Unmanned aerial vehicle cluster control method and system based on distributed reasoning

Abstract

The invention discloses an unmanned aerial vehicle cluster control method and system based on distributed reasoning, and the method comprises the steps: generating a task calculation instruction, and transmitting an own resource condition reporting instruction to all slave unmanned aerial vehicles; determining the number of slave unmanned aerial vehicles needing to cooperate according to the calculated amount of the generated task, the priority and the time delay requirement; evaluating the calculation capability of each slave unmanned aerial vehicle according to the received resource information of the slave unmanned aerial vehicles, and determining the cooperative slave unmanned aerial vehicles; based on a load balancing target, determining a task splitting result and a data splitting result model splitting result, splitting the task into a plurality of sub-tasks according to the task splitting result, and distributing the sub-tasks to corresponding slave unmanned aerial vehicles needing collaboration; receiving an execution result of executing the allocated sub-task by the cooperative slave unmanned aerial vehicle; and receiving a final execution result of the task, and issuing a corresponding execution instruction to the unmanned aerial vehicles in the unmanned aerial vehicle cluster according to the final execution result. According to the invention, adaptability and flexibility are improved.

Description

Technical Field

[0001] The present application relates to the field of drone control technology, and in particular to a drone cluster control method and system based on distributed reasoning. Background Art

[0002] Drone swarms have shown broad application prospects in many fields, including smart agriculture, post-disaster rescue, battlefield reconnaissance, etc. Through collaborative perception and computing, drones in the swarm can complete tasks autonomously without central command, improving the efficiency and accuracy of task execution.

[0003] Intelligent drone swarm is an important direction for the development of drone technology. It involves collaborative perception, calculation and decision-making between drones to achieve more efficient and intelligent task execution. Intelligent drone swarm technology is developing rapidly and is gradually being applied to multiple fields, including path planning, task collaboration, obstacle avoidance, swarm flight, etc. In the existing technology, intelligent control of drone swarms is mainly carried out in the following two ways:

[0004] The first method is to pre-assign and calculate tasks; each drone in the cluster performs tasks according to the planned algorithm or plan, such as collaborative path planning, collaborative perception, and collaborative task planning. These tasks are optimized and calculated before execution, and then assigned for execution. This method has a low degree of intelligence and requires planning and assigning tasks according to specific scenarios; and it has limitations in adaptability to dynamic task objectives and environmental changes.

[0005] The second method: the control method of biological swarms is a control strategy inspired by the behavior of biological groups in nature. It imitates the collective behavior of organisms such as bird flocks, fish schools, ant colonies, etc. to achieve the collaborative work of drone groups. For biological swarm control methods, different algorithms are needed to cope with different needs. Different algorithms are needed for path planning, task allocation, obstacle avoidance, formation, self-organization, etc. In general, there are also deficiencies in scalability and adaptability. Summary of the invention

[0006] Based on this, in response to the above technical problems, a UAV cluster control method based on distributed reasoning is provided to solve the problem that the existing technology is limited in its adaptability to dynamic mission objectives and environmental changes.

[0007] In a first aspect, a method for controlling a drone cluster based on distributed reasoning is applied to any drone in a drone cluster, and the method comprises:

[0008] Step S1: Generate a task calculation instruction and send a resource status report instruction to all slave drones; the task calculation instruction will be executed using a large model; the slave drones are other drones in the drone cluster except the drone that generates the task calculation instruction;

[0009] Step S2: determining the number of slave drones that need to be coordinated according to the computing amount, priority and delay requirements of the generated task; evaluating the computing power of each slave drone according to the received self-resource information of the slave drones, and determining the coordinated slave drones according to the computing power evaluation result of each slave drone and the number of slave drones that need to be coordinated;

[0010] Step S3: Based on the goal of load balancing, parallel planning is performed according to the requirements of the task and the resource conditions of the coordinated slave drones, and the task splitting results, data splitting results and model splitting results are determined. According to the task splitting results, the task is split into multiple subtasks and assigned to the corresponding slave drones that need to be coordinated. According to the data splitting results and the model splitting results, the model and the received data for processing the task are split and distributed to the corresponding coordinated slave drones, so that the coordinated slave drones perform the assigned subtasks; and the execution results of the assigned subtasks performed by the coordinated slave drones are received; the task splitting results include the subtasks performed by each coordinated slave drone; the data splitting results include the data required for the subtasks performed by each coordinated slave drone;

[0011] Step S4: receiving the final execution result of the task, and issuing corresponding execution instructions to the drones in the drone cluster according to the final execution result.

[0012] In the above scheme, optionally, the drone's own resource information includes computing power load, memory resources, and bandwidth and signal quality of the communication link.

[0013] In the above scheme, optionally, the method also includes: receiving the calculation delay results of the collaborative slave drones, and re-dividing the tasks according to the calculation delay results of each of the collaborative slave drones, reducing the task load of the slave drones with large calculation delay results, and increasing the task load of the slave drones with small calculation delay results.

[0014] In the above scheme, further optionally, after generating the task calculation instruction, it includes: determining multiple backup slave drones in the drone cluster, synchronizing their own process status and data with the multiple backup slave drones, and prioritizing the multiple backup slave drones; when a failure occurs, the backup slave drone ranked first takes over the work, and when a backup drone fails, the backup drone prioritized after the failure takes over the work.

[0015] In the above scheme, further optionally, the backup drone and the backup priority are determined according to the load of the slave drones in the drone cluster and the calculation delay results.

[0016] In the above scheme, optionally, after splitting the task according to the task splitting plan and assigning it to the corresponding collaborative slave drones, it also includes: receiving the failure timeout information of the collaborative slave drone, determining a substitute slave drone according to the computing power evaluation result of the slave drone that did not execute the task, and assigning the task performed by the collaborative slave drone with the failure timeout to the substitute slave drone.

[0017] In the above scheme, optionally, after splitting the task according to the task splitting plan and allocating it to the corresponding cooperative slave drones, it also includes: if the number of failure timeout information received from the cooperative slave drones exceeds the set number, re-send the self-resource status reporting instruction to all slave drones, and execute steps S2-S4.

[0018] In a second aspect, a drone cluster control system based on distributed reasoning, the system comprising:

[0019] Self-resource status reporting instruction module: used to generate task calculation instructions and send self-resource status reporting instructions to all slave drones; the task calculation instructions will be executed using the large model; the slave drones are other drones in the drone cluster except the drone that generates the task calculation instruction;

[0020] The slave drone determination module that needs to be coordinated is used to determine the number of slave drones that need to be coordinated according to the computing amount, priority and delay requirements of the generated task; evaluate the computing power of each slave drone according to the received self-resource information of the slave drone, and determine the coordinated slave drones according to the computing power evaluation result of each slave drone and the number of slave drones that need to be coordinated;

[0021] Task splitting module: used for parallel planning based on the goal of load balancing, according to the requirements of the task and the own resource conditions of the coordinated slave drones, determining the task splitting results, data splitting results and model splitting results, splitting the task into multiple subtasks according to the task splitting results and assigning them to the corresponding slave drones that need to be coordinated, splitting the model and the received data for processing the task according to the data splitting results and the model splitting results, and distributing them to the corresponding coordinated slave drones, so that the coordinated slave drones perform the assigned subtasks; and receiving the execution results of the assigned subtasks performed by the coordinated slave drones; the task splitting results include the subtasks performed by each coordinated slave drone; the data splitting results include the data required for the subtasks performed by each coordinated slave drone;

[0022] Execution result sending instruction: used to receive the final execution result of the task, and send corresponding execution instructions to the drones in the drone cluster according to the final execution result.

[0023] In a third aspect, a computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a drone cluster control method based on distributed reasoning described in the first aspect are implemented.

[0024] In a fourth aspect, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a drone cluster control method based on distributed reasoning as described in the first aspect.

[0025] This application has at least the following beneficial effects:

[0026] When any drone in a drone cluster faces environmental changes or dynamic mission targets, the present application obtains the computing resource status of other slave drones in the cluster, divides the tasks and allocates them to multiple drones, thereby allowing multiple drones to perform task calculations, making task calculations faster and allowing the drone cluster to be quickly adjusted, with improved adaptability and flexibility. At the same time, since the present application uses a large model to calculate task instructions, it can handle diverse and complex tasks more effectively, reduces dependence on specific algorithms, and improves the adaptability and flexibility of the system.

[0027] In the process of executing subtasks in a coordinated manner from the UAV, task allocation is dynamically adjusted according to the calculation delay, so that the task calculation can be performed faster and the final calculation results can be obtained faster, thus improving adaptability and flexibility.

[0028] After the drone initiates the task calculation instruction, multiple backup slave drones are determined in the drone cluster to synchronize the process status and data with themselves. When a failure occurs, the backup drone can be used to perform subsequent task allocation and transmission of calculation results, thereby ensuring the reliable execution of task calculations.

[0029] At the same time, it also monitors whether the coordinated slave drones fail. If a failure occurs, the idle slave drones in the cluster will reselect the coordinated drones and assign the tasks of the failed slave drones to the reselected coordinated slave drones to ensure the reliability of timely calculation of tasks.

[0030] Moreover, if multiple coordinated slave UAVs fail, the coordinated UAVs are re-determined to ensure timely calculation of the task. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1A flowchart of a drone cluster control system based on distributed reasoning provided for one embodiment of the present application;

[0032] Figure 2 A module diagram of a distributed reasoning framework in a drone is provided for one embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0034] In one embodiment, Figure 1 As shown, a UAV cluster control method based on distributed reasoning is provided, comprising the following steps:

[0035] Step S1: Generate a task calculation instruction and send a resource status report instruction to all slave drones; the task calculation instruction will be executed using a large model; the slave drones are other drones in the drone cluster except the drone that generates the task calculation instruction;

[0036] Step S2: determining the number of slave drones that need to be coordinated according to the computing amount, priority and delay requirements of the generated task; evaluating the computing power of each slave drone according to the received self-resource information of the slave drones, and determining the coordinated slave drones according to the computing power evaluation result of each slave drone and the number of slave drones that need to be coordinated;

[0037] Step S3: Based on the goal of load balancing, parallel planning is performed according to the requirements of the task and the resource conditions of the coordinated slave drones, and the task splitting results, data splitting results and model splitting results are determined. According to the task splitting results, the task is split into multiple subtasks and assigned to the corresponding slave drones that need to be coordinated. According to the data splitting results and the model splitting results, the model and the received data for processing the task are split and distributed to the corresponding coordinated slave drones, so that the coordinated slave drones perform the assigned subtasks; and the execution results of the assigned subtasks performed by the coordinated slave drones are received; the task splitting results include the subtasks performed by each coordinated slave drone; the data splitting results include the data required for the subtasks performed by each coordinated slave drone;

[0038] Step S4: receiving the final execution result of the task, and issuing corresponding execution instructions to the drones in the drone cluster according to the final execution result.

[0039] Specifically, the present application uses an integrated big model technology to achieve end-to-end perception, planning and execution. Regardless of the mission type and application scenario of the drone cluster, the end-to-end big model can support it.

[0040] Large models are used to achieve end-to-end cluster control, which means that the entire process from perception to control can be completed by a single model, improving efficiency and reducing error transmission between modules. This approach uses deep learning models to extract information directly from raw sensor data, achieving seamless connection from perception to control.

[0041] Drones typically fuse multiple sensor data, such as cameras, radar, GPS, and other information, to provide a more comprehensive understanding of the environment. This fusion of multimodal data enhances the system's perception capabilities and decision-making accuracy.

[0042] At the same time, distributed reasoning based on drone clusters is used to solve the problem of insufficient computing power and resources of a single drone. Large models have super-large parameters, and the required memory and computing resources are very large, which cannot be met by the resources on a single drone. The distributed reasoning framework needs to be able to allocate the storage and computing requirements of large models to different drones and use the communication links of drones to complete collaborative scheduling. Figure 2 As shown in the figure, it is a distributed reasoning framework, which includes the following:

[0043] Resource discovery: Discover the computing power, memory resources, network quality, bandwidth, and other information of each drone in the cluster through broadcast / multicast.

[0044] Parallel planning: The parallel planning module takes the result information of resource discovery (drone computing resources, storage resources, communication link quality and bandwidth, etc.) as input, and combines parallel technologies such as data parallelism and model parallelism to generate parallel planning. The planning results are synchronized to all drone devices.

[0045] Model splitting: The UAV performs model splitting according to the parallel planning configuration.

[0046] Data splitting: The input data is also split into multiple subsets according to the planned configuration. Each computing node processes a subset of the data and then merges the results.

[0047] Data and model scheduling: Drones perform calculations based on the split models and data subset distribution.

[0048] Load balancing: Ensure that all computing nodes have roughly the same workload, avoiding some nodes being idle while others are overloaded.

[0049] Collective communication: Use efficient communication mechanisms and optimize to reduce communication overhead.

[0050] The specific process of this application is as follows:

[0051] Drone distributed reasoning process: In a drone cluster without a master controller, each drone can initiate a reasoning task.

[0052] (1) The drone initiating the mission first discovers the resource status of surrounding drones through broadcast / multicast messages. The drones in the cluster respond to the broadcast / multicast messages and announce their own resource status.

[0053] (2) The UAV initiating the task selects the UAVs that need to collaborate based on the number of collaborative nodes required for the current task. The selection is based on the computing load of the UAV, the memory resources of the UAV, and the bandwidth and signal quality of the communication link.

[0054] The number of drones finally selected needs to meet the minimum requirements of the current task, and the recommended number of drones for collaboration is calculated based on the priority and latency requirements of the current task.

[0055] (3) The distributed reasoning framework will perform parallel planning based on the number of selected drones to determine the subsequent model and data splitting mode, as well as the role ID of each drone, and send the planning results and role ID to all drones.

[0056] (4) The drones participating in the distributed computing will split the computing tasks they are responsible for in the model based on the planning results and their current role ID.

[0057] (5) The drone that initiates reasoning starts reasoning, and the reasoning framework uses functions such as high-speed data distribution and collective communication to transmit reasoning data and collect results.

[0058] In the above-mentioned drone cluster control method based on distributed reasoning, when any drone in the drone cluster faces environmental changes or dynamic task targets, the computing resource status of other slave drones in the cluster is obtained, and the tasks are divided and allocated to multiple drones, so that multiple drones can perform task calculations, which makes the task calculation faster and allows the drone cluster to be quickly adjusted, thereby improving adaptability and flexibility; at the same time, since the present application uses a large model to calculate task instructions, it can handle diverse and complex tasks more effectively, reduce dependence on specific algorithms, and improve the adaptability and flexibility of the system.

[0059] In one embodiment, the drone's own resource information includes computing load, memory resources, and bandwidth and signal quality of the communication link.

[0060] In one embodiment, the method also includes: receiving the calculation delay results of the collaborative slave drones, and re-dividing the tasks according to the calculation delay results of each of the collaborative slave drones, reducing the task load of the slave drones with large calculation delay results, and increasing the task load of the slave drones with small calculation delay results.

[0061] Therefore, in the process of executing subtasks of the slave drones that need to cooperate, the task allocation is dynamically adjusted according to the calculation delay, so that the task calculation is carried out faster and the final calculation result is obtained faster, thereby improving adaptability and flexibility.

[0062] In one embodiment, the generation of the task calculation instruction includes: determining multiple backup slave drones in the drone cluster, synchronizing their own process states and data with the multiple backup slave drones, and prioritizing the multiple backup slave drones; when a failure occurs, the backup slave drone prioritized first takes over the work, and when a backup drone fails, the backup drone prioritized after the failure takes over the work.

[0063] In this embodiment, the inference initiating node acts as the coordinator of this task. Once a failure occurs, a backup node is needed to take over the work. The inference initiating node and its own backup node need to synchronize the process status and data. When the inference initiating node fails, the first backup node takes over the subsequent work and notifies other nodes.

[0064] The selection of backup nodes takes into account two factors: node load and average latency to computing nodes. Nodes in the cluster need to regularly measure the communication latency between other nodes and calculate the average latency to report to the inference initiating node. The inference initiating node selects the order of succession of backup nodes based on the load and average latency information.

[0065] It is also possible to reserve a separate unloaded node drone in the drone cluster specifically for backup to avoid the need for task replacement control when the node drone initiating the task fails.

[0066] Process state synchronization includes the following:

[0067] (1) Task status: The progress and stage of the current task, such as whether the task is in progress, paused, or completed.

[0068] (2) Data synchronization status: the data transmission and processing status between drones, including whether the data has been transmitted and the processing progress.

[0069] (3) Computation status: the role and status of each UAV in the computing task, including the completion status and computing results of the assigned tasks.

[0070] (4) Resource status: The resource usage of the drone, such as CPU and memory usage, remaining storage space, etc.

[0071] (5) Communication status: the quality of the communication link between drones, including signal strength, bandwidth usage, and connection stability.

[0072] (6) Error log: All errors and exceptions encountered during task execution, as well as the corresponding error codes or messages.

[0073] (7) Timestamp: The time stamp of key events during task execution, used for synchronization and troubleshooting.

[0074] (8) Task queue status: The queue status of pending tasks, including a list of tasks waiting to be executed, being executed, and completed.

[0075] Together, these pieces of information form the "process status" of the task in the drone cluster, providing the drone cluster with a detailed status snapshot of the task as it executes. With this information, the backup node can accurately understand the execution of the current task and take over the task when necessary.

[0076] Therefore, after the drone initiates the task calculation instruction, multiple backup slave drones are determined in the drone cluster to synchronize the process status and data with themselves. When a failure occurs, the backup drone can be used to perform subsequent task allocation and transmission of calculation results, thereby ensuring the reliable execution of task calculations.

[0077] In one embodiment, the backup drones and the backup priorities are determined based on the loads of the slave drones in the drone cluster and the calculation delay results.

[0078] In one embodiment, after splitting the task according to the task splitting plan and assigning it to the corresponding collaborative slave drones, it also includes: receiving the failure timeout information of the collaborative slave drone, determining a substitute slave drone according to the computing power evaluation result of the slave drone that did not execute the task, and assigning the task performed by the collaborative slave drone with the failure timeout to the substitute slave drone.

[0079] In this embodiment, when an intermediate computing node in the reasoning process times out due to a fault, the reasoning initiator needs to assign the role ID of the timed-out node to a substitute node and notify the replacement information within the cluster, and all related nodes will adjust the communication link. The substitute node is selected from the node with the lowest current load.

[0080] Therefore, it is also monitored whether the coordinated slave drones fail. If a failure occurs, the idle slave drones in the cluster reselect coordinated drones and assign the tasks of the failed slave drones to the reselected coordinated slave drones to ensure the reliability of timely calculation of tasks.

[0081] In one embodiment, after the task is split and assigned to the corresponding cooperative slave drones according to the task splitting plan, it also includes: if the received failure timeout information of the cooperative slave drones exceeds the set number, the self-resource status reporting instruction is re-sent to all slave drones, and steps S2-S4 are executed.

[0082] In this embodiment, when there are many node failures in the cluster and the available number is lower than the set threshold, the task master node will re-discover the resource process and use load balancing to share the tasks for the discovered nodes.

[0083] Therefore, if multiple coordinated slave UAVs fail, the coordinated UAVs are re-determined to ensure timely calculation of the task.

[0084] In one embodiment, a drone cluster control system based on distributed reasoning is provided, the system comprising:

[0085] Self-resource status reporting instruction module: used to generate task calculation instructions and send self-resource status reporting instructions to all slave drones; the task calculation instructions will be executed using the large model; the slave drones are other drones in the drone cluster except the drone that generates the task calculation instruction;

[0086] The slave drone determination module that needs to be coordinated is used to determine the number of slave drones that need to be coordinated according to the computing amount, priority and delay requirements of the generated task; evaluate the computing power of each slave drone according to the received self-resource information of the slave drone, and determine the coordinated slave drones according to the computing power evaluation result of each slave drone and the number of slave drones that need to be coordinated;

[0087] Task splitting module: used for parallel planning based on the goal of load balancing, according to the requirements of the task and the own resource conditions of the coordinated slave drones, determining the task splitting results, data splitting results and model splitting results, splitting the task into multiple subtasks according to the task splitting results and assigning them to the corresponding slave drones that need to be coordinated, splitting the model and the received data for processing the task according to the data splitting results and the model splitting results, and distributing them to the corresponding coordinated slave drones, so that the coordinated slave drones perform the assigned subtasks; and receiving the execution results of the assigned subtasks performed by the coordinated slave drones; the task splitting results include the subtasks performed by each coordinated slave drone; the data splitting results include the data required for the subtasks performed by each coordinated slave drone;

[0088] Execution result sending instruction: used to receive the final execution result of the task, and send corresponding execution instructions to the drones in the drone cluster according to the final execution result.

[0089] For the specific definition of a drone cluster control system based on distributed reasoning, please refer to the definition of a drone cluster control system based on distributed reasoning above, which will not be repeated here. Each module in the above-mentioned drone cluster control system based on distributed reasoning can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0090] In one embodiment, a computer device is provided, which may be a server, and includes a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above-mentioned method for controlling a drone cluster based on distributed reasoning is implemented.

[0091] In one embodiment, a computer program product is also provided, including a computer program / instruction, which, when executed by a processor, involves all or part of the process in the above-mentioned embodiment method.

[0092] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0093] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0094] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A UAV cluster control method based on distributed reasoning, characterized in that: Applied to any drone in a drone cluster, the method comprises: Step S1: Generate a task calculation instruction and send a resource status report instruction to all slave drones; the task calculation instruction will be executed using a large model; the slave drones are other drones in the drone cluster except the drone that generates the task calculation instruction; Step S2: determining the number of slave drones that need to be coordinated according to the computing amount, priority and delay requirements of the generated task; evaluating the computing power of each slave drone according to the received self-resource information of the slave drones, and determining the coordinated slave drones according to the computing power evaluation result of each slave drone and the number of slave drones that need to be coordinated; Step S3: Based on the goal of load balancing, parallel planning is performed according to the requirements of the task and the resource conditions of the coordinated slave drones, and the task splitting results, data splitting results and model splitting results are determined. According to the task splitting results, the task is split into multiple subtasks and assigned to the corresponding slave drones that need to be coordinated. According to the data splitting results and the model splitting results, the model and the received data for processing the task are split and distributed to the corresponding coordinated slave drones, so that the coordinated slave drones perform the assigned subtasks; and the execution results of the assigned subtasks performed by the coordinated slave drones are received; the task splitting results include the subtasks performed by each coordinated slave drone; the data splitting results include the data required for the subtasks performed by each coordinated slave drone; Step S4: receiving the final execution result of the task, and issuing corresponding execution instructions to the drones in the drone cluster according to the final execution result.

2. The method for controlling a drone cluster based on distributed reasoning according to claim 1 is characterized in that: The drone's own resource information includes computing power load, memory resources, and bandwidth and signal quality of the communication link.

3. The method for controlling a drone cluster based on distributed reasoning according to claim 1, characterized in that: The method also includes: receiving calculation delay results of the coordinated slave drones, and re-dividing tasks according to the calculation delay results of each of the coordinated slave drones, reducing the task amount of the slave drones with large calculation delay results, and increasing the task amount of the slave drones with small calculation delay results.

4. The method for controlling a drone cluster based on distributed reasoning according to claim 3 is characterized in that: The generation of the task calculation instruction includes: determining multiple backup slave drones in the drone cluster, synchronizing their own process states and data with the multiple backup slave drones, and prioritizing the multiple backup slave drones; when a failure occurs, the backup slave drone ranked first takes over the work, and when a backup drone fails, the backup drone prioritized after the failure takes over the work.

5. The method for controlling a drone cluster based on distributed reasoning according to claim 4 is characterized in that: The backup drone and the backup priority are determined according to the load of the slave drones in the drone cluster and the calculation delay results.

6. The method for controlling a drone cluster based on distributed reasoning according to claim 1, characterized in that: After splitting the task according to the task splitting plan and allocating it to the corresponding collaborative slave drones, the method also includes: receiving the failure timeout information of the collaborative slave drone, determining a substitute slave drone according to the computing power evaluation result of the slave drone that has not executed the task, and allocating the task performed by the collaborative slave drone with the failure timeout to the substitute slave drone.

7. The method for controlling a drone cluster based on distributed reasoning according to claim 1, characterized in that: After splitting the task according to the task splitting plan and allocating it to the corresponding cooperative slave drones, it also includes: if the number of failure timeout information received from the cooperative slave drones exceeds the set number, re-send the self-resource status reporting instruction to all slave drones, and execute steps S2-S4.

8. A UAV cluster control system based on distributed reasoning, characterized in that: The system comprises: Self-resource status reporting instruction module: used to generate task calculation instructions and send self-resource status reporting instructions to all slave drones; the task calculation instructions will be executed using the large model; the slave drones are other drones in the drone cluster except the drone that generates the task calculation instruction; The slave drone determination module that needs to be coordinated is used to determine the number of slave drones that need to be coordinated according to the computing amount, priority and delay requirements of the generated task; evaluate the computing power of each slave drone according to the received self-resource information of the slave drone, and determine the coordinated slave drones according to the computing power evaluation result of each slave drone and the number of slave drones that need to be coordinated; Task splitting module: used for parallel planning based on the goal of load balancing, according to the requirements of the task and the own resource conditions of the coordinated slave drones, determining the task splitting results, data splitting results and model splitting results, splitting the task into multiple subtasks according to the task splitting results and assigning them to the corresponding slave drones that need to be coordinated, splitting the model and the received data for processing the task according to the data splitting results and the model splitting results, and distributing them to the corresponding coordinated slave drones, so that the coordinated slave drones perform the assigned subtasks; and receiving the execution results of the assigned subtasks performed by the coordinated slave drones; the task splitting results include the subtasks performed by each coordinated slave drone; the data splitting results include the data required for the subtasks performed by each coordinated slave drone; Execution result sending instruction: used to receive the final execution result of the task, and send corresponding execution instructions to the drones in the drone cluster according to the final execution result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images