Method for trajectory optimization, data caching and resource optimization allocation of unmanned aerial vehicle MEC network considering task priority

Abstract

The application discloses a method for optimizing trajectory, data caching and resource allocation of a UAV MEC network considering task priority, and steps are as follows: S1, a multi-UAV data collection system model is established, priority of backhaul information is divided according to task requirements, a differentiated caching reward is set, and a hovering point is set by using a K-means clustering method; S2, under the condition of a given hovering trajectory, a task offloading decision is optimized by combining a branch and bound algorithm; meanwhile, a file caching strategy is determined based on priority and a 0-1 knapsack model, and caching reward is maximized; S3, under the condition of a given offloading and caching strategy, an optimization algorithm is used to optimize the position of the hovering point of the UAV with the minimum total endurance time of the UAV as a target; and S4, an iterative optimization algorithm based on block coordinate descent is used to iteratively execute sub-problem solving in steps S2 and S3 until a joint target of minimum total endurance time of the system and maximum caching benefit converges, and finally, an optimized information backhaul, trajectory and data caching scheme is obtained.

Description

Technical Field

[0001] This invention belongs to the field of information and communication engineering technology, and relates to mobile edge computing technology based on UAVs. Specifically, it is a method for optimizing UAV MEC network trajectory, data caching, and resource optimization allocation that takes into account task priority. Background Technology

[0002] In recent years, with the continuous maturation of unmanned aerial vehicle (UAV) technology and the expanding scope of its applications, its role in data acquisition, environmental monitoring, emergency communication, and disaster relief has become increasingly prominent. For remote areas with complex terrain, weak communication infrastructure, or insufficient signal coverage, UAVs, with their flexible deployment, high maneuverability, and wide coverage, have gradually become an important technological means for achieving data backhaul and communication relay. However, in such application environments, UAVs still face various limitations and challenges when performing information backhaul tasks. Therefore, under the premise of ensuring communication quality and data backhaul requirements, rationally planning the UAV flight path, reducing unnecessary flight and hovering overhead, and efficiently configuring onboard cache resources are of great significance for improving the overall system operating efficiency and task execution performance. Based on this, this invention proposes a method for optimizing UAV MEC network trajectory, data caching, and resource allocation that considers task priority. Summary of the Invention

[0003] To address the aforementioned problems in existing technologies, this invention discloses a method for optimizing UAV MEC network trajectory, data caching, and resource allocation while considering task priority. Under the condition of known user location, task priority, and task data volume, this invention constructs a mobile edge network data acquisition optimization model. Within the block coordinate descent (BCD) framework, it combines branch and bound (B&B) algorithms and knapsack models to jointly optimize the UAV's data acquisition strategy, flight trajectory planning, and cache resource allocation.

[0004] The present invention adopts the following technical solution:

[0005] A method for optimizing UAV MEC network trajectory, data caching, and resource allocation while considering task priority, comprising the following steps:

[0006] S1, System Modeling and Initialization: Establish a multi-UAV data collection system model, including communication model, latency model, etc., and prioritize the transmitted information according to the task requirements, set differentiated cache rewards, and use K-means clustering method to set hovering points.

[0007] S2, Unloading Decision and Cache Scheduling Optimization: Given a hovering trajectory, the task unloading decision is optimized by combining the branch and bound algorithm (B&B). At the same time, the file caching strategy is determined based on priority and the 0-1 knapsack model to maximize cache rewards.

[0008] S3, Hovering Trajectory Optimization: Given the unloading and caching strategies, optimization algorithms such as Multi-Subgroup Genetic Algorithm (MPGA) or Serial Convex Approximation (SCA) are used to optimize the hovering point position of the UAV with the goal of minimizing the total endurance time of the UAV (including hovering time and flight time).

[0009] S4, Iterative Joint Optimization: An iterative optimization algorithm based on block coordinate descent (BCD) is adopted to iteratively solve the subproblems in steps S2 and S3 until the joint objective of minimizing the total system endurance and maximizing cache efficiency converges, and the final information backhaul (i.e. resource optimization), trajectory and data cache optimization schemes are obtained.

[0010] Compared with the prior art, the technical solution provided by this invention has the following significant advantages:

[0011] 1. A precise priority scheduling mechanism for task processing is introduced: This invention introduces a priority weighting factor and a differentiated caching reward mechanism, enabling UAVs to prioritize bandwidth allocation and storage space for high-priority tasks under resource constraints. Compared to traditional indiscriminate scheduling algorithms, this invention significantly improves the latency guarantee capability for critical information transmission and the weighted efficiency of the system.

[0012] 2. Achieving Collaborative Scheduling and Overall Optimization of Multi-Dimensional Resources: This invention breaks away from the traditional approach of handling UAV trajectory planning, task offloading strategies, and data cache management separately, integrating all three into a unified optimization model for comprehensive analysis and solution. By introducing a block coordinate descent (BCD) iterative solution mechanism, the system continuously updates alternately among different decision variables, enabling dynamic coordination across multiple dimensions to achieve globally optimized results. For example, changes in the UAV's flight path directly affect communication link quality and data transmission rate, further influencing task offloading selection and cache resource utilization. Simultaneously, differences in the spatial distribution of tasks with different priorities also guide UAV path planning. Through this collaborative optimization mechanism among multi-dimensional variables, the system's task execution efficiency can be effectively improved, overall energy consumption reduced, and cache resource utilization efficiency further enhanced.

[0013] 3. Significantly improves the balance between system energy efficiency and timeliness: By deeply coupling trajectory planning (Greedy TSP) with task offloading and caching decisions, this invention enables the system to reduce UAV flight energy consumption while alleviating bandwidth pressure on the backhaul link through intelligent caching strategies. This collaborative optimization scheme effectively extends the UAV's operating time and shortens the total execution time of large-scale data acquisition tasks.

[0014] 4. Excellent computational efficiency and engineering applicability: This invention significantly reduces the search space through K-means clustering preprocessing and reduces the dimensionality of complex problems using block coordinate descent. Combined with a greedy trajectory planning strategy, the algorithm maintains low computational complexity even when handling large-scale user nodes, meeting the real-time scheduling requirements of UAV systems in dynamic environments. Attached Figure Description

[0015] Figure 1 This is a flowchart of a preferred embodiment of the present invention for a UAV MEC network trajectory optimization, data caching, and resource optimization allocation method that considers task priority;

[0016] Figure 2 This is an application scenario diagram of a preferred embodiment of the present invention, which considers task priority in UAV MEC network trajectory optimization, data caching, and resource optimization allocation method.

[0017] Figure 3 This invention provides a UAV flight trajectory diagram based on a preferred embodiment of a method for optimizing UAV MEC network trajectory, data caching, and resource optimization allocation that considers task priority.

[0018] Figure 4 The simulation results are shown to illustrate the changing trend of the system optimization objective under different numbers of users in the preferred embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings.

[0020] like Figure 1 As shown in the figure, this embodiment proposes a method for optimizing UAV MEC network trajectory, data caching, and resource allocation considering task priority. The specific steps are as follows:

[0021] Step S1, system modeling and initialization, is as follows:

[0022] S1.1 System Modeling: Establish a multi-UAV data collection system model, including communication model, latency model, flight energy consumption model, and buffer model;

[0023] Model building:

[0024] like Figure 2 As shown, this embodiment considers a cluster of U-shaped UAVs providing data collection services to K ground users distributed within a target area. The set of ground users is represented as... , No. The three-dimensional coordinates of each user are A collection of drones is represented as Each drone flies at a fixed altitude H, symbol The first drone There are L hovering points. The system allows a drone to have a maximum of L hovering points, and each drone has L different hovering points set. Therefore, the set of drone hovering points can be represented as... .in, The first character representing the drone u Coordinates of the hovering service point The x-coordinate of the drone's hovering point. Let H be the y-axis coordinate, and H be the fixed altitude of the drone. Let L be the Euclidean distance between user k and drone u at hovering point l. The calculation formula is as follows:

[0025]

[0026] The specific details of each model are as follows:

[0027] 1) Communication model

[0028] This embodiment employs an air-to-ground line-of-sight (LoS) channel model. Channel power gain. The model is as follows:

[0029]

[0030] in, Represents reference distance The channel gain constant at a distance of meters. The path loss exponent typically ranges from 2 to 4, depending on the propagation environment. To balance the acquisition latency for users at different distances, a bandwidth allocation strategy based on distance ratio is adopted. The bandwidth allocated to user k at hovering point l by the drone u is... for:

[0031]

[0032] in, Let B be the set of all ground users served by UAV u at its l-th hovering point, and let B be the total system bandwidth. It is an exponential factor used to adjust the fairness and efficiency of bandwidth allocation. According to the Shannon-Hartley theorem, it represents the maximum achievable offloading rate that user k can achieve when offloading tasks from a drone l located at hovering point u. Expressed as:

[0033]

[0034] Here, Let k be the transmit power during data transmission. This represents the additive white Gaussian noise power spectral density at the UAV receiver.

[0035] 2) Unloading and Delay Model

[0036] Introducing binary variables This indicates the offloading decision. If user k's data task is collected by drone u at point l, then... , on the contrary The set of binary decision variables is defined as follows: Each user's task is indivisible, therefore it must be offloaded to a single hovering point of a drone for processing, which means the following constraints must be met:

[0037]

[0038] Drone u at the lth hovering point Hover duration It depends on the maximum time required for data upload to be completed among all users responsible for that point:

[0039]

[0040] Among them, D k Let user k be the total amount of data to be collected. The drone u starts from its current hovering point. Fly at a constant speed to the next planned hovering point Required flight time Calculated as:

[0041]

[0042] in, Represents Euclidean distance. The maximum permissible average flight speed of the drone; the total execution time required for the drone to complete all data collection tasks assigned to it by all users. This includes the sum of hovering time at each hovering point and flight time along the path:

[0043]

[0044] in, This is the maximum flight speed.

[0045] 3) Flight energy consumption model

[0046] This invention employs a commonly used propulsion power consumption model for rotary-wing UAVs. Therefore, the propulsion power of the UAV at speed v is... The model is as follows:

[0047]

[0048] in, The blade profile power during hovering. The induced power during hovering, The tip linear velocity of the rotor blades, C represents the average rotor-induced airflow velocity during hovering. D The coefficient of parasitic drag of the fuselage. For air density, The equivalent frontal area of ​​the fuselage; the drone in hovering state. The power consumption at that time is The drone cruised at its maximum speed. Power consumption during flight is In addition, the drone's communication and computing modules also have basic operating power consumption. The total energy consumption of drone m is then... This includes total hovering energy consumption, total flight energy consumption, and basic operational energy consumption during the mission:

[0049]

[0050] The total energy consumption generated during the execution of a drone mission should be limited by its initial battery capacity. Restrictions, .

[0051] 4) Caching Model

[0052] This invention categorizes all user data files to be collected into different types based on their business type and timeliness requirements. There are three different priority levels: emergency rescue data is the highest priority, real-time environmental monitoring data is the medium priority, and routine data records are the lowest priority (i.e., priority classification); for each priority level... Allocate a cache reward value for a unit of data (e.g., per bit). , and satisfy , indicates that high-priority data has a higher caching value; if user k's data file If the file belongs to priority p, then the expected total cache reward for caching this file is... Defined as:

[0053]

[0054] The above settings include differentiated caching rewards.

[0055] Introducing binary cache decision variables , when The time indicates that the drone decides to store user k's data file in its onboard cache; otherwise... This indicates that the cache should not be cached or should be replaced when the cache is full; the maximum onboard cache storage capacity for each drone is [value missing]. (Unit: bits), therefore the total size of all cached data files cannot exceed this capacity limit:

[0056]

[0057] in It is the set of all users that the drone u may contact and consider caching its data.

[0058] The specific optimization objectives are as follows:

[0059] This invention comprehensively considers task timeliness and caching benefits. Its global optimization objective is to find the optimal set of unloading decisions. The optimal caching decision set , and the optimal set of drone hovering trajectories To minimize the system's global joint optimization objective function as follows:

[0060]

[0061] in, These are non-negative weighting coefficients corresponding to cache rewards, used to adjust the relative importance between different optimization objectives; (13b) and (13c) ensure that each user's data is uniquely allocated; (13d) and (13e) are cache decision and capacity constraints; (13f) is the safe distance constraint between UAVs; (13g) is the total energy consumption constraint of UAVs; and (13h) is the effective communication coverage radius constraint of UAVs. This indicates the effective communication coverage radius of the drone.

[0062] S1.2 System Initialization

[0063] S1.2.1. Parameter Settings: Set the number of users K, the number of drones U, and the number of hovering points per drone L. dThe simulation area is 300 meters × 300 meters, and the communication parameters are as follows (total bandwidth B, user transmit power P). k Noise power spectral density Channel gain factor Effective communication radius R of UAV cov ), UAV parameters (maximum flight speed v) max Flight altitude H, minimum safe distance d safe ), cache parameters (drone storage space limit) ).

[0064] S1.2.2. User and Data Initialization: Randomly generate the two-dimensional locations of K users within a specified area. A file of size D is randomly generated for each user. k (bits), and randomly assign priority p k (1, 2, or 3). Calculate the cache reward for each file based on priority and file size. .

[0065] S1.2.3. Initial Hovering Point Determination: To obtain a reasonable initial hovering position for the UAV and reduce the search space of subsequent optimization problems, the K-Means clustering algorithm is used to preprocess all user positions. Let the number of clusters be:

[0066] Nc = U × L d

[0067] This refers to the total number of hovering points for all drones in the system.

[0068] Where Nc is the total number of hovering points for all drones in the system.

[0069] Given a set of user locations:

[0070] W = {w1 , w2 , ... , w k}

[0071] The K-Means algorithm determines cluster centers by minimizing the Within-Cluster Sum of Squares (WCSS), and its optimization objective is expressed as:

[0072]

[0073] Where Sc represents the c-th user cluster, Indicates the corresponding cluster center.

[0074] In each iteration, users are first assigned to the nearest cluster center according to the principle of minimum Euclidean distance:

[0075]

[0076] in, For the first The center of each cluster, Corresponding to L d There are 10 hovering points, and the cluster center has a total of L. d indivual.

[0077] Then update the cluster center location:

[0078]

[0079] Repeat the above process (starting from assigning users to the nearest cluster center according to the least Euclidean distance principle in each iteration, and continuing until the cluster center positions are updated) until the clustering results converge. The final Nc cluster centers are used as the three-dimensional coordinates of the UAV's initial hovering point, i.e.:

[0080]

[0081] Where u represents the u-th drone, l represents the l-th hovering service point, and H is the fixed flight altitude of the drone. The first term of the drone u Coordinates of the hovering service point; The x-coordinate of the cluster center y is the y-coordinate of the cluster center.

[0082] S1.2.4. Initial Transmission Rate Calculation: Based on the initial user location, initial hovering point location, and communication model, calculate the initial achievable data offloading rate between all user-UAV-hovering point pairs. .

[0083] Step S2, Unloading Decision and Cache Scheduling Optimization, is as follows:

[0084] S2.1 Unloading decision optimization:

[0085] Fix the current hovering position of the drone And caching decisions The goal is to optimize the user uninstallation decision A and the hover time at each hover point to minimize the total hover time.

[0086] This subproblem (i.e., unloading decision and hover time optimization) can be modeled as a mixed-integer linear programming problem. This function uses the CVX toolbox to solve it (i.e., combining the branch and bound algorithm to optimize the task unloading decision; the CVX toolbox includes the branch and bound algorithm).

[0087] Variables: Binary unloading decision variables and continuous hovering time variable .

[0088] Objective function: .

[0089] Key constraints:

[0090] 1) Each user must connect to one and only one hover point.

[0091] 2) Users can only be assigned to feasible hovering points within their communication range.

[0092] 3) Hovering time constraint: Where, min_transmission_time(k,u,l)= Where u represents the u-th drone, l represents the l-th hovering service point, and k represents the k-th user. This means that the hovering time must be long enough to serve the most time-consuming user assigned to it.

[0093] 4) The hovering time is non-negative and has an upper limit (as shown in the code). ).

[0094] After solving, the unloading decision A for the current iteration is obtained. (i) and hovering time T hover(i) .

[0095] S2.2. Optimization of Cache Decisions (File Caching Strategy)

[0096] Fixed unloading decision A in the current iteration (i) and hover trajectory Q (i) The goal is to optimize the caching decision X for each drone, maximizing the total cache reward under cache capacity constraints.

[0097] For each drone u, this problem is a classic 0-1 knapsack problem.

[0098] Items: Files assigned to all users of the drone u service (according to A) (i) Determine which users are served by the drone (u service).

[0099] Item weight: File size D k .

[0100] Item value: File cache reward .

[0101] Backpack capacity: The cache capacity of the drone u .

[0102] For the 0-1 knapsack problem, dynamic programming can be used to obtain the optimal solution:

[0103] 2.2.1. Define the state:

[0104] set up This represents the maximum total cache reward that can be obtained by selecting from the top k candidate files (file1 to filek) and placing them into a cache with a current capacity limit of cap.

[0105] 2.2.2. Initialize the state transition matrix:

[0106] Z[0][cap] = 0 for all cap values ​​from 0 to 0. .

[0107] Z[k][0] = 0 for all k from 0 to K.

[0108] Z[k][cap] is a (K + 1) × ( A matrix of size +1).

[0109] 2.2.3. State transition equation:

[0110] For the k-th file under consideration (whose size is D) k Its value is And the current considered cache capacity limit cap:

[0111] Case 1: The current capacity cap is less than the size D of the k-th file. k

[0112] At this point, the k-th file definitely cannot be placed. Therefore, the maximum reward is the same as the reward for considering only the first k-1 files under the capacity cap. Z[k][cap] = Z[k-1][cap].

[0113] Case 2: The current capacity cap is greater than or equal to the size D of the k-th file. k

[0114] At this point, there are two options:

[0115] If the k-th file is not included: the maximum reward is the same as the reward for considering only the first k-1 files under the capacity cap, i.e., Z[k-1][cap].

[0116] Inserting the k-th file: The reward is the value of the k-th file. Add "from the previous k-1 files;

[0117] Select to add remaining capacity cap-D k The maximum reward that can be obtained is... .

[0118] In this embodiment, the option with the larger reward is selected.

[0119] Therefore, the state transition equation is:

[0120]

[0121] This represents the maximum total cache reward that can be obtained by selecting from the first k candidate files (file1 to filek) and placing them into a cache with a current capacity limit of cap. Item weight: file size D k Item value: File cache reward .

[0122] 2.2.4. Solving and Backtracking:

[0123] According to the above equation, from k = 1 to K, from cap = 1 to... Calculate sequentially to fill the entire Z matrix. Finally, This refers to the maximum cache reward that the drone can obtain within its total cache capacity. To determine which specific files are cached, it is necessary to... Start backtracking. Determine whether the k-th file is selected by comparing the values ​​of Z[k][cap] and Z[k-1][cap], and update cap and k accordingly until all files have been considered.

[0124] Step S3, hover trajectory optimization, is detailed below:

[0125] Drone hovering trajectory optimization: Fix the unloading decision A obtained in the current iteration (i) And caching decision X (i＿1) The goal is to optimize the hovering service point coordinates Q of all UAVs to minimize the total flight time while satisfying flight safety constraints. A Multi-Subgroup Genetic Algorithm (MPGA) is employed to enhance the global search capability. Details are as follows:

[0126] S3.1. Problem Definition and Individual Coding:

[0127] Task: U drones, each assigned L different hover points from the global candidate hover point set hover_points_global, and determine their respective access order.

[0128] Encoding: Each MPGA individual is a U × L matrix. Each row represents a scheme for a drone, which is a permutation of L global hover point indices, indicating the order in which the drone visits these hover points.

[0129] S3.2. Initialization:

[0130] Create multiple subpopulations. Each individual randomly or heuristically assigns L unique global hover point indices to each drone and randomizes their access order.

[0131] S3.3. Fitness Assessment:

[0132] For each individual (U × L scheme):

[0133] Decoding: Convert the hover point index and order it represents into a specific UAV flight path point set Q (connecting the start point, assigned hover points, and end point).

[0134] calculate:

[0135] Total flight time: usually the maximum time required for all drones to complete the mission (makespan), or the sum of the flight times of all drones.

[0136] Collision penalty: Detect whether the decoded path Q has a collision. If it does, apply a significant penalty.

[0137] Fitness function: The objective is to minimize total flight time + α × collision penalty.

[0138] S3.4. Genetic manipulation (performed independently within each subpopulation):

[0139] Selection: Using methods such as tournament selection, individuals with high fitness are given priority as parents.

[0140] Crossover: Using operators suitable for permutation encoding, such as order crossover, parent genes are combined to generate offspring.

[0141] Mutation: Mutate offspring with a low probability, such as swapping two hovering point indices in a drone's access sequence.

[0142] S3.5. Elite Migration Mechanism:

[0143] MPGA core: Periodically select the best individuals (elites) from each subpopulation and replicate them in other subpopulations to replace inferior individuals, thereby promoting the spread of superior genes and maintaining diversity.

[0144] S3.6. Termination Condition:

[0145] The maximum number of generations or fitness value is reached and convergence occurs (the change is less than the threshold).

[0146] S3.7. Output:

[0147] When the algorithm terminates, the individual with the highest fitness among all subpopulations is the found optimal hover point assignment and access order scheme. For example... Figure 3 As shown, this scheme can be directly used to generate the final UAV flight path Q and flight time T. tr .

[0148] Step S4, iterative joint optimization, is as follows:

[0149] S4.1. For example Figure 3 As shown, an iterative optimization algorithm based on block coordinate descent is adopted to iteratively solve the sub-problems (unloading decision and cache scheduling optimization problem and hovering trajectory optimization problem) in steps S2 and S3 until the joint objective of minimizing the total system endurance and maximizing cache efficiency converges, and the final information backhaul, trajectory and data cache optimization scheme is obtained.

[0150] S4.2. Convergence Judgment and Update:

[0151] Calculate the comprehensive objective function value for the current iteration. .

[0152] Determine whether the convergence condition is met or the maximum number of iterations has been reached.

[0153] If convergence is not achieved, update. Let i = i + 1, and return to step S2.1.

[0154] If convergence is achieved, the final optimized solution is output. (i.e., the final information feedback and trajectory-level caching optimization scheme).

[0155] The simulation experiment and result analysis are as follows:

[0156] To verify the effectiveness of the method proposed in this invention, a simulation experiment was conducted. The simulation parameters were set as described in the parameter setting stage.

[0157] System architecture: such as Figure 3 As shown, multiple drones provide data collection and transmission services to ground users within a designated area, with user data having different priorities.

[0158] Performance metrics: The main focus is on the overall objective function value, which combines total hover time, flight time, and cache reward.

[0159] Results analysis: such as Figure 4The simulation diagram (showing the change in optimization objective under different numbers of users) illustrates the variation of the average total objective value obtained by the optimization algorithm proposed in this invention under different numbers of users. The diagram typically includes a mean line and error bars representing the standard deviation. As can be seen from the diagram, the trend of the system's total objective value (e.g., total cost or total latency) as the number of users (or drones) increases can be used to evaluate the algorithm's scalability and performance under different loads.

[0160] Simulation results show that the proposed UAV MEC network trajectory optimization, data caching, and resource optimization allocation method considering task priority can effectively perform resource allocation and path planning, achieving good system performance while satisfying various constraints.

[0161] A preferred embodiment of the present invention also provides a UAV MEC network trajectory optimization, data caching, and resource optimization allocation system that considers task priorities, for implementing the above methods. This system can be integrated into a ground control station or distributed and partially implemented in a UAV swarm. Its main functional modules include:

[0162] System status awareness and parameter initialization module: responsible for collecting user location, task information, UAV status, channel conditions, etc., and initializing algorithm parameters.

[0163] The joint optimization module for task unloading decision and cache management executes the core algorithms of steps BCD-1 and BCD-3 to determine user task allocation and data caching strategies.

[0164] The UAV 3D flight trajectory dynamic planning module executes the core algorithm of step BCD-2 to plan or adjust the UAV's flight path and hovering point.

[0165] Iterative Optimization and Cooperative Control Module: Implements the BCD iterative framework, coordinates the operation of each module, controls the flow until convergence, and sends the final optimization instructions to the UAV and user terminal.

[0166] These modules work together to enable priority-based dynamic optimization of UAV information transmission and trajectory.

[0167] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

Claims

1. A method for optimizing UAV MEC network trajectory, data caching, and resource allocation considering task priority, characterized by: The specific steps are as follows: S1, System Modeling and Initialization: Establish a multi-UAV data collection system model, prioritize the transmitted information according to task requirements, set differentiated caching rewards, and use the K-means clustering method to set hovering points; S2, Unloading Decision and Cache Scheduling Optimization: Given a hovering trajectory, the task unloading decision is optimized by combining the branch and bound algorithm; at the same time, the file caching strategy is determined based on priority and the 0-1 knapsack model to maximize cache rewards; S3, Hovering Trajectory Optimization: Given the unloading and caching strategies, an optimization algorithm is used to optimize the hovering point position of the drone with the goal of minimizing the total flight time of the drone. S4, Iterative Joint Optimization: An iterative optimization algorithm based on block coordinate descent is adopted to iteratively solve the sub-problems in steps S2 and S3 until the joint objective of minimizing the total system endurance and maximizing cache efficiency converges, and the final optimization scheme for information backhaul, trajectory and data caching is obtained.

2. The method for optimizing and allocating UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 1, characterized in that, In step S1, a cluster of U-shaped UAVs is established to provide data collection services to K ground users distributed within the target area; the set of ground users is represented as... , No. The three-dimensional coordinates of each user are The collection of drones is represented as Each drone flies at a fixed altitude H. The first drone There are L hovering points. The system allows a maximum of L hovering points for each drone, with each drone having L distinct hovering points. The set of drone hovering points is represented as: ;in, The first term of the drone u The coordinates of a hovering service point; the Euclidean distance between user k and drone u at hovering point l. The calculation formula is as follows: 。 3. The method for optimizing UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 2, characterized in that, In step S1, the multi-UAV data collection system model includes a communication model, a latency model, a flight energy consumption model, and a cache model.

4. The method for optimizing UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 3, characterized in that, The specific communication model is as follows: An air-to-ground line-of-sight channel model is adopted; channel power gain. The model is as follows: in, Represents reference distance The channel gain constant at a distance of meters. The path loss index is used; a bandwidth allocation strategy based on distance ratio is adopted; the bandwidth allocated to user k at hovering point l by UAV u is... for: Where B is the total system bandwidth, Let U be the set of all ground users served by UAV u at its l-th hovering point. It is an exponential factor used to adjust the fairness and efficiency of bandwidth allocation; the maximum unloading rate that user k can achieve when unloading a task from a drone l located at hovering point u. Expressed as: in, Let k be the transmit power during data transmission. This represents the additive white Gaussian noise power spectral density at the UAV receiver.

5. The method for optimizing and allocating UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 4, characterized in that, The unloading model and the delay model are as follows: Introducing binary variables This indicates the offloading decision. If user k's data task is collected by drone u at point l, then... , on the contrary ; The binary decision variable set is defined as Each user's task is indivisible, therefore the following constraints must be met: Drone u at the lth hovering point Hover duration It depends on the maximum time required for data upload to be completed among all users responsible for that point: Among them, D k Let user k be the total amount of data to be collected; let drone u start from its current hovering point. Fly at a constant speed to the next planned hovering point Required flight time Calculated as: in, Represents Euclidean distance. The maximum permissible average flight speed of the drone; the total execution time required for the drone to complete all data collection tasks assigned to it by all users. This includes the sum of hovering time at each hovering point and flight time along the path: in, This is the maximum flight speed.

6. The method for optimizing UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 5, characterized in that, The specific flight energy consumption model is as follows: A propulsion power consumption model for a rotary-wing UAV is adopted; the propulsion power of the UAV at speed v. The model is as follows: in, The blade profile power during hovering. The induced power during hovering, The tip linear velocity of the rotor blades, C represents the average rotor-induced airflow velocity during hovering. D The coefficient of parasitic drag of the fuselage. For air density, The equivalent frontal area of ​​the fuselage; the power consumption of the drone in hovering state is... The drone cruised at its maximum speed. Power consumption during flight is The drone's communication and computing modules also have basic operating power consumption. The total energy consumption of drone m is then... This includes total hovering energy consumption, total flight energy consumption, and basic operational energy consumption during the mission: 。 7. The method for optimizing and allocating UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 6, is characterized in that... The caching model is as follows: All user data files to be collected are divided according to their business type and timeliness requirements. There are three different priority levels: emergency rescue data is the highest priority, real-time environmental monitoring data is the medium priority, and routine data records are the lowest priority. For each priority level... Allocate a cache reward value for one unit of data and satisfy , indicates that high-priority data has a higher caching value; if user k's data file If the file belongs to priority p, then the expected total cache reward for caching this file is... Defined as: Introducing binary cache decision variables , when When the time is right, it indicates that the drone has decided to store user k's data file in its onboard cache; otherwise... This indicates that the cache should not be cached or should be replaced when the cache is full; the maximum onboard cache storage capacity for each drone is [value missing]. Therefore, the total size of all cached data files cannot exceed this capacity limit. in, It is the set of all users that the drone u may contact and consider caching its data.

8. The method for optimizing and allocating UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 7, characterized in that the steps... In S1, the specific optimization objective is as follows: The global optimization objective is to find the optimal set of unloading decisions. The optimal caching decision set , and the optimal set of drone hovering trajectories To minimize the system's global joint optimization objective function as follows: in, This corresponds to the non-negative weighting coefficient of the cache reward; Equations (13b) and (13c) ensure that each user's data is uniquely allocated; Equations (13d) and (13e) are cache decision and capacity constraints; Equation (13f) is the safe distance constraint between UAVs; Equation (13g) is the total energy consumption constraint of the UAVs; and Equation (13h) is the effective communication coverage radius constraint of the UAVs. This indicates the effective communication coverage radius of the drone.

9. The method for optimizing and allocating UAV MEC network trajectory, data caching, and resource allocation considering task priority as described in claim 8, characterized in that, Step S2 is as follows: S2.1 Unloading Decision Optimization: Fix the current drone hovering point position And caching decisions ; The goal is to optimize the user uninstallation decision A and the hover time at each hover point in order to minimize the total hover time. S2.2 Cache Decision Optimization: Fix the unloading decision A of the current iteration (i) and hover trajectory Q (i) ; The goal is to optimize the caching decision X for each drone, maximizing the total cache reward under cache capacity constraints.

10. The method for optimizing and allocating UAV MEC network trajectory, data caching, and resources considering task priorities as described in claim 9, characterized in that, In step S3, the unloading decision A obtained in the current iteration is fixed. (i) And caching decision X (i＿1) ; The goal is to optimize the hovering service point coordinates Q of all UAVs to minimize the total flight time while satisfying flight safety constraints; a multi-subgroup genetic algorithm is used to optimize the hovering point positions of the UAVs.

Images