Agricultural unmanned aerial vehicle nest cluster control method for multi-machine cooperation

Abstract

The application discloses a kind of agricultural unmanned aerial vehicle nest cluster control methods for multi-machine cooperation, it is related to agricultural unmanned aerial vehicle cluster scheduling technical field.The method is monitored when the real-time residual capacity of plant protection unmanned aerial vehicle is lower than task generation threshold, and it is added to the agricultural task list for its construction virtual charging task, and the operation priority of virtual charging task monotonically increases with the decrease of residual capacity;When the highest priority task selected by a certain unmanned aerial vehicle is the virtual charging task associated with itself, and there is still a real agricultural task, the homing motion vector pointing to the target position of virtual charging task is weighted with the operation motion vector pointing to the second highest priority real agricultural task Vector synthesis is generated to generate the flight direction.The application realizes the continuous transition of plant protection unmanned aerial vehicle task execution and charging homing behavior, so that the unmanned aerial vehicle can respond to the task along the way on the way back.

Description

Technical Field

[0001] This invention relates to the field of agricultural drone swarm scheduling technology, and more specifically, to a method for controlling agricultural drone swarms in a multi-drone collaborative manner. Background Technology

[0002] In large-scale farmland plant protection operations, multiple fixed drone nests are typically deployed, with each nest controlling several plant protection drones. When multiple pests and diseases appear in the field, each drone needs to quickly travel to the task site to perform spraying operations and return to the nest to recharge when its battery is low.

[0003] Currently, drone mission scheduling and recharging / returning generally employ two independent logics: the task allocation module assigns a sequence of operational targets to the drone, while the energy management module monitors the battery level and forcibly interrupts the current task and generates a homing path when the battery level falls below a preset threshold. For example, CN117631657A discloses a method and system for controlling multiple drones based on a mobile portable drone nest, where drones execute inspection tasks according to a preset process, but there is a lack of a continuous behavioral connection mechanism between task execution and return. CN119148567A discloses a method and system for remote control of a drone nest, which generates scheduling and charging instructions by constructing a charging scheduling function. This method processes task scheduling and charging scheduling separately, and the homing decision and task execution remain independent processes.

[0004] In the above methods, the drone abruptly switches between "mission execution" and "interrupted return" states, forcibly abandoning the ongoing mission and failing to utilize the completed flight distance. The return path planning does not consider whether there are still pending tasks along the route, resulting in the drone being unable to respond to tasks along the way and incurring additional wasted flight. In multi-drone cluster scenarios, multiple drones may simultaneously trigger return and converge on the same cluster, leading to an imbalance in load among the clusters. Therefore, a method for controlling agricultural drone swarms in a collaborative manner with multiple drones is proposed to address the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for controlling agricultural drone swarms in a multi-drone collaborative manner, in order to solve the problems of disconnect between task execution and charging / homing behavior in the existing methods, the inability of drones to respond to tasks along the way during the return flight, and the uneven load among swarms when multiple drones return simultaneously.

[0006] To achieve the above objectives, this invention provides a method for controlling agricultural drone swarms in a multi-drone collaborative manner. The charging needs of agricultural drones are incorporated into a unified task list with real agricultural tasks, managed with priority, in the form of virtual charging tasks. A continuous transition between task execution and charging / homing behavior is achieved through weighted synthesis of dual motion vectors. The method includes the following steps: S1: Obtain the real-time remaining battery power and list of pending agricultural tasks for each agricultural drone in the agricultural drone cluster associated with multiple distributed drone nests within the farmland operation area. Each agricultural task includes the target location coordinates and the initial operation priority.

[0007] S2, monitor the real-time remaining power of each agricultural drone. When the real-time remaining power of any agricultural drone drops to a preset task generation threshold, construct a virtual charging task for that agricultural drone and add the virtual charging task to the agricultural task list.

[0008] As a result, charging needs are transformed into a task format managed in the same list as real agricultural tasks, allowing subsequent task selection and flight guidance to be carried out within a unified framework.

[0009] The priority of the virtual charging task increases monotonically as the remaining battery power of the agricultural drone decreases. During the period of sufficient battery power, the priority of the virtual charging task is at a low level and does not participate in priority competition; as the battery power gradually decreases, its priority gradually increases and it begins to gain response priority; when the battery power is almost exhausted, its priority reaches the highest level, triggering an emergency homing.

[0010] S3, each agricultural drone queries the job priority of all tasks in the current global task list at preset time intervals, and determines the task with the highest job priority as the current target of the drone.

[0011] When the current target selected by a certain agricultural drone is a virtual charging task associated with itself, and there are other real agricultural tasks in the global task list besides the virtual charging task, the real agricultural task with the highest priority is selected and recorded as the first real agricultural task.

[0012] The actual flight direction of the agricultural drone is obtained by vector synthesis of the first motion vector pointing the agricultural drone to the virtual charging task target position and the second motion vector pointing to the first real agricultural task target position.

[0013] By weighted synthesis of the homing trend represented by the first motion vector and the operational trend represented by the second motion vector, the agricultural drone has a deflection component that is directed toward the actual agricultural task superimposed on its flight direction toward the nest.

[0014] If there are unfinished tasks near the return path, the synthesized flight direction will guide the agricultural drone to approach and complete the task, thus performing tasks along the way on the way back to the homing site, achieving a continuous transition between task execution and charging / homing behavior.

[0015] Furthermore, the job priority of the virtual charging task in step S2 is determined as follows: When the real-time remaining battery power of the agricultural drone is higher than the preset safe battery power, the virtual charging task has the lowest priority value, and the agricultural drone will respond to the real agricultural task first. When the real-time remaining power is between the preset safe power and the preset warning power, the operation priority increases from the lowest priority value to the highest priority value as the remaining power decreases, and the homing tendency of the agricultural drone gradually increases. When the real-time remaining power is lower than the preset warning power, the operation priority is the highest priority value, and the agricultural drone completely switches to returning to the homing point to replenish power.

[0016] In one implementation, the preset safe battery level is 60% of the agricultural drone's full battery level, and the preset warning battery level is 20% of the agricultural drone's full battery level. The agricultural drone is fully operational when its remaining battery level is above 60% of its full battery level, and triggers the highest-priority homing mechanism when its remaining battery level is below 20% of its full battery level. This segmented mapping method ensures that the homing behavior is not triggered by a hard interruption, but rather by a decision-making process that evolves continuously with battery consumption.

[0017] Furthermore, the target location of the virtual charging task in step S2 is set in the following manner: Obtain the range supported by the current remaining power of the agricultural drone, select the drone nest with the smallest overall load from the nests within the range, and use its location coordinates as the target location of the virtual charging task.

[0018] The overall load is determined by the weighted sum of the number of agricultural drones included in the virtual charging task target in the drone nest, the number of agricultural drones currently charging, and the distance between the current location of the agricultural drone and the drone nest.

[0019] The first two items reflect the charging queue status of the drone nest, while the third item reflects the flight cost. The weighted sum of these three items yields a comprehensive load that takes into account both the load on the drone nest and the flight distance. This allows agricultural drones to prioritize nests with empty charging queues and moderate distances when returning to their nests, which helps balance the charging load of each nest.

[0020] Furthermore, as the agricultural drone travels to its associated virtual charging task target location, the overall load of each drone nest is periodically recalculated. When the difference between the overall load of another drone nest and the overall load of the current target drone nest exceeds a preset threshold, the target location of the virtual charging task is updated to the location coordinates of that other drone nest.

[0021] This dynamic adjustment mechanism allows agricultural drones to adjust their homing targets based on real-time load changes in each drone nest during flight, preventing them from continuing to fly to a particular nest due to a sudden increase in load, and further improving nest load balancing in multi-drone collaborative scenarios.

[0022] Furthermore, the task type identifier of the virtual charging task in step S2 is the same as the type identifier of the plant protection spraying task in the agricultural task list. At the data structure level, the virtual charging task is indistinguishable from the real agricultural task. Each plant protection drone selects tasks according to a unified priority mechanism, without needing to distinguish the task source, thus simplifying the task processing logic on the drone side.

[0023] Furthermore, the vector synthesis described in step S3 is composed of weighting coefficients. control, ,and It is positively correlated with the task priority of the first real agricultural task.

[0024] When the priority of the first real agricultural task is higher, the weight of the second motion vector in the synthesis is greater, and the plant protection drone is more likely to deviate in the direction of the real agricultural task. When the actual agricultural task has a low priority, the homing trend dominates, and agricultural drones almost return directly to their home base. This weighted mechanism ensures that the intensity of incidental operations during the homing journey matches the actual task requirements.

[0025] Furthermore, once the agricultural drone completes charging and its remaining battery level recovers to a preset task release threshold, the virtual charging task associated with the agricultural drone is removed, where the task release threshold is greater than the task generation threshold. This setting prevents the agricultural drone from repeatedly triggering the creation of virtual charging tasks near the generation threshold due to short-term battery fluctuations, ensuring the stability of the system.

[0026] Furthermore, once any agricultural drone completes a real agricultural task, that task is removed from the agricultural task list, the global task list is updated, and the task priorities are rearranged. Completed tasks no longer participate in subsequent priority competition and vector synthesis calculations, and the flight guidance of each agricultural drone is adjusted in real time based on the latest task status.

[0027] The technical effects and advantages of this invention are as follows: (1) Seamless connection between task execution and charging / homing In this method, when the remaining battery power of an agricultural drone drops to a preset task generation threshold, the edge computing node creates a virtual charging task for it and integrates this virtual charging task with the real agricultural tasks into the global task list for priority management. The operation priority of the virtual charging task monotonically increases as the drone's remaining battery power decreases, enabling the drone to prioritize real agricultural tasks when its battery power is sufficient, gradually increase its homing tendency as its battery power decreases, and completely switch to homing when its battery power is critically low. The entire process is driven by the same task priority mechanism, and the drone's behavior transitions naturally from full-efficiency operation to homing for recharging, eliminating explicit switching between task mode and return-to-home mode.

[0028] (2) On the way back, the operation was carried out by chance. Each agricultural drone periodically acquires a global task list and selects the highest-priority task as its current target. When the selected target is its corresponding virtual charging task and there are still real agricultural tasks, the homing motion vector pointing to the virtual charging task target location and the operation motion vector pointing to the second-highest priority real agricultural task are weighted and synthesized. The weighting coefficient is positively correlated with the priority of the real agricultural task. This method ensures that the drone does not abruptly interrupt its current operation as its battery level decreases, but instead adds a deflection component towards the real task in the homing direction. If there are unfinished tasks near the return path, the synthesized vector will guide the drone to approach and complete those tasks, thus performing operations on the homing journey and reducing return flight distance.

[0029] (3) Load balancing between racks In selecting the target location for the virtual charging task, the overall load is determined by a weighted sum of three factors: the number of drones already included in the virtual charging task target at each nest, the number of drones currently charging, and the distance between the drone's current location and the nest. The nest with the lowest overall load is selected as the target nest, and the overall load of each nest is periodically reassessed during flight to update the target location. This method allows multiple drones to select target nests in a distributed manner based on the real-time load status of each nest during the homing process, avoiding a concentrated rush to the same nest and improving the charging load balance among the nests. Attached Figure Description

[0030] Figure 1 The following is a flowchart illustrating the overall implementation of the control method of the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Example 1 As attached Figure 1 The method shown is a cluster control method for agricultural drones in a multi-drone collaborative manner, which is applied to plant protection operations in large-area farmland.

[0033] In this scenario, multiple drone nests are distributed and deployed around or within farmland areas. Each nest manages several agricultural drones. Edge computing nodes are installed within each nest. These edge computing nodes interact with their associated agricultural drones via wireless communication links. The edge computing nodes within each nest are interconnected via wired communication links to maintain the synchronization and consistency of the global task list.

[0034] This method is executed collaboratively by edge computing nodes and the onboard control unit of agricultural drones. The overall process is achieved through the following steps.

[0035] Step S1: Each agricultural drone sends status data packets to the edge computing nodes within its associated drone nest at a preset fixed reporting cycle.

[0036] The status data packet contains the identifier i of the agricultural drone and the real-time remaining battery power E. E is a normalized real value, ranging from 0 to 1. When E is 1, it indicates that the battery is fully charged. When E is 0, it indicates that the battery is completely depleted.

[0037] For a cluster of multiple agricultural drones, each drone independently reports its status data packet. The edge computing node receives and stores the most recently reported real-time remaining battery power E of each associated agricultural drone.

[0038] Meanwhile, edge computing nodes receive a list of pending agricultural tasks from the farmland monitoring system. This system transforms pest and disease distribution data collected by field sensors or remote sensing image analysis results into a structured list of agricultural tasks.

[0039] Each record in the list corresponds to a real agricultural task. Each real agricultural task includes at least the target location coordinates and the initial task priority. . For real numbers, The larger the value, the higher the urgency of the agricultural task.

[0040] In one example, three pest and disease sites were detected in a field. The agricultural task list contains three real agricultural tasks, with the target location coordinates of the three pest and disease sites in the field. The values ​​were assigned to 30, 20, and 10 respectively, based on the severity index of pests and diseases at each location.

[0041] Step S2: Edge computing nodes generate and manage virtual charging tasks. This includes triggering and constructing the task, determining the target location, calculating the job priority, and adding it to the global task list. Each step is explained below.

[0042] Triggering of the S2.1 virtual charging task: Edge computing nodes continuously monitor the real-time remaining battery power E of all associated agricultural drones at a preset monitoring period. This monitoring period is shorter than the status reporting period of the agricultural drones to ensure that the edge computing nodes can detect changes in battery status in a timely manner.

[0043] Let the task generation threshold be... When an edge computing node detects a certain agricultural drone... The real-time remaining power E satisfies At that time, a virtual charging task was created for the agricultural drone.

[0044] In one implementation, A fixed preset value is taken, for example, 0.3. In another implementation, The adjustment is dynamically made based on the total number m of actual agricultural tasks in the current agricultural task list. When m is large, The corresponding reduction allows agricultural drones to complete more agricultural tasks within a single operation cycle before triggering homing.

[0045] There are two optional methods for triggering and generating virtual charging tasks.

[0046] The first method is initiated proactively by edge computing nodes. Edge computing nodes iterate through the real-time remaining battery power of each agricultural drone according to a preset monitoring cycle, and when the triggering conditions are met, they proactively execute the construction process.

[0047] The second method is initiated proactively by the onboard control unit of the agricultural drone. Before each agricultural drone prepares to send a status data packet to the edge computing node, the onboard control unit first compares the drone's current real-time remaining battery power with... A comparison is performed. If the triggering condition is met, a creation request flag is appended to the status data packet. Upon receiving this flag, the edge computing node executes the virtual charging task construction process for the agricultural drone.

[0048] The following uses agricultural drones Taking this as an example, the complete construction process of a virtual charging task will be explained in detail.

[0049] set up The real-time remaining battery power E gradually decreased from 0.5 to 0.3. The edge computing node detected this. Triggered as Construct a virtual charging task.

[0050] S2.2 Determination of target location coordinates: First, for Created virtual charging task assigned task identifier In this context, subscript 2 corresponds to the agricultural drone that triggered the task. The naming convention establishes a unique and clear binding relationship between each agricultural drone and its corresponding virtual charging task.

[0051] Then, determine the target location coordinates for the virtual charging task. First, obtain... The range supported by the current remaining battery power E.

[0052] In one implementation, the remaining battery power E is used as input to query a pre-defined battery power-range mapping table for the agricultural drone model. The mapping table outputs the corresponding maximum flight distance, which is the flight range.

[0053] In another implementation, the flight range is directly calculated from the remaining battery power based on the agricultural drone's cruising speed and maximum flight time on a full charge. The calculation formula is as follows: ,in t represents the cruising speed, and t represents the maximum flight time on a full charge.

[0054] Suppose there are three machine nests within the farmland operation area, denoted as nest A, nest B, and nest C. Calculate... The current location is 800m, 1200m and 2000m away from the three nests, respectively.

[0055] Assuming the flight range is determined to be 1500m by referring to a table, nest A (800m) and nest B (1200m) are both within the flight range, while nest C (2000m) is outside the range. Therefore, the candidate nest set consists of nest A and nest B.

[0056] If the candidate nest set is empty (meaning there are no reachable nests within the flight range), the flight range is incremented by a preset step size. After each increment, the selection is repeated until the candidate nest set is not empty, or the number of increments reaches a preset limit. When the number of increments reaches the limit and the candidate nest set is still empty, distance is directly selected. The nearest nest is selected as the target nest, and subsequent integrated load calculation and comparison steps are skipped.

[0057] For each drone nest in the candidate nest set, its overall load is calculated. The overall load of a nest is determined by a weighted sum of three factors: the number of agricultural drones already included in the virtual charging task objective of that nest, the number of agricultural drones currently charging, and... The straight-line distance from the nest.

[0058] The calculation method for the comprehensive load L is as follows: Where n represents the number of agricultural drones included in the virtual charging task target of the drone nest. The specific counting method is as follows: traverse all incomplete virtual charging tasks in the current global task list, and count the number of tasks whose target location coordinates are equal to the location coordinates of the drone nest.

[0059] c represents the number of agricultural drones currently charging in the drone nest, i.e., the number of charging stations in the drone nest that are currently occupied, which is reported by the drone nest to the edge computing node in real time.

[0060] d is The straight-line distance between the current location and the nest.

[0061] , , The three weighting coefficients are preset, all of which are real numbers between 0 and 1, and satisfy the following conditions: .

[0062] In this example, Take 0.5, Take 0.3, Take 0.2. For nest A, according to statistics... , , m. For nest B, , , m.

[0063] In another implementation, the calculation of the overall load also incorporates the rated charging power p of each cell. The calculation method is as follows: The larger the p-value, the faster the charging speed of the device. The new weighting factor is added, and the sum of the four factors is 1. Under the same conditions, the charging cell with higher charging power is assigned a lower overall load value.

[0064] Compare the overall load values ​​of each nest in the candidate nest set, and select the nest with the smallest overall load value.

[0065] In this example, although nest B is slightly farther than nest A, nest B's charging queue is empty, and there are currently no virtual charging tasks targeting it. Its overall load value is lower than that of nest A, therefore nest B is selected as the target nest. The location coordinates of nest B are used as the basis for virtual charging tasks. The target location coordinates.

[0066] S2.3 Determining the priority of tasks: Virtual charging task Task priority ,in accordance with The real-time remaining battery power E is dynamically determined according to a preset mapping relationship. This mapping relationship satisfies the condition that as E decreases... Monotonically increasing.

[0067] Set preset safe power level and preset warning power . Greater than In this example, Take 0.6, Take 0.2.

[0068] In one implementation, the mapping relationship takes the form of a piecewise function.

[0069] when hour, .

[0070] when hour, .

[0071] when hour, .

[0072] in, The value is 100, which is greater than the upper limit of the initial task priority for all possible real agricultural tasks. current The value is between 0.2 and 0.6. Substituting these values ​​into the calculation yields... .

[0073] In another implementation, the mapping relationship takes the form of a nonlinear function. , where k is a preset positive real number parameter. When k is 2, The rise is more rapid in the range where E is close to 0.

[0074] S2.4 Setting and updating the task type identifier list: In the task list data structure maintained by the edge computing node, each task record contains a type identifier field. For all real agricultural tasks, such as plant protection spraying tasks, this field has a constant value of 1.

[0075] Virtual charging task The type identifier field is also assigned a value of 1. This ensures that the virtual charging task is completely consistent with the plant spraying task in terms of data format. Subsequently, when the plant spraying drone iterates through the task list, there is no need to distinguish the source of the task.

[0076] After completing the above settings, the virtual charging task will be completed. As a new task entry, it is added to the global task list currently maintained by the edge computing node.

[0077] The global task list here refers to the list of agricultural tasks received by the edge computing node from the farmland monitoring system in step S1, which is the current complete task list formed after the addition or deletion of virtual charging tasks.

[0078] After the addition operation is completed, the edge computing node immediately triggers a priority reordering operation. All task entries in the list are reordered according to their job priority values ​​from highest to lowest. The resulting global task list is then available for each agricultural drone to retrieve in the next fetch cycle.

[0079] Step S3: Each agricultural drone's onboard control unit makes behavioral decisions based on a global task list. This includes determining task priorities, selecting processing branches, and calculating flight direction. Each step is explained below.

[0080] S3.1 Obtaining the global task list and determining the highest priority task: Each agricultural drone's onboard control unit pulls a copy of the current global task list from the edge computing node at preset time intervals.

[0081] After obtaining the list, the airborne control unit iterates through the job priority field values ​​of all task entries. It then compares these values ​​to find the task record with the highest job priority. If multiple task entries have the same job priority value, the one that appears first in the list is selected.

[0082] In one example, the current global task list contains three real farming tasks and one virtual charging task. The three real-world farming tasks have priorities of 30, 20, and 10, respectively. (Virtual charging task) The task priority is 75. After comparing each item, the virtual charging task... Its task was determined as the highest priority task globally due to its priority of 75.

[0083] S3.2 Task type judgment and processing branch: After determining the highest priority task, the agricultural drone will enter one of three processing branches depending on the type of the task.

[0084] Branch 1: If the highest priority task is a real agricultural task, the agricultural drone will use the target location coordinates of that real agricultural task as its flight destination. The flight control system will calculate the direction from the agricultural drone's current position to the target location and control the agricultural drone to fly in that direction.

[0085] Second branch: If the highest priority task is a virtual charging task, but this virtual charging task was created for another agricultural drone (i.e., the associated agricultural drone identifier in its task identifier is not equal to the local identifier), then this agricultural drone ignores this task record and searches for the highest priority task in the remaining task entries until it finds a virtual charging task or a real agricultural task belonging to itself.

[0086] If all remaining entries are virtual charging tasks for other drones, and there are no real agricultural tasks or virtual charging tasks belonging to the drone itself, the agricultural drone will remain hovering in the current cycle, waiting for the next retrieval cycle to obtain the updated list before making a judgment.

[0087] In one example, an agricultural drone After retrieving the list, the highest priority task was found to be... Because its associated agricultural drone identifier is not equal to its own identifier 1, Will Ignore. Search again among the remaining three real farming tasks, and select the real farming task with priority 30 as the flight target.

[0088] Third branch: If the highest priority task happens to be the virtual charging task corresponding to the drone's own identifier. (This refers to the agricultural drone.) and its virtual charging task For example. Then Further determine whether there are any other items in the list besides... In addition to real agricultural tasks.

[0089] The method for determining this is to iterate through the list and skip certain sections. The current entry is used to filter out all real farm tasks from the remaining entries. The entry with the highest priority value is then selected. In this example, the priorities of the remaining three real farm tasks are 30, 20, and 10, respectively. The real farm task with priority 30 is selected and denoted as... .

[0090] like There are no other real agricultural tasks, indicating that all current agricultural tasks have been completed. directly with The target location coordinates are used as the sole flight destination to fly to the target nest for charging.

[0091] S3.3 Vector Synthesis-Guided Flight: If it exists , Entering the vector synthesis-guided flight phase.

[0092] set up The current position coordinates are (100, 200). The target location coordinates are (800, 300). The target location coordinates are (500, 600).

[0093] Calculate the first motion vector Second motion vector .

[0094] From Current position The unit direction vector of the target position. Its direction angle is determined by the arctangent of the difference between the coordinates of the two points. represent The homing trend toward the target nest.

[0095] From Current position The unit direction vector of the target position. Its direction angle is determined by the arctangent of the difference between the coordinates of the two points. represent The trend of operations moving towards real agricultural tasks.

[0096] Determine the weighting coefficients . A real number between 0 and 1 and Task priority They are positively correlated. In this example, , .

[0097] In one implementation, .

[0098] In another implementation, , where q is a preset positive real number parameter. When q is greater than 1, exist Growth was moderate in the lower range. near The growth rate is rapid.

[0099] Calculate the actual flight direction vector using the vector composition formula. .Will Substitute and get .

[0100] When F is a non-zero vector, it is normalized. This involves dividing each component of vector F by its magnitude to obtain a unit direction vector with a magnitude of 1. The direction angle is then extracted from this vector and sent as the flight direction command to the flight control system. Flies in that direction. When F is the zero vector, along Fly in the direction of.

[0101] In this example, because With a value of only 0.3, the direction of the synthesized vector F is closer to... The direction. Primarily flying towards target nest B, while also because The existence of quantity leads to The direction is deflected. Therefore, It did not fly along a straight line leading directly to Nest B, but rather along a path towards... Flying along a curved path.

[0102] along with It continues to fly, and its current position constantly changes. At the end of each pull cycle, The latest global task list is retrieved from the edge computing node, and priority judgment and vector synthesis calculation are re-executed.

[0103] because The current location has changed. The direction angle changes accordingly. At the same time, The remaining battery power is also continuously decreasing, leading to The angle of the composite vector calculated in each cycle continuously increases. These two factors work together to cause the direction angle of the composite vector to evolve dynamically, forming a continuous and smooth guiding trajectory.

[0104] exist During flight, if its onboard sensors detect that it has entered... Within the target location's work area, spraying operations are performed to complete the task. Once completed, the edge computing node removes the task entry from the global task list.

[0105] exist In the next fetch cycle, the global task list it retrieves will no longer contain... . Re-search The highest priority real agricultural task outside of the new real agricultural task will be selected, and its priority will be recalculated based on the new real agricultural task priority. And the composite vector F. If the priority of the new real agricultural task is lower than that of the previous one. ,but As it becomes smaller, the composite vector F becomes more biased towards direction, Its flight direction is closer to its direct homing direction.

[0106] In the above description, and All vectors are taken as unit vectors, and the flight direction is obtained after synthesis and normalization.

[0107] In another implementation, and The modulus does not take a unit value, but is associated with the job priority of the corresponding task. The modulus value is , The modulus value is After vector synthesis, F is not normalized. The direction of flight is taken as the azimuth angle F, and the flight speed is proportional to the magnitude of F.

[0108] When the priorities of two tasks are similar, the magnitude of the synthesized vector is larger, and the UAV flies at a higher speed. When the priorities of the two tasks differ significantly, the synthesized vector is more biased towards the higher-priority task, the magnitude approaches a unit value, and the UAV flies at a speed close to cruise speed.

[0109] S3.4 Dynamic update of target location: regardless Regardless of which vector synthesis scheme is used for flight as described in S3.3, the edge computing nodes periodically reassess whether the target position of the virtual charging task needs to be updated with a preset update cycle.

[0110] Each time an evaluation is performed, the edge computing node reacquires... The latest location coordinates. Recalculate the overall load using the same method as in step S2. The combined load value between each machine nest.

[0111] Let the total load of the nest corresponding to the current target location be... If there is another nest with a combined load ,satisfy If the load difference exceeds the preset threshold, it will be processed in the background. The target location coordinates are updated to the location coordinates of the nest.

[0112] This update operation is for transparent. In the list retrieved in the next pull interval, The target location is the updated coordinate.

[0113] In one example During flight, the overall load on drone nest B increased as multiple other agricultural drones converged there. Meanwhile, charging stations on nest C became available, significantly reducing the overall load. If the difference in overall load between the two nests exceeds a threshold, the edge computing node will... The target location has been updated from Hive B to Hive C. After obtaining the updated list in the next cycle, turn and fly towards Nest C.

[0114] To avoid frequent target location switching due to instantaneous load fluctuations, another implementation adds a minimum update interval restriction to the update operation. The time interval between two consecutive updates to the same virtual charging task target location must not be less than the preset minimum update interval.

[0115] In a reassessment, even if the update conditions are met, if the time interval since the last update operation has not reached the minimum update interval, the update will not be executed and will wait for the next reassessment cycle to make a judgment.

[0116] Step S4: Continue flying in the manner described above. When it performs a virtual charging task... Once the target location coordinates are reached and the drone has landed, the drone nest sends a landing notification message to the edge computing node. The message includes an identifier of the landed agricultural drone.

[0117] After receiving the notification, the edge computing node locates the task identified as [identified] in the global task list. Remove the task entry from the task list.

[0118] exist After landing and connecting to the charging station to begin charging, although its associated virtual charging task has been removed from the list, The remaining battery power E is still reported periodically. The edge computing node continuously monitors the changes in its E value during the charging process.

[0119] When E is detected to rise back to the preset task termination threshold At the above time, the edge computing node is deactivated for... The virtual charging task generates monitoring markers. Greater than In one example, The value is 0.8.

[0120] Even after the mark is removed, The real-time remaining battery power dropped again Or below, the system will no longer create new virtual charging tasks for it, until They were reassigned to work sequences and dispatched to perform tasks.

[0121] Except for the scenario described in S3.3 where a real agricultural task is completed incidentally during vector synthesis-guided flight, whenever any agricultural drone completes a real agricultural task while directly performing it or during other flight phases, its onboard task detection module will confirm the task completion and send a task completion message to the edge computing node. The message carries an identifier of the completed task.

[0122] Based on this identifier, the edge computing node locates the corresponding entry in the global task list and deletes it. After deletion, the edge computing node reorders all remaining task entries according to job priority from high to low, and updates the global task list so that each agricultural drone can retrieve them in the next interval.

[0123] Through the cyclical execution of steps S1 to S4, multiple drone nests deployed in large areas of farmland and their associated agricultural drone clusters achieve a continuous and natural transition between task execution and charging / endurance, without the need for explicit task allocation instructions and independent return-to-home scheduling logic, relying solely on the dynamic generation of virtual charging tasks and flight guidance based on dual-vector synthesis.

[0124] When the agricultural drone has ample power, it fully responds to real agricultural tasks, performing farmland operations with maximum efficiency. As the power gradually depletes, its behavior, driven by a unified task priority mechanism, smoothly transitions from full-efficiency operation to operation with a homing tendency. Finally, when the power is critically low, it completely switches to homing for recharging.

[0125] The virtual charging task triggering generation method, comprehensive load calculation method, priority mapping function form, and vector synthesis method described in the above embodiments can be combined and used according to the needs of actual application scenarios. They are not limited to the specific combinations described in each embodiment.

[0126] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of the present invention. These improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for controlling agricultural unmanned aerial vehicle (UAV) swarms for multi-UAV collaboration, characterized in that, Includes the following steps: S1, obtain the real-time remaining power and list of pending agricultural tasks of each agricultural drone in the cluster of agricultural drones associated with multiple distributed drone nests in the farmland operation area, where each agricultural task includes the target location coordinates and the initial operation priority. S2, monitor the real-time remaining power of each plant protection drone. When the real-time remaining power of any plant protection drone drops to a preset task generation threshold, construct a virtual charging task for the plant protection drone and add the virtual charging task to the agricultural task list. The operation priority of the virtual charging task increases monotonically as the real-time remaining power of the plant protection drone decreases. S3, each agricultural drone queries the job priority of all tasks in the current global task list at preset time intervals, and determines the task with the highest job priority as the current target of the drone. When a plant protection drone selects a virtual charging task associated with itself as its current target, and there are other real agricultural tasks in the global task list besides the virtual charging task, the real agricultural task with the highest priority is selected and recorded as the first real agricultural task. The first motion vector of the plant protection drone pointing to the target position of the virtual charging task is vector-combined with the second motion vector pointing to the target position of the first real agricultural task to obtain the actual flight direction of the plant protection drone.

2. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-UAV collaboration according to claim 1, characterized in that, The job priority of the virtual charging task in step S2 is determined as follows: When the real-time remaining battery power of the agricultural drone is higher than the preset safe battery power, the virtual charging task has the lowest priority value. When the real-time remaining power is between the preset safe power and the preset warning power, the operation priority increases from the lowest priority value to the highest priority value as the remaining power decreases. When the real-time remaining power is lower than the preset warning power, the operation priority is the highest priority value.

3. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-machine collaboration according to claim 2, characterized in that, The preset safe battery level is 60% of the full battery level of the agricultural drone, and the preset warning battery level is 20% of the full battery level of the agricultural drone.

4. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-machine collaboration according to claim 1, characterized in that, The target location for the virtual charging task in step S2 is set as follows: Obtain the range supported by the current remaining power of the agricultural drone, select the drone nest with the smallest overall load from the nests within the range, and use its location coordinates as the target location of the virtual charging task.

5. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-machine collaboration according to claim 4, characterized in that, The overall load is determined by the weighted sum of the number of agricultural drones included in the virtual charging task target in the drone nest, the number of agricultural drones currently charging, and the distance between the current location of the agricultural drone and the drone nest.

6. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-UAV collaboration according to claim 5, characterized in that, During the process of the agricultural drone traveling to its associated virtual charging task target location, the comprehensive load of each drone nest is periodically recalculated. When the difference between the comprehensive load of another drone nest and the comprehensive load of the current target drone nest exceeds a preset threshold, the target location of the virtual charging task is updated to the location coordinates of the other drone nest.

7. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-UAV collaboration according to claim 1, characterized in that, The task type identifier of the virtual charging task in step S2 is the same as the type identifier of the plant protection spraying task in the agricultural task list.

8. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-UAV collaboration according to claim 1, characterized in that, The vector synthesis described in step S3 is composed of weighting coefficients. control, ,and It is positively correlated with the task priority of the first real agricultural task.

9. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-machine collaboration according to claim 1, characterized in that, Once the agricultural drone has finished charging and its remaining power has recovered to a preset task release threshold, the virtual charging task associated with the agricultural drone is removed, wherein the task release threshold is greater than the task generation threshold.

10. The method for controlling agricultural unmanned aerial vehicle (UAV) swarms based on multi-UAV collaboration according to claim 1, characterized in that, Once any agricultural drone completes a real agricultural task, that task is removed from the agricultural task list, the global task list is updated, and the task priorities are rearranged.

Images