Method for allocating resources for mixed eMBB and URLLC traffic in 5G new radio

Abstract

The application discloses a method for allocating resources for mixed eMBB and URLLC services in a 5G new air interface, which comprises the following steps: a base station, K e eMBB users and K u URLLC users exist in a cell; eMBB uses a slot for transmission, which is composed of a plurality of OFDM symbols; URLLC uses a mini-slot for transmission, which is composed of a plurality of OFDM symbols; the slot of eMBB is divided into a group of K t mini-slots, which is represented as a set of P(k) = {k1, k2,..., kK} ; a puncturing strategy is used, that is, when URLLC services arrive, the physical resource block (PRB) where the eMBB services are being transmitted is punctured, and the URLLC data is immediately transmitted after being inserted; when no URLLC services arrive, the eMBB services use a proportional fair scheduling algorithm for PRB scheduling, that is, the weight of each user is calculated as follows: wherein r k,i (t) is the data transmission rate of the eMBB user i on the PRB k, R i (t) is the average transmission rate of the user i in the past period of time; max (ω k,i (t)), i = 1,..., K e is taken, and the PRB k is allocated to the user i. The application effectively supports the coexistence of eMBB / URLLC services, meets the quality of service requirements of different services, and can immediately schedule resources for URLLC services after the URLLC services arrive, thereby guaranteeing the delay requirement of the URLLC services.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of wireless communication, and particularly relates to a resource allocation method in a mixed scenario of enhanced broadband eMBB and ultra-reliable low-latency communication URLLC services in a 5G new air interface. BACKGROUND

[0002] There are three service scenarios in the 5G new air interface: enhanced broadband (eMBB), massive connectivity (mMTC), and ultra-reliable low-latency communication (URLLC). Coexistence of eMBB and URLLC will be a common scenario in the development of 5G, especially suitable for scenarios such as intelligent manufacturing, industrial Internet, and unmanned driving. Since URLLC has high latency requirements, and eMBB has high user data rate and large bandwidth occupation, coexistence of the two will inevitably cause competition for various resources in the system. Therefore, the goal of operator network design needs to specifically address the resource allocation problem when eMBB and URLLC coexist, improve the experience of eMBB users while ensuring URLLC services, and improve the efficiency of system resource utilization.

[0003] In the prior art, there are three categories of resource allocation mechanisms for coexistence of eMBB and URLLC: resource reservation, puncturing, and superimposed transmission. The resource reservation method will cause idle resources that could have been used for eMBB; the existing puncturing method sets the power allocated to eMBB to 0 on the time-frequency resources transmitting URLLC, which reduces the rate of eMBB users at the cost of ensuring the transmission of URLLC services, and sending puncturing instructions to the terminal will have signaling overhead; superimposed transmission transmits eMBB and URLLC services on the same time-frequency resources, which reduces the reliability of URLLC services.

[0004] In summary, how to optimize the resource allocation mechanism in the mixed scenario of eMBB and URLLC services is worthy of further in-depth research to find more feasible technical solutions. SUMMARY

[0005] To overcome the limitations of the prior art, the present application provides a mixed resource allocation method for coexistence of eMBB and URLLC users.

[0006] The present application adopts the following technical solutions:

[0007] A 5G new air interface eMBB and URLLC mixed service resource allocation method, according to the following steps: there is a base station in the cell, K e eMBB users and K uThere are 14 URLLC users. eMBB uses slots for transmission, consisting of several (preferably 14) Orthogonal Frequency Division Multiplexing (OFDM) symbols. URLLC uses mini-slots for transmission, consisting of several (preferably 2) OFDM symbols. The eMBB slots are divided into a group of K... t A set of mini-slots. The punching scheme is adopted, that is, after the URLLC service arrives, the physical resource block (PRB) where the eMBB service is being transmitted is punched, the URLLC data is inserted, and the transmission is carried out immediately.

[0008] When URLLC services have not yet arrived, eMBB services use the Proportional Fair Scheduler (PRB) algorithm for PRB scheduling, which calculates weights. Where r k,i (t) is the data transfer rate of eMBB user i that can be transmitted on PRB k (representing the k-th PRB), R i (t) is the average transmission rate of user i over a past period. Let max(ω) be the value of the transmission rate. k,i (t)), i=1,...,K e Assign PRB k to user i.

[0009] Preferably, an RBF neural network is trained based on historical data to fit the decoding probability function.

[0010] The decoding probability function can be expressed as: Among them, MCS i For the modulation and coding format of eMBB user i, this parameter determines the transport block length (TBS). i And the modulation method. It is the number of PRBs allocated to eMBB user i, which comes from the result of the proportional fair scheduling algorithm performing PRB scheduling on each eMBB user in the previous step. This refers to the number of punches performed by the URLLC service on eMBB user i.

[0011] The three variables that affect the decoding probability at the receiver are used as inputs to the RBF neural network, namely... x1, x2, and x3 are the expanded representations of vector X, corresponding to MCS respectively. i These three values ​​are then used to transform the input layer data to the hidden layer using a Gaussian kernel function:

[0012]

[0013] In the formula, X is the input sample vector; c j The kernel function center vector, It is the variance of the kernel function, ||Xcj || is the Euclidean distance between the sample and the center, and m is the number of hidden layers.

[0014] The output layer is set to 1 layer, and the output function is obtained after the hidden layer data is subjected to the following linear transformation: wherein, ω j is the connection weight from the hidden layer to the output layer, y i = ACK i is the feedback of the eMBB user i after the receiving end receives the transmission block, if the receiving end decodes successfully, ACK i = 1, otherwise ACK i = 0.

[0015] Under a large amount of data training, the decoding probability of the eMBB user under different puncturing numbers in the case of being allocated different modulation and coding formats MCS and PRB numbers can be fitted.

[0016] According to the prediction result of the decoding probability of the trained neural network, different puncturing strategies can be adopted, for example, the eMBB user with the minimum potential throughput loss caused by puncturing is punctured; the user is randomly selected for puncturing; the selection of the punctured eMBB user can also be combined with the current retransmission number, when a certain eMBB user is in the last HARQ retransmission, the user is punctured, which is equivalent to sacrificing the throughput of the user with poor channel conditions and multiple retransmissions to the last retransmission, and avoiding the throughput loss of other eMBB users.

[0017] The present application can effectively support eMBB / URLLC business coexistence, meet the quality of service requirements of different businesses, immediately schedule resources for the URLLC business after the URLLC business arrives, guarantee the delay requirement of the URLLC business, on this basis, reasonably puncture the eMBB business, as little as possible to affect the throughput of the eMBB user, so as to realize the reasonable compromise of eMBB / URLLC business, improve the efficiency of the whole wireless network, and enhance the user experience. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is the structure diagram of slot and mini-slot.

[0019] Figure 2 is an eMBB / URLLC puncturing coexistence resource scheduling model. DETAILED DESCRIPTION

[0020] In order to have a clearer understanding of the technical solutions of the present application, the present application will be described in detail in combination with specific embodiments.

[0021] Figure 1The structure of slot and mini-slot defined by 5G NR protocol is shown. One slot occupies 14 OFDM symbols, which is allocated to eMBB users; one mini-slot occupies 2 OFDM symbols, which is allocated to URLLC users.

[0022] Figure 2 The distribution of eMBB users and URLLC users in time-frequency grid is shown. Each eMBB user occupies several PRBs within one slot, and each URLLC user occupies several PRBs within one mini-slot.

[0023] In the eMBB and URLLC mixed traffic resource allocation method in this embodiment 5G new air interface, there is a base station in the cell, K e eMBB users and K u URLLC users. eMBB uses slot for transmission, which consists of 14 OFDM symbols. URLLC uses mini-slot for transmission, which consists of 2 OFDM symbols. The slot of eMBB is divided into a group of K t mini-slots, denoted as set The puncturing scheme is used, that is, when URLLC traffic arrives, the physical resource block (PRB) where the eMBB traffic is being transmitted is punctured, and the URLLC data is inserted and transmitted immediately after. URLLC users follow a Poisson process with arrival rate λ slot , and the arrival rate in the current mini-slot is λ t = λ slot / K t .

[0024] In this embodiment, when URLLC traffic does not arrive, eMBB traffic uses the proportional fair scheduling algorithm for PRB scheduling, that is, the weight ω is calculated, where r k,i (t) is the data transmission rate of eMBB user i that can be transmitted on PRB k, and R i (t) is the average transmission rate of user i in the past period of time. Take max(ω k,i (t)), i = 1,..., K e , and allocate PRB k to user i.

[0025] Next, the RBF neural network is trained based on historical data to fit the decoding probability function.

[0026] The decoding probability function can be expressed as where MCS i is the modulation and coding format of eMBB user i, which determines the length of the transport block TBSi and modulation mode. is the number of PRBs allocated to eMBB user i, which is the result of the PRB scheduling of each eMBB user by the proportional fair scheduling algorithm in the previous step. is the number of puncturing times of URLLC traffic to eMBB user i.

[0027] The three variables affecting the decoding probability of the receiving end are taken as the inputs of the RBF neural network, i.e. The input layer data is then transformed to the hidden layer by a Gaussian kernel function:

[0028]

[0029] In the formula, X is an input sample vector; c j is a kernel function center vector, is a kernel function variance, and ||X-c j || is the Euclidean distance between the sample and the center, and m is the number of hidden layers.

[0030] The output layer is set to one layer, and the hidden layer data is linearly transformed to obtain the output function after the following linear transformation: In the formula, ω j is the connection weight from the hidden layer to the output layer, and y i = ACK i is the feedback of the eMBB user i receiving end after receiving the transmission block, and ACK i = 1 if the receiving end decodes successfully, otherwise ACK i = 0.

[0031] After a large amount of data training, the decoding probability of the eMBB user under different puncturing numbers when the eMBB user is allocated different MCS and PRB numbers can be fitted.

[0032] According to the prediction result of the decoding probability of the trained neural network, the embodiment can adopt different puncturing strategies, specifically including the following three.

[0033] According to one embodiment of the present application, the eMBB user with the smallest potential throughput loss caused by puncturing is punctured, and the algorithm process is described as follows.

[0034] Step 1: Loop for each mini-slot: For t = 1 to K t ;

[0035] Step 2: Loop for URLLC traffic in a mini-slot: For m = 1 to λ t ;

[0036] Step 3: For each eMBB user coexisting in the mini-slot: For i = 1 to K e ;

[0037] Step 4: Prediction with RBF neural network

[0038] Step 5: Prediction with RBF neural network

[0039] Step 6: Calculate the potential throughput loss of eMBB user i due to puncturing by URLLC user m TBS i is the transport block length of user i, determined by MCS i ;

[0040] Step 7: End of eMBB user loop: End of FOR i;

[0041] Step 8: Update the puncturing weight matrix

[0042] Step 9: Let

[0043] Step 10: Loop for eMBB users: For j = 1 to K e ;

[0044] Step 11: If punctures eMBB user j,

[0045] Step 12: End of eMBB user loop: End of FOR j;

[0046] Step 13: End of URLLC service loop in the mini-slot: End of FOR m;

[0047] Step 14: End of mini-slot loop: End of FOR t.

[0048] According to another embodiment of the present application, eMBB users are punctured randomly, without bit error rate prediction.

[0049] According to another embodiment of the present application, when an eMBB user is in the last HARQ (Hybrid Automatic Repeat reQuest, which means that different versions are sent at initial transmission and each retransmission, and the receiving end combines these versions to improve transmission reliability. HARQ is well known to those skilled in the art) retransmission, puncturing is performed on this user, and the rest is punctured according to the minimum potential throughput loss principle. The specific steps are as follows:

[0050] Step 1 : Loop over each mini-slot: For t = 1 to K t ;

[0051] Step 2: Loop over URLLC traffic within a mini-slot: For m = 1 to λ t

[0052] Step 3: Loop over each eMBB user coexisting within a mini-slot: For i = 1 to K e ;

[0053] Step 4: Predict using RBF neural network

[0054] Step 5: Predict using RBF neural network

[0055] Step 6: Calculate potential throughput loss for eMBB user i due to puncturing by URLLC user m

[0056] Step 7: End of eMBB user loop: End of FOR i;

[0057] Step 8: Update puncturing weight matrix

[0058] Step 9: Set

[0059] Step 10: Loop over eMBB users: For j = 1 to K e ;

[0060] Step 11 : If eMBB user j is the last time HARQ retransmission, then puncture eMBB user j, Go to Step 13, otherwise go to Step 12;

[0061] Step 12: If Puncture eMBB user j,

[0062] Step 13: End of eMBB user loop: End of FOR j;

[0063] Step 14: End of URLLC traffic within a mini-slot loop: End of FOR m

[0064] Step 15: End of mini-slot loop: End of FOR t.

[0065] The application can effectively support eMBB / URLLC service coexistence, meet the quality of service requirements of different services, immediately schedule resources for the URLLC service after the URLLC service arrives, guarantee the delay requirement of the URLLC service, on this basis, reasonably puncture the eMBB service, as little as possible affect the throughput of the eMBB user, thereby realizing reasonable compromise of the eMBB / URLLC service, improving the efficiency of the whole wireless network, and enhancing the experience of the user.

[0066] The above are only the preferred embodiments of the present application and the technical principles applied. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A method for resource allocation for mixed eMBB and URLLC traffic in 5G new radio, characterized by There is one base station, K e eMBB users and K u URLLC users in the cell; eMBB uses slot for transmission, which is composed of several OFDM symbols; URLLC uses mini-slot for transmission, which is composed of several OFDM symbols; the slot of eMBB is divided into a group of K t mini-slots, denoted as set Punching strategy is adopted, that is, when URLLC service arrives, the physical resource block PRB where the eMBB service is being transmitted is punched, and the URLLC data is inserted and transmitted immediately; when the URLLC service does not arrive, the eMBB service uses the proportional fair scheduling algorithm for PRB scheduling, that is, the weight is calculated k,i (t) is the data transmission rate of eMBB user i transmitted on PRBk, R i (t) is the average transmission rate of user i in the past period of time; max(ω k,i (t)), i = 1,..., K e , PRBk is allocated to user i; The RBF neural network is trained based on historical data to fit the decoding probability function; The decoding probability function is expressed as wherein MCS i is the modulation and coding format of the eMBB user i, which determines the length of the transport block TBS i and the modulation mode; is the number of PRBs allocated to the eMBB user i, which is the result of the PRB scheduling of each eMBB user by the proportional fair scheduling algorithm; is the puncturing number of the URLLC service to the eMBB user i; The three variables affecting the decoding probability of the receiving end are taken as the inputs of the RBF neural network, namely The input layer data is transformed to the hidden layer through a Gaussian kernel function: where X is the input sample vector; c j is the kernel function center vector, is the kernel function variance, ||X-c j is the Euclidean distance between the sample and the center, and m is the number of hidden layers. The output layer is set to 1 layer, and the output function is obtained after the following linear transformation of the implicit layer data: where ω j is the connection weight from the implicit layer to the output layer, y i is the feedback of the eMBB user i receiving end after receiving the transmission block, if the receiving end decodes successfully, y i = 1, otherwise y i = 0.

2. The method of claim 1, wherein the method is characterized by: The puncturing strategy is to puncture the eMBB user with the minimum potential throughput loss caused by puncturing, and the specific process is described as follows: Step 1: Loop for each mini-slot: For t = 1 to K t ; Step 2: Loop for ULLRC traffic within a mini-slot: For m = 1 to λ t ; Step 3: For each eMBB user coexisting within a mini-slot: For i = 1 to K e Step 4: Prediction with RBF neural network Step 5: Prediction with RBF neural network Step 6: Calculate potential throughput loss for eMBB user i due to puncturing by URLLC user m Step 7: End of eMBB user loop: End of FOR i; Step 8: Update the puncturing weight matrix Step 9: Let Step 10: Loop over eMBB users: For j = 1 to K e ; Step 11: If Puncture for eMBB user j, Step 12: End of eMBB user loop: End of FOR j; Step 13: End of URLLC service in mini-slot loop: End of FOR m; Step 14: End of mini-slot loop: End of FOR t.

3. The method of claim 1, wherein the method is characterized by: The puncturing strategy is to randomly puncture the eMBB user without error rate prediction.

4. The method of claim 1, wherein the method is characterized by: The puncturing strategy is to puncture the eMBB user when the user is in the last HARQ retransmission, and the eMBB user puncturing is performed according to the minimum potential throughput loss principle in other cases; the specific process is as follows: Step 1: Loop for each mini-slot: For t = 1 to K t ; Step 2: Loop for ULLRC traffic within a mini-slot: For m = 1 to λ t ; Step 3: For each eMBB user coexisting within a mini-slot: For i = 1 to K e Step 4: Prediction with RBF neural network Step 5: Prediction with RBF neural network Step 6: Calculate potential throughput loss for eMBB user i due to puncturing by URLLC user m Step 7: End of eMBB user loop: End of FOR i; Step 8: Update the puncturing weight matrix Step 9: Let Step 10: Loop over eMBB users: For j = 1 to K e ; Step 11: If eMBB user j is the last HARQ retransmission, puncture eMBB user j, Go to step 13, else perform step 12; Step 12: If Puncture eMBB user j, Step 13: End of eMBB user loop: End of FOR j; Step 14: End of URLLC service in mini-slot loop: End of FOR m; Step 15: End of mini-slot loop: End of FOR t.

Images