 1/AC UID ) (13) Overall anti-collision speed, AC total = N n/t total = n/ (T TID + 1/AC UID ) where, N = Total number of different UID n = Total number of identical UID (14) The AC UID in the equations is defined as UID anti-collision speed; I.code SLI transponders have anti-collision rate of 60 transponders per second (transponders/s) and I.Code 1 transponders have anti-collision speed of 30 transponders/s [15,16]. With these equations and parameters, the overall anti-collision speed with n = 10 and n = 50 UID-identified transponders is determined as shown in Table 6. The obtained values are rounded down to integer numbers to make those more approximate for the practical scenarios. The tag identification time, T TID of Philips I.Code Protocol with different delay sessions (y = 4, y = 8, and y = 16) and different iterations are also presented in Table 6.Table6 shows that the fast mode identification duration for identification of each transponder with identical UID is about four to five times lower than the results obtained by using the standard mode. In general, the identification duration is considerably lower with single iteration as compared to those obtained with double iteration. Therefore, the increase of number of TID protocol iterations leads to only minimal changes on transponder identification duration. It can be also concluded from Table 6 that anti-collision speed for maximum delay of both the standard decreases with increases of number of iterations and number of delay sessions. A single iteration at fast mode with two delay sessions draws to the highest anti-collision rate that is about 1,500 transponders/s for I.Code SLI and 1,000 transponders/s for I.Code 1. However, these parameters are less practical to be adopted as single iteration and two bits delay data contribute to a huge probability of error if it is employed in inventory with a large amount of identical UIDs, saying n = 50 in above analysis. Furthermore, it can be noticed that the number of delay sessions does not much affect on the TID supported I.Code anti-collision speed.

18 706 A. W. Reza et al. Table 6 Identification duration and overall anti-collision speed for standard and fast mode with TID protocol Maximum T TID for one identified tag (ms) AC total transponders/s Iteration number, C n = 10 n = 50 y = SLI standard mode SLI fast mode Standard mode Fast mode y = SLI standard mode SLI fast mode Standard mode Fast mode y = SLI standard mode SLI fast mode Standard mode Fast mode Figure 10 is plotted for y = 8 and one to two iterations with 10 to 100 of tags in the operation. As observed from the simulated outcomes in Fig. 10, average access time proportionally increases with in field UID quantity, regardless of transponder amount for each of them. Indeed, the obtained value is integer N multiple of single TID anti-collision protocol cycle delay time that is achieved due to N different UIDs, implicitly indicates N cycles of TID protocol are performed. To determine the effect of different UIDs transponder population that is denoted as N over N cycles of TID protocol, any number of n may be chosen, as it will yield approximately the same result. To obtain constant mean access time for each TID protocol cycle, all following results analysis in this study implement n of 50 identical UID transponders. With respect to average time-to-last transponder, the results obtained by both I.Code SLI and I.Code 1 do not differ significantly as the obtained mean access time of I.Code 1 is somewhat lower than the values obtained by I.Code SLI. It is also found that the TID anti-collision algorithm gives anti-collision speed of less than ten transponders/s (for Philips I.Code standard mode) and ten transponders/s (for Philips I.Code fast mode), regardless of transponders type. However, if there are n of 50 transponders have been chosen for each UID, a large quantity of TID Philips I.Code transponders may be identified in short duration. Within one second, several hundred I.Code transponders (standard mode) and approximately thousand I.Code transponders (fast mode) may be recognized successfully. However, the overall anti-collision rate is dynamically polled up by the identical transponder population as presented in Fig. 11. Therefore, TID supported I.Code UID anti-collision can even provide powerful identification for larger amount of identical UIDs. The anti-collision performance does not depreciate with the increase of in field transponders quantity. However, the read error chances must be tolerated in this case to provide fast automatic identification for huge transponder population. Figure 11 indicates a comparison of anti-collision speed for double iterations TID protocol in two scenarios. The upper part of the figure is the output for I.Code SLI (fast mode) with

19 RFID Transponder Collision Control Algorithm 707 Average time-to-last transponder, s Average Time-to-Last Transponder for N Identical UIDs Group (I.Code SLI) Standard Mode (One Iteration) Fast Mode (One Iteration) Standard Mode (Two Iterations) Fast Mode (Two Iterations) Total number of transponders,n Each with n=50 identical transponders Average time-to-last transponder, s Average Time-to-Last Transponder for N Identical UIDs Group (I.Code 1) Standard Mode (One Iteration) Fast Mode (One Iteration) Standard Mode (Two Iterations) Fast Mode (Two Iterations) Total number of transponders,n Each with n=50 identical transponders Fig. 10 Average time-to-last transponder for N identical UIDs group (each with n = 50 identical transponders) several y delay sessions. The total possible delay session is two power of random delay data bits. The conclusion drawn from this figure is that the overall anti-collision speed is inversely proportional to number of delay sessions. Yet, if low value of delay session is implemented, it contributes to larger read error rate, especially if in field transponders population is large. Beside, the lower part of Fig. 11 illustrates that fast mode of uplink communication provides better anti-collision rate with I.Code SLI system and it provides even higher rate than I.Code 1 system. The comparison between several Philips I.Code types with several different transmission modes may be visualized in this part of the figure. Drawn to the conclusion, the overall mean access identification time posses a proportional relation with in field UID population. From the mathematical theory and simulation analysis, it can be concluded that the optimal three FSK data bits and three delay bits in two iterations of TID protocol cycle is sufficient to recognize a large amount of identical transponders within a second (in narrow bandwidth)

20 708 A. W. Reza et al. Anti-collision speed, transponders/sec Overall Anti-collision Speed of Different y Delay Sessions, transponders/sec (I.Code SLI Fast Mode) y=2 y=4 y=8 y= Total number of identical transponders,n Overall Anti-collision Speed of Different Standards, transponders/sec (with eight delay sections and two iterations) Anti-collision speed, transponders/sec I.Code SLI Standard Mode I.Code SLI Fast Mode I.Code 1 Standard Mode I.Code 1 Fast Mode Total number of identical transponders,n Fig. 11 Comparison of overall anti-collision speed on different parameters with the assurance read level. By extending the FSK data bits and the delay data bits (for example, 6 bits for FSK signal or 6 bits for delay data) or TID cycle iterations to further improve the identification performance does not cause a large effect on the TID anti-collision performance; nevertheless, the increase in number of random data bits (i.e., increase of FSK signal bits or delay data bits) only causes a increase in time-to-last transponder (i.e., transponder identification duration). Therefore, for the tolerance of bandwidth and response time, only three bits with additional one reservation bit is proposed (the conditions or constraints are analyzed and explained before) to assign FSK signal and delay data, respectively. Even random number of time slot or frequency bands is adopted, the collision can be still happened. Basically, the TID protocol neglects weak collision that is mainly due to the discrepancy in instantaneous power among the received signals from the transponders in different locations. Therefore, simulation will involve a tradeoff between ideal case and real world. In real environment, the stochastic nature of reading process may be influenced by the nature of wireless impairment. To overcome from the above mentioned issues, an extra 30 khz is adopted to

21 RFID Transponder Collision Control Algorithm 709 avoid the adjacent channels interference and at least three respective random FSK and data bits with minimum two iterations is proposed. Additionally, a short delay of µs is inserted after a command packet and a short delay of µs is padded after a response packet to provide sufficient electronic switching time at interrogator. Thus, the proposed TID protocol is in fact a flexible protocol that can be integrated into Philips I.Code SLI or I.Code 1 transponders of any uplink transmission modes with parameter chosen depending on the designated applications. 5 Conclusions This paper has proposed a new TID anti-collision algorithm based on I.Code to overcome overlapping occurs among transponders with identical UID, so that the RFID reader can successfully identify the quantity of transponders for each particular UID. The proposed anti-collision algorithm adopts a modified version of frequency domain method or FDMA by adding stochastic delays through the time domain method or TDMA. This algorithm has achieved optimum frequency bandwidth utilization and at the same time has low identification delay with almost negligible identification error by taking into consideration the effect of number of iterations. The computed results show that the identification accuracy of almost 100% with only two iterations for as many as 100 transponders with identical UID can be achieved. On the other hand, the identification duration analysis indicates that the T TID for each transponder with identical UID is about hundred of milliseconds (ms) for standard mode with single iteration, and increased by about 30% with two iterations. For fast mode identification, the identification delay is much lower and is about four to five times lower than the standard mode. The proposed algorithm is flexible and can be implemented for RFID system from different manufacturers. The technique presented in this paper may support a variety of applications, such as supply chain management (SCM), warehouse management system (WMS), and enterprise resource planning (ERP). References 1. Shepard, S. (2005). RFID: Radio frequency identification. New York: McGraw-Hill. 2. Finkenzeller, K. (2003). RFID handbook: Fundamentals and applications in contactless smart cards and identification (3rd ed.). Chichester: Wiley. 3. Bing, B. (2000). Broadband wireless access. Boston: Kluwer. 4. Forouzan, B. A. (2003). Data communications and networking (3rd ed.). New York: McGraw-Hill. 5. Zhou, S., Luo, Z., Wong, E., Tan, C. J., & Luo, J. (2007). Interconnected RFID reader collision model and its application in reader anti-collision. In IEEE international conference on RFID, Grapevine, Texas, USA (pp ). 6. Engels, D. W., & Sarma, S. E. (2002, October). The reader collision problem. In Proceedings of the IEEE international conference on systems, man and cybernetics, Hammamet, Tunisia (Vol. 3, pp. 6). 7. Leong, K. S., Ng, M. L., & Cole, P. H. (2005, August). The reader collision problem in RFID systems. In Proceedings of the IEEE international symposium on microwave, antenna, propagation and EMC technologies, Beijing (Vol. 1, pp ). 8. Hush, D. R., & Wood, C. (1998). Analysis of tree algorithms for RFID arbitration. In Proceedings of the IEEE international symposium on information theory, Cambridge, MA, USA (pp. 107). 9. Yuan, Y., Yang, Z., He, Z., & He, J. (2006). Taxonomy and survey of RFID anti-collision protocols. Computer Communications, 29, Su-Ryun, L., Sung-Don, J., & Chae-Woo, L. (2005). An enhanced dynamic framed slotted ALOHA algorithm for RFID tag identification. In The second annual international conference on mobile and ubiquitous systems: Networking and services (MobiQuitous 2005), San Diego, CA, USA (pp ).

22 710 A. W. Reza et al. 11. Leong, K. S., Ng, M. L., Grasso, A. R., & Cole, P. H. (2006, January). Synchronization of RFID readers for dense RFID reader environment. In Proceedings of the IEEE international symposium on applications and the internet workshops, Phoenix, Arizona (pp ). 12. Reza, A. W., & Tan, K. G. (2009). Objects tracking in a dense reader environment utilizing grids of RFID antenna positioning. International Journal of Electronics, 96(12), Reza, A. W., & Tan, K. G. (2009). Investigation of indoor location sensing via RFID reader network utilizing grid covering algorithm. Journal of Wireless Personal Communications, 49(1), Tan, K. G., Wasif, A. R., & Tan, C. P. (2008). Objects tracking utilizing square grid RFID reader network. Journal of Electromagnetic Waves and Applications, 22(1), Philips Semiconductors. (2005, January). I.Code1 label ICs protocol air interface, datasheet. 16. Philips Semiconductors. (2003, January). I.Code SLI smart label IC SL2 ICS20 functional specification, datasheet, from Philips Semiconductors. (2004, January). SL2 ICS11 I.Code UID smart label IC functional specification, datasheet, from Penton Media, Inc. (1999). RFID tag reader uses FSK to avoid collisions, from com/articles/. 19. Harald, V. (2002). Efficient object identification with passive RFID tags. In Proceedings of the first international conference on pervasive computing (pp ). London, UK: Springer. 20. Kreyszig, E. (1999). Advanced engineering mathematics. London: Wiley. Author Biographies Ahmed Wasif Reza received the B.Sc. (Hons.) degree in Computer Science and Engineering from Khulna University, Bangladesh in In 2009, he obtained Master of Engineering Science (M.Eng.Sc.) degree from Multimedia University (MMU), Malaysia. Currently, he is a Lecturer in University of Malaya (UM), Faculty of Engineering, Department of Electrical Engineering, Malaysia and also pursuing his Ph.D. in the area of Wireless Communication. Before this, he joined as a Research Officer and Research Scholar under E-Science Project (sponsored by Ministry of Science, Technology and Innovation) and Intel Research Grant, respectively in MMU, Malaysia. He also held a Lecturer position in Unity College International, Malaysia and University of Science and Technology Chittagong, Bangladesh. He is now a member of ICT and Computational Science Research Cluster and Wireless Communication Research Group, UM, Malaysia. His research area includes radio frequency identification (RFID) system, radio propagation for outdoor and indoor, smart items and wireless sensor network, wireless and mobile communication, biomedical image processing, and bioinformatics. He has authored and co-authored a number of international journal and conference papers. Tan Kim Geok received the B.E., M.E., and Ph.D. degrees all in electrical engineering from University of Technology Malaysia, in 1995, 1997, and 2000 respectively. He has been Senior R&D engineer in EPCOS Singapore in In , he joined DoCoMo Euro-Labs in Munich, Germany. He is currently academic staff in Multimedia University. His research interests include radio propagation for outdoor and indoor, RFID, multi-user detection technique for multicarrier technologies, and A-GPS.

23 RFID Transponder Collision Control Algorithm 711 Kiew Joh Chia received the Bachelor of Engineering (Hons.) degree in Electronics (majoring in Telecommunications) from Multimedia University (MMU), Malaysia in Kaharudin Dimyati received B.E. in Electrical Engineering with Honours in 1992 from University of Malaya, Malaysia and subsequently his Ph.D. in Communication Engineering in 1996 from University of Wales Swansea, United Kingdom. Since then, he joined University of Malaya as an academic staff where he is currently a Professor in the Department of Electrical Engineering. Being an academic staff, he is actively involves in teaching, research and consultation. Apart from that, he is also heavily involves in administration within the university and accreditation exercises within the country. His research interest includes wireless communication, optical communication and coding theory. He has published more than 100 papers in journals and conferences. He is a member of IEEE, US and IEICE, Japan.