A heterogeneous unmanned aerial vehicle scheduling method based on multi-machine nest multi-tasking

Abstract

This invention relates to the field of UAV swarm collaborative control technology, specifically disclosing a heterogeneous UAV scheduling method based on multi-nest, multi-task architecture. It aims to solve the scheduling optimization challenges of multi-nest deployment, heterogeneous UAV collaboration, and parallel multi-task handling in complex scenarios such as fire emergency rescue. The method first constructs a unified multi-nest-heterogeneous UAV-multi-task system model, defining the attributes and spatiotemporal constraints of UAVs, nests, and tasks. Then, an improved multi-objective genetic algorithm is used for static task pre-allocation to quickly generate a high-quality initial scheduling scheme. During the dynamic execution phase, a global state management mechanism monitors the system in real time, triggering reassignment based on event severity and filtering reassignable resources based on priority and distance to achieve rapid response and resource compensation. Finally, the scheduling performance is evaluated through a simulation verification platform. This invention, through a combined static and dynamic scheduling strategy, significantly improves the scheduling efficiency, resource utilization, and overall robustness of the system in dynamic and uncertain environments, providing an efficient solution for the application of large-scale UAV swarms in emergency patrols, disaster relief, and other scenarios.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) swarm control and intelligent decision-making technology, specifically relating to a multi-UAV collaborative scheduling method for emergency rescue scenarios. More specifically, this invention focuses on the static and dynamic task scheduling problem of heterogeneous UAV swarms in multi-nest deployment environments, covering technical directions such as optimization algorithms and multi-agent system collaborative control. Background Technology

[0002] Over the past decade, multi-UAV collaborative scheduling technology has gradually become an important research direction in the fields of urban public safety and disaster prevention. With the continuous improvement of the intelligence and swarming level of UAV systems, emergency rescue technology based on multi-UAV collaboration has shown significant advantages. Unlike the traditional operation mode that relies on a single UAV or a small-scale homogeneous UAV swarm, multi-UAV systems, through distributed collaborative decision-making, can achieve parallel handling of multiple areas and types of tasks at complex disaster sites, featuring rapid response, flexibility, and a unique perspective. This technological paradigm has shown great application potential in fire emergency tasks such as disaster assessment, personnel search and rescue, and emergency material delivery, significantly improving the situational awareness, decision-making speed, and operational safety of rescue operations. However, despite the broad prospects of multi-UAV collaborative technology, its practical application still faces severe challenges when facing large-scale, complex, and dynamically evolving mega-disaster sites. Most existing research is based on idealized assumptions of "single landing point" and "homogeneous UAV swarm," or focuses on relatively simple single-task-single-UAV allocation models, which differ significantly from the real-world scenarios of fire emergency rescue. The distributed multi-nest deployment strategy adopted to shorten emergency response time, along with heterogeneous UAV swarms equipped with different functional payloads, has transformed the task allocation problem into a high-dimensional, strongly constrained optimization challenge that requires consideration of multi-nest collaborative scheduling, heterogeneous resource matching, and complex task coupling. Existing methods struggle to quickly generate high-quality solutions during the static pre-allocation phase, and in the dynamic execution phase, they lack sufficient real-time response and system robustness to emergencies. They also fail to effectively coordinate heterogeneous resources across nests and ensure the temporal logic between collaborative tasks, thus hindering the improvement of overall rescue efficiency.

[0003] To address the aforementioned challenges, this invention proposes a heterogeneous UAV scheduling method based on multi-nest, multi-task architecture. The core innovation of this method lies in constructing a comprehensive mathematical model that can uniformly describe multi-nest deployment, functional differences among heterogeneous UAVs, various task types, and their spatiotemporal and collaborative requirements, laying a precise theoretical foundation for algorithm design. Regarding static task pre-allocation, this invention designs an improved multi-objective genetic algorithm, achieving rapid generation of initial task schemes under strong constraints through a hierarchical optimization framework and problem-oriented genetic operators. Regarding dynamic task reallocation, this invention develops a graded response mechanism based on event severity. For dynamic events of different levels, it triggers local rule adjustment strategies, ensuring the system's rapid response capability and operational stability in uncertain environments. This combined static and dynamic approach systematically solves the resource allocation and scheduling challenges under a "multi-nest-multi-type-multi-task" collaborative architecture. It not only improves adaptability to complex fire scenarios but also significantly optimizes task completion rate, response time, and resource utilization through efficient resource matching and flexible online adjustments, providing a practical technical approach to enhance the collaborative operation capabilities of UAV swarms in complex disaster environments. Summary of the Invention

[0004] The purpose of this invention is to provide a heterogeneous UAV scheduling method based on multi-nest multi-task, which solves the problems of difficulty in multi-nest collaborative scheduling, low efficiency of heterogeneous resource allocation, and poor adaptability to dynamic environments faced by UAV swarms in fire emergency rescue scenarios in the prior art.

[0005] The technical problems to be solved by this invention include: how to achieve unified modeling and resource description of heterogeneous UAVs under multi-nest deployment; how to quickly generate high-quality static task allocation schemes under strong spatiotemporal constraints; how to deal with sudden dynamic events and maintain system stability during task execution; and how to reduce system complexity and improve scalability through distributed architecture.

[0006] This invention solves the above-mentioned technical problems by adopting the following technical solution: a heterogeneous UAV scheduling method based on multi-nest multi-task, comprising the following steps.

[0007] S1 Multi-Nest Heterogeneous UAV Cooperative Scheduling System Model Construction

[0008] A unified mathematical model is established, encompassing three entity types: drones, drone nests, and tasks. The drone model defines its unique identifier, type, current location, flight speed, battery life, and onboard sensor types. The drone nest model describes its geographical location, capacity limitations, supported drone types, and current number of nested drones. The task model includes the target location, priority, required sensor types, and required number of drones. Physical constraints on system operation are described through constraints, including drone endurance constraints, sensor matching constraints, nest capacity constraints, task time window constraints, and drone exclusivity constraints. The objective function focuses on minimizing task completion time and minimizing penalties, providing a mathematical foundation for subsequent scheduling algorithms.

[0009] S2 is implemented based on a static task scheduling method using an improved multi-objective genetic algorithm.

[0010] To address the static task allocation problem, an improved multi-objective genetic algorithm is employed. The algorithm uses a variable-length encoding method based on task sequences, where the chromosome length equals the total number of tasks. Each gene position corresponds to a task, and the gene value represents the set of drones assigned to that task. Initial population generation combines heuristic rules and a random strategy, prioritizing the allocation of suitable drones from the nearest nest to high-priority tasks. The fitness function comprehensively considers task completion time, resource utilization efficiency, and constraint violation severity, handling multi-objective optimization and constraints through weighted summation and penalty function methods. Genetic operations include tournament selection, two-point crossover, and various mutation strategies, combined with an elite retention mechanism to ensure algorithm convergence. Through iterative evolution, a high-quality static scheduling scheme satisfying all constraints is ultimately output.

[0011] S3 implements a dynamic task reallocation mechanism based on global state management.

[0012] To address dynamic events during task execution, a global state manager is established to track the drone task allocation status in real time. When a new task is detected or a resource failure occurs, the scope and urgency of the event are first assessed. Based on priority comparison and spatial distance filtering principles, reassignable drone resources are identified, and reassignment decisions are made only for low-priority tasks and nearby drones. The reassignment process employs a real-time negotiation mechanism, combining central coordination and local optimization to ensure that the solution meets real-time requirements while maintaining system stability. For resource gaps caused by reassignment, a compensatory resource allocation strategy is activated, allocating alternative drones from the idle resource pool or low-priority areas to ensure the continuous execution of the original task.

[0013] Construction and Algorithm Verification of S4 Simulation Platform

[0014] A drone swarm scheduling simulation platform was built to simulate multi-nest deployment, heterogeneous drone resources, and various task requirements in urban fire emergency scenarios. Static scheduling tests were used to verify the efficiency of the initial scheme generation, while dynamic event injection was used to test the response speed and stability of the rescheduling mechanism.

[0015] The advantages of this invention include: Highly efficient scheduling capability. The algorithm of this invention achieves optimized allocation of static tasks through an improved multi-objective genetic algorithm, significantly improving task completion efficiency and resource utilization. In fire emergency rescue scenarios, it can effectively optimize task execution and fully meet the complex scheduling needs of multi-nest heterogeneous UAVs. Dynamic adaptability. Through a global state management mechanism and real-time rescheduling strategy, the system can respond quickly to emergencies, achieving efficient resource reallocation. Simultaneously, it utilizes an intelligent compensation mechanism to maintain task continuity, exhibiting good robustness and flexibility, suitable for dynamic and uncertain environments such as fire spread and equipment failure. Practicality and economy. Because this invention adopts a distributed architecture and optimization algorithm, it reduces dependence on high-performance hardware and complex communication systems, effectively controlling UAV hardware costs and system deployment costs. Furthermore, the modular design makes the system easy to scale and deploy on a large scale, making it particularly suitable for emergency scenarios requiring rapid response, such as fire patrols and disaster relief, providing a practical solution for the actual application of multi-UAV swarms. Attached Figure Description

[0016] Figure 1 This is a flowchart of the steps of the present invention.

[0017] Figure 2 This is a schematic diagram of the static task scheduling results of the present invention.

[0018] Figure 3 This is a schematic diagram of the redistribution results under the influence of dynamic events according to the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and beneficial effects of the embodiments of the present invention clearer, the present invention will be clearly and completely described below with reference to the embodiments. The embodiments described are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort all fall within the protection scope of the present invention.

[0020] This invention addresses the application requirements of "cooperative scheduling of multi-nest heterogeneous UAVs in fire emergency rescue scenarios" and proposes a closed-loop technical solution of "system modeling - static optimization - dynamic response - algorithm verification". In the task initialization phase, a unified system model is constructed, encompassing three types of entities: UAVs, nests, and tasks, comprehensively considering resource heterogeneity, spatiotemporal constraints, and collaborative requirements. An improved multi-objective genetic algorithm is used to achieve static task pre-allocation, generating a high-quality initial scheduling scheme that satisfies multiple constraints. A global state management mechanism enables dynamic event response, allowing for rapid reallocation in response to sudden new tasks, resource failures, and other situations. Finally, a simulation verification platform is used to comprehensively evaluate the system performance, providing reliable technical support for practical applications.

[0021] In the system modeling phase, the UAV type, performance parameters, and sensor configuration are first defined, and a nest network distribution and capacity constraint model is established to clarify task attributes and collaborative requirements. Based on this, a static scheduling algorithm is designed, which achieves efficient task-resource matching through chromosome encoding, fitness function optimization, and genetic operator design. For dynamic events during execution, a hierarchical response mechanism is established, and system stability is ensured by combining local adjustment and global replanning strategies. Finally, the effectiveness and practicality of the method are verified through multi-scenario simulation experiments, forming a complete scheduling solution.

[0022] To achieve the above objectives, such as Figure 1 , Figure 2 and Figure 3 As shown, the present invention provides an intelligent multi-UAV collaborative task execution scheduling method, the method comprising the following steps:

[0023] Step S1: Model Construction of Multi-Nest Heterogeneous UAV Cooperative Scheduling System

[0024] This step aims to establish a unified multi-factor system model, providing a mathematical foundation for subsequent algorithms. First, attribute parameters are defined for three types of entities: drones, nests, and tasks. The drone model covers key performance indicators such as type identification, flight speed, endurance, and sensor configuration. The nest model describes geographical distribution, capacity limitations, and supported drone types. The task model includes spatial location, priority weights, time windows, and coordination requirements. Based on this, multiple constraints are integrated, such as drone endurance, sensor matching rules, nest capacity limitations, and task timing logic. A multi-objective optimization function is constructed, with the core objective of minimizing task completion time and minimizing penalties. This model provides standardized input for static and dynamic scheduling algorithms by accurately characterizing resource heterogeneity, spatiotemporal coupling, and coordination requirements.

[0025] Step S2: Implementation of the static task scheduling method based on the improved multi-objective genetic algorithm

[0026] This step focuses on generating an initial task allocation scheme when disaster information is relatively certain. An improved multi-objective genetic algorithm is used as the core solver, employing a hierarchical optimization framework to handle high-dimensional combinatorial optimization problems. The algorithm design includes: a variable-length chromosome encoding mechanism based on task sequences, mapping each gene position to a task-UAV allocation relationship; an initial population generation method combining heuristics and stochastic strategies, prioritizing resource allocation for high-priority tasks; fitness function design; and parameterized implementation of genetic operators such as selection, crossover, and mutation. Through iterative optimization and an elite retention strategy, a high-quality static scheduling scheme satisfying all constraints is output, providing an initial blueprint for task execution.

[0027] Step S3: Implementation of a dynamic task reallocation mechanism based on global state management

[0028] This step addresses unforeseen dynamic events during mission execution by designing a real-time response and rescheduling mechanism. A global state manager tracks the drone mission allocation status in real time, establishing an event detection and evaluation process that triggers tiered response strategies based on event severity. The rescheduling process prioritizes available drones in low-priority mission areas, using spatial distance filtering and power constraint checks to achieve rapid resource reuse and compensatory allocation, ultimately generating a dynamically adjusted collaborative scheduling scheme.

[0029] Step S4: Construction of Simulation Verification Platform and Algorithm Verification

[0030] This step verifies the effectiveness and practicality of the above method through simulation experiments. A simulation platform oriented towards firefighting scenarios is built, integrating multi-nest network, heterogeneous UAV models, and dynamic task sequences. Test cases cover static scheduling verification and dynamic event response capability verification. Simulation results are output through a visual interface, providing data support for parameter optimization and actual deployment, ensuring the method's engineering feasibility.

[0031] Specifically, step S1 includes the following steps:

[0032] S11 UAV Model Definition and Parameter Configuration

[0033] This invention first establishes a drone model to describe the characteristics of various drones participating in collaborative tasks. This paper assumes that the set of all drones is [the following]. ,in Represents the total number of drones. Representing the A drone. The drone model includes the following core attribute parameters: Drone Number. Types of drones full flight time of drones drone speed The current nest of the drone drone location Drone battery power Types of sensors that drones can carry This model enables the system to accurately describe the capability differences of heterogeneous UAVs, providing a data foundation for subsequent task matching.

[0034] S12 Nest Model and Capacity Constraints Establishment

[0035] The nesting model describes the network characteristics of distributed take-off and landing bases. Let the set of UAV nests be... ,in Represents the total number of nests. Representing the Each drone nest contains a different type and a different number of drones. The drone nest model includes the following core attribute parameters: Drone Nest Number. Location of drone nests Unmanned aerial vehicle (UAV) nest capacity Types of drones that can land in drone nests The number of drones that can stay for each type Number of drones of various types currently docked in drone nests The drone number where the drone nest is located Since there may be multiple types of drones that can reside in a drone nest, the dimension of the number of each type of drone also changes with the type of drone, i.e., the types of drones that can reside in a drone nest. The number of times each type of drone can land. Number of drones of various types currently docked in the drone nest The capacity constraints ensure that the number of drones actually parked in the drone nest does not exceed its physical capacity limit, and the number of drones of each type does not exceed the corresponding sub-capacity limit.

[0036] S13 Mission Model and Spatiotemporal Attribute Definition

[0037] The task model describes various operational requirements in fire emergency rescue. Let the set of task objectives be... ,in Represents the total number of targets. Representing the There are several objectives. Each objective has a different priority. Each task attribute includes a task number. Task location Task Priority Task Type Types of sensors required for the task Number of drones required Task execution time window .

[0038] S14 System Constraint Integration

[0039] This invention integrates the following key constraints to ensure the feasibility of the solution: Battery life constraint: The total flight time of the drone in performing the mission shall not exceed the maximum flight time supported by its current battery power. ,in Indicates the first The total flight time of the drones performing the mission, of which The estimate is ,in This indicates the time the drone spends performing tasks within the designated area. This indicates the distance of the drone to the mission point or the nearest point to the mission area boundary, and it is expected that the drone will return to its starting point after completing the mission.

[0040] Sensor matching constraint: The types of sensors carried by the UAV ensemble performing the mission must cover the sensor set required for the mission. ,in, This represents the set of sensor types carried by all drones performing the task.

[0041] Task coverage constraint: The number of drones assigned to each task shall not be less than the task requirement. ,in, This indicates the total number of drones performing the task.

[0042] Nest type capacity constraints: When a drone returns to its nest, it must meet nest type compatibility and capacity availability requirements. .

[0043] Drone exclusivity constraint: A single drone can only be assigned to one task at a time. ,in For binary decision variables, when the decision variable Indicates that drones Assigned to task Conversely, decision variables Indicates that drones will not be used Assigned to task .

[0044] Time window constraint: The time window constraint requires the drone to arrive within the time allowed by the mission. And the time it takes for the drone to leave the mission after its completion. It must also be earlier than the end of the time window, i.e. The time it takes for the drone to arrive at the mission is... The time it takes for the drone to leave the mission after its completion .

[0045] In addition to the above constraints, since different tasks have different priorities, higher priority tasks have higher priority allocation rights, while priority allocation rights for tasks of the same priority remain consistent.

[0046] S15 Multi-objective optimization function design

[0047] To better reflect actual operational scenarios, several key factors were comprehensively analyzed. For the heterogeneous multi-UAV task allocation problem based on multi-objective optimization, a task time objective function was constructed. Different drones have different flight speeds, resulting in varying mission completion times depending on the mission allocation scheme. Due to power limitations, to reduce risk and improve mission efficiency, the mission allocation scheme should minimize the total mission execution time. The time consumed from drone departure to its return to base after completing the mission is primarily determined by path travel time. and task execution time Composed of two parts The path consumption time can be obtained by dividing the path from the drone to the disaster area by the drone's speed. Assuming that the time it takes for a drone to perform any task is directly proportional to the area of ​​the disaster zone, that is... ,in, This is a task time ratio coefficient. This represents the area affected by the disaster.

[0048] Therefore, a heterogeneous multi-UAV task allocation model based on multi-objective optimization can be constructed, including maximizing rewards. Minimize penalty and minimize task time .

[0049] Through the systematic implementation of the above five sub-steps, a complete multi-nest heterogeneous UAV collaborative scheduling system model was constructed, providing a rigorous mathematical foundation and constraint framework for subsequent static scheduling algorithms and dynamic reallocation mechanisms. This model fully considers the resource heterogeneity, spatiotemporal constraints, and collaborative needs in actual fire emergency rescue scenarios, ensuring that the scheduling scheme conforms to both physical reality and mission performance requirements.

[0050] Preferably, step S2, based on the static task scheduling method of the improved multi-objective genetic algorithm, includes the following steps:

[0051] S21 chromosome coding scheme design

[0052] This invention employs a variable-length chromosome encoding mechanism based on task sequences, mapping task allocation schemes to a chromosome structure of a genetic algorithm. Each chromosome represents a complete task allocation scheme, with its length equal to the total number of tasks. Gene positions are arranged in descending order of task priority, with each gene position corresponding to a specific task. The gene value represents the set of drones allocated to that task. For tasks requiring multi-drone collaboration, the gene value is a set of drone identifiers; for single-drone tasks, the gene value is a single drone identifier. This encoding method intuitively reflects the "task-resource" allocation relationship, facilitating the handling of heterogeneous drones and complex task requirements. The encoding process ensures that each drone appears only once in the chromosome, satisfying drone exclusivity constraints. Furthermore, the encoding scheme includes auxiliary information segments, recording the drone's departure nest and task execution sequence, providing complete information for subsequent decoding and evaluation.

[0053] S22 Initial Population Generation Strategy

[0054] To improve the algorithm's convergence speed and solution quality, a hybrid strategy combining heuristic rules and random generation is employed for initial population generation. The heuristic part processes tasks in descending order of priority, allocating the nearest and most capable available drones to high-priority tasks. Allocation considers drone endurance, sensor compatibility, and time window constraints to ensure feasible solutions are generated. The random part allocates available drones randomly, increasing population diversity. The ratio of these two strategies can be adjusted according to the problem size, typically with heuristic individuals comprising 30% of the population. The population size is set to 100 individuals to ensure sufficient diversity while controlling computational complexity. This hybrid initialization strategy maintains population diversity while ensuring initial solution quality, laying a solid foundation for subsequent evolution.

[0055] S23 Fitness Function Construction

[0056] The fitness function is the guiding core of algorithm optimization. This invention designs a multi-objective fitness function that comprehensively considers multiple performance indicators. The fitness function is designed as follows: The weighting coefficients can be dynamically adjusted. This is the penalty for constraint violation. If allocation scheme X violates any constraints, this will be a large positive number; if the constraints are fully satisfied, this will be 0. This is a penalty coefficient, ensuring that solutions that violate constraints have poor fitness. Each objective is transformed into a single-objective function through weighted summation, with the weight coefficients dynamically adjusted according to actual needs. For constraint handling, a penalty function method is used to transform the degree of constraint violation into a fitness penalty term, causing the algorithm to prioritize searching the feasible region. The design of the fitness function ensures that the algorithm evolves towards improving scheduling efficiency while satisfying various practical constraints.

[0057] S24 genetic operator implementation

[0058] The selection operator employs a strategy combining tournament selection and elite retention. Tournament selection randomly selects several individuals for comparison, retaining the most fit individuals for the next generation. This strategy ensures both selection pressure and population diversity. The elite retention strategy directly replicates the best individuals of the current generation to the next, preventing the loss of superior genes. The crossover operator is designed as a two-point crossover, randomly selecting two crossover points to exchange the task assignment information of the parent chromosomes. To address potential drone conflicts after crossover, a conflict resolution mechanism is designed to ensure the feasibility of the solution through redistribution. The mutation operator includes several strategies: single-point mutation randomly changes the drone assignment for a specific task; exchange mutation swaps the drone combination for two tasks; and local optimization mutation greedily improves the current solution. The mutation probability is adaptively adjusted, setting a higher value in the early stages of evolution to promote exploration and decreasing it later to facilitate convergence.

[0059] S25 Convergence Conditions and Elite Retention Mechanism

[0060] The algorithm sets a maximum of 200 iterations and a convergence threshold as termination conditions. When the fitness improvement of the optimal solution is less than 0.1% for 20 consecutive generations, the algorithm is considered converged and the iteration is terminated early. An elite retention mechanism ensures that the best 10% of individuals in each generation directly enter the next generation, preventing the loss of excellent solutions during evolution. A diversity preservation strategy is also introduced; when the population diversity falls below a threshold, the mutation probability is increased and random individuals are injected to avoid premature convergence. The algorithm also integrates a local search strategy, performing deep optimization on excellent individuals in the later stages of evolution to improve the quality of the solution. Through the combined effect of these mechanisms, the algorithm can converge to a high-quality solution within a reasonable time, meeting the real-time requirements of fire emergency rescue.

[0061] S26 Decoding and Scheme Output

[0062] After obtaining the optimal chromosome, a decoding operation is required to convert it into an executable task allocation scheme. The decoding process maps the gene sequences in the chromosome to specific UAV-task allocation relationships, generating a task sequence and execution schedule for each UAV. The scheme output includes key information such as a task allocation list and UAV flight paths. The system also verifies the feasibility of the scheme to ensure that all constraints are met. The final output static task scheduling scheme will serve as the basis for dynamic reallocation, providing a reliable initial action guide for fire emergency rescue.

[0063] Preferably, step S3, based on the dynamic task reallocation mechanism of global state management, includes the following steps:

[0064] S31 Global State Tracker Design

[0065] This invention designs a distributed global state tracker to maintain the system's operational status in real time. The core component is a UAV-task mapping database, which uses a hash table structure to store the real-time status information of each UAV, with an update frequency of 1Hz. The status record includes the UAV's current position coordinates. Remaining battery power Current task Simultaneously, a task status table is maintained, recording the start time, estimated completion time, actual progress, and assigned drone set for each task. A distributed consensus mechanism ensures data consistency across multiple nodes, and an alarm is automatically triggered and a recovery process initiated when an anomaly is detected.

[0066] S32 Dynamic Event Classification and Response Strategy

[0067] The system categorizes dynamic events that may occur during operation into three types and formulates corresponding response strategies for each type.

[0068] When the system receives a new task request, it first assesses the task's priority level. High-priority tasks immediately trigger a reallocation process, while low-priority tasks enter a waiting queue to await scheduling. The system monitors changes in the task queue in real time through event listeners to ensure timely responses to various task requests.

[0069] When the system receives a mission cancellation, it will return the scheduled drone to its nest.

[0070] The system employs a multi-detection mechanism to diagnose drone malfunctions, including heartbeat timeout detection, sensor data anomaly analysis, and flight trajectory deviation monitoring. Once a malfunction is confirmed, the system automatically removes the faulty drone from the available resource pool and reassigns its ongoing tasks to other available drones.

[0071] S33 Algorithm for Discovering Realizable Resources

[0072] The reassignable resource discovery algorithm designed in this invention employs a multi-stage screening mechanism, comprehensively considering multiple factors such as task priority, spatial distance, and resource status. Through a systematic evaluation process, the algorithm ensures rapid identification of the most suitable UAV resources for reassignment in dynamic environments. The algorithm first constructs an initial candidate set based on task priority comparison. The priority comparison function is defined as follows: ,in, Indicates task The higher the value, the higher the priority. For drones Location; The location of the new task; Calculate the Euclidean distance between two points; The maximum redistribution distance; This is the drone's current battery level. The minimum power required to perform the task. For the initially selected candidate drones, the algorithm further calculates the comprehensive redistribution cost. The algorithm execution process includes three main stages: first, a rapid spatial screening is performed to exclude drones that are too far away; then, a fine evaluation is performed to calculate the comprehensive cost of each candidate drone; finally, the top m optimal candidate drones are selected in ascending order of cost value and enter the redistribution queue.

[0073] S34 Compensatory Resource Allocation Strategy

[0074] The compensatory resource allocation strategy designed in this invention is a lightweight dynamic adjustment mechanism that is automatically triggered after a drone is reassigned to a new task, aiming to quickly fill resource gaps in the original task area. When the system detects a resource shortage in a certain area due to drone relocation, it first calculates the number of drones that need to be compensated using a simplified gap assessment model, i.e., by calculating the number of drones required for the original task. The number of drones remaining after redistribution The difference between the two values ​​yields the number of drones that need to be replenished. This model takes the smaller value between the number of drones that were relocated and the maximum gap between the task's requirements, ensuring that the compensation amount meets the task's needs without wasting resources.

[0075] The system then selects suitable compensation resources from a global pool of idle drones, based on three core factors: spatial proximity, capability matching, and battery adequacy. A reasonable search radius is set, prioritizing drones closer to the mission area; the matching degree between sensor configuration and mission requirements is evaluated to ensure technical feasibility; and the remaining battery power is checked to ensure it meets the minimum requirements for mission execution. This selection process is completed quickly through an efficient indexing structure, ensuring real-time performance.

[0076] The compensation allocation employs a nearest-first greedy strategy, selecting the closest eligible drone for allocation. While simple, this strategy demonstrates excellent performance in practical applications, completing decisions within milliseconds. During allocation, the system briefly considers global load balancing to prevent some drones from becoming overloaded, but avoids complex optimization calculations to maintain responsiveness.

[0077] The entire compensation mechanism is designed as a lightweight operation, ensuring that the system can quickly restore its task execution capability and maintain overall stability after resource reallocation. This strategy minimizes computational complexity while maintaining effectiveness, making it suitable for dynamic environments with high real-time requirements and effectively improving the adaptability and robustness of multi-UAV systems.

[0078] Through this efficient and simple compensation mechanism, the system can quickly adjust when resources change, ensuring the continuous execution of critical tasks and maintaining the overall load balance of the system, thus providing a reliable guarantee for the collaborative scheduling of multiple UAVs in complex environments.

[0079] Preferably, step S4 involves the construction of a simulation verification platform and algorithm verification.

[0080] S41 simulation environment configuration

[0081] This invention establishes a lightweight simulation verification platform based on Python, uses Matplotlib for visualization, and verifies algorithm performance through discrete event simulation. For simulation environment configuration, the platform constructs a 10km × 10km two-dimensional task area, sets up 10 distributed hives, and each hive is configured with 5 heterogeneous UAVs. UAV parameters, hive parameters, and task parameters are all uniformly configured and managed through JSON files. Environment modeling uses a simplified grid map, and the task area is defined by polygon vertices, supporting the dynamic addition and removal of regions.

[0082] S42 Static Scheduling Algorithm Test

[0083] Static scheduling performance testing is systematically verified through a dedicated evaluation module. The test case generation module creates task scenarios of varying scales, with three priority levels. The algorithm execution module runs the static scheduling algorithm to generate an initial task allocation scheme. The performance metric acquisition module records key data such as task completion time, resource utilization, and computational efficiency in real time. Task completion time statistics show the total time elapsed from task issuance to the last drone returning to its nest; resource utilization assesses the percentage of effective drone working time and sensor usage efficiency; and computational efficiency measures the algorithm's solution speed.

[0084] S43 Dynamic Reassignment Response Test

[0085] Dynamic event response testing is implemented through an event injection module. This module can dynamically insert preset events during simulation to simulate unexpected situations in a real environment. The test focuses on the system's response speed and stability, recording the replanning trigger time, scheme adjustment range, and system recovery status through a dedicated monitoring component. Stability assessment employs long-term runtime testing to observe the system's performance when continuously processing multiple dynamic events, detecting any performance degradation or oscillations.

[0086] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A heterogeneous UAV scheduling method based on multi-nest multi-task, characterized in that, Includes the following steps: Step S1: Construct a multi-nest heterogeneous UAV collaborative scheduling system model. Establish a unified mathematical model that includes UAVs, nests, and tasks. Define the performance parameters of UAVs, the capacity constraints of nests, and the spatiotemporal attributes of tasks. Integrate multiple conditions such as endurance constraints, sensor matching constraints, and nest capacity constraints to construct a multi-objective optimization function with the goal of minimizing task completion time and maximizing resource utilization. Step S2: Implement the static task scheduling method based on the improved multi-objective genetic algorithm. The method adopts a variable-length chromosome encoding mechanism based on task sequences, generates an initial population through heuristic and random strategies, designs a fitness function that comprehensively considers task completion time, resource utilization efficiency and constraint violation degree, uses genetic operators for iterative optimization, and outputs a high-quality static scheduling scheme. Step S3: Implement a dynamic task reallocation mechanism based on global state management. The system operation status is monitored in real time through a global state tracker. When a new task is detected or a resource failure is detected, UAV resources that can be reallocated are selected based on priority comparison and spatial distance. A compensatory resource allocation strategy is adopted to maintain system stability. Step S4: Simulation verification platform construction and algorithm verification. The task completion time and resource utilization are verified through static scheduling test, and the replanning response speed and system stability are tested through dynamic event injection test.

2. The method according to claim 1, characterized in that, Step S1 includes the following sub-steps: Step S11: Define the UAV model and configure parameters, defining the UAV set. Configure the drone's ID, type, flight time, speed, nest, location, battery level, and sensor type parameters; Step S12: Establish the nest model and capacity constraints, and define the nest set. Configure the nest number, location, capacity, types of drones that can stay, and the quantity of each type; Step S13: Define the task model and spatiotemporal attributes, and define the set of task objectives. Configure the task number, location, priority, type, required sensor type, required number of drones, and execution time window parameters; Step S14: Integrate system constraints, including UAV endurance constraints. Sensor matching constraints Task coverage constraints Nest capacity constraints Exclusivity constraints of drones Time window constraints; Step S15: Design of multi-objective optimization function, constructing the task time objective function. Path consumption time Task execution time .

3. The method according to claim 2, characterized in that, The specific parameters of the UAV model in step S11 include: Set flight speed Battery life, etc.; The nest capacity limit in step S12 includes the total capacity. ; Task priority in step S13 Divided into multiple levels, with time windows set as follows: .

4. The method according to claim 1, characterized in that, Step S2 includes the following sub-steps: Step S21: Design of chromosome coding scheme, adopting a variable-length chromosome coding mechanism based on task sequence, where the chromosome length is equal to the total number of tasks, and the gene positions are arranged in descending order of task priority; Step S22: Initial population generation strategy, which combines heuristic rules with random strategies, prioritizing resource allocation for high-priority tasks; Step S23: Construct the fitness function, designed as follows ; Step S24: Implementation of genetic operators, including tournament selection, two-point crossover, and various mutation strategies; Step S25: Set convergence conditions, maximum number of iterations, and terminate early if the fitness is low for several consecutive generations; Step S26: Decoding and Scheme Output: Decode the chromosome into an executable task allocation scheme.

5. The method according to claim 4, characterized in that, In step S25: Set the proportion of the population size to be selected for the tournament; Set the crossover probability at two points; Set the mutation probability in the early and late stages.

6. The method according to claim 1, characterized in that, Step S3 includes the following sub-steps: Step S31: Design a global state tracker, using a hash table structure to store the drone-task mapping relationship and setting the update frequency; Step S32: Implement dynamic event classification and response, classifying events into new task triggering, new task cancellation, and drone malfunction; Step S33: Execute the reallocatable resource discovery algorithm and filter candidate UAVs based on the priority comparison function; Step S34: Implement a compensatory resource allocation strategy and calculate the resource gap.

7. The method according to claim 6, characterized in that, In step S33: Set maximum redistribution distance ; Priority comparison uses a strict inequality, and reallocation is only triggered when a new task has a higher priority; The distance function D is calculated using Euclidean distance.

8. The method according to claim 1, characterized in that, Step S4 includes the following sub-steps: Step S41: Configure the simulation environment, construct a two-dimensional task region, and set up multiple distributed nests; Step S42: Static scheduling algorithm test to verify task completion time, resource utilization and computational efficiency; Step S43: Dynamic reallocation response test, inject dynamic tasks and verify whether the system responds.

9. An intelligent multi-UAV cooperative task execution scheduling system for implementing the method of any one of claims 1-8, characterized in that, include: A modeling module for a multi-nest heterogeneous UAV collaborative scheduling system is used to construct a unified mathematical model; An improved static scheduling module for the multi-objective genetic algorithm is used to generate an initial task allocation scheme. The global state management and dynamic reallocation module is used to handle dynamic events and resource reallocation. The simulation verification platform module is used for performance testing and evaluation.

10. The system according to claim 9, characterized in that, The system also includes: The visualization module is used to display the drone's location, task allocation status, and performance metrics in real time. The data recording and analysis module is used to store runtime data and generate performance reports.

Images