A non-orthogonal multiple access beamforming and power allocation method for UAV relay support

Abstract

The present invention proposes a non-orthogonal multiple access beamforming and power allocation method supported by drone relay. First, a non-orthogonal multiple access framework supported by drone relay is constructed, and the adverse effects of imperfect serial interference elimination in actual communication are considered. Then, a joint design algorithm of beamforming and power allocation for ground base stations is proposed, which alternately optimizes the beamforming and power allocation matrices, and can minimize the total power consumption of the base station while meeting the user service quality. Compared with the zero-forcing beam and fixed power allocation method, it effectively reduces the total power consumption at the base station and alleviates the adverse effects of user time-frequency resource reuse.

Description

Technical Field

[0001] The present invention belongs to the radio field of wireless communication technology, and specifically is a non-orthogonal multiple access beamforming and power allocation method supported by an unmanned aerial vehicle relay. Background Art

[0002] Unmanned aerial vehicle (UAV) communication refers to the wireless communication mode in which a UAV equipped with a base station provides ground users with wireless communication services or provides relay services for other UAVs. UAV communication can be divided into communication scenarios such as UAV air-to-ground (A2G), UAV air-to-air (A2A) and UAV air-to-ground integration. Compared with the complex ground channel environment, UAVs as base stations can establish better line-of-sight transmission channels in near space. The path gain between UAV base stations and users is high, with flexibility and mobility. They are widely used in emergency events such as fires, earthquakes, search and rescue, and load migration communication scenarios. At present, due to the limitation of the load of UAV base stations, it is urgent to solve the problem of limited airborne energy of UAV base stations. NOMA is used in UAV communication. The communication link is more likely to experience line-of-sight (LoS) propagation, reduce propagation attenuation, and provide access services with higher throughput for dense networks. In addition, cooperative communication, as a technology that uses relay nodes to combat channel path loss, fading and shadowing effects, has been widely used in UAV networks. It can not only improve the spectrum efficiency of the network, but also effectively expand the coverage of communication signals and enhance the reliability of communication systems.

[0003] The significance of massive MIMO-NOMA based on drone relay is:

[0004] 1. Higher spectrum efficiency: Compared with the impact on capacity caused by actively introducing interference between users, allowing different users to use the same information domain resources for data transmission has a greater effect on improving capacity.

[0005] 2. More user access: NOMA technology breaks the traditional orthogonal restrictions and makes it possible to serve more users on the same time-frequency code domain resources, thereby significantly improving the user access capability. At the same time, massive MIMO technology uses large antenna arrays to form multiple parallel high-speed data streams. Combined with NOMA, it can realize the reuse of massive user time-frequency resources.

[0006] 3. Stronger compromise between system throughput and user fairness: The "water injection theorem" power allocation method used in OFDM will allocate more power to users with better channel conditions, but in NOMA, users with poorer channel conditions will get more power. This method of improving the total system throughput while ensuring user fairness is more in line with the design goals of the actual system.

[0007] 4. Expanding the coverage of the network: The adaptability of drones in establishing communication links helps to improve the organizational flexibility and mobility of the network.

[0008] However, for massive MIMO-NOMA, the following requirements arise:

[0009] 1. The traditional NOMA system actively introduces interference through the transmitter, and the receiver uses serial interference cancellation technology to eliminate the interference. However, the interference cancellation problem becomes more complicated when combined with MIMO technology. In addition to the existing intra-cluster interference, the existence of different beams also requires consideration of inter-cluster interference. Therefore, it is necessary to study a precoding scheme suitable for Massive MIMO-NOMA. In addition, serial interference cancellation technology requires demodulation and judgment of user signals at the receiver. In dense access networks, imperfect serial interference cancellation cannot be ignored. Therefore, in order to ensure the effective improvement of quality of service (QoS) and the reasonable setting of system judgment conditions, how to properly allocate power is the focus of further research.

[0010] 2. In the NOMA system, both the source node and the relay need to reasonably allocate their power. The most commonly used power allocation schemes are static power allocation and dynamic power allocation. The static resource allocation method is no longer suitable for resource design under time-varying channels, while the dynamic power allocation requires processing of instantaneous channel state information, which has a large computational overhead. Therefore, inter-cluster interference can be reduced or even eliminated without adding additional wireless resources. If the base station (BS) has all the channel state information, the zero-forcing beam can completely suppress inter-cluster interference when the user is equipped with a single antenna. For multi-antenna NOMA downlinks, user clustering, beamforming, and power allocation are three effective methods to combat inter-cluster and intra-cluster interference. However, since all optimization parameters are coupled, it is a challenging problem to jointly design user clustering, beamforming, and power allocation schemes. Summary of the invention

[0011] The present invention proposes a non-orthogonal multiple access beamforming and power allocation method supported by an unmanned aerial vehicle relay.

[0012] The technical solution to achieve the purpose of the present invention is: a non-orthogonal multiple access beamforming and power allocation method supported by UAV relay, the specific steps are:

[0013] Step 1: Based on the needs of large-scale access and relay communication, a non-orthogonal multiple access large-scale access network supported by drone relay is constructed for ground base stations and ground users with communication interruptions;

[0014] Step 2: Based on the application mechanism of non-orthogonal multiple access, construct a probabilistic direct channel model and a user receiving signal model based on NOMA;

[0015] Step 3: Based on the techniques of Schur complement and semidefinite relaxation, the alternating iterative algorithm is used to solve the problem of minimizing the total power consumption in the MIMO-NOMA system, and the optimal solution of the beamforming and power allocation matrix of the constructed communication system is obtained.

[0016] Preferably, the non-orthogonal multiple access large-scale access network supported by the drone relay includes:

[0017] One equipment A ground base station with a uniform linear array of antennas, Q single-antenna users, and a A drone with a root antenna, the drone is used to relay messages from a ground base station; there is no direct channel between the ground base station and the ground user;

[0018] Divide Q users into M spatial clusters, and the mth cluster contains users, and , the kth user in the mth cluster is , ; Randomly generate a The unitary matrix , the matrix The mth column in As the basic direction vector of the m-th cluster, a cluster with the closest direction is selected for each user by comparing the M basic directions.

[0019] Preferably, based on the application mechanism of non-orthogonal multiple access, a probabilistic direct channel model and a NOMA-based user receiving signal model are constructed, specifically including:

[0020] Construct a wireless channel model for two-stage relay transmission, including:

[0021] Downlink channel vector between drone and user:

[0022]

[0023] In the formula, represents the small-scale fading vector, is the direct component, is the Ricean factor of the UAV to ground user link, is the large-scale fading coefficient;

[0024] Channel gain matrix from base station to drone:

[0025]

[0026] In the formula, is the large-scale fading per unit distance, is the distance from the ground base station to the drone, is the path loss factor corresponding to the distance from the ground base station to the UAV, , , is the distance from the i-th antenna of the ground base station to the n-th antenna of the drone, is the number of ground base station antennas, is the carrier wavelength, for A random phase shift uniformly distributed within .

[0027] Construct a base station transmission signal model and a user reception signal model, wherein the base station transmission signal model is specifically:

[0028]

[0029] Where M is the number of spatial clusters, is the base station transmit power, is the beamforming vector of the mth cluster, represents the signal of the kth user in the mth cluster, is the power allocation factor of the kth user in the mth cluster, is the number of users in the mth cluster;

[0030] The user receiving signal model includes:

[0031] In the first phase of communication, the base station sends a signal vector to the drone relay. , the drone receives the signal sent by the base station It is expressed as:

[0032]

[0033] In the formula, is the amount of additive Gaussian white noise at the drone, satisfying ;

[0034] In the second stage of communication, the drone uses amplification and forwarding to transmit the signal received from the base station to the ground user. The received signal at the user is expressed as:

[0035]

[0036] in, is the transmitting power of the UAV, is the downlink channel vector between the UAV and the user The conjugate transpose of is the Gaussian white noise at the kth user in the mth cluster.

[0037] Performing serial interference cancellation techniques on terrestrial users in the same cluster to identify signals with residual interference at the user and the signal-to-interference-to-noise ratio at the receiving end , specifically:

[0038]

[0039]

[0040] In the formula, is the imperfect successive interference cancellation (SIC) coefficient, represents the number of users in the jth cluster, The number of clusters to be collected.

[0041] Preferably, the total power consumption minimization problem in the MIMO-NOMA system is specifically expressed as:

[0042]

[0043] In the formula, C1 represents the QoS requirement of all ground users, C2 represents the power allocation principle that users in the cluster need to meet when performing serial interference cancellation (SIC), and C3 is the power allocation limit of users in the cluster. is the average achievable rate of the user, is the minimum average achievable rate of the kth user in the mth cluster, , is the power allocation factor vector of the mth cluster, is the total power consumption.

[0044] Preferably, based on technologies such as Schur complement and semidefinite relaxation, an alternating iterative algorithm is used to solve the problem of minimizing the total power consumption in the MIMO-NOMA system, and a specific method for obtaining the optimal solution of the beamforming and power allocation matrix of the constructed communication system is:

[0045] Problem P1 is decomposed into two sub-problems: beamforming design and power allocation. The two sub-problems are combined and optimized alternately to obtain the suboptimal solution. The total power consumption is calculated through multiple iterations. Satisfy the convergence conditions, including:

[0046] 1) Beamforming optimization design:

[0047] The power allocation factor As a fixed value to optimize the beamforming matrix , transform the C1 constraint in P1 as follows:

[0048]

[0049] in, .

[0050] Introduction As slack variables, the transformed C1 constraint is transformed into the following two inequality constraints:

[0051] The first inequality constraint:

[0052] The second inequality constraint:

[0053] in, , ;

[0054] The first inequality constraint is transformed using semi-definite relaxation technology, specifically:

[0055] definition , , changing the square term in the expression of the expected signal power within the cluster, we get: ;in represents the trace of the matrix, the matrix The restriction of rank=1 should be satisfied. After relaxation, the first inequality constraint is transformed into:

[0056]

[0057] Constructing a complementary matrix

[0058] in, , the second inequality constraint is reformulated as:

[0059]

[0060] Applying Schur's complement theorem to the re-expressed second inequality constraint, it is equivalent to a linear matrix inequality:

[0061]

[0062] in,

[0063] For beamforming matrix The power consumption minimization problem is transformed into:

[0064]

[0065] in, is the beamforming matrix of the mth cluster;

[0066] Use CVX to solve problem P2 and recover the beamforming vectors using the Gaussian random approximation method. ;

[0067] 2) Optimization design of power allocation factor

[0068] The beamforming matrix Set to a fixed value, then the power allocation factor Optimize, specifically:

[0069] The constraint C1 is transformed as follows:

[0070]

[0071] in is the coefficient vector of the user power allocation factor, satisfying:

[0072]

[0073] The power allocation factor optimization problem is expressed as: The optimal solution to the P3 problem is expressed as:

[0074] .

[0075] Compared with the prior art, the present invention has the following significant advantages: The present invention provides a novel non-orthogonal multiple access method for UAV empowerment. Compared with the existing beamforming and power control technologies, the present invention can effectively reduce the total power consumption at the base station and meet the service quality requirements of ground users; the present invention takes into account the situation of imperfect SIC, which is more in line with actual large-scale access scenarios, and has better adaptability and robustness than the existing transmission technology in MIMO-NOMA.

[0076] The present invention is described in further detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 Schematic diagram of massive MIMO-NOMA supported by drone relay.

[0078] Figure 2 It is a flow chart for solving the problem of minimizing the total power consumption in the MIMO-NOMA system using an alternating iterative algorithm. DETAILED DESCRIPTION

[0079] A non-orthogonal multiple access beamforming and power allocation method supported by UAV relay can achieve efficient large-scale access, which is suitable for emergency communication or urban hot spots with interrupted ground connection, including the following steps:

[0080] Step 1: Based on the needs of large-scale access and relay communication, a non-orthogonal multiple access large-scale access network supported by drone relay is constructed for ground base stations and ground users with communication interruptions.

[0081] In a further embodiment, the non-orthogonal multiple access large-scale access network supported by the drone relay includes:

[0082] One equipment There is a ground base station with a uniform linear array of antennas and Q ground users (GUEs) with single antennas. There is no direct channel between the ground base station and the ground user, such as emergency communications and urban hotspots. The UAV with a root antenna acts as a relay to forward messages from the ground base station. The UAV is located at a fixed coordinate position, and all ground users are randomly distributed in a ring area, such as Figure 1 As shown in the figure. In large-scale access networks, the number of terrestrial users is usually large, so MIMO-NOMA is used in combination with beamforming to support massive access needs. With respect to UAVs, it is assumed that users with similar spatial directions share a transmit beam. The purpose of this is to improve channel gain and alleviate user interference between clusters.

[0083] Divide Q users into M spatial clusters, and the mth cluster contains users, and , the kth user in the mth cluster is , . Randomly generate a The unitary matrix , then the matrix The mth column in As the basic direction vector of the mth cluster. By comparing the M basic directions, a cluster with the closest direction is selected for each user. Users within a cluster have different large-scale fading coefficients. The non-orthogonal multiple access large-scale access network supported by drone relay uses SIC in NOMA technology to eliminate the interference of users within the cluster caused by power reuse.

[0084] Step 2: Based on the application mechanism of non-orthogonal multiple access, construct a probabilistic direct channel model and a user receiving signal model based on NOMA, specifically:

[0085] Due to the limitation of MIMO spatial domain freedom, it is assumed that the number of antennas of the ground base station is not less than the number of antennas of the UAV, and the number of antennas of the UAV is greater than the number of user clusters, that is, It should be pointed out that due to the resource reuse of NOMA, the number of users can exceed the number of antennas of the base station. Obtaining complete channel state information is the premise of user clustering and beam design. Based on the existing channel estimation method, the present invention sets the system to be time division duplex, and assumes that the base station has the channel state information of the entire communication link.

[0086] 1) Constructing a wireless channel model for two-stage relay transmission

[0087] In the UAV-enabled massive MIMO-NOMA system under consideration, an effective model needs to be established to characterize the propagation channels from UAV to ground users and from UAV to base stations. First, consider the propagation link from UAV to ground users: As an aerial device, UAV can establish good line-of-sight propagation conditions with GUEs. This paper adopts the Rician probability density function to describe the A2G propagation channel, assuming For UAV and The downlink channel vector between is expressed as:

[0088] (1)

[0089] in, Represents the small-scale fading vector, satisfying ;

[0090] is the direct component;

[0091] in, For The random phase shifts are uniformly distributed inside. is the carrier wavelength, For the nth antenna of UAV The straight-line distance. The Rice factor representing the UAV-to-ground user link depends on the probability of the LoS link occurring:

[0092] (2)

[0093] in Altitude from UAV , the relative distance between nodes and the environment, defined as:

[0094] (3)

[0095] in, , , Represents UAV and The horizontal distance of the NLoS link is . is the large-scale fading coefficient:

[0096] (4)

[0097] In the formula, represents the free space path loss, is the carrier frequency, The standard deviation is Shadow fading component.

[0098] Next, consider the G2A propagation channel from the base station to the drone. Here, it is assumed that the ground base station antenna is higher than the ground environment, and the LoS component occupies the dominant component of the channel. Therefore, the channel gain matrix from the base station to the drone can be expressed as:

[0099] (5)

[0100] in, is the large-scale fading per unit distance, is the distance between the ground base station and the UAV, is the corresponding path loss factor, , , is the distance from the i-th antenna of the ground base station to the n-th antenna of the UAV.

[0101] 2) Construct a model of base station transmission signals and user reception signals

[0102] The BS constructs the transmission signal vector through superposition coding and beamforming, and defines the beamforming matrix at the BS as , in is the beamforming vector of the mth cluster, and the transmitted signal vector at the BS after superposition coding and beamforming is It can be expressed as:

[0103] (6)

[0104] in, is the BS transmit power, represents the signal of the kth user in the mth cluster, is the power allocation factor of the kth user in the mth cluster, satisfying: , is the power allocation factor vector of the mth cluster.

[0105] In the first stage of communication, the base station first sends a signal vector to the drone relay. , the signal sent by the base station BS received by the UAV is expressed as:

[0106] (7)

[0107] is the additive white Gaussian noise AWGN vector at the UAV, satisfying .

[0108] In the second stage of communication, the drone uses amplification and forwarding to transmit the signal received from the base station to the ground user. The received signal at can be expressed as:

[0109] (8)

[0110] in, is the transmitting power of the UAV, for AWGN at.

[0111] 3) Performing serial interference cancellation techniques among terrestrial users in the same cluster

[0112] Arrange the terrestrial users in the mth cluster in descending order of channel gain: , users with stronger channel gain will be assigned lower transmit power: , to ensure efficient SIC in downlink PD-NOMA. However, in practice, due to hardware errors and coherent pilot contamination, users with stronger channel gain cannot completely eliminate the information of weaker users in the cluster, resulting in SIC residual interference. The coefficient is used Modeling residual interference after imperfect SIC The signal with residual interference at can be expressed as:

[0113] (9)

[0114] The signal-to-interference-to-noise ratio at the receiving end is written as:

[0115] (10)

[0116] Step 3: Based on the techniques of Schur complement and semidefinite relaxation, the alternating iterative algorithm is used to solve the problem of minimizing the total power consumption in the MIMO-NOMA system, and the optimal solution of the beamforming and power allocation matrix of the constructed communication system is obtained. The specific method is as follows:

[0117] The beamforming matrix and power allocation factor are jointly optimized to meet the QoS requirements of all ground users while minimizing the total system power consumption. Therefore, the optimization problem of minimizing the total power consumption in the massive MIMO-NOMA system can be expressed as:

[0118] (11)

[0119] in, , for The average achievable rate (bits / Hz), C1 represents the QoS requirement of all ground users, for Minimum average achievable rate. C2 indicates that users in the cluster need to meet the power allocation principle when executing SIC, and C3 is the power allocation limit for users in the cluster.

[0120] The present invention decomposes the original problem into two sub-problems: beamforming design and power allocation. The two sub-problems are combined and optimized alternately to obtain a suboptimal solution, and the total power consumption is calculated through multiple iterations. Satisfy the convergence condition.

[0121] 1) Beamforming optimization:

[0122] First, the power allocation factor Treated as fixed values ​​to optimize the beamforming matrix , transform the C1 constraint in P1 as follows:

[0123] (12)

[0124] in, .

[0125] Note that constraint (12) is non-convex, so we introduce As slack variables, constraint (12) can be transformed into the following two inequality constraints:

[0126] (13)

[0127] (14)

[0128] in, , .

[0129] The non-convex constraint (13) is transformed using the semi-definite relaxation technique and defined as: , , changing the square term in the expression of the expected signal power within the cluster, we can get: .in represents the trace of a matrix, the matrix The restriction of rank=1 should be satisfied. After relaxation, constraint (13) can be transformed into:

[0130] (15)

[0131] Next, we deal with the quadratic term in the expression of inter-cluster interference power in constraint (14). First, we construct the complementary matrix ,in, , constraint (14) can be reformulated as:

[0132] (16)

[0133] Applying Schur's complement theorem to inequality (16), it can be equivalently transformed into a linear matrix inequality:

[0134] (17)

[0135] in, Through the above transformation, for the beamforming matrix The power consumption minimization problem can be transformed into:

[0136] (18)

[0137] in, Representation Matrix is semi-positive definite. After the above changes, problem (P2) is a standard convex optimization problem and can be solved using CVX. However, the rank of the optimal value of problem (P2) after SDR is usually not equal to 1, which means that the optimal value of problem (P2) is the lower bound of (P3). Therefore, an additional step is required to start from Mid-build The Gaussian random approximation method can quickly recover the beamforming vector .

[0138] 2) Power allocation factor optimization

[0139] In this subproblem, the beamforming matrix Set it to a fixed value, and then adjust the power allocation factor To optimize, first transform constraint C1 as follows:

[0140] (19)

[0141] in is the coefficient vector of the user power allocation factor, satisfying:

[0142] (20)

[0143] Therefore, the power allocation factor optimization problem can be expressed as:

[0144] (twenty one)

[0145] Lemma 1: The optimal solution to the P3 problem can be expressed as:

[0146] (twenty two)

[0147] Proof: Due to the power allocation factor There are multiple linear constraints, and the objective function is also linear, so (P3) is a linear programming problem. First, construct the Lagrangian function of problem (P3):

[0148] (twenty three)

[0149] in, are the Lagrange multipliers corresponding to the constraints respectively. Applying the KKT condition, we can get:

[0150] (twenty four)

[0151] (25)

[0152] By observation, it can be concluded that only When Equation (24) holds, then based on the KKT condition, After rearrangement, we can get formula (22).

[0153] Based on equation (22) and the equivalent channel gain, the power allocation factors of other users can be calculated in turn. In addition, For the infeasible solutions that may appear in the solution of problem (P3), some recursive algorithms, such as the obstacle method, can be used to find a solution that satisfies all constraints through repeated iterations.

[0154] Based on the above solution steps, the present invention proposes an alternating iterative optimization beamforming and power allocation algorithm, which is summarized in Algorithm 1. Specifically, given a feasible initial beam and power allocation factor, one of the variables is fixed to optimize the other, and the subproblems (P2) and (P3) are iteratively solved and the total power consumption of the system is updated until it converges or reaches the maximum number of iterations. The initial beam matrix Obtained through the maximum ratio transmission criterion, in order to ensure fairness among users within the cluster, the initial power allocation factor is defined as: .

[0155] The embodiments are described in more detail.

[0156] Algorithm 1 Beamforming and power allocation joint design algorithm:

[0157] enter: The feasible value, convergence coefficient ,calculate ;

[0158] Repeat the following steps:

[0159] 1: Solve problem P2 and get and , recovered by Gaussian randomization method ;

[0160] 2: Based on , solving problem P3 we get ;

[0161] 3: Update , make .

[0162] Repeat step termination condition: or ;

[0163] Output: and .

Claims

1. A non-orthogonal multiple access beamforming and power allocation method supported by UAV relay, characterized in that: The specific steps are: Step 1: Based on the needs of large-scale access and relay communication, a non-orthogonal multiple access large-scale access network supported by drone relay is constructed for ground base stations and ground users with communication interruption. The non-orthogonal multiple access large-scale access network supported by drone relay includes: One equipment A ground base station with a uniform linear array of antennas, Q single-antenna users, and a A drone with a root antenna, the drone is used to relay messages from a ground base station; there is no direct channel between the ground base station and the ground user; Divide Q users into M spatial clusters, and the mth cluster contains users, and , the kth user in the mth cluster is , ; Randomly generate a The unitary matrix , the matrix The mth column in As the basic direction vector of the mth cluster, a cluster with the closest direction is selected for each user by comparing the M basic directions; Step 2: Based on the application mechanism of non-orthogonal multiple access, construct a probabilistic direct channel model and a user receiving signal model based on NOMA; Step 3: Based on the techniques of Schur complement and semidefinite relaxation, the alternating iterative algorithm is used to solve the problem of minimizing the total power consumption in the MIMO-NOMA system, and the optimal solution of the beamforming and power allocation matrix of the constructed communication system is obtained.

2. The non-orthogonal multiple access beamforming and power allocation method supported by UAV relay according to claim 1, characterized in that: Based on the application mechanism of non-orthogonal multiple access, a probabilistic direct channel model and a NOMA-based user receiving signal model are constructed, including: Construct a wireless channel model for two-stage relay transmission, including: Downlink channel vector between drone and user: ， In the formula, represents the small-scale fading vector, is the direct component, is the Ricean factor of the UAV-to-ground user link, is the large-scale fading coefficient; Channel gain matrix from base station to drone: ， In the formula, is the large-scale fading per unit distance, is the distance from the ground base station to the drone, is the path loss factor corresponding to the distance from the ground base station to the UAV, , , is the distance from the i-th antenna of the ground base station to the n-th antenna of the drone, is the number of ground base station antennas, is the carrier wavelength, for Random phase shifts uniformly distributed within; Construct a base station transmission signal model and a user reception signal model, wherein the base station transmission signal model is specifically: ， Where M is the number of spatial clusters, is the base station transmit power, is the beamforming vector of the mth cluster, represents the signal of the kth user in the mth cluster, is the power allocation factor of the kth user in the mth cluster, is the number of users in the mth cluster; The user receiving signal model includes: In the first phase of communication, the base station sends a signal vector to the drone relay. , the drone receives the signal sent by the base station It is expressed as: ， In the formula, is the additive white Gaussian noise vector at the drone, satisfying ; In the second stage of communication, the drone uses amplification and forwarding to transmit the signal received from the base station to the ground user. The received signal at the user is expressed as: ， in, is the transmitting power of the UAV, is the downlink channel vector between the UAV and the user The conjugate transpose of is the Gaussian white noise at the kth user in the mth cluster; Performing serial interference cancellation techniques on terrestrial users in the same cluster to identify signals with residual interference at the user and the signal-to-interference-to-noise ratio at the receiving end , specifically: ， ， In the formula, is the imperfect serial interference cancellation coefficient, represents the number of users in the jth cluster, The number of clusters to be collected.

3. The non-orthogonal multiple access beamforming and power allocation method supported by UAV relay according to claim 2, characterized in that: The total power consumption minimization problem in the MIMO-NOMA system is specifically expressed as: ， Where C1 represents the QoS requirement of all ground users, C2 represents the power allocation principle that users within the cluster need to meet when executing SIC, and C3 represents the power allocation limit of users within the cluster. is the average achievable rate of the user, is the minimum average achievable rate of the kth user in the mth cluster, , is the power allocation factor vector of the mth cluster, is the total power consumption.

4. The non-orthogonal multiple access beamforming and power allocation method supported by UAV relay according to claim 3, characterized in that: Based on the techniques of Schur complement and semidefinite relaxation, the alternating iterative algorithm is used to solve the problem of minimizing the total power consumption in the MIMO-NOMA system. The specific method for obtaining the optimal solution of the beamforming and power allocation matrix of the constructed communication system is as follows: Problem P1 is decomposed into two sub-problems: beamforming design and power allocation. The two sub-problems are combined and optimized alternately to obtain the suboptimal solution. The total power consumption is calculated through multiple iterations. Satisfy the convergence conditions, including: 1) Beamforming optimization design: The power allocation factor As a fixed value to optimize the beamforming matrix , transform the C1 constraint in P1 as follows: ， in, ; Introduction As slack variables, the transformed C1 constraint is transformed into the following two inequality constraints: The first inequality constraint: , The second inequality constraint: , in, , ; The first inequality constraint is transformed using semi-definite relaxation technology, specifically: definition , , changing the square term in the expression of the expected signal power within the cluster, we get: ;in represents the trace of the matrix, the matrix The rank=1 restriction should be satisfied. After relaxation, the first inequality constraint is transformed into: ， Constructing a complementary matrix , in, , the second inequality constraint is reformulated as: ， Applying Schur's complement theorem to the re-expressed second inequality constraint, it is equivalent to a linear matrix inequality: ， in, , For beamforming matrix The power consumption minimization problem is transformed into: ， in, is the beamforming matrix of the mth cluster; Use CVX to solve problem P2 and recover the beamforming vectors using the Gaussian random approximation method. ; 2) Optimization design of power allocation factor The beamforming matrix Set to a fixed value, then the power allocation factor Optimize, specifically: The constraint C1 is transformed as follows: ， in is the coefficient vector of the user power allocation factor, satisfying: ， The power allocation factor optimization problem is expressed as: ， The optimal solution to the P3 problem is expressed as: 。

Images