Three-dimensional deployment optimization method based on unmanned aerial vehicle assisted MEC system

Abstract

The invention discloses a three-dimensional deployment optimization method based on an unmanned aerial vehicle assisted MEC system. The three-dimensional deployment optimization method comprises the following operation steps: step 1, initializing parameters; step 2, modeling the problem of maximizing the calculation efficiency of the terminal system as a nonlinear integer programming problem; 3, solving the 3D position of the unmanned aerial vehicle by using an unmanned aerial vehicle 3D position deployment algorithm based on IA and SEC; 4, solving the CPU frequency, the transmitting power and the terminal unloading decision of the mobile user equipment by using a Lagrange multiplier method; 5, solving the beam width of the unmanned aerial vehicle by using one-dimensional search; and step 6, maximizing the calculation rate of the system by using a three-stage alternating iterative algorithm (BSCEM). According to the invention, the computing and communication resources of the terminal can be reasonably distributed, dual optimization in the data transmission process is realized on the terminal equipment, the data transmission capability is improved, and the energy consumption is reduced, so that the overall efficiency and the energy utilization rate are balanced and improved.

Description

Technical Field

[0001] The present invention relates to the field of MEC systems, and in particular to a three-dimensional deployment optimization method based on an unmanned aerial vehicle (UAV)-assisted MEC system. Background Art

[0002] MEC, or Mobile Edge Computing (MEC), is a new distributed computing method based on mobile communication networks. It expands the functions of cloud computing by bringing it to the edge of the network. MEC can be regarded as a cloud server running on the edge of the mobile network and performing specific tasks. Mobile Edge Computing Mobile Edge Computing refers to the provision of IT services and cloud computing capabilities at the edge of the mobile network, within the wireless access network, and close to the user. Its basic idea is to migrate the cloud platform to the edge of the network, achieve deep integration of traditional telecommunications networks and Internet services, and use the information services, location services, identification services, and broadband management service capabilities provided by wireless networks to reduce end-to-end latency in service delivery, improve the flexibility of service deployment and user experience, and establish a new industrial chain and ecosystem.

[0003] Compared to cloud computing, mobile edge computing migrates computing tasks and data sources from the cloud or data center to the edge closest to the user, and uses wireless access networks to create a high-performance, low-latency and high-bandwidth usage environment to provide users with cloud content and computing capabilities nearby, allowing mobile users to enjoy a high-quality service experience. In mobile edge computing, devices (i.e. edge servers, usually built by infrastructure providers) are deployed at the edge of the network between the cellular network base station and the core network to provide external computing and storage services, so that mobile devices can send computing-intensive and latency-sensitive tasks to the nearest network edge device to complete, thereby reducing the transmission time of tasks in the network.

[0004] In a journal paper titled "Mobile Edge Computing (MEC)-enabled UAV Placement and Computation Efficiency Maximization in Disaster Scenario", a multi-stage offloading algorithm based on a learning algorithm and an interior point method was proposed. The basic principle of the multi-stage offloading algorithm based on a learning algorithm and an interior point method is as follows: The first stage of the algorithm is to use the K-Means algorithm to divide the space into M clusters according to the location of the user equipment and solve the optimal position of the drone. In the second stage, the connection between each UE and the drone is determined based on the drone position, the maximum number of connections allowed, and the distance matrix to ensure that the maximum number of connections is achieved as much as possible and the optimal offloading decision is obtained without exceeding the maximum number of connections for each drone. In the third stage, the optimal solution for the CPU frequency, transmit power, and beam width of the user equipment is solved by the interior point method.

[0005] However, in the above technologies, during the deployment of drone locations, the K-Means algorithm usually relies on the distance between data points for clustering, and cannot directly take into account the weight or priority of the user's service quality requirements. For example, for users with high service quality requirements, they may want drones to be deployed near them to ensure that the user's service quality requirements are met, but the K-Means algorithm cannot directly reflect this priority, which directly leads to the obtained drone deployment location being a suboptimal solution rather than an optimal solution.

[0006] That is, the disadvantage of the above technology is that, in the process of 3D position deployment of drones, considering the service quality requirements of different mobile user devices, the K-Means algorithm usually needs to deploy more drones to cover users with higher service quality requirements to ensure the service quality of more users. Although it can meet the service quality requirements of different mobile user devices, it also means increasing the deployment and operation and maintenance costs of the system. Summary of the invention

[0007] The technical problem to be solved by the present invention is to overcome the defect in the prior art that the service quality requirements of terminal devices cannot be met due to limited computing resources, and to provide a three-dimensional deployment optimization method based on a drone-assisted MEC system.

[0008] The present invention solves the above technical problems through the following technical solutions:

[0009] The present invention provides a three-dimensional deployment optimization method based on a drone-assisted MEC system, and the three-dimensional deployment optimization method includes the following operating steps:

[0010] Step 1: Initialization of parameters;

[0011] Step 2: Model the problem of maximizing the computational efficiency of the terminal system as a nonlinear integer programming problem;

[0012] Step 3: Use the UAV 3D position deployment algorithm based on IA and SEC to solve the UAV 3D position;

[0013] Step 4: Use Lagrange multiplier method to solve the CPU frequency, transmit power and terminal offloading decision of the mobile user equipment;

[0014] Step 5: Use one-dimensional search to solve the beam width of the drone;

[0015] Step 6: Use the three-stage alternating iterative algorithm (BSCEM) to maximize the calculation rate of the system.

[0016] In this technical solution, while ensuring the service quality of all mobile user devices, drones can be deployed in the least number and can be accurately deployed to the target area on demand to provide periodic services. The optimal mobile user device CPU frequency, transmission power and terminal offloading decision variables, the beam width of the drone and the terminal system computing efficiency can be maximized. The terminal's own computing and communication resources can be reasonably allocated, and dual optimization in the data transmission process can be achieved on the terminal device, which not only improves data transmission capabilities but also reduces energy consumption, thereby achieving a balance and improvement in overall efficiency and energy utilization.

[0017] Preferably, the initialization of the parameters in step 1 is to initialize the user distribution area, and the user distribution area is established as a two-dimensional circular plane, multiple users are randomly distributed within the circular range, and the base station is located in the edge area.

[0018] Preferably, the specific operation steps of step 2 are:

[0019] S1. Establish a wireless channel model;

[0020] S2, establish a local computing model;

[0021] S3. Establish communication model;

[0022] S4. Optimize model.

[0023] Preferably, in the wireless communication system in step S1, the channel model is crucial for simulating wireless communication in the real world.

[0024] Preferably, in the step S2, the mobile user equipment adopts a partial offloading method to assign the task to the drone for calculation.

[0025] Preferably, in step S3, the channel gain between the mobile user equipment and the drone is regarded as a free space wireless channel model.

[0026] Preferably, in the step S4, the computing efficiency of the terminal is maximized by optimizing the offloading association decision, transmission power, computing resource allocation and 3D position of the mobile user equipment while ensuring the user's service quality.

[0027] Preferably, the specific operation steps of step three are:

[0028] A1. Maximize the drone service radius;

[0029] A2. Use the drone 3D position deployment algorithm to optimize the drone 3D position.

[0030] Preferably, in step A1, a signal-to-noise ratio threshold is used as a standard for measuring whether the service quality can meet normal communication requirements.

[0031] Preferably, the specific operation steps of step six are to use a three-stage alternating iterative algorithm (BSCEM) to alternately iterate steps three, four, and five to solve the optimal mobile user device CPU frequency, transmit power and terminal offloading decision variables, the beam width of the drone, and maximize the terminal system computing efficiency.

[0032] On the basis of being in accordance with the common sense in the art, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present invention.

[0033] The positive and progressive effects of the present invention are:

[0034] The present invention can deploy drones with the least number while ensuring the service quality of all mobile user devices, and can accurately deploy them to the target area on demand to provide periodic services, and can obtain the optimal mobile user device CPU frequency, transmission power and terminal unloading decision variables, the beam width of the drone and the maximum terminal system computing efficiency, and can reasonably allocate the terminal's own computing and communication resources, and realize dual optimization in the data transmission process on the terminal device, which not only improves the data transmission capacity, but also reduces energy consumption, thereby achieving a balance and improvement in overall efficiency and energy utilization;

[0035] In order to determine the optimal position of the drone and ensure the service quality of the terminal user, on the one hand, the maximum service radius of the drone that meets the user's service quality requirements is derived based on the relationship between the minimum signal-to-noise ratio required for the user device to receive the signal and its service quality, which can ensure the different service quality requirements of different mobile user devices. On the other hand, according to the given maximum service radius, the optimal deployment position of the drone is obtained based on the minimum circumscribed circle and incremental algorithm, which can ensure the different service quality requirements of different mobile user devices and meet the service capacity constraints of the drone. That is, the present invention can ensure the different service quality requirements of different mobile user devices and meet the service capacity constraints of the drone while deploying a minimum number of drones. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The figure is a schematic diagram of the overall process of the three-dimensional deployment optimization method based on the drone-assisted MEC system according to an embodiment of the present invention.

[0037] Figure 2 for Figure 1 The specific operation flow diagram of step two of the three-dimensional deployment optimization method based on the drone-assisted MEC system is shown.

[0038] Figure 3 for Figure 1 The specific operation flow diagram of step three of the three-dimensional deployment optimization method based on the drone-assisted MEC system is shown.

[0039] Figure 4 for Figure 1 The figure shows a schematic diagram of the parameter initialization system model of the three-dimensional deployment optimization method based on the UAV-assisted MEC system.

[0040] Figure 5 for Figure 1 The C-based three-dimensional deployment optimization method for UAV-assisted MEC system is shown in i Schematic diagram of the coverage area.

[0041] Figure 6 for Figure 1 The diagram shown is a schematic diagram of the 22 UAV position deployment algorithm described in the present invention based on the three-dimensional deployment optimization method of the UAV-assisted MEC system.

[0042] Figure 7 for Figure 1 The figure shows a schematic diagram of a 40 UAV position deployment algorithm based on the K-Means algorithm of the three-dimensional deployment optimization method of the UAV-assisted MEC system.

[0043] Figure 8 for Figure 1The diagram shows the minimum number of drones required to be deployed to cover mobile user devices with different QoS requirements under the same algorithm of the three-dimensional deployment optimization method based on the drone-assisted MEC system as the number of mobile user devices increases.

[0044] Fig. 9 for Figure 1 Schematic diagram of the convergence of different algorithms for the three-dimensional deployment optimization method based on the UAV-assisted MEC system.

[0045] Fig.10 for Figure 1 The figure shows a schematic diagram of the influence of the CPU frequency of a mobile user device on the computing efficiency of the system based on the three-dimensional deployment optimization method of a drone-assisted MEC system.

[0046] Fig.11 Schematic diagram of the detailed steps of the UAV 3D position deployment algorithm based on IA and SEC.

[0047] Fig.12 This is a detailed diagram of step six. DETAILED DESCRIPTION

[0048] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0049] Figures 1 to 3 The structure diagram of an embodiment of the present invention is a three-dimensional deployment optimization method based on a drone-assisted MEC system. The three-dimensional deployment optimization method based on a drone-assisted MEC system includes the following steps:

[0050] Step 1: Initialization of parameters;

[0051] Step 2: Model the problem of maximizing the computational efficiency of the terminal system as a nonlinear integer programming problem;

[0052] Step 3: Use the UAV 3D position deployment algorithm based on IA and SEC to solve the UAV 3D position;

[0053] Step 4: Use Lagrange multiplier method to solve the CPU frequency, transmit power and terminal offloading decision of the mobile user equipment;

[0054] Step 5: Use one-dimensional search to solve the beam width of the drone;

[0055] Step 6: Use the three-stage alternating iterative algorithm (BSCEM) to maximize the calculation rate of the system.

[0056] In this technical solution, while ensuring the service quality of all mobile user devices, drones can be deployed in the least number and can be accurately deployed to the target area on demand to provide periodic services. The optimal mobile user device CPU frequency, transmission power and terminal offloading decision variables, the beam width of the drone and the terminal system computing efficiency can be maximized. The terminal's own computing and communication resources can be reasonably allocated, and dual optimization in the data transmission process can be achieved on the terminal device, which not only improves data transmission capabilities but also reduces energy consumption, thereby achieving a balance and improvement in overall efficiency and energy utilization.

[0057] The parameter initialization in step 1 is to initialize the user distribution area, and establish the user distribution area as a two-dimensional circular plane. Multiple users are randomly distributed within the circular range, and the base station is located in the edge area.

[0058] In this system, the mobile user equipment is located in a remote area and cannot establish an effective link with the remote base station. The drone is equipped with a small base station and a server to facilitate the drone to establish an effective link with the terminal and to handle the tasks unloaded by the mobile user equipment in a timely manner. Figure 4 shown.

[0059] The set of M UAVs and N mobile user devices is represented as M = {1, 2, ..., M}, N = {1, 2, ..., N}. Assume that the mobile user device is fixed on the ground and the UAV flies at a certain initial altitude. Then, a three-dimensional Cartesian coordinate system is constructed so that the positions of the mobile user device i and the UAV j are and In addition, each mobile user equipment i has a QoS requirement expressed in terms of signal-to-noise ratio (SINR), denoted by q i Indicates. Where Q = {q1,q2…,q n} represents the set of QoS requirements of mobile user equipment.

[0060] The specific operation steps of step 2 are:

[0061] S1. Establish a wireless channel model;

[0062] S2, establish a local computing model;

[0063] S3. Establish communication model;

[0064] S4. Optimize model.

[0065] In the wireless communication system in step S1, the channel model is crucial to simulating real-world wireless communications.

[0066] A2G (Air-to-Ground) links have a certain probability of line-of-sight (LoS) and non-line-of-sight (NLoS) transmission environments in practical applications. UAVs are deployed as BSs not only because of their high maneuverability, but also because they have a higher probability of producing line-of-sight and the resulting better channel gain. Under predefined environmental factors, the line-of-sight (LoS) channel probability between mobile user equipment i and UAVj is modeled as:

[0067]

[0068] Among them, α and β are constants that depend on the environment, θ i,j is the elevation angle between mobile user equipment i and UAVj, and The probability of non-line-of-sight (NLoS) channel is P NLoS =1-P LoS Since the A2G channel contains a free space segment and a ground segment, free space path loss and shadow fading exist simultaneously. Then, the loss (in dB) of the LoS and NLoS links can be derived separately:

[0069]

[0070] Among them, f c is the carrier frequency, d i,j is the distance between mobile user equipment i and UAVj, given by Given, c is the speed of light. In addition, η LoS and η NLoS are the average additional path losses in LoS and NLoS environments, respectively. Based on the above analysis, the average path loss (PL) can be written as:

[0071] PL=P LoS ×L LoS +P NLoS ×L NLoS . (5)

[0072] Substituting (2), (3) and (4) into (5), PL can be expressed as:

[0073]

[0074] Where A = η LoS -η NLoS and

[0075] In the step S2, the mobile user equipment uses a partial offloading method to assign tasks to the drone for calculation.

[0076] Assume that the system running time is T. The mobile user device adopts partial offloading mode, and the task can be assigned to the drone for calculation. It should be noted that the computing power of the drone far exceeds that of the mobile user device, and the task can be completed and the result can be returned within the system time T. In addition, the existing solar energy technology can enable the drone to fly continuously for more than 28 hours, and the drone can also be powered by a tether to achieve unlimited flight time. Therefore, the present invention assumes that the drone has sufficient energy to support its operation within the system time T. Each terminal device adopts dynamic frequency scaling (DFS) technology, which can reduce energy consumption and increase computing efficiency by adjusting the CPU frequency of local calculation. For the i-th mobile user device, within the system time T, the number of bits calculated locally is:

[0077]

[0078] where f i (cycles / second) represents the CPU frequency of the i-th mobile user equipment, W i represents the number of cycles required for the i-th mobile user equipment to calculate each bit of data. At the same time, the energy consumed by local calculation is as follows:

[0079]

[0080] Among them, k i Indicates the effective handover capability of a mobile user device. Due to the limited computing power of the terminal device, the CPU frequency in local computing is subject to the following constraints:

[0081]

[0082] In step S3, the channel gain between the mobile user equipment and the drone is regarded as a free space wireless channel model.

[0083] First, based on the system model, the horizontal distance between the i-th mobile user device and the j-th UAV is:

[0084]

[0085] Second, assume that each UAV is equipped with a directional antenna with adjustable beamwidth. For any j-th UAV, the antenna has equal half-power beamwidth in azimuth and elevation, both expressed as 2θ j ∈(0,π). In addition, with azimuth angle θ and elevation angle The antenna gain in a direction can be expressed as follows:

[0086]

[0087] Where Ψ represents the channel gain beyond the antenna beam width, and V0 = 2.2846. For simplicity, Ψ is set to 0. In addition, considering that the study scenario is located in a remote area with almost no large-scale building occlusion, it is reasonable to assume that a LoS communication channel is established between the UAV and the mobile user device. The present invention regards the channel gain between the mobile user device and the UAV as a free space wireless channel model, which can be expressed as follows:

[0088]

[0089] where μ is set to 2 according to

[38] , represents the path loss exponent, h0 is the channel power gain at the reference distance of 1 m, and d i,j represents the 3D distance between mobile user equipment i and UAV j.

[0090] If mobile user equipment i considers offloading the task to the jth UAV, mobile user equipment i must be within the service range of UAV j, and the calculation formula is:

[0091]

[0092] In the present invention, it is assumed that the link interference between UAVs caused by overlapping coverage areas can be resolved by allocating orthogonal frequency resources on the UAVs. In addition, it is assumed that the mobile user equipment adopts orthogonal multiple access as the uplink transmission mechanism, so that the interference between the mobile user equipment can be ignored when the mobile user equipment communicates with the same UAV. Therefore, for UAVj, the bandwidth is divided into subchannels, and each subchannel is occupied by a mobile user device. The transmission rate of the offload task from mobile user device i to UAV j can be modeled as follows:

[0093]

[0094] in, B represents the network bandwidth allocated to each UAV by the system, σ 2 represents Gaussian white noise. Then the number of bits of data unloaded and the energy consumed by any mobile user equipment i in system time T are expressed as follows:

[0095]

[0096] Since the transmission capability of mobile user equipment is limited, its transmission power is subject to the following restrictions:

[0097]

[0098] In addition, the present invention also takes into account that the number of terminal devices served by each drone is limited:

[0099]

[0100] in, is the maximum number of mobile user devices that UAV j can serve.

[0101] Therefore, the total number of bits calculated by the system consists of the number of bits calculated locally and the number of bits sent, and the energy consumed by the system consists of the energy calculated locally and the energy sent, which are expressed as follows:

[0102]

[0103] In addition, the focus of the present invention is to maximize the computing efficiency of the terminal, not the entire system. Therefore, it is assumed that the battery budget of the UAV is sufficient within the system time T. Based on this, the present invention reasonably ignores all energy consumption and data generated by the UAV.

[0104] At this point, the computational efficiency of the mobile user equipment system can be expressed as follows:

[0105]

[0106] In the step S4, under the premise of ensuring the user's service quality, the computing efficiency of the terminal is maximized by optimizing the offloading association decision, transmission power, computing resource allocation and 3D position of the mobile user equipment.

[0107] Based on the above system model analysis, we give the optimization objective function as shown in formula P1, that is, under the premise of ensuring the user's service quality, the terminal's computing efficiency is maximized by optimizing the offloading association decision, transmission power, computing resource allocation and 3D position of the mobile user device;

[0108] To simplify the mathematical expression, S is defined as the set of all optimization variables, namely θ={θ j} j∈M .

[0109]

[0110] stC1:α i,j ∈{0,1},i∈N,j∈M

[0111]

[0112] The specific operation steps of step three are:

[0113] A1. Maximize the drone service radius;

[0114] In the case of noise limitation, the traditional method of defining coverage is to use the signal-to-noise ratio (SINR). Therefore, the SINR threshold is used as a measure of whether the quality of service (QoS) can meet the normal communication requirements. The signal-to-noise ratio calculated by UAV j at mobile user equipment i is expressed as:

[0115] SINR i,j =P t -P n -PL. (28)

[0116] Among them, P t and P n They are transmission power and noise average power respectively, in dBm.

[0117] In order to meet the QoS requirements, the signal-to-noise ratio of UAV j at mobile user equipment i must meet SINR i,j ≥q i Therefore, the distance between UAV j and mobile user device i should satisfy:

[0118]

[0119] For any QoS requirement q expressed in terms of signal-to-noise ratio i , the elevation angle has an optimal constant value that maximizes the coverage area of ​​the drone and depends only on the environment. Therefore, once q is given i , all variables on the right side of (29) are considered constants, and the maximum allowed distance between any UAV j and mobile user equipment i is calculated as follows:

[0120]

[0121] For mobile user equipment i, the maximum QoS circle that meets the service quality requirements is defined as C i , where the maximum coverage radius is That is, as long as the mobile user equipment i is within a radius of The area of ​​the circle C i Within (including the border), normal communication can be achieved. and the drone altitude associated with it The relationship is as Figure 5 (Light-colored nodes indicate that their QoS requirements are met, while dark-colored nodes indicate that they are not met), the calculation formula is:

[0122]

[0123] A2. Use the drone 3D position deployment algorithm to optimize the drone 3D position.

[0124] Based on the above formula and derivation, in order to cover more mobile user devices, the present invention proposes a UAV 3D position deployment algorithm based on IA and SEC. The detailed steps of the algorithm are as follows: Fig.11 shown.

[0125] Specifically, in order to deploy drones with the least number of drones to cover more mobile user devices, this paper uses the IA algorithm and the SEC algorithm to optimize C i The IA algorithm will return the radius and coordinates of the minimum circumscribed circle that can cover a given set of mobile user devices, namely, C i The coordinates of .C i Mobile user equipment inside will become part of the UAV coverage set. In addition, Figure 2 As shown in the figure, since different mobile user devices have different QoS requirements, if the service capacity of the drone is not full, some mobile user devices will not be able to access the C i Internally, they can still meet their quality of service requirements through the same drone and be included in the drone's coverage area.

[0126] In step A1, the signal-to-noise ratio threshold is used as a standard for measuring whether the service quality can meet normal communication requirements.

[0127] Step 4: Use Lagrange multiplier method and KKT condition to solve the CPU frequency, transmit power and terminal offloading decision of the mobile user equipment;

[0128]

[0129] stC1~C7.(32)

[0130] When given the uninstallation-related variables of the mobile user device, the objective function of problem P3 and α i,j R i,j T relative to f i and p i,j is linear. In addition, -ηk i f i 3 T is the variable f i is a concave function. For the variable p i,j is concave. Since the objective function is to find the maximum value, the objective formula is convex. In addition, constraints C3 and C5 are relative to the variable f i and p i,j is linear, C1, C2, C4 and C6 have no variable f i and p i,j The constraint of C7 is relative to f i and pi,j is convex, so problem P2 is about f i and p i,j is convex. So far, for problem P3, the solution can be obtained by Lagrange multiplier method and conditions. The Lagrange function is defined as follows:

[0131] Question P3 relative to f i and p i,j is convex. So far, for problem P3, a solution can be found by the Lagrange multiplier method and the KKT condition. The Lagrange function is defined as follows:

[0132]

[0133] To simplify the problem, constraint C1 is omitted, but it will be strictly followed in the solution process. Where λ = [λ1,λ2,λ3,λ4,λ5,λ6] is the Lagrange multiplier. This paper considers using the subgradient method

[31] to update the Lagrange multiplier.

[0134] According to the KKT condition, the following expression can be obtained to obtain the optimal terminal CPU frequency:

[0135]

[0136] Solving the above equation, we can get the optimal mobile user device CPU frequency f i * The closed expression of is:

[0137]

[0138] Similarly, the following conditions are required to obtain the optimal terminal transmission power: The necessary and sufficient conditions.

[0139]

[0140] Solving equation (36) yields:

[0141]

[0142] At present, the closed-form expressions for the optimal solutions of CPU frequency and transmit power of mobile user equipment have been obtained.

[0143] After finding the optimal values ​​of terminal CPU frequency allocation and transmit power allocation, when the beam width of the drone is given again, we further consider solving the binary terminal offloading decision variables. First, relax the variables to [0,1]. Obviously, for the continuous offloading associated variable α i,j, the objective formula and constraints are convex. Then, a greedy algorithm is considered, so that the mobile user device i will give priority to the drone j with the best association effect to maximize the computational efficiency of the system. Therefore, the unloading association variable α is extracted from the objective formula (33): i,j :

[0144]

[0145] make Can get

[0146]

[0147] This formula indicates that for mobile user device i, while satisfying the carrying capacity of drone j, drone j that maximizes the system computational efficiency is selected in turn.

[0148] Step 5: Use one-dimensional search to solve the beam width of the drone;

[0149] After finding the terminal CPU frequency, transmit power, and unloading associated variables, the original problem P1 can be further transformed into the following pattern for any drone j:

[0150]

[0151] Finally, in order to obtain the optimal beam width θ j , this paper considers using a one-dimensional search method to solve it.

[0152] The specific operation steps of step 6 are to use the three-stage alternating iteration algorithm (BSCEM) to alternately iterate steps 3, 4, and 5 to solve the optimal mobile user equipment CPU frequency, transmission power and terminal offloading decision variables, the beam width of the drone, and the maximum terminal system computing efficiency. The detailed steps of the algorithm are as follows Fig.12 shown.

[0153] In the present invention, the 3D position deployment algorithm of the drone described in step three, in order to determine the optimal position of the drone and ensure the service quality of the terminal users, firstly derives the maximum service radius of the drone that meets the user's service quality requirements based on the relationship between the minimum signal-to-noise ratio required for the user device to receive the signal and its service quality, which can ensure the different service quality requirements of different mobile user devices. On the other hand, according to the given maximum service radius, the optimal deployment position of the drone is obtained based on the minimum circumscribed circle and incremental algorithm, which can ensure the different service quality requirements of different mobile user devices and meet the service capacity constraints of the drone. Therefore, compared with the K-Means algorithm used in the 'multi-stage unloading algorithm based on the learning algorithm and the interior point method' mentioned in the "most similar implementation scheme", the present invention can ensure the different service quality requirements of different mobile user devices and meet the service capacity constraints of the drone while deploying a minimum number of drones.

[0154] In order to verify the effectiveness of the drone 3D location deployment algorithm described in the present invention, firstly, 80 mobile user equipment users with different service quality requirements are randomly generated in a 3km×3km square area, and then the drone 3D location deployment algorithm and K-Means algorithm described in the present invention are used to cover these 80 mobile user equipment users respectively. The coverage results are shown in the figure. Figure 6 and Figure 7 As shown ( Figure 6 and Figure 7 In the figure, the square is 40dBm, the pentagon is 35dBm, the circle is 30dBm, and the triangle is 25dBm). In addition, in order to further verify the optimality of the drone 3D position deployment algorithm described in the present invention in minimizing the number of deployed drones, we added simulations of the minimum number of drones required to be deployed under different algorithms as the number of mobile user devices increases and to cover mobile user devices with different QoS requirements. The results are shown in the figure. Figure 8 shown.

[0155] The three-stage alternating iteration algorithm (BSCEM) described in step 6 alternately iterates steps 3, 4, and 5, which can maximize the computing efficiency of the terminal system and accelerate the convergence speed of the algorithm.

[0156] Through step three, the drone can be accurately deployed to the target area as needed. Through steps four and five, the terminal's own computing and communication resources can be reasonably allocated through one-dimensional search optimization, thereby increasing the amount of transmitted data and reducing its transmission energy consumption. Since the three-stage alternating iteration algorithm (BSCEM) proposed in the present invention can reasonably allocate the computing and communication resources of each terminal device after a finite number of iterations, compared with the "multi-stage unloading algorithm based on learning algorithm and interior point method mentioned in the most similar implementation scheme", the present invention can maximize the computing efficiency of the terminal system and accelerate the convergence speed of the algorithm. The convergence comparison diagram of the algorithm is shown in the figure below. Fig. 9 In addition, the comparison of the impact of the CPU frequency of mobile user equipment on system computing efficiency is shown in Fig.10 shown.

[0157] In order to verify the performance of the three-stage alternating iterative algorithm (BSCEM) proposed in the present invention, the proposed algorithm is compared with other benchmark schemes, including a fixed UAV 2D position scheme (Fix-XY), a fixed UAV altitude and beam width scheme (Fix-XY), a computational bit maximization scheme (CEBM), and a random scheme (AR, which means that the values ​​of offloading association decision, UAV 3D position, UAV beam width, mobile user device CPU frequency and transmit power are all random).

[0158] The convergence of the solution proposed by the present invention is as follows Fig. 9 As shown, from Fig. 9 It can be seen that the Fix-XY algorithm converges after about 7 iterations. The Fix-HB algorithm converges after about 10 iterations. The BSCEM algorithm proposed in this paper requires about 15 iterations to converge. This is because the BSCEM algorithm needs to optimize the three-dimensional position and beam width at the same time, which increases its complexity. However, it can be seen from the figure that the algorithm can still maintain good convergence performance when optimizing the three-dimensional position and beam width of the drone. The convergence of the AR algorithm is poor, and it takes about 30 iterations to converge. This is because the optimization variables of the algorithm are randomly generated. In this case, the speed of finding the optimal computational efficiency is slow, so the system computational efficiency of the algorithm is significantly lower than that of other optimization algorithms. In addition, the computational efficiency of the BSCEM algorithm is much higher than that of other algorithms, which proves that the algorithm in this paper plays an important role in improving the system computational efficiency by optimizing the drone position and beam width. The computational efficiency of the Fix-HB algorithm is slightly higher than that of the Fix-XY algorithm because it can find a better position to provide services for mobile user devices. The scheme proposed in the present invention achieves a balance between maximizing the computational bits and minimizing the energy consumption, which makes its computational efficiency the highest compared to other benchmark schemes.

[0159] Fig.10The influence of the CPU frequency of the mobile user device on the system computing efficiency is described. It can be seen that as the CPU frequency of the mobile user device increases, the system computing efficiency obtained by each algorithm continues to decrease. This is mainly because when the number of drones and network computing resources are limited, the mobile user device will speed up the processing by increasing the operating frequency of the local CPU in order to ensure that it can provide sufficient service quality. However, doing so will cause the device to consume more energy. At the same time, the energy consumption of the system increases faster than the rate of increase of the system computing power, resulting in a decrease in the computing efficiency of the system. In addition, as the CPU frequency of the mobile user device increases, the trend of the change in the computing efficiency of the system gradually decreases. This is because as the CPU frequency increases, more and more devices can meet the quality of service constraints (QoS), and they will not continue to increase the CPU frequency of the mobile user device, which will cause the computing efficiency of the system to decrease. However, the computing efficiency obtained by the AR algorithm maintains the same downward trend, which also proves the effectiveness of the algorithm proposed in this paper in optimizing various variables. In summary, optimizing the CPU frequency of mobile user devices can effectively improve the computing efficiency of the system.

[0160] Except for the BSCEM algorithm, the computational efficiency of each algorithm increases with the change of the maximum transmission power. This is because when the system computing and network resources are limited, in order to meet the minimum computing bit constraint, the devices covered by the drone will maximize the transmission power as much as possible to obtain more computing bits. At this time, the total computing bit number of the terminal system increases faster than the energy consumption, resulting in the continuous improvement of the system's computing efficiency. However, due to Fig.10 It can be seen that the speed at which the system's computational efficiency improves will gradually slow down, because as the transmit power increases, more and more mobile user devices have already reached the minimum computational bit requirement. In this case, further increasing the transmit power will result in a large amount of energy consumption, which in turn will reduce computational efficiency, so the system will not continue to increase the transmit power of mobile user devices.

[0161] Compared with other baseline algorithms, the BSCEM algorithm proposed in this paper obviously shows the above effects, which indirectly verifies the effectiveness of the algorithm proposed in this paper in optimizing transmit power and 3D position, that is, optimizing the transmit power of mobile user equipment can effectively improve the computational efficiency of the system.

[0162] In order to determine the optimal position of the drone and ensure the service quality of the terminal user, the present invention first derives the maximum service radius of the drone that meets the user service quality requirements based on the relationship between the minimum signal-to-noise ratio required for the user device to receive the signal and its service quality, which can ensure the different service quality requirements of different mobile user devices. On the other hand, according to the given maximum service radius, the optimal deployment position of the drone is obtained based on the minimum circumscribed circle and incremental algorithm, which can ensure the different service quality requirements of different mobile user devices and meet the service capacity constraints of the drone. That is, the present invention can ensure the different service quality requirements of different mobile user devices and meet the service capacity constraints of the drone while deploying a minimum number of drones.

[0163] The minimum circumscribed circle algorithm is a method for finding the minimum circumscribed circle of a given set of points. In the fields of computer graphics and computational geometry, the minimum circumscribed circle algorithm is widely used to solve various problems, such as enclosing geometric figures, fitting point clouds, and collision detection. The minimum circumscribed circle refers to a circle that can contain all points in a given set of points and has the smallest radius. In practical applications, the minimum circumscribed circle algorithm is usually based on a greedy strategy, which approximates the minimum circumscribed circle through continuous iteration.

[0164] The random incremental algorithm is an important algorithm in computational geometry. It does not require much theoretical knowledge, has low algorithmic time complexity, and has a wide range of applications. The idea of ​​the incremental method is similar to the first mathematical induction method. Its essence is to transform a problem into sub-problems of a smaller scale. After solving the sub-problem, add it to the current object. Written in recursive form:

[0165] T[n]=T[n-1]+g[n]. The incremental method is simple in form and can be applied to many geometry problems.

[0166] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that this is only for illustration and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A three-dimensional deployment optimization method based on a drone-assisted MEC system, characterized by: The three-dimensional deployment optimization method comprises the following steps: Step 1: Initialization of parameters; Step 2: Model the problem of maximizing the computational efficiency of the terminal system as a nonlinear integer programming problem; Step 3: Use the UAV 3D position deployment algorithm based on IA and SEC to solve the UAV 3D position; Step 4: Use Lagrange multiplier method to solve the CPU frequency, transmit power and terminal offloading decision of the mobile user equipment; Step 5: Use one-dimensional search to solve the beam width of the drone; Step 6: Use the three-stage alternating iterative algorithm (BSCEM) to maximize the calculation rate of the system.

2. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 1 is characterized in that: The parameter initialization in step 1 is to initialize the user distribution area, and establish the user distribution area as a two-dimensional circular plane. Multiple users are randomly distributed within the circular range, and the base station is located in the edge area.

3. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 1 is characterized in that: The specific operation steps of step 2 are: S1. Establish a wireless channel model; S2, establish a local computing model; S3. Establish communication model; S4. Optimize model.

4. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 3 is characterized in that: In the wireless communication system in step S1, the channel model is crucial to simulating real-world wireless communications.

5. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 3 is characterized in that: In the step S2, the mobile user equipment uses a partial offloading method to assign tasks to the drone for calculation.

6. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 3 is characterized in that: In step S3, the channel gain between the mobile user equipment and the drone is regarded as a free space wireless channel model.

7. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 3 is characterized in that: In the step S4, under the premise of ensuring the user's service quality, the computing efficiency of the terminal is maximized by optimizing the offloading association decision, transmission power, computing resource allocation and 3D position of the mobile user equipment.

8. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 1, characterized in that: The specific operation steps of step three are: A1. Maximize the drone service radius; A2. Use the drone 3D position deployment algorithm to optimize the drone 3D position.

9. The three-dimensional deployment optimization method based on the drone-assisted MEC system according to claim 8, characterized in that: In step A1, the signal-to-noise ratio threshold is used as a standard for measuring whether the service quality can meet normal communication requirements.

10. The three-dimensional deployment optimization method based on drone-assisted MEC system according to claim 1, characterized in that: The specific operation steps of step six are to use a three-stage alternating iterative algorithm (BSCEM) to alternately iterate steps three, four, and five to solve the optimal mobile user device CPU frequency, transmission power and terminal offloading decision variables, the beam width of the drone, and maximize the terminal system computing efficiency.

Images