Emergency scene unmanned aerial vehicle task allocation method based on multi-objective optimization

Abstract

The invention relates to the technical field of unmanned aerial vehicle scheduling, and discloses an emergency scene unmanned aerial vehicle task allocation method based on multi-objective optimization, and the method comprises the steps: obtaining a task parameter set of an unmanned aerial vehicle in real time through a standardized task interface, and obtaining a state data set of the unmanned aerial vehicle in real time through an unmanned aerial vehicle management platform; secondly, calculating an optimal task allocation scheme by adopting a multi-target particle swarm optimization algorithm and an improved PSO algorithm, and dynamically adjusting sudden emergency tasks and priority changes in combination with a task preemption mechanism; three task allocation plans including a time optimal scheme, a resource optimal scheme and a parameter optimal scheme are provided to ensure that tasks can be completed in the shortest time or executed at the lowest cost, and the final scheme is provided for a command center for decision making. According to the invention, the intelligence and flexibility of unmanned aerial vehicle scheduling are improved, and rapid response and global optimal resource configuration of unmanned aerial vehicle scheduling in an emergency scene are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle scheduling, specifically to an emergency scene unmanned aerial vehicle task allocation method based on multi-objective optimization. BACKGROUND

[0002] In recent years, with the acceleration of standardization and intelligentization, unmanned aerial vehicle scheduling has been deeply integrated into various aspects of social production and has become a key carrier of new-quality productivity. Unmanned aerial vehicle scheduling has always been a technical challenge, especially in complex environments. Traditional unmanned aerial vehicle scheduling methods often rely on fixed algorithms and cannot adapt to real-time changes in the environment or task requirements. Unmanned aerial vehicle scheduling systems based on artificial intelligence can consider weather, terrain, task priority, and other factors through deep learning and data analysis to optimize flight paths and scheduling strategies in real time. This not only improves the efficiency of unmanned aerial vehicles but also helps to reduce energy consumption and extend operating time.

[0003] The existing technology and its problems are as follows:

[0004] Scheduling method based on heuristic rules: Heuristic algorithms are methods that seek suboptimal solutions that meet computational efficiency and allocation goals by designing specific rules or strategies when it is difficult to directly obtain the optimal solution of the problem. Typical representatives include tabu search (TS), simulated annealing (SA), genetic algorithm (GA), etc. This type of algorithm has the advantages of easy implementation, fast calculation speed, and strong problem applicability. However, it also has inherent limitations such as high computational complexity, sensitivity to initial parameters, and inability to guarantee global optimal solutions. In the task allocation of unmanned aerial vehicles, if pre-set heuristic rules are used, when faced with complex and variable task requirements and limited unmanned aerial vehicle resources, this scheduling method based on empirical rules cannot achieve globally optimal resource allocation results.

[0005] Scheduling method based on centralized optimization algorithm: Integer programming or mixed integer programming optimization algorithms are used for unmanned aerial vehicle task scheduling, but as the task size increases, this method has high computational complexity and cannot meet real-time scheduling requirements.

[0006] Traditional Particle Swarm Optimization (PSO) based scheduling method: Particle Swarm Optimization (PSO) algorithm is derived from the bionic research on bird foraging behavior by Kennedy and Eberhart in 1995. Through dynamic adjustment of inertia weight, social learning factor and cognitive learning factor, combined with fitness function evaluation mechanism of individual extreme value and global extreme value, the algorithm realizes the optimal solution search under group cooperation. As an intelligent optimization method, particle swarm algorithm has the advantages of low computational complexity, strong robustness, outstanding global search ability and high convergence efficiency. However, due to the limitations of single objective optimization, it is difficult to meet multiple optimization objectives at the same time.

[0007] Unmanned aerial vehicle scheduling method based on static task allocation: This method first collects task demand and unmanned aerial vehicle state information, then calculates the task allocation scheme through optimization algorithm, and finally executes the scheduling. However, this method ignores the urgency of tasks, task dependency and dynamic adjustment of priority, making it difficult to respond to task changes in emergency or resource conflict situations.

[0008] Unmanned aerial vehicle scheduling method with weak task preemption mechanism: Some studies explore the task preemption mechanism, but usually only consider the insertion of emergency tasks, without comprehensive modeling of task dependency, dynamic adjustment of priority, multi-task conflict and other factors, resulting in insufficient system flexibility.

[0009] In summary, the existing technology has the following disadvantages:

[0010] 1) Lack of dynamic task preemption capability:

[0011] Traditional scheduling system uses fixed task queue and does not establish preemption mechanism based on real-time environmental changes (such as sudden tasks, resource shortage). Emergency tasks need to wait for the queue to be emptied, resulting in delayed response to critical tasks.

[0012] 2) Single optimization objective:

[0013] Existing technology usually uses single algorithm to optimize single index (time or resource consumption), lacks multi-objective collaborative optimization capability, and has low resource utilization or increased risk of task timeout.

[0014] 3) Static matching model → low matching degree between model and task demand:

[0015] Existing technology uses fixed weight allocation and does not dynamically adjust the matching degree model, resulting in decreased matching degree between unmanned aerial vehicle and task and increased failure rate of execution. SUMMARY

[0016] In view of the above problems, the purpose of the present application is to provide a multi-objective optimization-based unmanned aerial vehicle task allocation method for emergency scenarios, which combines multi-objective dynamic optimization and intelligent preemption decision-making to solve the multi-dimensional trade-off problem of unmanned aerial vehicle scheduling in complex scenarios and provide hierarchical decision support for emergency response, realize emergency task second-level response through a dynamic preemption mechanism, and reduce key task delays.

[0017] The multi-objective optimization-based unmanned aerial vehicle task allocation method for emergency scenarios comprises the following steps:

[0018] Step 1: Collect various data through an external emergency command system, including geographic information of disaster-stricken areas, weather conditions, traffic conditions, and distribution of rescue resources, obtain a task parameter set of the external emergency command system in real time through a standardized task interface, and obtain a unmanned aerial vehicle state data set in real time through a unmanned aerial vehicle management platform;

[0019] Step 2: Use a double-optimization channel including a multi-objective particle swarm optimization algorithm and an improved particle swarm optimization algorithm to perform task allocation optimization; the multi-objective particle swarm optimization algorithm defines a target function as minimizing the latest completion time of tasks and minimizing total resource consumption, maintains a Pareto front solution set through non-dominated sorting, and generates a time priority scheme and a resource saving scheme in combination with endurance, load, and time window constraints; the improved particle swarm optimization algorithm generates a dynamic parameter balancing scheme in combination with a penalty function;

[0020] Step 3: Based on a dynamic task preemption mechanism, the urgency of the task queue is evaluated in real time, when a high-priority task is inserted or environmental parameters are out of limits, the preemption mechanism is triggered, the unmanned aerial vehicle resources occupied by low-priority tasks are released, and unmanned aerial vehicles meeting the set conditions are preferentially allocated to emergency tasks;

[0021] Step 4: Push the generated three-level optimization scheme to the external emergency command system through a visual terminal for manual decision-making, then perform task allocation based on the decision-making result, feed back the unmanned aerial vehicle state and task progress in real time, dynamically trigger secondary optimization or abnormal fault tolerance, and form a closed-loop scheduling process.

[0022] The present application has the following advantages:

[0023] 1) The present application effectively balances efficiency and compliance in task allocation by combining the multi-objective particle swarm optimization (MOPSO) algorithm and the improved PSO algorithm with a multi-dimensional penalty function, quickly generating a feasible solution under complex constraints. The optimization of unmanned aerial vehicle emergency task scheduling is carried out, and efficient task allocation and reasonable resource scheduling are realized. Compared with the traditional static task allocation method, the present application introduces a double-interface interaction mechanism between the task and the state of the unmanned aerial vehicle, ensuring that the task information and the state of the unmanned aerial vehicle can be synchronized in real time, improving the accuracy and flexibility of the scheduling. At the same time, the improved MOPSO algorithm is used to balance between minimizing the task completion time and minimizing the total energy consumption of the unmanned aerial vehicle, ensuring that the scheduling scheme takes into account the task response speed and resource utilization efficiency.

[0024] 2) The present application designs a three-level optimal task scheduling scheme generation mechanism, provides time priority, energy optimization, and parameter scheme for different emergency scenarios, and pushes it to the command center through the decision interface. The anti-wind and endurance dynamic constraint model reduces the risk of task execution by quantifying the flight redundancy time and capability matching. Intelligent auxiliary decision and closed-loop feedback control are realized. Compared with the traditional heuristic scheduling method, the present application can dynamically adjust the optimization strategy in a complex task environment, improve the task execution capability of the unmanned aerial vehicle cluster, and provide reliable technical support for the application of unmanned aerial vehicles in emergency rescue scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The flowchart of the present application based on multi-objective optimization of unmanned aerial vehicle task allocation method in emergency scene. DETAILED DESCRIPTION

[0026] The present application will be further described in detail below in combination with the drawings and specific embodiments.

[0027] The present application proposes a kind of unmanned aerial vehicle task allocation method and system based on multi-objective particle swarm optimization (MOPSO) algorithm, and introduces a task preemption mechanism, which can optimize task scheduling in dynamic environment, improve task completion efficiency and unmanned aerial vehicle utilization.

[0028] The application processes real-time state data of unmanned aerial vehicles and tasks, including parameters such as speed, load capacity, wind resistance, endurance time, and single flight cost of the unmanned aerial vehicles, and information such as time limit, resource demand, emergency degree, and time window of the tasks. Secondly, a multi-objective particle swarm optimization (MOPSO) algorithm and an improved PSO algorithm are used to calculate an optimal task allocation scheme, and a task preemption mechanism is combined to dynamically adjust sudden emergency tasks and priority changes. By providing three task allocation plans of time optimization, resource optimization, and parameter optimization, it is ensured that the tasks can be completed in the shortest time or executed at the lowest cost, and the final scheme is provided to the command center for decision-making, improving the intelligence and flexibility of unmanned aerial vehicle scheduling and realizing rapid response and global optimal allocation of resources of unmanned aerial vehicle scheduling in emergency scenarios.

[0029] The scheme flow is roughly divided into the following steps:

[0030] Firstly, the standardized task interface provided by the system is used to inject emergency task parameters in real time, including time limit, resource demand, and emergency degree, and the real-time state data of the unmanned aerial vehicle management platform is synchronized, including location, endurance, and load capacity. The system dynamically evaluates the task queue based on a multi-dimensional preemption rule library, and when a high-priority task is inserted or the environmental parameter is out of limit, the preemption mechanism is triggered to release the resources of a low-priority task, and a high-score unmanned aerial vehicle is preferentially allocated to an emergency task.

[0031] Subsequently, a multi-objective particle swarm optimization (MOPSO) algorithm is used for parallel calculation: the unmanned aerial vehicle cluster is initialized as a particle swarm, a time optimization-minimizing task latest completion time and a resource optimization-minimizing total energy consumption double objective function are defined, a non-dominated sorting is used to maintain a Pareto front solution set, and a three-level plan-time priority plan, resource saving plan, and dynamic parameter balancing plan are generated in combination with endurance, load, time window constraints and a penalty function.

[0032] The optimization result is pushed to the command center visual terminal, supports manual decision selection, and provides real-time feedback of the unmanned aerial vehicle state and task progress during execution, and dynamically triggers secondary optimization or abnormal fault tolerance.

[0033] Finally, the task allocation data and algorithm parameters are archived to form a full-process dynamic scheduling system from real-time response, multi-objective optimization to closed-loop feedback, and realize efficient configuration and second-level preemption response of unmanned aerial vehicle resources in emergency scenarios.

[0034] The specific flowchart of the application is shown in Figure 1 .

[0035] 1. Task access stage:

[0036] A. The external emergency command system dynamically injects a task parameter set through a standardized task interface.

[0037] Let the task parameter set be: ; Wherein, is the task set, is the attribute set of the task, each task is composed of multiple attributes: ; Wherein, is the task unique identifier, is the task time limit, is the task required resources, is the task location coordinates, is the task type, is the task required wind resistance, is the task emergency degree, is the number of unmanned aerial vehicles required by the task, is the list of unmanned aerial vehicles assigned to the task.

[0038] At the same time, the unmanned aerial vehicle management platform uploads the unmanned aerial vehicle state data in real time through the interface, and the unmanned aerial vehicle state data set is: ; Wherein, is the unmanned aerial vehicle set, each unmanned aerial vehicle state parameter is: ; Wherein, is the unmanned aerial vehicle ID, is the unmanned aerial vehicle position, is the actual unmanned aerial vehicle endurance time, is whether the unmanned aerial vehicle is busy, is the unmanned aerial vehicle score, is the unmanned aerial vehicle endurance time (without counterweight), is the unmanned aerial vehicle flight speed, is the unmanned aerial vehicle load capacity, is the unmanned aerial vehicle executable task, is the unmanned aerial vehicle wind resistance, is the task single flight cost.

[0039] The main task type parameters are shown in Table 1, and the main unmanned aerial vehicle type parameters are shown in Table 2.

[0040] Table 1 Main task type parameter table .

[0041] Table 2 Main unmanned aerial vehicle type parameter table .

[0042] 2. Combined application performance module:

[0043] The emergency scene UAV combined application performance algorithm mainly provides UAV task priority analysis. According to the emergency scene and the urgency of the task, the algorithm analyzes the priority of different tasks to ensure that high-priority tasks are executed first. Considering the multi-machine cooperative execution of tasks, the algorithm realizes the cooperative planning of task allocation, improves the efficiency of emergency rescue, and uses the algorithm to intelligently analyze the technical capabilities, carried load and other information of the UAV, reasonably allocates tasks, and ensures that each UAV can play the maximum utility. This algorithm specifically dispatches UAVs with high performance and high resource consumption in an emergency situation, and dispatches UAVs with low resource utilization that can complete tasks within a specified time in a non-emergency situation.

[0044] A. The basic idea of the emergency scene UAV combined application performance algorithm mainly includes the following aspects:

[0045] a) UAV task allocation modeling;

[0046] Task allocation modeling is a multi-objective optimization problem, and the optimization objectives include minimizing task completion time and minimizing resource consumption.

[0047] Pareto optimal solutions are used to provide multiple candidate solutions to meet different scheduling needs.

[0048] b) Optimization method based on MOPSO;

[0049] MOPSO (Multi-Objective PSO) is used to solve the UAV scheduling problem. Through the individual optimal and global optimal updating strategy, the task allocation scheme is continuously optimized. Non-dominated sorting and congestion calculation are used to maintain the Pareto front, ensuring the diversity and coverage of the solution set.

[0050] c) Dynamic adjustment of UAV state and task adaptation;

[0051] During task execution, the UAV state is dynamically updated to ensure that the allocation scheme meets the real-time situation. Through the real-time feedback of the UAV management platform, the task allocation is adjusted.

[0052] d) Push and decision execution of optimized scheduling scheme;

[0053] Through the human-computer interaction terminal of the command center, time priority and resource priority are displayed for the commander to make decisions.

[0054] Through the above measures, the overall efficiency of large unmanned helicopters and medium-sized composite wing UAVs can be significantly improved and energy consumption can be reduced, thereby playing a greater role in various application scenarios.

[0055] B. Technical path:

[0056] a) Mission parameters and UAV state input;

[0057] External emergency command system injects a set of mission parameters through a standardized interface, including mission type, location, time constraints, resource requirements, etc.

[0058] UAV management platform uploads the current fleet status through the capability interface, including location, battery level, payload capacity, wind resistance, mission execution capability, etc.

[0059] b) Particle swarm initialization;

[0060] Available UAVs in the mission area are used as decision variables for particles, with each particle having a dimension of the number of UAVs. Initialize the position (mission allocation scheme) and velocity of the particle, and randomly assign tasks to UAVs.

[0061] In the multi-objective PSO (MOPSO) optimization process, the mission allocation scheme is modeled as a particle swarm, where each particle represents a mission allocation scheme.

[0062] The position of each particle is initialized as the task allocation of the UAV (one UAV can be assigned multiple tasks, and a task can be executed by multiple UAVs).

[0063] Set the initial velocity of the particle for subsequent update iterations.

[0064] Calculate the objective function of the particle:

[0065] Task completion time: the first objective is to minimize the latest task completion time, calculated as: ; where is the task completion time, is the completion time of the last task executed by UAV , calculated as: ; where is the start time of the task for UAV , is the distance from the current location to the task point for UAV , is the flight speed of UAV , is the time required for UAV to execute the task; ; where, The task distance is calculated precisely by the Harvard function.

[0066] The objective is to minimize the makespan of the entire set of tasks: .

[0067] Resource consumption: the objective is to minimize the total cost of all UAVs performing the tasks: ; where, is the resource consumption, is the cost of a single flight for a task.

[0068] c) Multi-objective PSO iterative optimization;

[0069] Individual best and global best calculation:

[0070] Calculate the current fitness of the particle (makespan + resource consumption) and update the individual best solution of each particle.

[0071] Maintain an external storage archive to store the Pareto front solutions (non-dominated solution set) of the current population.

[0072] Particle position and velocity update:

[0073] Use the velocity update formula of MOPSO: ; where, is the inertia weight, used to control the smoothness of particle movement; and are learning factors, used to determine the degree of influence of the individual best solution and the global best solution on the particle; and are random numbers, used to increase the randomness of the search; is the current position of the particle, is the current velocity of the particle, is the updated velocity of the particle; is the individual best solution, is the global best solution.

[0074] Update the position of the particle: ; Perform non-dominated sorting, maintain the Pareto front, and select the optimal solution set for the next round of optimization.

[0075] Iteration termination condition:

[0076] The iteration is terminated when the number of iterations reaches a pre-set upper limit, or the optimal solution does not change significantly for a number of consecutive generations. The fitness of each particle (task completion time + resource consumption) is calculated, and the individual historical optimum is updated. The Pareto front solution set is maintained, and non-dominated solutions are selected as the external archive for global guidance. The particle velocity is updated according to the particle velocity update formula, and the particle velocity and position are updated in combination with the individual optimum and global optimum. The iteration is performed until the maximum number of generations is reached, and the final optimization solution set is output.

[0077] d) Constraints;

[0078] In the optimization process, we need to ensure that the solution meets a series of physical constraints and task constraints.

[0079] Task constraints:

[0080] Each task must be performed by at least one UAV: ; where, is the first decision variable, representing whether the UAV performs the task , is the set of UAVs.

[0081] Endurance constraint:

[0082] The total energy consumption of the UAV in task execution cannot exceed its current remaining endurance: ; where, is the maximum available endurance of the UAV , is the time limit for task execution;

[0083] Payload constraint:

[0084] If any transportation materials are needed, the payload capacity of the UAV must be met: ; where, is the maximum payload capacity of the UAV ;

[0085] Task time window constraint:

[0086] Some tasks may have strict time windows that need to be guaranteed for task execution within the available time:

[0087] where, and are the start time and end time of the task The earliest and latest execution time, It's a drone The actual time the task was executed.

[0088] UAV scheduling uniqueness constraint:

[0089] A drone can only perform one mission at a time: ;

[0090] Decision variables:

[0091] In the MOPSO optimization process, we need to mainly define the decision variables applicable to the task scheduling problem:

[0092] Task Assignment Variables ; This variable determines the drone Whether the task is assigned .

[0093] Drone flight speed: ; Indicates the drone's flight speed, ranging between the minimum and maximum allowed values. Due to the complexity of emergency scenarios, the actual speed is controlled by the pilot.

[0094] Solution generation and decision push:

[0095] Generate a secondary plan:

[0096] Time-optimal solution: prioritize ensuring that tasks are completed in the shortest possible time.

[0097] Resource-optimal solution: Minimize the energy consumption of the drone within a limited time and also generate results beyond the time limit.

[0098] 3. Parameter optimization module:

[0099] A. Mapping drone indexes through particle encoding and combining penalty functions to solve the task-drone matching problem;

[0100] Particle encoding: ; The objective function is to minimize the total penalty value of the task assignment, which includes the following penalty terms:

[0101] Invalid index penalty : The drone index is out of range.

[0102] Penalty for lack of ability : The drone cannot meet the capacity, wind resistance, or time requirements of the mission.

[0103] Reuse penalty : Same UAV is assigned to multiple tasks.

[0104] Same task reuse penalty : Same UAV is assigned to the same task multiple times.

[0105] The final goal is to minimize the sum of all penalty terms, ensuring task assignment: ; Define second decision variable : ; where the second decision variable is a binary variable indicating whether a UAV is assigned to a task ; 1 if assigned, 0 if not assigned;

[0106] Invalid index penalty: ; Insufficient capability penalty: ; Reuse penalty: ; Same task reuse penalty: ;

[0107] B. Constraints;

[0108] Task demand constraint: ;

[0109] UAV capability constraint:

[0110] The UAV needs to support the task type;

[0111] The UAV's wind resistance capability needs to meet the task demand;

[0112] The UAV must be idle;

[0113] The task time limit needs to be less than the UAV's remaining endurance time (round-trip flight time plus redundancy ): ; ; where, , , and indicates whether the task type is supported, whether the wind resistance capability is reached, and whether the UAV is idle and the UAV endurance is greater than the task time limit, 1 is taken if yes, otherwise 0 is taken. is a redundancy.

[0114] Resource uniqueness constraint:

[0115] The UAV cannot be repeatedly allocated to multiple tasks, and the same UAV cannot be repeatedly allocated in the same task.

[0116] ;

[0117] Emergency task priority constraint:

[0118] Define emergency task: ; ; wherein, is a subset of the emergency task.

[0119] Emergency task resource guarantee constraint: ; wherein, is the UAV whether it is allocated to an emergency task , is the minimum number of UAVs required by the emergency task , is the total number of available UAVs;

[0120] Emergency task priority constraint: ;

[0121] wherein, the penalty coefficient of the emergency task is much larger than that of the non-emergency task, so that the algorithm preferentially meets the constraints of the emergency task.

[0122] C. Dynamic task preemption mechanism;

[0123] Through priority judgment and resource release rules, the UAV resources of high-priority tasks are dynamically preempted from low-priority tasks, and idle UAVs are preferentially allocated.

[0124] Preemption condition: and ; Resource release: (release the preempted UAV);

[0125] D. Plan generation and decision pushing;

[0126] Generate a plan: the parameter optimal plan.​

[0127] 4. The generated three-level optimization scheme is pushed to the external emergency command system through the visualization terminal for manual decision-making, task allocation is executed based on the decision-making result, the state of the unmanned aerial vehicle and the task progress are fed back in real time, secondary optimization or abnormal fault tolerance is triggered dynamically, and a closed-loop scheduling process is formed.

[0128] In summary, the present application effectively balances efficiency and compliance in task allocation by combining the multi-objective particle swarm optimization (MOPSO) algorithm and the improved PSO algorithm with a multi-dimensional penalty function, quickly generating a feasible solution under complex constraints. The present application optimizes the scheduling of unmanned aerial vehicle emergency tasks, achieving efficient task allocation and rational resource scheduling. Compared to traditional static task allocation methods, the present application introduces a double-interface interaction mechanism between tasks and unmanned aerial vehicle states, ensuring that task information and unmanned aerial vehicle states can be synchronized in real time, improving the accuracy and flexibility of scheduling. At the same time, the improved MOPSO algorithm is used to balance the minimization of task completion time and the minimization of total unmanned aerial vehicle energy consumption, ensuring that the scheduling scheme takes into account both task response speed and resource utilization efficiency.

Claims

1. A multi-objective optimization-based UAV task allocation method for emergency scenarios, characterized by: The following steps are involved: Step 1: Collect various data through the external emergency command system, including geographical information of the disaster area, weather conditions, traffic conditions, and the distribution of rescue resources. Obtain the task parameter set of the external emergency command system in real time through the standardized task interface, and obtain the drone status data set in real time through the drone management platform; Step 2: Task allocation optimization is performed using a dual optimization channel consisting of a multi-objective particle swarm optimization algorithm and an improved particle swarm optimization algorithm. The target particle swarm optimization algorithm defines the objective function as minimizing the latest task completion time and minimizing total resource consumption. The Pareto front solution set is maintained through non-dominated sorting. A time priority solution and a resource conservation solution are generated by combining endurance, load, and time window constraints. The improved particle swarm optimization algorithm is combined with a penalty function to generate a dynamic parameter balancing solution. Step 3: Based on the dynamic task preemption mechanism, the urgency of the task queue is evaluated in real time. When a high-priority task is inserted or the environmental parameters exceed the limit, the preemption mechanism is triggered to release the drone resources occupied by low-priority tasks and give priority to allocating drones that meet the set conditions to urgent tasks. Step 4: Push the generated plan to the external emergency command system through the visualization terminal for manual decision-making, and then execute task allocation based on the decision results, provide real-time feedback on the drone status and task progress, dynamically trigger secondary optimization or abnormal fault tolerance, and form a closed-loop scheduling process.

2. The emergency scenario UAV task allocation method based on multi-objective optimization according to claim 1 is characterized in that: In step 1, the task parameter set Defined as: ； in, is a set of tasks, each task It consists of several attributes: ； in, For the task The attribute set of Is the unique identifier of the task, To meet the timeliness requirement of the task, The resources required for the task, is the task location coordinate, is the task type, The wind resistance required for the mission, The urgency of the task, The number of drones required for the mission, A list of drones assigned to the mission; The drone status data set Defined as: ； in, is a set of drones, each drone state parameter for: ； in, is the drone ID, is the drone location, is the actual drone flight time, Is the drone busy? Score for the drone, For the drone's flight time, is the flight speed of the drone, is the payload capacity of the drone, For drones to perform tasks, For the UAV's wind resistance, The cost of a single mission flight.

3. The emergency scenario UAV task allocation method based on multi-objective optimization according to claim 2 is characterized in that: In step 3, the multi-objective particle swarm optimization algorithm is used to optimize task allocation, specifically including: Step 311: Use the available drones in the task area as the decision variables of the particles, with the dimension of each particle being the number of drones. Model the task allocation plan as a particle swarm, where each particle represents a task allocation plan. Initialize the position of each particle to the task allocation status of the drone, and set the initial velocity of the particle for subsequent update iterations. Step 312: Calculate the objective function of the particle; Minimize the latest completion time of the task. The calculation formula is: ； in, is the task completion time, For drones The completion time of its last task is calculated as follows: ； in, For drones The task start time, For drones The distance from the current position to the mission point, For drones The flight speed, For drones the time required to perform the task; ； in, is the Harvard function; Minimize the total cost of all drones performing tasks: ； in, For resource consumption, The cost of a single mission flight; Step 313: iterative optimization using a multi-objective particle swarm optimization algorithm; Individual optimal and global optimal calculation: Calculate the current fitness of the particle, i.e., task completion time and resource consumption, and update the individual optimal solution of each particle; Maintain an external storage file to store the Pareto front solution of the current population; Particle position and velocity update: The speed update formula using the multi-objective particle swarm optimization algorithm is: ； in, is the inertia weight, which is used to control the smoothness of particle movement; and is the learning factor, which is used to determine the degree to which the particle is affected by the individual optimal solution and the global optimal solution; and is a random number used to increase the randomness of the search; is the current position of the particle, is the current velocity of the particle, is the updated velocity of the particle; is the individual optimal solution, is the global optimal solution; Update the particle's position: ； in, The updated position of the particle; Perform non-dominated sorting, maintain the Pareto front, and select the optimal solution set for the next round of optimization; Iteration termination condition: The number of iterations reaches the preset upper limit, or the optimal solution is found in n consecutive d If the degree of change in each generation does not exceed the expected level, the iteration is terminated; the fitness of each particle is calculated and the individual historical optimal solution is updated; the Pareto front solution set is maintained and non-dominated solutions are selected as external archives for global guidance; according to the particle velocity update formula, the particle velocity and position are updated in combination with the individual optimal solution and the global optimal solution; the iteration is continued until the maximum number of generations is reached and the final optimized solution set is output; Step 314: Establish constraints; 1) Define decision variables; Define the first decision variable for the task scheduling problem : ； 2) Task constraints; Each mission must be performed by at least one drone and is represented as follows: ； 3) Endurance constraints; The total energy consumption of the drone during mission execution cannot exceed its current remaining flight time, as shown below: ； in, For drones The maximum available battery life, Timeliness requirements for tasks; 4) Load constraints; Any materials that need to be transported must meet the load capacity of the drone, as shown below: ； in, For drones Maximum load capacity; 5) Task time window constraints; The tasks are executed within the available time, which is represented as follows: ； in, and The tasks are The earliest and latest execution time, It's a drone The actual time to perform the task; 6) Uniqueness constraint for drone scheduling; A drone can only perform one mission at a time, as shown below: ； 7) UAV flight speed constraints; ； Indicates the flight speed of the drone The range is within the minimum allowed speed and maximum speed between.

4. The method for allocating emergency scenario UAV tasks based on multi-objective optimization according to claim 3 is characterized in that: In step 3, the improved particle swarm optimization algorithm is used to optimize task allocation, specifically including: Step 321: Mapping the drone index through particle coding and combining it with a penalty function to solve the task-drone matching problem; the particle coding is: ； in, For the A task allocation scheme, Assign positions to global drones; Step 322: Define the following penalty terms: Invalid index penalty : Drone index exceeds available range; Penalty for lack of ability : The drone cannot meet the mission’s capabilities, wind resistance, or time requirements; Reuse Penalty : The same drone is assigned to multiple tasks; Penalty for repeated assignment of the same task : The same drone is assigned to the same mission multiple times; The ultimate goal is to minimize the sum of all penalty terms: ； in, α 、 、 and is the weight of each item; Define the second decision variable : ； Among them, the second decision variable Is a binary variable representing the drone Is it assigned to a task? ; If it is 1, it means allocated; if it is 0, it means unallocated; then: ； ； ； ； in, For valid drone index, For the The drone was assigned to tasks; 、 、 and Indicates whether the drone is not a valid index, drone Is the task unable to be performed? , drones Is it assigned to missions and the same drone Whether the person has been assigned to the same task multiple times; Step 323: Set constraints: 1) Task requirement constraints: ； in, To carry out the mission The number of drones required; 2) UAV capability constraints: The drone must support the mission type, its wind resistance must meet the mission requirements, the drone must be idle, and the mission time limit must be less than the drone's remaining flight time, as shown in the following formula: ； ； in, 、 、 and Indicates whether the mission type is supported, whether the wind resistance is achieved, and whether the drone Is it idle and whether the drone's endurance is greater than the mission time limit? is the redundancy; 3) Resource uniqueness constraint: A drone cannot be assigned to multiple tasks repeatedly, and the same drone cannot be assigned to the same task repeatedly, as shown in the following example: ； 4) Urgent task constraints: Define urgent tasks: ； ； in, A subset of urgent tasks; The urgency of the task; Constraints on emergency task resource guarantee: ； in, For drones Is it assigned to an urgent task? , For urgent tasks Minimum number of drones required, is the total number of available drones; Urgent task priority constraints: ； in, 、 、 and They are invalid index penalty for urgent tasks, insufficient capacity penalty, reuse penalty, and repeated assignment penalty for the same task; 、 、 and are the corresponding weight coefficients respectively.

5. The method for allocating emergency scenario UAV tasks based on multi-objective optimization according to claim 4 is characterized in that: In step 3, a dynamic task preemption mechanism is developed, specifically: Through priority judgment and resource release rules, high-priority tasks can dynamically preempt drone resources from low-priority tasks, giving priority to allocating idle drones. Preemption conditions: and ; Resource release: , that is, releasing the seized drone; in, 、 、 and They are the newly arrived high priority task, the current low priority task, The endurance of each drone and the set of drones whose resources are preempted by high-priority tasks.

Images