Unmanned aerial vehicle cluster cooperative task real-time processing method and system based on edge computing

Abstract

The present application relates to the field of unmanned aerial vehicle cluster task processing, and discloses a method and system for real-time processing of cooperative tasks of an unmanned aerial vehicle cluster based on edge computing, which builds a communication link relying on a SkyEdge-5G air-based edge cooperative platform, and collects parameters of unmanned aerial vehicles and edge nodes; a dynamic temptation-punishment algorithm is used to generate a matching constraint coefficient; a double-target Pareto optimization model of energy consumption-time delay is constructed, candidate edge nodes are filtered by superimposing node reliability and communication distance physical constraints; resource conflict checking is performed through the power and storage utilization threshold, and overloaded nodes are removed to obtain target unloading nodes; distribution instructions are issued to control the fragmented transmission of task data by unmanned aerial vehicles; edge nodes process tasks in stages, compress and return results, and the power of nodes and the power of unmanned aerial vehicles are updated through a dynamic model, and data is synchronized into a database to form a scheduling closed loop. The present application balances the total energy consumption and processing time delay of the cluster, dynamically adapts to resource fluctuations, and effectively solves the problems of high time delay and node overload in traditional ground scheduling.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm task processing technology, and specifically to a real-time processing method and system for UAV swarm collaborative tasks based on edge computing. Background Technology

[0002] With the rapid development of drone and communication technologies, drone swarms are increasingly being used in environmental monitoring, emergency rescue, and power line inspection. These scenarios place higher demands on the real-time performance and stability of drone swarm task processing. Traditional drone swarm task processing relies heavily on ground-based central nodes for unified scheduling, which suffers from issues such as communication links being susceptible to interference and long data transmission latency, making it difficult to meet the demands of high real-time tasks. The rise of edge computing technology offers a solution to this problem. By deploying computing resources at edge nodes close to the drone swarm, data transmission distance and latency can be reduced. Furthermore, the SkyEdge-5G airborne edge collaboration platform, with its wide coverage and high bandwidth communication advantages, can further improve the communication stability between the drone swarm and edge nodes.

[0003] Existing technologies have two significant drawbacks in the collaborative task processing of drone swarms: First, the task allocation mechanism lacks dynamic adaptability. Most technologies allocate tasks based solely on the drone task type or edge node resources, without fully considering dynamic factors such as changes in edge node resource load and fluctuations in drone remaining battery power during task processing. This can lead to situations where some edge nodes are overloaded or some drones interrupt tasks due to insufficient battery power after task allocation, reducing the overall task processing efficiency of the swarm. Second, the ability to coordinate energy consumption and latency optimization is insufficient. Existing technologies often focus on single-dimensional performance optimization, or although they attempt multi-objective optimization, they lack effective constraint mechanisms, making it impossible to achieve a balance between energy consumption control and latency reduction. Especially when large-scale drone swarms process multiple types of tasks simultaneously, problems such as excessive energy consumption leading to shortened drone endurance or excessive latency affecting task real-time performance can easily occur. Summary of the Invention

[0004] The present invention aims to provide a real-time processing method and system for collaborative tasks of UAV swarms based on edge computing, in order to solve the problems of large communication interference and high transmission latency in the traditional ground center scheduling mode, and the inability of existing offloading schemes to dynamically balance the energy consumption and task processing latency of UAV swarms, and the easy occurrence of resource overload at edge nodes.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: Real-time processing methods for drone swarm collaborative tasks based on edge computing include: S1, through the SkyEdge-5G airborne edge collaboration platform, builds a communication link between the drone cluster and edge nodes, and obtains the task type, task data volume, remaining power, current location information of each drone in the drone cluster, as well as the computing resource load, storage capacity, and communication bandwidth parameters of the edge nodes; S2, based on the parameters of the drones and edge nodes collected in S1, uses a dynamic enticement-penalty mechanism task allocation algorithm to solve the task enticement coefficient and task delay penalty coefficient for each drone corresponding to different edge nodes. S3 adopts the energy consumption and time delay Pareto optimization unloading model, with the task temptation coefficient and task delay penalty coefficient as constraints, to establish a multi-objective optimization function for UAV task unloading and determine the candidate edge node set for each UAV task unloading. S4. For the candidate edge node set, resource conflict detection is performed from two dimensions: computing resources and storage capacity. Based on the resource utilization threshold, it is determined whether the node is overloaded and the target edge node without resource conflict is selected. S5 sends target edge node allocation instructions to each UAV through the SkyEdge-5G airborne edge collaboration platform, and controls the UAV to transmit mission data to the target edge node. S6, the target edge node completes real-time computation and processing of the task data uploaded by the drone. After processing, the results are fed back to the corresponding drone through the SkyEdge-5G airborne edge collaboration platform, while updating the edge node resource load parameters and the drone's remaining power parameters.

[0006] The principle and advantages of this solution are as follows: In practical applications, the SkyEdge-5G airborne edge collaboration platform serves as the carrier for full-domain communication and data acquisition, collecting real-time data on drone missions, power consumption, location, and edge node computing power, storage, and bandwidth. A dynamic enticement-penalty algorithm is used to quantify the matching relationship between drones and edge nodes from two perspectives: resource adaptation and latency / power consumption risk. Two sets of coefficients are used as constraints to build a dual-objective Pareto optimization model for energy consumption and latency, with multiple pre-constraints used to initially screen candidate nodes. Resource overload verification is performed based on computing power and storage utilization thresholds, eliminating conflicting nodes to obtain allocable target nodes. Allocation instructions are issued and segmented data transmission is managed via the 5G link. Edge nodes complete task calculations, transmit results, and dynamically refresh resource and power parameters, updating data back to the acquisition stage to form a closed-loop scheduling, thus fully realizing collaborative scheduling of the entire process of acquisition, matching optimization, resource verification, data transmission, and computational updates.

[0007] Beneficial effects: (1) By introducing a dynamic temptation-penalty mechanism task allocation algorithm, the temptation and penalty coefficients are calculated by combining multiple dimensions such as UAV task type, remaining power, edge node resource load, and communication bandwidth. This breaks the single-dimensional allocation mode and adapts in real time to changes in edge node resources and fluctuations in UAV power during task processing, effectively avoiding edge node overload and UAV task interruption. This solves the shortcomings of the task allocation mechanism lacking dynamic adaptability. (2) By constructing a multi-objective optimization function through the energy consumption and time delay Pareto optimization unloading model, the coefficients obtained from the task allocation algorithm are used as constraints to achieve coordinated optimization of energy consumption and time delay, rather than single-dimensional optimization. At the same time, by screening target edge nodes through resource conflict detection, the computing and storage resources of edge nodes are ensured to meet task requirements, balancing energy consumption control and time delay reduction, overcoming the problem of insufficient energy consumption and time delay coordinated optimization capability. (3) By relying on the SkyEdge-5G airborne edge collaboration platform to build a stable communication link, the real-time and stability of task data transmission and result feedback are guaranteed, further improving the cluster task processing efficiency and making up for the communication interference and long delay defects in ground center node scheduling.

[0008] Preferably, as an improvement, in the dynamic enticement-penalty mechanism task allocation algorithm, the calculation model for the task enticement coefficient of the edge node j corresponding to drone i is as follows:

[0009] in, Let be the attraction coefficient of edge node j corresponding to drone i. These are weighting coefficients for computing power and bandwidth. , , The remaining computing resources for edge node j, For the task computation requirements of drone i, Let j be the communication bandwidth of edge node j. The amount of task data for drone i; The calculation model for the task latency penalty coefficient of edge node j corresponding to drone i is as follows:

[0010] in, Let be the latency penalty coefficient for edge node j corresponding to drone i. For latency and power consumption weighting coefficients, , The task transmission latency from drone i to edge node j, The maximum allowable transmission delay of the system. The amount of power consumed by drone i. This is the initial battery level of drone i.

[0011] Technical effects: By constructing a positive incentive index based on computing power and bandwidth weighting, the adaptability of edge nodes to drone missions is quantified; a negative penalty index based on transmission latency and remaining power is designed to identify high latency and high power consumption matching schemes in advance; breaking through the limitations of traditional allocation based on a single resource dimension, the matching quality of each drone and edge node is accurately quantified, providing reliable constraint input for subsequent optimization models.

[0012] Preferably, as an improvement, the multi-objective optimization function of the energy consumption time delay Pareto optimization unloading model is:

[0013] in, For binary 0-1 decision variables, The total energy consumption of the drone swarm. Total task processing time;

[0014]

[0015] Where n is the number of drones and m is the number of edge nodes. The energy consumption for task transmission from UAV i to edge node j. The computational energy consumption for edge node j to process drone i's task. For the task transmission latency from UAV i to edge node j The computation time for edge node j to process drone i task.

[0016] Technical effect: Output multiple sets of balanced optimal unloading schemes to avoid the surge in energy consumption or excessive latency caused by single-objective optimization, and achieve a coordinated balance between energy consumption and latency when unloading massive drone cluster tasks.

[0017] Preferably, as an improvement, in step S3, a multi-objective optimization function is established in conjunction with physical constraints, including edge node reliability constraints and 5G communication distance constraints; Edge node reliability constraints include:

[0018] in, Let be the reliability coefficient of edge node j; The minimum reliability threshold set for the system. ; The calculation formula is:

[0019] in, For edge node j, specify the number of historical task failures. For edge node j, this represents the total number of historical task processing times. 5G communication distance constraints include:

[0020] in, Let be the straight-line distance between drone i and edge node j. Maximum communication distance supported by the SkyEdge-5G airborne edge collaboration platform.

[0021] Technical effects: Increased historical reliability verification of nodes eliminates unstable nodes with frequent failures; superimposed SkyEdge-5G maximum communication distance limit filters invalid matching combinations beyond the communication range, reduces the computational load of subsequent resource detection, and improves the overall stability and communication reachability of candidate edge nodes.

[0022] Preferably, as an improvement, step S3 includes the following sub-steps: S31, the task enticement coefficient of each drone's corresponding edge node With task delay penalty coefficient Standardize each linear mapping to interval; S32, configured with a pair of balanced energy consumption weighting factors With delay weighting coefficient , To construct a comprehensive evaluation index for energy consumption delay; S33. Sort the edge nodes according to the comprehensive evaluation index, and select the top k edge nodes, where k is the preset number of candidate nodes. m is the total number of edge nodes, forming the candidate edge node set for the corresponding UAV; S34, re-verify the available communication bandwidth of the edge nodes in the candidate edge node set, remove nodes whose communication bandwidth cannot support the data transmission requirements of the UAV mission, and update the candidate edge node set.

[0023] Technical effects: Standardizing coefficients eliminates dimensional differences, and generating comprehensive evaluation indicators through adjustable weights uniformly quantifies the merits of nodes; layered initial selection with secondary bandwidth filtering simplifies the candidate set, reduces the computational overhead of subsequent resource conflict detection, and improves the overall real-time performance of scheduling.

[0024] Preferably, as an improvement, in step S4, determining whether a node is experiencing resource overload based on a resource utilization threshold includes: Calculate the computational resource utilization of edge node j:

[0025] in, The computational resource utilization rate of edge node j. For the set of drone tasks that have been assigned to edge node j, For set The computational requirements of the UAV k mission The remaining computing resources for edge node j; when At that time, it was determined that there was no conflict of computing resources. Calculate the maximum resource utilization threshold for edge nodes. ; Calculate storage resource utilization:

[0026] in, The storage resource utilization rate of edge node j For set The amount of mission data for UAV k in China The remaining storage capacity of edge node j; when At that time, it was determined that there was no storage resource conflict. The threshold for maximum utilization of storage resources on edge nodes. .

[0027] Technical benefits: The system quantifies node load status using real-time resource utilization, sets dual thresholds for computing power and storage to distinguish between overload and normal states, replacing simple resource value comparisons, accurately predicts the risk of resource overflow after adding new tasks, and ensures the stable operation of edge nodes.

[0028] Preferably, as an improvement, S4 includes the following sub-steps: S41, the real-time remaining computing resources and storage capacity of each edge node in the candidate edge node set, as well as the task computing requirements and task data volume of the drone to be assigned; S42, For each candidate edge node, sum up the total computing requirements of all UAV tasks already allocated to that node, compare it with the node’s remaining computing resources, and determine whether it exceeds the node’s computing resource carrying capacity. S43. For each candidate edge node, sum up the total amount of all UAV mission data already allocated to that node, compare it with the node’s remaining storage capacity, and determine whether it exceeds the node’s storage resource carrying capacity. S44. After resource conflict detection is completed, if the computing resources and storage resources of a candidate edge node simultaneously meet the carrying capacity threshold requirements, it is retained as a target edge node without resource conflicts. If they do not meet the requirements simultaneously, the process returns to S3 to readjust the set of candidate edge nodes.

[0029] Technical effects: Simultaneously verify the carrying capacity of both computing power and storage resources, and only nodes that meet both indicators can be used as target nodes; when there are no qualified candidates, the selection will automatically fall back and be reselected, thus preventing tasks from being assigned to overloaded nodes and reducing the probability of task processing failure.

[0030] Preferably, as an improvement, S5 includes the following sub-steps: S51 generates allocation instructions containing UAV identifiers, target edge node identifiers, and data transmission protocols through the control module of the SkyEdge-5G airborne edge collaboration platform; S52 broadcasts the allocation command to the corresponding drone via the 5G communication link. After receiving the command, the drone parses the IP address and port information of the target edge node. S53: The UAV establishes a point-to-point data transmission channel with the target edge node based on the resolved IP address and port information, and dynamically configures two transmission parameters: transmission rate and data fragment size. S54: The UAV transmits mission data in segments to the target edge node according to the set parameters, and provides real-time feedback on the transmission progress to the SkyEdge-5G airborne edge collaboration platform during the transmission process.

[0031] Technical benefits: The binary broadcast command combined with the acknowledgment and retransmission mechanism ensures reliable command delivery; a point-to-point fragmented transmission channel is established, with dynamic rate limiting and bandwidth reservation, and real-time reporting of transmission progress, balancing transmission integrity and fairness in multi-machine concurrent communication.

[0032] Preferably, as an improvement, in step S6, the edge node updates its own resource load parameters using a dynamic update model as follows:

[0033] in, To update the remaining computing resources of edge node j, To update the remaining computing resources of the front edge node j, For the set of drones currently assigned to edge node j, For the task computation requirements of drone i, The edge node resource recovery coefficient is t, and the task processing time is t. The drone updates its remaining battery power parameters using a dynamic update model as follows:

[0034] in, The remaining battery level of drone i after the update. The remaining battery level of drone i before the update. The energy consumption for task transmission from UAV i to edge node j. Let t be the idle energy consumption rate of drone i, and t be the task processing time.

[0035] Technical effects: The introduction of a resource recovery coefficient dynamically corrects the remaining computing power of nodes, aligning with the natural release characteristics of hardware resources; the comprehensive power consumption during transmission and idle periods are used to accurately update the drone's power consumption, and the updated data is synchronized into the database to form a scheduling closed loop, ensuring that the resource and power data used for the next round of task allocation are accurate and in real time.

[0036] It also includes a real-time processing system for drone swarm collaborative tasks based on edge computing, which is applied to the real-time processing method for drone swarm collaborative tasks based on edge computing, including: SkyEdge-5G communication module, parameter acquisition module, task allocation module, optimization and unloading module, resource detection module, and result feedback module; The SkyEdge-5G communication module is used to build a communication link between the drone swarm and edge nodes, transmitting mission data and control commands; The parameter acquisition module is connected to the SkyEdge-5G communication module to acquire UAV mission parameters, UAV status parameters, and edge node resource parameters. The task allocation module is connected to the parameter acquisition module, and uses a dynamic temptation-penalty mechanism task allocation algorithm to calculate the temptation coefficient and the penalty coefficient. The connection between the unloading module and the task allocation module is optimized, and the energy consumption and time delay Pareto optimization unloading model is used to determine the candidate edge node set. The resource detection module is connected to the optimization and unloading module to detect resource conflicts at edge nodes and filter target edge nodes; The results feedback module connects the resource detection module and the SkyEdge-5G communication module. It is used to receive the task processing results of the edge node and feed them back to the UAV, while updating the resource and status parameters in the parameter acquisition module. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating a real-time processing method for collaborative tasks in a drone swarm based on edge computing. Figure 2 This is a schematic diagram of the structure of an embodiment of the present invention. Detailed Implementation

[0038] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 As shown, the real-time processing method for drone swarm collaborative tasks based on edge computing includes the following steps: S1 establishes a communication link between the UAV cluster and edge nodes through the SkyEdge-5G airborne edge collaboration platform, and obtains the task type, task data volume, remaining power, current location information of each UAV in the UAV cluster, as well as the computing resource load, storage capacity, and communication bandwidth parameters of the edge nodes.

[0039] Specifically, parameter acquisition is completed through the SkyEdge-5G airborne edge collaboration platform, providing a data foundation for subsequent task allocation and optimization. During implementation, the platform's communication capabilities and data acquisition functions are first used to establish a stable two-way data transmission channel between the UAV cluster and each edge node; then, the platform sends parameter acquisition commands to each UAV in the cluster, completing the information collection between the UAVs and the edge nodes.

[0040] The collected drone information mainly includes four categories: First, the task type, specifically including different operation categories such as environmental monitoring data collection, emergency material positioning, and power line inspection; second, the task data volume, which is dynamically set according to the task type. For example, when a single drone performs an environmental monitoring task, the amount of data generated in a single operation is 500~800MB; third, the remaining battery power, which is measured in milliampere-hours (mAh). The platform reads this status data in real time, such as 3000~5000mAh; and fourth, the current location information, which is recorded using latitude and longitude coordinates, with positioning accuracy controlled within 1~3 meters to ensure the accuracy of distance calculation.

[0041] The edge node information collected synchronously includes three core resource parameters: first, computing resource load, which is quantified by processor utilization and ranges from 0 to 100%; second, remaining storage capacity, in gigabytes (GB), such as 200 to 500 gigabytes; and third, real-time available communication bandwidth, in megabits per second (Mbps), which is the current available communication bandwidth obtained in real time, such as 50 to 100 megabits per second.

[0042] The parameters are collected once per second to ensure real-time data acquisition. After collection, the data is temporarily stored in the SkyEdge-5G space-based edge collaboration platform's built-in storage module, forming a complete parameter dataset to provide accurate and reliable data support for subsequent steps.

[0043] S2, based on the collected parameters of the drone and edge nodes, solves the task temptation coefficient and latency penalty coefficient respectively, and quantifies the matching relationship between the drone and each edge node from two dimensions: positive adaptation capability and reverse operation risk.

[0044] The task enticement coefficient quantifies the compatibility between edge node j and drone i for the task. A higher coefficient indicates that the node's computing power and bandwidth are more compatible with the drone's task requirements. For tasks with high computational demands, such as high-definition image recognition, priority is given to edge node computing resource load: when the edge node's processor utilization is below 30%, its computing power is considered sufficient, increasing the enticement of that edge node for drones with high computational demands. For large-volume data transmission tasks, the focus is on node communication bandwidth. If the available communication bandwidth of an edge node is higher than 80 megabits per second, the enticement coefficient is further increased for drones with task data exceeding 600 megabytes.

[0045] The calculation model for the task enticement coefficient of edge node j corresponding to drone i is as follows:

[0046] in, Let be the attraction coefficient of edge node j corresponding to drone i. These are weighting coefficients for computing power and bandwidth. , , The remaining computing resources for edge node j, For the task computation requirements of drone i, Let j be the communication bandwidth of edge node j. The amount of mission data for drone i.

[0047] Configure based on task resource dependency characteristics: For example, for computationally intensive tasks (image recognition, object detection), where the task relies more on computing resources, the computing power weight will be adjusted accordingly. The values ​​are 0.6 and 0.4 respectively; for large data transmission tasks, which rely more on communication bandwidth, then... The values ​​are 0.4 and 0.6 respectively.

[0048] The task latency penalty coefficient is used to quantify the latency and battery life loss risks associated with unloading drone i from edge node j. It acts as a reverse constraint, avoiding inferior nodes. The lower the task latency penalty coefficient, the lower the risk of transmission latency and battery loss. Based on the system's preset maximum allowable transmission latency for drone tasks (e.g., 50 milliseconds for emergency rescue tasks), the task latency penalty coefficient is calculated when the actual transmission latency between the drone and the target edge node exceeds 30 milliseconds. Simultaneously, the drone's battery status is considered. If the remaining battery is below 2000mAh, an additional battery penalty weight is applied. This approach avoids high-risk allocation schemes from both latency and battery life perspectives, preventing mission interruptions.

[0049] The calculation model for the task latency penalty coefficient of edge node j corresponding to drone i is as follows:

[0050] in, Let be the latency penalty coefficient for edge node j corresponding to drone i. For latency and power consumption weighting coefficients, , The task transmission latency from drone i to edge node j, The maximum allowable transmission delay of the system. The amount of power consumed by drone i. This is the initial battery level of drone i.

[0051] Configure weights based on the sensitivity dimension of the task: for latency-sensitive tasks (emergency rescue, real-time inspection). The values ​​are 0.7 and 0.3 respectively; for long-endurance inspections and low real-time tasks, The values ​​are 0.3 and 0.7 respectively.

[0052] After calculating the task temptation coefficient and task delay penalty coefficient, the task temptation coefficient and task delay penalty coefficient corresponding to each drone and each edge node are summarized and stored in the algorithm result database in a matrix data structure so that they can be directly called by the next process S3, providing constraints and basic calculation data for the energy consumption delay Pareto optimization unloading model.

[0053] S3 adopts the energy consumption and time delay Pareto optimization unloading model, and establishes a multi-objective optimization function for UAV task unloading with the task temptation coefficient and task penalty coefficient as constraints, and determines the candidate edge node set for each UAV task unloading.

[0054] Specifically, before constructing the multi-objective optimization function, the task temptation coefficient and task penalty coefficient are used as admission constraints to filter obviously mismatched nodes: the task temptation coefficient is used as a positive constraint, that is, only edge nodes with a coefficient higher than a set threshold (such as 0.6) are included in the optimization solution range; the task penalty coefficient is used as a negative constraint, that is, edge nodes with a coefficient higher than a set threshold (such as 0.4) are excluded.

[0055] After filtering, a multi-objective optimization function is established based on physical constraints. Using binary 0-1 variables as unloading decision indicators, a dual-objective multi-objective optimization function is constructed to minimize the total energy consumption of the cluster and the total processing latency of the task. The Pareto optimization algorithm is used to solve the problem and output feasible edge node schemes that balance energy consumption and latency, providing a calculation basis for selecting candidate edge node sets.

[0056] Set up a binary 0-1 decision variable Used to mark drone mission unloading matching relationships, a variable of 0 or 1. This indicates that the drone i's task is offloaded to edge node j, where it participates in the calculation of energy consumption and latency accumulation. This indicates that you will not uninstall.

[0057] The multi-objective optimization function of the energy consumption time delay Pareto optimization unloading model is:

[0058] in, For binary 0-1 decision variables, The total energy consumption of the drone swarm. Total task processing time;

[0059]

[0060] Where n is the number of drones (typically 10-50), and m is the number of edge nodes (typically 5-20). The energy consumption for data transmission from UAV i to edge node j is 50 to 80 milliamp-hours per 100 megabytes of data transmitted. The computational energy consumption for edge node j to process drone i tasks is 30 to 50 mAh per 1 gigahertz of computation. The task transmission latency from UAV i to edge node j is calculated based on the communication bandwidth. For example, the latency for transmitting 500 megabytes of data with a bandwidth of 50 megabits per second is 80 to 100 milliseconds. The computation time for edge node j to process drone i task is calculated based on the edge node's computing resources. For example, if the remaining computing resources of 2 gigahertz are used to process the 1 gigahertz requirement, the latency is 40 to 60 milliseconds.

[0061] By filtering out nodes with poor stability and unreachable communication links through physical constraints, the quality of the candidate set is optimized, reducing the failure probability of subsequent S4 resource conflict detection and improving the overall reliability and communication feasibility of task allocation. Physical constraints include edge node reliability constraints and 5G communication distance constraints.

[0062] Edge node reliability constraints include:

[0063] in, Let be the reliability coefficient of edge node j; The minimum reliability threshold set for the system. The standard value for this engineering approach is 0.8. Nodes with values ​​below this threshold lack stability and are therefore not used. During verification, the historical task database of edge nodes needs to be retrieved, requiring a total historical task volume for each node. No fewer than 100 times, to ensure that the reliability coefficient has statistical reference value and to avoid calculation deviations caused by a small amount of historical data. The calculation formula is:

[0064] in, For edge node j, specify the number of historical task failures. This represents the total number of historical task processing times for edge node j.

[0065] 5G communication distance constraints include:

[0066] in, Let be the straight-line distance between drone i and edge node j. The maximum communication distance supported by the SkyEdge-5G airborne edge collaboration platform is typically set at 50. Exceeding this distance will prevent the establishment of a stable transmission link. Distance measurement relies on the location information of the UAV's positioning module and edge nodes; positioning accuracy must be controlled within 1-3 units to ensure accurate distance data.

[0067] S3 includes the following steps: S31, the task enticement coefficient of each drone's corresponding edge node With task delay penalty coefficient Standardize each linear mapping to The interval is used to eliminate the impact of differences in the magnitude of different parameters on the optimization model.

[0068] S32, configured with a pair of balanced energy consumption weighting factors With delay weighting coefficient , Construct a comprehensive evaluation index for energy consumption and latency; adjust according to task scenarios. , Emergency tasks should place greater emphasis on latency, such as Taking values ​​of 0.4 and 0.6 respectively, the energy consumption weighting for long-distance navigation inspection tasks is increased, such as... Take values ​​of 0.6 and 0.4 respectively.

[0069] S33. Sort the edge nodes according to the comprehensive evaluation index, and select the top k edge nodes, where k is the preset number of candidate nodes. m is the total number of edge nodes, forming a candidate edge node set for the corresponding UAV; in this embodiment, the top 3 to 5 edge nodes are preferred as the candidate set, and the number of candidates is adjusted according to the total number of edge nodes. If the total number of edge nodes is less than 3, all of them are included.

[0070] S34. Re-verify the available communication bandwidth of the edge nodes in the candidate edge node set. Typically, the bandwidth is required to be no less than 50. Remove nodes whose communication bandwidth cannot support the data transmission requirements of the UAV mission and update the candidate edge node set.

[0071] After completing the candidate edge node set, communication stability is verified: the SkyEdge-5G platform is used to detect the communication packet loss rate of the candidate nodes in the past 10 seconds. If the packet loss rate is higher than 2%, the link is determined to be unstable and it is directly removed from the candidate set, thus forming the effective candidate edge node set for each drone.

[0072] S4. For the candidate edge node set, resource conflict detection is performed from two dimensions: computing resources and storage capacity. Based on the resource utilization threshold, it is determined whether the node is overloaded. Nodes that meet the resource carrying requirements of both computing resources and storage capacity are selected as the final target edge nodes.

[0073] Specifically, before detection, the remaining computing resources and storage capacity of edge nodes are collected in real time, requiring that the remaining computing resources of the nodes be ≥0.5GHz and the remaining storage capacity be ≥100GB. Based on this, the computing resource utilization rate and storage resource utilization rate are calculated respectively, and a preset threshold is used to determine whether there is a conflict.

[0074] The formula for calculating the computational resource utilization rate of edge node j is:

[0075] in, The computational resource utilization rate of edge node j. For the set of drone tasks that have been assigned to edge node j, For set The computational requirements of the UAV k mission The remaining computing resources for edge node j. When At that time, it was determined that there was no conflict of computing resources. Calculate the maximum resource utilization threshold for edge nodes. The standard value for engineering projects is 0.9.

[0076] The formula for calculating storage resource utilization is:

[0077] in, The storage resource utilization rate of edge node j For set The amount of mission data of UAV k in China The remaining storage capacity of edge node j; when At that time, it was determined that there was no storage resource conflict. The threshold for maximum utilization of storage resources on edge nodes. The standard value for engineering projects is 0.85.

[0078] For example, in computing resource conflict detection, if an edge node has 1.5 GHz of remaining computing resources, the total computing requirement of allocated tasks is 0.8 GHz, and the computing requirement of the currently pending drone task is 0.6 GHz, the total requirement is 1.4 GHz. Since the total requirement is less than the remaining computing resources, there is no computing resource conflict. However, if the requirement of the pending task increases to 0.8 GHz, the total requirement becomes 1.6 GHz, exceeding the remaining computing resources, then a computing resource conflict exists. Storage resource conflict detection works similarly. The remaining storage capacity of the candidate edge node is obtained, for example, 300 gigabytes. The data volume of allocated tasks and the data volume of the currently pending task are added together. For example, if the allocated data volume is 180 gigabytes and the currently pending data volume is 100 gigabytes, the total data volume is 280 gigabytes, which does not exceed the remaining storage capacity, so there is no storage resource conflict. If the pending data volume increases to 130 gigabytes, the total data volume becomes 310 gigabytes, exceeding the remaining capacity, then a storage resource conflict is determined.

[0079] S4 includes the following sub-steps: S41: Obtain the real-time remaining computing resources and storage capacity of each edge node in the candidate edge node set obtained in S3, as well as the task computing requirements and task data volume of the drone to be assigned. The parameter acquisition frequency is set to once per second to ensure that the data can reflect the resource and task status in real time.

[0080] S42, for each candidate edge node, sums up the total computing requirements of all UAV tasks already allocated to that node. Typically, it is required that the remaining computing resources retain at least 0.2 gigahertz of redundancy to avoid overload caused by sudden load. It is compared with the node's remaining computing resources to determine whether the computing power capacity limit is exceeded after adding new tasks.

[0081] S43. For each candidate edge node, sum up the total amount of all UAV mission data already allocated to that node and compare it with the node's remaining storage capacity. The remaining storage capacity usually needs to retain 50 gigabytes of redundancy to determine whether the storage capacity limit is exceeded after adding new missions.

[0082] S44. After resource conflict detection is completed, a candidate edge node is retained as a target edge node without resource conflicts only if both its computing resources and storage resources meet the carrying capacity threshold requirements. If a node only meets the computing resource requirements or only has sufficient storage resources, the node is not included in the range of available target nodes.

[0083] If none of the candidate edge nodes in this round can simultaneously meet the computing and storage resource requirements, and no qualified target node exists, the first step is to verify whether the node resources and task parameters used in this detection were collected accurately. If there are collection errors in the parameters, the parameters are corrected and the resource conflict detection is re-executed. If the parameter data is accurate, the process reverts to S3, adjusts the weight coefficients of the energy consumption delay Pareto optimization unloading model (e.g., increasing the weights related to computing resources), recalculates to generate a new set of candidate edge nodes, and sends them back to S4 for resource conflict verification. This loop continues until a target edge node that simultaneously meets the computing power and storage requirements is selected.

[0084] S5 sends target edge node allocation commands to each UAV through the SkyEdge-5G airborne edge collaboration platform, controlling the UAVs to transmit mission data to the target edge nodes. S5 includes the following steps: The S51, through the control module of the SkyEdge-5G airborne edge collaboration platform, generates unique assignment instructions for each UAV. The instructions include the UAV's unique identifier (such as UAV number UAV-001 to UAV-100), the target edge node identifier (such as edge node number EN-01 to EN-20), and the data transmission protocol (using TCP / IP protocol to ensure reliability). The instruction format uses binary encoding to reduce the amount of data transmitted. The instructions use a fixed format encoding, with the length controlled between 100 and 200 bytes to ensure that the amount of data transmitted is small and the information is complete.

[0085] The S52 SkyEdge-5G airborne edge collaboration platform uses 23 dBmW power to push and distribute commands via a 5G broadcast channel, with a communication coverage range of 10-50 kilometers, ensuring stable reception by drones. After receiving the command, the drone resolves the IP address and port corresponding to the target edge node, with the resolution time controlled within 50 milliseconds. After successful reception, the drone sends back a confirmation signal via the feedback channel. If the platform does not receive a confirmation within 100ms, it retransmits the command, up to 3 times. If there is still no confirmation, the drone is marked as abnormal and the record is kept.

[0086] The S53 drone establishes a dedicated point-to-point transmission link with the target edge node based on the resolved IP and port information. It dynamically configures two transmission parameters: transmission rate and data fragment size, balancing transmission efficiency and bandwidth fairness among multiple devices. The transmission rate is adjusted according to the real-time available bandwidth of the edge node. For example, if the available bandwidth of the edge node is 60 megabits per second, the transmission rate is set to 40 to 50 megabits per second to avoid occupying all the bandwidth and affecting other data transmissions. The data fragment size is uniformly set to 10 to 20 megabytes. After each fragment is transmitted, the edge node sends a fragment confirmation signal to the drone. The drone transmits the next fragment only after receiving the confirmation.

[0087] S54: The UAV transmits mission data in segments to the target edge node according to the set parameters. During the transmission process, it reports the transmission progress to the SkyEdge-5G airborne edge collaboration platform after each segment is completed. The feedback interval is 100 milliseconds. The UAV continuously reports the progress to the SkyEdge-5G airborne edge collaboration platform throughout the transmission process. The platform monitors the transmission status in real time, such as the percentage of data already transmitted relative to the total data volume. If the transmission is interrupted, a reconnection mechanism is triggered. The number of reconnections does not exceed 3 times to ensure that the data is transmitted completely to the target edge node.

[0088] In S6, the target edge node completes real-time computation and processing of the task data uploaded by the UAV. The results are then fed back to the corresponding UAV through the SkyEdge-5G airborne edge collaboration platform. At the same time, the edge node resource load parameters and the UAV's remaining power parameters are updated, and the updated real-time parameters are stored in the database to provide the latest data for the next round of S1 parameter collection, realizing full-process cyclic scheduling.

[0089] Specifically, during the task processing phase, after the target edge node receives the UAV task data, it starts a dedicated processing module and calls the corresponding processing program according to the task type. For example, the environmental monitoring task calls the image recognition algorithm to process the collected image data, and the emergency material positioning task calls the coordinate parsing program to process the positioning data. During the processing, the edge node monitors the processor load in real time. If the load exceeds 80%, task priority scheduling is started to prioritize high-priority tasks (such as emergency rescue tasks having a higher priority than regular monitoring tasks) to ensure that the processing is not interrupted.

[0090] During the results feedback phase, the generated results data (such as pollutant concentration values ​​from environmental monitoring and the specific coordinates of emergency supplies) are transmitted to the corresponding drone through the feedback channel of the SkyEdge-5G platform. The feedback data is compressed (such as ZIP compression) to reduce the amount of data transmitted. After the transmission is completed, the drone sends a result reception confirmation to the platform, and the platform records the feedback completion status.

[0091] During the parameter update phase, resource load parameters and remaining power parameters are updated synchronously, and the updated data overwrites the old data in the database in real time to support the next round of scheduling.

[0092] When updating the resource load parameters of an edge node, first determine the set of drone tasks currently assigned to that node, sum the computational requirements of all tasks in the set, and then set the resource recovery coefficient. The value of the resource recovery coefficient is adjusted according to the hardware performance of the edge node, and is usually set between 0.001 and 0.005. During the update process, the task processing time needs to be monitored in real time, which is usually between 100 and 500 milliseconds, to ensure that the product of the recovery coefficient and the processing time is calculated accurately.

[0093] Edge nodes update their own resource load parameters using a dynamic update model as follows:

[0094] in, To update the remaining computing resources of edge node j, To update the remaining computing resources of the front edge node j, For the set of drones currently assigned to edge node j, For the task computation requirements of drone i, t represents the resource recovery coefficient for edge nodes, and t represents the task processing time.

[0095] Edge nodes update their own resource load parameters. For example, after consuming 0.3 GHz of computing resources to process a task, the remaining computing resources are updated from 1.5 GHz to 1.2 GHz. After consuming 50 gigabytes of storage capacity, the remaining storage capacity is updated from 300 gigabytes to 250 gigabytes.

[0096] When updating the remaining battery power of a drone, it is necessary to first obtain the energy consumption of the drone transmitting mission data and set the idle energy consumption rate of the drone. The idle energy consumption rate is usually between 0.1 and 0.3. The transmission energy consumption needs to be calculated based on the amount of mission data and the power of the drone's communication module. Usually, the energy consumption is between 50 and 100 for each certain amount of data transmitted.

[0097] The drone updates its remaining battery power parameters using a dynamic update model as follows:

[0098] in, The remaining battery level of drone i after the update. The remaining battery level of drone i before the update. The energy consumption for task transmission from UAV i to edge node j. Let t represent the idle energy consumption rate of drone i, and t represent the task processing time. The drone updates its remaining power parameters. For example, after consuming 80 mAh for data transmission and 30 mAh for waiting for processing, the remaining power is updated from 3500 mAh to 3390 mAh. The updated parameters are synchronized to the parameter database in real time through the SkyEdge-5G platform, overwriting the original data and providing the latest parameters for step S1 of the next task, forming a closed-loop process for task processing.

[0099] This invention effectively addresses the shortcomings of existing technologies in task allocation, which lack dynamic adaptability. Instead of relying on a single-dimensional parameter, it comprehensively considers multiple factors, including the drone's task type, remaining battery power, and the resource load and communication bandwidth of edge nodes, to calculate the corresponding enticement and penalty coefficients. This multi-dimensional calculation method can adapt to dynamic changes during task processing in real time. When the resource load of edge nodes fluctuates or the drone's remaining battery power changes, the task allocation strategy can be adjusted promptly, avoiding edge node overload due to resource misjudgment and preventing drones from interrupting tasks due to insufficient battery power. This significantly improves the rationality and flexibility of task allocation.

[0100] The energy consumption and latency Pareto optimization offloading model overcomes the shortcomings of existing technologies in co-optimizing energy consumption and latency. The model uses coefficients derived from a dynamic temptation-penalty mechanism as constraints to construct a multi-objective optimization function, rather than being limited to single-dimensional performance optimization. After determining the candidate edge node set, resource conflict detection is used to verify whether the computing resources and storage capacity of the edge nodes meet the task requirements, filtering out target edge nodes without resource conflicts. This process avoids energy waste caused by resource overload while ensuring that task processing latency is controlled within a reasonable range, achieving a balance between energy consumption control and latency reduction, and meeting the core requirements of real-time processing of UAV swarm tasks.

[0101] The SkyEdge-5G airborne edge collaboration platform overcomes the shortcomings of traditional ground-based central node scheduling, such as communication interference and long latency. The platform establishes a stable communication link between the UAV swarm and edge nodes, playing a crucial role in the entire task processing workflow: from initially acquiring various parameters from the UAVs and edge nodes, to sending target edge node allocation commands to the UAVs, and then to the edge nodes feeding back the processing results to the UAVs, all are efficiently completed through this platform. The stable communication link reduces the impact of external interference on data transmission, shortens the transmission time of commands and data, and, combined with the proximity processing advantages of edge computing, further reduces overall task processing latency, ensuring the stability and efficiency of real-time processing of UAV swarm collaborative tasks.

[0102] like Figure 2 As shown, it also includes a real-time processing system for drone swarm collaborative tasks based on edge computing, applied to the real-time processing method for drone swarm collaborative tasks based on edge computing, including: SkyEdge-5G communication module, parameter acquisition module, task allocation module, optimization and unloading module, resource detection module, and result feedback module.

[0103] The SkyEdge-5G communication module is used to establish communication links between the UAV cluster and edge nodes, transmitting task data and control commands. The parameter acquisition module, connected to the SkyEdge-5G communication module, is used to acquire UAV task parameters, UAV status parameters, and edge node resource parameters. The task allocation module, connected to the parameter acquisition module, uses a dynamic enticement-penalty mechanism task allocation algorithm to calculate the enticement and penalty coefficients. The optimization and unloading module, connected to the task allocation module, uses an energy consumption and time delay Pareto optimization unloading model to determine the candidate edge node set. The resource detection module, connected to the optimization and unloading module, is used to detect edge node resource conflicts and filter target edge nodes. The result feedback module, connected to the resource detection module and the SkyEdge-5G communication module, is used to receive the task processing results from the edge nodes and feed them back to the UAVs, while simultaneously updating the resource and status parameters in the parameter acquisition module.

[0104] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A real-time processing method for collaborative tasks of unmanned aerial vehicle (UAV) swarms based on edge computing, characterized in that, include: S1, through the SkyEdge-5G airborne edge collaboration platform, builds a communication link between the drone cluster and edge nodes, and obtains the task type, task data volume, remaining power, current location information of each drone in the drone cluster, as well as the computing resource load, storage capacity, and communication bandwidth parameters of the edge nodes; S2, based on the parameters of the drones and edge nodes collected in S1, uses a dynamic enticement-penalty mechanism task allocation algorithm to solve the task enticement coefficient and task delay penalty coefficient for each drone corresponding to different edge nodes. S3 adopts the energy consumption and time delay Pareto optimization unloading model, with the task temptation coefficient and task delay penalty coefficient as constraints, to establish a multi-objective optimization function for UAV task unloading and determine the candidate edge node set for each UAV task unloading. S4. For the candidate edge node set, resource conflict detection is performed from two dimensions: computing resources and storage capacity. Based on the resource utilization threshold, it is determined whether the node is overloaded and the target edge node without resource conflict is selected. S5 sends target edge node allocation instructions to each UAV through the SkyEdge-5G airborne edge collaboration platform, and controls the UAV to transmit mission data to the target edge node. S6, the target edge node completes real-time computation and processing of the task data uploaded by the drone. After processing, the results are fed back to the corresponding drone through the SkyEdge-5G airborne edge collaboration platform, while updating the edge node resource load parameters and the drone's remaining power parameters.

2. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 1, characterized in that: In the dynamic enticement-penalty mechanism task allocation algorithm, the calculation model for the task enticement coefficient of edge node j corresponding to drone i is as follows: in, Let be the attraction coefficient of edge node j corresponding to drone i. These are weighting coefficients for computing power and bandwidth. , , The remaining computing resources for edge node j, For the task computation requirements of drone i, Let j be the communication bandwidth of edge node j. The amount of task data for drone i; The calculation model for the task latency penalty coefficient of edge node j corresponding to drone i is as follows: in, Let be the latency penalty coefficient for edge node j corresponding to drone i. For latency and power consumption weighting coefficients, , The task transmission latency from drone i to edge node j, The maximum allowable transmission delay of the system. The amount of power consumed by drone i. This is the initial battery level of drone i.

3. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 1, characterized in that, The multi-objective optimization function of the energy consumption time delay Pareto optimization unloading model is: in, For binary 0-1 decision variables, The total energy consumption of the drone swarm. Total task processing time; Where n is the number of drones and m is the number of edge nodes. The energy consumption for task transmission from UAV i to edge node j. The computational energy consumption for edge node j to process drone i's task. For the task transmission latency from UAV i to edge node j The computation time for edge node j to process drone i task.

4. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 3, characterized in that: In S3, a multi-objective optimization function is established by combining physical constraints, including edge node reliability constraints and 5G communication distance constraints. Edge node reliability constraints include: in, Let be the reliability coefficient of edge node j; The minimum reliability threshold set for the system. ; The calculation formula is: in, For edge node j, specify the number of historical task failures. For edge node j, this represents the total number of historical task processing times. 5G communication distance constraints include: in, Let be the straight-line distance between drone i and edge node j. Maximum communication distance supported by the SkyEdge-5G airborne edge collaboration platform.

5. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 4, characterized in that, S3 includes the following steps: S31, the task enticement coefficient of each drone's corresponding edge node With task delay penalty coefficient Standardize each linear mapping to interval; S32, configured with a pair of balanced energy consumption weighting factors With delay weighting coefficient , To construct a comprehensive evaluation index for energy consumption delay; S33. Sort the edge nodes according to the comprehensive evaluation index, and select the top k edge nodes, where k is the preset number of candidate nodes. m is the total number of edge nodes, forming the candidate edge node set for the corresponding UAV; S34, re-verify the available communication bandwidth of the edge nodes in the candidate edge node set, remove nodes whose communication bandwidth cannot support the data transmission requirements of the UAV mission, and update the candidate edge node set.

6. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 1, characterized in that, In step S4, determining whether a node is experiencing resource overload based on a resource utilization threshold includes: Calculate the computational resource utilization of edge node j: in, The computational resource utilization rate of edge node j. For the set of drone tasks that have been assigned to edge node j, For set The computational requirements of the UAV k mission The remaining computing resources for edge node j; when At that time, it was determined that there was no conflict of computing resources. Calculate the maximum resource utilization threshold for edge nodes. ; Calculate storage resource utilization: in, The storage resource utilization rate of edge node j For set The amount of mission data for UAV k in China The remaining storage capacity of edge node j; when At that time, it was determined that there was no storage resource conflict. The threshold for maximum utilization of storage resources on edge nodes. .

7. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 6, characterized in that, S4 includes the following sub-steps: S41, the real-time remaining computing resources and storage capacity of each edge node in the candidate edge node set, as well as the task computing requirements and task data volume of the drone to be assigned; S42, For each candidate edge node, sum up the total computing requirements of all UAV tasks already allocated to that node, compare it with the node’s remaining computing resources, and determine whether it exceeds the node’s computing resource carrying capacity. S43. For each candidate edge node, sum up the total amount of all UAV mission data already allocated to that node, compare it with the node’s remaining storage capacity, and determine whether it exceeds the node’s storage resource carrying capacity. S44. After resource conflict detection is completed, if the computing resources and storage resources of a candidate edge node simultaneously meet the carrying capacity threshold requirements, it is retained as a target edge node without resource conflicts. If they do not meet the requirements simultaneously, the process returns to S3 to readjust the set of candidate edge nodes.

8. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 1, characterized in that, S5 includes the following steps: S51 generates allocation instructions containing UAV identifiers, target edge node identifiers, and data transmission protocols through the control module of the SkyEdge-5G airborne edge collaboration platform; S52 broadcasts the allocation command to the corresponding drone via the 5G communication link. After receiving the command, the drone parses the IP address and port information of the target edge node. S53: The UAV establishes a point-to-point data transmission channel with the target edge node based on the resolved IP address and port information, and dynamically configures two transmission parameters: transmission rate and data fragment size. S54: The UAV transmits mission data in segments to the target edge node according to the set parameters, and provides real-time feedback on the transmission progress to the SkyEdge-5G airborne edge collaboration platform during the transmission process.

9. The real-time processing method for UAV swarm collaborative tasks based on edge computing according to claim 1, characterized in that: In step S6, the edge nodes update their own resource load parameters using a dynamic update model as follows: in, To update the remaining computing resources of edge node j, To update the remaining computing resources of the front edge node j, For the set of drones currently assigned to edge node j, For the task computation requirements of drone i, The edge node resource recovery coefficient is t, and the task processing time is t. The drone updates its remaining battery power parameters using a dynamic update model as follows: in, The remaining battery level of drone i after the update. The remaining battery level of drone i before the update. The energy consumption for task transmission from UAV i to edge node j. Let t be the idle energy consumption rate of drone i, and t be the task processing time.

10. A real-time processing system for drone swarm collaborative tasks based on edge computing, applied in any one of the real-time processing methods for drone swarm collaborative tasks based on edge computing as described in claims 1-9, characterized in that, include: SkyEdge-5G communication module, parameter acquisition module, task allocation module, optimization and unloading module, resource detection module, and result feedback module; The SkyEdge-5G communication module is used to build a communication link between the drone swarm and edge nodes, transmitting mission data and control commands; The parameter acquisition module is connected to the SkyEdge-5G communication module to acquire UAV mission parameters, UAV status parameters, and edge node resource parameters. The task allocation module is connected to the parameter acquisition module, and uses a dynamic temptation-penalty mechanism task allocation algorithm to calculate the temptation coefficient and the penalty coefficient. The connection between the unloading module and the task allocation module is optimized, and the energy consumption and time delay Pareto optimization unloading model is used to determine the candidate edge node set. The resource detection module is connected to the optimization and unloading module to detect resource conflicts at edge nodes and filter target edge nodes; The results feedback module connects the resource detection module and the SkyEdge-5G communication module. It is used to receive the task processing results of the edge node and feed them back to the UAV, while updating the resource and status parameters in the parameter acquisition module.

Images