CN103955657B - A kind of frame slot ultrahigh frequency RFID system anti-collision method based on blind separation - Google Patents

Abstract

A frame-slot UHF RFID system anti-collision method based on blind separation. For ultra high frequency (UHF) radio frequency signals in the frequency range of 860‑960MHz, an anti-collision algorithm for RFID systems based on Independent Component Analysis (ICA) and frame slots is proposed. The present invention analyzes the ICA algorithm steps in the RFID system and establishes an antenna system model for blind source separation, uses a specific frame time slot number selection to make the number of tags in each time slot not greater than the number of antennas of the reader, and satisfies blind source separation identification The condition of the label, and then use the ICA algorithm to realize the simultaneous identification of multiple labels. The simulation results show that, compared with the traditional tag anti-collision algorithm and the blind separation multi-tag anti-collision algorithm based on slot dynamic grouping, the present invention has obvious advantages in tag recognition rate, and the recognition time of the algorithm is less, further verification In order to ensure the feasibility and efficiency of applying blind source separation technology to label recognition, it has potential application value in engineering fields that require high efficiency and intelligent management.

Description

A Blind Separation Based Anti-collision Method for Frame-Slot UHF RFID System

technical field

The invention belongs to the multi-label reading technology in the communication field, and relates to a label anti-collision method.

Background technique

Radio Frequency Identification (RFID) is a non-contact automatic identification technology. Compared with traditional identification methods, it can complete information input and processing without contact, non-optical visibility, and non-manual intervention. With the advantages of convenient operation, large storage capacity, good confidentiality, short response time, and strong adaptability to the environment, it has been widely used in the fields of access control, transportation, food safety, and logistics. In the RFID communication system, there are often multiple tags coexisting within the range of the reader. If the reader sends a query command, it will cause multiple tags to respond at the same time, resulting in the response signal of the tag cannot be quickly recognized by the reader, resulting in RFID system failure. The reduction of recognition efficiency. The anti-collision algorithm of the tag can realize the normal communication between multiple tags and the reader, and the pros and cons of the algorithm are closely related to the tag throughput rate of the reader. Common RFID system anti-collision algorithms can be divided into two types: deterministic and non-deterministic. The ALOHA anti-collision algorithm based on probability statistics is non-deterministic. Tags reduce collisions by randomly selecting time slots for sending information, and only when the number of tags is equal to the number of time slots, the algorithm can maintain a high recognition rate, with a maximum value of 36.8 %. Even the improved dynamic frame time slot ALOHA algorithm that can adaptively adjust the frame length has a maximum recognition rate of only 58%; among the deterministic algorithms, the maximum The recognition rate is just over 50%. EPC Gen2 stipulates that the Q algorithm based on the dynamic frame time slot random algorithm is used to solve the label collision problem, which has certain adaptability and shows good throughput performance. The Q algorithm can change the number of time slots by adjusting the Q value at any time in an inventory cycle, so that the unrecognized tag can reselect the response time slot and enter the response of the next frame, but the possible repeated changes of the Q value will affect The recognition efficiency of the algorithm and the tag recognition rate are basically maintained at about 50%. Therefore, in order to further improve the tag recognition rate of the anti-collision algorithm, it is necessary to seek a new algorithm that can recognize multiple tags at the same time, and this algorithm came into being.

Blind Source Separation (BSS) of signals refers to recovering the original signal that cannot be directly observed from several observed mixed signals. Since the original signals come from different signal sources, it can be considered that the original signals are independent of each other. Independent Component Analysis (ICA), a statistical and computer technique developed in the 1980s, is one of the most popular methods in blind source separation.

Contents of the invention

The present invention aims at the ultra-high frequency (UHF) RFID system in the frequency range of 860-960MHz, combines the frame time slot ALOHA algorithm, by establishing the blind source separation antenna system model of the reader, uses the simple but effective FastICA algorithm (fast ICA algorithm) ) performs blind source separation on tag mixed signals, and proposes a frame-slot UHF RFID system anti-collision algorithm (Blind Separation and Framed-slot Algorithm, BSFA) based on blind separation, which achieves the ability to identify multiple tags at the same time Purpose. Compared with the existing tag anti-collision algorithm, the present invention has higher tag recognition rate, less time consumption and excellent performance, and has strong application value in warehouses and logistics of medium and large enterprises.

1. Combination of RFID system and blind source separation technology

1.1 Analysis of FastICA algorithm steps of RFID system

Assuming that there are n tags within the range of the reader, the response signal (original signal) sent by the tag to the reader is S=[s ₁ ,s ₂ ,…,s _n ] ^T , where s _j =[s _j1 ,s _j2 ,...,s _jL ], 1≤j≤n is the sampling value of the signal sent by the jth tag, and the number of samples is L. The reader has m antennas, and the mixed signal received by the antenna is X=[x ₁ ,x ₂ ,…,x _m ] ^T , where x _i =[x _i1 ,x _i2 ,…,x _iL ], 1≤i ≤m is the sampling value of the i-th antenna, and the relationship between the mixed signal and the tag signal is:

where i=1,2,...,m. a _ij is the mixing coefficient, and n _i is the observation noise. This formula can be expressed in vector as:

X=AS+N (2)

When the noise is ignored, the FastICA algorithm model of the RFID system is shown in Figure 1.

The problem of blind source separation of label signals is to use the statistical independence between the observed mixed signal X and the original signal S, and at the same time recover the original signal with the help of prior knowledge of the probability distribution of the original signal. In order to make the blind source separation problem solvable, the following two conditions generally need to be met:

(1) The number m of mixed signals is greater than or equal to the number n of original signals, that is, A is a full-rank random matrix of order m×n.

(2) More than two signals in the original signal S are not allowed to be Gaussian signals.

The probability distribution of the RFID signal is a super-Gaussian distribution, and the non-Gaussianity of the signal is estimated by the function G(u).

g(u)=G'(u) (4)

Where α≈1, after the reader antenna receives the tag signal, it first preprocesses the observed signal X, removes its mean value and performs whitening, so that the covariance matrix is an identity matrix, and a new signal Z is obtained. Let the unmixing matrix be W=[w ₁ ,w ₂ ,…,w _n ] ^T , where w _i =[w _i1 ,w _i2 ,…,w _im ] ^T , 0≤i≤n is the output estimated signal Y= [y ₁ ,y ₂ ,…,y _n ] Vector of unmixing coefficients for ^T.

Orthogonalize the vector w _i , if w _i converges, calculate the estimated signal Y by formula (6). If w _i does not converge, repeat the operation formula (5) until the algorithm converges.

Y=WX (6)

In order to successfully perform blind source separation on tag signals, it is required that the number of tag signals is less than or equal to the number of reader antennas. When the number of tags in the working range of the reader is large, the number of tags responding in each time slot can meet this requirement by choosing a reasonable number of frame time slots.

1.2 Antenna system model with blind source separation

The blind source separation antenna system is composed of a tag group, a multi-antenna reader and a computer system (as shown in Figure 2). The tags within the range of the reader send identification signals to the reader and are received by multiple antennas. The blind source separation processing unit identifies these tags through antenna mixed signals and stores their relevant information, and finally sends the information to the computer system for relevant data processing.

2. Description of the inventive method

2.1 The relationship between tag collision and time slot

Assuming that the number of tags is n, the number of slots is fs, and the number of antennas is A, then the probability of k tags in a slot is:

The probability that the number of tags in a slot is less than or equal to the number of antennas A is:

Then the probability that the number of tags in a time slot is more than the number of antennas A is:

p ₃ (fs,n,k)=1-p ₂ (fs,n,k)(k>A) (9)

2.2 The time slot number selection of the inventive method

The BSFA algorithm dynamically selects the number of time slots fs according to the number of tags n, so that the number of tags k in each time slot is less than or equal to the number of antennas A of the reader, and the number of time slots fs selected is:

Among them, α≈10, β≈0.8, γ≈0.66, and round() indicates rounding.

As shown in Figure 3 and Figure 4, when the number of tags n=50~500, the number of antennas A=8 and A=12, the number of time slots selected by the BSFA algorithm makes the probability p that the number of tags in a time slot is more than the number of antennas ₃ (fs,n,k) is very close to 0, so the selection of the number of time slots in the algorithm is reasonable. Table 1 shows how the probability value of p ₃ (fs,n,k) changes with the number of antennas when it is stable. It can be seen that p ₃ (fs,n,k) decreases with the increase of the number of antennas, which further proves that the algorithm The correct selection of the number of time slots.

Table 1p ₃ (fs,n,k) changes with the number of antennas

2.3 Flow process of the inventive method

(S1), the reader sends a Query command to the tag

First, the reader sends the command Query to the tag entering the identification range, starts the inventory cycle (the time interval between two consecutive Query commands), and the tag enters the ready state.

(S2), the tag responds to the Query command of the reader

(1) During the tag identification process, each tag executes a 16-bit time slot counter. After receiving the Query command, all unidentified tags (the number of tags is n) start from 0 to round(n(A+α)/(βA ² Randomly select a time slot from +γA)) time slots and store it in their respective time slot counters, the tag enters the response state, and the tag with a time slot counter of 0 responds.

(2) The response tag generates a 16-bit random sequence RN16 through a random or pseudo-random data generator (RNG) (for a tag group with a total of 10,000 tags, the probability that two or more tags generate the same RN16 sequence should be less than 0.1%, so it is feasible to use the RN16 sequence as the original signal that is statistically independent between the tags), and send the respective RN16 sequence to the reader, according to the number of response tags, it can be divided into the following two situations:

① When the number of response tags is 0, it indicates that the time slot is free, and go to (S5).

② If the number of response tags is equal to or more than 1, go to (S3) to execute the FastICA algorithm.

(S3), the reader uses the FastICA algorithm to realize signal separation

(1) The signal received by the reader antenna is a mixed signal of the original signal of the tag, and the blind source separation of the mixed signal can be performed more accurately by estimating the number of original signals:

① When there is noise, the RFID system can determine the number of original signals sent by the tag by observing the number of main eigenvalues of the correlation matrix of the mixed signal X;

② When there is no noise, it can be determined by observing the rank of the mixed signal X.

(2) Use the FastICA algorithm to unmix the mixed signal X. By selecting a reasonable unmixing coefficient matrix W (random matrix), the observed estimated signal Y is clearly identifiable, and the estimated signal Y=WX is established by the reader and the response tag. Contact, tag enters confirmation state, go to (S4).

(S4), the reader sends an ACK command to the response tag

After receiving the ACK instruction, the responding tag that enters the confirmation state sends its own PC, EPC and CRC-16 information. The reader uses the unmixing coefficient matrix W of (S3) to continue to separate and store the identification information of these tags. At this time, the tag is successfully identified, and finally the successfully identified tag is removed, and the process goes to (S5).

(S5), the reader sends a QueryRep command to the tag

The reader sends a QueryRep command, enters the ready state and decrements the slot counter value of unrecognized tags by 1, and jumps to (S2) to continue tag identification until all tag identification is completed.

Since the number of time slots selected by the BSFA algorithm can make the probability that the number of tags in a time slot is less than or equal to the number of antennas close to 0%, and the tags can be correctly separated by the FastICA algorithm, the impact of the algorithm performance on the simulation results is very weak. It will affect the correct separation of labels in the BSFA algorithm.

3. the recognition rate of the inventive method

The probability p ₄ (fs,n,k) of knowing that the number of labels in a time slot is 0 through the algorithm description is:

Then the probability that the number of labels in a slot is not 0 is p ₅ (fs,n,k):

The total number of queries N required to separate these tags is:

Wherein, "+1" means that the selection of the number of time slots has been performed only once. Then the label recognition rate of n labels is:

Assuming that the number of tags is n=50-500 and the number of antennas A=8, the number of time slots selected by the BSFA algorithm and the change of the tag recognition rate with the number of tags are shown in Figure 5 and Figure 6, and Figure 7 shows the BSFA algorithm and slot-based Comparison of the tag recognition rate of Blind Separation and Dynamic Bit-slot Grouping (BSDBG) when the number of antennas A = 2 to 32. Figure 8 and Figure 9 show the number of tags n = 50 to 256 , BSFA algorithm and BSDBG algorithm compare the total number of queries when the number of antennas A = 2 to 32, and Figure 10 shows the change of the number of time slots during the execution of the Q algorithm.

It can be seen from Figure 5 that although the number of time slots of the BSFA algorithm increases with the number of tags, it has obvious advantages over the traditional anti-collision algorithm. It can be seen from Figure 6 and Figure 7 that as the number of antennas increases, the advantage of the BSFA algorithm in the tag recognition rate is more obvious than that of the BSDBG algorithm. It can be seen from Fig. 8 and Fig. 9 that as the number of antennas increases, the total number of queries required by the BSFA algorithm proposed by the present invention is less than that of the BSDBG algorithm. Since the identification time of the algorithm is proportional to the total number of queries, the identification of the BSFA algorithm The time is also shorter than the BSDBG algorithm, and it has a faster recognition rate in a multi-antenna system. Figure 10 shows that the number of time slots selected by the Q algorithm not only increases and decreases repeatedly, but also often exceeds the number of time slots in the BSFA algorithm, resulting in waste of time slot resources and greatly reducing system performance.

The present invention makes the number of collided tags in each time slot less than or equal to the number of antennas of the reader through reasonable time slot number selection, and satisfies the condition of using the FastICA algorithm. Simulation results show that this method can effectively increase the recognition rate of tags and the stability of the algorithm.

Description of drawings

Fig. 1 is the FastICA algorithm model of the RFID system.

Fig. 2 is the antenna system model of blind source separation.

Fig. 3 shows the probability that the number of tags in one time slot is more than the number of antennas when the number A of antennas in the present invention is 8.

Fig. 4 shows the probability that the number of tags in one time slot is more than the number of antennas when the number A of antennas in the present invention is 12.

Fig. 5 shows the variation of the number of time slots with the number of labels in the algorithm (BSFA) of the present invention.

Fig. 6 shows the variation of the label recognition rate of the algorithm (BSFA) of the present invention with the number of labels.

Figure 7 is a comparison of the tag recognition rate between the algorithm of the present invention (BSFA) and the BlindSeparation and Dynamic Bit-slot Grouping (BSDBG) algorithm when the number of antennas A=2-32.

Fig. 8 is the total query times of the (BSFA) algorithm of the present invention.

Figure 9 shows the total query times of the BSDBG algorithm.

Figure 10 shows the changes in the number of time slots during the execution of the Q algorithm.

detailed description

The invention will be further illustrated by the following examples.

(1) The reader sends a Query command to the tag

The RFID tag anti-collision system of the BSFA algorithm consists of a reader and multiple tags. First, the reader sends the command Query to the tags entering the identification range, starts the inventory cycle, and the tags enter the ready state.

(2) The tag responds to the Query command of the reader

During the tag identification process, each tag executes a 16-bit time slot counter. After receiving the Query command, all tags that enter the ready state start from 0 to round(n(A+α)/(βA ² +γA)) time slots A time slot is randomly selected and stored in the respective time slot counter, the tag enters the response state, and the tag with the time slot counter of 0 responds. Then, the response tag generates a 16-bit random sequence RN16 through a random or pseudo-random data generator (RNG), and sends the respective RN16 sequence to the reader. According to the number of response tags, it can be divided into the following two situations:

① When the number of response tags is 0, it indicates that the time slot is free, and go to (5).

② If the number of response tags is equal to or more than 1, go to (3) to execute the FastICA algorithm.

(3) The reader uses the FastICA algorithm to realize signal separation

The signal received by the reader antenna is a mixed signal of the original signal of the tag. In order to accurately perform blind source separation on the mixed signal, it is first necessary to estimate the number of original signals. The RFID system can observe the correlation matrix of the mixed signal X when there is noise The number of main eigenvalues to determine the number of original signals sent by the tag, and then use the FastICA algorithm to unmix the mixed signal X. By selecting a reasonable unmixing coefficient matrix W, the observed estimated signal Y is clear and variable. By estimating the signal Y =The WX reader establishes contact with the responding tag, the tag enters the confirmation state, and goes to (4).

(4) The reader sends an ACK command to the response tag

After receiving the ACK command, the responding tag that enters the confirmation state sends its own PC, EPC and CRC-16 information. The reader uses the unmixing coefficient matrix W of (3) to continue to separate and store the identification information of these tags. At this time, the tag is successfully identified, and finally remove the successfully identified tag, and go to (5).

(5) The reader sends a QueryRep command to the tag

The reader sends the QueryRep command, enters the ready state and decrements the slot counter value of unrecognized tags by 1, and jumps to (2) to continue tag identification until all tag identification is completed.

Claims (1)

1. a kind of frame slot ultrahigh frequency RFID system anti-collision method based on blind separation, it is characterized in that passing through following steps reality It is existing：

(S1), reader sends Query orders to the label entered in identification range, starts the cycle of taking inventory, and label enters ready State；

(S2), label responds the Query orders of reader, and randomly chooses time slot and be stored in respective time slot counter；Specifically According to the following steps：

(1) each label is performed both by 16 digit time slot counters during tag recognition, and all unidentified labels receive Query lives After order, from 0~round (n (A+ α)/(β A²+ γ A)) time slot is randomly choosed in individual time slot it is stored in respective time slot counter In, label enters responsive state, and time slot counter is 0 label response；Wherein A is antenna number, and n is number of tags, α ≈ 10, β ≈ 0.8, γ ≈ 0.66, round () are round function；

(2) responsive tags produce the random sequence RN16 of 16 by randomly or pseudo-randomly number generator, and will be respective RN16 sequences be sent to reader, the number according to responsive tags is divided into following two situations：

1. when the number of responsive tags is 0, show that time slot is idle, go to (S5)；

2. the number of responsive tags is equal to or more than 1, goes to (S3) and performs FastICA algorithms；

(S3), reader utilizes FastICA algorithm separation tags mixed signals, while set up estimating signal and label primary signal Between corresponding relation, label enters acknowledgement state；Specifically according to the following steps：

(1) signal that reader antenna is received is label mixed signal, in order to accurately carry out blind source to mixed signal Separate, it is necessary first to estimate the number of primary signal, be divided into following two situations：

When 1. having noise can by observe mixed signal X correlation matrix dominant eigenvalue number come determine label send it is original Signal number；

2. can be by the number of observing the order of mixed signal X to determine primary signal during noiseless；

(2) it is using FastICA algorithms that mixed signal X solutions is mixed, by the selection mixed coefficient matrix W of rational solution, make what is observed Estimate that signal Y is clear and legible, contacted by estimating that signal Y=WX readers are set up with responsive tags, label enters acknowledgement state；

(S4), reader send ACK instruct to the label under acknowledgement state, label receive ACK instruction after, send oneself PC, EPC and CRC-16 information, reader utilizes the mixed coefficient matrix of solution in (S3) to continue to separate and store the mark letter of these labels Breath, now label is successfully identified, and goes to (S5)；

(S5), reader sends QueryRep orders to label, into ready state and the time slot counter of unrecognized label Value subtracts 1, jumps to (S2) and proceeds tag recognition, untill all tag recognitions are completed.