A method for downlink user grouping and power allocation in a multi-carrier NOMA system

Abstract

The present application relates to the technical field of wireless communication, and especially relates to a method for user grouping and power allocation in downlink of a multi-carrier NOMA system. Multiple users share a subcarrier to simultaneously transmit signals in the same frequency, and a receiving end uses a serial interference cancellation technology to demodulate the user's own signal. Assuming that the base station can obtain the instantaneous channel state information from the base station to each user, the obtained CSI is used to sort the channel gains, and then the user grouping is performed, and users in the same group share the same subcarrier to communicate. The channel gain of the largest user in the group is equivalent to the channel gain of the subcarrier, and the power allocation among the subcarriers is performed according to the equivalent channel gain of the subcarriers by using a deep learning algorithm. The power allocation of the users in the subcarriers is performed according to the lowest transmission rate constraint condition. The present application avoids grouping users with similar channel gains. The deep learning algorithm is used for power allocation of the NOMA system, and the maximum transmission rate can be realized under the condition of guaranteeing the lowest transmission rate of the users.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication technology, in particular to a user grouping and power allocation method for downlink of multi-carrier NOMA system. BACKGROUND

[0002] Non-Orthogonal Multiple Access (NOMA) is a promising technology to improve the capacity and spectrum efficiency of wireless communication system. Unlike the Orthogonal Multiple Access (OMA) technology, in which a single radio resource block serves only one user equipment, the basic idea of NOMA technology is to allow multiple users to occupy the same resource for transmission. At the sending end, the data of multiple users occupies the same time / frequency domain resource and is sent out after superposition coding in the power domain. At the receiving end, the interference between different users is eliminated by successive interference cancellation technology, so as to decode the data information of each user. In order to ensure the fairness of users, less power is allocated to users with good channel conditions, and more power is allocated to users with poor channel conditions. The essence of NOMA is to exchange the complexity of the receiver for higher spectrum efficiency.

[0003] In the NOMA system, multiple users share the same resource block to transmit information, and the purpose of distinguishing users is achieved by differentiating in the power domain. Allocating users with different channel conditions on the same subcarrier will inevitably lead to different total sum rates of system information transmission, and thus affect the performance of the entire system. Therefore, in the resource allocation of the NOMA system, user grouping is a problem worth paying attention to and studying. Zhiguo Ding et al. published in IEEE Transactions on Vehicular Technology journal Impact of User Pairing on 5G Nonorthogonal Multiple-Access Downlink Transmissions (DOI: 10.1109 / TVT.2015.2480766) proposed a user grouping method based on the maximum difference of channel gain. In the case of two user pairing, compared with the OMA system, the greater the difference in channel gain between users in the same group, the more obvious the improvement in system performance. However, this method will group users with similar or even identical channel gains as a group as the grouping process progresses, increasing the difficulty of SIC at the receiving end and increasing the bit error rate.

[0004] In the research of power allocation, the full space search algorithm can achieve the best performance of the NOMA system. It searches the entire power allocation space according to the set minimum search interval, and obtains the optimal power allocation scheme by traversing all power allocation conditions of users. It can achieve the optimal theoretical throughput, but the calculation complexity is high, and it is difficult to apply to the actual system. Yuya Saito et al. proposed a fractional transmit power allocation (FTPA) algorithm in the article "System-Level Performance Evaluation of Downlink Non-orthogonal Multiple Access (NOMA)" (DOI: 10.1109 / PIMRC.2013.6666209). According to the channel conditions of users, the power allocation is reduced, and the system performance is better. However, this algorithm is a local optimization algorithm, and it is difficult to obtain the optimal performance of the system.

[0005] Nagisa Otao et al. proposed a fixed power allocation (FPA) algorithm in the article "Performance of Non-orthogonal Access with SIC in Cellular Downlink Using Proportional Fair-Based Resource Allocation" (DOI: 10.1109 / ISWCS.2012.6328413). The algorithm does not consider the channel conditions of users, and allocates power to users according to a fixed power allocation factor. The algorithm complexity is low, but the system performance is poor.

[0006] The invention patent CN 112469113A proposes to use genetic algorithm to allocate power to users in the NOMA system. Genetic algorithm can obtain global optimal solution, and does not need to be divided into steps to directly obtain the power of users, which effectively improves the total transmission rate of the system. However, this algorithm needs multiple iterations, which will lead to long convergence time, and does not consider the minimum transmission rate requirement of users, which cannot guarantee the quality of service of users.

[0007] The invention patent CN 110856247A uses an iterative water injection power allocation algorithm to allocate power among subbands, and completes power allocation on each subcarrier. On each subcarrier, based on the quality of service of users, the KKT condition is used to solve the optimization problem of power allocation in the subcarrier, and the power allocation coefficient of each user is calculated to complete the power allocation of each user. However, this invention still needs iterative calculation when allocating power among subcarriers. SUMMARY

[0008] The purpose of the present application is to overcome the deficiencies of the prior art, and provide a multi-carrier NOMA system downlink user grouping and power allocation method. An optimization problem is established with the system and rate as the objective function, and the total transmission power and the minimum transmission rate as the constraint condition. Since the complexity of directly obtaining the global optimal solution is high, the idea of step-by-step optimization is adopted to obtain the suboptimal solution of the model optimization problem. A new user grouping method is proposed, which solves the shortcomings of the user grouping method based on the maximum channel gain difference. The power allocation between subcarriers is obtained using a deep learning algorithm, and then the power allocation of the superimposed users within the subcarrier is derived according to the minimum transmission rate constraint condition. The technical problems of poor system performance, complex algorithm and difficult practical application of the prior art are solved.

[0009] The present application provides the following technical scheme: a multi-carrier NOMA system downlink user grouping and power allocation method, comprising the following steps.

[0010] Step 1: Obtain the instantaneous channel state information (CSI) of each user in the cell at the base station, and sort the channel gains of each user;

[0011] Consider the multi-carrier NOMA system downlink scenario, there is one base station and M users in the cell, the base station transmits information to the M users through N subcarriers, two users share the same subchannel to transmit information at the same time, M=2N. Assuming that the instantaneous CSI of all users can be obtained at the base station, the channel coefficient from the base station to the user is h, h represents Rayleigh fading, and the channel gain h 2 is arranged in ascending order.

[0012] Step 2: Group users according to the channel gain arrangement order in step 1, and the users in the same group share the same subcarrier to transmit signals;

[0013] Assuming that the total number of users M is even, the user grouping is divided into two steps. In the first step, the two users with the largest channel gain difference are grouped, that is, the user with the largest channel gain and the user with the smallest channel gain are paired in the sorted users, the user with the second largest channel gain and the user with the second smallest channel gain are paired, and the above process is repeated until the remaining 6 users in the middle. In the second step, when there are 6 users left in the middle, another matching method is used, that is, according to the original sorting, the 1st user and the 4th user are paired, the 2nd user and the 5th user are paired, and the 3rd user and the 6th user are paired, and the user grouping is completed. If M is odd, select one user to occupy one subcarrier, and the remaining users are grouped according to the above steps.

[0014] Step 3: The channel gain of the largest user in the group is equivalent to the channel gain of the subcarrier, and the power allocation between subcarriers is performed using a deep learning algorithm according to the equivalent channel gain on the subcarrier;

[0015] Comparing two user channel gains h n,1 2 and h n,2 2 The maximum channel gain is selected as the equivalent channel gain of the subcarrier n, and the equivalent channel gains of N subcarriers are obtained. The N equivalent channel gains are taken as the input of the neural network, and the output is the normalized power of each subcarrier. The neural network is a 3-layer fully connected network structure, the input layer and the output layer are both N neurons, the hidden layer is two layers, the first layer of the hidden layer has 15 neurons, the second layer of the hidden layer has 10 neurons, and the activation function is a sigmoid function. In order to obtain the normalized power of the subcarrier, a softmax layer is connected after the output layer. The loss function is the inverse of all subcarriers and rates, and the loss function is minimized by the back propagation algorithm, that is, the maximum sum rate.

[0016] Step 4: After obtaining the power on the subcarrier, the power of the superimposed user in the subcarrier that maximizes the sum rate is obtained according to the minimum transmission rate constraint;

[0017] After obtaining the power on each subcarrier, the power allocated to the two superimposed users on the subcarrier is solved. According to the set minimum transmission rate R min , that is, R n,1 ≥ R min , R n,2 ≥ R min , the power allocation coefficient of the two users that maximize R n =R n,1 +R n,2 is derived.

[0018] Compared with the prior art, the beneficial effects achieved by the user grouping and power allocation method of the multi-carrier NOMA system of the present application include:

[0019] Firstly, the user grouping method provided by the present application solves the shortcoming that users with similar or identical channel gains may be grouped in the user grouping method based on the maximum channel gain difference, ensuring the SIC performance. Secondly, compared with the traditional power allocation method, the deep learning power allocation algorithm proposed by the present application can significantly improve the system sum rate. Thirdly, according to the constraint condition of the minimum transmission rate, the closed-form solution of the power allocation of the superimposed users in the subcarrier is derived, ensuring the minimum transmission rate of the user. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The flowchart of the user grouping and power allocation method of the multi-carrier NOMA system of the present application is shown.

[0021] Figure 2 A multi-carrier NOMA system downlink model diagram.

[0022] Figure 3 A user grouping schematic diagram.

[0023] Figure 4 A subcarrier power allocation algorithm neural network structure diagram.

[0024] Figure 5 A subcarrier power allocation algorithm performance comparison diagram under different total transmit powers.

[0025] Figure 6 A subcarrier power allocation algorithm performance comparison diagram under different numbers of users. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only one of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0027] As can be seen from the drawings, the multi-carrier NOMA system downlink user grouping and power allocation method provided by the present application comprises the following steps, as shown in Figure 1 .

[0028] Step 1: Consider the multi-carrier NOMA system downlink scenario, as shown in Figure 2 . Assume that there is one base station and M users in the cell, the base station and each user are equipped with an antenna, the total bandwidth of the system is B0, the number of orthogonal subcarriers used for transmission is N, and the average allocated bandwidth of each subcarrier is B=B0 / N. In order to ensure the accuracy of the SIC at the receiving end and reduce the signal separation delay, only two user signals are superimposed on each subcarrier, M=2N. Assume that the transmission signal of user m on subcarrier n is xm n,m , n={1,2,…,N}, m={1,2}, the transmission power is pm n,m , then the superimposed signal xm n sent on subcarrier n is

[0029]

[0030] The superimposed signal xm n after channel transmission to the receiving end user m can be expressed as

[0031]

[0032] where w n,m represents the mean of 0, variance of σ n 2 of additive white Gaussian noise, h n,m represents the channel coefficient from base station to user m. At the receiving end, the signal received by the user includes its own useful signal and interference signal, in order to obtain its own signal, the receiving end needs to perform SIC technology. Without loss of generality, it is assumed that the channel gain of each user in each subcarrier is arranged in descending order, that is, h n,1 2 ≥ h n,2 2 . After SIC detection processing, according to Shannon formula, the information transmission rate of user 1 and user 2 on subcarrier n is respectively

[0033]

[0034]

[0035] The total transmission rate on subcarrier n is

[0036]

[0037] Maximizing the system sum rate and its constraint conditions are represented as

[0038]

[0039]

[0040] C1 represents the constraint of total transmission power of the system, C2 represents the constraint that the allocated power of the user is not less than 0, and C3 represents the constraint of the minimum rate acceptable by the user. Since this problem is a non-convex optimization problem and is NP-hard, it is very complex to directly solve the global optimal solution, therefore, a step-by-step optimization method is adopted to seek a suboptimal solution, so as to reduce the calculation complexity. First, the user grouping on the subcarrier is performed, and then the power allocation of the user is performed.

[0041] According to the CSI of each user obtained by the base station, the channel gain H = |h 2 is arranged in ascending order, that is, H1≤H2≤...≤H M .

[0042] Step 2: assuming that the total number of users M is even, the user grouping is performed according to the sorting result of the channel gain, such as Figure 3As shown. If M is odd, a single user can occupy a single subcarrier. User grouping is done in two steps. First, the two users with the largest channel gain differences are grouped together; that is, the user with the largest channel gain is paired with the user with the smallest, the user with the second largest channel gain is paired with the user with the second smallest, and so on. Second, for the remaining six users, another matching method is used. The first user is paired with the fourth user, the second user with the fifth user, the third user with the sixth user, and so on. This completes the user grouping process.

[0043] Step 3: After determining the user groups, perform power allocation. Based on the user grouping results in Step 2, compare the channel gains |h| of the two users within subcarrier n. n,1 | 2 and |h n,2 | 2 The maximum channel gain is selected as the equivalent channel gain |h| of the subcarrier n. n | 2 We obtain the equivalent channel gain for N subcarriers. Then, we divide the N equivalent channel gains |h n | 2 The input to the neural network is the normalized power of each subcarrier, and the output is the normalized power of each subcarrier. The neural network has a 3-layer fully connected network structure, such as... Figure 4 As shown. Both the input and output layers have N neurons, and there are two hidden layers: the first layer has 15 neurons, and the second layer has 10 neurons. The activation function is the sigmoid function. (Where x is the input to the activation function, and y is the output of the activation function). To obtain the normalized power allocated to each subcarrier, a softmax layer is connected after the original output layer of the neural network. and The normalized power β on each subcarrier is obtained in the final output layer of the neural network. n Then the power allocated to each subcarrier is P. n =β n P T The loss function is the reciprocal of the sum of all subcarriers and rates, as shown in Equation (8). The loss function is minimized by backpropagation, which is equivalent to the sum of rates maximizing. The optimizer is Adam, and the learning rate is set to 0.01.

[0044]

[0045] Step 4: Based on the power allocation results of each subcarrier obtained in Step 3, perform power allocation for the two superimposed users within the subcarrier. Assume that the channel conditions for user 1 on subcarrier n are better than those for user 2, i.e., |h n,1 | 2 ≥|h n,2 |2 According to the detection order of SIC, User 2 does not need to perform the SIC detection process, and directly demodulates the received signal by regarding the interference of User 1 to it as noise. User 1 needs to perform SIC, first demodulates the signal of User 2, then subtracts the demodulated signal of User 2 from the received signal, and finally demodulates the received signal of User 1. Therefore, the information transmission rate of User 1 is represented as

[0046]

[0047] The information transmission rate of User 2 is represented as

[0048]

[0049] When the power P n on the subcarrier n is determined, assuming that the power allocation coefficient of User 1 is α, the power p n,1 allocated to User 1 is αP n , and the power p n,2 allocated to User 2 is (1-α)P n . According to the power allocation principle of the NOMA system, the user with good channel condition is allocated less power, and the user with poor channel condition is allocated more power, so The total transmission rate of User 1 on the subcarrier n can be represented as

[0050]

[0051] The latter half of formula (12) is a certain value, let

[0052] then

[0053] Simplifying it obtains

[0054]

[0055] Since |h n,1 | 2 ≥ |h n,2 | 2 ≥ 0, f'(α) > 0, f(α) is a monotonically increasing function in the domain, so it is concluded that R n (α) is also a monotonically increasing function in the domain. According to the minimum transmission rate R min , that is, R n,1 ≥ R min and R n,2 ≥ R min , it is concluded that

[0056]

[0057] Therefore there is The power allocated to user 1 is P. n,1 =α * P n The power allocated to user 2 is P. n,2 =(1-α) * )P n .

[0058] To verify the effectiveness of the method of the present invention, the following comparative experiments are provided:

[0059] The system's total transmission bandwidth B0 is 10kHz, with M=20 users evenly distributed within the cell, N=10 subcarriers, and a noise power spectral density N0=-70dBm / Hz. The channel noise variance on each subcarrier is... The key parameter α of the FTPA method FTPA The key parameter α of the FPA method is set to 0.7. FPA Take 0.5.

[0060] Figure 5 The figures present a comparison of the system and rate outputs of the Deep Learning Power Allocation (DLPA), FTPA, and FPA methods under different total transmit power conditions and using the same user grouping method. The base station transmit power ranges from 5W to 10W. As can be seen from the figures, the sum rate of all three methods improves with increasing base station transmit power. The DLPA method in this embodiment achieves a significantly higher sum rate than the other comparison schemes, improving upon the FTPA and FPA methods by approximately 6.3 × 10⁻⁶. 3 bit / s and 2.1×10 4 bit / s.

[0061] Figure 6 A comparison chart of the sum rate generated by three power allocation methods under different numbers of users is presented. The total transmit power of the base station is 10W, and the number of users in the cell is 10, 20, 30, 40, and 50, respectively, with other parameters remaining unchanged. As the number of users increases, the number of subcarriers also increases. With the total system bandwidth remaining constant, the bandwidth occupied by each subcarrier decreases. Therefore, the sum rate of all three methods shows a decreasing trend, but the DLPA method in this embodiment still achieves the best sum rate.

[0062] For power allocation between two superimposed users within a subcarrier, this invention derives a closed-form expression for power allocation that maximizes the sum of the transmission rates while ensuring the minimum transmission rate for each user. Therefore, it can guarantee the quality of service for users, which is an advantage compared to other power allocation methods within a subcarrier.

[0063] The application discloses a multi-carrier NOMA system downlink user grouping and power allocation method. Multiple users share a subcarrier for simultaneous transmission of signals at the same frequency, and a receiving end demodulates user's own signals by using a successive interference cancellation (SIC) technology. Assuming that instantaneous channel state information (CSI) from a base station to each user can be acquired at the base station, the acquired CSI is used for channel gain sorting, and then user grouping is performed, and users in the same group share the same subcarrier for communication. The channel gain of the largest user in the group is equivalent to the channel gain of the subcarrier, and a deep learning algorithm is used for power allocation between subcarriers according to the equivalent channel gain of the subcarrier. After the power on the subcarrier is determined, power allocation of users in the subcarrier is performed according to a minimum transmission rate constraint condition. The application proposes a new user grouping method aiming at the shortcomings of a user grouping method based on the largest channel gain difference, and avoids grouping users with similar channel gains. A deep learning algorithm is used for power allocation of the NOMA system, and the maximum transmission rate can be realized under the condition of guaranteeing the minimum transmission rate of users.

[0064] The above only describes the preferred embodiments of the application. It should be noted that those skilled in the art can make some improvements and modifications without departing from the technical principles of the application, and these improvements and modifications should also be considered as the protection scope of the application.

Claims

1.A method for user grouping and power allocation in a multi-carrier NOMA system downlink, characterized in that, The method comprises the following steps: Step 1: acquiring instantaneous channel state information (CSI) of each user in a cell at a base station, and sorting channel gains of each user; Step 2: grouping users according to the channel gain arrangement order of step 1, and the users in the same group sharing the same subcarrier for transmitting signals; Step 3: the maximum user channel gain in the group is equivalent to the channel gain of the subcarrier, and a deep learning algorithm is used for power distribution among subcarriers according to the equivalent channel gain on the subcarriers; Step 4: after obtaining the power on the subcarriers, the power of the superimposed users in the subcarrier that maximizes the sum rate is obtained according to the minimum transmission rate constraint; The step 1 comprises: In a multi-carrier NOMA system downlink scenario, there is one base station and M users in a cell, the base station transmits information to the M users through N subcarriers, in order to reduce the decoding delay of the receiving end, two users share the same subcarrier to transmit information at the same time, M=2N, assuming that the base station obtains the instantaneous CSI of all users, the channel coefficient from the base station to the user is h, h represents Rayleigh fading, and the channel gain |h| is denoted as h 2 in ascending order; The step 2 specifically comprises: In step 2, the total number of users M is even, and the user grouping is divided into two steps, in the first step, the two users with the largest channel gain difference are grouped, among all the users arranged in ascending order of channel gain, the user with the largest channel gain is paired with the user with the smallest channel gain to form a group, the user with the second largest channel gain is paired with the user with the second smallest channel gain to form a second group, and the above process is repeated until the remaining six users in the middle, in the second step, when there are six users in the middle, the first user and the fourth user in the six users are paired, the second user and the fifth user are paired, and the third user and the sixth user are paired, and the user grouping is completed; if M is odd, one user is selected to occupy one subcarrier, and the remaining users are grouped according to the above steps. 2.The method of user grouping and power allocation for downlink of multi-carrier NOMA system according to claim 1, characterized in that The step 3 specifically comprises: Comparing the channel gains of two users in subcarrier n, |h n ,1 | 2 and |h n ,2 | 2 The maximum channel gain is selected as the channel gain of the subcarrier n, and the equivalent channel gains of N subcarriers are obtained. The equivalent channel gains of N subcarriers are used as the input of the neural network, and the output is the normalized power of each subcarrier. The neural network is a 3-layer fully connected network structure, the input layer and the output layer each have N neurons, the hidden layer is set to two layers, the first layer of the hidden layer has 15 neurons, the second layer of the hidden layer has 10 neurons, the activation function is a sigmoid function, the output layer is connected with a softmax layer to obtain the normalized power of each subcarrier, and the loss function is the reciprocal of the sum rate of all subcarriers. The neural network is trained by the back propagation algorithm to minimize the loss function and maximize the sum rate. 3.The method of user grouping and power allocation for downlink of multi-carrier NOMA system according to claim 1, characterized in that The step 4 specifically comprises: After the power on each sub-carrier is obtained, the power allocated to two superimposed users on the sub-carrier is solved, and the minimum transmission rate R of the user is set according to the set user min , that is, R n ,1 ≥ R min , R n ,2 ≥ R min , the power allocation coefficient of the two users that make R n = R n ,1 + R n ,2 is obtained.

Images