Lightweight slice resource allocation method and system for unmanned aerial vehicle auxiliary network

Abstract

The invention belongs to the technical field of unmanned aerial vehicles, discloses a lightweight slice resource allocation method for an unmanned aerial vehicle auxiliary network, and aims to maximize the minimum user rate by jointly optimizing channel allocation, power allocation and unmanned aerial vehicle deployment. The UAWN slicing problem is modeled as non-convex mixed integer nonlinear programming. Considering that the computing power of the unmanned aerial vehicle is limited, a low-complexity decomposition framework is provided, and the problem is solved by combining problem decomposition, integer relaxation and a batch coordinate descent method. In addition, an optimal rounding algorithm is designed to obtain a relaxation solution, and the feasibility and optimality of the solution after rounding are ensured. In order to verify the effectiveness of the scheme, a large number of numerical experiments are performed. A simulation result shows that the convergence speed and the performance of the scheme are superior to those of some existing algorithms.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicles, and in particular, relates to a lightweight slice resource configuration method and system for unmanned aerial vehicle-assisted networks. Background Art

[0002] Non-terrestrial networks (NTNs) are considered to be one of the pillars of the upcoming 6G network, which can provide services for a variety of critical applications such as environmental monitoring, public safety, disaster search and rescue, and procurement delivery. In particular, UAWNs have attracted considerable attention in academia and industry due to their significant features such as flexible deployment, fast relay, and line-of-sight (LoS) links. In recent years, with the rise of the low-altitude economy, the diversity of UAWN services has shown a rapid growth trend. Since one network architecture does not apply to all situations, it is unrealistic to expect a single UAWN to meet these diverse service requirements. Network slicing is a virtualized independent logical network created on a common physical infrastructure to meet the specific needs of different services. By introducing network slicing technology into UAWN, it is possible to avoid deploying independent UAWNs for each service type, thereby rationally allocating UAWN resources.

[0003] Because network slicing can eliminate the need to deploy separate networks for each service type, it can reduce the number of drones on the basis of providing differentiated services. The reduction in the number of drones will lead to a proportional decrease in the capital and operating costs of UAWN, which also means that the probability of interference and collision between drones is reduced, thereby simplifying the coordination of drones in the drone network. In addition, with the help of technologies that support network slicing, such as Network Functions Virtualization (NFV) and Software Defined Network (SDN), the physical network equipment installed on drones can be significantly reduced, and drones can use more energy for data transmission, thereby extending their service time.

[0004] To fully unleash the benefits of network slicing, several challenges need to be addressed. First, the aerial base station (BS) should consider the deployment of drones and the allocation of network resources to adapt to the ever-changing three-dimensional network. Second, due to the limited computing power and battery capacity of drones, low-complexity network slicing solutions must be designed. Finally, in resource-constrained UAWNs, guaranteeing the service level agreement (SLA) of the slices is also a challenging task. However, in addition to the scarcity of research on these issues, existing research has failed to provide a comprehensive solution to overcome all these challenges. Although some recent studies have jointly considered drone deployment and resource allocation, these solutions are highly complex, slow to converge, and lack SLA guarantees, which limits their application in drone networks.

[0005] First, the complexity of existing solutions is too high, which exceeds the computing power of UAVs. Second, although some researchers have adopted certain decomposition techniques to simplify the problem, these decompositions are usually based on heuristics and cannot guarantee the equivalence of the decomposed problem with the original problem, resulting in their poor performance. Third, these studies lack SLA modeling, which will lead to SLA violations that are unacceptable to slice users. In contrast, the proposed decomposition method is theoretically reasonable, in a sense, the decomposed problem is equivalent to the original problem. At the same time, the model combines the SLA constraints of the network slice, so that the proposed solution can guarantee the user's Quality of Service (QoS). Finally, it should be noted that the proposed solution framework is lightweight and can converge within a few milliseconds, making it suitable for UAWNs with limited computing power in rapidly changing environments.

[0006] Through the above analysis, the problems and defects of the prior art are as follows:

[0007] (1) The complexity of existing solutions is too high, exceeding the computing power of drones.

[0008] (2) Although some researchers have adopted certain decomposition techniques to simplify the problem, these decompositions are usually based on heuristics and cannot guarantee the equivalence of the decomposed problem with the original problem, resulting in their poor performance.

[0009] (3) These studies lack SLA modeling, which may lead to SLA violations that are unacceptable to slice users. Summary of the invention

[0010] In view of the problems existing in the prior art, the present invention provides a lightweight slice resource configuration method for drone-assisted networks.

[0011] The present invention is implemented as follows: a lightweight slice resource configuration method for a drone-assisted network includes:

[0012] Step 1: Divide the problem into resource allocation problem and drone deployment problem equivalently;

[0013] Step 2, by analyzing the properties of the sub-problems, deriving explicit solutions to the sub-problems, and deriving explicit expressions of the main problem;

[0014] Step 3, prove that the relaxed main problem is a pseudo-concave problem, whose stable point corresponds to the global optimal solution, and use the KKT condition to find a lightweight decomposition algorithm for the global optimal solution, and design the optimal rounding algorithm to minimize the rounding loss of the relaxed solution.

[0015] Further, the derivation method is as follows:

[0016] Basal network model

[0017] First, we focus on downlink transmission, where the UAV is equipped with a maximum transmission power of P max The bandwidth available to the drone is denoted as B tot ; Assume that the projection position of the drone on the horizontal plane is the center of the drone network; let θ be the beam width and h be the height of the drone; then the following coverage conditions must be met:

[0018]

[0019] R in the above formula is the radius of the UAWN; considering that the flight capability of the UAWN is limited, its altitude is limited by the following conditions:

[0020]

[0021] In the above formula, c is the flight speed of the drone. is the flight altitude of the UAV in the previous time slot, T s is the duration of each time slot in the network;

[0022] Network Slicing Model

[0023] use represents the set of network slices, and the set of user devices in slice i is recorded as u ij represents the jth user equipment in slice i, user equipment u ij The two-dimensional coordinates of User equipment ij The square distance from the drone is in Assume g0 is the channel power gain at a reference distance of one meter; according to existing literature, user equipment u ijThe channel power gain with the drone can be modeled as:

[0024]

[0025] Where, α is the path loss exponent; in represents the small-scale fading channel coefficient of the Rayleigh distribution, that is, Follows a non-central chi-square distribution;

[0026] Use B i represents the channel bandwidth of slice i; it should be noted that due to service differentiation, the channel bandwidth of different slices is different; a non-negative integer x ij Indicates the allocation to user equipment u ij The number of channels; obviously, x ij The following capacity constraints should be met:

[0027]

[0028] Among them, the drone sends a signal to the user device u ij The transmitted power is denoted as p ij , the following capacity constraints must be met:

[0029]

[0030] To guarantee the SLA, each slice should provide a guaranteed data rate to the subscriber, namely:

[0031]

[0032] Among them, σ 2 is the power of the background noise, is the minimum rate guaranteed by slice i; in addition, each user device can request a customized service rate Considering the tractability of the problem, this constraint is modeled as:

[0033]

[0034] in, is a positive integer that satisfies

[0035] UAWN network slicing model

[0036] The UAWN network slicing problem can be expressed as:

[0037]

[0038] st(1)-(2),(4)-(7)(9)

[0039] h min ≤h≤h max (10)

[0040]

[0041] Among them, h min is the minimum permitted flight altitude of the UAV, h max is the maximum allowed flight altitude of the UAV. Further, the problem is equivalently divided into:

[0042] Breakdown of USP

[0043] The objective function of USP is recorded as f(x,p,h)=min i,j f ij (x ij ,p ij ,h); the so-called problem partitioning means projecting the objective function and constraint set of USP onto the (x,p) axis at the same time. After projecting USP onto the (x,p) axis, the following main problem can be obtained:

[0044]

[0045] st(x,p)∈V(14)

[0046] Where X and P are defined as: X:={x|x satisfies (4) and (7)}, and P:={p|p satisfies (5) and (11)}; the set V is defined as:

[0047] V: = {p| exists h∈H B So that p satisfies (6)} (15)

[0048] In the above formula, H B is the set of intervals defined by constraints (1), (2) and (10), that is, in and They are and

[0049] In MP, v(x,p) is defined as a subproblem:

[0050]

[0051] Explicit solution to subproblem

[0052] The explicit solution of SP can be derived directly, which is independent of x and p, and the following theorem about SP can be derived:

[0053] For each fixed pair If the corresponding SP is feasible, then its optimal solution must be

[0054] Explicit Formulation of the Master Problem and Its Relaxation

[0055] Let n represent the index pair (i,j), Represents all user devices in the network, that is, The number of user devices in the system is denoted by N, and the relaxed MP can be expressed as the following problem:

[0056]

[0057] std T x≤B tot (19)

[0058] 1 T p≤P max (20)

[0059] x≥ x ,p≥ p (twenty one)

[0060] Where d and 1 are the coefficients of constraints (4) and (5), respectively. β n is the number of B in the corresponding index pair (i, j) i , For a clearer representation, the objective function of MP-Relax is represented by ω(x, p); and it can be shown that the inequality constraints in constraints (19) and (20) can be replaced by equality constraints.

[0061] Furthermore, the lightweight decomposition algorithm for finding the global optimal solution using the KKT condition is:

[0062] Relaxed master problem solution

[0063] The BCD method will be used to solve MP-Relax. Based on the pseudo-convex optimization theory and the characterization of the MP-Relax solution, the KKT condition is used to solve two sub-problems, namely the power allocation sub-problem and the sub-channel allocation sub-problem, and then the BCD algorithm for solving the global optimal solution of MP-Relax is proposed.

[0064] Power allocation subproblem

[0065] By fixing a stationary point x on the domain, we can obtain the following power allocation subproblem:

[0066]

[0067] st1 T p≤P max (twenty three)

[0068] P≥ P (twenty four)

[0069] Among them, a n =β n x n ; It can be seen that the power allocation subproblem is a non-smooth convex programming. Since its objective function is strictly convex, it has a unique solution; assuming that the optimal solution is (p * ,z * ),definition:

[0070]

[0071] Subchannel allocation subproblem;

[0072] BCD algorithm for the relaxed master problem.

[0073] Furthermore, the sub-channel allocation sub-problem is:

[0074] At the x coordinate, the following subchannel allocation subproblem needs to be solved for a fixed p:

[0075]

[0076] std T x≤B max (28)

[0077] x≥ x (29)

[0078] where c n =β n ln(1+b n p n );because Sub x The objective function of is strictly convex, which indicates the uniqueness of its optimal solution; similarly, assuming that the optimal value is t * , assuming the optimal solution is It can be deduced that:

[0079]

[0080] Further, the BCD algorithm of the relaxed master problem:

[0081] Based on KKT x and KKT p Algorithm,A BCD algorithm based on loop iteration rules is proposed to solve MP-Relax.

[0082] Another object of the present invention is to provide a lightweight slice resource configuration system for drone-assisted networks, including:

[0083] A partitioning module is used to divide the problem into resource allocation problem and drone deployment problem;

[0084] The derivation module is used to derive explicit solutions to sub-problems by analyzing the properties of sub-problems and derive explicit expressions of the main problem;

[0085] The solution module is used to prove that the relaxed main problem is a pseudo-concave problem, whose stable point corresponds to the global optimal solution. Based on this, the KKT condition is used to find a lightweight decomposition algorithm for the global optimal solution, and the optimal rounding algorithm is designed to minimize the rounding loss of the relaxed solution.

[0086] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the lightweight slice resource configuration method for drone-assisted networks.

[0087] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the lightweight slice resource configuration method for drone-assisted networks.

[0088] Another object of the present invention is to provide an information data processing terminal, which is used to implement the lightweight slice resource configuration system for the drone-assisted network.

[0089] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0090] First, through the study of existing technical solutions, we found the following technical problems: the complexity of existing solutions is too high, which exceeds the computing power of drones; although some researchers have adopted certain decomposition techniques to simplify the problem, these decompositions are usually based on heuristics, and the equivalence of the decomposed problem with the original problem cannot be guaranteed, resulting in poor performance; these studies lack SLA modeling, which will lead to SLA violations that slice users cannot accept. The key to solving these problems is to comprehensively consider drone deployment, slice resource allocation and SLA modeling, reduce the complexity of the solution, and improve the practical application.

[0091] This paper studies the UAWN Network Slicing Problem (USP) by considering UAV deployment and resource allocation, with the goal of maximizing the minimum user rate. Multiple parameters such as the flight altitude, power allocation, and channel allocation of the UAV are jointly optimized. Studies have shown that this problem is a large-scale non-convex mixed integer nonlinear programming problem that is extremely difficult to solve. Therefore, a new Batch Coordinate Descent-based Decomposition (BCD) is designed based on the structure of the problem. 2 ) framework to solve it.

[0092] Different from the existing heuristic decomposition methods, the present invention is based on a rigorous theoretical foundation. First, the USP is decomposed into a master problem (MP) and a subproblem (SP) by projection technology, where SP is explicitly solved. It is proved that the relaxed MP is a pseudo-convex problem, and the batch coordinate descent (BCD) method is used to solve it. By utilizing the Karush-Kuhn-Tucker (KKT) condition, a corresponding fast algorithm is designed for each subproblem under each coordinate, and its complexity is mainly determined by the array sorting. Finally, a rounding algorithm is designed to round the relaxed solution to its integer counterpart to ensure that the rounding loss is minimized. The main innovations of the present invention are as follows:

[0093] (1) The UAWN slicing problem is modeled as a nonconvex mixed-integer nonlinear programming problem, and a low-complexity decomposition framework is proposed. This decomposition framework demonstrates the feasibility of designing a log-linear algorithm that can solve large-scale nonconvex and nonsmooth problems in NTNs within milliseconds.

[0094] (2) The scheme discovers that the throughput maximization problem has a good pseudo-concave property, which can be used to alleviate the non-convex challenges that are prevalent in NTNs.

[0095] (3) Through a large number of numerical experiments, the scheme reveals the complex relationship between slice resource allocation and drone deployment, providing guidance for drone deployment and network slice configuration in UAWN.

[0096] The present invention proposes a low-complexity optimization framework for the slicing problem in drone-assisted networks, thereby jointly optimizing resource allocation and drone deployment. The problem is equivalently divided into a resource allocation problem and a drone deployment problem. By analyzing the properties of the sub-problems, the explicit solutions of the sub-problems are derived, thereby deriving an explicit expression of the main problem. It is then proved that the relaxed main problem is a pseudo-concave problem, and its stationary point corresponds to the global optimal solution. Based on this, a lightweight decomposition algorithm using the KKT condition to find the global optimal solution is proposed. In addition, an optimal rounding algorithm is designed to minimize the rounding loss of the relaxed solution. Numerical results show that compared with existing algorithms, the algorithm proposed in the present invention has significant advantages such as fast convergence speed and high user throughput.

[0097] Second, as auxiliary evidence of the inventiveness of the present invention, it is also reflected in the following important aspects:

[0098] (1) The expected benefits and commercial value of the technical solution of the present invention after transformation are:

[0099] With the commercialization of network slicing technology and the gradual maturity of drone-assisted network research, the present invention can be applied to future commercial drone-assisted networks to provide operators with low-complexity customized business solutions, which has certain commercial value.

[0100] (2) The technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0101] There is an urgent demand for drone-assisted network slicing technology both at home and abroad. This invention proposes a low-complexity decomposition framework for the first time, which enables the optimization model of drone-assisted network slicing to converge quickly, with performance better than some known algorithms, and provides guidance for drone deployment and slice resource allocation.

[0102] (3) The technical solution of the present invention solves the technical problems that people have been eager to solve but have never been able to solve successfully:

[0103] Although the technical solutions for drone-assisted network slicing at home and abroad can jointly consider drone deployment and slice resource allocation, these solutions have the defects of high complexity and slow convergence, which poses a challenge to the limited computing power of drones. The present invention reduces the complexity of the drone-assisted network slicing problem through problem decomposition and mathematical methods, and takes into account SLA requirements, which has higher practical application value than traditional solutions.

[0104] The present invention breaks the industry's technical prejudice that the problem of drone-assisted network resource allocation cannot be solved quickly.

[0105] Third, the technical solution of the present invention solves the following technical problems of the prior art in industrial applications and has achieved significant technical progress:

[0106] 1. Problems with the prior art:

[0107] Unreasonable resource allocation: Existing drone-assisted networks have problems with inefficiency and imprecision in resource allocation, and cannot effectively respond to different needs and environmental changes. Especially in the context of multi-channel signal processing and high-frequency requirements, traditional methods often cannot adjust resources according to dynamic environments, resulting in resource waste or poor signal quality.

[0108] Insufficient optimization of drone deployment: Traditional resource allocation methods often ignore the deployment efficiency of drones and their signal coverage capabilities, resulting in the failure of drones to maximize their communication auxiliary role during actual deployment, affecting the performance of the overall network.

[0109] Complex optimization algorithms: Existing optimization algorithms are often complex and computationally intensive, resulting in slow processing speeds, especially in large-scale networks or real-time communication environments, and cannot meet real-time requirements.

[0110] 2. Technical advancement of the present invention:

[0111] Effective integration of resource allocation and drone deployment: This invention effectively improves the efficiency of resource allocation and the rationality of drone deployment by dividing the resource allocation problem and the drone deployment problem into two sub-problems and solving them through an optimization algorithm. This solution can adjust resource configuration and the deployment location of drones in real time, maximize the signal quality and resource utilization of the network, and solve the problem of unreasonable resource allocation in traditional technologies.

[0112] Implementation of global optimal solution: By adopting the KKT condition and relaxation technology of pseudo-concave problem, the present invention can quickly calculate the global optimal solution in a complex multi-channel communication environment, ensuring the optimization of resource allocation and UAV deployment, and significantly improving the performance and stability of the network.

[0113] Lightweight decomposition algorithm and optimal rounding algorithm: In order to improve the computational efficiency, the present invention designs a lightweight decomposition algorithm and an optimal rounding algorithm, which effectively reduces the computational complexity and reduces the rounding error. This technical solution is particularly suitable for real-time data processing and resource scheduling of large-scale networks, so that it can respond to changes quickly and reduce delays in practical applications.

[0114] 3. Advantages of industrial applications:

[0115] Optimized signal coverage: The present invention can achieve optimal deployment of drones and optimal allocation of resources in a dynamic environment, effectively improving the signal quality and coverage of the communication network. It is particularly suitable for application scenarios that require high signal strength and high bandwidth, such as drone communications, smart cities, and industrial Internet of Things.

[0116] Improved system efficiency: By adopting the optimization method of the present invention, network resources are maximized, avoiding the problem of resource waste in traditional methods and improving the working efficiency of the entire system.

[0117] Enhanced real-time processing capability: The optimization algorithm design of the present invention enables the system to quickly respond to and process real-time data, greatly improving the real-time performance of the system, adapting to the needs of dynamic network environments, and meeting the high standards for real-time and high efficiency in industrial applications.

[0118] In summary, the technical solution of the present invention overcomes multiple technical bottlenecks in resource allocation and optimization algorithms on the basis of existing technologies, significantly improves the performance of drone-assisted networks, and has broad industrial application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0119] Figure 1 It is a flow chart of a lightweight slice resource configuration method for drone-assisted networks provided in an embodiment of the present invention.

[0120] Figure 2 This is a structural block diagram of a lightweight slice resource configuration system for drone-assisted networks provided in an embodiment of the present invention.

[0121] Figure 3 This is a diagram of a UAWN slice application scenario provided by an embodiment of the present invention.

[0122] Figure 4 It is a solution framework diagram provided by an embodiment of the present invention.

[0123] Figure 5 It is a convergence time diagram of different implementations provided by the embodiments of the present invention.

[0124] Figure 6 This is a comparison diagram of the maximum and minimum rates of users in different network scenarios provided by an embodiment of the present invention.

[0125] (a) The impact of the number of users on the maximum-minimum user rate; (b) The impact of total bandwidth on the maximum-minimum user rate; (c) The impact of maximum power on the maximum-minimum user rate; (d) The algorithm convergence process and the change of target value. DETAILED DESCRIPTION

[0126] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0127] like Figure 1As shown, a lightweight slice resource configuration method for a drone-assisted network provided by an embodiment of the present invention includes the following steps:

[0128] S101, divide the problem into resource allocation problem and drone deployment problem equivalently;

[0129] S102, by analyzing the properties of the sub-problems, deriving explicit solutions to the sub-problems, and deriving explicit expressions of the main problem;

[0130] S103, prove that the relaxed main problem is a pseudo-concave problem, whose stable point corresponds to the global optimal solution, and use the KKT condition to find a lightweight decomposition algorithm for the global optimal solution, and design an optimal rounding algorithm to minimize the rounding loss of the relaxed solution.

[0131] 1. Signal data processing for resource allocation problem

[0132] In step 1, after the problem is equivalently divided into resource allocation problem and drone deployment problem, the resources in the network (such as spectrum, bandwidth and computing power) are first accurately quantified. The processing of signal data at this stage mainly involves mapping the spectrum allocation of the wireless communication network with the deployment requirements of the actual drone. Through the optimization model based on signal power control and spectrum utilization, the requirements of each communication link (such as bandwidth, delay, throughput) are compared with the current resource configuration to obtain the allocation strategy of signal data. At this time, the collection and analysis of data depends on real-time channel information, drone location and mobility, and the signal processing algorithm is used to estimate the interference and signal strength of each link to form a preliminary plan for resource allocation.

[0133] 2. Sub-problem analysis and derivation of solutions

[0134] In step 2, after analyzing the properties of the resource allocation problem and the drone deployment problem, their respective explicit solutions are derived. At this stage, the processing of signal data requires the solution of each subproblem to be related through the calculation of signal interference and network topology. Specifically, the resource allocation subproblem requires the optimal allocation of bandwidth and computing resources in the drone-assisted network, taking into account the interference of different frequency channels, the signal reception strength of users, and power control. This process involves the application of signal enhancement and noise reduction techniques, and the solution of each subproblem is optimized through digital signal processing algorithms. At the same time, in the derivation of the solution, the coverage of the wireless signal and the adaptability of the space-time signal are taken into account to derive the best configuration of the overall network.

[0135] 3. Relaxation of pseudo-concave problem and calculation of global optimal solution

[0136] In step 3, after relaxing the main problem to a pseudo-concave problem, the processing of signal data begins to enter the optimization stage. At this stage, by analyzing the structural characteristics of the problem, the explicit expression of the relaxed problem is used to form the calculation of stable points, which correspond to the global optimal solution of the system. The key to signal data processing is to use the KKT conditions (Karush-Kuhn-Tucker Conditions) to find the global optimal solution. Through the optimization relationship between the constraints and the objective function, the signal data will be input into the KKT condition solver to obtain the optimal resource allocation and drone deployment plan under the current network configuration. In this process, the signal data undergoes multiple iterations and adjustments to ensure that the signal strength and transmission quality of each wireless link are maximized.

[0137] 4. Application of lightweight decomposition algorithm and optimal rounding algorithm

[0138] Finally, the signal data is refined through a lightweight decomposition algorithm and an optimal rounding algorithm. The rounding loss of the relaxed solution is minimized to ensure the application of the global optimal solution in practical implementation. The processing of signal data at this stage involves the actual mapping of the calculated theoretical optimal value and the discretization of resource allocation using an approximate algorithm. The designed optimal rounding algorithm solves the rounding error caused by the discretization operation, ensuring that the resource allocation of each signal link will not suffer excessive performance loss in actual deployment. In this process, signal interference control and reasonable spectrum division are fully guaranteed, and ultimately the optimal solution for resource allocation in the drone-assisted network is achieved.

[0139] The derivation method provided by the embodiment of the present invention is as follows:

[0140] Basal network model

[0141] First, we focus on downlink transmission, where the UAV is equipped with a maximum transmission power of P max The bandwidth available to the drone is denoted as B tot ; Assume that the projection position of the drone on the horizontal plane is the center of the drone network; let θ be the beam width and h be the height of the drone; then the following coverage conditions must be met:

[0142] h·tanθ≥R (1)

[0143] R in the above formula is the radius of the UAWN; considering that the flight capability of the UAWN is limited, its altitude is limited by the following conditions:

[0144]

[0145] In the above formula, c is the flight speed of the drone. is the flight altitude of the UAV in the previous time slot, T sis the duration of each time slot in the network;

[0146] Network Slicing Model

[0147] use represents the set of network slices, and the set of user devices in slice i is recorded as u ij represents the jth user equipment in slice i, user equipment u ij The two-dimensional coordinates of User equipment ij The square distance from the drone is in Assume g0 is the channel power gain at a reference distance of one meter; according to existing literature, user equipment u ij The channel power gain with the drone can be modeled as:

[0148]

[0149] Where, α is the path loss exponent; in represents the small-scale fading channel coefficient of the Rayleigh distribution, that is, Follows a non-central chi-square distribution;

[0150] Use B i represents the channel bandwidth of slice i; it should be noted that due to service differentiation, the channel bandwidth of different slices is different; a non-negative integer x ij Indicates the allocation to user equipment u ij The number of channels; obviously, x ij The following capacity constraints should be met:

[0151]

[0152] Among them, the drone sends a signal to the user device u ij The transmitted power is denoted as p ij , the following capacity constraints must be met:

[0153]

[0154] To guarantee the SLA, each slice should provide a guaranteed data rate to the subscriber, namely:

[0155]

[0156] Among them, σ 2 is the power of the background noise, is the minimum rate guaranteed by slice i; in addition, each user device can request a customized service rate Considering the tractability of the problem, this constraint is modeled as:

[0157]

[0158] in, is a positive integer that satisfies

[0159] UAWN network slicing model

[0160] The UAWN network slicing problem can be expressed as:

[0161]

[0162] st(1)-(2),(4)-(7)(9)

[0163] h min ≤h≤h max (10)

[0164]

[0165] Among them, h min is the minimum permitted flight altitude of the UAV, h max is the maximum allowed flight altitude of the UAV, the problem provided by the embodiment of the present invention is equivalently divided into:

[0166] Breakdown of USP

[0167] The objective function of USP is recorded as f(x,p,h)=min i,j f ij (x ij ,p ij ,h); the so-called problem partitioning means projecting the objective function and constraint set of USP onto the (x,p) axis at the same time. After projecting USP onto the (x,p) axis, the following main problem can be obtained:

[0168]

[0169] st(x,p)∈V(14)

[0170] Where X and P are defined as: X:={x|x satisfies (4) and (7)}, and P:={p|p satisfies (5) and (11)}; the set V is defined as:

[0171] V: = {p| exists h∈H B So that p satisfies (6)} (15)

[0172] In the above formula, H B is the set of intervals defined by constraints (1), (2) and (10), that is, in and They are and

[0173] In MP, v(x,p) is defined as a subproblem:

[0174]

[0175] Explicit solution to subproblem

[0176] The explicit solution of SP can be derived directly, which is independent of x and p, and the following theorem about SP can be derived:

[0177] For each fixed pair If the corresponding SP is feasible, then its optimal solution must be

[0178] Explicit Formulation of the Master Problem and Its Relaxation

[0179] Let n represent the index pair (i,j), Represents all user devices in the network, that is, The number of user devices in the system is denoted by N, and the relaxed MP can be expressed as the following problem:

[0180]

[0181] std T x≤B tot (19)

[0182] 1 T p≤P max (20)

[0183] x≥ x ,p≥ p (twenty one)

[0184] Where d and 1 are the coefficients of constraints (4) and (5), respectively. β n is the number of B in the corresponding index pair (i, j) i , For a clearer representation, the objective function of MP-Relax is represented by ω(x, p); and it can be shown that the inequality constraints in constraints (19) and (20) can be replaced by equality constraints.

[0185] The lightweight decomposition algorithm for finding the global optimal solution using the KKT condition provided in the embodiment of the present invention is:

[0186] Relaxed master problem solution

[0187] The BCD method will be used to solve MP-Relax. Based on the pseudo-convex optimization theory and the characterization of the MP-Relax solution, the KKT condition is used to solve two sub-problems, namely the power allocation sub-problem and the sub-channel allocation sub-problem, and then the BCD algorithm for solving the global optimal solution of MP-Relax is proposed.

[0188] Power allocation subproblem

[0189] By fixing a stationary point x on the domain, we can obtain the following power allocation subproblem:

[0190]

[0191] st1 T p≤P max (twenty three)

[0192] P≥ P (twenty four)

[0193] Among them, a n =β n x n ; It can be seen that the power allocation subproblem is a non-smooth convex programming. Since its objective function is strictly convex, it has a unique solution; assuming that the optimal solution is (p * ,z * ),definition:

[0194]

[0195] Subchannel allocation subproblem;

[0196] BCD algorithm for the relaxed master problem.

[0197] The sub-channel allocation sub-problem provided by the embodiment of the present invention is as follows:

[0198] At the x coordinate, the following subchannel allocation subproblem needs to be solved for a fixed p:

[0199]

[0200] std T x≤B max (28)

[0201] x≥ x (29)

[0202] where c n =β n ln(1+b n p n ); due to c n ≠0, Sub x The objective function of is strictly convex, which indicates the uniqueness of its optimal solution; similarly, assuming that the optimal value is t * , assuming the optimal solution is It can be deduced that:

[0203]

[0204] The BCD algorithm for the relaxed master problem provided by the embodiment of the present invention is as follows:

[0205] Based on KKT x and KKT p Algorithm,A BCD algorithm based on loop iteration rules is proposed to solve MP-Relax.

[0206] like Figure 2 As shown, a lightweight slice resource configuration system for a drone-assisted network provided by an embodiment of the present invention includes:

[0207] A partitioning module is used to divide the problem into resource allocation problem and drone deployment problem;

[0208] The derivation module is used to derive explicit solutions to sub-problems by analyzing the properties of sub-problems and derive explicit expressions of the main problem;

[0209] The solution module is used to prove that the relaxed main problem is a pseudo-concave problem, whose stable point corresponds to the global optimal solution. Based on this, the KKT condition is used to find a lightweight decomposition algorithm for the global optimal solution, and the optimal rounding algorithm is designed to minimize the rounding loss of the relaxed solution.

[0210] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the lightweight slice resource configuration method for drone-assisted networks.

[0211] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the lightweight slice resource configuration method for drone-assisted networks.

[0212] Another object of the present invention is to provide an information data processing terminal, which is used to implement the lightweight slice resource configuration system for the drone-assisted network.

[0213] The present invention is specifically implemented:

[0214] 1. System model

[0215] Basal network model

[0216] like Figure 3 As shown, we focus on a cellular UAV-assisted wireless network (i.e., UAWN 1), which consists of a UAV and several ground small base stations. Each cell consists of multiple user equipment (UE) and a ground base station that provides services for it. In the scenario considered, it is assumed that the base station cannot provide satisfactory QoS for the user equipment, so UAVs are deployed to supplement the base station. The designed framework works in a time slot manner, and in each time slot, the altitude of the UAV is periodically updated to adapt to the network status. The duration of each time slot is denoted as T s .

[0217] In the present invention, we first focus on downlink transmission, where the drone is equipped with a maximum transmission power of P max The bandwidth available to the drone is denoted as B tot Assume that the projection position of the drone on the horizontal plane is the center of the drone network. Let θ be the beam width and h be the height of the drone. Then the following coverage conditions must be met:

[0218] h·tanθ≥R (1)

[0219] R in the above formula is the radius of the UAWN. Considering that the flight capability of the UAWN is limited, its altitude is restricted by the following conditions:

[0220]

[0221] In the above formula, c is the flight speed of the drone. is the flight altitude of the UAV in the previous time slot.

[0222] Network Slicing Model

[0223] Due to the heterogeneity of user devices, the basic network can be divided into several network slices, each of which will be customized to support a certain service type. represents the set of network slices, and the set of user devices in slice i is recorded as u ij represents the jth user equipment in slice i, user equipment u ij The two-dimensional coordinates of User equipment ij The square distance from the drone is in Assume g0 is the channel power gain at a reference distance of one meter. According to existing literature, user equipment u ij The channel power gain with the drone can be modeled as:

[0224]

[0225] Where α is the path loss exponent. in represents the small-scale fading channel coefficient of the Rayleigh distribution, that is, Follows a noncentral chi-square distribution.

[0226] Use B i represents the channel bandwidth of slice i. It should be noted that due to service differentiation, the channel bandwidths of different slices are different. ij Indicates the allocation to user equipment u ij Obviously, x ij The following capacity constraints should be met:

[0227]

[0228] Among them, the drone sends a signal to the user device u ij The transmitted power is denoted as p ij , the following capacity constraints must be met:

[0229]

[0230] 6G end-to-end network slicing spans multiple areas and adjusts multi-dimensional service level agreements in terms of latency, throughput, reliability, etc. Considering that the main part of UAWN is the Radio Access Network (RAN), the model uses transmission rate as a key indicator of SLA. In order to guarantee the SLA, each slice should provide a guaranteed data rate for subscribed users, namely:

[0231]

[0232] Among them, σ 2 is the power of the background noise, is the minimum rate guaranteed by slice i. In addition, each user device can request a customized service rate Considering the tractability of the problem, this constraint is modeled as:

[0233]

[0234] in, is a positive integer that satisfies

[0235] UAWN network slicing model

[0236] In the UAWN network slicing problem, our goal is to find the optimal values ​​of channel allocation x, power allocation p, and drone height h, and maximize the minimum user rate. The UAWN Network Slicing Problem (USP) can be expressed as:

[0237]

[0238] st(1)-(2),(4)-(7)(9)

[0239] h min ≤h≤h max (10)

[0240]

[0241] In USP, constraint (10) ensures that the altitude of the drone is subject to a lower limit h min to avoid collision with ground obstacles, and it is also subject to the upper limit h max The constraint is the maximum allowed flight height of the drone. It can be seen that USP is a large-scale non-convex mixed integer problem. Even if x is relaxed to a continuous variable, the objective function (8) is still non-smooth and non-convex. In addition, USP contains a large number of non-convex constraints as shown in formula (6), which makes the solution of USP extremely difficult.

[0242] 2. Problem decomposition and drone deployment optimization

[0243] This section will explain how to solve the BCD of USP 2 The main steps of the framework are as follows Figure 4 The framework consists of five steps. First, the USP is divided into the main problem and subproblems using projection techniques. Second, the structure of the subproblems is checked and its closed-form solution is derived to obtain the optimal solution h for each fixed (x, p) * , so the explicit expression of the master problem is derived. The third step is to relax the integer constraints of the master problem to obtain MP-Relax, which is proved to be a pseudo-concave problem and each stationary point corresponds to the global optimal solution. The fourth step is to use the batch coordinate descent method to find the global optimal solution of MP-Relax: decompose it into two sub-problems and solve them iteratively in a cyclic order; by analyzing the KKT conditions of the two sub-problems, the scheme designs two low-complexity algorithms to solve them, thereby accelerating BCD 2 In the last step, an optimal rounding algorithm (ORA) is proposed to round the relaxed solution.

[0244] Breakdown of USP

[0245] The objective function of USP is recorded as f(x,p,h)=min i,j f ij (x ij ,p ij ,h). The so-called problem partitioning refers to projecting the objective function and constraint set of USP onto the (x,p) axis at the same time. After projecting USP onto the (x,p) axis, the following main problem can be obtained:

[0246]

[0247] st(x,p)∈V(14)

[0248] Where X and P are defined as: X:={x|x satisfies (4) and (7)}, and P:={p|p satisfies (5) and (11)}. The set V is defined as:

[0249] V: = {p| exists h∈H B So that p satisfies (6)} (15)

[0250] In the above formula, H B is the set of intervals defined by constraints (1), (2) and (10), that is, in and They are and

[0251] In MP, v(x,p) is defined as a subproblem:

[0252]

[0253] Explicit solution to subproblem

[0254] The explicit solution of SP can be derived directly, which is independent of x and p, and the following theorem about SP can be derived:

[0255] For each fixed pair If the corresponding SP is feasible, then its optimal solution must be

[0256] Explicit Formulation of the Master Problem and Its Relaxation

[0257] Bring h into h *, MP can be equivalently transformed into a joint channel and power allocation problem. In this case, MP is a non-smooth and non-convex mixed integer problem, which is extremely difficult to solve. The present invention does not directly process it, but first solves its relaxed version (i.e., MP-Relax) to obtain a relaxed optimal solution. Then, the relaxed solution is taken as an integer value by the optimal rounding algorithm. To simplify the notation, n is used to represent the index pair (i, j), Represents all user devices in the network, that is, The number of user devices in the system is denoted by N, and the relaxed MP can be expressed as the following problem:

[0258]

[0259] std T x≤B tot (19)

[0260] 1 T p≤P max (20)

[0261] x≥ x ,p≥ p (twenty one)

[0262] Where d and 1 are the coefficients of constraints (4) and (5), respectively. β n is the number of B in the corresponding index pair (i, j) i , For a clearer representation, the objective function of MP-Relax is represented by ω(x, p). It can be shown that the inequality constraints in constraints (19) and (20) can be replaced by equality constraints.

[0263] 3. Relax the main problem and solve it

[0264] The BCD method will be used to solve MP-Relax. Based on the pseudo-convex optimization theory and the characterization of the MP-Relax solution, the KKT condition is used to solve two sub-problems, namely the power allocation sub-problem and the sub-channel allocation sub-problem, and then the BCD algorithm for solving the global optimal solution of MP-Relax is proposed.

[0265] Power allocation subproblem

[0266] By fixing a stationary point x on the domain, we can obtain the following power allocation subproblem:

[0267]

[0268] st1 T p≤P max (twenty three)

[0269] P≥ P (twenty four)

[0270] Among them, a n =β n x n It can be seen that the power allocation subproblem is a non-smooth convex programming. Since its objective function is strictly convex, it has a unique solution. Assume that the optimal solution is (p * ,z * ),definition:

[0271]

[0272] In order to find the optimal solution, a KKT-based power allocation algorithm is proposed, referred to as KKT p , which is used to solve the power allocation subproblem.

[0273]

[0274] Subchannel allocation subproblem

[0275] At the x coordinate, the following subchannel allocation subproblem needs to be solved for a fixed p:

[0276]

[0277] std T x≤B max (28)

[0278] x≥ x (29)

[0279] where c n =β n ln(1+b n p n ). Due to c n ≠0, Sub x The objective function of is strictly convex, which indicates the uniqueness of its optimal solution. Similarly, assuming that the optimal value is t * , assuming the optimal solution is It can be deduced that:

[0280]

[0281] In order to find the optimal value, a subchannel allocation algorithm based on KKT is proposed, referred to as KKT x , which is used to solve the subchannel allocation subproblem.

[0282]

[0283] BCD Algorithm for Relaxed Master Problem

[0284] Based on KKT x and KKT p Algorithm,The following BCD algorithm is proposed to solve MP-Relax.

[0285]

[0286]

[0287] 4. Simulation results

[0288] Figure 5 Convergence time for different implementations.

[0289] Figure 6 Comparison of maximum and minimum rates of users in different network scenarios.

[0290] The present invention proposes a low-complexity optimization framework for the slicing problem in drone-assisted networks, thereby jointly optimizing resource allocation and drone deployment. The problem is equivalently divided into a resource allocation problem and a drone deployment problem. By analyzing the properties of the sub-problems, the explicit solutions of the sub-problems are derived, thereby deriving an explicit expression of the main problem. It is then proved that the relaxed main problem is a pseudo-concave problem, and its stationary point corresponds to the global optimal solution. Based on this, a lightweight decomposition algorithm using the KKT condition to find the global optimal solution is proposed. In addition, an optimal rounding algorithm is designed to minimize the rounding loss of the relaxed solution. Numerical results show that compared with existing algorithms, the algorithm proposed in the present invention has significant advantages such as fast convergence speed and high user throughput. Figure 6 Convergence curves of the algorithm of the present invention and the existing algorithms. USP is a large-scale non-convex mixed integer programming, and there is a lack of solutions in existing research at home and abroad. Considering that metaheuristics can escape the local optimal phenomenon in non-convex problems, the present invention selects the following three algorithms as the comparison benchmarks of the present invention: Quasi-subgradient Projection Algorithm (QPA); Adaptive Inertia Weighted Particle Swarm Optimization (AIW-PSO); Success History Intelligent Optimizer (SHIO). Describes BCD 2 Compared with the convergence curves of the three existing algorithms, the present invention sets the number of UEs in UAWN to 60. It can be seen that BCD 2 Only two iterations are needed to achieve convergence, which is equivalent to one round of BCD iterations. In contrast, QPA requires about 100 iterations to converge, while AIW-PSO and SHIO do not even converge. Therefore, the present invention concludes that the BCD2 of the present invention is lightweight and converges quickly.

[0291] Table 4 Comparison between the algorithm of the present invention and the QPA algorithm

[0292]

[0293]

[0294] To quantify the convergence speed of BCD, the table above compares BCD 2 and the execution time and optimal value of QPA, the number of UEs varies from 20 to 100. It can be seen that BCD 2 converges within a few milliseconds, which is about 0.02% of QPA. In addition, it can be observed that BCD 2 The optimal value of is slightly larger than QPA. This result shows that BCD 2 It is an exact algorithm in the sense that it can find the exact solution in a finite number of iterations, which is not possible with QPA. In addition, the convergence time increases linearly with the problem size, which verifies the log-linear convergence of BCD2.

[0295] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0296] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A lightweight slice resource configuration method for drone-assisted networks, characterized in that: The following steps are involved: Step 1: The problem is equivalently divided into a resource allocation problem and a drone deployment problem. A preliminary resource allocation plan is formed by quantifying the resources in the network and using signal processing algorithms to estimate the interference and signal strength of each link. Step 2: Analyze the properties of the resource allocation problem and the drone deployment problem, derive their respective explicit solutions, and optimize the solution of each sub-problem through signal enhancement, noise reduction techniques and digital signal processing algorithms, considering the coverage of wireless signals and space-time signal adaptability; Step 3: By relaxing the main problem into a pseudo-concave problem, the KKT condition is used to solve the global optimal solution, and the signal data is calculated through the optimization relationship to ensure the optimal resource allocation and drone deployment plan under the network configuration; Step 4: Through the lightweight decomposition algorithm and the optimal rounding algorithm, the signal data is refined to minimize the rounding loss of the relaxed solution, ensure the practical application of the global optimal solution, solve the errors caused by the discretization operation, and optimize the signal link resource allocation and interference control.

2. The lightweight slice resource configuration method for drone-assisted networks according to claim 1, characterized in that: The derivation method is as follows: Basal network model First, we focus on downlink transmission, where the UAV is equipped with a maximum transmission power of P max The bandwidth available to the drone is denoted as B tot ; Assume that the projection position of the drone on the horizontal plane is the center of the drone network; let θ be the beam width and h be the height of the drone; then the following coverage conditions must be met: h·tanθ≥R (1) R in the above formula is the radius of the UAWN; considering that the flight capability of the UAWN is limited, its altitude is limited by the following conditions: In the above formula, c is the flight speed of the drone. is the flight altitude of the UAV in the previous time slot; Network Slicing Model use represents the set of network slices, and the set of user devices in slice i is recorded as u ij represents the jth user equipment in slice i, user equipment u ij The two-dimensional coordinates of User equipment ij The square distance from the drone is in Assume g0 is the channel power gain at a reference distance of one meter; according to existing literature, user equipment u ij The channel power gain with the drone can be modeled as: Where, α is the path loss exponent; in represents the small-scale fading channel coefficient of the Rayleigh distribution, that is, Follows a non-central chi-square distribution; Use B i represents the channel bandwidth of slice i; it should be noted that due to service differentiation, the channel bandwidth of different slices is different; a non-negative integer x ij Indicates the allocation to user equipment u ij The number of channels; obviously, x ij The following capacity constraints should be met: Among them, the drone sends a signal to the user device u ij The transmitted power is denoted as p ij , the following capacity constraints must be met: To guarantee the SLA, each slice should provide a guaranteed data rate to the subscriber, namely: Among them, σ 2 is the power of the background noise, is the minimum rate guaranteed by slice i; in addition, each user device can request a customized service rate Considering the tractability of the problem, this constraint is modeled as: in, is a positive integer that satisfies UAWN network slicing model The UAWN network slicing problem can be expressed as: st(1)-(2),(4)-(7)(9) h min ≤h≤h max (10) 3. The lightweight slice resource configuration method for drone-assisted networks as claimed in claim 1, characterized in that: The problem can be divided into: Breakdown of USP The objective function of USP is recorded as f(x,p,h)=min i,j f ij (x ij ,p ij ,h); the so-called problem partitioning means projecting the objective function and constraint set of USP onto the (x,p) axis at the same time. After projecting USP onto the (x,p) axis, the following main problem can be obtained: st(x,p)∈V(14) Where X and P are defined as: X:={x|x satisfies (4) and (7)}, and P:={p|p satisfies (5) and (11)}; the set V is defined as: V: = {p| exists h∈H B So that p satisfies (6)} (15) In the above formula, H B is the set of intervals defined by constraints (1), (2) and (10), that is, in and They are and In MP, v(x,p) is defined as a subproblem: Explicit solution to subproblem The explicit solution of SP can be derived directly, which is independent of x and p, and the following theorem about SP can be derived: For each fixed pair If the corresponding SP is feasible, then its optimal solution must be Explicit Formulation of the Master Problem and Its Relaxation Let n represent the index pair (i,j), Represents all user devices in the network, that is, The number of user devices in the system is denoted by N, and the relaxed MP can be expressed as the following problem: s.t.d T x≤B tot (19) 1 T p≤P max (20) x≥ x ,p≥ p (21) Where d and 1 are the coefficients of constraints (4) and (5), respectively. β n is the number of B in the corresponding index pair (i, j) i , For a clearer representation, the objective function of MP-Relax is represented by ω(x, p); and it can be shown that the inequality constraints in constraints (19) and (20) can be replaced by equality constraints.

4. The lightweight slice resource configuration method for drone-assisted networks according to claim 1, characterized in that: The lightweight decomposition algorithm for finding the global optimal solution using the KKT condition: Relaxed master problem solution The BCD method will be used to solve MP-Relax. Based on the pseudo-convex optimization theory and the characterization of the MP-Relax solution, the KKT condition is used to solve two sub-problems, namely the power allocation sub-problem and the sub-channel allocation sub-problem, and then the BCD algorithm for solving the global optimal solution of MP-Relax is proposed. Power allocation subproblem By fixing a stationary point x on the domain, we can obtain the following power allocation subproblem: s.t.1 T p≤P max (23) P≥ P (24) Among them, a n =β n x n ; It can be seen that the power allocation subproblem is a non-smooth convex programming. Since its objective function is strictly convex, it has a unique solution; assuming that the optimal solution is (p * ,z * ),definition: Subchannel allocation subproblem; BCD algorithm for the relaxed master problem.

5. The lightweight slice resource configuration method for drone-assisted networks as claimed in claim 4, characterized in that: The sub-channel allocation sub-problem: At the x coordinate, the following subchannel allocation subproblem needs to be solved for a fixed p: s.t.d T x≤B max (28) x≥ x (29) where c n =β n ln(1+b n p n ); due to c n ≠0, Sub x The objective function of is strictly convex, which indicates the uniqueness of its optimal solution; similarly, assuming that the optimal value is t * , assuming the optimal solution is It can be deduced that:

6. The lightweight slice resource configuration method for drone-assisted networks as claimed in claim 4, characterized in that: The BCD algorithm for the relaxed master problem: Based on KKT x and KKT p Algorithm,A BCD algorithm based on loop iteration rules is proposed to solve MP-Relax.

7. A lightweight slice resource configuration system for drone-assisted networks implementing the lightweight slice resource configuration method for drone-assisted networks as described in any one of claims 1 to 6, characterized in that: The lightweight slice resource configuration system for drone-assisted networks includes: A partitioning module is used to divide the problem into resource allocation problem and drone deployment problem; The derivation module is used to derive explicit solutions to sub-problems by analyzing the properties of sub-problems and derive explicit expressions of the main problem; The solution module is used to prove that the relaxed main problem is a pseudo-concave problem, whose stable point corresponds to the global optimal solution. Based on this, the KKT condition is used to find a lightweight decomposition algorithm for the global optimal solution, and the optimal rounding algorithm is designed to minimize the rounding loss of the relaxed solution.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the lightweight slice resource configuration method for drone-assisted networks as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the lightweight slice resource configuration method for drone-assisted networks as described in any one of claims 1 to 6.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the lightweight slice resource configuration system for drone-assisted networks as described in claim 7.

Images