Joint optimization method of UAV trajectory and time based on non-orthogonal multiple access technology

The joint optimization of UAV trajectory, time, and resource allocation in NOMA-UAV networks using clustering and iterative convex optimization addresses the challenges of energy efficiency and interference, enhancing system throughput and spectral efficiency.

CN114501647BActive Publication Date: 2024-12-06DALIAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210030227.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-12
Publication Date
2024-12-06
Estimated Expiration
2042-01-12

AI Technical Summary

Technical Problem

The challenge of efficiently allocating resources in non-orthogonal multiple access (NOMA) unmanned aerial vehicle (UAV) communication networks is compounded by limited energy capacity and significant user interference, making it difficult to maximize system throughput due to coupled, non-convex variables.

Method used

A joint optimization method for UAV trajectory and time allocation using K-means clustering, genetic algorithms, and alternating iterative convex optimization to optimize transmission power, hovering position, and communication time, ensuring user service quality while maximizing system throughput.

Benefits of technology

The method effectively maximizes system throughput by optimizing resource allocation, ensuring user communication requirements are met and improving spectral efficiency in NOMA-UAV networks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114501647B_ABST
    Figure CN114501647B_ABST
Patent Text Reader

Abstract

The present invention provides a method for jointly optimizing the trajectory and time of a drone based on non-orthogonal multiple access technology, and belongs to the field of resource optimization of drone downlink communication. The drone serves multiple users as a single-antenna mobile base station, and these users can be divided into multiple user clusters. Starting from a starting point, the drone passes through each user group in turn and performs hovering communication above them. In order to improve spectrum efficiency, the drone uses non-orthogonal multiple access technology to serve all users. On the basis of meeting the communication requirements of each user, resources are optimized to maximize the total throughput of the system. Based on the model, the present invention is a design method for jointly optimizing the transmission power, hovering position and working time of the drone, which can meet the communication requirements of the user and maximize the throughput of the system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of resource allocation optimization of unmanned aerial vehicle multi-user wireless communication networks, and relates to a joint allocation strategy of power, unmanned aerial vehicle positions and working time of an unmanned aerial vehicle communication network with multiple user groups. Background Art

[0002] UAV communication has attracted much attention as an important supplement to current ground communication due to its flexible configuration, high mobility, low cost and portability. UAVs can provide emergency communication services when the ground network fails partially or completely. In fact, UAV-assisted wireless communication is a very promising solution that can provide wireless connections for devices that lack infrastructure or whose infrastructure is damaged. At the same time, the air-to-ground communication channel of the UAV can be approximated as a high-quality line-of-sight link (LoS). The UAV can act as a base station or relay to provide short-term high-speed wireless connections for ground users. In addition, compared with communications based on fixed infrastructure, the flexible location deployment of UAVs can further improve the performance of the system. Therefore, UAVs have great development potential in future communication systems.

[0003] On the other hand, with the development of communication technology, communication resources are becoming increasingly scarce. Non-orthogonal multiple access (NOMA) has attracted great attention from academia and industry in the fifth generation (5G) and later mobile communication networks because it can provide high spectrum utilization, large-scale connectivity and user fairness. The basic idea of ​​NOMA is to serve multiple users in the same orthogonal resource block in a non-orthogonal domain. For example, power domain NOMA uses the different channel conditions of multiple users to allocate different sizes of transmission power to them and provide services for them on the same orthogonal resource block. At the receiving end, serial interference cancellation technology (SIC) is used to gradually eliminate interference and gradually separate signals from different users. Based on the above advantages, using NOMA to assist drones in wireless communications is a promising solution. Compared with traditional orthogonal multiple access technology (OMA), NOMA can bring higher gain to drone communication systems.

[0004] However, the energy that a drone can carry is very limited, and since drones work in the air, it is difficult to recharge their energy, so it is necessary to reasonably configure system resources to utilize these energies more efficiently. At the same time, since the users of NOMA services use the same orthogonal resource block, there is serious mutual interference between them, so in the NOMA system, special attention must be paid to power allocation. Therefore, resource allocation has always been a difficult problem in NOMA-UAV communication networks. This is because the variables in this problem are generally coupled with each other, and the problem is usually non-convex, making it difficult to solve directly. Based on the above problems, the present invention considers a NOMA-UAV communication system with multiple user groups, and maximizes the total system throughput by optimizing the resources in the system. Summary of the invention

[0005] The present invention provides a new resource allocation strategy to maximize the system throughput. In a non-orthogonal multiple access UAV network with K users, the K users can be divided into M user groups, each with N m The drone is used as a single-antenna mobile base station to serve ground users. The drone starts from the starting point, passes through each user group according to the optimal flight route and hovers above it. In order to avoid the Doppler effect, communication is only carried out when the drone is hovering. In order to achieve higher spectrum efficiency and access volume, the drone uses NOMA technology during the communication process. On the basis of meeting the service quality of each user, the transmission power, hovering position and working time of the drone are optimized to maximize the total throughput of the system.

[0006] In order to achieve the above objectives, the technical solution adopted by the present invention to solve the technical problem is as follows:

[0007] A method for jointly optimizing the trajectory and time of a UAV based on non-orthogonal multiple access technology comprises the following steps:

[0008] The first step is to build a system model:

[0009] 1) If Figure 1 In a UAV communication network, a rotorcraft UAV equipped with a single antenna serves as a base station to serve K ground users. These K users can be divided into M user groups, and the set of user groups is represented by Λ={1,2,…,M}. Each user group includes N m users, with Γ m ={1,2,…,N m}, m∈Λ represents the set of users in the mth user group. The drone starts from the starting point, passes through each user group according to the optimal flight route and hovers above it to communicate based on NOMA. In order to achieve the maximum throughput of the system, the transmission power, hovering position and working time of the drone are jointly optimized on the basis of satisfying the service quality of each user.

[0010] 2) In the system model, bold letters represent matrices or sets, and non-bold letters represent scalars; Represents a matrix space with a dimension of 2×1; where Θ={Γ m |Γ m ={1,2,…,N m},m∈Λ} represents the set of all users.

[0011] 3) The total operational time T0 of the drone can be divided into the flight time T S And the hovering communication time, the hovering communication time of the UAV above the mth user group is expressed as τ m , we can get:

[0012]

[0013] Define the nth user in the mth user group as U m,n , set the fixed working height of the drone to H0, and the horizontal hovering position of the drone above the mth user group and the position of the user are respectively and Indicates that user U m,n The distance to the drone can be expressed as:

[0014]

[0015] Without loss of generality, assume that the distance between the user and the drone in each user group satisfies the following relationship:

[0016]

[0017] Since the UAV air-to-ground channel can be regarded as a LoS link, user U m,n The channel coefficient h m,n | 2 It can be expressed as:

[0018]

[0019] Where ρ0 represents the channel gain at unit distance.

[0020] According to NOMA, users with better channel quality will be allocated less transmission power.

[0021]

[0022] Where P m,n Represents user U m,n transmission power.

[0023] For User U m,n ,2≤n∈Γ m , the signal to interference plus noise ratio (SINR) is:

[0024]

[0025] in, Represents user U m,n The signal of user U m,w SINR at the receiver; Indicates U m,n SINR when decoding its own signal; σ 2 Represents the noise power.

[0026] In particular, when n=1, user U m,1 The SINR can be expressed as:

[0027]

[0028] In order to successfully achieve serial interference cancellation, the user's SINR needs to meet the following requirements:

[0029]

[0030] Among them, η m,n Represents user U m,n Minimum SINR requirement. Then the user's achievable rate is expressed as:

[0031]

[0032] In the second step, according to the specific settings of the first step, determine the objective function and list the optimization problem:

[0033] For the convenience of description, the transmission power of the user is defined here. The set of the hovering position and the hovering time of the drone is P = {P m,n |n∈Γ m ,m∈Λ},L={L m |m∈Λ} and T={τ m |m∈Λ}. When the drone is hovering for communication, the user's service quality requirements need to be met. Therefore, the optimization goal of the present invention is to maximize the total throughput of the system, and the optimization problem can be expressed as:

[0034]

[0035]

[0036]

[0037] R m,n τ m ≥δ m,n , (10d)

[0038]

[0039]

[0040]

[0041] where δ m,n Is user U m,n The communication requirements of P sum is the total transmission power of the UAV.

[0042] The third step is to design an algorithm to solve the optimization problem:

[0043] In the optimization problem (P1), the variables are coupled together and the optimization problem is non-convex, which is difficult to solve directly. Therefore, a solution is proposed to simplify the problem (P1). First, the K-means algorithm is used to divide K users into M user groups, and then the genetic algorithm is used to find the optimal route. Finally, an alternating iterative algorithm is proposed to optimize P, L and T. The specific process is:

[0044] 1) Optimization of user clustering

[0045] In clustering problems, the K-means algorithm has been widely used because of its advantages of simple principle and fast convergence speed. Therefore, the present invention uses the K-means algorithm to cluster users. The specific process is as follows:

[0046] 1.1) First randomly select the positions of M users as the center points of the initial M clusters, and then assign each of the remaining users to the cluster whose center position is closest to the user.

[0047] 1.2) Recalculate the center point μ of the cluster based on the division results m , which is calculated as follows:

[0048]

[0049] 1.3) Re-divide the users according to the center position calculated in 1.2).

[0050] Repeat the above two steps until the center position of each user group does not change. In the above process, the objective function of the algorithm is square error and It can be expressed as:

[0051]

[0052] Generally speaking, the smaller the function value, the better the clustering effect. The curve with increasing M can determine the number of clusters M based on the elbow rule. Through the above process, the number of user clusters can be determined, that is, the optimal grouping of drone users.

[0053] 2) Optimal route optimization

[0054] After obtaining the grouping of drone users, the centroid of each user group is set as the initial hovering position of the drone. Next, we need to find the optimal route for the drone to fly. This problem can be converted into a traveling salesman problem (TSP). Generally speaking, the best one can be found through exhaustive search. The complexity of this method is the factorial of the number of clusters, that is, O(M!). It can be seen that when the number of user groups increases, the algorithm complexity of this method will also increase sharply. Therefore, in order to cope with the situation of more user groups, a genetic algorithm is used to find the optimal route G and the corresponding shortest total distance S. min , where G is a (M+1)×1 matrix that stores the current optimal route. The specific steps are as follows:

[0055] 2.1) Encode the problem. Define the take-off position of the drone as 0. In this problem, the gene of each individual in the genetic algorithm represents the order in which the drone passes through the user group. Set the total number of individuals in each generation of the genetic algorithm to Z. Define the set of individuals in the tth generation to be Ω t ={1,2,…,Z}.

[0056] 2.2) Calculate the fitness value of the individual and define the fitness value ψ of the zth individual in the tth generation t (z) is:

[0057]

[0058] in, represents the longest total distance in the tth generation, is the shortest total distance, λ t (z) is the total distance corresponding to the zth individual. ε1 is a constant greater than 0, which is used to ensure that the denominator of the fitness function is not equal to zero. The larger the value of the fitness function, the more the corresponding individual meets our requirements. Find the best individual of the tth generation and compare it with the current best individual. If the best individual is better, update the current best route and the shortest total distance.

[0059] 2.3) The roulette wheel method is used to select the parent generation for the next generation of the population. The first step is to calculate the cumulative probability of the zth individual in the tth generation.

[0060]

[0061] Randomly generate a number ε2∈(0,1], if the random number satisfies Then the zth individual is selected as the parent.

[0062] 2.4) Genetic or mutation to produce the next generation. Genetic algorithms require a certain mutation probability to prevent falling into the local optimal solution. The mutation probability is defined as Randomly generate a random number ε3∈(0,1), if the number satisfies Then select mutation to generate the next generation, otherwise use genetic method. Repeat 2.3) and 2.4) until enough next generation individuals are obtained.

[0063] Repeat the above steps until a sufficient number of generations are reached. The optimal flight route and the shortest total distance of the UAV can be obtained through the above genetic algorithm.

[0064] 3) Alternately iterate and optimize P, L and T.

[0065] Through the above method, the user grouping and the UAV flight route are determined. In addition, in order to prevent the total hovering communication time from being reduced when optimizing the hovering position of the UAV, the following constraints are introduced:

[0066]

[0067] Among them, L G(j) represents the horizontal position of the drone hovering above the G(j)th user group. Therefore, problem (P1) can now be transformed into (P2) to optimize P, L and T:

[0068]

[0069] sqF m,n τ m ≥δ m,n , (16b)

[0070]

[0071]

[0072]

[0073]

[0074]

[0075] In (P2), R m,nis a non-convex and non-concave function, so constraints (16b) and (16c) are both non-convex constraints. (P2) is a non-convex problem as a whole and the variables P, L and T are coupled together, which makes the problem difficult to be solved directly. Based on the idea of ​​alternating iteration, the present invention divides the problem into three sub-problems, uses convex optimization theory and convex approximation technology to convert each non-convex sub-problem into a convex problem, and finally proposes an iterative algorithm to alternately optimize the three sub-problems to solve the problem (P1).

[0076] 3.1) Transmission power P optimization

[0077] When optimizing the power P, L and T remain unchanged, and the problem (P2) can be decomposed into:

[0078]

[0079] sqF m,n τ m ≥δ m,n , (17b)

[0080]

[0081]

[0082]

[0083] In this problem, the optimization objective and constraints (17b) and (17c) are both non-convex and non-concave. To solve this problem, the present invention makes an approximation based on the concave-convex process theory and converts them into convex constraints. Specifically, we first use the properties of the logarithmic function to convert R m,n Expand it into the form of two concave functions subtracted, and then use the first-order Taylor expansion to approximate it as a concave function, so that (17b) is converted into a convex constraint. For (17c), rearrange it to separate the variables to one side of the inequality, so that (17c) is also converted into a convex constraint. In summary, problem (P3) is approximated as a convex problem, which can be solved using the convex optimization toolbox CVX.

[0084] 3.2) Hover position L optimization

[0085] When optimizing the hovering position L, P and T remain unchanged, so the problem can be transformed into:

[0086]

[0087] sqF m,n τ m ≥δ m,n , (18b)

[0088]

[0089]

[0090] In (P4), constraints (17b) and (17c) are both non-convex and non-concave. First, the convex function ||q m,n -L m || 2 As a variable, the same is true for R m,n Using the properties of the logarithmic function, it is converted into a convex function minus a convex function. The first convex function is converted into a concave function about the variable L using the first-order Taylor expansion. The second convex function introduces the slack variable V. m,n Convert it to V m,n Convex function. For (18c), the variables are also separated to one side of the equation. In this way, (P4) is also transformed into a convex optimization problem that can be solved using CVX.

[0091] 3.3) Optimization of hovering time T

[0092] Based on the optimized P and L, optimize T. First, set the maximum flight speed of the drone during flight to ν. Assume that the drone starts from the current position and first accelerates to the maximum speed, then moves at a constant speed, and finally decelerates to the next hovering position. Define the acceleration of the drone during flight to be α. Then the time taken by the drone during flight is:

[0093]

[0094] Among them, S α and T α The distance and time that the drone accelerates and decelerates during the flight can be expressed as:

[0095]

[0096] So now problem (P2) can be transformed into:

[0097]

[0098] sqF m,n τ m ≥δ m,n , (21b)

[0099]

[0100] This problem is a linear optimization problem and can be solved directly.

[0101] 3.4) Design an alternating optimization algorithm;

[0102] Through the above method, the three sub-problems have been transformed into convex problems, which can be solved using CVX tools. In order to solve problem (P2), an algorithm is proposed to alternately optimize the three sub-problems to solve the original problem. The specific steps are as follows:

[0103] 3.4.1) Initialize P 0 , L 0 and T 0 , r and k represent the number of inner and outer layer iterations, and the initial value is set to k = 0;

[0104] 3.4.2) Set the initial value r = 0;

[0105] 3.4.3) Based on L r and T k , optimize P r , and get P r+1 ;

[0106] 3.4.4) Based on P r+1 and T k , optimize L r , and get L r+1 ;

[0107] 3.4.5) According to P r+1 and L r+1 Update the user decoding order and use the genetic algorithm to recalculate the optimal path G and the shortest total distance S min ;

[0108] 3.4.6) Update r = r + 1;

[0109] 3.4.7) Repeat 3.4.3)-3.4.6) until P and L converge;

[0110] 3.4.8) Based on L r and P r , optimize T k , and get T k+1 ;

[0111] 3.4.9) Update k = k + 1;

[0112] 3.4.10) Repeat 3.4.2)-3.4.9) until P, L and T converge.

[0113] Through the above optimization algorithm, (P2) can be solved, and the optimal transmission power and time allocation method can be given. The hovering position of the UAV is reasonably arranged to maximize the system throughput.

[0114] The beneficial effects of the present invention are as follows: the present invention considers a UAV communication network with multiple user groups, and proposes an optimization problem of maximizing throughput in this network. A new solution is provided to optimize resource allocation in UAV communication. First, the K-means algorithm is used to reasonably group the users in the network and the genetic algorithm is used to arrange the flight routes of the UAV. Through reasonable resource allocation, the communication requirements of the users in the system are firstly ensured to be met, and the throughput of the system can also be maximized. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 Diagram of a non-orthogonal multiple access-UAV network system with multiple user groups.

[0116] Figure 2 This is a K-means algorithm example diagram with 19 users.

[0117] Figure 3 The optimal hovering position and route map of the drone under different service quality requirements.

[0118] Figure 4 is the throughput of each user when the service quality requirement is 1.0.

[0119] Figure 5 is the communication rate and time of each user group when the service quality requirement is 1.0.

[0120] Figure 6 The change of total throughput of each scheme with the increase of total transmission power under different service quality requirements.

[0121] Figure 7 Optimize the hovering position and optimal route for a UAV communication system with 7 user groups.

[0122] Figure 8 The throughput of each user in the UAV communication system with 7 user groups. DETAILED DESCRIPTION

[0123] In order to better understand the above technical solution, a specific analysis is given below in conjunction with the accompanying drawings and specific implementation methods.

[0124] Example 1

[0125] The first step is to build a system model:

[0126] 1) If Figure 1 For the UAV communication system model, in this embodiment, the initial hovering position of the UAV is set to (0, 0, 150). The fixed working height of the UAV is H0 = 150m, where the maximum flight speed of the UAV is 8m / s, and the acceleration α is 4m / s 2. Due to the limited energy carried, it is assumed that the drone can work for 415s. In the embodiment, K=19, and the horizontal positions of each user are [112,56], [348,100], [49,333], [202,310], [554,-45], [623,-206], [763,5], [864,-115], [654,0], [1005,263], [1054,542], [1206,306], [0,-500], [100,-400], [-94,-430], [1040,-456], [1106,-360], [1194,-430], and [1123,-565]. These users can be divided into M user groups, each of which includes N m The drone starts from the starting point, passes through each user group according to the optimal flight route and hovers above them to communicate based on NOMA.

[0127] 2) In the system model, bold letters represent matrices and sets, and non-bold letters represent scalars; represents a matrix space with a dimension of 2×1; the user group and the user set can be represented as Λ={1,2,…,M} and Θ={Γ m |Γ m ={1,2,…,N m},m∈Λ};

[0128] In the second step, the objective function is determined based on the specific settings of the first step and the inferences of formulas (1) to (9), and the optimization problem is listed as follows:

[0129] For the convenience of description, the transmission power of the user is defined here. The set of the hovering position and the hovering time of the drone is P = {P m,n |n∈Γ m ,m∈Λ},L={L m |m∈Λ} and T={τ m |m∈Λ}. When the drone is hovering for communication, the user's service quality requirements need to be met. Therefore, the optimization goal of the present invention is to maximize the total throughput of the system, and the optimization problem can be expressed as:

[0130]

[0131]

[0132]

[0133] R m,n τ m ≥δ m,n , (22d)

[0134]

[0135]

[0136]

[0137] For convenience, the communication requirements of each user are set to be the same, that is, m,n =δ0=3bit / Hz and η m,n =η0. Set the total transmission power of the drone P sum =0.2W.

[0138] The third step is to design an algorithm to solve the optimization problem:

[0139] In the optimization problem (P1), the variables are coupled together and the optimization problem is non-convex, which is difficult to solve directly. Therefore, a solution is proposed to simplify the problem (P1). First, the K-means algorithm is used to divide K users into M user groups, and then the genetic algorithm is used to find the optimal route. Finally, an alternating iterative algorithm is proposed to optimize P, L and T. Specifically:

[0140] 1) Optimization of user clustering

[0141] In clustering problems, the K-means algorithm has been widely used because of its advantages of simple principle and fast convergence speed. Therefore, this paper uses the K-means algorithm to cluster users. According to steps 1.1)-1.3), users are classified, and the results are as follows: Figure 2 As shown. Figure 2 .b By drawing the final As M increases, the number of clusters can be determined to be M=5 based on the elbow rule.

[0142] 2) Optimal route optimization

[0143] After obtaining the grouping of drones, the centroid of each user group is set as the initial hovering position of the drone. Next, we need to find the optimal route for the drone to fly. This problem can be converted into a traveling salesman problem (TSP). Generally speaking, an exhaustive search can be performed to find the best option among all possible choices. The complexity of this method is O(M!). It can be seen that when the number of user groups increases, the algorithm complexity will also increase sharply. Therefore, in order to cope with the situation of more user groups, a genetic algorithm is used to find the optimal route G and the corresponding shortest total distance S. min , where G is a (M+1)×1 matrix storing the current optimal route. Based on steps 2.1)-2.4), the optimal flight route and the shortest total distance of the drone are obtained through a genetic algorithm.

[0144] 3) Alternately iterate and optimize P, L and T.

[0145] Through the above method, user clustering and drone flight routes are determined. Therefore, problem (P6) can now be transformed into (P7) to optimize P, L and T:

[0146]

[0147] sqF m,n τ m ≥δ m,n , (23b)

[0148]

[0149]

[0150]

[0151]

[0152]

[0153] In (P7), R m,n is a non-convex and non-concave function, so constraints (23b) and (23c) are both non-convex constraints. (P7) is a non-convex problem as a whole and the variables P, L and T are coupled together, which makes the problem difficult to solve directly. Based on the idea of ​​alternating iteration, the present invention divides the problem into three sub-problems, uses convex optimization theory and convex approximation technology to convert each non-convex sub-problem into a convex problem, and finally proposes an iterative algorithm to alternately optimize the three sub-problems to solve the problem (P7).

[0154] 3.1) Transmission power P optimization

[0155] When optimizing the power P, L and T remain unchanged, and the problem (P7) can be decomposed into:

[0156]

[0157] sqF m,n τ m ≥δ m,n , (24b)

[0158]

[0159]

[0160]

[0161] In this problem, the optimization objective and constraints (24b) and (24c) are both non-convex and non-concave. To solve this problem, the present invention makes an approximation based on the concave-convex process theory and converts them into convex constraints. Specifically, we first use the properties of the logarithmic function to convert R m,n Expand it into the form of two concave functions subtracted, and then use the first-order Taylor expansion to approximate it as a concave function, so that (24b) is converted into a convex constraint. For (24c), rearrange it to separate the variables to one side of the inequality, so that (24c) is also converted into a convex constraint. In summary, problem (P8) is approximated as a convex problem, which can be solved using the convex optimization toolbox CVX.

[0162] 3.2) Hover position L optimization

[0163] When optimizing the hovering position L, P and T remain unchanged, so the problem (P7) can be transformed into:

[0164]

[0165] sqF m,n τ m ≥δ m,n , (25b)

[0166]

[0167]

[0168] In (P9), constraints (25b) and (25c) are both non-convex and non-concave. First, the convex function ||q m,n -L m || 2 As a variable, the same is true for R m,n Using the properties of the logarithmic function, it is converted into a convex function minus a convex function. The first convex function is converted into a concave function about the variable L using the first-order Taylor expansion. The second convex function introduces the slack variable V. m,n Convert it to V m,n Convex function. For (25c), the variables are also separated to one side of the equation. In this way, (P9) is also transformed into a convex optimization problem that can be solved using CVX.

[0169] 3.3) Optimization of hovering time T

[0170] Based on the optimized P and L, T is optimized. Using formula 19) and formula 20) we can calculate the time T used by the drone during the flight process: S , so now problem (P7) can be transformed into:

[0171]

[0172] sqF m,n τ m ≥δ m,n , (26b)

[0173]

[0174] This problem is a linear optimization problem and can be solved directly.

[0175] 3.4) Design an alternating optimization algorithm;

[0176] Through the above method, the three sub-problems can be transformed into convex problems respectively, which can be solved using CVX tools. In order to solve problem (P7), an algorithm is proposed to alternately optimize the three sub-problems to solve the original problem. Execute 3.4.1)-3.4.10) and the results are as follows:

[0177] Figure 3 The optimal hovering position and route map of the drone under different service quality requirements are shown. It can be seen that when the service quality requirements are high, the drone will hover closer to users with good channel quality. This is to allocate more power to users with poor channel quality in order to achieve higher service quality requirements. Therefore, users with good channel quality at the hovering position can compensate for their reduced transmission power.

[0178] Figure 4 This is the throughput of each user when the service quality requirement is 1.0. It can be seen that this solution can meet the communication requirements of users.

[0179] Figure 5 The communication rate and time of each user group when the service quality requirement is 1.0. It can be seen that this solution will tilt system resources toward high-quality users on the basis of ensuring user communication requirements during optimization.

[0180] In order to reflect the advantages of the present invention, this embodiment sets the present invention as Scheme 1, and also sets Scheme 2 and Scheme 3. Both schemes adopt alternating optimization algorithms, but Scheme 2 considers traditional OMA technology to optimize the configuration of system resources, and the method used is consistent with the method of this scheme; in order to reflect the advantages of position optimization, the hovering position of the drone in Scheme 3 is fixed as the center position of each user group, and NOMA technology is still used, and the remaining steps are consistent with this scheme. Figure 6 The performance of schemes 1 to 3 in different scenarios will be compared. It should be noted that, unless otherwise specified, the simulation parameters all use the parameters in the above steps.

[0181] Figure 6This is a comparison of the three solutions. It can be seen that the performance of this solution is better than the other two solutions. It can also be seen that when the service quality requirements are improved, the throughput of the NOMA solution will be slightly improved. This is because high service quality means that the user's minimum achievable rate is also higher, which makes it possible to complete the user's communication requirements in a shorter time. Therefore, there will be more time for high-quality users, which in turn improves the overall performance of the system. The service quality has little impact on the OMA solution.

[0182] Example 2

[0183] In order to demonstrate the effectiveness of this solution when facing more users, 5 users are added as Example 2 based on Example 1. The coordinates of the five users are: [405,-591], [347,-682], [513,-734], [500,475], [612,394]. In addition, the total working time of the drone is reset to T0=489s, and the user service quality is η0=1.0. The other parameters are consistent with Example 1. The results of this example are Figure 7 as well as Figure 8 Displayed in.

[0184] Figure 7 The optimized positions and routes when the number of users is 25 are displayed. It can be seen that these users are divided into 7 user groups. The optimal route and hovering position can be effectively solved.

[0185] Figure 8 The throughput of each user is displayed, and it can be seen that the communication requirements of each user are well met. Figure 7 and Figure 8 , it can be seen that this solution can also handle situations with more users.

[0186] The above-described embodiments merely express the implementation methods of the present invention, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A method for joint optimization of UAV trajectory and time based on non-orthogonal multiple access technology, characterized in that: The following steps are involved: The first step is to build a system model: 1) In a UAV communication network, a rotary-wing UAV equipped with a single antenna serves as a base station, serving K ground users. The K users are divided into M user groups, and Λ={1,2,…,M} represents the set of user groups; each user group includes N m users, with Γ m ={1,2,…,N m }, m∈Λ represents the set of users in the mth user group; the drone starts from the starting point, passes through each user group according to the optimal flight route and hovers above it to communicate based on NOMA; 2) In the system model, bold letters represent matrices or sets, and non-bold letters represent scalars; Represents a matrix space with a dimension of 2×1; where Θ={Γ m |Γ m ={1,2,…,N m },m∈Λ} represents the set of all users; 3) The total operational time T0 of the drone can be divided into the flight time T S And the hovering communication time, the hovering communication time of the UAV above the mth user group is expressed as τ m ,get: Define the nth user in the mth user group as U m,n , set the fixed working height of the drone to H0, and the horizontal hovering position of the drone above the mth user group and the position of the user are represented by and Indicates that user U m,n The distance to the drone is expressed as: Assume that the distance between the user and the drone in each user group satisfies the following relationship: The UAV air-to-ground channel is considered as a LoS link, so user U m,n The channel coefficient h m,n 2 It is expressed as: Where ρ0 represents the channel gain at unit distance; According to NOMA, users with better channel quality will be allocated less transmission power; therefore: Among them, P m,n Represents user U m,n The transmission power of For User U m,n ,2≤n∈Γ m , the signal to interference plus noise ratio SINR is: in, Represents user U m,n The signal of user U m,w SINR at the receiver; Indicates U m,n SINR when decoding its own signal; σ 2 represents the noise power; In order to successfully achieve serial interference cancellation, the user's SINR needs to meet the following requirements: Among them, η m,n Represents user U m,n Minimum SINR requirement; then the user's achievable rate is expressed as: In the second step, according to the specific settings of the first step, determine the objective function and list the optimization problem: The set of user's transmission power, drone hovering position and hovering time is P = {P m,n |n∈Γ m ,m∈Λ},L={L m |m∈Λ} and T={τ m |m∈Λ}; When the drone is hovering for communication, the user's service quality requirements need to be met; the optimization goal is to maximize the total throughput of the system, then the optimization problem is expressed as: R m,n t m ≥d m,n ,(10d) Among them, δ m,n Is user U m,n The communication requirements of P sum is the total transmission power of the UAV; The third step is to design an algorithm to solve the optimization problem: First, the K-means algorithm is used to divide K users into M user groups, and then the genetic algorithm is used to find the optimal route. Finally, an alternating iterative algorithm is proposed to optimize P, L and T. The specific process is as follows: 1) Optimization of user clustering Use K-means algorithm to cluster users; 2) Optimal route optimization After obtaining the grouping of drone users, the centroid of each user group is set as the initial hovering position of the drone; find the optimal route for the drone to fly, which can be converted into a traveling salesman problem TSP; use genetic algorithm to find the optimal route G and the corresponding shortest total distance S min , where G is a (M+1)×1 matrix storing the current optimal route; the specific steps are as follows: 2.1) Encode the problem; define the take-off position of the drone as 0, then in this problem, the gene of each individual in the genetic algorithm represents the order in which the drone passes through the user group; set the total number of individuals in each generation of the genetic algorithm to Z; define the set of individuals in the tth generation as Ω t ={1,2,…,Z}; 2.2) Calculate the fitness value of the individual and define the fitness value ψ of the zth individual in the tth generation t (z) is: in, represents the longest total distance in the tth generation, is the shortest total distance, λ t (z) is the total distance corresponding to the zth individual; ε1 is a constant greater than 0; find the best individual of the tth generation and compare it with the current best individual. If the best individual is better, update the current best route and the shortest total distance; 2.3) The roulette wheel method is used to select the parent generation of the next generation population; the first step is to calculate the cumulative probability of the zth individual in the tth generation Randomly generate a number ε2∈(0,1], if the random number satisfies Then the zth individual is selected as the parent; 2.4) Genetic or mutation to produce the next generation; genetic algorithms require a certain mutation probability to prevent falling into the local optimal solution; the probability of mutation is defined as Randomly generate a random number ε3∈(0,1), if the number satisfies Then select mutations to generate the next generation, otherwise use genetic methods; repeat 2.3) and 2.4) until enough next generation individuals are obtained; Repeat the above steps until a sufficient number of generations are iterated; the optimal flight route and the shortest total distance of the UAV can be obtained through the above genetic algorithm; 3) Alternately iterate and optimize P, L and T; Through the above method, the user grouping and the UAV flight route are determined, and the following constraints are introduced: Among them, L G(j) represents the horizontal position of the drone hovering above the G(j)th user group; therefore, problem (P1) can now be transformed into (P2) to optimize P, L and T: s.t.R m,n τ m ≥δ m,n ,(16b) Based on the idea of ​​alternating iteration, the problem (P2) is divided into three sub-problems. Each non-convex sub-problem is transformed into a convex problem using convex optimization theory and convex approximation technology. Finally, an iterative algorithm is proposed to alternately optimize the three sub-problems to solve the problem (P1). 3.1) Transmission power P optimization When optimizing the power P, L and T remain unchanged, and the problem (P2) can be decomposed into: s.t.R m,n τ m ≥δ m,n , (17b) Based on the concave-convex process theory, constraints (17b) and (17c) are converted into convex constraints. Finally, the above problem (P3) is approximated as a convex problem and solved using the convex optimization toolbox CVX. 3.2) Hover position L optimization When optimizing the hovering position L, P and T remain unchanged, and the problem can be transformed into: s.t.R m,n τ m ≥δ m,n , (18b) In (P4), constraints (17b) and (17c) are both non-convex and non-concave; first, the convex function ||q m,n -L m || 2 As a variable, the same is true for R m,n Using the properties of the logarithmic function, it is converted into a convex function minus a convex function. The first convex function is converted into a concave function about the variable L using the first-order Taylor expansion. The second convex function introduces the slack variable V. m,n Convert it to V m,n Convex function; for (18c), the variables are also separated to one side of the equation; thus (P4) is also transformed into a convex optimization problem that can be solved using CVX; 3.3) Optimization of hovering time T Based on the optimized P and L, T is optimized. First, the maximum flight speed of the drone during flight is set to ν. Assume that the drone starts from the current position and first accelerates to the maximum speed, then moves at a constant speed, and finally decelerates to the next hovering position. The acceleration of the drone during flight is defined as α. Then the time taken by the drone during flight is: Among them, S α and T α The distance and time that the drone accelerates and decelerates during the flight can be expressed as: So now problem (P2) can be transformed into: stR m,n t m ≥d m,n ,(21b) 3.4) Design an alternating optimization algorithm; In order to solve problem (P2), an algorithm is proposed to alternately optimize the three sub-problems to solve the original problem. The specific steps are as follows: 3.4.1) Initialize P 0 , L 0 and T 0 , r and k represent the number of inner and outer layer iterations, and the initial value is set to k = 0; 3.4.2) Set the initial value r = 0; 3.4.3) Based on L r and T k , optimize P r , and get P r+1 ; 3.4.4) Based on P r+1 and T k , optimize L r , we get L r+1 ; 3.4.5) According to P r+1 and L r+1 Update the user decoding order and use the genetic algorithm to recalculate the optimal path G and the shortest total distance S min ; 3.4.6) Update r = r + 1; 3.4.7) Repeat 3.4.3)-3.4.6) until P and L converge; 3.4.8) Based on L r and P r , optimize T k , and get T k+1 ; 3.4.9) Update k = k + 1; 3.4.10) Repeat 3.4.2)-3.4.9) until P, L and T converge; The above optimization algorithm can be used to solve (P2) and provide the optimal transmission power and time allocation method.

2. According to claim 1, a method for jointly optimizing the trajectory and time of a UAV based on non-orthogonal multiple access technology is characterized in that: In the first step 3) formula (6), when n=1, user U m,1 The SINR can be expressed as:

3. The method for jointly optimizing the trajectory and time of a UAV based on non-orthogonal multiple access technology according to claim 1 or 2, characterized in that: The specific process of the third step 1) is as follows: 1.1) First randomly select the positions of M users as the center points of the initial M clusters, and then assign each of the remaining users to the cluster whose center position is closest to the user; 1.2) Recalculate the center point μ of the cluster based on the division results m , which is calculated as follows: 1.3) Re-divide users according to the center position calculated in 1.2); Repeat the above two steps until the center position of each user group does not change; in the above process, the square error sum of the objective function of the algorithm It can be expressed as: By drawing the final As M increases, the number of clusters M can be determined based on the elbow rule; through the above process, the number of user clusters can be determined, that is, the optimal grouping of drone users.