Unmanned aerial vehicle cluster distributed task allocation method and system under communication limitation

Abstract

The invention belongs to the technical field of unmanned aerial vehicle cluster cooperative control, and discloses an unmanned aerial vehicle cluster distributed task allocation method and system under communication limitation, and the method comprises the steps: constructing a communication prediction model; performing priority calculation and importance evaluation on the cache tasks of the unmanned aerial vehicle; based on the cache tasks after priority calculation and importance evaluation, a distributed task cache space is constructed, the unmanned aerial vehicle judges a future communication state by using a communication prediction model, multiple rounds of negotiation comparison are performed on each unmanned aerial vehicle by using a distributed auction mechanism, and if the future communication state is limited, the distributed auction mechanism determines that the communication state is limited. The unmanned aerial vehicle performs autonomous execution based on the cache task after priority calculation and importance evaluation, and updates the execution state of the unmanned aerial vehicle; and performing completion degree judgment on the task area, and if all local tasks in the task area are completed, generating a task report. According to the method, the resource utilization efficiency can be improved, and environment changes and newly discovered clues can be quickly responded based on a dynamic task redistribution mechanism of real-time information.

Description

Technical Field

[0001] This invention relates to the field of collaborative control technology for unmanned aerial vehicle (UAV) swarms, and in particular to a method and system for distributed task allocation in UAV swarms under communication constraints. Background Technology

[0002] With the development of drone technology, drones are being used in a wider range of fields. For example, drones can be used in complex terrain environments such as mountain search and rescue, forest fire prevention, and border patrol to replace human resources, which can improve work efficiency and ensure personal safety.

[0003] However, existing drone swarm systems generally employ a centralized architecture, relying on stable communication links for task coordination. In complex terrain environments such as mountainous areas, terrain obstruction leads to frequent communication interruptions, causing the system to malfunction. Traditional methods pre-allocate search areas before the mission begins, failing to dynamically adjust task allocation based on real-time clues or environmental changes, resulting in inefficient resource allocation. When communication is limited, the lack of effective information-sharing mechanisms among drones easily leads to multiple drones repeatedly searching the same area, wasting time and energy.

[0004] Therefore, how to provide a method and system for distributed task allocation of UAV swarms under communication constraints is an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a method for distributed task allocation in a drone swarm under communication constraints, in order to solve the problems mentioned above in the prior art.

[0006] According to a first aspect of the present invention, a method for distributed task allocation in unmanned aerial vehicle (UAV) swarms under communication constraints is provided.

[0007] In one embodiment, a method for distributed task allocation in a communication-constrained UAV swarm includes: Obtain the pre-defined task area and construct a communication prediction model using a pre-configured digital elevation model and long short-term memory network model; Based on a pre-defined task area, the task area is divided into multiple local tasks, and the local tasks are used as cached tasks for the drone. Combining the pre-acquired drone performance parameters and the drone's historical execution results, the priority of the drone's cached tasks is calculated and the importance is evaluated. A distributed task cache space is built based on cached tasks, a trust score is assigned to cached tasks, and a dynamic eviction mechanism is used to adjust cached tasks based on the score results. The drone uses a communication prediction model to determine the future communication status. If the future communication status is normal, based on the adjusted cached task, a distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among various drones to determine the drone that will execute the adjusted cached task and update the drone's execution status. If future communication is restricted, the drone will autonomously execute cached tasks based on priority calculations and importance assessments, and update the drone's execution status. The task area is assessed for completion. If all local tasks within the task area are completed, a task report is generated; otherwise, the drone is re-executed to assess the future communication status.

[0008] In one embodiment, obtaining a predefined task area and constructing a communication prediction model using a pre-configured digital elevation model and a long short-term memory network model includes: Topographic data was extracted from the digital elevation model and a mathematical model was established using wireless propagation theory to characterize the impact of terrain on communication. The pre-defined task area is divided into grids. For any two grids, the line-of-sight obstruction is detected by the ray tracing algorithm, and a position-communication quality mapping table is established. The mathematical model and the current location data of the UAV are used as input features of the long short-term memory network model, and a communication prediction model is constructed by combining the UAV's motion trajectory and historical records.

[0009] In one embodiment, extracting terrain data from a digital elevation model and using wireless propagation theory to establish a mathematical model to characterize the impact of terrain on communication includes: Extract elevation, slope, and aspect data from the terrain data; Calculate the initial attenuation value of the reference signal according to the free space path loss formula; Using the knife-edge diffraction model in wireless propagation theory, the additional attenuation value of the mountain to the reference signal is calculated.

[0010] In one embodiment, a distributed task cache space is constructed based on the cached tasks after priority calculation and importance assessment, and trust scores are assigned to the cached tasks. A dynamic eviction mechanism is then used to adjust the cached tasks, including: The distributed task cache space includes the main cache task, the backup cache task, and the neighbor cache task; Each cached task is assigned a trust score based on the reliability of the source of the cached task and the historical execution results. The historical execution results are the success rate and quality assessment records of each UAV in the past when performing the same type of task. Based on the storage space limitations of drones and the importance of cached tasks, a dynamic eviction mechanism is used to adaptively evict cached tasks.

[0011] In one embodiment, the distributed task cache space includes a primary cache task, a backup cache task, and neighbor cache tasks, including: The primary cache task stores currently executing tasks and high-priority tasks to be executed, and the primary cache task has a high trust score. The backup cache task stores alternative tasks and medium-priority tasks, and the backup cache task has a medium trust score. The neighbor cache task stores the area search status and neighbor drone task information, and the neighbor cache task has a low trust score. A hierarchical compression strategy is implemented for cache tasks. The main cache task uses a lossless compression algorithm to retain complete task parameters, the backup cache task uses a lossy compression algorithm to retain key field information, and the neighboring cache task only retains task summary information. When the drone detects that the cache utilization rate exceeds the preset merging threshold, it will combine spatially adjacent and temporally similar tasks into task clusters and store them. When the drone detects that the storage utilization rate is less than 50%, a tiered compression strategy is used for cached tasks. When the storage utilization rate is between 50% and 80%, all cached tasks are subjected to medium compression and selective merging. When the storage utilization rate is greater than 80%, all cached tasks are subjected to high compression and deep merging, while retaining critical task information.

[0012] In one embodiment, the primary cache task stores currently executing tasks and high-priority tasks awaiting execution, and the primary cache task has a high trust score. The backup cache task stores alternative tasks and medium-priority tasks, and the backup cache task has a medium trust score. The neighbor cache task stores area search status and neighbor drone task information, and the neighbor cache task has a low trust score, including: When a drone receives task information from a neighboring drone through the neighbor cache task, it performs a trust comparison. If the neighbor cache task is higher than the current primary cache task, the neighbor cache task is promoted to the primary cache task for execution. The current primary cache task is downgraded to a backup cache task, and the original backup cache task is shared with neighboring cache tasks.

[0013] In one embodiment, adaptively eviction of cached tasks using a dynamic eviction mechanism, based on drone storage space limitations and the importance of the cached tasks, includes: Evict cache tasks in the neighboring cache tasks that have been in existence for more than 60 minutes and have not been assigned. The cached tasks are calculated using a comprehensive value judgment formula. If the number of cached tasks is less than the eviction threshold, they are evicted. When the drone detects that the cache usage rate exceeds the preset threshold, the cached tasks are sorted according to their overall value, and the cached tasks with the lowest overall value are eliminated. When multiple drones detect that the overlap rate between the task areas corresponding to cached tasks exceeds a preset threshold, the cached tasks are compared in terms of comprehensive value, and only the cached task with the highest comprehensive value is retained. At the same time, if the number of execution failures of a cached task reaches a preset threshold, the cached task is eliminated.

[0014] In one embodiment, the UAV uses a communication prediction model to determine the future communication status. If the future communication status is normal, based on the adjusted cached task, a distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among various UAVs to determine the UAV that will execute the adjusted cached task, and the execution status is updated, including: When the drone detects the adjusted cached task, it calculates the bid value based on its own parameters. A distributed experience pool is pre-configured for each UAV, and the UAV's Q-value table is constructed using this experience pool. When drones establish communication with each other, the drones determine whether to use the bid value to participate in the distributed auction negotiation based on the Q-value table. If multiple drones select the same cache task and / or their search areas overlap, a spatiotemporal conflict detection method and a priority comparison method are used to process each drone.

[0015] In one embodiment, when communication is established between drones, the drones determine whether to use their bid value to participate in the distributed auction negotiation based on the Q-value table, including: When drones compete for bids, they exchange their Q-value table updates, and the receiving drone uses a weighted fusion strategy to optimize its bidding decision. Introducing a load balancing factor during the bidding process encourages cached tasks to move to areas with lower loads when the load on nearby drones is too high. Set an initial negotiation round. Calculate the load variance after each negotiation round. If the load variance is greater than a preset threshold and the negotiation round is less than the maximum negotiation round, proceed to the next negotiation round. If the load variance decrease rate is less than the decrease value for two consecutive rounds, terminate the negotiation.

[0016] According to a second aspect of the present invention, a distributed task allocation system for unmanned aerial vehicle (UAV) swarms under communication constraints is provided.

[0017] In one embodiment, a distributed task allocation system for a drone swarm under communication constraints includes: The communication prediction module is used to obtain a pre-defined task area and construct a communication prediction model using a pre-configured digital elevation model and a long short-term memory network model. The calculation and evaluation module is used to divide the task area into multiple local tasks and use the local tasks as cached tasks of the UAV. Combining the pre-acquired UAV performance parameters and the UAV's historical execution results, the module performs priority calculation and importance evaluation of the UAV's cached tasks. The cache space module is used to build a distributed task cache space based on the priority calculation and importance assessment of the cache tasks, assign trust scores to the cache tasks, and adjust the cache tasks according to the score results using a dynamic eviction mechanism. The distributed auction module is used by UAVs to determine the future communication status using a communication prediction model. If the future communication status is normal, based on the adjusted cached task, the distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among various UAVs to determine the UAV that will execute the adjusted cached task and update the UAV execution status.

[0018] The autonomous execution module is used to enable the drone to autonomously execute cached tasks based on priority calculation and importance assessment if future communication is restricted, and to update the drone's execution status. The progress judgment module is used to judge the completion status of the task area. If all local tasks in the task area are completed, a task report is generated; otherwise, the drone is re-executed to judge the future communication status.

[0019] According to a third aspect of the present invention, a computer device is provided for a method of distributed task allocation for unmanned aerial vehicle (UAV) swarms under communication constraints.

[0020] In some embodiments, the computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0021] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: Through prediction and caching mechanisms, the drones can maintain over 90% mission execution capability even in the event of communication interruption. Intelligent task allocation and conflict detection mechanisms reduce the duplicate search rate from 25% in traditional methods to 8%, improving resource utilization efficiency. A dynamic task reallocation mechanism based on real-time information improves search efficiency by 35% compared to static allocation methods, enabling rapid response to environmental changes and newly discovered clues. In unstable communication environments, the task completion rate increases from less than 60% in traditional methods to over 85%, enhancing the robustness of this approach. Through a hierarchical architecture and regional grouping strategy, it supports cluster sizes of 3-100 drones, ensuring efficient coordination for large-scale clusters.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0024] Figure 1 This is a flowchart illustrating a distributed task allocation method for a drone swarm under communication constraints, according to an exemplary embodiment. Figure 2 This is a schematic diagram illustrating the principle of a distributed task allocation system for unmanned aerial vehicle (UAV) swarms under communication constraints, according to an exemplary embodiment. Figure 3 This is a schematic diagram of the structure of a computer device according to an exemplary embodiment; Figure 4 This is a flowchart illustrating a distributed auction process in a distributed task allocation method for UAV swarms under communication constraints, according to an exemplary embodiment. Figure 5 This is a schematic diagram illustrating the flow of cached tasks in a distributed task allocation method for UAV swarms under communication constraints, according to an exemplary embodiment. Figure 6 This is a structural diagram of the task cache space in a distributed task allocation method for UAV swarms under communication constraints, according to an exemplary embodiment. Figure 7 This is an application flowchart illustrating a distributed task allocation method for UAV swarms under communication constraints, according to an exemplary embodiment. Detailed Implementation

[0025] The following description and accompanying drawings fully illustrate specific embodiments described herein to enable those skilled in the art to practice them. Some portions and features of certain embodiments may be included in or replace portions and features of other embodiments. The scope of the embodiments herein includes the entire scope of the claims and all available equivalents thereof. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments; similar or identical parts between embodiments can be referred to interchangeably.

[0026] The modules in the apparatus or system of this application can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0027] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0028] Figure 1 and Figure 7 An embodiment of the distributed task allocation method for UAV swarms under communication constraints according to the present invention is shown.

[0029] In this optional embodiment, a method for distributed task allocation in a communication-constrained UAV swarm includes: Step S101: Obtain the pre-defined task area and construct a communication prediction model using the pre-configured digital elevation model and long short-term memory network model; In this optional embodiment, obtaining a pre-defined task area and constructing a communication prediction model using a pre-configured digital elevation model and a long short-term memory network model includes: Topographic data was extracted from the digital elevation model and a mathematical model was established using wireless propagation theory to characterize the impact of terrain on communication. In this optional embodiment, extracting terrain data from the digital elevation model and establishing a mathematical model using wireless propagation theory to characterize the impact of terrain on communication includes: Extract elevation, slope, and aspect data from the terrain data; Calculate the initial attenuation value of the reference signal according to the free space path loss formula; Using the knife-edge diffraction model in wireless propagation theory, the additional attenuation value of the mountain to the reference signal is calculated.

[0030] Specifically, terrain data is extracted from the Digital Elevation Model (DEM), and the initial attenuation value FSPL of the reference signal is calculated using the free space path loss formula. The calculation expression is as follows: ; In the formula, d Distance (km) f Frequency (MHz).

[0031] Next, the additional signal attenuation caused by the mountain is calculated using a knife-edge diffraction model, which consists of the following parts: The geometric parameter calculation module is used to determine the spatial geometric relationship between the launch point, the receiver point, and obstacles; The Fresnel zone determination module is used to calculate the radius of the first Fresnel zone and determine the degree of obstacle occlusion; The diffraction loss calculation module calculates the additional attenuation of the signal as it passes through the edge of an obstacle based on the knife-edge diffraction theory. The specific process is as follows: First, based on the coordinates of the launch point and receiving point coordinates and the coordinates of the obstacle vertices Calculate the radius of the first Fresnel zone The calculation expression is: ; In the formula, λ is the signal wavelength. The distance from the launch point to the obstacle. This is the distance from the receiving point to the obstacle.

[0032] in, The calculation expression is: ; The expression is: ; Then, calculate the height difference of the obstacle penetrating the first Fresnel zone: ; In the formula, This represents the height of the line-of-sight path at the obstacle.

[0033] Next, calculate the Fresnel-Kirchhoff diffraction parameter v: ; In the formula, Let λ be the height difference and λ be the signal wavelength. The distance from the launch point to the obstacle. This is the distance from the receiving point to the obstacle.

[0034] Finally, the additional attenuation caused by diffraction is calculated based on parameter v. : ; In the formula, v is the Fresnel-Kirchhoff diffraction parameter.

[0035] The pre-defined task area is divided into grids. For any two grids, the line-of-sight obstruction is detected by the ray tracing algorithm, and a position-communication quality mapping table is established. Specifically, the task area is divided into a 50m×50m grid. For any two grid points, the intersection of the ray and the terrain surface is detected. If there is occlusion, the diffraction loss is calculated, and a location-communication quality mapping table is established.

[0036] The mathematical model and the current location data of the UAV are used as input features of the long short-term memory network model, and a communication prediction model is constructed by combining the UAV's motion trajectory and historical records.

[0037] Specifically, historical flight data of the UAV, including position, speed, and acceleration, is collected. An input sequence with a time window of 10 time steps is constructed and fed into a Long Short-Term Memory (LSTM) network model for feature extraction and temporal modeling. The LSM model includes an input layer, a hidden layer, and an output layer. The input layer has 128 dimensions, the hidden layer has 256 dimensions, and the output layer has 3 dimensions. Based on the output of the LSM model, the UAV's position information at future moments is predicted. The predicted position information is then used to query a position-communication quality mapping table to obtain the corresponding communication quality index. The communication quality index is compared with a preset communication threshold. When the communication quality is lower than the preset threshold, such as -90 (dBm), the communication status at future moments is determined to be a communication interruption state.

[0038] Step S102: Divide the task area into multiple local tasks and use the local tasks as cached tasks for the UAV. Combine the pre-acquired UAV performance parameters and the UAV's historical execution results to calculate the priority and assess the importance of the UAV's cached tasks. Specifically, the importance of caching tasks is calculated using a Bayesian probabilistic framework, and its expression is as follows: ; In the formula, Let be the prior probability of position (x, y). Let (x, y) be the posterior probability of the target at position (x, y). To observe sensor data at position (x,y) The likelihood probability, It is the product of all sensor observations.

[0039] in, Using a logistic regression model combined with terrain features such as slope data, vegetation data, and water source distance data, the task area was ultimately divided into three levels: high (I>0.7), medium (0.3≤I≤0.7), and low (I<0.3).

[0040] Next, the task priority is calculated, and the calculation expression is as follows: ; In the formula, Priority Score the task priority. to This represents the weighting coefficient of each factor, where, I(R) is the importance index of region R. To ensure the suitability of the mission for the drone's capabilities, Indicates the urgency of the task. The cost of performing the task.

[0041] Next, the NSGA-II algorithm is used for multi-objective optimization, with the optimization objectives being to maximize search efficiency and minimize resource consumption.

[0042] Specifically, maximizing search efficiency is as follows: ; Minimize resource consumption as follows: ; In the formula, The objective function is the search efficiency. Let the objective function be resource consumption. Let I(R) represent the covered search area, and let I(R) be the importance index of region R. Total search time Energy already consumed For flight time, Cost per unit of time This is the summation over all tasks.

[0043] The specific process is as follows: First, initialize the population for the genetic algorithm, setting the population size to 100. The initial population is constructed using a random generation strategy, randomly assigning a task to each drone and constraining that each task is assigned to at least one drone. After completing the above assignment, for the drones with remaining unassigned tasks, randomly select tasks from the task set for assignment. In this way, each solution represents a search path, ensuring that all tasks can be assigned. Perform non-dominated sorting, and then generate offspring through tournament selection, simulated binary crossover, and polynomial mutation. After 200 iterations, select a compromise solution.

[0044] The tournament selection process involves randomly selecting k individuals from the current population (k=3 in this method), comparing the fitness values ​​of the selected individuals, and choosing the individual with the highest fitness as the parent. This process is repeated until a sufficient number of parent individuals are selected.

[0045] Next, simulated binary crossover is performed on any two parent individuals, specifically as follows: For each gene position i, generate a random number. Then calculate the distribution index: When u≤0.5 ; when u >0.5 ; In the formula, β is the distribution index. This is the cross-distribution index, and its value ranges from 20.

[0046] Finally, based on the distribution index, two offspring individuals are generated. Its expression is: ; In the formula, β is the distribution index. p For the parent generation, For the first parent individual in the first i Gene values ​​at each gene location For the second parent individual in the first i Gene values ​​at each gene location.

[0047] For each gene x of individual x i Perform polynomial mutation, specifically, with mutation probability p m Perform a mutation operation to generate random numbers. Then calculate the variation parameters. : When r < 0.5 ; When r ≥ 0.5 ; In the formula, This is the variation distribution index, with a value of 20.

[0048] Finally, the mutated gene value is calculated, and its expression is: ; In the formula, Gene value, The upper limit of a gene. The lower bound of a gene. These are the variation parameters.

[0049] Step S103: Based on the priority calculation and importance assessment of the cached tasks, construct a distributed task cache space, assign trust scores to the cached tasks, and adjust the cached tasks using a dynamic eviction mechanism according to the score results. In this optional embodiment, a distributed task cache space is constructed based on the cached tasks after priority calculation and importance assessment, and trust scores are assigned to the cached tasks. A dynamic eviction mechanism is then used to adjust the cached tasks, including: The distributed task cache space includes the main cache task, the backup cache task, and the neighbor cache task; In this optional embodiment, the distributed task cache space includes a primary cache task, a backup cache task, and neighbor cache tasks, including: The primary cache task stores currently executing tasks and high-priority tasks to be executed, and the primary cache task has a high trust score. The backup cache task stores alternative tasks and medium-priority tasks, and the backup cache task has a medium trust score. The neighbor cache task stores the area search status and neighbor drone task information, and the neighbor cache task has a low trust score. In this optional embodiment, the primary cache task stores currently executing tasks and high-priority tasks to be executed, and the primary cache task has a high trust score. The backup cache task stores alternative tasks and medium-priority tasks, and the backup cache task has a medium trust score. The neighbor cache task stores the area search status and neighbor drone task information, and the neighbor cache task has a low trust score, including: When a drone receives task information from a neighboring drone through the neighbor cache task, it performs a trust comparison. If the neighbor cache task is higher than the current primary cache task, the neighbor cache task is promoted to the primary cache task for execution. The current primary cache task is downgraded to a backup cache task, and the original backup cache task is shared with neighboring cache tasks.

[0050] like Figure 5 and Figure 6 As shown, a hierarchical compression strategy is implemented for cache tasks. The main cache task uses a lossless compression algorithm to retain complete task parameters, the backup cache task uses a lossy compression algorithm to retain key field information, and the neighboring cache task only retains task summary information. When the drone detects that the cache utilization rate exceeds the preset merging threshold, it will combine spatially adjacent and temporally similar tasks into task clusters and store them. When the drone detects that the storage utilization rate is less than 50%, a tiered compression strategy is used for cached tasks. When the storage utilization rate is between 50% and 80%, all cached tasks are subjected to medium compression and selective merging. When the storage utilization rate is greater than 80%, all cached tasks are subjected to high compression and deep merging, while retaining critical task information.

[0051] Specifically, key fields include task coordinates, priority, and timestamp, while task summary information includes task ID, type, and approximate location.

[0052] Lossless compression algorithms can include LZ77 variants, Zstandard, and LZ4, while lossy compression algorithms can include differential coding and other algorithms. Task digest information can use spatial rasterization coding and hash digest algorithms.

[0053] In addition, the task cluster includes the cluster center coordinates, the number of tasks, the priority distribution, and the expected completion time window.

[0054] Each cached task is assigned a trust score based on the reliability of the source of the cached task and the historical execution results. The historical execution results are the success rate and quality assessment records of each UAV in the past when performing the same type of task. Specifically, trust rating The calculation is performed using a multi-factor weighted model, and its expression is as follows: ; In the formula, To ensure the reliability of the mission source, To increase the success rate of execution, As a time-sensitive factor, b and z as well as t The weighting coefficient has a value of b =0.4, z =0.35, t =0.25.

[0055] The rating levels are as follows: high trust level ≥ 0.8, medium trust level 0.5-0.8, and low trust level < 0.5.

[0056] Based on the storage space limitations of drones and the importance of cached tasks, a dynamic eviction mechanism is used to adaptively evict cached tasks.

[0057] In one embodiment, adaptively eviction of cached tasks using a dynamic eviction mechanism, based on drone storage space limitations and the importance of the cached tasks, includes: Evict cache tasks in the neighboring cache tasks that have been in existence for more than 60 minutes and have not been assigned. The cached tasks are calculated using a comprehensive value judgment formula. If the number of cached tasks is less than the eviction threshold, they are evicted. When the drone detects that the cache usage rate exceeds the preset threshold, the cached tasks are sorted according to their overall value, and the cached tasks with the lowest overall value are eliminated. When multiple drones detect that the overlap rate between the task areas corresponding to cached tasks exceeds a preset threshold, the cached tasks are compared in terms of comprehensive value, and only the cached task with the highest comprehensive value is retained. At the same time, if the number of execution failures of a cached task reaches a preset threshold, the cached task is eliminated.

[0058] Specifically, the formula for comprehensive value judgment is as follows: ; In the formula, For the overall value of the task, Priority As a task priority, Score the time taken to complete the task. This represents the probability that the task has been completed.

[0059] Step S104: The UAV uses a communication prediction model to determine the future communication status. If the future communication status is normal, based on the adjusted cache task, the distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among the various UAVs to determine the UAV that will execute the adjusted cache task and update the execution status of the UAV.

[0060] In this optional embodiment, the UAV uses a communication prediction model to determine the future communication status. If the future communication status is normal, based on the adjusted cached task, a distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among various UAVs to determine the UAV that will execute the adjusted cached task, and the execution status is updated, including: When the drone detects the adjusted cached task, it calculates the bid value based on its own parameters. A distributed experience pool is pre-configured for each UAV, and the UAV's Q-value table is constructed using this experience pool. When drones establish communication with each other, the drones determine whether to use the bid value to participate in the distributed auction negotiation based on the Q-value table. In this optional embodiment, when communication is established between drones, the drones determine whether to use their bid value to participate in the distributed auction negotiation based on the Q-value table, including: When drones compete for bids, they exchange their Q-value table updates, and the receiving drone uses a weighted fusion strategy to optimize its bidding decision. Introducing a load balancing factor during the bidding process encourages cached tasks to move to areas with lower loads when the load on nearby drones is too high. Set an initial negotiation round. Calculate the load variance after each negotiation round. If the load variance is greater than a preset threshold and the negotiation round is less than the maximum negotiation round, proceed to the next negotiation round. If the load variance decrease rate is less than the decrease value for two consecutive rounds, terminate the negotiation.

[0061] If multiple drones select the same cache task and / or their search areas overlap, a spatiotemporal conflict detection method and a priority comparison method are used to process each drone.

[0062] Specifically, the overlapping area of ​​the drone's execution region is calculated and the overlapping of the drone's execution time window is checked. If the preset regional overlap rate and time window overlap rate are met at the same time, a conflict log is recorded. The difference between the bid values ​​of the two drones is calculated. If the difference is less than a preset threshold, the winning drone is selected based on the timestamp. If the difference is greater than and / or equal to the preset threshold, the drone with higher priority is selected, and the drone with lower priority cancels its bid and the conflict log is recorded.

[0063] Specifically, such as Figure 4As shown, each drone calculates its bid value based on its own capabilities, distance cost, current load, and other factors. : ; In the formula, Utility represents the utility component, and Cost represents the cost component. To reward ability, Priority Score based on priority.

[0064] The cost component includes flight costs. C flight Time cost C time and opportunity cost C opportunity Finally, normalize to the [0.1] interval.

[0065] Next, a Q-value table is constructed based on the distributed experience pool, which records historical task allocation decisions and their execution results, including completion time, energy consumption, and success rate.

[0066] Specifically, the state space and action space of the drone are obtained. The state space includes the drone's remaining energy, current position, mission type, and number of surrounding drones. The action space is bidding or giving up.

[0067] The reward function R used to update the Q-value table is designed as follows: ; In the formula, For task completion time, For energy consumption costs, For successful rewards, α, β, and γ are weighting coefficients.

[0068] In the bidding process among various drones, an ε-greedy strategy is adopted to balance exploration and utilization.

[0069] When drones establish communication, they exchange their respective Q-value table updates. The receiving drone updates its Q-value table using a weighted fusion strategy, expressed as follows: ; In the formula, A table of Q-values ​​for local drones. This is a table of received Q-values ​​for neighboring drones, where λ is a weighting coefficient. The weight λ is dynamically adjusted based on the historical success rate of the information source. Through this mechanism, the overall strategy can continuously learn from individual experiences, gradually optimize bidding decisions, and reduce inefficient competition and resource waste.

[0070] In addition, a load balancing factor is introduced during the bidding process to avoid excessive drone load. Adjust the bid value, load balance factor The expression is: ; In the formula, For the current drone load, This represents the average load for the group.

[0071] Revised bid value for: ; In the formula, It is a neighborhood cooperation penalty factor that increases when a neighboring drone is detected to be overloaded, prompting missions to move to areas with lower loads.

[0072] Furthermore, upon restoration of communication, each UAV exchanges its execution status and discovery information. A negotiation mechanism unifies the global mission status to avoid information inconsistencies. This negotiation mechanism employs an improved contract network protocol: First, a conflict message is sent, and all participating drones enter negotiation mode and initialize negotiation parameters, setting the maximum number of negotiation rounds to [value missing]. r maxr In this method, the number of negotiation rounds is 5. During each round of negotiation, each UAV adjusts its original bid value according to the current negotiation round to reflect the negotiation concession strategy. After each round of negotiation, the bid values ​​of the UAVs participating in the negotiation are compared. The UAV with the lower bid value makes a concession decision, abandons the bidding for the current task, and selects the next highest priority alternative task from its cached tasks for subsequent execution or bidding. When the negotiation process fails to reach a consensus decision within the preset number of rounds, a failure handling mechanism is triggered, and the task is reported to the ground station for centralized adjudication, or the UAVs participating in the negotiation use a random allocation method to determine the UAV to execute the task, so as to ensure the continuous progress of the task.

[0073] Step S105: If future communication is restricted, the drone will autonomously execute the cached tasks based on priority calculation and importance assessment, and update the drone's execution status. Step S105: Determine the completion status of the task area. If all local tasks within the task area are completed, generate a task report; otherwise, re-execute the UAV to determine the future communication status.

[0074] Figure 2 An embodiment of a communication-constrained distributed task system for unmanned aerial vehicle (UAV) swarms is shown.

[0075] In this optional embodiment, a distributed task allocation system for UAV swarms under communication constraints includes: The communication prediction module 201 is used to obtain a pre-defined task area and construct a communication prediction model using a pre-configured digital elevation model and a long short-term memory network model. The calculation and evaluation module 202 is used to divide the task area into multiple local tasks and use the local tasks as cached tasks of the UAV. Combining the pre-acquired UAV performance parameters and the UAV's historical execution results, it performs priority calculation and importance evaluation on the UAV's cached tasks. The cache space module 203 is used to construct a distributed task cache space based on the cache tasks after priority calculation and importance evaluation, assign trust scores to cache tasks, and adjust cache tasks using a dynamic eviction mechanism based on the score results. The distributed auction module 204 is used by the UAV to determine the future communication status using a communication prediction model. If the future communication status is normal, based on the adjusted cached task, the distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among the various UAVs to determine the UAV that will execute the adjusted cached task and update the execution status of the UAV.

[0076] The autonomous execution module 205 is used to enable the UAV to autonomously execute cached tasks based on priority calculation and importance assessment if future communication is restricted, and to update the UAV execution status. The progress judgment module 206 is used to judge the completion status of the task area. If all local tasks in the task area are completed, a task report is generated; otherwise, the drone is re-executed to judge the future communication status.

[0077] In one embodiment, a computer device is provided for a method of distributed task allocation for UAV swarms under communication constraints. This computer device may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores static and dynamic information data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0078] Those skilled in the art will understand that Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0079] In addition, the present invention also provides a method and system computer device for distributed task allocation of UAV swarms under communication constraints, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0080] In addition, the present invention also provides a method and system for distributed task allocation of unmanned aerial vehicle (UAV) swarms under communication constraints. The computer-readable storage medium thereon stores a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0081] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory in a communication-constrained UAV swarm distributed task allocation method and system. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0082] This invention is not limited to the structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this invention is limited only by the appended claims.

Claims

1. A method for distributed task allocation in a UAV swarm under communication constraints, characterized in that, include: Obtain the pre-defined task area and construct a communication prediction model using a pre-configured digital elevation model and long short-term memory network model; The task area is divided into multiple local tasks, and these local tasks are used as cached tasks for the drone. Based on the pre-acquired drone performance parameters and the drone's historical execution results, the priority of the drone's cached tasks is calculated and its importance is evaluated. Based on the cached tasks calculated in terms of priority and evaluated in terms of importance, a distributed task cache space is constructed. The cached tasks are assigned trust scores, and the cached tasks are adjusted using a dynamic eviction mechanism based on the score results. The drone uses a communication prediction model to determine the future communication status. If the future communication status is normal, based on the adjusted cached task, a distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among various drones to determine the drone that will execute the adjusted cached task and update the drone's execution status. If future communication is restricted, the drone will autonomously execute cached tasks based on priority calculations and importance assessments, and update the drone's execution status. The task area is assessed for completion. If all local tasks within the task area are completed, a task report is generated; otherwise, the drone is re-executed to assess the future communication status.

2. The method for distributed task allocation in a UAV swarm under communication constraints according to claim 1, characterized in that, The step of obtaining a pre-defined task area and constructing a communication prediction model using a pre-configured digital elevation model and a long short-term memory network model includes: Topographic data was extracted from the digital elevation model and a mathematical model was established using wireless propagation theory to characterize the impact of terrain on communication. The pre-defined task area is divided into grids. For any two grids, the line-of-sight obstruction is detected by the ray tracing algorithm, and a position-communication quality mapping table is established. The mathematical model and the current location data of the UAV are used as input features of the long short-term memory network model, and a communication prediction model is constructed by combining the UAV's motion trajectory and historical records.

3. The method for distributed task allocation in a UAV swarm under communication constraints according to claim 2, characterized in that, The extraction of terrain data from the digital elevation model and the establishment of a mathematical model using wireless propagation theory to characterize the impact of terrain on communication include: Extract elevation, slope, and aspect data from the terrain data; Calculate the initial attenuation value of the reference signal according to the free space path loss formula; Using the knife-edge diffraction model in wireless propagation theory, the additional attenuation value of the mountain to the reference signal is calculated.

4. The method for distributed task allocation in a UAV swarm under communication constraints according to claim 1, characterized in that, The process of constructing a distributed task cache space based on priority calculation and importance assessment of the cached tasks, assigning trust scores to the cached tasks, and adjusting the cached tasks using a dynamic eviction mechanism includes: The distributed task cache space includes the main cache task, the backup cache task, and the neighbor cache task; Each cached task is assigned a trust score based on the reliability of the source of the cached task and the historical execution results. The historical execution results are the success rate and quality assessment records of each UAV in the past when performing the same type of task. Based on the storage space limitations of drones and the importance of cached tasks, a dynamic eviction mechanism is used to adaptively evict cached tasks.

5. The method for distributed task allocation in a UAV swarm under communication constraints according to claim 4, characterized in that, The distributed task cache space includes a primary cache task, a backup cache task, and neighbor cache tasks, including: The primary cache task stores currently executing tasks and high-priority tasks to be executed, and the primary cache task has a high trust score. The backup cache task stores alternative tasks and medium-priority tasks, and the backup cache task has a medium trust score. The neighbor cache task stores the area search status and neighbor drone task information, and the neighbor cache task has a low trust score. A hierarchical compression strategy is implemented for cache tasks. The main cache task uses a lossless compression algorithm to retain complete task parameters, the backup cache task uses a lossy compression algorithm to retain key field information, and the neighboring cache task only retains task summary information. When the drone detects that the cache utilization rate exceeds the preset merging threshold, it will combine spatially adjacent and temporally similar tasks into task clusters and store them. When the drone detects that the storage utilization rate is less than 50%, a tiered compression strategy is used for cached tasks. When the storage utilization rate is between 50% and 80%, all cached tasks are subjected to medium compression and selective merging. When the storage utilization rate is greater than 80%, all cached tasks are subjected to high compression and deep merging, while retaining critical task information.

6. A method for distributed task allocation in a communication-constrained UAV swarm according to claim 5, characterized in that, The primary cache task stores currently executing tasks and high-priority tasks awaiting execution, and the primary cache task has a high trust score. The backup cache task stores alternative tasks and medium-priority tasks, and the backup cache task has a medium trust score. The neighbor cache task stores area search status and neighbor drone task information, and the neighbor cache task has a low trust score, including: When a drone receives task information from a neighboring drone through the neighbor cache task, it performs a trust comparison. If the neighbor cache task is higher than the current primary cache task, the neighbor cache task is promoted to the primary cache task for execution. The current primary cache task is downgraded to a backup cache task, and the original backup cache task is shared with neighboring cache tasks.

7. A method for distributed task allocation in a UAV swarm under communication constraints as described in claim 4, characterized in that, The adaptive eviction mechanism for cached tasks based on UAV storage space limitations and the importance of cached tasks includes: Evict cache tasks in the neighboring cache tasks that have been in existence for more than 60 minutes and have not been assigned. The cached tasks are calculated using a comprehensive value judgment formula. If the number of cached tasks is less than the eviction threshold, they are evicted. When the drone detects that the cache usage rate exceeds the preset threshold, the cached tasks are sorted according to their overall value, and the cached tasks with the lowest overall value are eliminated. When multiple drones detect that the overlap rate between the task areas corresponding to cached tasks exceeds a preset threshold, the cached tasks are compared in terms of comprehensive value, and only the cached task with the highest comprehensive value is retained. If the number of execution failures of a cached task reaches a preset threshold, the cached task is eliminated.

8. A method for distributed task allocation in a UAV swarm under communication constraints as described in claim 1, characterized in that, The drone uses a communication prediction model to determine the future communication status. If the future communication status is normal, based on the adjusted cached task, a distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among various drones to determine the drone that will execute the adjusted cached task, and the execution status is updated, including: When the drone detects the adjusted cached task, it calculates the bid value based on its own parameters. A distributed experience pool is pre-configured for each UAV, and the UAV's Q-value table is constructed using this experience pool. When drones establish communication with each other, the drones determine whether to use the bid value to participate in the distributed auction negotiation based on the Q-value table. If multiple drones select the same cache task and / or their search areas overlap, a spatiotemporal conflict detection method and a priority comparison method are used to process each drone.

9. A method for distributed task allocation in a UAV swarm under communication constraints as described in claim 8, characterized in that, When communication is established between drones, the process of determining whether to use the bid value to participate in the distributed auction negotiation based on the Q-value table includes: When drones compete for bids, they exchange their Q-value table updates, and the receiving drone uses a weighted fusion strategy to optimize its bidding decision. Introducing a load balancing factor during the bidding process encourages cached tasks to move to areas with lower loads when the load on nearby drones is too high. Set an initial negotiation round. Calculate the load variance after each negotiation round. If the load variance is greater than a preset threshold and the negotiation round is less than the maximum negotiation round, proceed to the next negotiation round. If the load variance decrease rate is less than the decrease value for two consecutive rounds, terminate the negotiation.

10. A distributed task allocation system for unmanned aerial vehicle (UAV) swarms under communication constraints, characterized in that, The system includes: The communication prediction module is used to obtain a pre-defined task area and construct a communication prediction model using a pre-configured digital elevation model and a long short-term memory network model. The calculation and evaluation module is used to divide the task area into multiple local tasks and use the local tasks as cached tasks of the UAV. Combining the pre-acquired UAV performance parameters and the UAV's historical execution results, the module performs priority calculation and importance evaluation of the UAV's cached tasks. The cache space module is used to build a distributed task cache space based on the priority calculation and importance assessment of the cache tasks, assign trust scores to the cache tasks, and adjust the cache tasks according to the score results using a dynamic eviction mechanism. The distributed auction module is used by UAVs to determine the future communication status using a communication prediction model. If the future communication status is normal, based on the adjusted cached task, the distributed auction mechanism is used to conduct multiple rounds of negotiation and comparison among various UAVs to determine the UAV that will execute the adjusted cached task and update the UAV execution status. The autonomous execution module is used to enable the drone to autonomously execute cached tasks based on priority calculation and importance assessment if future communication is restricted, and to update the drone's execution status. The progress judgment module is used to judge the completion status of the task area. If all local tasks in the task area are completed, a task report is generated; otherwise, the drone is re-executed to judge the future communication status.

Images