Unmanned aerial vehicle remote joint defense task scheduling method based on multi-objective optimization

Abstract

The invention discloses an unmanned aerial vehicle remote joint defense task scheduling method based on multi-objective optimization. The method comprises the following steps: S1, task graph construction and parameter initialization; s2, calculating the system temperature; s3, initializing path search; s4, generating path disturbance; s5, perturbation path acceptance judgment; s6, evaluating a fitness function; s7, convergence judgment and local guide state updating; and S8, outputting a final scheduling result. According to the method, the problems of many path conflicts, high energy consumption and slow response of the unmanned aerial vehicle in a complex task environment are effectively solved, the probability of communication interference and task overlapping is remarkably reduced, good convergence stability and scheduling performability are achieved, and the method can be widely applied to various multi-vehicle cooperative task scenes such as forest fire prevention, emergency monitoring and urban inspection.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) joint defense, and particularly to a UAV remote joint defense task scheduling method based on multi-objective optimization. Background Art

[0002] With the wide application of multi-UAV systems in scenarios such as public safety, military reconnaissance, environmental monitoring, and emergency response, how to achieve efficient, collaborative, and dynamic UAV task scheduling has become a research and engineering application hotspot in this field. Especially in remote joint defense tasks facing large ranges and complex terrains, there are many tasks, strong task correlations, resource limitations, and high real-time requirements, making it difficult for traditional single-objective scheduling models to meet practical needs. Therefore, researching multi-objective task scheduling methods with intelligent optimization capabilities has become a key technical path to promote the development of the intelligent collaboration capabilities of multi-UAV systems.

[0003] In the prior art, the multi-UAV task scheduling problem is usually modeled as a combinatorial optimization problem, and the main objectives include minimizing the total flight energy consumption, minimizing the response delay, maximizing the task completion rate, or minimizing the communication conflict probability, etc. Since the problem itself is a typical NP-hard problem, it is difficult for traditional exact algorithms to solve high-dimensional task scenarios within an acceptable time. Therefore, in recent years, researchers have widely adopted heuristic or meta-heuristic algorithms for solving, especially genetic algorithms, particle swarm optimization algorithms, artificial bee colony algorithms, simulated annealing algorithms, and wolf pack optimization algorithms, etc., which have shown certain practicality in scheduling optimization.

[0004] Among them, the simulated annealing algorithm is applied to problems such as UAV path planning and task allocation due to its global search ability and strong convergence control ability. This method simulates the physical annealing process of solids, and controls the acceptance probability of perturbed solutions through temperature, so as to achieve the purpose of jumping out of local optimal solutions. However, the classical simulated annealing method has problems such as a single perturbation strategy, a rigid acceptance mechanism, and a difficult-to-adjust convergence process, and it is difficult to adapt to multi-UAV collaborative scheduling scenarios with complex tasks and sensitive priorities. At the same time, the annealing process often lacks spatial information guidance, and is prone to defects such as non-directional path perturbations and low efficiency.

[0005] The wolf pack optimization algorithm simulates the mechanism of group cooperation and information sharing in the foraging process of wolf packs, and has good local guidance ability and the ability to utilize individual difference information, and is suitable for solving optimization problems such as path search. Especially in UAV scheduling, the wolf pack algorithm can select the path search direction through the guiding wolf mechanism, improving the search efficiency and path quality. However, the classical wolf pack algorithm also has problems such as too strong search locality, being easily trapped in local extrema, lack of population diversity, etc., and has limited means in multi-objective trade-off processing, and cannot achieve effective dynamic balance between different objectives.

[0006] On the other hand, in existing task scheduling systems, most research only focuses on the core optimization process of algorithms, but pays insufficient attention to the "scheduling result output" level. In common systems, the final scheduling path is mostly a simplified task number sequence, lacking structured information such as task response time, path distance segmentation, and remaining energy status, making it difficult to directly deploy and execute the scheduling results and difficult to use them for subsequent system evaluation or task re-optimization. In addition, there are redundant situations such as path overlap and task repeated execution in the scheduling results of multi-UAV systems. Existing methods lack a deduplication and screening mechanism in the scheduling output stage, which is likely to lead to resource waste or execution conflicts.

[0007] Existing scheduling systems also lack a unified and executable scheduling output structure organization strategy, mostly staying in the process mainly of "obtaining the solution by optimization → directly outputting", without reflecting structure optimization strategies such as task priority, task dependencies, and path conflict avoidance in the output results. This makes it difficult for the scheduling system to be deployed and implemented in complex systems although it can generate relatively optimal paths, restricting the engineering practical value of the algorithm.

[0008] To sum up, existing multi-UAV scheduling technologies have problems such as single algorithms, weak target coordination, and inflexible disturbance control at the method level, and problems such as unclear result output structure, lack of execution controllability and deployment friendliness at the system integration level. Therefore, there is an urgent need to propose a task scheduling method with a path search guidance mechanism, a dynamic disturbance acceptance strategy, a multi-dimensional fitness evaluation ability, and a structured output ability to achieve multi-objective collaborative optimization of multi-UAVs in the remote joint defense scenario and improve the system closed-loop nature of the scheduling process and the executability of the scheduling results.

[0009] The present invention precisely aims to solve the above problems, and proposes a multi-objective UAV task scheduling method combining a simulated annealing mechanism and an improved wolf pack optimization strategy. By introducing various disturbance methods, an annealing acceptance function based on the logistic distribution, a task graph-driven temperature modeling method, and a sub-strategy scheduling result output structure, an intelligent joint defense scheduling full-process system covering "path generation - disturbance control - acceptance judgment - state update - scheduling output" is constructed, significantly improving the usability, robustness, and engineering implementation ability of the scheduling scheme.

[0010] Therefore, how to provide a UAV remote joint defense task scheduling method based on multi-objective optimization is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0011] An object of the present invention is to propose a method for remotely coordinated defense mission scheduling of unmanned aerial vehicles based on multi-objective optimization. The present invention integrates a dynamic annealing control strategy, a local guided path search mechanism, a multi-objective fitness function evaluation system, and a sub-strategy structured scheduling output scheme, and systematically constructs the entire process of unmanned aerial vehicle cluster mission scheduling covering path generation, perturbation optimization, acceptance judgment, state update, and result output. This method has the advantages of high task coverage rate, high energy utilization rate, few path conflicts, and strong executability of the scheduling result, and is particularly suitable for intelligent cooperative scheduling of multiple unmanned aerial vehicles to perform high-density and high-priority coordinated defense missions in complex geographical areas.

[0012] The method for remotely coordinated defense mission scheduling of unmanned aerial vehicles based on multi-objective optimization according to an embodiment of the present invention includes the following steps: S1. Construct a task graph and form a task set according to standard classification; S2. Collect the current position, remaining energy, communication ability, and task load information of each unmanned aerial vehicle, construct a scheduling search space, and set the initial system temperature to control the perturbation acceptance probability, where the temperature is an adaptive parameter in the annealing algorithm; S3. Map each unmanned aerial vehicle to an individual wolf, divide local cooperation areas according to the task geographical distance, state similarity, and communication topology, dynamically select local guided wolves in each area, and perform search, summoning, and siege operations respectively to generate an initial path population; S4. In each local area, perturb the individual paths, including task order swapping, insertion, or replacement operations, to generate candidate path solutions; S5. Adopt the simulated annealing algorithm, adjust the system temperature according to the variance of the current path fitness change, calculate the acceptance probability based on the perturbation path fitness difference and the updated temperature, and perform the simulated annealing path acceptance judgment; S6. Calculate the fitness values of all unmanned aerial vehicle paths, and the fitness function combines task completion rate, total system energy consumption, response delay, and communication interference indicators for weighted evaluation; S7. Select the path with the optimal fitness from the current path population as the global optimal solution, and record the local guided wolves and individual states to guide the next round of path search and perturbation adjustment; S8. Judge whether the system temperature has dropped to the termination threshold or the population fitness has converged. If the condition is met, output the final scheduling scheme for each unmanned aerial vehicle.

[0013] Optionally, S1 specifically includes: extracting scheduling classification criteria from historical task execution data and monitoring results, and defining classification rules for high-priority tasks, medium-priority tasks, and low-priority tasks based on parameters such as task location distribution density, task time sensitivity, resource consumption intensity, and task path correlation. Among them: low-priority tasks are tasks with a resource consumption intensity ≤ 70% of the system average and a time sensitivity ≤ 0.3; medium-priority tasks are tasks with a resource consumption intensity between 70% and 120% of the system average or a time sensitivity between 0.3 and 0.7; high-priority tasks are tasks with a resource consumption intensity > 120% of the system average or a time sensitivity > 0.7. Based on this classification result, construct a task graph and label the scheduling attribute parameters of each task node as the input basis for the scheduling search space.

[0014] Optionally, S2 specifically includes: establishing a UAV cluster set, obtaining the current position coordinates, remaining energy, communication capabilities, and task load capabilities of each UAV. Among them, the current position coordinates are used to calculate the spatial distance relationship between the UAV and the task node, the remaining energy is used to determine the flight range that can be supported within the current scheduling cycle, the communication capabilities are used to define the information interaction area that it can participate in, and the task load capabilities are used to constrain the maximum task resources that can be carried during the scheduling process. Jointly analyze the UAV state parameters and the constructed task graph, and filter out task nodes that meet the scheduling conditions based on spatial reachability, energy consumption matching, and task priority thresholds. Establish an initial mapping relationship between the UAV and the task node, set the initial temperature of the system, and jointly form the scheduling search space with the UAV state parameters, the set of executable tasks, and the initial temperature.

[0015] Optionally, S3 specifically includes: S31. Map each UAV to a search individual, and initialize the task path vector. The path is composed of the serial numbers of the schedulable task nodes in the execution order. S32. Calculate the correlation score between task nodes , which is weighted and constructed based on the geographical distance, priority difference, and resource requirement difference between tasks. The scoring function is: ; Among them, , , is the weight coefficient, satisfying , is the task node and in the Euclidean distance on the two-dimensional map, is the maximum distance between any two nodes in the current task graph, and are respectively the task nodes The task priority value of is the highest priority value in the current task graph. The task priority value is affected by task classification. , are the resource demand values of task nodes and respectively, and is the maximum resource demand value in the current task graph; S33. Based on the correlation score, divide the task graph into multiple local task regions. Each search individual performs task path search according to the region where the current position belongs, replacing the traditional global search method, and improving the search parallelism and convergence stability; S34. In each region, select the local optimal individual according to the fitness value of the current round as the "local guiding wolf", and the path vector is used as a reference for the path update of other individuals; S35. Replace the traditional siege behavior in the wolf pack optimization algorithm with a "difference-driven update" mechanism. Ordinary individuals adjust their structures according to the path difference from the guiding individual, and the update formula is: ; where represents the current task path vector of the th UAV in the th round, represents the current task path vector of the th UAV in the th round, represents the task path vector of the local guiding individual in the th round, represents the task sequence insertion or sequential exchange operation, represents the path difference extraction, is the update proportionality coefficient; ; where is the task coverage rate function, is the energy consumption function, is the communication conflict degree function, is the delay response function, , , , are the weight coefficients, and the sum is 1; S37. Introduce an improved elimination update mechanism, and calculate the fitness standard deviation of the current path population , if is less than the dynamic threshold, it is recognized that the current population tends to converge, and the individual perturbation operation is executed. A new individual is randomly generated to replace the individual with the worst fitness to prevent premature convergence.

[0016] Optionally, the S4 specifically includes: S41. Select a perturbation target path vector, and the path vector is screened by an improved elimination and update mechanism; S42. Set the path perturbation probability , and perform a structural perturbation operation on the path vector according to the perturbation probability. The specific perturbation methods include: Task exchange perturbation: Randomly select two positions in the path and exchange the corresponding task orders; Task insertion perturbation: Randomly select a task node and insert it into a new position in the path; Subsequence reversal perturbation: Randomly select a continuous subsequence in the path and reverse the task order; Local rearrangement perturbation: Randomly rearrange the task order within a specified subsequence range; Determine the perturbation intensity according to the following perturbation control function: ; Among them, represents the perturbation intensity ratio, is the maximum perturbation ratio factor, is the current system temperature, is the initial system temperature; S43. Generate a candidate path vector after the perturbation , and ensure its path legality, that is, all task nodes are not repeated, the order is reasonable, and the resource load does not exceed the task load capacity to form an effective path solution.

[0017] Optionally, the S5 specifically includes: S51. Let the fitness value be , the perturbed path vector be the fitness value be , and the fitness function be a multi-objective fitness function; S52. Define the current path fitness difference as: ; At the same time, let the current expected acceptance difference of the system be: ; Among them, is the initial acceptance expected difference, is the current scheduling iteration round, is the maximum number of iterations; S53. Calculate the acceptance probability using a perturbation acceptance probability function based on the logistic distribution function. : ; where is the system temperature at the current th iteration, is the soft decision adjustment coefficient used to control the change amplitude of the acceptance boundary; S54. Generate a uniform random number in the interval [0, 1]. If , then accept the perturbed path and update it to: ; otherwise, keep the original path unchanged: ; S55. Update the system temperature using an adaptive temperature decay strategy based on the global perturbation behavior and fitness fluctuation: ; where is the decay adjustment factor, is the standard deviation of the global individual path fitness in the current round, is the th round perturbation path acceptance rate, defined as the ratio of the number of individuals with the current accepted path to the total number of individuals; Finally, output the current task path vector of the th unmanned aerial vehicle in the updated round together with the system temperature as the input for path search and perturbation judgment in subsequent scheduling rounds.

[0018] Optionally, the S6 specifically includes: performing a multi-objective comprehensive evaluation on the scheduling path vector of each unmanned aerial vehicle, and using the following fitness function: ; where represents the task path vector of the th unmanned aerial vehicle, is the objective function weighting coefficient, satisfying , the task coverage rate function represents the ratio of the number of tasks completed by the unmanned aerial vehicle to the total number of schedulable tasks, and the energy consumption function represents the inverse proportional normalized value of the sum of the flight costs between consecutive tasks in the path to the maximum energy capacity, and the communication conflict degree function Indicates the degree of interference caused by communication time or communication radius conflicts between tasks in the path, and the response delay function Indicates the normalized deviation value between the task completion time and the task timeliness requirement, and is weighted and averaged in combination with the task priority. The above fitness function is used for ranking the pros and cons of each UAV path plan during the path search stage, for judging whether the new path is better after path perturbation, as the basis for calculating the perturbation acceptance probability during the simulated annealing process, and as the standard for individual screening and elimination during the path population update process, providing an evaluation basis for the entire process of the UAV task scheduling strategy.

[0019] Optionally, the S7 specifically includes: Let the set of all paths in the current round be , and the fitness set be , where is the number of UAVs. Set the individual with the best fitness in each task area as the local guiding individual in that area for guiding the next round of path search. At the same time, calculate the global optimal path vector of the current round , satisfying: ; Let its fitness be ; Record the difference in fitness of the optimal paths in the previous and current rounds: ; Define as the preset threshold for the difference in fitness of the optimal path, as the preset threshold for the acceptance rate of the perturbed path. If and the acceptance rate of the perturbed path in the current round , or the current round satisfies , then it is determined that the scheduling converges, and the optimal individual path in the path set and its fitness are output as the final scheduling result. Otherwise, accumulate the energy consumption corresponding to the task execution order and the distance between tasks in the current path of each UAV as the path cost of this round, update its remaining energy , and store together, providing input for the next round of path search and perturbation feasibility judgment.

[0020] Optionally, the S8 specifically includes: after meeting the scheduling convergence condition, extracting the final task path, task execution order, start time and end time of each task, flight distance between adjacent task nodes in the path, and remaining energy information after completing the scheduling task of each UAV, structuring and organizing the data to generate a scheduling output record for each UAV, and summarizing the scheduling output records of all UAVs to form a final scheduling plan, which is used to guide the task scheduling and deployment of multiple UAVs, and at the same time provides a basis for path visualization display, task schedule generation, and resource consumption analysis.

[0021] Optionally, the output of the final scheduling plan adopts a sub-strategy structured organization method, and the output strategies include the following three categories: the scheduling result sorting strategy driven by task priority, the path screening strategy under energy safety constraints, and the cluster consistency strategy for task deduplication and fusion. Among them, the task priority driven strategy is used to sort the tasks involved in the scheduling path of each UAV in ascending or descending order according to the priority field of the task node, and give priority to completing key tasks with high response requirements. The energy safety constraint strategy is used to screen and output paths based on the remaining energy threshold of each UAV, and retain the path records that still have the lowest remaining energy margin after completing the current path tasks. The task deduplication and fusion strategy is used to detect duplicate task nodes in the set of all UAV paths, and by cross-verifying the start and end time sequences, communication conflicts, and energy consumption costs of the paths, select the only executor and eliminate redundant records; finally, generate a unified scheduling data set output file according to the above strategies for the scheduling results, and record the task sequence, task schedule, flight segment distance, task coverage rate, and remaining energy of each UAV in the output file.

[0022] The beneficial effects of the present invention are as follows: (1) By integrating and structurally improving the simulated annealing mechanism and the wolf pack optimization algorithm, introducing a logistic function to construct a perturbation acceptance probability model, and dynamically modeling the system temperature in combination with the task graph structure, the present invention realizes an adaptive scheduling convergence control mechanism based on target difference and perturbation history. Compared with traditional annealing or single heuristic algorithms, this method has better search jump-out ability and convergence flexibility, and can effectively improve the scheduling stability and the quality of scheduling results when facing scheduling problems with high task priority, resource tension, and complex paths.

[0023] (2) The present invention designs a multi-objective weighted fitness function, covering core indicators such as task coverage rate, energy consumption, communication conflict, and response delay, clarifies the evaluation methods and weight adjustment mechanisms of each sub-objective, and combines a local guidance mechanism to drive the path search direction, making the path generation process have a clear optimization goal orientation and constraint satisfaction ability, and significantly improving the rationality of UAV task allocation, path diversity, and goal coordination of the scheduling process.

[0024] (3) In the scheduling result output stage, the present invention introduces a multi-strategy structured output mechanism such as task priority-driven, energy threshold screening, and duplicate task fusion, which not only optimizes the structural organization of the scheduling scheme, but also ensures the feasibility, safety, and system deployment friendliness of task execution. The scheduling output result has the capabilities of direct executability, visual display, and re-optimization support, which helps to achieve the closed-loop operation of the scheduling system and the engineering of deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings: Figure 1 is a flowchart of the method for remotely coordinated defense task scheduling of unmanned aerial vehicles based on multi-objective optimization proposed by the present invention; Figure 2 is a schematic diagram of the individual search and local guidance mechanism of the improved wolf pack optimization algorithm proposed by the present invention; Figure 3 is a structural block diagram of the logistic distribution type perturbation acceptance probability function and the dynamic temperature annealing mechanism proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0027] Refer to Figures 1-3 , the method for remotely coordinated defense task scheduling of unmanned aerial vehicles based on multi-objective optimization includes the following steps: S1. Construct a task graph, and form a task set according to standard classification; S2. Collect the current position, remaining energy, communication ability, and task load information of each unmanned aerial vehicle, construct a scheduling search space, and set the initial temperature of the system to control the perturbation acceptance probability. The temperature is an adaptive parameter in the annealing algorithm; S3. Map each unmanned aerial vehicle to an individual wolf, divide the local cooperation area according to the task geographical distance, state similarity, and communication topology, and dynamically select a local guiding wolf in each area to perform search, summon, and siege operations respectively to generate an initial path population; S4. In each local area, perturb the individual path, including task order swapping, insertion, or replacement operations, to generate candidate path solutions; S5. Adopt the simulated annealing algorithm, adjust the system temperature according to the variance of the current path fitness change, calculate the acceptance probability based on the perturbation path fitness difference and the updated temperature, and perform the simulated annealing path acceptance judgment; S6. Calculate the fitness values for all UAV paths. The fitness function combines the task completion rate, total system energy consumption, response delay, and communication interference metrics for weighted evaluation. S7. Select the path with the optimal fitness from the current path population as the global optimal solution, and record the local guiding wolves and individual states to guide the next round of path search and perturbation adjustment. S8. Determine whether the system temperature has dropped to the termination threshold or the population fitness has converged. If the condition is met, output the final scheduling scheme for each UAV.

[0028] Through the construction of a full-process scheduling method integrating task modeling, path search, perturbation optimization, fitness evaluation, annealing control, and scheduling output, the present invention forms a highly modular remote joint defense multi-UAV scheduling scheme with a clear algorithm-data-execution closed-loop. Compared with the existing scheduling methods that only focus on the optimization of a certain stage, the present invention provides a complete system structure, has a clear information flow and feedback mechanism, supports task-level priority control and multi-objective balance of energy consumption and delay, and significantly improves the system-level scheduling performance.

[0029] In this embodiment, the S1 specifically includes: extracting the scheduling classification criteria from the historical task execution data and monitoring results, and defining the classification rules for high-priority tasks, medium-priority tasks, and low-priority tasks according to parameters such as task location distribution density, task time sensitivity, resource consumption intensity, and path correlation between tasks. Among them: low-priority tasks are tasks with a resource consumption intensity ≤ 70% of the system average and a time sensitivity ≤ 0.3; medium-priority tasks are tasks with a resource consumption intensity between 70% and 120% of the system average or a time sensitivity between 0.3 and 0.7; high-priority tasks are tasks with a resource consumption intensity > 120% of the system average or a time sensitivity > 0.7. Construct a task graph based on the classification results and label the scheduling attribute parameters of each task node as the input basis for the scheduling search space.

[0030] By introducing task priority, node connectivity, energy consumption model, and communication constraint information in the task graph initialization stage, the present invention completes a multi-dimensional, task-graph-driven parameter initialization mechanism. Compared with the existing methods that only perform static initialization based on topology or location, the present invention can perceive task heterogeneity and node importance in real time, provide accurate support for subsequent path search and perturbation determination, and improve the adaptability and robustness of the scheduling system to complex constraint scenarios.

[0031] In this embodiment, step S2 specifically includes: establishing a UAV cluster set, obtaining the current position coordinates, remaining energy, communication ability, and task load capacity of each UAV, where the current position coordinates are used to calculate the spatial distance relationship between the UAV and the task node, the remaining energy is used to determine the flight range that can be supported within the current scheduling period, the communication ability is used to define the information interaction area that it can participate in, the task load capacity is used to restrict the maximum amount of task resources that it can carry during the scheduling process, jointly analyzing the UAV state parameters and the constructed task graph, screening out the task nodes that meet the scheduling conditions according to the spatial reachability, energy consumption matching, and task priority threshold, establishing an initial mapping relationship between the UAV and the task node, setting the initial temperature of the system, and jointly forming a scheduling search space with the UAV state parameters, the set of executable tasks, and the initial temperature.

[0032] The present invention innovatively combines the task graph structure characteristics and the task priority distribution characteristics to construct the system initial temperature parameter, realizing the adaptive initialization of the temperature of the simulated annealing algorithm. Different from the traditional method that uses a fixed or empirical initial temperature, the present invention directly binds the temperature to the problem complexity, making the annealing mechanism more semantically consistent and task-aware for scheduling, effectively enhancing the accuracy and dynamic adjustment ability of the perturbation acceptance control, and improving the adaptation effect of the system in different-scale task scenarios.

[0033] In this embodiment, step S3 specifically includes: S31. Map each UAV to a search individual, initialize the task path vector, and the path consists of the serial numbers of the schedulable task nodes in the execution order; S32. Calculate the correlation score between task nodes , which is weighted and constructed according to the geographical distance, priority difference, and resource requirement difference between tasks. The scoring function is: ; where , , are weight coefficients that satisfy , is the Euclidean distance between task nodes and in the two-dimensional map, is the maximum distance between any two nodes in the current task graph, and are respectively the task priority values of task nodes and , is the highest priority value in the current task graph, and the task priority value is affected by task classification, , are respectively the task nodes The resource demand value of and is the maximum resource demand value in the current task graph; S33. Based on the correlation score, divide the task graph into multiple local task regions. Each search individual performs task path search according to the region where the current position belongs, replacing the traditional global search method to improve the search parallelism and convergence stability; S34. In each region, select the local optimal individual as the "local guiding wolf", and the path vector is used as a reference for the path update of other individuals; S35. Replace the traditional siege behavior in the wolf pack optimization algorithm with a "difference-driven update" mechanism. Ordinary individuals adjust their structures according to the path differences from the guiding individuals, and the update formula is: ; where represents the current task path vector of the th UAV in the th round, represents the current task path vector of the th UAV in the th round, represents the task path vector of the local guiding individual in the th round, represents the task sequence insertion or sequential exchange operation, represents the path difference extraction, and is the update ratio coefficient; where is the task coverage function, is the energy consumption function, is the communication conflict degree function, is the delay response function, , , , is the weight coefficient, satisfying the sum of 1; S37. Introduce an improved elimination and update mechanism, calculate the fitness standard deviation of the current path population. If is less than the dynamic threshold, it is recognized that the current population tends to converge, and an individual perturbation operation is performed to randomly generate a new individual to replace the individual with the worst fitness to prevent premature convergence.

[0034] The present invention realizes the structural enhancement of the traditional wolf pack optimization algorithm by introducing a local task domain-guided individual update mechanism, a dynamic path offset function, and a stability adjustment factor. Different from the fixed update mechanism in the existing algorithms, the present invention constructs a guiding closed-loop between local optimum and individual evolution, significantly enhancing the directionality and controllability of path search, and improving the individual diversity and solution set distribution quality of the algorithm in a multi-objective trade-off environment.

[0035] In this embodiment, S4 specifically includes: S41. Select a perturbed target path vector, which is screened by an improved elimination update mechanism; S42. Set a path perturbation probability , and perform a structural perturbation operation on the path vector according to the perturbation probability. The specific perturbation methods include: Task exchange perturbation: Randomly select two positions in the path and exchange the corresponding task orders; Task insertion perturbation: Randomly select a task node and insert it into a new position in the path; Subsequence reversal perturbation: Randomly select a continuous subsequence in the path and reverse the task order; Local rearrangement perturbation: Randomly rearrange the task orders within a specified subsequence range; Determine the perturbation intensity according to the following perturbation control function: ; Wherein, represents the perturbation intensity ratio, is the maximum perturbation ratio factor, is the current system temperature, is the initial system temperature; S43. After the perturbation is completed, generate a candidate path vector , and ensure that it satisfies path legality, that is, all task nodes are not repeated, the order is reasonable, and the resource load does not exceed the task load capacity, forming an effective path solution.

[0036] The present invention designs a variety of path structure-based perturbation operators, introduces a perturbation intensity function to dynamically control the perturbation range, and constructs a controllable, phased, and highly adaptable path perturbation mechanism. Compared with the problems of large randomness and fixed perturbation amplitude in traditional perturbation operations, the present invention can select the perturbation form and intensity by combining historical path changes and the current scheduling state, making the path search have both local refinement ability and global exploration flexibility, and improving the scheduling search efficiency and solution quality.

[0037] In this embodiment, S5 specifically includes: S51. Let the fitness value be , the perturbation path vector is The fitness value is , and the fitness function is a multi-objective fitness function; S52. Define the current path fitness difference as: ; Meanwhile, set the current expected acceptance difference of the system as: ; where, is the initial acceptance expected difference, is the current scheduling iteration round, is the maximum number of iterations; S53. Adopt a perturbation acceptance probability function based on the logistic distribution function to calculate the acceptance probability : ; where, is the system temperature at the current th iteration, is the soft decision adjustment coefficient, which is used to control the change range of the acceptance boundary; S54. Generate a uniform random number that follows the interval [0, 1]. If , then accept the perturbed path and update it to: ; Otherwise, keep the original path unchanged: ; S55. Update the system temperature , and adopt an adaptive temperature decay strategy based on the global perturbation behavior and fitness fluctuation: ; where, is the decay adjustment factor, is the standard deviation of the global individual path fitness in the current round, is the th round perturbation path acceptance rate, which is defined as the ratio of the number of individuals of the currently accepted path to the total number of individuals; Finally, output the current task path vector of the th UAV in the updated round together with the system temperature as the input for path search and perturbation judgment in subsequent scheduling rounds.

[0038] The present invention innovatively introduces the Logistic distribution as the perturbation acceptance probability function, and combines the expected acceptance difference in the current round with the path volatility to dynamically adjust the temperature update strategy. Compared with the traditional exponential function annealing model, the annealing mechanism constructed by the present invention has three mechanisms: "soft boundary", "target guidance" and "adaptive temperature regulation", which can more precisely control the perturbation acceptance behavior, realize the dynamic transition from wide-area search to fine convergence, and greatly improve the convergence quality and search efficiency of the scheduling system.

[0039] In this embodiment, S6 specifically includes: performing a multi-objective comprehensive evaluation on the scheduling path vector of each unmanned aerial vehicle (UAV), and using the following fitness function: ; where represents the task path vector of the th UAV, is the weighted coefficient of the objective function, satisfying , the task coverage rate function represents the ratio of the number of tasks completed by the UAV to the total number of schedulable tasks, the energy consumption function represents the inverse proportional normalization value of the sum of the flight costs between consecutive tasks in the path to the maximum energy capacity, the communication conflict degree function represents the degree of interference caused by communication time or communication radius conflicts between tasks in the path, the response delay function represents the normalized deviation value between the task completion time and the task timeliness requirement, and is weighted and averaged in combination with the task priority. The above fitness function is used for ranking the pros and cons of each UAV path scheme in the path search stage, for judging whether the new path is better after path perturbation, as the basis for calculating the perturbation acceptance probability in the simulated annealing process, and as the standard for individual screening and elimination in the path population update process, providing an evaluation basis for the entire process of the UAV task scheduling strategy.

[0040] The present invention constructs a multi-objective fitness function model with clear structure, independent objectives and adjustable weights, clearly quantifying the task completion rate, energy consumption, communication conflict degree and response delay indicators. Different from the existing multi-objective scheduling methods that only use weighted summation or heuristic scoring, the present invention functions and parameterizes the fitness sub-functions, improving the evaluation accuracy and flexibility, facilitating the adjustment of the scheduling strategy and the tracking of system-level performance, and enhancing the engineering implementation ability of path evaluation.

[0041] In this embodiment, S7 specifically includes: Let the set of all paths in the current round be , and the fitness set be , where Let \(n\) be the number of drones. Set the individual with the optimal fitness in each task area as the local guiding individual for the next round of path search guidance. Meanwhile, calculate the global optimal path vector for the current round. , satisfying: ; Let its fitness be ; Record the difference in fitness between the optimal paths of the previous and current rounds: ; Define as the preset threshold for the difference in fitness of the optimal path, and as the preset threshold for the acceptance rate of the perturbed path. If is satisfied and the acceptance rate of the perturbed path in the current round , or the current round satisfies , then it is determined that the scheduling has converged, and the optimal individual path in the path set and its fitness are output as the final scheduling result; otherwise, the energy consumption corresponding to the task execution order and the distance between tasks in the current path of each drone is accumulated as the path cost for this round, and its remaining energy is updated, and

[0042] are stored together to provide input for the next round of path search and the judgment of the feasibility of perturbation.

[0043] In this embodiment, step S8 specifically includes: after the scheduling convergence condition is satisfied, extract the final task path, task execution order, start time and end time of each task, flight distance between adjacent task nodes in the path, and remaining energy information of each drone after the scheduling task is completed, organize the data in a structured manner, generate a scheduling output record for each drone, and summarize the scheduling output records of all drones to form a final scheduling plan, which is used to guide the task scheduling deployment of multiple drones, and at the same time provide a basis for path visualization display, task schedule generation, and resource consumption analysis.

[0044] The present invention constructs a structured scheduling result output mechanism after scheduling is terminated. The output content covers key scheduling elements such as path sequence, time allocation, flight distance, and energy consumption residual, and supports control execution and result visualization analysis. Unlike existing scheduling schemes that only output task numbers or result sets, the output structure of the present invention is designed for engineering deployment scenarios, with clear field boundaries, timing consistency, and resource constraint annotations, which is conducive to the rapid implementation of the scheduling system and closed-loop control of tasks.

[0045] In this embodiment, the output of the final scheduling plan adopts a strategy-based structured organization method, and the output strategies include the following three categories: a scheduling result sorting strategy driven by task priority, a path screening strategy under energy safety constraints, and a cluster consistency strategy for task deduplication fusion. Among them, the task priority driven strategy is used to arrange the tasks involved in the scheduling path of each drone in ascending or descending order according to the priority field of the task node, and give priority to completing key tasks with high response requirements. The energy safety constraint strategy is used to screen the output path based on the remaining energy threshold of each drone, and retain the path records that still have the lowest remaining energy margin after completing the current path task. The task deduplication fusion strategy is used to detect repeated task nodes in the set of all drone paths, and cross-validate the path start and end timing, communication conflicts and energy consumption costs, and select the only executor and eliminate redundant records. Finally, the scheduling results are generated into a unified scheduling data set output file according to the above strategies, and the output file records the task sequence, task schedule, flight segment distance, task coverage, and remaining energy of each drone.

[0046] The present invention introduces scheduling result organization strategies based on the scheduling output structure, including task priority sorting, residual energy threshold screening, and task de-duplication and merging mechanisms, to build an intelligent scheduling output post-processing module. Compared with the traditional direct result output mode, the present invention improves the systematicness and execution robustness of scheduling output results, reduces the probability of task conflicts and the risk of energy exhaustion, and has the ability to support later expansion, which is an important part of the closed-loop management of the scheduling system.

[0047] Embodiment 1: In order to verify the feasibility and superiority of the present invention in implementation, the present invention is applied to the scenario of "forest fire prevention and emergency monitoring in a certain city". The city is located in the western hilly forest area, with a high forest coverage rate and a long dry period throughout the year, which is prone to fire. In order to improve the efficiency of daily aerial monitoring, the local government plans to deploy a remote joint defense inspection system based on multiple drones to replace traditional manual duty and fixed-point flight methods, and perform three types of tasks: daily regional inspections, high-frequency inspections of key risk points, and high-response processing of abnormal events.

[0048] In this scenario, 8 autonomous flying drones are deployed, all equipped with visual recognition, real-time communication, and medium-range endurance modules. The mission area covers an area of approximately 156 square kilometers, and the total number of missions is 112. The mission priorities are automatically generated based on parameters such as terrain, forest density, and historical fire point frequency, and are divided into three categories: high, medium, and low. There are communication constraints at each mission point, and inspection data needs to be returned within a certain time window. At the same time, there are some dependencies and connection relationships between missions. In addition, the flight resources of the drones are limited, with a maximum flight time of 40 minutes per single flight and a maximum load power of 800 Wh.

[0049] Deploy the scheduling method described in the present invention in this area. First, the task graph construction module performs topological modeling on all tasks and constructs the task graph structure based on priorities and dependency relationships. Then, by initializing the states of each drone and the system temperature, where the temperature is jointly calculated by the number of task nodes, connection density, and priority fluctuation coefficient, and the initial value is 0.91. Subsequently, the system combines local guiding wolves and multi-disturbance operators to generate an initial path and introduces an improved annealing acceptance mechanism to dynamically evaluate path changes. After each round of path evolution, according to the multi-objective fitness function, calculate the performance of each path scheme in four aspects: task completion rate, energy consumption, communication conflict rate, and response delay, and perform convergence judgment and individual state update. When the system meets the criteria that the path perturbation acceptance rate is less than 3% and the path fitness change is less than 0.0025, it automatically converges, and the scheduling result is structurally output and the execution strategy is reorganized.

[0050] To demonstrate the performance advantages of the method of the present invention, comparative experiments are carried out with two typical drone scheduling methods: Method A is a traditional scheduling strategy based on genetic algorithms, and Method B is an improved particle swarm algorithm. Each of the three methods schedules and executes the above task scenario 20 times, and records the average task completion time, total energy consumption, path repetition rate, task conflict rate, and high-priority task response timeliness indicators.

[0051] The results are shown in the following table: Table 1: Comparative performance results of the present invention and other scheduling algorithms in the forest remote joint defense mission ; From the perspective of task completion efficiency, under the same number of tasks and scheduling constraints, the average task completion time of the method of the present invention is 35.2 minutes, which is significantly better than 42.6 minutes of the genetic algorithm-based method and 39.8 minutes of the improved particle swarm algorithm, indicating that it is more efficient in task path generation and drone allocation. Combining the analysis of the task completion rate, the method of the present invention achieves a high completion rate of 98.2%, while the comparative methods A and B are 90.7% and 94.3% respectively, showing that the present invention not only has a shorter scheduling time, but also has a stronger ability to cover tasks, and has a more perfect task global allocation and resource matching mechanism.

[0052] In terms of task response priority, the present invention shows significant advantages in processing high-priority tasks. Its average response time is 12.3 minutes, which is significantly improved compared to 19.5 minutes of Method A and 15.8 minutes of Method B. This result stems from the introduction of a task priority-driven mechanism in the present invention, which fully considers the urgency of high-weight tasks during the path perturbation and output result sorting processes, thereby achieving rapid response and priority allocation for critical tasks.

[0053] In terms of energy consumption control, the task path generated by the present invention only consumes a total flight energy of 590.4 Wh, which is much lower than 648.1 Wh of Method A and 622.6 Wh of Method B. This indicates that the present invention takes into account the rationality of flight distance and task execution order in path optimization, avoiding problems of repeated flight and path winding. In addition, its path repetition rate is only 1.8%, significantly lower than 6.3% of Method A and 4.9% of Method B, reflecting that effective task deduplication and path fusion strategies are introduced in the scheduling output of the present invention, which helps to reduce repeated scheduling and resource waste among drones.

[0054] In terms of communication conflict rate, the present invention controls it at an extremely low level of 0.7%, while traditional methods are generally above 2.8%. This shows that it effectively considers the conflict relationship between task communication time windows and coverage ranges during the scheduling process, improving the operation stability and communication reliability of the entire scheduling system.

[0055] In terms of the algorithm efficiency of the scheduling system itself, the present invention can converge to the optimal solution in an average of only 43 rounds, while Method A and Method B require 68 rounds and 57 rounds respectively. This demonstrates the high coordination of its annealing mechanism and perturbation path acceptance function, which can quickly drive individuals towards the global optimum and reduce the number of ineffective iteration rounds. At the same time, the present invention can maintain an average system remaining battery utilization rate of 83.4% after scheduling, which is much higher than other methods, indicating that it is more resource-saving and has a larger safety margin for the use of drones, providing flexibility and expansion space for subsequent supplementary tasks.

[0056] In summary, by comparing the key indicators in the table, it can be clearly seen that the method of the present invention is superior to existing mainstream scheduling algorithms in multiple dimensions such as task scheduling efficiency, response ability, energy consumption control, path exclusivity, conflict avoidance, and system stability, verifying its significant technical advantages and engineering feasibility in practical applications. This method is particularly suitable for multi-drone collaborative scenarios with limited resources, complex targets, and heterogeneous tasks, and has broad promotion value in applications such as forest fire prevention and control, urban inspection, and emergency response.

[0057] As described above, it is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A method for scheduling UAV remote joint defense tasks based on multi-objective optimization, characterized in that: The steps include: S1. Construct a task map according to standard classification; S2, collect data information parameters, and jointly analyze the drone status parameters and the mission map; S3, mapping each drone to an individual wolf, selecting a local guide wolf, and performing path search and difference-driven update mechanism operations respectively to generate an initial path group; S4. In each local area, perturb the individual paths to generate candidate path solutions; S5, using the simulated annealing algorithm, adjusting the system temperature according to the variance of the current path fitness change, calculating the acceptance probability based on the perturbed path fitness difference and the updated temperature, and performing the simulated annealing path acceptance judgment; S6. Calculate the fitness value for all UAV paths. The fitness function combines the task completion rate, system total energy consumption, response delay and communication interference indicators for weighted evaluation. S7, select the path with the best fitness from the current path group as the global optimal solution, record the local guiding wolf and individual status, and guide the next round of path search and disturbance adjustment; S8: Determine whether the system temperature drops to the termination threshold or the group fitness converges. If the conditions are met, output the final scheduling plan for each drone.

2. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: The S1 specifically includes: extracting scheduling classification standards from historical task execution data and monitoring results, and defining classification rules for high-priority tasks, medium-priority tasks, and low-priority tasks based on task location distribution density, task time sensitivity, resource consumption intensity, and path correlation parameters between tasks, wherein: a low-priority task is a task with a resource consumption intensity ≤ 70% of the system average and a time sensitivity ≤ 0.3; a medium-priority task is a task with a resource consumption intensity between 70% and 120% of the system average, or a time sensitivity between 0.3 and 0.7; a high-priority task is a task with a resource consumption intensity > 120% of the system average or a time sensitivity > 0.7, and constructing a task graph based on the classification results.

3. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: The S2 specifically includes: establishing a drone cluster set, obtaining the current position coordinates, remaining energy, communication capability and task load capacity of each drone, wherein the spatial distance relationship between the drone and the task node is calculated based on the current position coordinates, the flight range that can be supported in the current scheduling cycle is determined based on the remaining energy, the information interaction area in which it can participate is limited based on the communication capability, and the maximum amount of task resources that can be carried during the scheduling process is constrained based on the task load capacity. The drone state parameters are jointly analyzed with the constructed task graph, and the task nodes that meet the scheduling conditions are screened out based on spatial accessibility, energy consumption matching and task priority thresholds, an initial mapping relationship between the drone and the task node is established, and the initial temperature of the system in simulated annealing is set, and the drone state parameters, the executable task set and the initial temperature together form a scheduling search space.

4. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: The S3 specifically includes: S31, mapping each UAV as a search individual, initializing the task path vector, the path is composed of the schedulable task node numbers in the execution order; S32. Calculate the correlation scores between task nodes , based on the weighted construction of geographical distance, priority difference and resource requirement difference between tasks, the scoring function is: ； in, , , is the weight coefficient, satisfying , For task nodes and Euclidean distance in a two-dimensional map, is the maximum distance between any two nodes in the current task graph, and Task nodes and The task priority value, It is the highest priority value in the current task graph. The task priority value is affected by the task classification. , Task nodes and The resource demand value, is the maximum resource requirement value in the current task graph; S33, based on the relevance score, the task graph is divided into multiple local task areas, and each search individual performs a task path search according to the area to which the current position belongs, replacing the traditional global search method; S34. In each region, select the local optimal individual according to the current round fitness value. As a local guide wolf, the path vector is used as a reference for path updates of other individuals; S35. Replace the traditional siege behavior in the wolf pack optimization algorithm with the "difference-driven update" mechanism. Ordinary individuals make structural adjustments based on the path differences with the guiding individuals. The update formula is: ； in, Indicates Round The current mission path vector of the UAV, Indicates Round The current mission path vector of the UAV, Indicates The task path vector of the local guiding individual in the round, Indicates a task sequence insertion or sequence exchange operation. represents path difference extraction, To update the scale factor; S36. For each path Calculate the multi-objective fitness function, which is as follows: ； in, is the task coverage function, is the energy consumption function, is the communication conflict degree function, is the delayed response function, , , , is the weight coefficient, satisfying the sum of 1; S37, introduce an improved elimination update mechanism to calculate the standard deviation of the fitness of the current path group ,like If it is less than the dynamic threshold, the current population is identified to be converging, and individual perturbation operations are performed to randomly generate new individuals to replace the individuals with the worst fitness.

5. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: The S4 specifically includes: S41, selecting a disturbance target path vector, where the path vector is screened by an improved elimination update mechanism; S42. Setting the path disturbance probability , perform structural perturbation operations on the path vector according to the perturbation probability. The specific perturbation methods include: Task exchange perturbation: randomly select two positions in the path and exchange the corresponding task order; Task insertion perturbation: randomly select task nodes and insert them into new positions in the path; Subsequence reversal perturbation: randomly select a continuous subsequence in the path and reverse the order of the tasks; Local reordering perturbation: randomly reorder the order of tasks within a specified subsequence range; The disturbance intensity is determined according to the following disturbance control function: ； in, represents the disturbance intensity ratio, is the maximum disturbance proportional factor, is the current system temperature, is the initial system temperature; S43, generating a candidate path vector after completing the perturbation , to ensure that the path legitimacy is met, that is, all task nodes are not repeated, the order is reasonable, and the resource load does not exceed the task load capacity, forming a valid path solution.

6. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: The S5 specifically includes: S51, set The fitness value is , the perturbation path vector is The fitness value is , the fitness function is a multi-objective fitness function; S52, define the current path fitness difference as: ； At the same time, let the system's current expected acceptance difference be: ； in, is the initial acceptance expected difference, is the current scheduling iteration round, is the maximum number of iterations; S53, using the disturbance acceptance probability function based on the logistic distribution function to calculate the acceptance probability : ； in, For the current The system temperature of the iteration, is the soft decision adjustment coefficient, which is used to control the change range of the acceptance boundary. is an exponential function; S54, generate uniform random numbers in the interval [0,1] ,like , then accept the perturbation path for update, otherwise keep the original path unchanged; S55, update the system temperature, using an adaptive temperature decay strategy based on global disturbance behavior and fitness fluctuations: ； in, is the attenuation adjustment factor, is the standard deviation of the global individual path fitness in the current round, For the The perturbation path acceptance rate is defined as the ratio of the number of individuals currently accepting the path to the total number of individuals. Round The current mission path vector of the UAV With system temperature Output together.

7. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: S6 specifically includes: performing a multi-objective comprehensive evaluation on the scheduling path vector of each UAV, using the following fitness function: ； in, Indicates The mission path vector of the UAV, is the weighted coefficient of the objective function, satisfying , task coverage function It represents the ratio of the number of tasks completed by the UAV to the total number of tasks that can be scheduled. The energy consumption function Indicates the inverse proportional normalization value of the sum of the flight costs between consecutive tasks in the path to the maximum energy capacity, the communication conflict function Indicates the degree of interference between tasks in the path due to communication time or communication radius conflicts, and the response delay function It represents the normalized deviation between the task completion time and the task timeliness requirement, and is weighted averaged based on the task priority.

8. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: S7 specifically includes: assuming that the set of all paths in the current round is , the fitness set is ,in is the number of drones, and the individual with the best fitness in each mission area is set as the local guiding individual in the area for the next round of path search guidance. At the same time, the global optimal path vector of the current round is calculated. ,satisfy: ； Assume its fitness is ; Record the fitness difference of the optimal path before and after the two rounds: ； definition is the preset optimal path fitness difference threshold, is the preset disturbance path acceptance rate threshold. If And the current round perturbation path acceptance rate , or the current round satisfies , then the scheduling converges and the optimal individual path in the output path set is determined Its fitness As the final scheduling result; otherwise, the energy consumption corresponding to the task execution order and the distance between tasks in the current path of each drone is accumulated as its current round path cost, and its remaining energy is updated , and The data are stored together to provide input for the next round of path search and disturbance feasibility judgment.

9. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 1 is characterized in that: The S8 specifically includes: after the scheduling convergence conditions are met, extracting the final task path of each UAV, the task execution order, the start time and end time of each task, the flight distance between adjacent task nodes in the path, and the remaining energy information after completing the scheduling task, organizing the data in a structured manner, generating a scheduling output record for each UAV, and aggregating the scheduling output records of all UAVs to form a final scheduling plan.

10. The method for scheduling UAV remote joint defense tasks based on multi-objective optimization according to claim 9 is characterized in that: The output of the final scheduling scheme adopts a strategy-based structured organization method. The output strategies include the following three categories: a scheduling result sorting strategy driven by task priority, a path screening strategy under energy safety constraints, and a cluster consistency strategy for task deduplication fusion. Among them, the task priority driven strategy is used to arrange the tasks involved in the scheduling path of each drone in ascending or descending order according to the priority field of the task node, and give priority to completing key tasks with high response requirements. The energy safety constraint strategy is used to screen the output path based on the remaining energy threshold of each drone, and retain the path records that still have the lowest remaining energy margin after completing the current path task. The task deduplication fusion strategy is used to detect repeated task nodes in the set of all drone paths, and cross-validate the path start and end timing, communication conflicts and energy consumption costs, and select the only executor and eliminate redundant records. Finally, the scheduling results are generated into a unified scheduling data set output file according to the above strategies. The output file records the task sequence, task schedule, flight segment distance, task coverage, and remaining energy of each drone.

Images