Distributed combat mission scheduling method and system for unmanned equipment

Abstract

The invention discloses a distributed combat task scheduling method and system for unmanned equipment, and relates to the technical field of task scheduling. The method comprises the steps of evaluating execution complexity and task importance according to task attribute information of a target job task, and correcting a reference scheduling constraint condition to generate an adaptive scheduling constraint condition; based on the comprehensive load state data and the position coordinates of the unmanned equipment execution nodes, taking adaptive scheduling constraint conditions as scheduling quality constraints, taking task remaining time windows as scheduling response constraints, and taking the number of equipment types required by tasks as equipment demand constraints, performing random scheduling simulation to generate a plurality of qualified equipment scheduling schemes; establishing a multi-objective optimization function to evaluate a scheme scheduling quality index; and optimizing the qualified equipment scheduling scheme by taking the maximum scheme scheduling quality index as an optimization target, outputting an optimal equipment scheduling scheme, and performing scheduling control. According to the invention, dynamic balanced scheduling and collaborative allocation of unmanned equipment cluster tasks are realized.

Description

A distributed combat mission scheduling method and system for unmanned equipment Technical Field

[0001] This invention relates to the field of mission scheduling technology, and more specifically to a distributed combat mission scheduling method and system for unmanned equipment. Background Technology

[0002] Traditional single-node unmanned equipment mission scheduling systems struggle to support parallel task processing for multiple types of large-scale unmanned equipment clusters. When multiple types of combat missions occur concurrently within an unmanned equipment cluster, single-node scheduling systems are prone to issues such as instruction issuance delays, mission conflict blocking, and chaotic cluster coordination. Furthermore, single-node scheduling architectures are susceptible to operational losses; if a node is destroyed, the entire unmanned equipment scheduling link will be interrupted, causing the unmanned combat cluster to fall into a state of loss of control.

[0003] To alleviate the bottleneck caused by single-node scheduling, many existing technologies have proposed distributed scheduling schemes, but they still have significant shortcomings. First, scheduling strategies generally lack specific design for the characteristics of unmanned equipment, often employing generic load balancing methods that cannot effectively adapt to the unique priorities, timeliness, tactical relevance, and swarm coordination requirements of unmanned equipment combat missions. Second, the collection of unmanned equipment load status is not comprehensive enough, failing to incorporate unmanned equipment-specific indicators such as the status of mounted equipment and the quality of remote control links, leading to biased judgments on load balancing. Simultaneously, the mission fault tolerance mechanisms of existing solutions are not compatible with the operational rules of unmanned equipment, lacking the adaptability to rapidly transfer missions, replenish redundant equipment, and dynamically prioritize urgent missions when dealing with damaged or lost unmanned equipment, making it difficult to cope with the complex and ever-changing battlefield situations in unmanned equipment swarm operations. Summary of the Invention

[0004] This invention addresses the technical problems of existing distributed scheduling schemes in unmanned equipment combat scenarios, such as poor tactical adaptability, inaccurate load status judgment, and insufficient fault tolerance, and provides a distributed combat mission scheduling method and system for unmanned equipment.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: Firstly, this invention provides a distributed combat mission scheduling method for unmanned equipment, comprising: on a centralized scheduling management platform, evaluating and obtaining the execution complexity and importance of the target task based on the task attribute information of the target task; dynamically correcting preset baseline scheduling constraints to generate adaptive scheduling constraints; based on the comprehensive load status data and location coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, the remaining time window of the task as scheduling response constraints, and the type and quantity of equipment required by the task as equipment demand constraints, performing random scheduling simulation on the several unmanned equipment execution nodes to generate multiple qualified equipment scheduling schemes; establishing a multi-objective optimization function to evaluate the scheduling quality of the multiple qualified equipment scheduling schemes and determine multiple scheme scheduling quality indices; with maximizing the scheme scheduling quality index as the optimization objective, using an optimization algorithm to optimize the scheduling of the multiple qualified equipment scheduling schemes based on the multiple scheme scheduling quality indices, outputting the optimal equipment scheduling scheme, and performing scheduling control on the several unmanned equipment execution nodes.

[0006] Secondly, this invention provides a distributed combat mission scheduling system for unmanned equipment, comprising: a constraint analysis module, used to evaluate the execution complexity and importance of the target mission based on the mission attribute information of the target mission on a centralized scheduling management platform, dynamically correct preset baseline scheduling constraints, and generate adaptive scheduling constraints; a scheme generation module, used to perform random scheduling simulation on the several unmanned equipment execution nodes based on the comprehensive load status data and location coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, the remaining time window of the mission as scheduling response constraints, and the type and quantity of equipment required by the mission as equipment demand constraints, to generate multiple qualified equipment scheduling schemes; a multi-objective optimization evaluation module, used to establish a multi-objective optimization function, evaluate the scheduling quality of the multiple qualified equipment scheduling schemes, and determine multiple scheme scheduling quality indices; and an optimization and scheduling control module, used to optimize the multiple qualified equipment scheduling schemes based on the multiple scheme scheduling quality indices using an optimization algorithm, output the optimal equipment scheduling scheme, and perform scheduling control on the several unmanned equipment execution nodes, with the goal of maximizing the scheme scheduling quality index.

[0007] The beneficial effects of this invention are as follows: Compared with existing technologies, firstly, this invention improves the accuracy of adapting scheduling strategies to combat scenarios by evaluating the execution complexity and importance of tasks and dynamically correcting baseline scheduling constraints. Secondly, by combining random scheduling simulation with multi-objective optimization, multiple qualified schemes are generated and their quality is evaluated under multiple constraints, overcoming the resource allocation imbalance problem of traditional single-index scheduling and achieving global optimization of combat resources. Thirdly, by using the scheme scheduling quality index as the optimization objective and employing optimization algorithms for iterative optimization, it can quickly converge to the optimal scheme, effectively addressing dynamic changes in battlefield situation and time urgency requirements. Finally, by outputting the optimal scheduling scheme and precisely controlling execution nodes, this invention achieves load balancing for different types of unmanned equipment while ensuring mission success rate, enhancing the reliability and stability of unmanned equipment clusters executing continuous combat missions. Attached Figure Description

[0008] Figure 1 is a flowchart illustrating the distributed combat mission scheduling method for unmanned equipment provided by the present invention; Figure 2 is a structural diagram illustrating the distributed combat mission scheduling system for unmanned equipment provided by the present invention.

[0009] In the attached diagram, the components represented by each number are as follows: constraint analysis module 11, scheme generation module 12, multi-objective optimization evaluation module 13, optimization and scheduling control module 14. Detailed Implementation

[0010] Example 1, as shown in Figure 1, provides a distributed combat mission scheduling method for unmanned equipment, including: S10: In a centralized scheduling management platform, the execution complexity and task importance of the target task are evaluated and obtained based on the task attribute information of the target task, and the preset baseline scheduling constraints are dynamically corrected to generate adaptive scheduling constraints; First, in the centralized scheduling management platform, the execution complexity and task importance of the target task are evaluated and obtained based on the task attribute information of the target task. The centralized scheduling management platform is the core control unit of the entire unmanned equipment scheduling system, responsible for receiving externally issued task instructions, collecting and aggregating real-time status data of each unmanned equipment execution node, executing scheduling decision algorithms, and finally issuing control instructions to the unmanned equipment.

[0011] The target task is a specific operational activity to be assigned to unmanned equipment, such as area patrol, material delivery, and environmental monitoring. Because different tasks vary significantly in terms of geographical features of the execution area, on-site weather conditions, electromagnetic interference intensity, specified completion deadlines, and the task's own level requirements, a comprehensive evaluation of task attribute information is necessary to extract two core indicators that quantify the difficulty and urgency of the task: execution complexity and task importance. These indicators provide a basis for subsequent scheduling decisions. Execution complexity represents the inherent difficulty of completing the target task under the current environmental conditions and task type; task importance represents the overall importance of the target task relative to other tasks in terms of priority and timeliness.

[0012] Furthermore, the preset baseline scheduling constraints are dynamically corrected. These baseline scheduling constraints are a set of standard scheduling thresholds pre-set under normal conditions to ensure the stable operation of the unmanned equipment cluster, representing a general limitation on the load capacity of various types of unmanned equipment in typical operational scenarios. These baseline scheduling constraints are preset during system initialization based on factors such as the design parameters of the unmanned equipment, historical operational data, and cluster size. Correcting these baseline scheduling constraints yields adaptive scheduling constraints.

[0013] Specifically, the adaptive scheduling constraint represents a new set of scheduling thresholds formed after dynamic adjustment under the influence of the specific complexity and importance of the current task to be executed. Its purpose is to make the scheduling constraints fit the execution requirements of specific tasks, so as to prioritize the resource requirements of high-complexity and high-importance tasks while ensuring the overall stability of the cluster.

[0014] Specifically, the evaluation and acquisition of the execution complexity and task importance of the target task based on its task attribute information includes: acquiring the task attribute information of the target task, wherein the task attribute information includes at least the task execution area, task characteristics, environmental conditions, task level, and remaining time window; inputting the task characteristics into a preset task type-to-basic complexity mapping table to determine the basic complexity of the task type corresponding to the current task characteristics; extracting the weather condition score, electromagnetic interference intensity score, and terrain complexity score from the environmental conditions, respectively, and weighting and summing them according to preset environmental weight coefficients to calculate the environmental complexity reflecting the severity of the current working environment; weightedly fusing the task type-to-basic complexity and environmental complexity to obtain the execution complexity; and inputting the task level into a preset level-to-importance mapping table. The importance component is calculated using a function, where a higher task level results in a larger importance component value. The ratio of the remaining time window for the task to a preset standard work duration is calculated and input into a preset time urgency nonlinear function to calculate a time urgency component. This time urgency nonlinear function ensures that the time urgency component remains low when the remaining time window is sufficient and increases when the remaining time window approaches zero. The urgency weight of the time urgency component is dynamically determined based on the remaining time window, where the urgency weight is inversely proportional to the remaining time window. The level weight is obtained by subtracting the urgency weight from 1. Based on the level weight and the urgency weight, the importance component and the time urgency component are weighted and summed to obtain the task importance.

[0015] First, the task attribute information of the target operation is obtained. This information includes at least the task execution area, task characteristics, environmental conditions, task level, and remaining time window. Specifically, the task execution area refers to the specific geographical location range that the target operation needs to cover or reach; task characteristics describe the activity type of the operation; environmental conditions encompass the weather conditions, electromagnetic interference levels, and topographic features of the work site; the task level is a pre-defined classification based on the importance of the task; and the remaining time window refers to the available time from the current moment until the task's scheduled completion time. Obtaining this multi-dimensional task attribute information provides fundamental data support for the subsequent quantitative assessment of execution complexity and task importance.

[0016] Secondly, the task characteristics are input into a pre-defined task type-to-basic complexity mapping table to determine the basic complexity of the task type corresponding to the current task characteristics. This mapping table is established in advance based on historical operational data or expert experience, recording the basic complexity values ​​corresponding to different task characteristics. Specifically, task types include inspection tasks, surveying tasks, transportation tasks, and rescue tasks, with different basic complexity values ​​corresponding to different task types.

[0017] For example, inspection tasks typically have fixed routes or paths, and their operation modes are relatively simple, so they can correspond to a low basic complexity, such as setting the basic complexity value to 0.2; surveying tasks require comprehensive data collection according to a predetermined area, and have certain requirements for route planning and data accuracy, so they can correspond to a medium-low basic complexity, such as setting the basic complexity value to 0.4; transportation tasks involve multiple stages such as material loading, route planning, and precise landing at the destination, and have high requirements for the payload and endurance of unmanned equipment, so they can correspond to a medium-high basic complexity, such as setting the basic complexity value to 0.6; rescue tasks often require rapid response in complex environments, and have high requirements for the mobility, environmental perception, and emergency response capabilities of unmanned equipment, so they can correspond to a high basic complexity, such as setting the basic complexity value to 0.8.

[0018] By inputting the currently acquired task characteristics into the task type and searching and matching them with the basic complexity mapping table, the basic complexity value reflecting the inherent difficulty of the task type can be obtained.

[0019] Then, the weather condition score, electromagnetic interference intensity score, and terrain complexity score are extracted from the environmental conditions, and weighted and summed according to the preset environmental weight coefficients to calculate the environmental complexity that reflects the severity of the current working environment.

[0020] Optionally, a weather condition score is used to quantify the favorable or unfavorable nature of current weather conditions for unmanned equipment to perform its mission. The score is set based on three real-time meteorological data points: visibility, wind speed, and precipitation intensity. First, the visibility, wind speed, and precipitation intensity values ​​for the operational area are collected in real time. Then, each data point is compared with preset scoring rules: for visibility, a score of 1.0 is given if visibility is greater than 10 kilometers; 0.8 if visibility is between 5 and 10 kilometers; 0.5 if visibility is between 1 and 5 kilometers; and 0.2 if visibility is less than 1 kilometer. Regarding wind speed, a score of 1.0 is given for wind speeds less than 5 meters per second; 0.7 for wind speeds between 5 and 10 meters per second; 0.4 for wind speeds between 10 and 15 meters per second; and 0.1 for wind speeds greater than 15 meters per second. Regarding precipitation intensity, a score of 1.0 is given for no precipitation; 0.8 for light rain; 0.5 for moderate rain; and 0.2 for heavy rain or torrential rain. Finally, the average of the visibility score, wind speed score, and precipitation intensity score is used as the final weather condition score.

[0021] Electromagnetic interference (EMI) intensity scoring quantifies the degree of interference of the electromagnetic environment within the work area on the communication and navigation of unmanned equipment. The score is set based on the real-time power spectral density of the interference signal. Optionally, the background noise level within the work area is acquired using spectrum monitoring equipment, and this level is compared with a preset interference level threshold. If the background noise level is below -110 dBm, it is determined to be no interference or slight interference, with an EMI intensity score of 1.0; if the background noise level is between -110 dBm and -90 dBm, it is determined to be moderate interference, with an EMI intensity score of 0.6; and if the background noise level is above -90 dBm, it is determined to be strong interference, with an EMI intensity score of 0.2.

[0022] The terrain complexity score quantifies the impact of factors such as surface undulation and obstacle distribution within the operational area on the passage or operation of unmanned equipment. The score is calculated based on digital elevation model (DEM) data or geographic information system (GIS) data. First, terrain elevation data for the task execution area is acquired, and the elevation standard deviation within that area is calculated. A larger elevation standard deviation indicates more pronounced terrain undulation and higher terrain complexity. Optionally, the calculated elevation standard deviation is compared with a preset complexity level threshold. If the elevation standard deviation is less than 10 meters, it is classified as flat terrain with a terrain complexity score of 1.0; if the elevation standard deviation is between 10 and 30 meters, it is classified as undulating terrain with a terrain complexity score of 0.6; and if the elevation standard deviation is greater than 30 meters, it is classified as rugged terrain with a terrain complexity score of 0.3.

[0023] The preset environmental weighting coefficients correspond to the relative importance of weather conditions, electromagnetic interference intensity, and terrain complexity in the overall environmental complexity. These weighting coefficients are set based on statistical analysis of the impact of each environmental factor on unmanned equipment operations. For example, the weighting coefficient for weather conditions is set to 0.3, for electromagnetic interference intensity to 0.4, and for terrain complexity to 0.3. Multiplying the weather condition score, electromagnetic interference intensity score, and terrain complexity score by their respective weighting coefficients and then summing them yields an environmental complexity value ranging from 0.1 to 1.0. This environmental complexity comprehensively reflects the severity of the operating environment; a lower value indicates a more severe environment.

[0024] Furthermore, the execution complexity is obtained by weighted fusion of the acquired task type basic complexity and environmental complexity. The task type basic complexity reflects the inherent difficulty of the task itself, while the environmental complexity reflects the degree of influence of the external environment on task execution. By weighted summing of the two indicators, the combined effect of the task's inherent attributes and external environmental factors can be comprehensively considered, resulting in a quantitative indicator that comprehensively reflects the difficulty of executing the target task, namely, execution complexity.

[0025] The weighting coefficients for the basic complexity of the task type and the environmental complexity can be preset or dynamically adjusted according to the actual application scenario. For example, in routine operation scenarios, the weighting coefficient for the basic complexity of the task type can be set to 0.6, and the weighting coefficient for the environmental complexity can be set to 0.4 to reflect the dominant role of the task's own attributes in the execution difficulty. When performing tasks in areas with frequent severe weather or strong electromagnetic interference, the weighting coefficients can be dynamically adjusted, reducing the weighting coefficient for the basic complexity of the task type to 0.3 and increasing the weighting coefficient for the environmental complexity to 0.7 to highlight the significant impact of external environmental factors on the execution difficulty. For time-sensitive scenarios such as emergency rescue, the weighting coefficient for the basic complexity of the task type and the weighting coefficient for the environmental complexity can be set to 0.5 to balance the inherent difficulty of the task with environmental constraints.

[0026] Then, the task level is input into a preset level-to-importance mapping function to calculate the level importance component, where the higher the task level, the larger the value of the level importance component. The level-to-importance mapping function is a pre-established mathematical relationship used to convert discrete task levels into continuous importance values. Optionally, task levels can be divided into three levels according to their importance, from high to low: Level 1, Level 2, and Level 3. Level 1 tasks correspond to the highest level importance component, with a value set to 1.0; Level 2 tasks correspond to a medium importance component, with a value set to 0.6; and Level 3 tasks correspond to the lowest importance component, with a value set to 0.3. This level-to-importance mapping function quantifies the basic importance represented by different task levels.

[0027] Secondly, the ratio of the remaining time window of the task to the preset standard operation time is calculated, and this ratio is input into the preset time urgency nonlinear function to calculate the time urgency component. The time urgency nonlinear function makes the time urgency component keep a low value when the remaining time window of the task is sufficient, and increase sharply when the remaining time window of the task approaches zero.

[0028] Specifically, the preset standard operation time refers to the standard time required to complete this type of task under ideal environmental conditions, representing the normal operation time of the task without external interference. This standard operation time is preset according to different task types; for example, the standard operation time for inspection tasks is set to 30 minutes, for surveying tasks to 60 minutes, for transportation tasks to 45 minutes, and for rescue tasks to 20 minutes. The ratio of the remaining task time window to the standard operation time reflects the sufficiency of available time relative to the time required for normal operation. A ratio greater than 1 indicates ample time, a ratio less than 1 indicates time pressure, and a ratio approaching 0 indicates extreme time pressure. Using this ratio as input, a time pressure component reflecting the degree of time urgency can be obtained through mapping using a time pressure nonlinear function.

[0029] Optionally, the nonlinear function for time urgency is constructed using a negative exponential function. Specifically, the time urgency component is equal to the negative power of the natural constant e, then taking the reciprocal. That is, the time urgency component equals 1 divided by the negative power of e. This function is designed to output lower values ​​when the ratio is large, indicating ample time. For example, when the ratio is 2, the time urgency component is approximately 0.14; when the ratio is 1, indicating just enough time, the time urgency component is approximately 0.37; when the ratio is 0.5, the time urgency component is approximately 0.61; and when the ratio approaches 0, indicating extremely tight time, the time urgency component approaches 1.0. This accurately reflects the nonlinear, rapid increase in time urgency as the remaining time decreases.

[0030] Furthermore, the urgency weight of the time urgency component is dynamically determined based on the remaining time window of the task, where the urgency weight is inversely proportional to the remaining time window. Specifically, the urgency weight is used to adjust the contribution of the time urgency component to the final task importance in subsequent calculations. This urgency weight is dynamically determined based on the remaining time window of the task; the shorter the remaining time, the larger the urgency weight value, and the longer the remaining time, the smaller the urgency weight value.

[0031] Optionally, the urgency weight is calculated using an inverse linear function, specifically expressed as urgency weight equal to 1 minus the ratio of the remaining task time window to the preset maximum remaining time window. The preset maximum remaining time window can be set according to task management rules, for example, to 120 minutes. When the remaining task time window is 120 minutes, the urgency weight is 0; when the remaining task time window is 60 minutes, the urgency weight is 0.5; when the remaining task time window is 30 minutes, the urgency weight is 0.75; and when the remaining task time window approaches 0 minutes, the urgency weight approaches 1.0.

[0032] This calculation method allows the urgency weight to take a smaller value when there is ample remaining time and a larger value when there is limited remaining time, thus accurately reflecting the time-varying characteristics that the degree of time urgency should have in the assessment of task importance.

[0033] Secondly, the urgency weight is obtained by subtracting the urgency weight from 1. The rank weight is a coefficient used to adjust the contribution of the rank importance component to the final task importance. Through the complementary relationship between the urgency weight and the rank weight, it can be ensured that the relative importance of the rank importance component and the time urgency component achieves a dynamic balance when the time urgency changes.

[0034] Finally, based on the rank weight and urgency weight, the rank importance component and the time urgency component are weighted and summed to obtain the task importance. Specifically, the rank importance component is multiplied by its rank weight, and the time urgency component is multiplied by its urgency weight. The products are then added together to obtain a quantitative index that comprehensively reflects both the basic importance of the task rank and the urgency of the task time—that is, the task importance. This task importance provides an important basis for the dynamic adjustment of subsequent scheduling constraints.

[0035] Further, the preset baseline scheduling constraints are dynamically corrected to generate adaptive scheduling constraints, including: obtaining the baseline scheduling constraints, wherein the baseline scheduling constraints include multiple node load margin lower limits corresponding to multiple equipment types and multiple overall average load upper limits for the same type of equipment; using the ratio of the execution complexity to the preset baseline execution complexity as a first correction coefficient; using the ratio of the task importance to the preset baseline task importance as a second correction coefficient; performing a weighted summation of the first and second correction coefficients, using the weighted summation result as a node load margin compensation coefficient, and using the reciprocal of the weighted summation result as an overall average load upper limit compensation coefficient; correcting the multiple node load margin lower limits according to the node load margin compensation coefficient, and correcting the multiple overall average load upper limits for the same type of equipment according to the overall average load upper limit compensation coefficient, to obtain multiple adaptive node load margin lower limits and multiple adaptive overall average load upper limits for the same type of equipment, which serve as adaptive scheduling constraints.

[0036] First, the baseline scheduling constraints are obtained. These constraints include the lower limits of node load margins for multiple equipment types and the upper limits of the overall average load for multiple equipment of the same type. Specifically, the lower limit of node load margin refers to the minimum remaining load capacity that a single unmanned equipment execution node must retain before undertaking a new task to ensure the reliability of task execution. This lower limit of node load margin is used to prevent task failure due to node overuse. The upper limit of the overall average load for equipment of the same type refers to the upper limit that the overall average load of all unmanned equipment execution nodes of the same type must not exceed after undertaking a new task. This upper limit of the overall average load for equipment of the same type is used to prevent the overall load imbalance of the cluster due to a certain type of equipment undertaking too many tasks.

[0037] Secondly, the ratio of the execution complexity to the preset baseline execution complexity is used as the first correction coefficient. The preset baseline execution complexity refers to a reference complexity value pre-set under standard operating conditions, for example, 0.5. When the execution complexity of the target task is higher than the baseline execution complexity, the first correction coefficient is greater than 1; when the execution complexity is lower than the baseline execution complexity, the first correction coefficient is less than 1. This first correction coefficient is used to quantify the degree of deviation of the actual task complexity from the standard situation.

[0038] Simultaneously, the ratio of the task importance to the preset benchmark task importance is used as a second correction coefficient. The preset benchmark task importance refers to a reference importance value pre-set under normal circumstances, for example, 0.5. When the target task's importance is higher than the benchmark task importance, the second correction coefficient is greater than 1; when the task's importance is lower than the benchmark task importance, the second correction coefficient is less than 1. This second correction coefficient is used to quantify the degree of deviation of the actual task importance from the standard situation.

[0039] Furthermore, the first and second correction coefficients are weighted and summed. The weighted sum is used as the node load margin compensation coefficient, and the reciprocal of the weighted sum is used as the overall average load upper limit compensation coefficient. The weighting coefficients used for the weighted summation of the first and second correction coefficients are set according to the relative importance of execution complexity and task importance in scheduling constraint correction. For example, in normal operation scenarios, the weighting coefficient corresponding to execution complexity can be set to 0.4, and the weighting coefficient corresponding to task importance can be set to 0.6, to reflect the dominant role of task importance in adjusting scheduling constraints. In scenarios with harsh environmental conditions or significant task execution difficulty, the weighting coefficient corresponding to execution complexity can be increased to 0.7, and the weighting coefficient corresponding to task importance can be decreased to 0.3, to highlight the decisive impact of execution complexity on node load capacity requirements. In time-sensitive scenarios such as emergency rescue, the weighting coefficient corresponding to execution complexity can be set to 0.5, and the weighting coefficient corresponding to task importance can be set to 0.5, to balance the combined effects of inherent task difficulty and task importance on scheduling constraint correction.

[0040] Specifically, the node load margin compensation coefficient is used to subsequently correct the lower limit of node load margin. Its value comprehensively reflects the combined impact of task complexity and importance on node load requirements. The greater the task complexity and importance, the larger the node load margin compensation coefficient, thus correcting the lower limit of node load margin upwards. This ensures that nodes executing high-complexity, high-importance tasks have sufficient remaining load capacity, guaranteeing task execution quality. The overall average load upper limit compensation coefficient is the reciprocal of the node load margin compensation coefficient. Its value comprehensively reflects the combined impact of task complexity and importance on the overall cluster load limit. The greater the task complexity and importance, the smaller the overall average load upper limit compensation coefficient, thus correcting the overall average load upper limit of similar equipment downwards. This prevents the cluster from becoming overloaded due to undertaking high-requirement tasks, ensuring the overall stability of cluster operation.

[0041] Finally, the lower limit of the load margin of multiple nodes is corrected according to the node load margin compensation coefficient, and the upper limit of the overall average load of multiple similar equipment is corrected according to the overall average load upper limit compensation coefficient, so as to obtain the lower limit of the load margin of multiple adapted nodes and the upper limit of the overall average load of multiple adapted similar equipment, which serve as the adaptation scheduling constraints.

[0042] The specific correction method is as follows: The original lower limit of node load margin for each type of equipment is multiplied by a node load margin compensation coefficient to obtain the adapted lower limit of node load margin for each type of equipment; the original upper limit of the overall average load of the same type of equipment for each type of equipment is multiplied by an overall average load upper limit compensation coefficient to obtain the adapted overall average load upper limit of the same type of equipment for each type of equipment. Through the above correction, the scheduling constraints can be dynamically adapted to the specific requirements of the current task. When the task is highly complex and important, by increasing the single-node load threshold and tightening the overall cluster load limit, sufficient resources are reserved for high-requirement tasks, thereby ensuring the quality of task execution and the stability of cluster operation.

[0043] S20: Based on the comprehensive load status data and location coordinates of several unmanned equipment execution nodes, using the aforementioned adaptive scheduling constraints as scheduling quality constraints, the remaining task time window as scheduling response constraints, and the equipment type and quantity required by the task as equipment demand constraints, a random scheduling simulation is performed on the several unmanned equipment execution nodes to generate multiple qualified equipment scheduling schemes. The comprehensive load status data of the several unmanned equipment execution nodes is a set of quantitative indicators reflecting the current remaining operational capacity of each unmanned equipment, including parameters such as remaining battery power, remaining available task payload, communication link quality margin, and availability of the autonomous navigation system. Location coordinates are the spatial location of each unmanned equipment execution node at the current moment, typically expressed in latitude and longitude coordinates or grid coordinates.

[0044] Since different unmanned equipment execution nodes have different load states and spatial distributions, and the target operation task has clear requirements on equipment type, quantity, and arrival time limit, the aforementioned adaptive scheduling constraints are used as scheduling quality constraints to ensure that the selected nodes have sufficient remaining load capacity to guarantee the quality of task execution; the remaining time window of the task is used as scheduling response constraints to ensure that the selected nodes can arrive at the task execution area within the specified time limit; and the equipment type and quantity required by the task are used as equipment demand constraints to ensure that the type and quantity of the selected nodes meet the basic operational requirements of the task.

[0045] By conducting random scheduling simulations on multiple unmanned equipment execution nodes under the above triple constraints, multiple qualified equipment scheduling schemes that meet the basic constraints are generated, providing a rich set of candidate schemes for subsequent multi-objective optimization.

[0046] Specifically, based on the comprehensive load status data and position coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, the remaining time window of the task as scheduling response constraints, and the type and quantity of equipment required by the task as equipment demand constraints, a random scheduling simulation is performed on the several unmanned equipment execution nodes to generate multiple qualified equipment scheduling schemes. This includes: selecting unmanned equipment execution nodes of various types that match the type of equipment required by the task from the several unmanned equipment execution nodes, as the initial node set corresponding to each type; selecting nodes whose current node load margin is greater than or equal to the corresponding type's adaptive node load margin lower limit from the initial node set corresponding to each type, based on the multiple adaptive node load margin lower limits in the adaptive scheduling constraints, forming a first-level candidate node set corresponding to each type; calculating the time required to reach the task execution area based on the distance between the position coordinates of each candidate node and the task execution area, combined with the travel speed of each candidate node, and selecting the first-level candidate node set corresponding to each type. From the primary candidate node set, candidate nodes capable of reaching the task execution area within the remaining time window of the task are selected to form secondary candidate node sets corresponding to each type. Based on the equipment type and quantity required by the task, a specified number of candidate nodes are randomly selected from the secondary candidate node sets corresponding to each type, and these selected candidate nodes are combined to generate a candidate equipment scheduling scheme. This random selection and combination process is repeated to generate multiple candidate equipment scheduling schemes. For each candidate equipment scheduling scheme, the overall average load of each type of unmanned equipment execution node in the current cluster is obtained. Based on the upper limit of the overall average load of multiple adaptable equipment of the same type in the adaptation scheduling constraints, it is determined whether the overall average load of each type of unmanned equipment execution node after the candidate equipment scheduling scheme is adopted is less than or equal to the upper limit of the overall average load of the corresponding type of adaptable equipment of the same type. Candidate equipment scheduling schemes that meet the above overall average load upper limit requirements are considered qualified equipment scheduling schemes, until a preset number of qualified equipment scheduling schemes are obtained.

[0047] First, based on the equipment types required for the mission, unmanned equipment execution nodes matching the required equipment types are selected from a pool of unmanned equipment execution nodes, forming the initial node set for each type. The equipment type required for the mission refers to one or more categories of unmanned equipment necessary to complete the current objective task. For example, if the mission requires drones for aerial reconnaissance and unmanned vehicles for ground support, the required equipment types include both drones and unmanned vehicles. By classifying and filtering all unmanned equipment execution nodes according to equipment type, initial node sets corresponding to each required equipment type can be obtained.

[0048] Secondly, based on the lower limits of the load margin of multiple adaptable nodes in the adaptation scheduling constraints, nodes whose current load margin is greater than or equal to the lower limit of the load margin of the corresponding type of adaptable node are selected from the initial node set corresponding to each type, forming the first-level candidate node set for each type. Here, node load margin is a quantitative indicator reflecting the current remaining operational capacity of a single unmanned equipment execution node. The lower limit of the adaptable node load margin is a single-node load threshold value obtained after correction for the current task. This step, by removing nodes from the initial node set that do not meet the lower limit of the adaptable node load margin, ensures that all nodes entering subsequent screening have sufficient remaining load capacity to guarantee task execution quality.

[0049] Then, based on the distance between the position coordinates of each candidate node and the mission execution area, and combined with the sailing speed of each candidate node, the time required to reach the mission execution area is calculated. Candidate nodes that can reach the mission execution area within the remaining time window of the mission are selected from the first-level candidate node set corresponding to each type, forming the second-level candidate node set corresponding to each type.

[0050] The arrival time is calculated by dividing the straight-line distance or planned path distance between the node and the mission execution area by the node's average sailing speed. This time is compared with the remaining time window of the mission, and only nodes whose arrival time is less than or equal to the remaining time window are retained, thus ensuring that the selected nodes can arrive at the mission site within the specified time limit and meet the mission's timeliness requirements.

[0051] Furthermore, according to the type and quantity of equipment required for the mission, a specified number of candidate nodes are randomly selected from the second-level candidate node set corresponding to each type, and the selected candidate nodes are combined to generate a candidate equipment scheduling scheme.

[0052] The required equipment quantity for a task refers to the number of unmanned equipment units needed for each equipment type to complete the current task. For example, if a task requires two drones and one unmanned vehicle, two nodes are randomly selected from the second-level candidate node set corresponding to the drone type, and one node is randomly selected from the second-level candidate node set corresponding to the unmanned vehicle type. Combining these three nodes constitutes a candidate equipment scheduling scheme. This random selection method ensures the diversity and exploratory nature of the generated schemes.

[0053] Finally, the above random selection and combination process is repeated to generate multiple candidate equipment scheduling schemes. Through multiple independent and repeated random selection and combination operations, multiple candidate equipment scheduling schemes covering different node combination methods can be obtained, providing a sufficient sample space for subsequent qualification screening and optimization evaluation.

[0054] Furthermore, for each candidate equipment scheduling scheme, the overall average load of each type of unmanned equipment execution node in the current cluster is obtained. Based on the upper limit of the overall average load of multiple adapted equipment of the same type in the adaptation scheduling constraints, it is determined whether the overall average load of each type of unmanned equipment execution node is less than or equal to the upper limit of the overall average load of the corresponding type of adapted equipment of the same type after the candidate equipment scheduling scheme is adopted. Here, the overall average load refers to the average load level of all unmanned equipment execution nodes of a certain type at the current moment, which can be calculated based on the load margin of each node. Considering the adoption of a candidate scheme, the overall average load of each type of equipment needs to be recalculated, that is, after taking into account the newly added task load in the scheme, it is determined whether the new overall average load exceeds the threshold specified by the upper limit of the overall average load of adapted equipment of the same type. This step aims to prevent a certain type of equipment from taking on too many tasks, thus ensuring the overall load balance and operational stability of the cluster.

[0055] Finally, candidate equipment scheduling schemes that meet the aforementioned overall average load ceiling requirements are selected as qualified equipment scheduling schemes, until a preset number of qualified equipment scheduling schemes are obtained. The preset number is set based on the scheme diversity requirements for scheduling decisions and the convergence speed requirements of subsequent optimization algorithms; for example, it may be set to 50. The final qualified equipment scheduling scheme is a set of feasible schemes that simultaneously meet the triple constraints of individual node load capacity thresholds, task timeliness requirements, and overall cluster load balancing constraints. Each scheme represents a candidate equipment combination capable of completing the target task. This set of qualified equipment scheduling schemes serves as the input sample space for subsequent multi-objective optimization evaluation. By selecting the scheme with the highest scheduling quality index from this set, a decision-making basis is provided for the final output of the optimal equipment scheduling scheme.

[0056] S30: Establish a multi-objective optimization function to evaluate the scheduling quality of the multiple qualified equipment scheduling schemes and determine multiple scheme scheduling quality indices; further, establish a multi-objective optimization function to evaluate the scheduling quality of the aforementioned multiple qualified equipment scheduling schemes and determine multiple scheme scheduling quality indices. The multi-objective optimization function is a mathematical expression used to comprehensively measure the performance of a qualified equipment scheduling scheme on multiple conflicting optimization objectives. The multiple qualified equipment scheduling schemes are evaluated separately using this multi-objective optimization function, and the scheme scheduling quality index corresponding to each scheme is calculated. This scheme scheduling quality index is a dimensionless quantitative indicator, representing the overall excellence of the corresponding qualified equipment scheduling scheme when comprehensively considering multiple objectives such as resource consumption and load balancing. A higher value indicates a better overall performance of the scheme in achieving resource conservation and cluster load balancing.

[0057] Specifically, a multi-objective optimization function is established to evaluate the scheduling quality of the multiple qualified equipment scheduling schemes and determine multiple scheduling quality indices. This includes: establishing a multi-objective optimization function that includes the objectives of minimizing resource consumption and maximizing load balancing; for each qualified equipment scheduling scheme, calculating the resource consumption target value based on the total travel distance of all unmanned equipment execution nodes from their current positions to the task execution area, and the estimated total energy consumption calculated based on the unit distance energy consumption of each type of unmanned equipment; for each qualified equipment scheduling scheme, calculating the load balancing target value based on the variance of node load margin between execution nodes of the same type of unmanned equipment and the variance of node load margin between execution nodes of different types of unmanned equipment; multiplying the resource consumption target value and the load balancing target value of each qualified equipment scheduling scheme by the first weighting coefficient and the second weighting coefficient, respectively, and summing the results, using the reciprocal of the calculation results as the scheduling quality index of each qualified equipment scheduling scheme.

[0058] Specifically, a time urgency factor is calculated based on the remaining time window of the task, and a cluster resource stress factor is calculated based on the average load margin of all unmanned equipment execution nodes in the current cluster. The time urgency factor and the cluster resource stress factor are input into a preset weight allocation model, and the first weight coefficient corresponding to the goal of minimizing resource consumption and the second weight coefficient corresponding to the goal of maximizing load balancing are output.

[0059] First, a multi-objective optimization function is established, comprising the objectives of minimizing resource consumption and maximizing load balancing. The objective of minimizing resource consumption aims to reduce the energy consumption required to execute the task, while the objective of maximizing load balancing aims to make the load distribution among the execution nodes of the unmanned equipment more even, avoiding overload of some nodes while other nodes are idle.

[0060] Secondly, for each qualified equipment scheduling scheme, the resource consumption target value of the scheme is calculated comprehensively based on the total travel distance of all unmanned equipment execution nodes from their current positions to the task execution area, and the estimated total energy consumption calculated based on the energy consumption per unit distance of each type of unmanned equipment. The total travel distance is obtained by adding the distances between the position coordinates of each node in the scheme and the task execution area. The estimated total energy consumption is obtained by multiplying the energy consumption per unit distance of each type of unmanned equipment by the travel distance of the corresponding node and then summing the results. The resource consumption target value can be directly represented by the estimated total energy consumption. The smaller this target value, the less energy the scheme consumes during task execution, and the higher the degree of achievement of the resource consumption minimization goal.

[0061] Simultaneously, for each qualified equipment scheduling scheme, the load balancing target value of the scheme is calculated comprehensively based on the variance of node load margin between execution nodes of the same type of unmanned equipment and the variance of node load margin between execution nodes of different types of unmanned equipment. Among them, the node load margin variance is used to measure the dispersion of load margin among a group of nodes. The larger the variance, the greater the load difference between nodes, that is, the more unbalanced the load; the smaller the variance, the closer the load between nodes, that is, the more balanced the load.

[0062] Specifically, the variance of the load margin of each node within the same type in the scheme is first calculated to obtain the degree of load balancing within the same type. Then, the variance of the average load margin between different types is calculated to obtain the degree of load balancing between different types. The load balancing target value, reflecting the overall load balancing degree of the scheme, is obtained by weighted summation of the variances within the same type and between different types. In the weighted summation process, the weighting coefficients are set according to the relative importance of load balancing within the same type and load balancing between different types in the overall load balancing assessment. For example, the weighting coefficient for variance within the same type is set to 0.6, and the weighting coefficient for variance between different types is set to 0.4 to reflect the dominant role of load balancing within the same type of equipment in the overall stability of the cluster. In scenarios involving the collaborative operation of multiple types of equipment, the weighting coefficient for variance within the same type can be adjusted to 0.5, and the weighting coefficient for variance between different types can be adjusted to 0.5 to balance the combined impact of internal balance and inter-type balance on the overall load balancing. Since a larger variance indicates a more unbalanced load, and the goal of maximizing load balancing requires the load to be as balanced as possible, the target value of load balancing should be as small as possible.

[0063] Furthermore, the resource consumption target value and load balancing target value of each qualified equipment scheduling scheme are multiplied by the first weighting coefficient and the second weighting coefficient, respectively, and then summed. The reciprocal of the result is used as the scheme scheduling quality index for each qualified equipment scheduling scheme. The smaller the resource consumption target value and the smaller the load balancing target value, the smaller the weighted sum, and the larger its reciprocal, i.e., the scheme scheduling quality index, is, indicating that the scheme performs better in terms of both resource consumption and load balancing objectives. By taking the reciprocal, the original minimization problem can be transformed into a scheme scheduling quality index maximization problem, which facilitates subsequent optimization operations.

[0064] Specifically, the time urgency factor is calculated based on the remaining time window of the task, and the cluster resource stress factor is calculated based on the average load margin of all unmanned equipment execution nodes in the current cluster. The time urgency factor and the cluster resource stress factor are input into the preset weight allocation model, and the first weight coefficient corresponding to the goal of minimizing resource consumption and the second weight coefficient corresponding to the goal of maximizing load balancing are output.

[0065] Specifically, the time urgency factor quantifies the time urgency of the current task and can be represented as the reciprocal of the remaining time window of the task; the more urgent the time, the larger the value of this factor. The cluster resource stress factor quantifies the stress of the overall resources of the current cluster and can be represented as the reciprocal of the average load margin of all nodes; the smaller the average load margin, the higher the overall load of the cluster, and the larger the value of this factor.

[0066] The time urgency factor and cluster resource stress factor are input into a preset weighting model, which dynamically adjusts the relative importance of the two optimization objectives based on the current task urgency and cluster resource stress. Specifically, the time urgency factor is represented by the normalized reciprocal of the remaining task time window. For example, when the remaining task time window is 30 minutes and the standard time threshold is set to 60 minutes, the time urgency factor is calculated as 0.67; when the remaining task time window is 120 minutes, the time urgency factor is calculated as 0.33. The cluster resource stress factor is represented by the normalized reciprocal of the current average cluster load margin. When the current average cluster load margin is 0.3 and the baseline load margin is set to 0.5, the cluster resource stress factor is calculated as 0.625; when the current average cluster load margin is 0.8, the cluster resource stress factor is calculated as 0.385.

[0067] The time urgency factor and cluster resource stress factor are input into a preset weighting model, which calculates the first and second weighting coefficients using a linear weighting method. The first weighting coefficient is calculated as the time urgency factor divided by the sum of the time urgency factor and the cluster resource stress factor; the second weighting coefficient is calculated as the cluster resource stress factor divided by the sum of the time urgency factor and the cluster resource stress factor. For example, when the time urgency factor is 0.67 and the cluster resource stress factor is 0.625, the first weighting coefficient is calculated as 0.517 and the second weighting coefficient as 0.483; when the time urgency factor is 0.33 and the cluster resource stress factor is 0.385, the first weighting coefficient is calculated as 0.462 and the second weighting coefficient as 0.538; when the time urgency factor is 0.8 and the cluster resource stress factor is 0.3, the first weighting coefficient is calculated as 0.727 and the second weighting coefficient as 0.273.

[0068] Through the above quantitative calculation method, the first and second weight coefficients can be dynamically adjusted according to the urgency of the task and the scarcity of cluster resources, thereby realizing the adaptive capability of the multi-objective optimization function to changes in the scenario.

[0069] S40: With maximizing the scheme scheduling quality index as the optimization objective, based on the multiple scheme scheduling quality indices, use an optimization algorithm to optimize the multiple qualified equipment scheduling schemes, output the optimal equipment scheduling scheme, and perform scheduling control on the multiple unmanned equipment execution nodes.

[0070] Ultimately, with the goal of maximizing the scheduling quality index of the scheme, an optimization algorithm is used to optimize the scheduling of multiple qualified equipment schemes based on the scheduling quality indices of multiple schemes.

[0071] Specifically, with maximizing the scheduling quality index as the optimization objective, the optimization algorithm is used to optimize the scheduling of the multiple qualified equipment scheduling schemes based on the multiple scheduling quality indices. This includes: encoding each qualified equipment scheduling scheme as a particle position in a particle swarm optimization algorithm, where the particle position is represented by a multi-dimensional vector, and the value of each dimension corresponds to the index number of the unmanned equipment execution node assigned to the equipment type required for a task in the candidate node set; initializing a particle swarm containing a preset number of particles, using the multiple qualified equipment scheduling schemes as the initial particle positions, and randomly initializing a velocity vector for each particle; and using the scheduling quality index of each particle as the particle position. The fitness value is used to determine the individual optimal position of each particle and the global optimal position of the entire particle swarm. Based on the velocity and position update formulas of the particle swarm optimization algorithm, combined with inertia weights, individual learning factors, and social learning factors, the velocity and position of each particle are iteratively updated. In each iteration, the scheduling quality index of the corresponding scheduling scheme is recalculated as the fitness value based on the updated particle position, and the individual optimal position and the global optimal position are updated. This iterative update is repeated until the preset maximum number of iterations is reached or the fitness value converges to a stable state. The particle position corresponding to the global optimal position at the end of the iteration is decoded to obtain the optimal equipment scheduling scheme.

[0072] The particle swarm optimization algorithm incorporates an inertia weight in its speed update formula. This inertia weight decreases linearly from the initial inertia weight to the final inertia weight as the number of iterations increases, thus emphasizing global search in the early stages of iteration and local fine-grained search in the later stages. Furthermore, the particle swarm optimization algorithm dynamically adjusts the individual learning factor and the social learning factor based on the remaining time window of the task. When the remaining time window is less than a preset time threshold, the social learning factor is increased and the individual learning factor is decreased to accelerate particle convergence to the global optimum, meeting the rapid optimization requirements in time-sensitive scenarios.

[0073] First, each qualified equipment scheduling scheme is encoded as a particle position in the particle swarm optimization algorithm. The particle position is represented by a multi-dimensional vector, where each dimension corresponds to the index number of the unmanned equipment execution node assigned to the equipment type required for a task within the candidate node set. Through this encoding method, each qualified equipment scheduling scheme is mapped to a particle position in the search space, thus transforming the scheduling scheme optimization problem into a particle swarm optimization problem within the search space.

[0074] Secondly, a particle swarm containing a preset number of particles is initialized. Multiple qualified equipment scheduling schemes are used as the initial particle positions, and a velocity vector is randomly initialized for each particle. The initial particle swarm consists of qualified equipment scheduling schemes generated in the preceding steps, ensuring that all search starting points are feasible solutions. The initial velocity vector of each particle is randomly generated within a preset velocity range, used to control the particle's movement step size and direction in the search space. The velocity range refers to the maximum allowable change in a particle's position across all dimensions. This velocity range is set based on the value range of each dimension of the particle's position; for example, the lower velocity limit is set to the absolute value of the difference between the maximum and minimum negative index numbers, and the upper velocity limit is set to the absolute value of the difference between the maximum and minimum index numbers.

[0075] For example, if the candidate node set corresponding to a certain equipment type contains 10 nodes, the index number ranges from 1 to 10, and the absolute value of the maximum difference between index numbers is 9. The velocity range of this dimension can be set from -9 to +9 to ensure that the particle does not cross too many candidate node indices in a single iteration, thereby maintaining the stability of the search process.

[0076] Furthermore, the scheduling quality index of each particle is used as its fitness value. Based on the fitness value, the individual optimal position of each particle and the global optimal position of the entire particle swarm are determined. Specifically, the individual optimal position is the position with the highest fitness value that each particle has experienced during the historical search process, and the global optimal position is the position with the highest fitness value among the individual optimal positions of all particles in the entire particle swarm. Guided by the individual optimal and global optimal positions, the particle swarm can gradually gather towards regions with higher fitness values.

[0077] Furthermore, based on the velocity and position update formulas of the particle swarm optimization algorithm, and combined with inertia weights, individual learning factors, and social learning factors, the velocity and position of each particle are iteratively updated. Specifically, the velocity update formula is: the particle's velocity at the (k+1)th iteration equals the inertia weight multiplied by the particle's velocity at the kth iteration, plus the individual learning factor multiplied by a random number multiplied by the difference between the individual's optimal position and the particle's current position, plus the social learning factor multiplied by a random number multiplied by the difference between the global optimal position and the particle's current position. This velocity update formula comprehensively considers the inertial continuation of the particle's current velocity, the particle's motion component towards its own historical optimal position, and the particle's motion component towards the swarm's global optimal position. The inertia component maintains the particle's original motion trend, the individual cognition component guides the particle towards its previously discovered optimal position, and the social cognition component guides the particle to converge towards the swarm's currently discovered optimal position.

[0078] The position update formula is that the particle's position at the (k+1)th iteration equals the particle's position at the (k)th iteration plus its velocity at the (k+1)th iteration. This position update formula adds the updated velocity to the current position, thereby enabling the particle to move in the search space and gradually move towards regions with higher fitness values.

[0079] Specifically, the inertia weight is a coefficient used to control the degree to which the particle's previous velocity affects its current velocity. A larger inertia weight is beneficial for the particle to explore a wide area, while a smaller inertia weight is beneficial for the particle to perform a fine search near its current optimal region. The individual learning factor is a coefficient used to adjust the degree to which the particle learns towards its own optimal position; the larger this coefficient, the stronger the particle's tendency to move towards its individual optimal position. The social learning factor is a coefficient used to adjust the degree to which the particle learns towards the global optimal position of the group; the larger this coefficient, the stronger the particle's tendency to move towards the global optimal position. The specific values ​​of the inertia weight, individual learning factor, and social learning factor need to be reasonably configured according to the characteristics of the scheduling decision scenario to balance the algorithm's global search capability and local exploration capability.

[0080] In the particle swarm optimization algorithm, an inertia weight is introduced into the velocity update formula. This inertia weight decreases linearly with the number of iterations, from the initial inertia weight to the final inertia weight. This allows for a focus on global search in the early stages of iteration and on fine-grained local search in the later stages. For example, the initial inertia weight can be set to 0.9, and the final inertia weight to 0.4. As the number of iterations increases, the inertia weight decreases linearly, enabling particles to explore a wide area with larger step sizes in the early stages of the search, avoiding premature entrapment in local optima. In the later stages, smaller step sizes are used for fine-grained search near the current optimal region, improving convergence accuracy.

[0081] In the Particle Swarm Optimization (PSO) algorithm, individual learning factors and social learning factors are dynamically adjusted based on the remaining time window of the task. When the remaining time window is less than a preset time threshold, the social learning factor is increased while the individual learning factor is decreased to accelerate particle convergence towards the global optimum, meeting the need for rapid optimization in time-sensitive scenarios. For example, if the time threshold is set to 30 minutes, when the remaining time window is greater than 30 minutes, both the individual and social learning factors can be set to 2.0 to maintain a balance between the particle's own experience and the group's experience. When the remaining time window is less than or equal to 30 minutes, the social learning factor is increased to 2.5 while the individual learning factor is decreased to 1.5, making the particles more inclined to move towards the global optimum, reducing their reliance on their own historical experience, thereby accelerating the convergence speed and ensuring a better scheduling scheme is obtained within a limited time.

[0082] By rationally configuring inertia weights, individual learning factors, and social learning factors, a balance can be achieved between global exploration capabilities and local development capabilities, thereby improving the search efficiency and convergence accuracy of the particle swarm optimization algorithm.

[0083] In each iteration, the scheduling quality index of the corresponding scheduling scheme is recalculated as the fitness value based on the updated particle positions, and the individual optimal position and the global optimal position are updated. As the iteration progresses, the particle swarm continuously moves towards regions with higher fitness values, gradually approaching the optimal solution.

[0084] Finally, the iterative updates are repeated until the preset maximum number of iterations is reached or the fitness value converges to a stable state. The particle positions corresponding to the globally optimal positions at the end of the iteration are decoded to obtain the optimal equipment scheduling scheme. The maximum number of iterations is the maximum number of update rounds allowed by the particle swarm optimization algorithm, set according to the available computation time for scheduling decisions and the required accuracy of the problem solution; for example, the maximum number of iterations may be set to 200. When the actual number of iterations reaches this preset maximum number, the algorithm is forcibly terminated and outputs the current optimal solution, ensuring that the scheduling decision can be completed within a finite time. Fitness value convergence to a stable state means that in multiple consecutive iterations, the change in the fitness value corresponding to the globally optimal position is less than a preset threshold; for example, in 20 consecutive iterations, the change in the globally optimal fitness value is less than 0.001. When this convergence state occurs, it indicates that the algorithm can no longer improve the quality of the current solution, and further iteration is of limited significance; therefore, the iteration can be terminated early to save computational resources.

[0085] Specifically, when the iterative update reaches the preset maximum number of iterations or the fitness value converges to a stable state, the particle position corresponding to the global optimal position at the end of the iteration is decoded. This decoding process is the inverse of the encoding process, restoring the multi-dimensional vector form of the particle position to the specific unmanned equipment execution node allocation combination, which is the final output optimal scheduling scheme.

[0086] Furthermore, the particle swarm optimization algorithm introduces a constraint handling mechanism during its iterative update process, including: after obtaining a new particle position in each iteration, analyzing the allocation of execution nodes of each type of unmanned equipment corresponding to the new particle position; verifying whether the current node load margin of each unmanned equipment execution node allocated in the new particle position is greater than or equal to the corresponding type of adaptation node load margin lower limit based on multiple adaptation node load margin lower limits in the adaptation scheduling constraints; and verifying whether the overall average load of each type of unmanned equipment execution node in the new particle position is less than or equal to the corresponding type of adaptation equipment overall average load upper limit based on multiple adaptation equipment of the same type overall average load upper limits in the adaptation scheduling constraints. Based on the location coordinates of each unmanned equipment execution node and its distance from the mission execution area, and combined with the travel speed of each node, the time required to reach the mission execution area is calculated. It is then verified whether each assigned unmanned equipment execution node in the new particle position can reach the mission execution area within the remaining mission time window. If the new particle position simultaneously satisfies the node load margin verification, the overall average load verification of similar equipment, and the mission arrival time verification, the particle position is retained and its fitness value is calculated in the conventional way for subsequent iterations. If the new particle position fails any of the above verifications, a penalty is applied to the particle position, and a preset penalty term is added when calculating the particle fitness value, so that particles that do not meet the constraints are gradually eliminated in subsequent iterations.

[0087] First, after obtaining new particle positions in each iteration, the allocation of execution nodes for each type of unmanned equipment corresponding to the new particle position is analyzed. Since the particle position uses multi-dimensional vector encoding, the value of each dimension corresponds to the index number of the unmanned equipment execution node assigned to a specific equipment type in the candidate node set. Therefore, it is necessary to determine the specific combination of unmanned equipment execution nodes represented by the particle position through decoding operations.

[0088] Secondly, based on the lower limits of the load margin of multiple adaptable nodes in the adaptation scheduling constraints, it is verified whether the current node load margin of each unmanned equipment execution node assigned in the new particle position is greater than or equal to the lower limit of the load margin of the corresponding type of adaptable node. This step ensures that each selected unmanned equipment execution node in the scheme has sufficient remaining load capacity to reliably undertake the assigned tasks and avoids task execution failure due to excessive node load. Simultaneously, based on the upper limits of the overall average load of multiple adaptable equipment of the same type in the adaptation scheduling constraints, it is verified whether the overall average load of each type of unmanned equipment execution node in the new particle position is less than or equal to the upper limit of the overall average load of the corresponding type of adaptable equipment of the same type. This step, from a cluster-wide perspective, verifies whether the overall load level of each type of equipment is still within a reasonable range after the scheme is adopted, preventing the overall cluster load imbalance caused by one type of equipment undertaking too many tasks.

[0089] Then, based on the location coordinates of each unmanned equipment execution node and its distance from the mission execution area, combined with the travel speed of each node, the time required to reach the mission execution area is calculated. This verifies whether each assigned unmanned equipment execution node in the new particle position can reach the mission execution area within the remaining time window of the mission. This step ensures that all selected nodes in the plan have the ability to arrive at the mission site within the specified time limit, meeting the mission's timeliness requirements.

[0090] Specifically, if a new particle position simultaneously satisfies node load margin verification, overall average load verification for similar equipment, and task arrival time verification, then the particle position is retained and its fitness value is calculated in the conventional way to participate in subsequent iterations. A particle position that passes these three verifications represents a feasible scheduling scheme and can normally participate in fitness value calculation and subsequent optimization processes.

[0091] Furthermore, if a new particle position fails any of the above checks, a penalty is applied to the particle position. A preset penalty term is added when calculating the particle's fitness value, causing particles that do not meet the constraints to be gradually eliminated in subsequent iterations. The penalty term can be a large constant value or a dynamic value related to the degree of constraint violation, making the particle's fitness value significantly worse than a feasible solution, thus gradually eliminating it during the updates of individual and global optima. For example, the penalty term can be set to a constant much larger than the normal fitness value range, such as 10000.

[0092] By introducing the above constraint handling mechanism, it can be ensured that the particle swarm optimization algorithm always seeks optimization within the feasible solution space throughout the entire search process. The particle position corresponding to the final output global optimal position must simultaneously satisfy the node load capacity constraint, the overall cluster load constraint, and the task timeliness constraint, thereby ensuring the actual executability of the optimal equipment scheduling scheme.

[0093] In summary, the embodiments of this application have at least the following technical effects: First, the present invention evaluates the execution complexity and importance of the task through in-depth analysis of task attribute information, and dynamically corrects the baseline scheduling constraints based on this, generating scheduling constraints highly adapted to the current combat mission, thereby improving the adaptability and relevance of the scheduling strategy to the complex and ever-changing battlefield environment. Second, considering the comprehensive load status and position coordinates of the unmanned equipment execution nodes, the present invention comprehensively utilizes scheduling quality constraints, scheduling response constraints, and equipment demand constraints to conduct random scheduling simulation, generating diverse qualified equipment scheduling schemes, providing a rich feasible solution space for subsequent optimization, and avoiding the local optimum problem that may exist in a single scheme. Third, by establishing a multi-objective optimization function, the present invention evaluates the scheduling quality of the multiple generated qualified scheduling schemes, realizing a comprehensive trade-off between multiple conflicting optimization objectives such as resource consumption and load balancing, overcoming the limitations of traditional methods that only focus on a single performance index.

[0094] Finally, this invention aims to maximize the scheduling quality index of the scheme, uses an optimization algorithm to globally optimize qualified scheduling schemes, outputs the optimal scheduling scheme, and precisely controls the execution nodes of unmanned equipment. Thus, under the premise of meeting complex constraints, it realizes efficient, balanced and coordinated scheduling of unmanned equipment cluster combat missions, and improves the combat effectiveness and resource utilization of unmanned equipment clusters.

[0095] Example 2, as shown in Figure 2, based on the same inventive concept as the distributed combat mission scheduling method for unmanned equipment provided in Example 1, this embodiment of the invention also provides a distributed combat mission scheduling system for unmanned equipment, including: a constraint analysis module 11, used to evaluate and obtain the execution complexity and task importance of the target task based on the task attribute information of the target task in a centralized scheduling management platform, dynamically correct the preset benchmark scheduling constraints, and generate adaptive scheduling constraints; and a scheme generation module 12, used to generate a scheme based on the comprehensive load status data and position coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, and using any... The remaining time window serves as the scheduling response constraint, and the type and quantity of equipment required for the task serve as the equipment demand constraint. Random scheduling simulation is performed on the several unmanned equipment execution nodes to generate multiple qualified equipment scheduling schemes. A multi-objective optimization evaluation module 13 is used to establish a multi-objective optimization function to evaluate the scheduling quality of the multiple qualified equipment scheduling schemes and determine multiple scheme scheduling quality indices. An optimization and scheduling control module 14 is used to optimize the multiple qualified equipment scheduling schemes based on the multiple scheme scheduling quality indices, using an optimization algorithm to optimize the scheduling of the multiple qualified equipment scheduling schemes, output the optimal equipment scheduling scheme, and perform scheduling control on the several unmanned equipment execution nodes.

[0096] The constraint analysis module 11 is specifically used for: evaluating and obtaining the execution complexity and task importance of the target task based on the task attribute information of the target task, including: obtaining the task attribute information of the target task, wherein the task attribute information includes at least the task execution area, task characteristics, environmental conditions, task level, and remaining task time window; inputting the task characteristics into a preset task type and basic complexity mapping table to determine the basic complexity of the task type corresponding to the current task characteristics; extracting the weather condition score, electromagnetic interference intensity score, and terrain complexity score from the environmental conditions, respectively, and weighting and summing them according to preset environmental weight coefficients to calculate the environmental complexity reflecting the severity of the current working environment; weightedly fusing the basic complexity of the task type and the environmental complexity to obtain the execution complexity; and inputting the task level into a preset... A grade-importance mapping function is used to calculate the grade importance component, where a higher task grade corresponds to a larger grade importance component value. The ratio of the remaining task time window to a preset standard operation time is calculated and input into a preset time urgency nonlinear function to calculate the time urgency component. This time urgency nonlinear function ensures the time urgency component remains low when the remaining task time window is sufficient and increases sharply when the remaining task time window approaches zero. The urgency weight of the time urgency component is dynamically determined based on the remaining task time window, where the urgency weight is inversely proportional to the remaining task time window. The grade weight is obtained by subtracting the urgency weight from 1. Based on the grade weight and the urgency weight, the grade importance component and the time urgency component are weighted and summed to obtain the task importance.

[0097] Further, the preset baseline scheduling constraints are dynamically corrected to generate adaptive scheduling constraints, including: obtaining the baseline scheduling constraints, wherein the baseline scheduling constraints include multiple node load margin lower limits corresponding to multiple equipment types and multiple overall average load upper limits for the same type of equipment; using the ratio of the execution complexity to the preset baseline execution complexity as a first correction coefficient; using the ratio of the task importance to the preset baseline task importance as a second correction coefficient; performing a weighted summation of the first and second correction coefficients, using the weighted summation result as a node load margin compensation coefficient, and using the reciprocal of the weighted summation result as an overall average load upper limit compensation coefficient; correcting the multiple node load margin lower limits according to the node load margin compensation coefficient, and correcting the multiple overall average load upper limits for the same type of equipment according to the overall average load upper limit compensation coefficient, to obtain multiple adaptive node load margin lower limits and multiple adaptive overall average load upper limits for the same type of equipment, which serve as adaptive scheduling constraints.

[0098] Specifically, the scheme generation module 12 is used to: Based on the comprehensive load status data and position coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, the remaining time window of the task as scheduling response constraints, and the type and quantity of equipment required by the task as equipment demand constraints, perform random scheduling simulation on the several unmanned equipment execution nodes to generate multiple qualified equipment scheduling schemes. This includes: selecting unmanned equipment execution nodes of various types that match the type of equipment required by the task from the several unmanned equipment execution nodes, as the initial node set corresponding to each type; selecting nodes whose current node load margin is greater than or equal to the corresponding type's adaptive node load margin lower limit from the initial node set corresponding to each type, based on the multiple adaptive node load margin lower limits in the adaptive scheduling constraints, to form a first-level candidate node set corresponding to each type; and calculating the time required to reach the task execution area based on the distance between the position coordinates of each candidate node and the task execution area, combined with the sailing speed of each candidate node. From the first-level candidate node set corresponding to each type, candidate nodes that can reach the task execution area within the remaining time window of the task are selected to form a second-level candidate node set corresponding to each type. According to the equipment type and quantity required by the task, a specified number of candidate nodes are randomly selected from the second-level candidate node set corresponding to each type, and the selected candidate nodes are combined to generate a candidate equipment scheduling scheme. The above random selection and combination process is repeated to generate multiple candidate equipment scheduling schemes. For each candidate equipment scheduling scheme, the overall average load of each type of unmanned equipment execution node in the current cluster is obtained. According to the upper limit of the overall average load of multiple adapted equipment of the same type in the adaptation scheduling constraints, it is determined whether the overall average load of each type of unmanned equipment execution node is less than or equal to the upper limit of the overall average load of the corresponding type of adapted equipment of the same type after the candidate equipment scheduling scheme is adopted. The candidate equipment scheduling scheme that meets the above overall average load upper limit requirement is taken as a qualified equipment scheduling scheme, until a preset number of qualified equipment scheduling schemes are obtained.

[0099] The multi-objective optimization evaluation module 13 is specifically used for: establishing a multi-objective optimization function to evaluate the scheduling quality of the multiple qualified equipment scheduling schemes and determine multiple scheme scheduling quality indices, including: establishing a multi-objective optimization function that includes the objectives of minimizing resource consumption and maximizing load balancing; for each qualified equipment scheduling scheme, comprehensively calculating the resource consumption target value of the scheme based on the total travel distance of all unmanned equipment execution nodes from their current positions to the task execution area, and the estimated total energy consumption calculated based on the unit distance energy consumption of each type of unmanned equipment; for each qualified equipment scheduling scheme, comprehensively calculating the load balancing target value of the scheme based on the variance of the node load margin between execution nodes of the same type of unmanned equipment and the variance of the node load margin between execution nodes of different types of unmanned equipment; multiplying the resource consumption target value and the load balancing target value of each qualified equipment scheduling scheme by the first weighting coefficient and the second weighting coefficient respectively, and summing them, and using the reciprocal of the calculation result as the scheme scheduling quality index of each qualified equipment scheduling scheme.

[0100] Specifically, a time urgency factor is calculated based on the remaining time window of the task, and a cluster resource stress factor is calculated based on the average load margin of all unmanned equipment execution nodes in the current cluster. The time urgency factor and the cluster resource stress factor are input into a preset weight allocation model, and the first weight coefficient corresponding to the goal of minimizing resource consumption and the second weight coefficient corresponding to the goal of maximizing load balancing are output.

[0101] The optimization and scheduling control module 14 is specifically used for: optimizing the scheduling of multiple qualified equipment scheduling schemes based on the multiple scheduling quality indices, with the goal of maximizing the scheme scheduling quality index; including: encoding each qualified equipment scheduling scheme as a particle position in a particle swarm optimization algorithm, wherein the particle position is represented by a multi-dimensional vector, and the value of each dimension corresponds to the index number of the unmanned equipment execution node allocated to the equipment type required for a task in the candidate node set; initializing a particle swarm containing a preset number of particles, using the multiple qualified equipment scheduling schemes as the initial particle positions, and randomly initializing a velocity vector for each particle; and adjusting the scheduling of each particle's scheme... The fitness index is used as the fitness value of a particle. Based on the fitness value, the individual optimal position of each particle and the global optimal position of the entire particle swarm are determined. According to the velocity update formula and position update formula of the particle swarm optimization algorithm, combined with inertia weight, individual learning factor and social learning factor, the velocity and position of each particle are iteratively updated. In each iteration, the scheme scheduling quality index of the corresponding scheduling scheme is recalculated as the fitness value based on the updated particle position, and the individual optimal position and the global optimal position are updated. The iterative update is repeated until the preset maximum number of iterations is reached or the fitness value converges to a stable state. The particle position corresponding to the global optimal position at the end of the iteration is decoded to obtain the optimal equipment scheduling scheme.

[0102] The particle swarm optimization algorithm incorporates an inertia weight in its speed update formula. This inertia weight decreases linearly from the initial inertia weight to the final inertia weight as the number of iterations increases, thus emphasizing global search in the early stages of iteration and local fine-grained search in the later stages. Furthermore, the particle swarm optimization algorithm dynamically adjusts the individual learning factor and the social learning factor based on the remaining time window of the task. When the remaining time window is less than a preset time threshold, the social learning factor is increased and the individual learning factor is decreased to accelerate particle convergence to the global optimum, meeting the rapid optimization requirements in time-sensitive scenarios.

[0103] Specifically, the particle swarm optimization algorithm introduces a constraint handling mechanism during its iterative update process, including: after obtaining a new particle position in each iteration, analyzing the allocation of execution nodes of each type of unmanned equipment corresponding to the new particle position; verifying whether the current node load margin of each unmanned equipment execution node allocated in the new particle position is greater than or equal to the corresponding type of adaptation node load margin lower limit based on the multiple adaptation node load margin lower limits in the adaptation scheduling constraints; and verifying whether the overall average load of each type of unmanned equipment execution node in the new particle position is less than or equal to the corresponding type of adaptation equipment overall average load upper limit based on the multiple adaptation equipment overall average load upper limits in the adaptation scheduling constraints. Based on the location coordinates of each unmanned equipment execution node and the distance to the mission execution area, and combined with the travel speed of each node, the time required to reach the mission execution area is calculated. It is then verified whether each unmanned equipment execution node assigned in the new particle position can reach the mission execution area within the remaining time window of the mission. If the new particle position simultaneously satisfies the node load margin verification, the overall average load verification of similar equipment, and the mission arrival time verification, the particle position is retained and its fitness value is calculated in the conventional way to participate in subsequent iterations. If the new particle position fails to pass any of the above verifications, a penalty is applied to the particle position. A preset penalty term is added when calculating the particle fitness value, so that particles that do not meet the constraints are gradually eliminated in subsequent iterations.

[0104] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Clearly, those skilled in the art can make various alterations and variations to the invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the invention and its equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A distributed combat mission scheduling method for unmanned equipment, characterized in that, The method includes: on a centralized scheduling management platform, evaluating the execution complexity and importance of the target task based on its task attribute information, dynamically correcting preset baseline scheduling constraints, and generating adaptive scheduling constraints; based on the comprehensive load status data and location coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, the remaining time window of the task as scheduling response constraints, and the type and quantity of equipment required by the task as equipment demand constraints, performing random scheduling simulation on the several unmanned equipment execution nodes to generate multiple qualified equipment scheduling schemes; establishing a multi-objective optimization function to evaluate the scheduling quality of the multiple qualified equipment scheduling schemes and determine multiple scheme scheduling quality indices; using the maximization of the scheme scheduling quality index as the optimization objective, optimizing the multiple qualified equipment scheduling schemes using an optimization algorithm based on the multiple scheme scheduling quality indices, outputting the optimal equipment scheduling scheme, and performing scheduling control on the several unmanned equipment execution nodes.

2. The distributed combat mission scheduling method for unmanned equipment according to claim 1, characterized in that, The evaluation of the execution complexity and importance of the target task based on its task attribute information includes: acquiring the task attribute information of the target task, wherein the task attribute information includes at least the task execution area, task characteristics, environmental conditions, task level, and remaining time window; inputting the task characteristics into a preset task type-to-basic complexity mapping table to determine the basic complexity of the task type corresponding to the current task characteristics; extracting the weather condition score, electromagnetic interference intensity score, and terrain complexity score from the environmental conditions, respectively, and weighting and summing them according to preset environmental weight coefficients to calculate the environmental complexity reflecting the severity of the current working environment; weightedly fusing the basic complexity of the task type and the environmental complexity to obtain the execution complexity; and inputting the task level into a preset level-to-importance mapping function. The importance component is calculated, where a higher task level results in a larger importance component value. The ratio of the remaining task time window to a preset standard work duration is calculated and input into a preset time urgency nonlinear function to calculate the time urgency component. This nonlinear function ensures the time urgency component remains low when the remaining task time window is sufficient and increases when the remaining time window approaches zero. The urgency weight of the time urgency component is dynamically determined based on the remaining task time window, where the urgency weight is inversely proportional to the remaining task time window. The level weight is obtained by subtracting the urgency weight from 1. Based on the level weight and the urgency weight, the importance component and the time urgency component are weighted and summed to obtain the task importance.

3. The distributed combat mission scheduling method for unmanned equipment according to claim 2, characterized in that, The process involves dynamically correcting preset baseline scheduling constraints to generate adaptive scheduling constraints. This includes: obtaining baseline scheduling constraints, where the baseline constraints include multiple node load margin lower limits corresponding to multiple equipment types and multiple overall average load upper limits for the same type of equipment; using the ratio of the execution complexity to a preset baseline execution complexity as a first correction coefficient; using the ratio of the task importance to a preset baseline task importance as a second correction coefficient; performing a weighted summation of the first and second correction coefficients, using the weighted summation result as a node load margin compensation coefficient, and using the reciprocal of the weighted summation result as an overall average load upper limit compensation coefficient; correcting the multiple node load margin lower limits according to the node load margin compensation coefficient, and correcting the multiple overall average load upper limits for the same type of equipment according to the overall average load upper limit compensation coefficient, to obtain multiple adaptive node load margin lower limits and multiple adaptive overall average load upper limits for the same type of equipment, which serve as adaptive scheduling constraints.

4. The distributed combat mission scheduling method for unmanned equipment according to claim 3, characterized in that, Based on the comprehensive load status data and position coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, the remaining time window of the task as scheduling response constraints, and the type and quantity of equipment required by the task as equipment demand constraints, a random scheduling simulation is performed on the several unmanned equipment execution nodes to generate multiple qualified equipment scheduling schemes. This includes: selecting unmanned equipment execution nodes of various types that match the type of equipment required by the task from the several unmanned equipment execution nodes, as the initial node set corresponding to each type; selecting nodes whose current node load margin is greater than or equal to the corresponding type's adaptive node load margin lower limit from the initial node set corresponding to each type, based on the multiple adaptive node load margin lower limits in the adaptive scheduling constraints, forming a first-level candidate node set corresponding to each type; calculating the time required to reach the task execution area based on the distance between the position coordinates of each candidate node and the task execution area, combined with the travel speed of each candidate node, and selecting from the first-level... Candidate nodes that can reach the task execution area within the remaining time window of the task are selected from the candidate node set to form a second-level candidate node set corresponding to each type. According to the equipment type and quantity required by the task, a specified number of candidate nodes are randomly selected from the second-level candidate node set corresponding to each type, and the selected candidate nodes are combined to generate a candidate equipment scheduling scheme. The above random selection and combination process is repeated to generate multiple candidate equipment scheduling schemes. For each candidate equipment scheduling scheme, the overall average load of each type of unmanned equipment execution node in the current cluster is obtained. According to the upper limit of the overall average load of multiple adapted equipment of the same type in the adaptation scheduling constraints, it is determined whether the overall average load of each type of unmanned equipment execution node is less than or equal to the upper limit of the overall average load of the corresponding type of adapted equipment of the same type after the candidate equipment scheduling scheme is adopted. The candidate equipment scheduling scheme that meets the above overall average load upper limit requirement is taken as a qualified equipment scheduling scheme, until a preset number of qualified equipment scheduling schemes are obtained.

5. The distributed combat mission scheduling method for unmanned equipment according to claim 1, characterized in that, A multi-objective optimization function is established to evaluate the scheduling quality of multiple qualified equipment scheduling schemes and determine multiple scheduling quality indices. This includes: establishing a multi-objective optimization function that includes the objectives of minimizing resource consumption and maximizing load balancing; for each qualified equipment scheduling scheme, calculating the resource consumption target value based on the total travel distance of all unmanned equipment execution nodes from their current positions to the mission execution area, and the estimated total energy consumption calculated based on the unit distance energy consumption of each type of unmanned equipment; for each qualified equipment scheduling scheme, calculating the load balancing target value based on the variance of node load margin between execution nodes of the same type of unmanned equipment and the variance of node load margin between execution nodes of different types of unmanned equipment; multiplying the resource consumption target value and load balancing target value of each qualified equipment scheduling scheme by the first weighting coefficient and the second weighting coefficient, respectively, and summing the results, using the reciprocal of the calculation results as the scheduling quality index of each qualified equipment scheduling scheme.

6. The distributed combat mission scheduling method for unmanned equipment according to claim 5, characterized in that, Calculate the time urgency factor based on the remaining time window of the task, calculate the cluster resource stress factor based on the average load margin of all unmanned equipment execution nodes in the current cluster, input the time urgency factor and the cluster resource stress factor into the preset weight allocation model, and output the first weight coefficient corresponding to the goal of minimizing resource consumption and the second weight coefficient corresponding to the goal of maximizing load balancing.

7. The distributed combat mission scheduling method for unmanned equipment according to claim 1, characterized in that, With maximizing the scheduling quality index as the optimization objective, an optimization algorithm is used to optimize the scheduling of multiple qualified equipment scheduling schemes based on the multiple scheduling quality indices. This includes: encoding each qualified equipment scheduling scheme as a particle position in a particle swarm optimization algorithm, where each particle position is represented by a multi-dimensional vector, and the value of each dimension corresponds to the index number of the unmanned equipment execution node assigned to the equipment type required for a task in the candidate node set; initializing a particle swarm containing a preset number of particles, using the multiple qualified equipment scheduling schemes as the initial particle positions, and randomly initializing a velocity vector for each particle; and using the scheduling quality index of each particle as the particle's fitness index. Fitness values ​​are used to determine the individual optimal position of each particle and the global optimal position of the entire particle swarm. Based on the velocity and position update formulas of the particle swarm optimization algorithm, and combined with inertia weights, individual learning factors, and social learning factors, the velocity and position of each particle are iteratively updated. In each iteration, the scheduling quality index of the corresponding scheduling scheme is recalculated as the fitness value based on the updated particle positions, and the individual optimal position and the global optimal position are updated. This iterative update is repeated until the preset maximum number of iterations is reached or the fitness value converges to a stable state. The particle positions corresponding to the global optimal positions at the end of the iteration are decoded to obtain the optimal equipment scheduling scheme.

8. The distributed combat mission scheduling method for unmanned equipment according to claim 7, characterized in that, The particle swarm optimization algorithm introduces an inertia weight in its speed update formula. This inertia weight decreases linearly from the initial inertia weight to the final inertia weight as the number of iterations increases, thus emphasizing global search in the early stages of iteration and local fine-grained search in the later stages. The particle swarm optimization algorithm dynamically adjusts the individual learning factor and the social learning factor based on the remaining time window of the task. When the remaining time window is less than a preset time threshold, the social learning factor is increased and the individual learning factor is decreased to accelerate particle convergence to the global optimum, meeting the rapid optimization requirements in time-sensitive scenarios.

9. The distributed combat mission scheduling method for unmanned equipment according to claim 7, characterized in that, The particle swarm optimization algorithm introduces a constraint handling mechanism during its iterative update process, including: after obtaining a new particle position in each iteration, analyzing the allocation of execution nodes of each type of unmanned equipment corresponding to the new particle position; verifying whether the current node load margin of each unmanned equipment execution node allocated in the new particle position is greater than or equal to the corresponding type of adaptation node load margin lower limit based on the multiple adaptation node load margin lower limits in the adaptation scheduling constraints; verifying whether the overall average load of each type of unmanned equipment execution node in the new particle position is less than or equal to the corresponding type of adaptation node average load upper limit based on the multiple overall average load upper limits of the same type of adaptation equipment in the adaptation scheduling constraints; and further verifying whether the overall average load of each type of unmanned equipment execution node in the new particle position is less than or equal to the corresponding type of adaptation equipment average load upper limit. Based on the location coordinates of each unmanned equipment execution node and its distance from the mission execution area, and combined with the travel speed of each node, the time required to reach the mission execution area is calculated. It is then verified whether each assigned unmanned equipment execution node in the new particle position can reach the mission execution area within the remaining mission time window. If the new particle position simultaneously satisfies the node load margin verification, the overall average load verification of similar equipment, and the mission arrival time verification, the particle position is retained and its fitness value is calculated in the conventional way for subsequent iterations. If the new particle position fails any of the above verifications, a penalty is applied to the particle position, and a preset penalty term is added when calculating the particle fitness value, so that particles that do not meet the constraints are gradually eliminated in subsequent iterations.

10. A distributed combat mission scheduling system for unmanned equipment, characterized in that: The method for implementing the distributed combat mission scheduling method for unmanned equipment as described in any one of claims 1-9 includes: a constraint analysis module, used to evaluate and obtain the execution complexity and importance of the target operation task based on the task attribute information of the target operation task on a centralized scheduling management platform, dynamically correct the preset benchmark scheduling constraints, and generate adaptive scheduling constraints; a scheme generation module, used to perform random scheduling simulation on the several unmanned equipment execution nodes based on the comprehensive load status data and position coordinates of several unmanned equipment execution nodes, using the adaptive scheduling constraints as scheduling quality constraints, the remaining time window of the task as scheduling response constraints, and the type and quantity of equipment required by the task as equipment demand constraints, to generate multiple qualified equipment scheduling schemes; a multi-objective optimization evaluation module, used to establish a multi-objective optimization function, evaluate the scheme scheduling quality of the multiple qualified equipment scheduling schemes, and determine multiple scheme scheduling quality indices; and an optimization and scheduling control module, used to optimize the multiple qualified equipment scheduling schemes based on the multiple scheme scheduling quality indices using an optimization algorithm, output the optimal equipment scheduling scheme, and perform scheduling control on the several unmanned equipment execution nodes, with the goal of maximizing the scheme scheduling quality index.

Images