Identity-less Multiple Access Method for Massive Machine-Type Communications

Abstract

The present invention discloses an identity-less multiple access method for massive machine type communication, which is as follows: Step 1, determine the coding scheme for active users: encode the information sent by active users among a total of K users at the sending end tot ; Step 2, determine the decoding scheme for active users: at the receiving end, a two-stage iterative algorithm is proposed to jointly recover user data. This method solves the problems of high decoding complexity in the current scheme and poor performance when there are many active users.

Description

Technical Field

[0001] The present invention belongs to the field of communication technologies, and particularly relates to an identity-free multiple access method for massive machine type communications. Background Art

[0002] Massive Machine-Type Communications (mMTC) is one of the three application scenarios of the 5th generation mobile communication system (5G) defined by 3GPP. There are several notable characteristics in the mMTC scenario. For example, 1) sporadic communication: in a certain transmission time slot, only a small part of the devices are active, and the other devices are inactive. 2) short packet transmission: usually, each active device only transmits a small part of the data. Based on the above characteristics, in order to reduce the transmission delay and signaling overhead, in the mMTC scenario, we often adopt the Grant-Free (GF) access scheme, that is, active devices can directly transmit data without prior request and approval. Since the interaction process between the device and the base station is omitted, the base station cannot obtain the specific information of the active devices. Therefore, in the traditional GF access scheme, there are the following two problems: 1) Each user needs to be assigned a unique pilot sequence. For a large number of devices, a large codebook needs to be designed to support the pilot sequence assignment, resulting in a large amount of resource waste. 2) At the receiving end, active detection is required to identify the identities of active devices.

[0003] Due to the requirements of the mMTC scenario for large-scale connection and low latency, the design of random access schemes has received extensive attention. In the traditional grant-free random access scheme, active users directly transmit pilots and data to the Base Station (BS) without authorization. Due to the high density and large number of connections, this random access scheme is not suitable for the mMTC scenario. In 2017, Polyanskiy proposed a new random access scheme called identity-free random access in "A perspective on massive random-access". In this scenario, the number of active users is relatively small, the transmitted data packets are small, and the base station only cares about recovering the set of sent messages and does not need to identify the identities of active users.

[0004] In response to the problems existing in traditional access schemes in the mMTC scenario, the research on the unlabeled multiple access scenario has gradually emerged. Due to its significant feature that the receiving end does not need to identify which user the information comes from, various schemes have been proposed by scholars in recent years to approach the achievable bound. V.K. Amalladinne and A. Fengler et al. proposed a message splitting coding scheme in "A Coded Compressed Sensing Scheme for Unsourced Multiple Access" and "SPARCs for Unsourced Random Access". Each active user divides its message into sub-blocks, which are separately encoded, transmitted, and recovered, and then stitched together using a tree-based algorithm. A.K. Pradhan et al. proposed in "Sparse IDMA: A Joint Graph-Based Coding Scheme for Unsourced Random Access" to encode the data as a whole, and use the cascade of coding and spreading to divide the message of each active user into two parts. The first part is used to determine a spreading sequence, and the second encoded part is transmitted using this sequence. Z. Han et al. adopted a tail-biting convolutional code (CC) sparse spreading scheme in "Sparse Kronecker-Product Coding for Unsourced Multiple Access", and this scheme recovers the message through bilinear generalized approximate message passing (BiG-AMP). Further, A.K. Pradhan adopted a low density parity check code (LDPC) with random diffusion in "LDPC Codes with Soft Interference Cancellation for Uncoordinated Unsourced Multiple Access", used density evolution to match the iterative process, and adopted MMSE detection at the receiver. This scheme currently has the best performance, but due to the matrix inversion operation in MMSE, the complexity of this scheme is relatively high.

[0005] Taking the best-performing solution in the existing solutions as an example: In "LDPC Codes with Soft Interference Cancellation for Uncoordinated Unsourced Multiple Access", the authors proposed a solution that cascades low-density parity-check codes with dense spread spectrum. The information of each active user is divided into two parts. One part is used to determine a spread-spectrum sequence, and the other part is transmitted according to the spreading sequence after LDPC coding. The design of the LDPC code uses density evolution to match the iterative process. In the first stage of the receiver, due to the use of dense spread spectrum, this solution uses the Minimum Mean-Squared Error (MMSE) to detect the spread-spectrum sequence, and then recovers the first part of the information. In the second stage, the present invention uses a Soft Input Soft Output (SISO) iterative multi-user detection and decoding algorithm to recover the second part of the information. This solution is the best-performing solution among the existing solutions. Compared with the best-performing solution in the previous solutions, this solution drops by about 0.4 dB when the number of active devices is 25 - 275. However, when the number of active users is large, the gap between this solution and the achievable bound is still more than 2 dB, and the MMSE detection requires matrix inversion operations, which has a high complexity. The current solution has problems of poor performance and high complexity when there are many active users. Summary of the Invention

[0006] The first object of the present invention is to provide an unlabeled multiple access method for large-scale machine-type communication, which solves the problems of high decoding complexity and poor performance when there are many active users in the current solution.

[0007] The first technical solution adopted by the present invention is an unlabeled multiple access method for large-scale machine-type communication, which is specifically as follows:

[0008] Step 1: Determine the coding scheme for active users: At the transmitter, encode the information sent by the active users among the total number of users K tot ; Step 2: Determine the decoding scheme for active users: At the receiver, a two-stage iterative algorithm is proposed to jointly recover the user data.

[0009] The features of the present invention also lie in that

[0010] Step 1 is specifically as follows:

[0011] Step 1.1: The total number of users at the transmitter is K tot , and each user can be active or inactive; define s k ∈ {0, 1}, k = 1,..., K totAs an index of user activity, when s k = 1 indicates that user k is active, and s k = 0 indicates that user k is inactive; assuming that the receiver knows the total number of active users k, defined as K a , K a << K tot , each active user k transmits B bit information in n transmissions; when user k is active, its encoding process is as Figure 1 shown. Divide the information u k = (u ks , u kc ) of length B into two parts, u ks and u kc , with lengths B s and B c = B - B s ;

[0012] Step 1.2, randomly generate a binary regular sparse matrix as the switching pattern codebook, where the Hamming weights of the rows and columns of the switching pattern codebook are n c and respectively. Convert the first part u k = (u ks , u kc ) of the information u to the decimal number M, and select the Mth row of A as the switching pattern codebook a ks of user k through index modulation; M ;

[0013] Step 1.3, generate a codeword kc by channel encoding for the second part of the information u . The code rate of this encoding is B c / n c ; the codeword v k is obtained after BPSK modulation as w k，j = 1 - 2v k，j ∈ {-1, 1};

[0014] Step 1.4, transmit the n c codewords in the n M channels determined by the switching pattern codebook a c , and the remaining n - n c channels are idle channels, where a M = (a M，1 , a M，2 , …, a M，n ), a M，j ∈ {0, 1}. When a M，j = 1, it indicates that the jth channel is used to transmit information, and finally generate the transmission signal xk Transmit.

[0015] Step 2 is as follows:

[0016] Step 2.1: At the receiver, a two-stage iterative algorithm is proposed to jointly recover the encoded data in Step 1; in the first stage, based on the received signal y, the encoded data is regarded as a random variable to detect the switching pattern so as to recover the first part of the user's message u ks , k ∈ D;

[0017] Step 2.2: In the second stage, based on the switching pattern detected in the first stage, a joint iterative multi-user decoding scheme is designed to recover the second part of the user's message u kc , k ∈ D, and the decoded output information is passed to the first stage for iteration;

[0018] Step 2.3: For the joint iterative process of the two stages, according to the output information of the second stage gradually update the switching pattern obtained in the first stage and then implement the two-stage iterative process, and terminate the iteration when all user data is fully recovered or the maximum number of iterations is reached.

[0019] In Step 2.1, in the switching pattern detection of the first stage, if multiple active users select the same switching pattern, the decoding fails. Therefore, only consider the case where the switching patterns selected by active users are all different, and represent the received signal y as:

[0020]

[0021] where, b i ∈ {0, 1} is the switching pattern indicator. When b i = 1, it means that this switching pattern is active, that is, selected by an active user. b i = 0 means that this switching pattern is inactive, and at this time, it is considered that all-zero signals are transmitted Since is unknown, it can be regarded as independent random variables. Using Bayes' rule, the log-likelihood ratio (LLR) of b i can be expressed as follows:

[0022]

[0023] For the convenience of calculating Equation (2), Equation (1) is expressed as:

[0024]

[0025] According to the central limit theorem, ξ i,j can be regarded as a Gaussian variable with mean E(ξ i,j ) and variance Var(ξ i,j ), and we get:

[0026]

[0027] where m = 0, 1, q = 0, ±1. In the first stage of the initial iteration, E[ξ i,j = 0, Var[ξ i,j = K a +σ 2 and these values are continuously updated during the iteration. Select the largest K i in L(b a ) to obtain the estimated active switching mode set

[0028] In step 2.2, in the second stage, a joint iterative multi-user decoding scheme is designed to recover information. The joint iterative multi-user decoding scheme is specifically as follows: perform multi-user decoding according to the active switching mode set obtained in the first stage . For the j-th channel for information transmission of the active user k, that is, when a k,j = 1, the channel node performs multi-user detection based on the received signal y j and transmits the LLR to the variable node x k , j , j ∈ {j|a k,j = 1, j = 1,..., n};

[0029] E(ξ i,j ) and Var(ξ i,j ) are obtained from the first stage, and is used as the input, and the output

[0030]

[0031] The beneficial effects of the present invention are:

[0032] Aiming at the problem of high decoding complexity of the traditional scheme, the method of the present invention realizes low decoding complexity by using the ultra-sparse characteristic of the switching multiple access mode; aiming at the problem of poor performance when there are many active users in the traditional scheme, the soft interference cancellation technology of parallel messages is used to improve the decoding reliability, thereby improving the detection performance when the number of active users is large and enhancing the reliability of the system. Brief Description of the Drawings

[0033] Figure 1It is the encoding process diagram of active user k in step 1 of the method of the present invention;

[0034] Figure 2 It is the flow chart of two-stage iterative switch mode detection and data decoding in the inventive method;

[0035] Figure 3 It is the performance comparison diagram in the unlabeled random access scenario;

[0036] Figure 4 It is the comparison diagram of E b / N0 values of different schemes when PUPE = 0.05. Specific embodiments

[0037] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0038] The present invention provides a label-free multiple access method for massive machine type communication, specifically as follows:

[0039] Step 1: Determine the encoding scheme for active users:

[0040] Step 1.1: The total number of users at the sending end is K tot , and each user can be active or inactive. Encode the active users in the label-free multiple access system with the number of users being K tot . Define s k ∈ {0, 1}, k = 1,..., K tot as the activity index of the user. When s k = 1, it means that user k is active, and when s k = 0, it means that user k is inactive; assume that the receiving end knows the total number of active users k, defined as K a , K a << K tot . Each active user k transmits B bit information in n transmissions; when user k is active, its encoding process is as Figure 1 shown. Divide the information u k of length B into u ks and u kc two parts, with lengths of B ks and B kc = B - B s and B c respectively; s ;

[0041] Step 1.2: Randomly generate a binary regular sparse matrix as the switch mode codebook. The Hamming weights of the rows and columns of this switch mode codebook are n c and respectively. Divide the information uk =(u ks , u kc )'s first part u ks is converted to a decimal number M, and the M-th row of A is selected as the switching mode codebook a of user k through index modulation M ;

[0042] Step 1.3: Generate a codeword from the second part of the information u kc through channel coding The code rate of this coding is B c / n c ; The codeword v k is obtained after BPSK modulation w k，j = 1 - 2v k，j ∈{-1, 1};

[0043] Step 1.4: Transmit the n c codewords in the n M channels determined by the switching mode codebook a c , and the remaining n - n c channels are idle channels, where a M =(a M，1 , a M，2 , …, a M，n ), a M，j ∈{0, 1} When a M，j = 1, it means that the j-th channel is used to transmit information, and finally the transmission signal x k is transmitted.

[0044] Step 2: Determine the decoding scheme for active users. At the receiving end, a two-stage iterative algorithm is proposed to jointly recover user data, as Figure 2 shown:

[0045] Step 2.1: At the receiving end, a two-stage iterative algorithm is proposed to jointly recover the data encoded in Step 1. In the first stage, based on the received signal y, the encoded data is regarded as a random variable to detect the switching mode so as to recover the first part of the message u of the user ks , k ∈ D;

[0046] In Step 2.1, in the switching mode detection in the first stage, if multiple active users select the same switching mode, the decoding fails. Therefore, only consider the case where the switching modes selected by active users are all different. The received signal y is expressed as:

[0047]

[0048] where, b i∈{0, 1} is the switch mode indicator. When b i = 1, it means that this switch mode is active, that is, selected by an active user. b i = 0 means that this switch mode is inactive, and at this time, it is considered to transmit an all-zero signal. Due to the unknown nature of, it can be regarded as independent random variables. Using Bayes' rule, the log-likelihood ratio (LLR) of b i can be expressed as follows:

[0049]

[0050] For the convenience of calculating Equation (2), Equation (1) is expressed as:

[0051]

[0052] According to the central limit theorem, ξ i,j can be regarded as a Gaussian variable with a mean of E(ξ i,j ) and a variance of Var(ξ i,j ). Therefore, we get:

[0053]

[0054] where m = 0, 1, q = 0, ±1. In the first stage of the initial iteration, E[ξ i,j = 0, Var[ξ i,j = K a + σ 2 and the above values are continuously updated during the iteration. Select the largest K i in L(b a ) to obtain the estimated set of active switch modes

[0055] Step 2.2. In the second stage, based on the switch modes detected in the first stage, a joint iterative multi-user decoding scheme is designed to recover the second part of the user's message u kc , k ∈ D, and the decoded output information is transmitted to the first stage for iteration;

[0056] In Step 2.2, in the second stage, a joint iterative multi-user decoding scheme is designed to recover the information. The joint iterative multi-user decoding scheme is specifically as follows: Perform multi-user decoding according to the set of active switch modes obtained in the first stage. For the jth channel through which the active user k transmits information, that is, when a k,j = 1, the channel node, based on the received signal y jPerform multi-user detection and pass the LLR to variable node x k , j , j ∈ {j | a k,j = 1, j = 1, ..., n}.

[0057] E(ξ i,j ) and Var(ξ i,j ) are obtained in the first stage. Taking as the input, the output

[0058]

[0059] Step 2.3: For the joint iterative process of the two stages, update gradually according to the output information of the second stage E(ξ i,j ) and Var(ξ i,j ), so as to update L(b i ) and the detection of the active switch mode, and then implement the two-stage iterative process. The iteration is terminated when all user data is fully recovered or the maximum number of iterations is reached.

[0060] In the case of unlabeled random access scenarios and the same channel coding, compared with the existing scheme that requires separate recovery of the preamble transmission mode and data, when the Per-User Probability of Error (PUPE) is 0.01, the proposed method of the present invention has performance improvements of 0.5 dB, 0.7 dB, and 0.9 dB respectively when the number of active users is 50, 150, and 250, as Figure 3 shown, and the performance of the proposed method of the present invention is almost the same as that of the scheme with known mode. Therefore, the performance of the proposed method of the present invention is better than the scheme with separate recovery of mode and data in the case of different numbers of active users.

[0061] Figure 4 gives the E b / N0 required by the existing scheme and the present invention when the user error probability is 0.05. When K a < 350, the gap between the present invention and the achievable bound is less than 2 dB, while the existing scheme differs from the achievable bound by more than 2 dB when K a > 200. In addition, due to the ultra-sparse access characteristics of the switch multiple access mode and the advantages of joint iterative sparse graph multi-user decoding, the decoding complexity of the present invention is much lower than that of the LDPC-MMSE scheme and is similar to the BiG-AMP scheme.

Claims

1. An identity-less multiple access method for massive machine type communication, characterized in that The details are as follows: Step 1. Determine the coding scheme for active users: At the sending end, encode the information sent by the active users among the total number of users K tot ; The specific steps of Step 1 are as follows: Step 1.1: Divide the information u of length B k =(u ks , u kc ) into two parts, u ks and i kc , with lengths B s and B c = B - B s ; Step 1.

2. Randomly generate a binary regular sparse matrix as the switching pattern codebook, where the Hamming weights of the rows and columns of the switching pattern codebook are n c and Convert the first part u k =(u ks , u kc ) of the information u ks into a decimal number M, and select the M-th row of A as the switching pattern codebook a M ; Step 1.3: Generate a codeword from the second part of information u kc through channel coding The code rate of this coding is B c / n c ; The codeword v k is obtained after BPSK modulation Step 1.4: Place n c codewords for transmission over the n M channels determined by the switching-mode codebook a c , leaving the remaining n - n c channels as idle channels, where a M = (a M,1 , a M,2 ,..., a M,n ), a M,j ∈ {0, 1}, and when a M,j = 1, it means that the j-th channel is used to transmit information, and finally generate the transmission signal x k for transmission; Step 2. Determine the decoding scheme for active users: At the receiving end, a two-stage iterative algorithm is proposed to jointly recover the user data.

2. The identity-less multiple access method for massive machine type communication according to claim 1, wherein The specific steps of Step 2 are as follows: Step 2.

1. At the receiver, a two-stage iterative algorithm is proposed to jointly recover the encoded data in Step 1; in the first stage, based on the received signal y, the encoded data is regarded as a random variable to detect the switching pattern so as to recover the first part of the user's message u ks , k ∈ D; Step 2.

2. In the second stage, based on the switching mode detected in the first stage, a joint iterative multi-user decoding scheme is designed to recover the second part of the user's message u kc , k ∈ D, and the decoded output information is passed to the first stage for iteration; Step 2.3: Based on the output information of the second stage Gradually update the switching mode obtained in the first stage, thereby realizing a two-stage iterative process, and terminate the iteration when all user data is fully restored or the maximum number of iterations is reached.

3. The identity-less multiple access method for massive machine type communication according to claim 2, wherein In Step 2.1, in the first-stage switch-mode detection, if multiple active users select the same switch mode, the decoding fails. Therefore, only the case where the switch modes selected by active users are all different is considered, and the received signal y is expressed as: where b i ∈ {0, 1} is the switch mode indicator. When b i = 1, it means that the switch mode is active, that is, selected by an active user. b i = 0 means that the switch mode is inactive, and at this time, it is considered to transmit all-zero signals. Due to its unknownness, it is regarded as independent random variables. Using Bayes' rule, the log-likelihood ratio of b i is expressed as follows: For the convenience of calculating Equation (2), Equation (1) is expressed as: According to the central limit theorem, ξ i,j is regarded as a Gaussian variable with mean E(ξ i,j ) and variance Var(ξ i,j ), and we get: where m = 0, 1, q = 0, ±1; E[ξ i,j = 0, Var[ξ i,j = K a +σ 2 And continuously update the above values during iteration; select the largest K i in L(b a ) to obtain the estimated switching mode set 4. The identity-less multiple access method for massive machine type communication according to claim 2, wherein In step 2.2, in the second stage, a joint iterative multi-user decoding scheme is designed to recover information. Specifically, the joint iterative multi-user decoding scheme is as follows: According to the active switch mode set obtained in the first stage perform multi-user decoding; for the j-th channel through which the active user k transmits information, that is, when a k,j = 1, the channel node performs multi-user detection based on the received signal y j and transmits the LLR to the variable node x k,j , j ∈ {j|a k,j = 1, j = 1,..., n}; E(ξ i,j ) and Var(ξ i,j ) are obtained in the first stage. Taking as the input, the output

Images