A three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and air base station deployment position

Abstract

This invention relates to the field of wireless communication network optimization technology, specifically a three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of airborne base stations. Based on an integrated air-ground network, this method designs a two-layer optimization architecture for network optimization. By handling the location selection of airborne base stations and the adjustment of antenna tilt angles of ground base stations separately, it avoids strong coupling between discrete and continuous variables, thereby preventing superlinear expansion of the solution space. Simultaneously, by introducing SPSA simulated gradient technology and the ZO-Adam optimization algorithm, it overcomes the non-smoothness and non-convexity problems of the system's spectral efficiency function, ensuring the stability and convergence of the optimization process. Furthermore, the two-layer optimization architecture effectively alleviates the path dependency problem of greedy algorithms, achieving a balance between global search and timeliness.

Description

Technical Field

[0001] This invention relates to the field of wireless communication network optimization technology, specifically a three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of air base stations. Background Technology

[0002] Low-altitude unmanned aerial vehicles (UAVs) possess the advantages of maneuverability and flexibility, serving as aerial base stations (ABS) to enhance network coverage in scenarios such as emergency communications and urban hotspot coverage. However, in Air-Ground Integrated Networks (AGINs), user distribution has expanded from the traditional planar to three-dimensional space. How to coordinate and optimize the deployment of ground and aerial base stations to ensure reliable signal coverage for users in three-dimensional space, especially in the coverage of low-altitude users and the selection of deployment locations for UAV base stations, has become a key issue facing AGINs.

[0003] Air-ground integrated networks combine the advantages of terrestrial cellular base stations and airborne base stations, enabling wide-area, three-dimensional coverage communication services. To fully leverage the advantages of air-ground synergy, the coordinated optimization of both terrestrial and airborne base stations needs to be considered simultaneously. Currently, in cellular networks, each terrestrial base station is typically fixed in deployment and has a limited coverage sector. Its antenna pointing (tilt / downtilt angle, etc.) is generally pre-configured for covering surface users. With the emergence of low-altitude users (such as drone terminals), the vertical coverage of traditional networks is insufficient. Furthermore, the introduction of drone base stations provides dynamic coverage capabilities, but how to select the deployment location and altitude of drones to complement the existing terrestrial network is also a complex decision-making problem.

[0004] CN117082536B discloses a reinforcement learning-based method for air-to-ground network cooperative coverage, comprising the following steps: S1, obtaining the geographical location and base station parameter domain of the base station; the geographical location refers to the latitude and longitude information of the base station; the base station parameter domain includes the types of base station parameters and the range of each type of base station parameter; S2, based on UAV measured data, obtaining the low-altitude and ground coverage of the base station under different base station parameters, and forming a three-dimensional antenna pattern; S3, modeling the air-to-ground network cooperative problem as a reinforcement learning task, and finding the optimal base station parameter configuration state based on reinforcement learning, thereby achieving air-to-ground network cooperative coverage. By obtaining the relationship between base station configuration parameters and base station coverage, and then constructing a reinforcement learning task to find the optimal base station configuration in the base station parameter domain, cooperative coverage of low-altitude and ground is achieved.

[0005] Existing systems typically employ a hierarchical or joint optimization strategy: First, candidate locations for airborne nodes are selected individually using heuristic algorithms (such as greedy algorithms or simulated annealing), generating an initial deployment plan based on predefined coverage rules (such as user demand heatmaps). Second, the tilt parameters of ground base stations are optimized using numerical methods such as gradient descent to maximize the local spectral efficiency objective function. Finally, deployment and parameter values ​​are adjusted in iterative feedback. However, this approach does not fully consider the real-time interactive characteristics of integrated air-ground networks, and the solution space formed by the high-dimensional continuous parameter space of airborne node candidate locations and ground tilt angles exhibits superlinear expansion, making real-time solutions infeasible. Furthermore, the user dynamic association mechanisms in existing methods (such as switching based on the strongest reference signal received power) cause the system's spectral efficiency function to exhibit non-smooth and non-convex characteristics, thus gradient-based algorithms often fail to converge. Traditional greedy algorithms are limited by the path dependency of deployment order, while global search methods lack timeliness guarantees and cannot converge quickly in large-scale network optimization. Summary of the Invention

[0006] To address the problem that existing optimization strategies in the prior art struggle to balance efficiency and global applicability due to complex solution spaces, non-convex problem models, and dependence on order, this invention provides a three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of air base stations. It employs a two-layer optimization architecture: the outer layer iteratively processes discrete deployment decisions for air base stations, while the inner layer iteratively calculates the tilt angle parameters of ground base stations through stochastic perturbation gradient estimation and adaptive optimization algorithms. This solves the problem of superlinear expansion of the solution space caused by strong coupling between discrete and continuous variables, thereby reducing computational complexity and enabling real-time optimization.

[0007] This invention is achieved through the following technical solution: A three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of airborne base stations includes: Obtain the fixed location set of ground base stations, the deployed and non-deployed sets of air base stations, and the user grid point set; A two-layer optimization architecture is constructed, and a two-layer alternating optimization method is adopted to deploy air base stations and adjust the antenna downtilt angle of ground base stations with the maximum weighted total spectrum efficiency of all users as the optimization index. The dual-layer optimization architecture consists of an outer layer and an inner layer. In the outer layer, a greedy algorithm with a backtracking mechanism is used to perform discrete deployment of undeployed airborne base stations in a loop. In the inner layer, the tilt angle parameters of ground base stations are obtained iteratively using stochastic perturbation gradient estimation and adaptive optimization algorithm.

[0008] Preferably, a two-layer optimization architecture is constructed, using the maximization of the weighted total spectral efficiency for all users as the optimization metric for deploying airborne base stations and adjusting the antenna downtilt angles of ground base stations, including: S1: Select any air base station from the undeployed set as a candidate base station, and move the candidate base station into the deployed set to generate a temporary deployment set; S2, based on the fixed location set, temporary deployment set and user grid point set of ground base stations, outputs the optimal antenna tilt angle set through random perturbation gradient estimation and adaptive optimization algorithm, and obtains the system spectral efficiency gain based on the optimal antenna tilt angle set; S3, iterate through the system spectral efficiency gain of all airborne base stations, and use the system spectral efficiency gain of any airborne base station compared with a preset gain threshold as a criterion to determine whether to move the airborne base station into the deployed set: If the maximum system spectral efficiency gain of an airborne base station exceeds a preset gain threshold, the airborne base station will be moved into the deployed set and a new historical gain set for the current round will be generated after its removal. If not, then select from all airborne base stations the airborne base station whose maximum system spectral efficiency gain exceeds the preset replacement threshold as the candidate point; wherein, the preset gain threshold is greater than the preset replacement threshold. S4. If the current round is the first round, move the candidate point into the deployed set; if the current round is not the first round, compare the system spectrum efficiency gain of the candidate point with the airborne base station with the maximum value in the historical gain set of the previous round, and select the airborne base station corresponding to the larger value of the system spectrum efficiency gain to move into the deployed set. S5. If there is no candidate in the current round, a backtracking mechanism is introduced to retrieve the historical gain set of the previous round and execute S3~S4 operations. If no candidate is found after the execution is completed, the backtracking is determined to have failed. S6, update the deployment set and re-execute the loop to move into the deployed set until the undeployed set is empty or backtracking fails, and output the deployed set and the optimal antenna tilt set.

[0009] Preferably, in S2, the optimal antenna tilt angle set is output iteratively through random perturbation gradient estimation and adaptive optimization algorithm, including: S201, Generate the channel sample matrix This matrix contains the channel gain from each air base station to the user grid point; S202, with the initial downtilt angle parameter set of the antenna as the independent variable, and the channel sample matrix as the independent variable. Construct a system spectral efficiency function for the conditional parameters; S203, by applying random perturbations, zero-order gradient simulation is performed on the system spectral efficiency function to obtain gradient estimation components; S204 uses the gradient estimation components of the current iteration as the historical gradient vector and employs the Adama stochastic algorithm to update the antenna downtilt angle.

[0010] Preferably, in step S203, the zero-order gradient simulation of the system's spectral efficiency function is performed by applying a random perturbation to obtain the gradient estimation components, including: S2031, Generate perturbation vector: Adjust antenna downtilt parameters Perform random perturbations to generate perturbation directions. Its elements are drawn independently from a uniform or normal distribution. Random values; S2032, Calculate the spectral efficiency function of the system after the disturbance: in the... In the next iteration, the antenna downtilt angle Apply a perturbation and calculate the objective function values ​​before and after the perturbation:

[0011] in, To control the magnitude of the disturbance, It is the perturbation vector. This represents the calculated spectral efficiency; S2033, SPSA estimates antenna downtilt angle based on objective function values ​​before and after perturbation. gradient:

[0012] In the formula, It is a simulated estimation of the gradient of the objective function.

[0013] Preferably, in S3, a greedy algorithm is introduced to progressively select the airborne base station that can bring the maximum gain from the candidate base stations using gain difference as the evaluation index, including: For each candidate base station, obtain the difference in system spectral efficiency gain before and after deploying the candidate base station. ; A greedy algorithm is used to obtain the gain difference of each candidate base station. In each round of selection, the remaining candidate set is selected. Choose the air base station location that brings the greatest gain difference. :

[0014] The selected airborne base station location is added to the deployed set. In the process, the spectral efficiency of the updated system is calculated.

[0015] Preferably, in S5, if a certain candidate position The gain difference did not reach the preset threshold. ,Right now: If this happens, the backtracking mechanism will be triggered, and the backtracking mechanism will retrieve data from the historical gain record set. Choose an alternative position with a higher gain and reselect.

[0016] The preferred backtracking selection formula is: .

[0017] A system for implementing the three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of airborne base stations, comprising: The data acquisition module is used to acquire the fixed location set of ground base stations, the deployed and non-deployed sets of air base stations, and the user grid point set; The deployment optimization module is used to deploy air base stations and adjust the antenna downtilt angle of ground base stations by using a two-layer alternating optimization method with the maximum weighted total spectral efficiency of all users as the optimization index.

[0018] An electronic device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method.

[0019] A storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0020] Compared with the prior art, the present invention has the following beneficial effects: This invention presents a three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of airborne base stations. Based on an integrated air-ground network, this method employs a two-layer optimization architecture for network optimization. By handling the location selection of airborne base stations and the antenna tilt angle adjustment of ground base stations separately, it avoids strong coupling between discrete and continuous variables, thereby preventing superlinear expansion of the solution space. Simultaneously, by introducing SPSA simulated gradient technology and the ZO-Adam optimization algorithm, it overcomes the non-smoothness and non-convexity of the system's spectral efficiency function, ensuring the stability and convergence of the optimization process. Furthermore, the two-layer optimization architecture effectively alleviates the path dependency problem of greedy algorithms, achieving a balance between global search and timeliness.

[0021] Furthermore, in the inner layer, symmetric stochastic perturbation gradient estimation and adaptive learning rate adjustment are introduced to solve the problem of non-smooth and non-convex characteristics of the function caused by the stochastic association mechanism, thereby ensuring the convergence of the algorithm and improving the spectral efficiency.

[0022] Furthermore, at the outer layer, a greedy algorithm is used to gradually and quickly select the location of the airborne base station. To overcome the problem that the greedy algorithm may get stuck in a local optimum due to the selection strategy based on the current optimal location, a backtracking mechanism is introduced. By backtracking historical gain records, a better airborne base station deployment location is selected again, thereby ensuring the global optimum. This solves the problems of path dependence and low global search efficiency caused by deployment order sensitivity, and achieves synergistic optimization of decision robustness and real-time response capability. Attached Figure Description

[0023] Figure 1 This is a flowchart of a three-dimensional communication coverage method for jointly optimizing the downtilt angle of a ground base station and the deployment location of an air base station according to the present invention; Figure 2 This is a schematic diagram of a three-dimensional communication coverage system based on the joint optimization of the downtilt angle of a ground base station and the deployment location of an airborne base station, according to the present invention. Detailed Implementation

[0024] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification.

[0025] Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0026] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0027] The present invention will be further described in detail below with reference to specific embodiments. These descriptions are for explanation purposes only and are not intended to limit the scope of the invention.

[0028] This invention is based on an integrated air-ground network, which includes several fixed-location ground base stations (GBS) and deployable airborne base stations (ABS). The ground base stations typically have a three-sector structure, with each sector covering... The antenna has an adjustable vertical downtilt angle within a horizontal angular range; each airborne base station (such as a small base station carried by a UAV) uses a single downward-facing conical beam antenna sector to provide coverage within a certain range. Users in the system include ground users and low-flying airborne users, distributed in three-dimensional coordinates. All base stations and user terminals operate within a unified frequency band (full frequency reuse), and users are automatically connected to the base station sector with the strongest signal based on the maximum received signal strength (RSS) criterion.

[0029] In this invention, the signal-to-interference-plus-noise ratio (SINR) of all users is used to calculate the spectral efficiency, and these are accumulated to form an optimization index. Given a set of candidate airborne base station deployment locations, several ABS are selected for deployment and the antenna downtilt angle of GBS is adjusted to maximize the weighted total spectral efficiency of airborne and ground users.

[0030] The system spectral efficiency function is:

[0031] In the formula, Let be the system spectral efficiency function. The total number of channel sample matrices. The total number of user grid points. The signal-to-interference-plus-noise ratio (SINR) function. These are the channel sample matrix index number and the user grid point set index number, respectively. This refers to the tilt angle parameter.

[0032] The signal-to-interference-plus-noise ratio (SIR / NOT) function is:

[0033] in, To serve the base station's transmission power, For the first The transmission power of the interfering base station For the service base station channel coefficient, For the first The channel coefficient of each interfering base station The noise power of the channel sample matrix, This is the antenna gain function. The angle between the line connecting the service base station and the user grid point and the horizontal plane. For the first The angle between the line connecting the interfering base station and the user grid point and the horizontal plane.

[0034] The antenna gain function is: ; in, This represents the maximum antenna gain, which can be set to 15-18 dB. The antenna's half-power beamwidth can be taken as... , For comparison before and after, the value can be 20-30dB.

[0035] The path loss formula from the ground base station to the user grid point is:

[0036] in, , is the carrier frequency, It is a three-dimensional Euclidean distance.

[0037] The path loss formula from the airborne base station to the user grid point is:

[0038] in, , is the carrier frequency, It is a three-dimensional Euclidean distance.

[0039] Referring to the channel modeling method of ITU-R M.2135-1, this embodiment generates small-scale fading coefficients that follow a Rayleigh distribution for each communication link, and generates random fading coefficients by superimposing shadow fading coefficients that follow a log-normal distribution.

[0040] It should be noted that, due to the significant multipath propagation effect in urban non-line-of-sight (NLoS) scenarios, a Rayleigh distribution is used to simulate the rapid random fluctuations in signal amplitude. This distribution can accurately characterize the multipath signal interference phenomenon caused by reflections from dense building clusters. Referring to the 3GPP TR 38.901 standard, this embodiment independently generates random variables following a Rayleigh distribution for each communication link, with the scale parameter set to... The corresponding distribution mean is The generated random variables are expressed in decibels (dB) after logarithmic transformation.

[0041] It should also be noted that the shadow fading coefficient, which characterizes the long-term slow fading characteristics caused by building obstruction in a macrocell, uses a log-normal distribution to simulate the slow spatial variation of signal strength; and generates an independent normally distributed random variable for each communication link. Then through exponential conversion The variable is converted to a log-normal distribution and ultimately expressed in decibels (dB); among which, the urban environment can be taken as... Rural environment .

[0042] In existing scenarios of dynamic coverage optimization in heterogeneous air-ground networks, the strong coupling effect between discrete deployment decisions and continuous parameter optimization leads to a superlinear expansion of the solution space, making real-time solutions infeasible. The dynamic user association mechanism causes the system's spectral efficiency function to exhibit non-smooth and non-convex characteristics, resulting in a lack of convergence for gradient-based algorithms. Traditional methods use heuristic algorithms to handle air node location separately and then optimize the ground base station tilt angle through gradient descent, without fully considering the real-time interaction characteristics of network components, leading to strong path dependence of deployment order and low global search efficiency.

[0043] To address the aforementioned problems, this invention discloses a three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of airborne base stations, comprising: Obtain the fixed location set of ground base stations, the deployed and non-deployed sets of air base stations, and the user grid point set; A two-layer optimization architecture is constructed, and a two-layer alternating optimization method is adopted to deploy air base stations and adjust the antenna downtilt angle of ground base stations with the maximum weighted total spectrum efficiency of all users as the optimization index. The dual-layer optimization architecture consists of an outer layer and an inner layer. In the outer layer, a greedy algorithm with a backtracking mechanism is used to perform discrete deployment of undeployed airborne base stations in a loop. In the inner layer, the tilt angle parameters of ground base stations are obtained iteratively using stochastic perturbation gradient estimation and adaptive optimization algorithm.

[0044] Preferably, a two-layer optimization architecture is constructed, using the maximization of the weighted total spectral efficiency for all users as the optimization metric for deploying airborne base stations and adjusting the antenna downtilt angles of ground base stations, including: S1: Select any air base station from the undeployed set as a candidate base station, and move the candidate base station into the deployed set to generate a temporary deployment set; S2, based on the fixed location set, temporary deployment set and user grid point set of ground base stations, outputs the optimal antenna tilt angle set through random perturbation gradient estimation and adaptive optimization algorithm, and obtains the system spectral efficiency gain based on the optimal antenna tilt angle set; S3, iterate through the system spectral efficiency gain of all airborne base stations, and use the system spectral efficiency gain of any airborne base station compared with a preset gain threshold as a criterion to determine whether to move the airborne base station into the deployed set: If the maximum system spectral efficiency gain of an airborne base station exceeds a preset gain threshold, the airborne base station will be moved into the deployed set and a new historical gain set for the current round will be generated after its removal. If not, then select from all airborne base stations the airborne base station whose maximum system spectral efficiency gain exceeds the preset replacement threshold as the candidate point; wherein, the preset gain threshold is greater than the preset replacement threshold. S4. If the current round is the first round, move the candidate point into the deployed set; if the current round is not the first round, compare the system spectrum efficiency gain of the candidate point with the airborne base station with the maximum value in the historical gain set of the previous round, and select the airborne base station corresponding to the larger value of the system spectrum efficiency gain to move into the deployed set. S5. If there is no candidate in the current round, a backtracking mechanism is introduced to retrieve the historical gain set of the previous round and execute S3~S4 operations. If no candidate is found after the execution is completed, the backtracking is determined to have failed. S6, update the deployment set and re-execute the loop to move into the deployed set until the undeployed set is empty or backtracking fails, and output the deployed set and the optimal antenna tilt set.

[0045] In one embodiment, based on a fixed location set, a temporary deployment set, and a user grid point set, an optimal antenna tilt angle set is output iteratively through random perturbation gradient estimation and adaptive optimization algorithms, including: S201, Generate the channel sample matrix This matrix contains the channel gain from each air base station to the user grid point; S202, Considering the user grid point set and airborne base station deployment And optimized antenna downtilt angle parameters for ground base stations Calculate the channel coefficients from each user grid point to each base station, using the initial antenna downtilt angle parameter set as the independent variable and the channel sample matrix as the input. Construct a system spectral efficiency function for the conditional parameters; S203, by applying random perturbations, zero-order gradient simulation is performed on the system spectral efficiency function to obtain gradient estimation components; S204 uses the gradient estimation components of the current iteration as the historical gradient vector and employs the Adama stochastic algorithm to update the antenna downtilt angle.

[0046] When generating the channel matrix, factors such as spatial location, path loss, and fading effects among different users need to be considered. Zero-order gradient simulation is used because the dynamic user correlation mechanism causes the system's spectral efficiency function to exhibit non-smooth and non-convex characteristics, resulting in a lack of convergence for gradient-based algorithms.

[0047] The core of zero-order gradient simulation is to perform zero-order gradient simulation on the system's spectral efficiency function by applying random perturbations to obtain gradient estimation components, specifically including: S2031, Generate perturbation vector: Adjust antenna downtilt parameters Perform random perturbations to generate perturbation directions. Its elements are drawn independently from a uniform or normal distribution. The random value; the probability of each perturbation component. .

[0048] S2032, Calculate the spectral efficiency function of the system after the disturbance: in the... In the next iteration, the antenna downtilt angle Apply a perturbation and calculate the objective function values ​​before and after the perturbation:

[0049] in, To control the magnitude of the disturbance, It is the perturbation vector. This represents the calculated spectral efficiency; S2033, SPSA estimates antenna downtilt angle based on objective function values ​​before and after perturbation. gradient:

[0050] In the formula, It is a simulated estimation of the gradient of the objective function.

[0051] Compared with existing technologies, traditional gradient estimation methods rely on analytical differentiation or one-sided perturbation approximation, which are prone to estimation bias under non-smooth objective functions (i.e., system spectral efficiency functions). In contrast, this invention uses symmetric random perturbation combined with difference quotient calculation to effectively eliminate gradient direction error and improve estimation accuracy. Existing optimization algorithms usually use a fixed learning rate, which is prone to oscillation or convergence stagnation in complex terrain scenarios. This invention can achieve adaptive adjustment of the update step size through standardization processing and dynamic constraint mechanism, thereby enhancing the robustness of the algorithm.

[0052] This invention combines the Adam optimizer's SPSA (Simultaneous Perturbation Stochastic Approximation) gradient simulation technique with a proposed zero-order optimization algorithm (ZO-Adam). This algorithm can efficiently solve non-smooth, non-convex optimization problems, and is particularly suitable for the complex optimization challenges faced in the undertilt optimization of terrestrial base station antennas. Unlike traditional gradient optimization methods, ZO-Adam does not rely on the explicit derivative of the objective function. Instead, it approximates the gradient by applying random perturbations to the parameters and utilizes the adaptive characteristics of the Adam optimizer to update the optimization parameters. Therefore, it can still perform optimization efficiently without explicit gradient information, while maintaining high stability and convergence. ZO-Adam not only solves the optimization failure problem caused by inaccurate gradient estimation under non-smooth objective functions, but also reduces the estimation variance through a symmetric perturbation mechanism, avoiding getting trapped in local optima. The standardization processing and dynamic learning rate adjustment mechanism can adapt to different base station deployment densities and terrain conditions, improving the algorithm's convergence speed and stability. The physical constraint mechanism of the tilt parameters ensures the engineering feasibility of the solution set, avoiding the problem of invalid solutions due to mechanical limitations.

[0053] In one embodiment, S3, a greedy algorithm is introduced to progressively select the airborne base station that can bring the maximum gain from the candidate base stations using gain difference as the evaluation index, including: S301, Gain Difference Calculation: For each candidate base station, the benefit of each location is evaluated by calculating the gain difference, and the gain difference in system spectral efficiency before and after deploying the candidate base station is obtained. The calculation formula is:

[0054] in, Indicates the current candidate position Post-system spectrum benefits It is the spectral efficiency of the currently deployed airborne base station set. By calculating the gain difference, the contribution of each candidate location can be enhanced, thereby selecting the airborne base station location that brings the maximum gain.

[0055] S302, Stepwise selection of the optimal location: A greedy algorithm is used to obtain the gain difference of each candidate base station. In each round of selection, the remaining candidate set is used to select the optimal location. Choose the air base station location that brings the greatest gain difference. :

[0056] The selected airborne base station location is added to the deployed set. In this process, the spectral efficiency of the system is calculated and updated. In each round of selection, the gain is maximized, thereby improving the overall spectral efficiency of the system.

[0057] Greedy algorithms can quickly select the location of each airborne base station, but because their selection strategy is based on the current optimal location, they may get stuck in local optima. Therefore, a backtracking mechanism is introduced. When the gain difference is small or cannot be further improved, historical gain records are backtracked to reselect a better airborne base station deployment location, thereby ensuring a globally optimal solution.

[0058] Specifically, if a certain candidate position The gain difference did not reach the preset threshold. ,Right now: If this happens, the backtracking mechanism will be triggered, and the backtracking mechanism will retrieve data from the historical gain record set. Choose an alternative position with a higher gain and reselect.

[0059] The backtracking selection formula is: .

[0060] To enable backtracking, the greedy algorithm stores the gain difference and the selected position in the historical gain record set after each selection. Therefore, when backtracking is triggered, the greedy algorithm can use historical records to select an alternative position with greater gain, thereby optimizing the selection process.

[0061] This invention discloses a three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and airborne base station deployment location. First, a spatial model including ground base stations, airborne base stations, and user grid points is established. In each iteration, airborne base stations are selected from the undeployed set and simulated for addition to the deployed set, generating a temporary deployment scheme. The gradient direction of the ground base station downtilt angle parameters is calculated by applying symmetric random perturbations, and the downtilt angle parameters are updated using an adaptive learning rate until convergence, obtaining the optimal downtilt configuration under the current deployment scheme. By comparing the spectral efficiency gain of each candidate point, points that meet a preset gain threshold are selected and added to the deployed set. When no candidate points are available, the historical gain set is called back for re-decision to avoid getting trapped in local optima. This process is repeated until all candidate points are evaluated or backtracking fails, finally outputting the optimal deployed set and the corresponding antenna downtilt angle set.

[0062] This invention is a joint optimization method for selecting the deployment location of airborne base stations and optimizing the antenna downtilt angle of ground base stations. It not only solves the deployment problem of airborne base stations, but also maximizes the spectral efficiency of the network by dynamically adjusting the antenna downtilt angle. It adopts a two-layer optimization architecture to reduce the dimensionality of the solution space, decomposes the high-dimensional coupled optimization problem into two sub-problems: discrete deployment and continuous parameters, and solves them iteratively. The adaptive optimization algorithm dynamically adjusts the preset learning rate according to the convergence status to improve the efficiency of parameter update. The backtracking mechanism avoids the waste of resources in invalid search paths and ensures that an approximate global optimal solution is obtained under limited computing resources. In scenarios with dynamically changing user distribution, this method can quickly generate an optimization scheme that takes into account both coverage and communication quality, thereby improving the deployment efficiency and system performance of heterogeneous air-ground networks.

[0063] In the inner layer, SPSA simulated gradient technology is combined with the Adam optimizer to form a zero-order optimization algorithm (ZO-Adam), achieving efficient optimization, especially suitable for non-smooth, non-convex objective functions, which can improve the stability and convergence of the optimization process. Simultaneously, the inner and outer optimization modules are called alternately. Each time a candidate location is evaluated, the outer optimization calls the inner optimization, and the inner optimization returns the optimized antenna configuration, influencing the outer optimization's decision. Furthermore, a protection backtracking mechanism is implemented. When the air base station location selected by the outer optimization fails to achieve the expected gain, the backtracking mechanism is activated to re-evaluate historical gains, ensuring that each selected air base station location brings maximum gain. The backtracking process also relies on the inner optimization module for antenna configuration adjustments.

[0064] This invention also discloses a system for implementing the three-dimensional communication coverage method for jointly optimizing the downtilt angle of ground base stations and the deployment location of airborne base stations, comprising: The data acquisition module is used to acquire the fixed location set of ground base stations, the deployed and non-deployed sets of air base stations, and the user grid point set; The deployment optimization module is used to deploy air base stations and adjust the antenna downtilt angle of ground base stations by using a two-layer alternating optimization method with the maximum weighted total spectral efficiency of all users as the optimization index.

[0065] Specifically, the deployment optimization module includes a greedy algorithm. This invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, and is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions from a computer storage medium to implement the corresponding method flow or corresponding function. This invention also discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method. The computer-readable storage medium is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space containing the terminal's operating system. Furthermore, this storage space also contains one or more instructions suitable for loading and execution by a processor; these instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium.

[0066] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0068] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0069] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0070] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the functions specified in one or more boxes. Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

[0071] The above description is merely a preferred embodiment of the present invention and is not intended to limit the technical solution of the present invention in any way. Those skilled in the art should understand that, without departing from the spirit and principles of the present invention, the technical solution can be modified and replaced in several simple ways, and these modifications and replacements are all within the scope of protection covered by the claims.

Claims

1. A three-dimensional communication coverage method that jointly optimizes the downtilt angle of ground base stations and the deployment location of airborne base stations, characterized in that, include: Obtain the fixed location set of ground base stations, the deployed and non-deployed sets of air base stations, and the user grid point set; A two-layer optimization architecture is constructed, and a two-layer alternating optimization method is adopted to deploy air base stations and adjust the antenna downtilt angle of ground base stations with the maximum weighted total spectrum efficiency of all users as the optimization index. The dual-layer optimization architecture consists of an outer layer and an inner layer. In the outer layer, a greedy algorithm with a backtracking mechanism is used to perform discrete deployment of undeployed airborne base stations in a loop. In the inner layer, the tilt angle parameters of ground base stations are obtained iteratively using stochastic perturbation gradient estimation and adaptive optimization algorithm.

2. The three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and air base station deployment location according to claim 1, characterized in that, A two-layer optimization architecture is constructed, using the maximization of the weighted total spectral efficiency for all users as the optimization metric for deploying airborne base stations and adjusting the antenna downtilt angles of ground base stations. This includes: S1: Select any air base station from the undeployed set as a candidate base station, and move the candidate base station into the deployed set to generate a temporary deployment set; S2, based on the fixed location set, temporary deployment set and user grid point set of ground base stations, outputs the optimal antenna tilt angle set through random perturbation gradient estimation and adaptive optimization algorithm, and obtains the system spectral efficiency gain based on the optimal antenna tilt angle set; S3, iterate through the system spectral efficiency gain of all airborne base stations, and use the system spectral efficiency gain of any airborne base station compared with a preset gain threshold as a criterion to determine whether to move the airborne base station into the deployed set: If the maximum system spectral efficiency gain of an airborne base station exceeds a preset gain threshold, the airborne base station will be moved into the deployed set and a new historical gain set for the current round will be generated after its removal. If not, then select from all airborne base stations the airborne base station whose maximum system spectral efficiency gain exceeds the preset replacement threshold as the candidate point; wherein, the preset gain threshold is greater than the preset replacement threshold. S4. If the current round is the first round, move the candidate point into the deployed set; if the current round is not the first round, compare the system spectrum efficiency gain of the candidate point with the airborne base station with the maximum value in the historical gain set of the previous round, and select the airborne base station corresponding to the larger value of the system spectrum efficiency gain to move into the deployed set. S5. If there is no candidate in the current round, a backtracking mechanism is introduced to retrieve the historical gain set of the previous round and execute S3~S4 operations. If no candidate is found after the execution is completed, the backtracking is determined to have failed. S6, update the deployment set and re-execute the loop to move into the deployed set until the undeployed set is empty or backtracking fails, and output the deployed set and the optimal antenna tilt set.

3. The three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and air base station deployment location according to claim 1, characterized in that, In S2, the optimal antenna tilt angle set is output iteratively through random perturbation gradient estimation and adaptive optimization algorithm, including: S201, Generate the channel sample matrix This matrix contains the channel gain from each air base station to the user grid point; S202, with the initial downtilt angle parameter set of the antenna as the independent variable, and the channel sample matrix as the independent variable. Construct a system spectral efficiency function for the conditional parameters; S203, by applying random perturbations, zero-order gradient simulation is performed on the system spectral efficiency function to obtain gradient estimation components; S204 uses the gradient estimation components of the current iteration as the historical gradient vector and employs the Adama stochastic algorithm to update the antenna downtilt angle.

4. The three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and air base station deployment location according to claim 3, characterized in that, S203, by applying a random perturbation to perform a zero-order gradient simulation of the system's spectral efficiency function, obtains the gradient estimation components, including: S2031, Generate perturbation vector: Adjust antenna downtilt parameters Perform random perturbations to generate perturbation directions. Its elements are drawn independently from a uniform or normal distribution. Random values; S2032, Calculate the spectral efficiency function of the system after the disturbance: in the... In the next iteration, the antenna downtilt angle Apply a perturbation and calculate the objective function values ​​before and after the perturbation: in, To control the magnitude of the disturbance, It is the perturbation vector. This represents the calculated spectral efficiency; S2033, SPSA estimates antenna downtilt angle based on objective function values ​​before and after perturbation. gradient: In the formula, It is a simulated estimation of the gradient of the objective function.

5. The three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and air base station deployment location according to claim 2, characterized in that, In S3, a greedy algorithm is introduced to progressively select the airborne base station that can bring the maximum gain from the candidate base stations, using gain difference as the evaluation index. This includes: For each candidate base station, obtain the difference in system spectral efficiency gain before and after deploying the candidate base station. ; A greedy algorithm is used to obtain the gain difference of each candidate base station. In each round of selection, the remaining candidate set is selected. Choose the air base station location that brings the greatest gain difference. : The selected airborne base station location is added to the deployed set. In the process, the spectral efficiency of the updated system is calculated.

6. The three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and air base station deployment location according to claim 5, characterized in that, In S5, if a certain candidate position The gain difference did not reach the preset threshold. ,Right now: If this happens, the backtracking mechanism will be triggered, and the backtracking mechanism will retrieve data from the historical gain record set. Choose an alternative position with a higher gain and reselect.

7. The three-dimensional communication coverage method for joint optimization of ground base station downtilt angle and air base station deployment location according to claim 1, characterized in that, The backtracking selection formula is: 。 8. A system for implementing a three-dimensional communication coverage method for jointly optimizing the downtilt angle of ground base stations and the deployment location of airborne base stations as described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire the fixed location set of ground base stations, the deployed and non-deployed sets of air base stations, and the user grid point set; The deployment optimization module is used to deploy air base stations and adjust the antenna downtilt angle of ground base stations by using a two-layer alternating optimization method with the maximum weighted total spectral efficiency of all users as the optimization index.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images