Unmanned aerial vehicle edge computing method and device based on non-orthogonal multiple access

Abstract

This application relates to a method and apparatus for UAV edge computing based on non-orthogonal multiple access. It utilizes an airborne mobile edge computing system based on Internet of Things (IoT) applications to construct UAV system models, communication models, and computational models. By jointly optimizing channel correlation coefficients, UAV trajectories, and computational resource allocation, the minimum computational throughput of all ground devices is maximized. Furthermore, this application employs an indeterminate block coordinate descent (inexact BCD) method to avoid getting trapped in strong local optima, and combines it with Successive Convex Approximation (SCA) technology to handle non-convex constraints. This method exhibits significant performance advantages in terms of computational throughput.

Description

Technical Field

[0001] This application relates to the field of wireless communication, and in particular to a method and apparatus for edge computing of unmanned aerial vehicles based on non-orthogonal multiple access. Background Technology

[0002] In recent years, latency-sensitive and computationally intensive applications have become increasingly prevalent, such as facial recognition, autonomous driving, video and image processing, and real-time online gaming. With the development of the Internet of Things (IoT), there is a need to connect various smart devices to the internet for data exchange and computation. However, IoT devices typically have limited resources, particularly in terms of computing power and energy, making it difficult for them to respond instantly.

[0003] Mobile edge computing (MEC) allows computing tasks to be performed locally by moving computing resources to the edge of the network, eliminating the need to transmit data to remote cloud servers. Compared to cloud computing, MEC brings computing resources closer to ground devices (GDs), significantly reducing latency in computing tasks. However, ground devices located at the edge of terrestrial networks or in remote areas often struggle to reliably access MEC servers and may even be difficult to cover by existing communication infrastructure.

[0004] Integrating unmanned aerial vehicles (UAVs) into MEC systems is an effective solution to these problems. UAVs equipped with high-performance servers are increasingly being used to provide compute offloading services for resource-constrained ground-based IoT devices. Due to their high maneuverability, UAVs can approach IoT devices, increasing the probability of line-of-sight (LoS) air-to-ground links. Therefore, the quality of air-to-ground communication is improved, which facilitates compute offloading for ground devices.

[0005] Non-orthogonal multiple access (NOMA) technology is gaining increasing attention in modern wireless communication systems. Specifically, NOMA enables multiple ground devices to share communication resources and utilizes Successive Interference Cancellation (SIC) technology to decode superimposed signals, thereby improving spectrum utilization and increasing throughput. Using NOMA, wireless networks can accommodate more users while maintaining high-quality service. However, the implementation of air-to-ground computing cooperation in UAV MEC networks via uplink NOMA is still in its early stages. Summary of the Invention

[0006] This application provides a UAV edge computing method and apparatus based on non-orthogonal multiple access, maximizing the minimum computational throughput of all ground devices. The technical solution of this application is as follows:

[0007] According to a first aspect of the embodiments of this application, a method for edge computing of unmanned aerial vehicles (UAVs) based on non-orthogonal multiple access is provided, the method comprising:

[0008] S1: Construct UAV system models, communication models, and computing models using an airborne mobile edge computing system based on IoT applications. ；

[0009] In this scenario, the drone provides computing services to K ground devices, and the set of ground devices is represented as... Ground equipment k The horizontal coordinate is represented as 1≤k≤K, the UAV maintains an altitude of H above the ground, and its horizontal position is determined by... t∈[0,T] means that all computational unloading tasks are completed within the time range T;

[0010] S2: Based on the joint optimization of computing resource allocation strategy, channel correlation coefficient and UAV trajectory, construct the minimum computing throughput model problem (P1);

[0011] S3: Penalize non-integer solutions of the problem by adding a penalty term to the objective function of the minimum computational throughput model problem (P1); and iteratively solve the obtained optimization problem model using a nondeterministic block coordinate descent method until convergence.

[0012] Optionally, the aerial mobile edge computing system based on Internet of Things applications constructs a drone system model, specifically including:

[0013] The time range T is uniformly discretized into N time intervals of length δ. t The time slot, δ t If the value is set small enough so that the position of the UAV remains constant in each time slot, then the flight trajectory of the UAV is represented as a sequence {q[n], 1≤n≤N}, where This indicates the horizontal position of the drone in the nth time slot;

[0014] Furthermore, the flight trajectory of the UAV satisfies the constraint ||q[n+1]-q[n]||≤V max δ t , Where V max This indicates the maximum speed of the drone.

[0015] Optionally, the aerial mobile edge computing system based on Internet of Things applications, used to construct the UAV system model, further includes:

[0016] Ground equipment k Channel gain modeling between the drone and the UAV is Where β0 represents the average power gain at a distance of d0 = 1 meter. Indicates device s k The distance to the drone in the nth time slot;

[0017] Assuming all ground equipment transmits the calculation offloading data at the same power P, that is Where p k s k Maximum transmission power.

[0018] Optionally, the communication model constructed by the airborne mobile edge computing system based on IoT applications specifically includes:

[0019] Define a binary variable x k，l [n]∈{0,1} represents the channel correlation coefficient of the nth time slot, where x k，l [n] = 1 indicates that the ground equipment s l The computational task offloading is considered to be for ground equipment closer to the drone. k The interference of unloading computational tasks satisfies d k，U [n]≤d l，U [n], otherwise x k，l [n] = 0;

[0020] Set x k，l [n]+x l，k [n] = 1, to avoid [n] in d k，U [n] = d l，U In the case of [n], simultaneously ground equipment s k and s l If this is considered interference with the user, then the ground equipment s k The achievable unloading rate of the drone is expressed as:

[0021]

[0022] Where B is the channel bandwidth, σ 2 This indicates the noise power of the receiver.

[0023] Optionally, the computing model for an airborne mobile edge computing system based on IoT applications includes:

[0024] Definition l local，k [n] represents ground equipment s k The number of task input bits required to perform local computation in time slot n is then: Among them, C k This represents the number of CPU cycles required to process 1 bit of input data. and These represent drones and ground equipment, respectively. k CPU peak frequency;

[0025] Define f U，k [n] allocates resources for computing ground devices s for the drone. k The CPU frequency for unloading tasks is as follows:

[0026] make in This indicates that during the cutoff time slot n, the data from ground equipment s... k The amount of input data, This represents the amount of data that the drone performs in the corresponding calculations, so that the aerial mobile edge computing system can meet the information causal constraints.

[0027] Define μ k For ground equipment k The effective capacitance coefficient, then s k The energy consumed in performing local computation in time slot n is expressed as follows:

[0028] Optionally, the computing model for building an airborne mobile edge computing system based on IoT applications also includes:

[0029] make So that ground equipment s k The total energy consumption for local computation and unloading of task information to the UAV within time T does not exceed its energy limit. Indicates ground equipment s k Energy limitations;

[0030] Using computational throughput as a metric to measure the system's computing power, this metric represents the total number of bits processed by the UAV edge server and ground equipment. Therefore, the ground equipment throughput is s. k The total computing throughput is expressed as

[0031] Optionally, based on the joint optimization of computational resource allocation strategy, channel correlation coefficient, and UAV trajectory, a minimum computational throughput model problem (P1) is constructed, specifically including:

[0032] By jointly optimizing the channel correlation coefficient {x k，l [n]}、CPU calculation frequency {f U，k [n]}、Local computation data size {l local，kGiven the drone trajectory {q[n]} and the drone trajectory {q[n]}, construct a minimum computational throughput model (P1).

[0033] Alternatively, the minimum computational throughput model problem (P1) can be represented as follows:

[0034]

[0035] The constraints include:

[0036]

[0037]

[0038]

[0039] x k，l [n]+x l，k [n] = 1, (4)

[0040]

[0041]

[0042]

[0043]

[0044] q[1]=q[N], (9)

[0045]

[0046] Optionally, a penalty term is added to the objective function of the minimum computational throughput model problem (P1) to penalize non-integer solutions; and a nondeterministic block coordinate descent method is used to iteratively solve the obtained optimization problem model until convergence, specifically including:

[0047] Constraint (2) is transformed into the intersection of the following regions:

[0048]

[0049]

[0050] Constraint (3) is equivalently transformed into the following form:

[0051]

[0052] Introducing a penalty term into the minimum computational throughput model problem (P1) to penalize cases where constraint (12) is not satisfied, we obtain the following first penalty problem model (P1.1):

[0053]

[0054] The constraints are given by equations (1), (4)-(11), and (13), and the set... function Where λ>>1 is a penalty parameter;

[0055] The variable set Z is optimized using a nondeterministic block coordinate descent method, and it is partitioned into Z1 = {x}. k，l [n], f U，k [n], l local，k [n]}, Z2={q[n],f U，k [n], l local，k [n]};

[0056] Wherein, variable {f U，k [n], l local，k [n]} is included in both Z1 and Z2 to avoid getting trapped in strong local optima;

[0057] By optimizing the channel correlation coefficient, CPU computation frequency, and local computation resource allocation given q[n], and given {x} k，l Under [n]}, optimize the drone trajectory, CPU computing frequency and local computing resource allocation, and alternately optimize Z1 and Z2 until the target value converges.

[0058] According to a second aspect of the embodiments of this application, a drone edge computing device based on non-orthogonal multiple access is provided, comprising:

[0059] processor;

[0060] Memory for storing executable instructions of the processor ；

[0061] The processor is configured to execute the instructions to implement the UAV edge computing method based on non-orthogonal multiple access as described in any of the methods mentioned in the first aspect above.

[0062] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:

[0063] This application relates to a method and apparatus for UAV edge computing based on non-orthogonal multiple access, aiming to maximize computational throughput. By jointly optimizing channel correlation coefficients, UAV trajectories, and computational resource allocation, the minimum computational throughput of all ground devices is maximized. Secondly, an indeterminate block coordinate descent (inexact BCD) method is used to avoid getting trapped in strong local optima, and a Successive Convex Approximation (SCA) technique is combined to handle non-convex constraints. This method exhibits significant performance advantages in computational throughput.

[0064] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0065] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application, and do not constitute an undue limitation of this application.

[0066] Figure 1 This is a schematic diagram illustrating an architecture for constructing communication between a drone and ground equipment, according to an exemplary embodiment.

[0067] Figure 2 This is a flowchart illustrating an edge computing method for unmanned aerial vehicles based on non-orthogonal multiple access, according to an exemplary embodiment.

[0068] Figure 3 This is a schematic diagram of a drone trajectory according to an exemplary embodiment;

[0069] Figure 4(a) illustrates the interference relationship between different ground devices under the NOMA scheme when T = 30s.

[0070] Figure 4(b) illustrates the channel correlation coefficients between ground equipment and other ground equipment;

[0071] Figure 5(a) illustrates an exemplary performance comparison of Algorithm 1 with three benchmark schemes in terms of maximum-minimum computational throughput;

[0072] Figure 5(b) exemplifies the maximum-minimum computational throughput versus energy limitations achieved at T = 40 s. A diagram illustrating the relationship between them. Detailed Implementation

[0073] To enable those skilled in the art to better understand the technical solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0074] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects, not to describe a specific order or sequence. Therefore, such data can be interchanged where appropriate, and the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0075] The application scenarios described in this application are for the purpose of more clearly illustrating the technical solutions of this application, and do not constitute a limitation on the technical solutions provided in this application. As those skilled in the art will know, with the emergence of new application scenarios, the technical solutions provided in this application are also applicable to similar technical problems.

[0076] This application relates to a method and apparatus for UAV edge computing based on non-orthogonal multiple access. It utilizes an airborne mobile edge computing system based on Internet of Things (IoT) applications to construct UAV system models, communication models, and computational models. By jointly optimizing channel correlation coefficients, UAV trajectories, and computational resource allocation, the minimum computational throughput of all ground devices is maximized. This application employs the indeterminate block coordinate descent (inexact BCD) method to avoid getting trapped in strong local optima and combines it with Successive Convex Approximation (SCA) technology to handle non-convex constraints. This method exhibits significant performance advantages in terms of computational throughput.

[0077] like Figure 2 As shown, this application provides a UAV edge computing method based on non-orthogonal multiple access, specifically including:

[0078] Step 1: Build the UAV system model, communication model, and computing model using an airborne mobile edge computing system based on Internet of Things applications.

[0079] In this scenario, the drone provides computing services to K ground devices, and the set of ground devices is represented as... Ground equipment k The horizontal coordinate is represented as 1≤k≤K. The UAV maintains an altitude of H above the ground, and its horizontal position is determined by... t∈[0,T] indicates that all computational unloading tasks are completed within the time range T.

[0080] In some embodiments of this application, a system model is constructed for a UAV mobile edge computing network. For example... Figure 1 As shown, this application provides an aerial mobile edge computing (MEC) system utilizing IoT-based applications. It is assumed that the set of ground devices represents... Ground equipment k The horizontal coordinate is represented as 1 ≤ k ≤ K. Let T be the total time range, which can be interpreted as the flight time of the UAV, meaning that the UAV needs to complete all computational unloading tasks within the time range T. The horizontal position of the UAV at time t can be represented as... t∈[0,T]. Define z(t) as the altitude of the UAV at time t, where z(t) ranges from Z... min To Z max Z min ≤z(t)≤Z max The final drone position can be represented as [q(t)]. T z(t)] T ,in[·] T This represents the transpose of the matrix. Assume q(0) = q(T) and z(0) = z(T) to ensure that the UAV can periodically provide services to ground equipment.

[0081] To simplify the problem, the time range T is uniformly discretized into N intervals of length δ. t The time slot, δ t This value can be set small enough that the drone's position can be considered constant within each time slot. Therefore, the drone's trajectory can be approximated as a sequence {q[n], 1≤n≤N}, where Let q[n] represent the horizontal position of the UAV in the nth time slot. To ensure that the UAV can provide periodic services to ground equipment, we assume that the UAV returns to its starting position after completing its mission within the time range T, i.e., q[1] = q[N]. In addition, due to the physical performance of the UAV itself, the trajectory of the UAV should satisfy the constraint ||q[n+1]-q[n]||≤V. max δ t , Where V max This indicates the maximum speed of the drone.

[0082] Air-to-ground communication channels are generally considered to be LoS channels, therefore, in each time slot n, ground equipment s can be... kChannel gain modeling between the drone and the UAV is Where β0 represents the average power gain at a distance of d0 = 1 meter. Indicates device s k The distance to the drone in the nth time slot. We use p k s k The transmission power, which corresponds to the device's s k The maximum transmission power. To ensure fairness among all ground devices, we assume all devices transmit the computational offload data at the same power P, i.e.

[0083] In some embodiments of this application, a communication model is also constructed for the UAV mobile edge computing network. Specifically, a NOMA mechanism is used to offload computing tasks from ground devices, enabling multiple ground devices to simultaneously transmit tasks to the UAV using the same bandwidth and time resources. The UAV uses Successive Interference Cancellation (SIC) technology to decode the received information in descending order of channel gain. That is, offloading signals from ground devices with weaker channel gain (i.e., devices farther from the UAV) are considered interference to signals offloading from ground devices with stronger channel gain (i.e., devices closer to the UAV).

[0084] Define a binary variable x k，l [n]∈{0,1} represents the channel correlation coefficient of the nth time slot, where x k，l [n] = 1 indicates that the ground equipment s l The computational task offloading is considered to be for ground equipment closer to the drone. k The interference of unloading computational tasks satisfies d k，U [n]≤d l，U [n]. Otherwise x k，l [n] = 0. To avoid [n] in d k，U [n] = d l，U In the case of [n], simultaneously ground equipment s k and s l Considered as interfering with the user, set x k，l [n]+x l，k [n] = 1. Based on the above model, ground equipment s k The achievable unloading rate of the drone can be expressed as: Where B is the channel bandwidth, σ 2 This indicates the noise power of the receiver.

[0085] In some embodiments of this application, a partial offloading computation strategy is employed to construct the computational model. That is, each ground device s... kSome tasks are performed locally, while the remainder is offloaded to the drone for in-flight calculations. local，k [n] is defined as ground equipment s k The number of task input bits required to perform local computation in time slot n. For ground equipment s k , will C k Defined as the number of CPU cycles required to process 1 bit of input data (i.e., the computational density of the task), its value is usually application-dependent. and These represent drones and ground equipment, respectively. k The CPU's peak frequency. Therefore, there is

[0086] Define f U，k [n] allocates resources for computing ground devices s for the drone. k The CPU frequency of the unloading task. Therefore, there is... In order for the UAV to perform computational tasks unloaded from ground equipment, the system must satisfy information causality constraints; that is, the UAV can only begin computation after the corresponding data has been unloaded from the ground equipment. Therefore, we have The left side of the inequality represents the time slot n from ground equipment s. k The right side represents the amount of input data, while the right side represents the amount of data the drone performs the corresponding calculations. Define μ. k For ground equipment k The effective capacitance coefficient, then s k The energy consumed in performing local computation in time slot n can be calculated as follows:

[0087] It is worth noting that ground equipment s k The total energy consumption for local computation and unloading of task information to the drone within time T must not exceed its energy limit, therefore... in Indicates ground equipment s k Energy limitations. Computational throughput is used as a metric to measure the system's computing power; this metric represents the total number of bits processed by the UAV edge server and ground equipment. Therefore, the energy consumption of the ground equipment can be calculated. k The total computational throughput is expressed as

[0088] Step 2: Based on the joint optimization of computing resource allocation strategy, channel correlation coefficient and UAV trajectory, construct the minimum computing throughput model problem (P1);

[0089] In some embodiments, to achieve fairness among all ground equipment, our goal is to jointly optimize the channel correlation coefficient {x}. k，l[n]}、CPU calculation frequency {f U，k [n]}、Local computation data size {l local，k Given [n]} and the drone trajectory {q[n]}, we need to maximize the minimum computational throughput of all ground devices while satisfying the energy constraints of the ground devices. The minimum computational throughput model problem (P1) can be represented as follows:

[0090]

[0091] The constraints include:

[0092]

[0093]

[0094]

[0095] x k，l [n]+x l，k [n] = 1, (4)

[0096]

[0097]

[0098]

[0099]

[0100] q[1]=q[N], (9)

[0101]

[0102] Because of the presence of non-convex constraints (7) and binary constraints (2) and (3) in (P1), it is difficult to obtain the optimal solution.

[0103] Step 3: Penalize non-integer solutions of the problem by adding a penalty term to the objective function of the minimum computational throughput model problem (P1); and iteratively solve the obtained optimization problem model using a nondeterministic block coordinate descent method.

[0104] First, constraint (2) is transformed into the intersection of the following regions:

[0105]

[0106]

[0107] It can be verified that only when x k，lOnly when [n]∈{0,1} can constraints (11) and (12) be satisfied simultaneously. Specifically, if constraint (11) is satisfied, then we have Thus obtain Combining this with constraint (12), we obtain Therefore, there is

[0108] On the other hand, constraint (3) can be equivalently transformed into the following form:

[0109]

[0110] By introducing a penalty term into the minimum computational throughput model problem (P1) to penalize the case where constraint (12) is not satisfied, the following first penalty problem model (P1.1) is obtained:

[0111]

[0112] The constraints are given by equations (1), (4)-(11), and (13), and the set... function λ >> 1 is a penalty parameter. Related techniques can prove that when the penalty factor λ is sufficiently large, model (P1) is equivalent to the first penalty problem model (P1.1).

[0113] However, the first penalty problem model (P1.1) is still a non-convex problem. To address this issue, an exemplary embodiment of this application employs a nondeterministic block coordinate descent method. Specifically,

[0114] First, the optimization variable Z is divided into two blocks, namely Z1 = {x} k，l [n], f U，k [n], l local，k [n]}, Z2={q[n],f U，k [n], l local，k [n]}. Unlike traditional block coordinate descent methods, the variable {f U，k [n], l local，k [n]} is included in both Z1 and Z2, thus avoiding getting trapped in strong local optima. Next, the variables of these two blocks are alternately optimized using the following two optimization subproblems until the objective value converges.

[0115] A. Given q[n], optimize the channel correlation coefficient, CPU computation frequency, and local computation resource allocation.

[0116] Given a fixed q[n], the optimization subproblem of Z1 can be simplified to a second penalty problem model (P2.1). Specifically,

[0117]

[0118] The constraints are given by equations (1), (4)-(8), (11), and (13).

[0119] Using the Successive Convex Approximation (SCA) technique, we can achieve this by... The objective function F is derived by first-order Qinle approximation. λ The lower bound expression for (ξ), that is:

[0120]

[0121]

[0122] Where f λ (ξ) is a linear function.

[0123] Similarly, the first-order Cunner approximation of the logarithmic function can be used at a given point. The nonconvex term R is derived here. k，U The lower bound expression for [n].

[0124] Specifically, definition and get:

[0125]

[0126] in

[0127] Therefore, the second penalty problem model (P2.1) can be approximated as the second sub-penalty problem model (P2.2). Specifically,

[0128]

[0129] The constraints are as follows:

[0130] And equations (1), (4)-(6), (8), (11), and (13).

[0131] It can be verified that the second sub-penalty problem model (P2.2) contains a linear objective function and convex constraints. Therefore, the second sub-penalty problem model (P2.2) is a convex optimization problem, which can be efficiently solved using standard convex optimization tools (such as CVX). B. Optimize UAV trajectory, CPU computation frequency, and local computing resource allocation.

[0132] Given {x k，l[n]}, the optimization subproblem of Z2 can be simplified to a third penalty problem model (P3.1), specifically,

[0133]

[0134] st(1), (5)-(10), (13).

[0135] For constraint (7) R k，U [n] can be represented as:

[0136]

[0137] in,

[0138]

[0139] By introducing slack variables, we can equivalently describe the third penalty problem model (P3.1) as the third sub-penalty problem model (P3.2). Specifically,

[0140]

[0141] The constraints are equations (1), (5), (6), (8)-(10), and

[0142]

[0143]

[0144]

[0145] Related techniques can verify that the equation of constraint (16) must hold in the optimal solution of the third sub-penalty problem model (P3.2). This is because it can be achieved by continuously increasing y. l The value of [n] and the objective value of the third sub-penalty problem model (P3.2) remain unchanged, while all other constraints are still satisfied.

[0146] To handle the non-convex terms in equation (15) Applying a first-order Cunningham expansion of the logarithmic function to a given local point qr[n] yields an approximate expression for the lower bound:

[0147]

[0148]

[0149] in,

[0150] Similarly, by using the local point q r[n] for ||q[n]-w l || 2 By approximating the term using a first-order Qinle expansion, we obtain the following inequality:

[0151] ||q[n]-w l || 2 ≥||q r [n]-w l || 2 +2(q r [n]-w l ) T ×(q[n]-q r [n]), (19)

[0152] in[·] T This represents the transpose of the matrix. For the right-hand side of constraint (17), the non-convex terms are handled in a similar manner, yielding...

[0153]

[0154] Therefore, the third sub-penalty problem model (P3.2) can be approximated as the third sub-penalty problem model (P3.3), specifically:

[0155]

[0156] The constraints are equations (1), (5), (6), (8)-(10), and

[0157]

[0158]

[0159]

[0160] Thus, the third sub-penalty problem model (P3.3) is a standard convex optimization problem, which can be solved efficiently using CVX.

[0161] C. Overall Algorithm, Convergence, and Complexity

[0162] Based on the exemplary discussion in this application, the detailed process of solving the minimum computational throughput model problem (P1) can be summarized in Algorithm 1. Next, a convergence analysis of Algorithm 1 is given. Given the channel correlation coefficient... and drone trajectory Let the objective value of the minimum computational throughput model problem (P1) be denoted as... Therefore, we have the following inequality:

[0163]

[0164] Inequality (a) is due to the monotonically convergent nature of the SCA technique when solving the second penalty problem model (P2.1), and inequality (b) is due to the monotonically convergent nature of the SCA technique when solving the third sub-penalty problem model (P3.2).

[0165] Therefore, the objective function value will continuously increase during the iteration process, and due to limited computing resources, the total number of bits required to compute the objective function value has an upper bound, thus ensuring convergence of Algorithm 1. Furthermore, in Algorithm 1, we transform the minimum computational throughput model problem (P1) into a series of convex optimization problems for solution, the complexity of which depends on the number of optimization variables. Therefore, the total complexity of Algorithm 1 can be expressed as O(L(K)). 2 N) 3.5 ), where L is the number of iterations and has The magnitude.

[0166]

[0167] This application relates to a method and apparatus for UAV edge computing based on non-orthogonal multiple access, aiming to maximize computational throughput. By jointly optimizing channel correlation coefficients, UAV trajectories, and computational resource allocation, the minimum computational throughput of all ground devices is maximized. Secondly, an indeterminate block coordinate descent (inexact BCD) method is used to avoid getting trapped in strong local optima, and a Successive Convex Approximation (SCA) technique is combined to handle non-convex constraints. This method exhibits significant performance advantages in computational throughput.

[0168] According to one specific embodiment of this application, such as Figure 3 As shown, an aerial MEC system supporting unmanned aerial vehicles (UAVs) is considered, where the number of ground devices K = 5. They are uniformly and randomly distributed within an area of ​​size 1.6 × 1.6 km. 2 Within the square area, such as Figure 3 As shown. The system's total computing power is evaluated using computing throughput. This assumes all ground equipment has the same energy limitations and peak CPU frequency, i.e. Unless otherwise specified, we set the relevant parameters to the following default values: B = 1MHz, H = 100m, σ 2 =-110dBm, β0=-60dB, P=0.1W, V max =50m / s, δt=1s, λ=10 6 C k =10 3 , μk =10 -30 ,

[0169] Figure 3 The trajectory optimization results at different time points T are presented. It can be observed that as T increases, the UAV flexibly adjusts its path and gradually approaches the ground equipment. This adjustment reduces the distance between the ground equipment and the UAV, improving the air-to-ground communication quality. This is because offloading computational tasks to the UAV is more energy-efficient than performing local computation at the ground equipment, thus prompting the ground equipment to offload computational tasks to the UAV.

[0170] On the other hand, Figure 4(a) shows the interference relationship between different ground devices under the NOMA scheme when T = 30s. It is worth noting that a penalty function is introduced to ensure that the optimized channel correlation coefficient is a binary solution. Therefore, according to the results obtained from Algorithm 1, the channel correlation coefficient can only take the values ​​0 or 1. Specifically, x... k，l [n] represents ground equipment s l Does time slot n correspond to ground equipment s? k Interference occurs. As can be observed from Figure 4(a), as the UAV flies, s1 and s4 interfere with s2, while s3 and s5 do not. This is because the ground equipment uses NOMA for data transmission, while the UAV uses SIC to decode the received signals in descending order of channel gain. According to Equation (3), the computational offload signals of ground equipment s1 and s4, which have lower channel gain, are considered to interfere with the computational offload signal of s2, which has higher channel gain. Furthermore, as can be observed from Figure 4(b), s5 always interferes with the other four ground equipment, namely s1, s2, s3, and s4. This is because s5 has the lowest channel gain during UAV flight, and therefore always interferes with the other four ground equipment with higher channel gain.

[0171] Figures 5(a) and (b) show the performance comparison of the proposed algorithm with several benchmark schemes in terms of maximum-minimum computational throughput. Specifically, computational throughput is defined as the total number of bits processed by the UAV edge server and ground equipment, and maximum-minimum computational throughput refers to maximizing the minimum computational throughput of all ground equipment. The benchmark schemes include: 1) a TDMA benchmark scheme, which uses TDMA to avoid interference between different ground equipment during offloading; 2) a UAV-only computation benchmark scheme, which requires all computational tasks to be offloaded to the UAV, and does not allow any local computation by ground equipment; 3) a local computation benchmark scheme, where all computational tasks are processed locally by ground equipment without being offloaded to the UAV. In Figure 5(a), we observe that the proposed scheme outperforms the three benchmark schemes. Furthermore, the computational throughput increases with time T. Compared with the TDMA scheme, the proposed scheme utilizes the advantages of NOMA transmission, allowing multiple users to share the same communication resources; compared with the UAV-only computation scheme, the proposed scheme reduces the computational burden on the UAV and improves the system's computational throughput; compared with the local computation scheme, the proposed scheme improves computational efficiency by offloading some computational tasks to the UAV through air-ground cooperation. Figure 5(b) illustrates the maximum-minimum computational throughput and energy limitations achieved at T=40s. The relationship between them. It can be observed that, with... With the increase in energy, the computational throughput shows an upward trend. This is because the increased energy constraints allow ground equipment to have more local computation and computational offloading capabilities, thereby improving the overall system's computational performance. The performance gap compared to the baseline scheme indicates that the proposed penalty-based and nondeterministic block coordinate descent method has greater advantages in trajectory design and resource allocation.

[0172] This application studies NOMA-based unmanned aerial vehicle (UAV) networks, aiming to provide edge computing services for multiple ground devices. By jointly optimizing UAV trajectories, channel correlation coefficients, and computational resource allocation, the minimum computational throughput of all ground devices is maximized. To address the aforementioned mixed-integer non-convex optimization problem, an efficient iterative algorithm is proposed. This algorithm utilizes a penalty-based method and a nondeterministic block coordinate descent method, and employs a continuous convex approximation technique to handle non-convex constraints.

[0173] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0174] In some possible implementations, the electronic device according to this application may include at least one processor and at least one memory. The memory stores program code that, when executed by the processor, causes the processor to perform the operational data management methods according to the various exemplary embodiments of this application described above. For example, the processor may perform steps such as those in the operational data management method.

[0175] Furthermore, the UAV edge computing device based on non-orthogonal multiple access according to this embodiment of the present application can perform the steps in the UAV edge computing method based on non-orthogonal multiple access mentioned in the above embodiments.

[0176] In exemplary embodiments, various aspects of the UAV edge computing method and apparatus based on non-orthogonal multiple access provided in this application can also be implemented as a program product, which includes program code. When the program product is run on a computer device, the program code is used to cause the computer device to perform the steps in the method for maximizing experience quality in a multi-antenna UAV video transmission system according to various exemplary embodiments of this application described above.

[0177] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0178] Furthermore, although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0179] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application 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.

[0180] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. 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 image scaling device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable image scaling device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0181] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable image scaling 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.

[0182] These computer program instructions can also be loaded onto a computer or other programmable image scaling device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device 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.

[0183] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0184] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A UAV edge computing method based on non-orthogonal multiple access, characterized in that, The method includes: S1: Construct UAV system models, communication models, and computing models using an airborne mobile edge computing system based on Internet of Things applications; Among them, drones are A set of ground devices provides computing services, assuming the set of ground devices represents... Ground equipment The horizontal coordinate is represented as The drone maintains an altitude of [missing information]. , It indicates that all computational unloading tasks are completed within a time range T; the trajectory of the UAV is represented as a sequence. , Indicates that the drone is in The horizontal position of each time slot; S2: Based on the joint optimization of computational resource allocation strategy, channel correlation coefficient, and UAV trajectory, construct the minimum computational throughput model problem (P1); specifically including: through joint optimization of channel correlation coefficient { CPU calculation frequency Locally computed data size { } and drone trajectory { }, Construct a minimum computational throughput model (P1); S3: Penalize the non-integer solutions of the problem by adding a penalty term to the objective function of the minimum computational throughput model problem (P1); and iteratively solve the resulting optimization problem using a nondeterministic block coordinate descent method until convergence. The minimum computational throughput model problem (P1) is represented as follows: The constraints include: Constraint (1) represents the ground equipment Total computing throughput Greater than the system's minimum computing throughput ; The number of CPU cycles required to process 1 bit of input data. N represents the time slot length, and N represents the time range. The length of the uniformly discretized object is The number of time slots; In constraints (2) and (3), Used to indicate the first The channel correlation coefficient for each time slot, where Indicates ground equipment The computational task offloading is considered to be for ground equipment closer to the drone. The interference of unloading computational tasks, satisfying ,otherwise ; Indicates ground equipment With drones in The distance in each time slot Indicates ground equipment With drones in Distance in each time slot; Constraint (4) is used to avoid in In the case of simultaneously ground equipment and This is considered interference with the user; In constraint (5), For ground equipment In the time slot The number of task input bits required to perform local computation. Indicates ground equipment CPU peak frequency; In constraint (6), Specifically, the drone at time slot n is allocated for computing ground equipment. CPU frequency for unloading tasks. This indicates the peak CPU frequency of the drone; In constraint (7), ] indicates ground equipment The achievable unloading rate of the drone, Indicates the deadline slot From ground equipment The amount of input data, This represents the amount of data that the drone performs in the corresponding calculations, so that the aerial mobile edge computing system can meet the information causal constraints. Constraint (8) is used to enable ground equipment In time The total energy consumption for local computation and offloading of task information to the drone shall not exceed its energy limit, of which Indicates ground equipment Energy limitation, Indicates ground equipment Maximum transmission power For ground equipment The effective capacitance coefficient; Constraint (9) indicates that the UAV is within a certain time range Return to the starting position after completing the task; Constraint (10) is the flight trajectory constraint for the UAV, where This indicates the maximum flight speed of the drone.

2. The method according to claim 1, characterized in that, The method of constructing a drone system model using an airborne mobile edge computing system based on Internet of Things applications also includes: The ground equipment The channel gain between the UAV and the UAV is modeled as follows ,in Indicates distance Average power gain at meters.

3. The method according to claim 2, characterized in that, By adding a penalty term to the objective function of the minimum computational throughput model problem (P1), the non-integer solutions of the problem are penalized; The obtained optimization problem model is then iteratively solved using a nondeterministic block coordinate descent method until convergence, specifically including: Constraint (2) is transformed into the intersection of the following regions: Constraint (3) is equivalently transformed into the following form: Introducing a penalty term into the minimum computational throughput model problem (P1) to penalize cases where constraint (12) is not satisfied, we obtain the following first penalty problem model. : The constraints are given by equations (1), (4) - (11), (13), and the set... , ,function ,in, It is a penalty parameter; Optimize the variable set using a nondeterministic block coordinate descent method. It is divided into: }， ； Among them, the variable { } is included simultaneously and In order to avoid getting trapped in strong local optima; By giving Under these conditions, optimize the channel correlation coefficient, CPU computation frequency, and local computing resource allocation, and within a given... Next, optimize the drone trajectory, CPU computing frequency, and local computing resource allocation, and optimize them alternately. and Until the target value converges.

4. A UAV edge computing device based on non-orthogonal multiple access, characterized in that, include: processor; A memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the UAV edge computing method based on non-orthogonal multiple access as described in any one of claims 1 to 3.

Images