A method for joint optimization of unmanned aerial vehicle location deployment, task scheduling and resource allocation

Abstract

This invention belongs to the field of UAV edge computing technology and discloses a joint optimization method for UAV location deployment, task scheduling, and resource allocation. The method includes the following steps: S3, unfolding the coordinate codes of all UAVs sequentially and concatenating them with the codes of all task scheduling in order to form a one-dimensional vector; S4, calculating the total task latency and UAV flight energy consumption using a fitness function, and evaluating the merits of each one-dimensional vector; S5, based on the obtained one-dimensional vector, total task latency, and UAV flight energy consumption, performing real-time joint optimization of UAV location deployment, task scheduling, and resource allocation using a dynamic multi-objective evolutionary algorithm. This invention provides efficient, balanced, and flexibly adaptable decision support for UAV edge computing systems in dynamic and changing scenarios through a dynamic adaptive mechanism, global joint optimization, fast convergence strategy, multi-objective equilibrium decision-making, and fair and efficient resource allocation.

Description

Technical Field

[0001] This invention belongs to the field of edge computing technology for unmanned aerial vehicles (UAVs), and in particular relates to a joint optimization method for UAV location deployment, task scheduling, and resource allocation. Background Technology

[0002] In recent years, with the rapid development of the Internet of Things (IoT) and intelligent sensing technologies, the demand for real-time, high-precision geospatial information has become increasingly urgent. As an emerging paradigm, UAV edge computing endows UAV platforms with edge computing capabilities, upgrading them from simple data collectors to aerial intelligent nodes integrating perception, computing, and decision-making. UAV edge computing is a deep integration of UAV mobility and the low-latency characteristics of edge computing. By offloading computing tasks to the UAV's local location, ground edge nodes, or nearby UAV clusters, it reduces the latency of data transmission to the cloud, enabling real-time data processing and decision-making. Leveraging the advantages of UAVs' maneuverability and unique perspective, combined with the low latency and high bandwidth of edge computing, it can achieve near-ground real-time perception and processing with centimeter-level accuracy and minute-level response, effectively compensating for the shortcomings of long satellite remote sensing cycles and narrow coverage of ground-based sensors. This provides revolutionary technical support for applications such as precision agriculture monitoring, disaster relief, logistics distribution, smart cities, and industrial inspection.

[0003] In UAV edge computing systems, UAV deployment, task scheduling, and resource allocation are three core and interdependent dimensions that determine the overall system performance and energy efficiency. Deployment location determines the quality of perception coverage and the status of communication links; task scheduling affects the order and strategy of distributing computing load between the UAV and ground facilities; and resource allocation directly involves the effective utilization of limited onboard computing and communication resources. Independent optimization of these three aspects often leads to suboptimal overall system performance or even conflicts between them. Therefore, joint optimization is a necessary requirement and key to unlocking the full potential of UAV edge computing and achieving the optimal balance between multiple objectives such as perception service quality, task processing latency, and system energy consumption in dynamic and complex environments.

[0004] Currently, academia and industry have conducted relevant research. Existing work mainly focuses on two levels: First, static or single-objective optimization, such as optimizing task offloading strategies to minimize latency under fixed deployment, or optimizing resource allocation to save energy under a given task; Second, step-by-step or decoupled optimization, that is, first determining the UAV deployment, and then performing task scheduling and resource allocation based on this. The above research provides valuable theoretical foundations and algorithmic tools for system optimization. However, facing dynamic and complex real-world application scenarios, existing technologies still face the following fundamental challenges: 1. Lack of dynamism: Most models assume that the environment and tasks are static and unchanging, failing to consider dynamic factors such as random task arrival, time-varying channels, and UAV endurance degradation, resulting in poor adaptability of optimization strategies in practice; 2. Insufficient joint optimization: Decoupling or sequentially optimizing deployment, scheduling, and allocation problems ignores their inherent strong coupling relationship, making it difficult to obtain the globally optimal solution; 3. Multi-objective trade-offs: When simultaneously optimizing multiple objectives such as latency, energy consumption, and even coverage, there is a lack of efficient dynamic multi-objective collaborative decision-making mechanisms. These challenges make it difficult for existing technologies to support the long-term, autonomous, and efficient operation of UAV edge computing systems. Summary of the Invention

[0005] To address the aforementioned issues, the present invention aims to provide a joint optimization method for UAV location deployment, task scheduling, and resource allocation based on dynamic multi-objective evolution for UAV edge computing scenarios. This method enables the UAV edge computing system to continuously maintain Pareto optimal multi-objective performance amidst dynamic changes, thereby significantly improving the overall adaptability, efficiency, and robustness of the UAV edge computing system.

[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: A joint optimization method for UAV location deployment, task scheduling, and resource allocation includes the following steps: S1. Unfold the coordinate codes of all UAVs in sequence and concatenate them with the codes of all task scheduling in order to form a one-dimensional vector; S2. Calculate the total mission delay and UAV flight energy consumption using the fitness function, and evaluate the merits of each one-dimensional vector in step S1. S3. Based on the one-dimensional vector obtained in step S1, the total task delay and UAV flight energy consumption obtained in step S2, perform real-time joint optimization of UAV location deployment, task scheduling and resource allocation based on dynamic multi-objective evolutionary algorithm.

[0007] Furthermore, in step S1 above, the encoding method is as follows: Assuming there is a total The deployment location of each drone is represented by two-dimensional coordinates. express, Coordinate values ​​are represented by real numbers or integers, and are ordered according to the UAV number sequence. Arranged sequentially, forming a coordinate sub-vector with a length of ; The task scheduling section has a total of There are 1 task, and each task corresponds to a binary decision variable. , Decision variables A value of 1 indicates that the task is processed by the drone base station that received the request, while a value of 0 indicates that the task is offloaded to the cloud computing platform for processing. All binary variables are arranged in order of task number to obtain the scheduling sub-vector, with a length of [missing value]. ; Concatenate the coordinate subvector with the scheduling subvector, forming a vector of length [length missing]. A one-dimensional vector.

[0008] Further, step S2 above includes: First, the one-dimensional vector is decoded into drone deployment locations and task scheduling schemes: based on the number of drones. and number of tasks Extract the first... from the starting position of the one-dimensional vector sequentially. These values ​​correspond in sequence. Restore the coordinates of each drone; remaining The given values ​​form a task scheduling sub-vector, where each value is a binary variable ordered by task number. Assign values ​​to the corresponding tasks respectively; Then, based on the deployment location and task scheduling scheme obtained from the decoding, the performance indicators of the entire system are calculated, including total task latency and UAV flight energy consumption; the operation is as follows: Based on task scheduling variables The tasks are divided into two categories: a set of tasks to be handled by drones. and sets of tasks processed in the cloud Define each drone Set of tasks to be processed ; Set up drones There are a total of The first task, the... The bandwidth ratio obtained by each task on this drone and computing resource ratio The allocations will be made according to the following formulas to ensure that the drones are assigned accordingly. The tasks on each platform have the same data transmission latency and computation latency: (1) (2) in, and The first The amount of data transmitted and the amount of computation required for each task; For the first Data transfer volume per task; For the first The computational requirements of each task; Based on the above ratio, drones Total latency of all tasks for: (3) in, and drones Total bandwidth resources and total computing resources; The total transmission latency for all tasks; Total computation delay; For tasks processed in the cloud, all cloud tasks share the cloud's bandwidth and computing resources, and the total number of cloud tasks... Total latency of all cloud tasks for: (4) in, and These refer to the bandwidth resources and computing resources of the cloud computing platform, respectively. Total task latency of the entire system for: (5) Drone flight energy consumption Calculated by the sum of the energy required for all drones to move from their initial positions to their new positions: (6) in, For drones The initial deployment location; The new, optimized location; Energy consumption per unit distance traveled.

[0009] Furthermore, step S3 above includes the following sub-steps: S3.1, Randomly initialize a... A population of individuals, each with a genotype represented by a one-dimensional vector obtained in step S1; evaluate the fitness of each individual: total task latency. and drone flight energy consumption Based on fitness, individuals in the population are sorted quickly without being dominated. This sorting divides individuals into several front layers with different levels of dominance. Individuals in the higher front layers dominate individuals in the lower front layers. The highest-level front individuals in the sorting are selected to form the Pareto optimal solution set, and their corresponding fitness set is taken as the Pareto front. S3.2 The algorithm periodically detects whether the service status of the current UAV edge computing system has changed. The detection method is as follows: First, calculate the average latency of each UAV's processing task at the current moment; then, calculate the coefficient of variation of the average latency of all UAVs. ,in, The standard deviation of the average latency of each drone. The mean of the average latency across all drones is given; finally, this coefficient of variation is calculated. With preset threshold If a comparison is made, If the environment changes, the subsequent evolutionary process is triggered; otherwise, periodic checks continue. S3.3 When an environmental change is detected, individuals are randomly selected from the current population according to a preset ratio, and mutation operations are performed on each selected individual. S3.4, Evolution iteration count Initialize to 0; continue evolutionary iterations to optimize the population, with the number of iterations continuing until the preset maximum value is reached. ; S3.5 After the iteration is completed, the algorithm outputs the latest Pareto optimal solution set and Pareto front, providing a basis for decision-making to dynamically adjust UAV deployment, task scheduling and resource allocation; S3.6 In each generation of evolution, generate four offspring for each individual in the current population; S3.7. Evaluate the fitness of each offspring generated in step S3.6; S3.8. Based on the fitness obtained in step S3.7, perform a fast non-dominated sort on all offspring and all individuals in the current population; update the Pareto optimal solution set to the set of individuals at the highest front layer of the sort, and update the Pareto front to its corresponding fitness set; select individuals from the top of the sort with the same number as the original population, and update them into a new generation population based on the non-dominated sort and crowding distance; iteration count: [number missing]. Return to step S3.6 until the desired result is reached. .

[0010] Furthermore, in step S3.3 above, the mutation operation includes: for the real numbers in the coordinate part, a polynomial mutation is used; for the binary bits in the task scheduling part, a bit-flip mutation is used.

[0011] Furthermore, step S3.6 above is performed as follows: S3.61. Randomly select an individual from the current population that is different from the current individual, and perform cross-operation between the two; S3.62. Randomly select an individual from the current Pareto optimal solution set and perform crossover operations with the two offspring generated in step S3.61 to obtain a total of four offspring. This operation aims to introduce the superior genes from the Pareto optimal solution set into the offspring to accelerate convergence. S3.63. Perform the same mutation operation as in step S3.3 on the four offspring generated in step S3.62 to obtain the final new offspring.

[0012] Furthermore, in step S3.61 above, the crossover operation is applied to the mixed encoding: for the coordinate part, simulated binary crossover is used; for the task scheduling part, uniform crossover is used; after crossover, two offspring individuals are generated.

[0013] Due to the adoption of the technical solution described above, the present invention has the following advantages: This invention presents a joint optimization method for UAV location deployment, task scheduling, and resource allocation. By constructing a dynamic environment detection mechanism, it monitors system load fluctuations in real time and adaptively triggers the optimization process. This enables the system to respond quickly to dynamic scenarios such as changes in task distribution and sudden changes in computational demands, maintaining a consistently high-efficiency operating state and significantly enhancing the system's robustness and adaptability in complex environments. A unified hybrid coding strategy is employed to integrate UAV location deployment and task scheduling decisions within the same optimization framework for co-evolution. Resources are allocated based on task demand ratios, achieving global joint optimization of location deployment, task scheduling, and resource allocation. This unified hybrid coding strategy effectively avoids the local optimum problem caused by variable fragmentation in traditional step-by-step optimization methods, ensuring the global optimality of the overall solution under the dual objectives of latency and energy consumption. Pareto optimality is introduced during the evolutionary process. The crossover mechanism guided by the optimal solution set enables the rapid propagation of superior genes, significantly improving convergence speed while maintaining population diversity and meeting the stringent real-time requirements of edge computing scenarios. With total task latency and UAV flight energy consumption as dual optimization objectives, a multi-objective evolutionary algorithm generates a Pareto optimal solution set, providing decision-makers with multiple non-dominant equilibrium solutions. In practical applications, it can flexibly select between low-latency-first or low-energy-consumption-first adaptation strategies based on business needs, without the need for preset weights, greatly improving the system's flexibility and applicability in different scenarios. A resource allocation method based on task demand ratios ensures that each task on the same UAV receives bandwidth and computing resources matching its own needs, giving all tasks the same transmission and computation latency, achieving optimal average latency. This ensures fairness in task processing while avoiding resource idleness and waste.

[0014] This invention provides a joint optimization method for UAV location deployment, task scheduling, and resource allocation. Through dynamic adaptive mechanisms, global joint optimization, fast convergence strategies, multi-objective balanced decision-making, and fair and efficient resource allocation, it provides efficient, balanced, and flexible decision support for UAV edge computing systems in dynamic and ever-changing scenarios. Attached Figure Description

[0015] Figure 1 This is a flowchart of the joint optimization method for UAV location deployment, task scheduling, and resource allocation according to the present invention; Figure 2 This is a flowchart illustrating the real-time joint optimization of UAV location deployment, task scheduling, and resource allocation based on a dynamic multi-objective evolutionary algorithm, as described in this invention. Detailed Implementation

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can fully understand and implement the present invention.

[0017] Example 1: As Figure 1 , 2 As shown, a joint optimization method for UAV location deployment, task scheduling, and resource allocation includes the following steps: S1. Unfold the coordinate codes of all UAVs sequentially and concatenate them with the codes of all task scheduling in order to form a one-dimensional vector; the encoding method is as follows: Assuming there is a total The deployment location of each drone is represented by two-dimensional coordinates. express, Coordinate values ​​are represented using real numbers or integers (e.g., real numbers for continuous space, integers for discrete grids), and are ordered by UAV number. Arranged sequentially, forming a coordinate sub-vector with a length of ; The task scheduling section has a total of There are 1 task, and each task corresponds to a binary decision variable. , A value of 1 indicates that the task is processed by the drone base station that received the request, while a value of 0 indicates that the task is offloaded to the cloud computing platform for processing. All binary variables are arranged in order of task number to obtain the scheduling sub-vector, with a length of [value missing]. ; Concatenate the coordinate subvector with the scheduling subvector, forming a vector of length [length missing]. A one-dimensional vector is used as the genotype of an individual in the evolutionary algorithm; this encoding method facilitates subsequent decoding and evolutionary operations. The coordinates and binary variables adopt a hybrid encoding strategy, which needs to be handled separately in crossover and mutation operations. S2. The fitness function is used to evaluate the quality of each one-dimensional vector in step S1, which guides the search direction of the evolutionary algorithm; the steps are as follows: First, the one-dimensional vector is decoded into UAV deployment locations and task scheduling schemes. The decoding process is the inverse operation of encoding: based on the number of UAVs... and number of tasks Extract the first... from the starting position of the one-dimensional vector sequentially. These values ​​correspond in order. This allows for the recovery of the coordinates of each drone; the remaining... The numerical values ​​constitute the task scheduling subvector, where each value is a binary variable (0 or 1), ordered by task number. Assign values ​​to the corresponding tasks respectively; Then, based on the deployment location and task scheduling scheme obtained from the decoding, the performance indicators of the entire system are calculated, including total task latency and UAV flight energy consumption, as two optimization objectives; the operation is as follows: Based on task scheduling variables The tasks are divided into two categories: a set of tasks to be handled by drones. and sets of tasks processed in the cloud For each drone Define the set of tasks it processes. ; Calculate the proportion of resources allocated to each task on the drone: Let the drone... There are a total of The first task, the... The bandwidth ratio obtained by each task on this drone and computing resource ratio The allocations will be made according to the following formulas to ensure that the drones are assigned accordingly. The tasks on each platform have the same data transmission latency and computation latency: (1) (2) in, and The first The data transfer and computational requirements of each task are allocated in a way that ensures each task receives resources proportional to its needs, thus maximizing the efficiency of all tasks performed by the drone. The tasks on the same platform have the same data transmission latency and computation latency to achieve optimal average latency; For the first Data transfer volume per task; For the first The computational requirements of each task; Based on the above ratio, drones Total latency of all tasks for: (3) in, and drones Total bandwidth resources and total computing resources; The total transmission latency for all tasks; Total computation delay; For tasks processed in the cloud, all cloud tasks share the cloud's bandwidth and computing resources, and the total number of cloud tasks... Total latency of all cloud tasks for: (4) in, and These refer to the bandwidth resources and computing resources of the cloud computing platform, respectively. Therefore, the total task latency of the entire system for: (5) Drone flight energy consumption Calculated by the sum of the energy required for all drones to move from their initial positions to their new positions: (6) in, For drones The initial deployment location; The new, optimized location; Energy consumption per unit distance traveled; At this point, each individual corresponds to a solution, and its merits are determined by two objective values. Measure the total task latency and drone flight energy consumption Smaller is always better; S3. Real-time joint optimization of UAV location deployment, task scheduling, and resource allocation based on dynamic multi-objective evolutionary algorithm; constructing an adaptive evolutionary optimization framework for the dynamic characteristics of UAV edge computing systems, specifically including the following sub-steps: S3.1, Randomly initialize a... A population consisting of individuals, each individual's genotype is generated as described in step S1: the coordinates are randomly generated within the feasible deployment area, and the task scheduling part is... Randomly select 0 or 1; evaluate the fitness of each individual, that is, calculate its two objective function values ​​according to formulas (5) and (6) in step S2: total task delay and drone flight energy consumption A fast non-dominated sort is performed on the individuals in the population based on fitness. This sort divides the individuals into several front layers with different dominance levels. Individuals in the higher front layers are all dominant (better) than individuals in the lower front layers. The individuals in the highest front layer of the sort are selected to form the Pareto optimal solution set (PS), and the fitness set corresponding to it is taken as the Pareto front (PF). S3.2 The algorithm periodically detects whether the service status of the current UAV edge computing system has changed. The detection method is as follows: First, calculate the average latency of each UAV's processing task at the current moment; then, calculate the coefficient of variation of the average latency of all UAVs. ,in, The standard deviation of the average latency of each drone. The mean of the average latency for all drones; finally, the coefficient of variation. With preset threshold If a comparison is made, If the environment changes, the subsequent evolutionary process is triggered; otherwise, periodic checks continue. S3.3 When an environmental change is detected, individuals are randomly selected from the current population according to a preset ratio, and mutation operations are performed on each selected individual to introduce diversity to adapt to the environmental change; for real numbers in the coordinate part, polynomial mutation is used; for binary bits in the task scheduling part, bit-flip mutation is used.

[0018] S3.4, Evolution iteration count Initialize to 0; continue evolutionary iterations to optimize the population, with the number of iterations continuing until the preset maximum value is reached. ; S3.5 After the iteration is completed, the algorithm outputs the latest Pareto optimal solution set (PS) and Pareto front (PF) found so far, providing a basis for decision-making to dynamically adjust UAV deployment, task scheduling and resource allocation; S3.6 In each generation of evolution, generate four offspring for each individual in the current population, as follows: S3.61. Randomly select an individual from the current population that is different from the current individual, and perform a crossover operation between the two individuals; the crossover operation is for mixed encoding: for the coordinate part, simulated binary crossover is used; for the task scheduling part, uniform crossover is used; after the crossover, two offspring individuals are generated. S3.62. Randomly select an individual (representing a member of the current Pareto optimal solution set (PS)) from the current Pareto optimal solution set (PS) and perform crossover operations with the two offspring generated in step S3.61 (the crossover operation method is the same as in step S3.61) to obtain a total of four offspring; this operation aims to introduce the superior genes in the Pareto optimal solution set (PS) into the offspring to accelerate convergence; S3.63. Perform the same mutation operation as in step S3.3 on the four offspring generated in step S3.62 to obtain the final new offspring; S3.7. Evaluate the fitness of each offspring generated in step S3.6, i.e., calculate its total task delay according to formulas (5) and (6). and drone flight energy consumption ; S3.8. Based on the fitness obtained in step S3.7, perform a fast non-dominated sort on all offspring and all individuals in the current population; update the Pareto optimal solution set (PS) to the set of individuals at the highest front layer in the sort, and update the Pareto front (PF) to its corresponding fitness set; select individuals from the top of the sort that are the same number as the original population, i.e., the front... Each element is updated to a new generation population based on non-dominated sorting and crowding distance; the number of iterations is [number missing]. Return to step S3.6 until the desired result is reached. .

[0019] Through the aforementioned dynamic multi-objective evolutionary algorithm, the UAV edge computing system can respond to environmental changes in real time, continuously optimize UAV location deployment, task scheduling strategies, and resource allocation ratios, and achieve a balance between minimizing total task latency and UAV flight energy consumption.

[0020] Example 2: The present invention also provides a computer-readable storage medium for storing one or more programs, the one or more programs including instructions, which, when executed by the electronic device of the present application, cause the electronic device to perform the above-described joint optimization method for UAV location deployment, task scheduling and resource allocation.

[0021] Example 3: The present invention also provides an electronic device, including: a processor and a memory; wherein, the memory is used to store one or more programs, the one or more programs including computer execution instructions, and when the electronic device is running, the processor executes the computer execution instructions stored in the memory to enable the electronic device to execute the above-mentioned joint optimization method for UAV location deployment, task scheduling and resource allocation.

[0022] Example 4: The present invention also provides a computer program product containing instructions, which, when executed on a computer, cause the electronic device of the present application to perform the above-described joint optimization method for UAV location deployment, task scheduling and resource allocation.

[0023] The scope of protection of this invention is not limited to the specific embodiments described above. Any changes made based on the core principles of this invention, including but not limited to equivalent substitutions of technical solutions and structural improvements, should be considered to fall within the scope of protection of the claims of this invention. Various modifications and adjustments made to the implementation schemes by those skilled in the art without departing from the design concept of this invention also fall within the scope of protection of this invention.

Claims

1. A joint optimization method for UAV location deployment, task scheduling, and resource allocation, characterized by: It includes the following steps: S1. Unfold the coordinate codes of all UAVs in sequence and concatenate them with the codes of all task scheduling in order to form a one-dimensional vector; S2. Calculate the total mission delay and UAV flight energy consumption using the fitness function, and evaluate the merits of each one-dimensional vector in step S1. S3. Based on the one-dimensional vector obtained in step S1, the total task delay and UAV flight energy consumption obtained in step S2, perform real-time joint optimization of UAV location deployment, task scheduling and resource allocation based on dynamic multi-objective evolutionary algorithm.

2. The joint optimization method for UAV location deployment, task scheduling, and resource allocation according to claim 1, characterized in that: In step S1, the encoding method is as follows: Assuming there is a total The deployment location of each drone is represented by two-dimensional coordinates. express, Coordinate values ​​are represented by real numbers or integers, and are ordered according to the UAV number sequence. Arranged sequentially, forming a coordinate sub-vector with a length of ; The task scheduling section has a total of There are 1 task, and each task corresponds to a binary decision variable. , Decision variables A value of 1 indicates that the task is processed by the drone base station that received the request, while a value of 0 indicates that the task is offloaded to the cloud computing platform for processing. Arrange all binary variables in order of task number to obtain the scheduling subvector with a length of [length missing]. ; Concatenate the coordinate subvector with the scheduling subvector, forming a vector of length [length missing]. A one-dimensional vector.

3. The joint optimization method for UAV location deployment, task scheduling, and resource allocation according to claim 2, characterized in that: Step S2 includes: First, the one-dimensional vector is decoded into drone deployment locations and task scheduling schemes: based on the number of drones. and number of tasks Extract the first... from the starting position of the one-dimensional vector sequentially. These values ​​correspond in sequence. Restore the coordinates of each drone; remaining The given values ​​form a task scheduling sub-vector, where each value is a binary variable ordered by task number. Assign values ​​to the corresponding tasks respectively; Then, based on the deployment location and task scheduling scheme obtained from the decoding, the performance indicators of the entire system are calculated, including total task latency and UAV flight energy consumption; the operation is as follows: Based on task scheduling variables The tasks are divided into two categories: a set of tasks to be handled by drones. and sets of tasks processed in the cloud Define each drone Set of tasks to be processed ; Set up drones There are a total of The first task, the... The bandwidth ratio obtained by each task on this drone and computing resource ratio The allocations will be made according to the following formulas to ensure that the drones are assigned accordingly. The tasks on each platform have the same data transmission latency and computation latency: (1) (2) in, and The first The amount of data transmitted and the amount of computation required for each task; For the first Data transfer volume per task; For the first The computational requirements of each task; Based on the above ratio, drones Total latency of all tasks for: (3) in, and drones Total bandwidth resources and total computing resources; The total transmission latency for all tasks; Total computation delay; For tasks processed in the cloud, all cloud tasks share the cloud's bandwidth and computing resources, and the total number of cloud tasks... Total latency of all cloud tasks for: (4) in, and These refer to the bandwidth resources and computing resources of the cloud computing platform, respectively. Total task latency of the entire system for: (5) Drone flight energy consumption Calculated by the sum of the energy required for all drones to move from their initial positions to their new positions: (6) in, For drones The initial deployment location; The new, optimized location; Energy consumption per unit distance traveled.

4. The joint optimization method for UAV location deployment, task scheduling, and resource allocation according to claim 3, characterized in that: Step S3 includes the following sub-steps: S3.1, Randomly initialize a... A population of individuals, each with a genotype represented by a one-dimensional vector obtained in step S1; evaluate the fitness of each individual: total task latency. and drone flight energy consumption Based on fitness, individuals in the population are sorted quickly without being dominated. This sorting divides individuals into several front layers with different levels of dominance. Individuals in the higher front layers dominate individuals in the lower front layers. The highest-level front individuals in the sorting are selected to form the Pareto optimal solution set, and their corresponding fitness set is taken as the Pareto front. S3.2 The algorithm periodically detects whether the service status of the current UAV edge computing system has changed. The detection method is as follows: First, calculate the average latency of each UAV's processing task at the current moment; then, calculate the coefficient of variation of the average latency of all UAVs. ,in, The standard deviation of the average latency of each drone. The mean of the average latency across all drones is given; finally, this coefficient of variation is calculated. With preset threshold If a comparison is made, If the environment changes, the subsequent evolutionary process is triggered; otherwise, periodic checks continue. S3.3 When an environmental change is detected, individuals are randomly selected from the current population according to a preset ratio, and mutation operations are performed on each selected individual. S3.4, Evolution iteration count Initialize to 0; continue evolutionary iterations to optimize the population, with the number of iterations continuing until the preset maximum value is reached. ; S3.5 After the iteration is completed, the algorithm outputs the latest Pareto optimal solution set and Pareto front, providing a basis for decision-making to dynamically adjust UAV deployment, task scheduling and resource allocation; S3.6 In each generation of evolution, generate four offspring for each individual in the current population; S3.

7. Evaluate the fitness of each offspring generated in step S3.6; S3.

8. Based on the fitness obtained in step S3.7, perform a fast non-dominated sort on all offspring and all individuals in the current population; update the Pareto optimal solution set to the set of individuals at the highest front layer of the sort, and update the Pareto front to its corresponding fitness set; select individuals from the top of the sort with the same number as the original population, and update them into a new generation population based on the non-dominated sort and crowding distance; iteration count: [number missing]. Return to step S3.6 until the desired result is reached. .

5. The joint optimization method for UAV location deployment, task scheduling, and resource allocation according to claim 4, characterized in that: In step S3.3, the mutation operation includes: for the real numbers in the coordinate part, a polynomial mutation is used; for the binary bits in the task scheduling part, a bit-flip mutation is used.

6. The joint optimization method for UAV location deployment, task scheduling, and resource allocation according to claim 4, characterized in that: Step S3.6 is performed as follows: S3.

61. Randomly select an individual from the current population that is different from the current individual, and perform cross-operation between the two; S3.

62. Randomly select an individual from the current Pareto optimal solution set and perform crossover operations with the two offspring generated in step S3.61 to obtain a total of four offspring. This operation aims to introduce the superior genes from the Pareto optimal solution set into the offspring to accelerate convergence. S3.

63. Perform the same mutation operation as in step S3.3 on the four offspring generated in step S3.62 to obtain the final new offspring.

7. The joint optimization method for UAV location deployment, task scheduling, and resource allocation according to claim 6, characterized in that: In step S3.61, the crossover operation is applied to the mixed encoding: for the coordinate part, simulated binary crossover is used; for the task scheduling part, uniform crossover is used; after crossover, two offspring individuals are generated.

8. A computer-readable storage medium for storing one or more programs, characterized in that: The one or more programs include instructions that, when executed by an electronic device, cause the electronic device to perform the joint optimization method for UAV location deployment, task scheduling, and resource allocation as described in claims 1 to 7.

9. An electronic device, characterized in that: include: A processor and a memory; wherein the memory is used to store one or more programs, the one or more programs including computer execution instructions, and when the electronic device is running, the processor executes the computer execution instructions stored in the memory to cause the electronic device to perform the joint optimization method for UAV location deployment, task scheduling and resource allocation as described in claims 1 to 7.

10. A computer program product containing instructions, characterized in that: When the instruction is executed on a computer, it causes the electronic device to perform the joint optimization method for UAV location deployment, task scheduling, and resource allocation as described in claims 1 to 7.

Images