A multi-uav flight task reassignment optimization method and system

Abstract

The application discloses a multi-unmanned aerial vehicle flight task reassignment optimization method and system, and belongs to the technical field of unmanned aerial vehicle intelligent control. In order to solve the problem of low task matching efficiency in multi-unmanned aerial vehicle task assignment, the original task assignment result of a fault unmanned aerial vehicle is inquired to find all tasks undertaken by the fault unmanned aerial vehicle, and an initial failure task set is formed; all candidate unmanned aerial vehicles with high fusion scores of the initial failure tasks and not currently executing the initial failure tasks are found by traversing the unmanned aerial vehicle set, and a propagation unmanned aerial vehicle set is obtained; the tasks currently executed by the candidate unmanned aerial vehicles are inquired, the influence of the initial failure tasks on the currently executed tasks is considered, and a propagation task set is constructed; a participation unmanned aerial vehicle set of task reassignment is obtained; a candidate combination set of task reassignment is formed, then each combination in the candidate combination set is scored again, and an optimal matching is performed by using a task assignment algorithm, so that a multi-unmanned aerial vehicle flight task reassignment optimization scheme is obtained.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of unmanned aerial vehicle intelligent control, and particularly relates to a multi-unmanned aerial vehicle flight task reassignment optimization method and system. BACKGROUND

[0002] As an important carrier of low-altitude economy, unmanned aerial vehicles have been widely used in logistics transportation, emergency rescue, city inspection, agricultural plant protection, environmental monitoring and other industries, and have become a key technical equipment for promoting intelligent aerial operation and efficient use of low-altitude resources. In the future, with the continuous evolution of flight control systems, communication navigation, artificial intelligence and other technologies, unmanned aerial vehicles will develop from single-machine operation to multi-machine cooperation and autonomous decision-making, becoming a basic execution unit in the low-altitude economy system.

[0003] In the process of multi-unmanned aerial vehicle cooperation in executing complex tasks, sudden situations such as task interruption, unmanned aerial vehicle failure, insufficient power or communication interruption often occur, resulting in the inability to complete some tasks as planned. In order to ensure the stability and task completion rate of the overall task system, it is necessary to timely perform task reassignment and reassign the affected tasks to other available unmanned aerial vehicles for execution. However, most existing methods use global recalculation strategies, lack precise identification of the impact range of tasks, resulting in high computational complexity and slow response speed. At the same time, the task reassignment process often ignores changes in task priority and timeliness requirements, making it difficult to adapt to dynamic, real-time and high-risk application scenarios. Therefore, there is an urgent need for an efficient multi-unmanned aerial vehicle task reassignment optimization method that can adapt to changes in flight state, in order to improve the real-time response capability and overall scheduling efficiency of unmanned aerial vehicle task assignment. SUMMARY

[0004] The application aims to solve the problems of insufficient task matching efficiency and real-time response in multi-unmanned aerial vehicle task assignment, and proposes a multi-unmanned aerial vehicle flight task reassignment optimization method and system.

[0005] To achieve the above-mentioned purpose, the application realizes the following technical solutions:

[0006] A multi-unmanned aerial vehicle flight task reassignment optimization method, comprising the following steps:

[0007] S1. Collecting key attribute data of unmanned aerial vehicles to construct a set of unmanned aerial vehicles, and collecting task data of unmanned aerial vehicles to construct a set of tasks of unmanned aerial vehicles;

[0008] S2. When a certain unmanned aerial vehicle fails, querying the original task assignment result of the failed unmanned aerial vehicle, finding all tasks undertaken by the failed unmanned aerial vehicle, and constituting an initial invalid task set;

[0009] S3. Based on the initial failed task set obtained in step S2, traverse the drones in the drone set to find all candidate drones with high fusion scores for the initial failed tasks and which have not yet executed the initial failed tasks, and obtain the propagation drone set.

[0010] S4. For the candidate drones in the propagation drone set obtained in step S3, find the tasks currently being executed by the candidate drones, consider the impact of the initial failed task on the currently executed task, and construct a propagation task set;

[0011] S5. Merge the initial set of failed tasks obtained in step S1 and the set of propagated tasks obtained in step S4 into a complete set of affected tasks. Then find all the drones that are assigned to the complete set of affected tasks to obtain the set of drones participating in task redistribution.

[0012] S6. Pair the tasks in the complete set of influencing tasks obtained in step S5 with the drones in the set of participating drones for task redistribution to form a candidate combination set for task redistribution. Then, re-score each combination in the candidate combination set and use the task allocation algorithm to perform optimal matching to obtain the task redistribution optimization scheme for multi-drone flight.

[0013] Furthermore, in step S2, the k-th drone If a failure occurs, the initial set of failed tasks will be... The expression is:

[0014]

[0015] in, This is the initial set of failed tasks. In the original task allocation Execute The corresponding task, with a value of 1, indicates execution. Let be the set of key attributes for the j-th task.

[0016] Furthermore, the expression for the propagation drone set in step S3 is:

[0017]

[0018] in, For the union operator, Indicates all that belong to the set The task in Perform a union operation; A fusion score for drone missions; Intersection symbol; This indicates that the task is not currently being performed. The scoring threshold; To propagate the UAV set.

[0019] Further, the specific calculation method of the fusion score value of the UAV performing the task in step S3 includes the following steps:

[0020] S3.1. Calculate the total cost of the UAV performing the task by considering the time required for the flight of the UAV, energy consumption and the degree of danger on the path, and then construct a feasibility constraint discrimination method for the UAV and the task of the UAV to screen the UAV and the task pair of the UAV that meet the feasibility constraint;

[0021] S3.2. Establish a dynamic updating mechanism for the priority of the task of the UAV, update the initial priority, and obtain the priority of the task;

[0022] S3.3. For the UAV and the task pair of the UAV that meet the feasibility constraint obtained in step S3.1, construct a rule score function model of the UAV performing the task by considering the energy utilization efficiency under the unit execution cost, the matching degree between the current load capacity of the UAV and the required load of the task, the priority of the task, and the energy and speed adaptation efficiency in the unit task time;

[0023] S3.4. Construct a score function model of the UAV performing the task based on a deep neural network, extract deep-level correlations between the state of the UAV and the attributes of the task by learning historical samples, and then fuse the rule score function model of the UAV performing the task obtained in step S3.3 to obtain the fusion score of the UAV performing the task.

[0024] Further, the expression of the propagated task set in step S4 is:

[0025]

[0026] Wherein, is the propagated task set; is the key attribute set of the i-th task; is the existence quantifier, indicates that the UAV belongs to the set; indicates that the UAV is currently performing the corresponding task.

[0027] Further, the specific implementation method of step S5 includes the following steps:

[0028] S5.1. Merge the initial invalid task set obtained in step S1 and the propagated task set obtained in step S4 to obtain a complete impact task set, and the expression is: ​

[0029] ;

[0030] S5.1. Based on the complete influence task set, extract the task reassignment participating UAV set, expressed as:

[0031]

[0032] Wherein, is the task reassignment participating UAV set; indicates that there is a certain The corresponding task belongs to Set; Indicates that the task j has been assigned to the UAV i.

[0033] Further, the specific implementation method of step S6 includes the following steps:

[0034] S6.1. Pair the tasks in the complete influence task set obtained in step S5 and the UAVs in the task reassignment participating UAV set, to form a task reassignment candidate combination set, expressed as:

[0035]

[0036] Wherein, is the task reassignment candidate combination set;

[0037] S6.2. Score each combination in the task reassignment candidate combination set obtained in step S6.1, and use the greedy matching or linear programming algorithm or the Hungarian algorithm to obtain the optimal matching, to obtain the local allocation result;

[0038] S6.3. Merge the local allocation result obtained in step S6.2 back into the original task allocation result to form the final allocation scheme; the merging rule is as follows:

[0039]

[0040] Wherein, is the final task allocation result; is the original task allocation; is the optimized local result.

[0041] A multi-UAV flight task reassignment optimization method system, comprising a processor, a memory and a computer program stored in the memory and executable on the processor, the computer program realizes the steps of the multi-UAV flight task reassignment optimization method when running.

[0042] The beneficial effects of the present application are:

[0043] The multi-unmanned aerial vehicle flight task reassignment optimization method provided by the application solves the problem of large influence range and high calculation cost of task reassignment under the condition of unmanned aerial vehicle failure. The application designs a task influence recognition and local task reassignment mechanism, which can quickly recognize the affected unmanned aerial vehicles and tasks when the flight state or task state changes, and only update the score and distribution in a local range, thereby avoiding the high calculation cost and system instability problems caused by global reassignment.

[0044] The multi-unmanned aerial vehicle flight task reassignment optimization method provided by the application has high efficiency and stability. The local task reassignment and dynamic priority adjustment mechanism is adopted, only the minimum range adjustment is performed on the affected area when the task or flight state changes, which significantly reduces the calculation complexity and system disturbance, and guarantees the continuity and reliability of the multi-unmanned aerial vehicle system task execution. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The flowchart of the multi-unmanned aerial vehicle flight task reassignment optimization method provided by the application is shown in the figure.

[0046] Figure 2 The final score column chart corresponding to the task reassignment result of the application is shown in the figure. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the application more clear and understandable, the application will be further described in detail below in combination with the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and are not used to limit the application, that is, the described specific embodiments are only a part of the embodiments of the application, but not all the specific embodiments. The components of the specific embodiments of the application described and shown in the drawings herein can be arranged and designed in various different configurations, and the application can also have other embodiments.

[0048] Therefore, the detailed description of the specific embodiments of the application provided in the drawings below is not intended to limit the scope of the claimed application, but only represents selected specific embodiments of the application. Based on the specific embodiments of the application, all other specific embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0049] In order to further understand the inventive content, characteristics and effects of the application, the following specific embodiments are exemplified, and the drawings are Figure 1 -attached Figure 2 The detailed description is as follows:

[0050] Example 1:

[0051] A multi-UAV flight task reassignment optimization method, comprising the following steps:

[0052] S1. Collecting the key attribute data of the UAVs to construct a UAV set, and collecting the task data of the UAVs to construct a task set of the UAVs;

[0053] S2. When a UAV fails, querying the original task assignment result of the failed UAV, finding all tasks undertaken by the failed UAV, and constituting an initial failed task set;

[0054] Further, in step S2, the kth UAV fails, and the expression of the initial failed task set is:

[0055]

[0056] wherein, is the initial failed task set, is whether the corresponding task is executed in the original task assignment, equal to 1 indicating execution, is the key attribute set of the jth task. S3. Based on the initial failed task set obtained in step S2, traversing the UAVs in the UAV set, finding all candidate UAVs with high fusion score for the initial failed task and not currently executing the initial failed task, and obtaining a propagation UAV set;

[0057] Further, the expression of the propagation UAV set in step S3 is:

[0058]

[0059]

[0060] wherein, is a union operator, indicates that the tasks in the set are subjected to a union operation; is a fusion score value of the UAV executing the task; is an intersection symbol; indicates that the task is not currently executed; is a score threshold value; is the propagation UAV set. Further, the specific calculation method of the fusion score value of the UAV executing the task in step S3 comprises the following steps:

[0061]

[0062] ​​​S3.1. Calculate the total cost of the UAV mission by considering the flight time, energy consumption, and danger level along the path. Then, construct a feasibility constraint judgment method for UAVs and UAV missions, and screen UAVs and UAV mission pairs that meet the feasibility constraints.

[0063] Furthermore, the specific implementation method includes the following steps:

[0064] S3.1.1. Collect key attribute data of UAVs to construct a UAV set, and collect UAV mission data to construct a UAV mission set;

[0065] Furthermore, the key attribute data of the drones collected includes the current location of the i-th drone. Remaining battery power Flight speed Maximum load capacity The mission data collected from the UAV includes the target location of the j-th mission. Initial priority Time required Required load capacity .

[0066] Furthermore, each UAV is represented by a state vector, which includes key attributes such as spatial location, battery power, payload capacity, and speed; each task includes key parameters such as target location, time window, payload requirements, and priority. This modeling serves as the information input basis for the entire task allocation.

[0067] drone collection Recorded as: ;

[0068] in, Let i be the set of key attributes of the i-th drone. The current location, remaining battery power, flight speed, and total number of drones can be obtained by the flight control system; the maximum payload capacity is preset; the target location, initial priority, mission time, and mission payload are all directly given by the mission system.

[0069] S3.1.2. By integrating the flight time, energy consumption, and hazard levels along the path of the UAV, the total cost of the UAV's mission is comprehensively assessed, resulting in the following expression:

[0070]

[0071] in, The total cost of performing the j-th task for the i-th drone; The time required for the i-th drone to perform the j-th task; The path risk value for the i-th drone performing the j-th task. For the time derivative; , , These are the flight time weighting coefficient, energy consumption weighting coefficient, and path risk weighting coefficient, respectively. , , , Determined by expert experience; ;

[0072] Flight time reflects efficiency, speed squared term reflects energy consumption, and risk value characterizes safety. The three are flexibly adjusted through weighting coefficients to finally arrive at a comprehensive cost assessment value.

[0073] S3.1.3. Consider three constraints to construct a feasibility constraint judgment method for UAVs and UAV missions. The first constraint is whether the remaining battery power of the UAV is sufficient to complete the path cost. The second constraint is whether the maximum payload capacity of the UAV meets the mission requirements. The third constraint is whether the UAV can complete the execution within the time required for the mission at the current time. Select UAVs and UAV mission pairs that simultaneously meet the three constraints.

[0074] Furthermore, only task pairs that simultaneously satisfy the above three types of constraints will proceed to the next stage of the scoring and allocation process to ensure the physical feasibility and time accessibility of the task allocation results. At the same time, the method has been optimized by introducing factors such as path risk discount, power utilization efficiency, and task time window smoothing penalty, making the judgment process more reasonable and continuous, avoiding false rejection or incorrect allocation of tasks due to fluctuations in boundary conditions, and significantly improving the robustness and fault tolerance of task scheduling.

[0075] The feasibility determination formula is as follows:

[0076]

[0077] in, The remaining battery power of the i-th drone is obtained by the flight control system; The path risk value for the i-th drone performing the j-th task; This represents the maximum value of the path risk. The power safety factor is determined by expert experience; The total cost of performing the j-th task for the i-th drone; Let be the maximum payload capacity of the i-th drone; The required load for the j-th task; The time required for the j-th task; Let be the current position of the i-th drone. Let be the target location of the j-th task. Let be the flight speed of the i-th drone. It is a tiny constant, determined by expert experience. The carrying capacity threshold is determined by expert experience; The time matching threshold is determined by expert experience;

[0078] Furthermore, It involves determining the feasible power supply, which must meet the task execution cost while also considering path risks and the urgency of the task time. This involves assessing the carrying capacity. The drone's carrying capacity should be proportional to the payload required for the mission, which can improve the model's fault tolerance and flexibility and avoid boundary misjudgments. It is a time-matching judgment to determine whether the drone can complete the mission within the required time.

[0079] S3.2. Establish a dynamic priority update mechanism for UAV missions, update the initial priority, and obtain the mission priority;

[0080] Furthermore, the priority dynamic update mechanism for UAV missions in step S3.2 uses two key indicators—joint mission delay and urgency—to dynamically update the initial mission priority. The expression is:

[0081]

[0082] in, Let the priority of the j-th task be... , These are the delay time coefficient and the urgency index coefficient, respectively, determined by expert experience; Let be the delay time for the j-th task, which is obtained by comparing the task's required deadline with the current time; As an indicator of urgency, .

[0083] Furthermore, task priorities can dynamically evolve over time during the task allocation cycle. In particular, for tasks that are already delayed or about to expire, their urgency level should be appropriately increased.

[0084] S3.3. For the UAVs and UAV task pairs that satisfy the feasibility constraints obtained in step S3.1, construct a rule scoring function model for UAV task execution, considering the energy utilization efficiency per unit execution cost, the degree of matching between the current payload capacity of the UAV and the payload required by the task, the task priority, and the energy and speed adaptation efficiency per unit task time.

[0085] Furthermore, step S3.3 constructs an explicit scoring function model for UAV mission execution, introducing nonlinear functions and combined feature factors to enhance the comprehensive expression of UAV energy consumption, payload capacity, priority, and urgency. The resulting expression for the rule-based scoring of UAV mission execution is:

[0086]

[0087] in, The rule score for the i-th drone to perform the j-th task; The task priority of the j-th task; The energy efficiency coefficient. This is the load matching coefficient. This is the priority coefficient. To achieve a comprehensive fit coefficient, It is a constant. ~ and Determined by expert experience; As a comprehensive work efficiency indicator; .

[0088] Furthermore, the explicit scoring function model for UAV mission execution not only considers the energy efficiency ratio per unit cost but also introduces a logarithmic weighting of power consumption on mission cost and a nonlinear mapping for payload matching. Simultaneously, to better reflect mission adaptability and execution efficiency, a composite feature exponential function fusing speed, power consumption, and required time is designed. This model is structurally more discriminative and robust, capable of adapting to the mission allocation decision-making needs in complex dynamic environments. In the above equation, Used to characterize energy utilization efficiency per unit execution cost; Used to characterize the degree of matching between the current payload capacity of the drone and the payload required for the mission; Used to characterize task priority; It comprehensively reflects the energy-speed adaptation efficiency within a unit of task time, demonstrating the overall execution capability of the UAV for the current task.

[0089] S3.4. Construct a scoring function model for UAV task execution based on deep neural networks. By learning from historical samples, extract the deep-level correlation between UAV state and task attributes, and then fuse it with the rule-based scoring function model for UAV task execution obtained in step S3.3 to obtain the fused score for UAV task execution.

[0090] Furthermore, the specific implementation method of step S3.4 includes the following steps:

[0091] S3.4.1. Define the UAV-mission feature vector as follows:

[0092] ;

[0093] in, The drone-task feature vector for the i-th drone performing the j-th task;

[0094] S3.4.2. After normalizing the parameters in the UAV-mission feature vector using a numerical normalization method, the normalized parameters are input into the constructed deep neural network-based UAV mission execution scoring function model. The output is the deep neural network-based UAV mission execution score, expressed as:

[0095]

[0096] in, The output score value of the scoring function model for a drone performing a task based on a deep neural network; It is a multilayer perceptron neural network.

[0097] Furthermore, considering that traditional explicit functions are insufficient to characterize nonlinear task allocation strategies in complex environments, a scoring prediction model based on deep neural networks is introduced. This model learns from historical samples to extract deep-level correlations between UAV states and task attributes, thereby estimating the potential rewards for task completion. This model can adapt to different scenario policy preferences, improving the task allocation method's adaptability to high-dimensional and variable inputs.

[0098] To facilitate understanding, the specific implementation process of the model is explained as follows:

[0099] (1) Feature engineering and data preprocessing.

[0100] To ensure the effectiveness of model training and its generalization ability, systematic preprocessing of the original feature data is necessary. This step primarily addresses the problem of imbalanced data, providing high-quality training samples for subsequent deep learning models.

[0101] The input feature vector is ;

[0102] Using numerical normalization methods, Scale to the [0, 1] interval;

[0103] (2) Neural network structure design

[0104] The architecture of a neural network directly impacts its expressive power and learning performance. Through a multi-layer perceptron structure, deep correlations between features are extracted layer by layer, ultimately outputting a standardized score. Each layer of the network is designed to balance model complexity and computational efficiency.

[0105] The architecture of a multilayer perceptron neural network is as follows:

[0106] 1. Input layer: 10 neurons, corresponding to ;

[0107] 2. Hidden layer 1: 32 neurons, ReLU activation function;

[0108] 3. Hidden layer 2: 16 neurons, ReLU activation function;

[0109] 4. Output layer: 1 neuron, corresponding to , Sigmoid activation function.

[0110] The inter-layer connections adopt a fully connected approach.

[0111] (3) Model training process

[0112] Model training is the core of the entire learning process, requiring the appropriate setting of training parameters and the establishment of an effective monitoring mechanism. Through well-designed training strategies, it is essential to ensure that the model can fully learn the patterns in historical data while avoiding overfitting.

[0113] The training parameters for the model are set as follows:

[0114] 1. Batch size: 64;

[0115] 2. Learning rate: 0.001, using the Adam optimizer;

[0116] 3. Training rounds: 100, using an early stop strategy;

[0117] 4. Loss function: Binary cross-entropy;

[0118] Once the model is trained, the following can be achieved: [parameter settings] The input is fed into a multilayer perceptron neural network, and the output is... .

[0119] S3.4.3. Calculate the fusion score of the UAV's mission execution, expressed as:

[0120]

[0121] in, A fusion score for drone missions; The fusion ratio is determined by expert experience.

[0122] S3.4.4. Based on the obtained fusion score of the UAVs performing tasks, the task allocation problem of multi-UAV flights is transformed into solving the problem of maximizing the total score of all UAV task pairs, thus obtaining the task allocation optimization scheme for multi-UAV flights.

[0123] Furthermore, the task allocation for multi-UAV flights is modeled as a constrained optimization problem, with the goal of maximizing the total score of all assigned task pairs. By constructing binary allocation variables and imposing resource and uniqueness constraints, the optimal solution is obtained by employing greedy matching, linear programming, or the Hungarian algorithm.

[0124] The optimization goals are as follows:

[0125]

[0126] The constraints are as follows:

[0127]

[0128] in, This indicates whether task j should be assigned to drone i. If yes, the value is 1; otherwise, the value is 0. This represents the maximum mission capacity limit for the i-th drone, determined by expert experience.

[0129] S4. For the candidate drones in the propagation drone set obtained in step S3, find the tasks currently being executed by the candidate drones, consider the impact of the initial failed task on the currently executed task, and construct a propagation task set;

[0130] Furthermore, the expression for the propagation task set in step S4 is:

[0131]

[0132] in, For the dissemination of a set of tasks; For the first A set of key attributes for each task; It is an existential quantifier. This indicates the presence of drones. belong gather; Indicates drone Currently in progress The corresponding task.

[0133] S5. Merge the initial set of failed tasks obtained in step S1 and the set of propagated tasks obtained in step S4 into a complete set of affected tasks. Then find all the drones that are assigned to the complete set of affected tasks to obtain the set of drones participating in task redistribution.

[0134] Furthermore, the specific implementation method of step S5 includes the following steps:

[0135] S5.1. The initial set of failure tasks obtained in step S1 and the set of propagation tasks obtained in step S4 are combined to form the complete set of influencing tasks, expressed as:

[0136] ;

[0137] S5.1. Based on the complete set of impacting tasks, extract the set of participating UAVs for task redistribution, expressed as:

[0138]

[0139] in, The ensemble of participating drones for task reassignment; This indicates the existence of a certain The corresponding task belongs to gather; This indicates that task j has been assigned to drone i.

[0140] S6. Pair the tasks in the complete set of influencing tasks obtained in step S5 with the drones in the set of participating drones for task redistribution to form a candidate combination set for task redistribution. Then, re-score each combination in the candidate combination set and use the task allocation algorithm to perform optimal matching to obtain the task redistribution optimization scheme for multi-drone flight.

[0141] Furthermore, the specific implementation method of step S6 includes the following steps:

[0142] S6.1. Pair the tasks in the complete set of influencing tasks obtained in step S5 with the drones in the set of participating drones for task reassignment to form a candidate combination set for task reassignment, expressed as:

[0143]

[0144] in, A set of candidate combinations for task reassignment;

[0145] S6.2. For each combination in the candidate combination set for task reassignment obtained in step S6.1, score it and use a greedy matching algorithm, linear programming algorithm, or Hungarian algorithm to perform optimal matching and obtain the local allocation result;

[0146] S6.3. Merge the local allocation results obtained in step S6.2 back into the original task allocation results to form the final allocation scheme; the merging rules are as follows:

[0147]

[0148] in, Assign results to the final task; Assign the original task; This is the optimized local result.

[0149] Based on this embodiment, the practical application is as follows:

[0150] Given the initial drone task assignments as shown in Table 1, if drone u3 malfunctions, its assigned task t2 cannot continue. In this case, task t2 needs to be reassigned, and the propagation-affected tasks (i.e., other tasks originally performed by the propagating drone) need to be identified. A local task reassignment optimization model is constructed, calculating the initial set of failed tasks, the set of propagating drones, the set of propagating tasks, the set of all tasks requiring reassignment, and the set of combinations requiring re-evaluation. Scoring is then recalculated, and task reassignment is performed. The final task reassignment results are shown in Table 2. Figure 2 As shown:

[0151] Table 1

[0152]

[0153] Table 2

[0154]

[0155] Example 2:

[0156] A system for optimizing task reassignment in multi-UAV flight includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements the steps of the task reassignment optimization method for multi-UAV flight as described in Embodiment 1.

[0157] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0158] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for optimizing task redistribution during multi-UAV flights, characterized in that, Includes the following steps: S1. Collect key attribute data of drones to construct a drone set, and collect task data of drones to construct a task set of drones; S2. When a drone malfunctions, query the original task assignment results of the malfunctioning drone, find all the tasks undertaken by the malfunctioning drone, and form an initial set of failed tasks; S3. Based on the initial failed task set obtained in step S2, traverse the drones in the drone set to find all candidate drones with high fusion scores for the initial failed tasks and which have not yet executed the initial failed tasks, and obtain the propagation drone set. The specific calculation method for the fusion score of the UAV's mission in step S3 includes the following steps: S3.

1. Calculate the total cost of the UAV mission by considering the flight time, energy consumption, and danger level along the path. Then, construct a feasibility constraint judgment method for UAVs and UAV missions, and screen UAVs and UAV mission pairs that meet the feasibility constraints. S3.

2. Establish a dynamic priority update mechanism for UAV missions, update the initial priority, and obtain the mission priority; S3.

3. For the UAVs and UAV task pairs that satisfy the feasibility constraints obtained in step S3.1, construct a rule scoring function model for UAV task execution, considering the energy utilization efficiency per unit execution cost, the degree of matching between the current payload capacity of the UAV and the payload required by the task, the task priority, and the energy and speed adaptation efficiency per unit task time. S3.

4. Construct a scoring function model for UAV task execution based on deep neural networks. By learning from historical samples, extract the deep-level correlation between UAV state and task attributes. Then, fuse it with the rule-based scoring function model for UAV task execution obtained in step S3.3 to obtain the fused score for UAV task execution. S4. For the candidate drones in the propagation drone set obtained in step S3, find the tasks currently being executed by the candidate drones, consider the impact of the initial failed task on the currently executed task, and construct a propagation task set; S5. Merge the initial set of failed tasks obtained in step S1 and the set of propagated tasks obtained in step S4 into a complete set of affected tasks. Then find all the drones that are assigned to the complete set of affected tasks to obtain the set of drones participating in task redistribution. S6. Pair the tasks in the complete set of influencing tasks obtained in step S5 with the drones in the set of participating drones for task redistribution to form a candidate combination set for task redistribution. Then, re-score each combination in the candidate combination set and use the task allocation algorithm to perform optimal matching to obtain the task redistribution optimization scheme for multi-drone flight.

2. The task redistribution optimization method for multi-UAV flight according to claim 1, characterized in that, The k-th drone in step S2 If a failure occurs, the initial set of failed tasks will be... The expression is: ； in, This is the initial set of failed tasks. In the original task allocation Execute The corresponding task, with a value of 1, indicates execution. Let be the set of key attributes for the j-th task.

3. A task redistribution optimization method for multi-UAV flight according to claim 1 or 2, characterized in that, The expression for the propagation drone set in step S3 is: ； in, For the union operator, Indicates all that belong to the set The task in Perform a union operation; A fusion score for drone missions; Intersection symbol; This indicates that the task is not currently being performed. The scoring threshold; To spread information about drone assemblies; u i This refers to drones.

4. The task redistribution optimization method for multi-UAV flight according to claim 3, characterized in that, The expression for the propagation task set in step S4 is: ； in, For the dissemination of a set of tasks; For the first A set of key attributes for each task; It is an existential quantifier. This indicates the presence of drones. belong gather; Indicates drone Currently in progress The corresponding task.

5. The task redistribution optimization method for multi-UAV flight according to claim 4, characterized in that, The specific implementation method of step S5 includes the following steps: S5.

1. The initial set of failure tasks obtained in step S1 and the set of propagation tasks obtained in step S4 are combined to form the complete set of influencing tasks, expressed as: ； S5.

2. Based on the complete set of impacting tasks, extract the set of participating UAVs for task redistribution, expressed as: ； in, The ensemble of participating drones for task reassignment; This indicates the existence of a certain The corresponding task belongs to gather; This indicates that task j has been assigned to drone i.

6. The task redistribution optimization method for multi-UAV flight according to claim 5, characterized in that, The specific implementation method of step S6 includes the following steps: S6.

1. Pair the tasks in the complete set of influencing tasks obtained in step S5 with the drones in the set of participating drones for task reassignment to form a candidate combination set for task reassignment, expressed as: ； in, A set of candidate combinations for task reassignment; S6.

2. For each combination in the candidate combination set for task reassignment obtained in step S6.1, score it and use a greedy matching algorithm, linear programming algorithm, or Hungarian algorithm to perform optimal matching and obtain the local allocation result; S6.

3. Merge the local allocation results obtained in step S6.2 back into the original task allocation results to form the final allocation scheme; the merging rules are as follows: ； in, Assign results to the final task; Assign the original task; This is the optimized local result.

7. A system for optimizing task redistribution in multi-UAV flight, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed, implements the steps of a task reassignment optimization method for multi-UAV flight as described in any one of claims 1-6.

Images