Data processing method, apparatus, device, system, and storage medium

Abstract

The present disclosure provides a data processing method, device, equipment, system and storage medium, and relates to the technical field of unmanned aerial vehicles. The method comprises the following steps: a core network element acquires an inference model sent by a model enabling server, then divides the inference model into multiple sub-models, and sends the corresponding sub-models to each unmanned aerial vehicle in a group of unmanned aerial vehicles; and each unmanned aerial vehicle uses the received sub-models to perform inference on sensing data to obtain inference results of each sub-model. In the present scheme, the core network element divides the inference model into multiple sub-models and sends them to the corresponding unmanned aerial vehicles, and the unmanned aerial vehicles use the sub-models to perform inference on the sensing data, thereby avoiding the unmanned aerial vehicles from sending a large amount of sensing data to the ground user end, reducing the time delay of data inference, and reducing the time delay of unmanned aerial vehicle sensing services.

Description

Technical Field

[0001] This disclosure relates to the field of wireless communication technology, and in particular to a data processing method, apparatus, device, system and storage medium. Background Technology

[0002] Unmanned Aerial Vehicle (UAV) services refer to a comprehensive range of technologies and services that utilize unmanned aerial vehicle (UAV) platforms to perform specific tasks through remote control or autonomous control.

[0003] In related technologies, a ground-based user terminal is typically responsible for controlling and managing a swarm of drones consisting of multiple drones. The drone swarm transmits a large amount of sensing data to the ground-based user terminal, which then performs inference on this data.

[0004] Ground-based user terminals can only rely on local computing resources to process large amounts of sensing data. Due to the rudimentary nature of the ground-based user terminal equipment, it is unable to provide powerful computing resources for the integration, analysis, and inference of large amounts of sensing data, resulting in long inference latency. Summary of the Invention

[0005] This disclosure provides a data processing method, apparatus, device, system, and storage medium that at least partially overcomes the problem of prolonged inference time in related technologies.

[0006] Other features and advantages of this disclosure will become apparent from the following detailed description, or may be learned in part from practice of this disclosure.

[0007] According to one aspect of this disclosure, a data processing method is provided, which is applied to a core network element, comprising: acquiring an inference model sent by a model enablement server; dividing the inference model into multiple sub-models; and sending the corresponding sub-models to each drone in a drone swarm, so that the drones can use the received sub-models for data processing.

[0008] In one exemplary embodiment of this disclosure, dividing the inference model into multiple sub-models includes: dividing the inference model into multiple sub-models according to at least one of the configuration information of the drone swarm, the configuration information of the inference model, and the task requirement information.

[0009] In an exemplary embodiment of this disclosure, dividing the inference model into multiple sub-models includes: dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and drone identifiers; broadcasting the correspondence to the drone group so that the drones determine their corresponding sub-model identifiers; and sending the corresponding sub-model to each drone in the drone group includes: receiving a sub-model request message sent by the drone for each drone in the drone group, wherein the sub-model request message includes a sub-model identifier.

[0010] Send the sub-model identifier corresponding to the sub-model to the drone.

[0011] In an exemplary embodiment of this disclosure, the configuration information of the drone group includes the priority of each drone in the drone group; determining the correspondence between sub-model identifiers and drone identifiers includes: determining the correspondence between sub-model identifiers and drone identifiers based on the granularity of the sub-model and the priority of the drone.

[0012] In one exemplary embodiment of this disclosure, the priority of a drone is determined by the drone's remaining energy and / or the distance between the drone and core network elements.

[0013] In one exemplary embodiment of this disclosure, the granularity of the sub-model is inversely proportional to the priority of the drone.

[0014] In one exemplary embodiment of this disclosure, obtaining the inference model transmitted by the model enable server includes: determining the model identifier of the inference model matching the drone group; sending a model selection signaling to the model enable server, wherein the model selection signaling includes the model identifier of the inference model; and obtaining the inference model corresponding to the model identifier sent by the model enable server.

[0015] In one exemplary embodiment of this disclosure, obtaining the inference model sent by the model enabling server includes: obtaining the updated inference model sent by the model enabling server, wherein the updated inference model is determined by the updated configuration information of the drone swarm received by the model enabling server.

[0016] In one exemplary embodiment of this disclosure, determining the model identifier of the inference model matching the drone swarm includes: determining the model identifier of the inference model matching the drone swarm based on at least one of the following: task requirement information, drone swarm configuration information, and model enablement server configuration information.

[0017] In an exemplary embodiment of this disclosure, determining the model identifier of the inference model matching the drone group based on at least one of the following: task requirement information, drone group configuration information, and model enablement server configuration information, includes: receiving a drone group registration request message sent by a drone, wherein the drone group registration request message includes: task requirement information and drone group configuration information; sending the drone group registration request message to the model enablement server; after receiving a drone group registration reply message sent by the model enablement server, determining the model identifier of the inference model matching the drone group based on at least one of the following: task requirement information, drone group configuration information, and model enablement server configuration information; and sending a drone group registration reply message to the drone group.

[0018] In one exemplary embodiment of this disclosure, the configuration information of the drone swarm includes user information to which the drone swarm belongs; sending a drone swarm registration request message to a model enabling server includes: determining a model enabling server that meets the requirements based on the user information to which the drone swarm belongs; and sending a drone swarm registration request message to the model enabling server that meets the requirements.

[0019] In one exemplary embodiment of this disclosure, the method further includes: sending the inference model to a ground user terminal corresponding to the drone group, so that the ground user terminal receives and aggregates the inference results sent by each drone.

[0020] According to one aspect of this disclosure, a data processing method is provided, applied to an unmanned aerial vehicle (UAV), comprising: receiving a sub-model sent by a core network element, the sub-model being obtained by the core network element from a segmented inference model, the inference model being a model matched with a UAV group obtained by the core network element from a model enabling server; acquiring perception data; performing inference based on the sub-model and the perception data to obtain the inference result of the sub-model; and sending the inference result of the sub-model to a ground user terminal, so that the ground user terminal can aggregate the inference results of each sub-model.

[0021] In some exemplary embodiments of this disclosure, receiving a sub-model sent by a core network element includes: receiving a correspondence between a sub-model identifier and a UAV identifier broadcast by the core network element; determining the sub-model identifier corresponding to the UAV based on the correspondence; sending a sub-model request message to the core network element, wherein the sub-model request message includes a sub-model identifier; and receiving the sub-model corresponding to the sub-model identifier sent by the core network element.

[0022] In some exemplary embodiments of this disclosure, the method further includes: obtaining shared information of each drone in the drone swarm; determining configuration information of the drone swarm based on the shared information of each drone; sending a drone swarm registration request message to a core network element, wherein the drone swarm registration request message is used to instruct the core network element to forward the drone swarm registration request message to a model enabling server, so that after receiving a drone swarm registration reply message sent by the model enabling server, the core network element obtains the model identifier of the inference model matching the drone swarm based on at least one of the task requirement information, the configuration information of the drone swarm, and the configuration information of the model enabling server; and receiving the drone swarm registration reply message sent by the core network element.

[0023] In some exemplary embodiments of this disclosure, the method further includes: receiving updated configuration information of the drone swarm; sending the updated configuration information to a model enabling server, so that the model enabling server determines an updated inference model based on the updated configuration information of the drone swarm, and sends the updated inference model to the core network.

[0024] In some exemplary embodiments of this disclosure, the method further includes: periodically sending configuration information of the drone swarm to the model enabling server, so that the model enabling server can determine whether the configuration information of the drone swarm has been updated; when the configuration information of the drone swarm has been updated, determining the updated inference model based on the updated configuration information of the drone swarm, and sending the updated inference model to the core network.

[0025] According to another aspect of this disclosure, a data processing method is provided, which is applied to a model enabling server, comprising: determining an inference model that matches a drone swarm; transmitting the inference model to a core network element, such that the core network element divides the inference model into multiple sub-models and sends the corresponding sub-models to each drone in the drone swarm, so that the drones can use the received sub-models for data processing.

[0026] In some exemplary embodiments of this disclosure, determining the inference model matching the drone swarm includes: receiving model selection signaling sent by a core network element, wherein the model selection signaling includes a model identifier of the inference model matching the drone swarm; and determining the inference model corresponding to the model identifier.

[0027] In some exemplary embodiments of this disclosure, determining the inference model that matches the drone swarm includes: receiving updated configuration information of the drone swarm; and determining an updated inference model based on the updated configuration information of the drone swarm.

[0028] In some exemplary embodiments of this disclosure, determining the inference model that matches the drone swarm includes: periodically receiving configuration information of the drone swarm sent by the drones; and when the configuration information of the drone swarm is updated, determining the updated inference model based on the updated configuration information of the drone swarm.

[0029] In some exemplary embodiments of this disclosure, the method further includes: receiving a drone group registration request message sent by a core network element; sending a drone group registration reply message to the core network element, so that after receiving the drone group registration reply message sent by the model enablement server, the core network element obtains the model identifier of the inference model matching the drone group based on at least one of the task requirement information, the configuration information of the drone group, and the configuration information of the model enablement server, and forwards the drone group registration reply message to the drone group.

[0030] According to another aspect of this disclosure, a data processing system is provided, comprising a core network element, a model enabling server, unmanned aerial vehicles (UAVs), and a ground user terminal; the model enabling server is used to determine an inference model matching a UAV swarm; the core network element is used to acquire the inference model sent by the model enabling server; divide the inference model into multiple sub-models; send the corresponding sub-models to each UAV in the UAV swarm; the UAVs are used to perform data processing using the received sub-models to obtain the inference results of each sub-model; and the ground user terminal is used to receive and aggregate the inference results of each sub-model.

[0031] According to another aspect of this disclosure, a data processing apparatus is provided, configured in a core network element, comprising: a model acquisition module for acquiring an inference model sent by a model enablement server; a model segmentation module for dividing the inference model into multiple sub-models; and a sub-model sending module for sending the corresponding sub-models to each UAV in a UAV group, so that the UAVs can use the received sub-models for data processing.

[0032] According to another aspect of this disclosure, a data processing apparatus is provided, configured on an unmanned aerial vehicle (UAV), comprising: a sub-model receiving module for receiving sub-models sent by a core network element, wherein the sub-models are obtained by the core network element from a segmented inference model, and the inference model is a model that matches the UAV group obtained by the core network element from a model enabling server; a perception data acquisition module for acquiring perception data; a data inference module for performing inference based on the sub-models and perception data to obtain inference results of the sub-models; and an inference result sending module for sending the inference results of the sub-models to a ground user terminal, so that the ground user terminal can aggregate the inference results of each sub-model.

[0033] According to another aspect of this disclosure, a data processing apparatus is provided, configured in a model enabling server, comprising: an inference model determination module for determining an inference model matching a drone swarm; and an inference model transmission module for transmitting the inference model to a core network element, so that the core network element divides the inference model into multiple sub-models and sends the corresponding sub-models to each drone in the drone swarm, so that the drones can use the received sub-models for data processing.

[0034] According to another aspect of this disclosure, an electronic device is provided, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform any of the above-described data processing methods by executing the executable instructions.

[0035] According to another aspect of this disclosure, a computer-readable storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements any of the above-described data processing methods.

[0036] According to another aspect of this disclosure, a computer program product is provided, comprising: a computer program or instructions, wherein the computer program or instructions are executed by a processor and any of the above-described data processing methods are provided.

[0037] The data processing method provided in the embodiments of this disclosure involves a core network element acquiring an inference model sent by a model enablement server, then dividing the inference model into multiple sub-models; and sending the corresponding sub-models to each drone in the drone group; the drones then use the received sub-models to process data and obtain the inference results of each sub-model. In this scheme, the core network element divides the inference model into multiple sub-models and sends them to the corresponding drones. The drones use the sub-models to infer the perception data, avoiding the drones sending a large amount of perception data to the ground user terminal, reducing the latency of data inference, and lowering the latency of drone perception services.

[0038] 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 this disclosure. Attached Figure Description

[0039] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0040] Figure 1 This diagram illustrates a data processing method applied in a core network element according to an embodiment of the present disclosure.

[0041] Figure 2 This illustration shows a flowchart of another data processing method applied in a core network element according to an embodiment of the present disclosure;

[0042] Figure 3 This illustration shows a flowchart of another data processing method applied in a core network element according to an embodiment of the present disclosure;

[0043] Figure 4 This invention discloses a flowchart of another data processing method applied in a core network element according to an embodiment of the present disclosure;

[0044] Figure 5 This diagram illustrates a flowchart of a drone swarm registration method according to an embodiment of the present disclosure;

[0045] Figure 6 This diagram illustrates a data processing method applied in a drone according to an embodiment of the present disclosure.

[0046] Figure 7This illustration shows a flowchart of another data processing method applied in a drone according to an embodiment of the present disclosure;

[0047] Figure 8 This illustration shows a flowchart of another data processing method applied in a drone according to an embodiment of the present disclosure;

[0048] Figure 9 This illustration shows a flowchart of another data processing method applied in a drone according to an embodiment of the present disclosure;

[0049] Figure 10 This diagram illustrates a data processing method applied in a model enablement server according to an embodiment of the present disclosure.

[0050] Figure 11 This illustration shows a flowchart of another data processing method applied in a model enablement server according to an embodiment of the present disclosure;

[0051] Figure 12 This illustration shows a flowchart of another data processing method applied in a model enablement server according to an embodiment of the present disclosure;

[0052] Figure 13 This illustration shows a flowchart of another data processing method applied in a model enablement server according to an embodiment of the present disclosure;

[0053] Figure 14 This diagram illustrates a data interaction flowchart applied in a data processing system according to an embodiment of the present disclosure.

[0054] Figure 15 This diagram illustrates a block diagram of an application in a data processing system according to an embodiment of the present disclosure;

[0055] Figure 16 A schematic diagram illustrating a drone swarm registration process according to an embodiment of this disclosure is shown;

[0056] Figure 17 A schematic diagram illustrating a drone swarm management process according to an embodiment of this disclosure is shown;

[0057] Figure 18 This diagram illustrates a data processing apparatus according to an embodiment of the present disclosure.

[0058] Figure 19 This diagram illustrates another data processing apparatus according to an embodiment of the present disclosure;

[0059] Figure 20 This diagram illustrates yet another data processing apparatus according to an embodiment of the present disclosure;

[0060] Figure 21 A structural block diagram of a computer device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0061] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0062] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0063] To facilitate understanding, before introducing the embodiments of this disclosure, the terms involved in the embodiments of this disclosure are explained as follows:

[0064] Data sensing involves detecting and sensing the surrounding environment through user devices and / or base stations, then storing the data locally or uploading it to the network for processing to obtain deeper insights. Data sensing includes, but is not limited to, target location, speed, and trajectory tracking, and is widely used in areas such as vehicle-to-everything (V2X) communication, drone control, smart homes and health monitoring, environmental monitoring, and large-scale Internet of Things (IoT).

[0065] Figure 1 This illustration shows a flowchart of a data processing method according to an embodiment of the present disclosure. This embodiment provides a data processing method that can be executed by any core network element with computing capabilities. For example... Figure 1 As shown, the data processing method provided in this embodiment includes the following steps.

[0066] S102. Obtain the inference model sent by the model enable server.

[0067] Core network elements can be understood as the basic units or components that constitute a telecommunications network, including but not limited to: base stations, switches, etc. Each network element has its specific functions, such as signal processing and user data management. Optionally, in this embodiment, the core network element is taken as NWDAF (Network Data Analytics Function). NWDAF is a standardized functional entity in the core network used to collect and analyze network data to provide intelligent decision support.

[0068] In the fields of artificial intelligence and machine learning, a model can be understood as a computational system generated from training data, capable of making predictions or decisions based on perceived data. Reasoning refers to the process of using a trained model to predict or classify new perceived data. A reasoning model can be understood as a pre-trained machine learning or deep learning model used to perform reasoning tasks. Reasoning models include, but are not limited to, image recognition models, classification models, language models, etc. Optionally, the aforementioned reasoning models are AI (Artificial Intelligence) models.

[0069] A model enabling server can be understood as a server or service framework in the fields of artificial intelligence and machine learning that provides specific support and services to facilitate the deployment, operation, and management of models. Model enabling servers are used to provide the necessary environment and support for the use of models, such as distributing inference models, optimizing model parameters, and monitoring their status. For example, a model enabling server may include an AI enabling server.

[0070] In some exemplary embodiments of this disclosure, the model enablement server is used to store and manage inference models. When a core network element needs to use an inference model, the core network element sends a model selection instruction to the model enablement server, and the model enablement server sends the inference model corresponding to the model selection instruction to the core network element. The core network element's use of the inference model can be understood as the core network element forwarding the inference model to the drone swarm.

[0071] In some exemplary embodiments of this disclosure, when the model enabling server determines that the configuration information of the drone group has been updated, it updates the model parameters of the inference model according to the updated configuration information to obtain the updated inference model, and sends the updated inference model to the core network element.

[0072] S104. Divide the inference model into multiple sub-models.

[0073] A sub-model can be understood as dividing a large inference model into several smaller parts or modules. Each sub-model is used to process a specific task or part of the data. In the context of a drone swarm, different drones receive different sub-models to perform specific data processing tasks.

[0074] In this embodiment, each sub-model corresponds to a drone, meaning that each sub-model will be sent to a drone and deployed in the drone, enabling the drone to perform specific inference tasks.

[0075] In one exemplary embodiment of this disclosure, the inference model is divided into multiple sub-models according to the number of drones in the drone swarm, wherein the number of sub-models is the same as the number of drones in the drone swarm.

[0076] In one exemplary embodiment of this disclosure, the inference model is divided into multiple sub-models according to the model's hierarchical structure. For example, a CNN (Convolutional Neural Networks) model is divided into three sub-models according to the model's hierarchical structure: sub-model 1 is used for feature extraction in the first few layers; sub-model 2 is used for feature combination in the middle layers; and sub-model 3 is used for classification or regression tasks in the last few layers.

[0077] In one exemplary embodiment of this disclosure, a complete inference model is split into multiple sub-models based on the user's task requirements in actual application. For example, the user's requirements are: quickly detect abnormal behavior; accurately analyze the specific type of abnormal behavior. The inference model is divided into sub-model 1 and sub-model 2. Sub-model 1 is a lightweight model used for quickly detecting whether moving objects are abnormal; it is a low-precision, low-latency model. Sub-model 2 is a heavyweight model used for detailed analysis of the specific type of abnormal behavior; it is a high-precision, high-latency model.

[0078] The inference model is divided into multiple sub-models, each sub-model corresponding to a drone in the drone group.

[0079] S106. Send the corresponding sub-model to each drone in the drone swarm so that the drones can use the received sub-model for data processing.

[0080] A drone swarm can be understood as a group of drone systems interconnected by a network and working collaboratively. The individual drones in a swarm can communicate with each other and jointly complete specific tasks according to preset algorithms or instructions.

[0081] Data processing using received sub-models by drones can be understood as the process by which drones use received sub-models to analyze, classify, and identify the collected data. This includes, but is not limited to, various tasks such as image recognition, environmental perception, and path planning.

[0082] The core network elements transmit multiple sub-models to the corresponding UAVs via communication links. After receiving the corresponding sub-model, each UAV loads the sub-model into its local computing environment and uses it to infer the perceived data to obtain the inference result.

[0083] The communication technologies used between core network elements and drones include, but are not limited to: cellular networks, satellite communication, Wi-Fi, point-to-point communication, ad hoc networks, etc.

[0084] In this embodiment, the core network element obtains the inference model sent by the model enable server, then divides the inference model into multiple sub-models and sends the corresponding sub-models to each drone in the drone swarm. The drones use the received sub-models to process data and obtain the inference results of each sub-model. Since each drone can perform a specific inference task using its own deployed sub-model, it is not necessary to send perception data to the ground user terminal for perception data inference. This avoids drones sending a large amount of perception data to the ground user terminal, reducing the latency of data inference and lowering the latency of drone perception services.

[0085] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 2 As shown, the optimized data processing method includes the following steps.

[0086] S202, Obtain the inference model sent by the model enable server.

[0087] S204. Based on at least one of the following information: configuration information of the drone swarm, configuration information of the inference model, and task requirement information, divide the inference model into multiple sub-models.

[0088] The configuration information of a drone swarm can be understood as information describing the overall and individual capabilities, status, and related parameters of the drone swarm, including but not limited to: the location information of the drone swarm, the number of drones in the swarm, the type of drone, the identification of the drone, the hardware resources of the drone, the flight endurance information of the drone, etc.

[0089] The configuration information of an inference model can be understood as information describing the structure, function, performance, and resource requirements of the inference model itself, including but not limited to: the model architecture, input and output parameters, computational resource requirements, communication requirements, etc.

[0090] Task requirements information refers to the goals, constraints, and evaluation criteria set to complete a specific task. This includes, but is not limited to: task objectives, geographical scope, time limits, accuracy requirements, resource constraints, environmental conditions, etc.

[0091] In one exemplary embodiment of this disclosure, the segmentation method based on the configuration information of the drone group includes, but is not limited to: segmentation based on the hardware resources of each drone and segmentation based on sensor type.

[0092] Based on the hardware resource allocation of each drone, if there are differences in hardware resources within the drone swarm, the model can be divided into a lightweight sub-model and a complex sub-model. The lightweight sub-model is suitable for running on drones with limited computing resources, while the complex sub-model is suitable for running on drones with strong computing capabilities.

[0093] Segmentation by sensor type: different types of sensors correspond to different sub-models. Drones are equipped with different types of sensors, such as cameras, radar, and lidar. The model can be segmented based on the characteristics of the sensor input data. The segmentation results include: an image processing sub-model for processing data from cameras, a signal processing sub-model for processing data from radar or lidar, and so on.

[0094] In one exemplary embodiment of this disclosure, the segmentation method based on inference model configuration information includes, but is not limited to: segmentation by functional modules and segmentation by real-time requirements.

[0095] If the inference model contains multiple independent functional modules, such as feature extraction, classification, and regression, it can be segmented according to the functional boundaries of the modules. The segmentation results include: feature extraction sub-models for extracting high-level features from the original data, classification / regression sub-models for classification or regression prediction, etc.

[0096] If some modules in the inference model have high real-time requirements, they can be treated as separate sub-models, and it should be ensured that this part can run under the condition of meeting the real-time requirements.

[0097] In one exemplary embodiment of this disclosure, the segmentation method based on task requirement information includes, but is not limited to: segmentation by task priority and segmentation by task complexity.

[0098] If a task comprises multiple subtasks with different priorities, the model can be divided into high-priority and low-priority sub-models. High-priority sub-models are used for urgent tasks, such as rapid target identification. Low-priority sub-models are used for non-urgent tasks, such as background analysis.

[0099] For complex and simple tasks, sub-models of different complexities can be created. Simple task sub-models handle basic tasks, such as obstacle detection. Complex task sub-models handle more challenging tasks, such as multi-object tracking.

[0100] Segmenting the inference model based on the configuration information of the drone swarm, the configuration information of the inference model, and the task requirements information can not only improve the overall performance of the drone swarm, but also enhance its adaptability and flexibility, and better meet diverse task requirements.

[0101] S206. Determine the correspondence between the sub-model identifier and the drone identifier.

[0102] A sub-model identifier can be understood as a unique identifier for each sub-model segmented from the inference model. Sub-model identifiers are used to distinguish different sub-models and to manage and schedule them. Similarly, a drone identifier can be understood as a unique identifier for each drone in a drone swarm. Each drone may have different hardware configurations, location information, or mission statuses; drone identifiers are used to identify and locate individual drones within the swarm.

[0103] The correspondence can be understood as establishing an association between each sub-model identifier and the identifier of the drone that it is suitable for operation. This includes, but is not limited to: matching the computational requirements of the sub-model with the computational capabilities of the drone; adapting the data processing type of the sub-model to the type of sensors carried by the drone; and assigning different sub-models to drones covering different areas.

[0104] Analyze the resource requirements of each sub-model and the available resources of each drone to ensure that the sub-model can operate efficiently on the corresponding drone. Based on the specific requirements of the mission, determine the drones on which the sub-model should be deployed first. For example, for sub-models requiring rapid response, select drones with strong computing power and located close to the mission area.

[0105] Create a mapping table or use a data structure to record the correspondence between each sub-model identifier and the drone identifier.

[0106] S208. Broadcast the correspondence to the drone group so that the drones can determine their corresponding sub-model identifier.

[0107] Broadcasting is a one-to-many communication mode. Using broadcasting in a drone swarm allows for the efficient delivery of information to all drone members. Broadcasting methods include, but are not limited to: radio signal broadcasting, dedicated protocol broadcasting, multi-hop network broadcasting, and internet-based broadcasting.

[0108] In one exemplary embodiment of this disclosure, if the drone swarm is relatively concentrated and operates within a short distance, radio signals can be used to broadcast directly from the core network element to all drones. For more widely distributed drone swarms, multi-hop network technology can be employed, which utilizes the method of drones forwarding information to each other to extend the signal coverage.

[0109] In one exemplary embodiment of this disclosure, a message containing the correspondence between sub-model identifiers and UAV identifiers is sent out via direct radio broadcast or multi-hop network propagation. To ensure that each UAV successfully receives the correspondence, an acknowledgment mechanism can be added to the communication protocol. For example, after receiving the correspondence, the UAV sends back an acknowledgment signal to the core network element, indicating that it has received the correspondence.

[0110] After receiving the corresponding relationship, each drone parses the message content and extracts the sub-model identifier corresponding to its own drone identifier.

[0111] S210. For each drone in the drone group, receive a sub-model request message sent by the drone, wherein the sub-model request message includes a sub-model identifier.

[0112] After receiving the broadcast mapping, each drone generates a sub-model request message. The sub-model request message includes, but is not limited to, a sub-model identifier and a drone identifier. The sub-model identifier indicates the sub-model that the drone needs to load, and the drone identifier identifies the drone that sent the request.

[0113] The drones send sub-model request messages directly to the core network elements, and the core network elements receive the sub-model request messages sent by each drone.

[0114] S212. Send the sub-model identifier corresponding to the sub-model to the drone.

[0115] The core network element retrieves the corresponding sub-model data from its local storage based on the sub-model identifier in the sub-model request message. The sub-model data includes, but is not limited to, information such as model structure, weight parameters, and configuration files, so that the UAV can correctly load and run the sub-model.

[0116] After locating the corresponding sub-model data, the core network element sends the sub-model data to the drone. Upon receiving the sub-model data from the core network element, the drone performs an integrity check. If the verification passes, the drone saves the sub-model data to local storage and prepares to load the sub-model to perform inference tasks. If the verification fails, the drone notifies the core network element, requesting a retransmission of the problematic data.

[0117] In this embodiment, by broadcasting the correspondence, the drone requests the sub-model it needs, avoiding the indiscriminate acquisition of the entire inference model's data, reducing the amount of data transmitted over the network, and lowering the risk of network congestion.

[0118] Based on the above embodiments, this embodiment further optimizes step S206, such as... Figure 3 As shown, the method for determining the optimized correspondence includes the following steps.

[0119] S302. Obtain the priority of each drone in the drone group from the configuration information of the drone group. The priority of a drone is determined by the drone's remaining energy and / or the distance between the drone and the core network element.

[0120] The priority of drones can be understood as the degree to which each drone is sorted or classified when a drone swarm is performing a task, determining whether a drone should perform certain tasks, obtain resources, or receive instructions in a priority manner.

[0121] The remaining energy of a drone can be understood as the level of energy remaining in the drone's current battery or energy storage system. During sub-model allocation, drones with higher remaining energy levels are given higher priority because they can perform tasks for longer periods without needing to return for charging or battery replacement, and can handle more critical or time-consuming tasks. Conversely, drones with low remaining energy levels are assigned lower priority to ensure that their limited energy resources are used only for necessary operations, avoiding mission failure or drone loss due to energy depletion.

[0122] The distance between a drone and a core network element refers to the distance from the drone to the core network element responsible for command, control, and data transmission. Drones that are closer to the core network element have higher priority for more efficient communication, reduced signal latency, and lower energy consumption for data transmission. Drones that are farther from the core network element have lower priority due to higher communication costs, unless a specific mission requires a drone located further away from the core network to complete the task.

[0123] In one exemplary embodiment of this disclosure, the drones in the drone swarm are networked and share information with each other. When a drone in the drone swarm receives information shared by other drones, it sorts the drones according to their remaining energy from largest to smallest to obtain the priority of each drone, and sends the priority of each drone to the core network element.

[0124] In one exemplary embodiment of this disclosure, the drones in the drone swarm are networked and share information with each other. When a drone in the drone swarm receives information shared by other drones, it sorts the drones according to the distance between each drone and the core network element from smallest to largest to obtain the priority of each drone, and sends the priority of each drone to the core network element.

[0125] In this embodiment, the priority of drones is determined by considering their remaining energy, allowing drones with sufficient power to download larger-granular sub-models first, maximizing the utilization of each drone's energy resources and avoiding mission interruptions or failures due to insufficient power. The distance between the drone and core network elements is also considered as a priority factor, assigning tasks requiring frequent communication to drones that are closer to them. This reduces communication latency and minimizes the energy consumption increase caused by long-distance transmission.

[0126] S304. Determine the correspondence between sub-model identifiers and drone identifiers based on the granularity of the sub-model and the priority of the drone. The granularity of the sub-model is inversely proportional to the priority of the drone.

[0127] The granularity of a sub-model can be understood as the size of the functionality, data volume, or computational complexity contained in each sub-model when a large inference model is divided into multiple sub-models. Granularity can be divided into coarse granularity and fine granularity. Coarse granularity means that each sub-model contains more functionality or data, while fine granularity means that each sub-model contains less functionality or data.

[0128] When determining the correspondence between sub-model identifiers and UAVs, the granularity of each sub-model is first assessed, and then matched with the priority of each UAV. For sub-models with larger granularity and higher complexity, they are assigned to UAVs with more remaining energy and closer to the core network to ensure effective task completion. For sub-models with smaller granularity and relatively simpler features, they can be flexibly assigned to UAVs with slightly lower priority based on the actual situation, which can make full use of resources and ensure that all tasks are handled properly.

[0129] In this embodiment, task offloading at different granularities is achieved in the drone swarm, which effectively reduces the latency of integrated drone communication, perception and inference, optimizes the user experience, and reduces the energy consumption of drone services to a certain extent.

[0130] In this embodiment, the optimal allocation of resources is achieved based on the granularity of the sub-model and the priority of the UAV, ensuring the effective execution of the task while improving overall operational efficiency and flexibility.

[0131] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 4 As shown, the optimized data processing method includes the following steps.

[0132] S402. Determine the model identifier of the inference model that matches the drone swarm.

[0133] A model identifier can be understood as a unique identifier or name for an inference model, used to distinguish different models. Each model identifier corresponds to a specific model structure, parameter configuration, and other characteristics, including but not limited to: model name, ID number, version number, or other forms of identifiers.

[0134] The inference model that matches the drone swarm can be understood as a model that needs to be deployed in the drone swarm and performs data inference based on the perception data collected by the drone swarm.

[0135] In some exemplary embodiments of this disclosure, an inference model corresponding to the drone swarm is determined based on the task requirements information that the drone swarm needs to perform. For example, if the drone swarm needs to conduct large-scale farmland monitoring, an inference model that processes large amounts of image data and identifies crop health status is selected as the inference model matching the drone swarm.

[0136] In some exemplary embodiments of this disclosure, an inference model matching the drone swarm is determined based on factors such as the drone's sensor type, computing power, battery life, flight speed, and communication capabilities. For example, if the drone swarm has strong local computing power, a more complex inference model is selected as the inference model matching the drone swarm; if computing resources are limited, a more efficient lightweight model is selected as the inference model matching the drone swarm.

[0137] In some exemplary embodiments of this disclosure, the computation of the inference model may depend on a model enabling server. Therefore, the inference model that matches the drone swarm can be determined according to the processor performance, memory size, storage capacity, network bandwidth, etc. of the model server. For example, if the model enabling server is configured to be very powerful, it can support complex and computationally intensive models; if the server resources are limited, a simplified model that is more suitable for remote execution needs to be selected.

[0138] S404. Send a model selection signaling message to the model enable server, wherein the model selection signaling message includes the model identifier of the inference model.

[0139] Model selection signaling is a communication instruction used to notify the model enablement server to select and issue an inference model.

[0140] The core network element embeds the model identifier into the model selection signaling and sends the model selection signaling to the model enable server through the network or other communication technologies. After receiving the model selection signaling, the model enable server parses the model selection signaling to obtain the model identifier and sends the inference model corresponding to the model identifier to the core network element.

[0141] S406. Obtain the inference model corresponding to the model identifier transmitted by the model enable server.

[0142] The core network element sends a model selection signaling message to the model enablement server. The model selection signaling message includes the model identifier of the inference model and is used to request the inference model corresponding to the model identifier.

[0143] After the model enable server selects the model, it finds the corresponding inference model based on the model identifier. The model enable server then sends the found inference model to the core network element via the network or other communication methods.

[0144] The model enabling server sends the inference model, including: the model architecture, model parameters, model identifier, information about the drone swarm, and information about the ground user terminal corresponding to the drone swarm.

[0145] In this embodiment, by sending model selection signaling to the model enable server, all inference models are uniformly managed and distributed by the server, reducing the need for local storage and maintenance of models by the UAV and reducing the complexity of the system.

[0146] S408. Divide the inference model into multiple sub-models.

[0147] S410: Send the corresponding sub-model to each drone in the drone swarm so that the drones can use the received sub-model to perform perception data inference.

[0148] S412. Send the inference model to the ground user terminal corresponding to the drone group so that the ground user terminal can receive and aggregate the inference results sent by each drone.

[0149] Ground user terminals can refer to user equipment or software terminals located on the ground. Ground user terminals can communicate with core network elements and also with drones. Ground user terminals include, but are not limited to: smartphones, tablets, laptops, desktop computers, dedicated remote controls, vehicle terminals, wearable devices, industrial control systems, etc.

[0150] The inference results are the output generated by the drone based on its sensor data and sub-models. For example, the result of target detection might be "Target A detected," or the result of path planning might be "Recommended optimal flight route."

[0151] Aggregation refers to integrating, analyzing, and processing inference results from multiple drones to generate a global inference result. Aggregation methods include, but are not limited to: fusion of inference results, fusion of inference parameters, and fusion of multi-dimensional inference parameters and inference results.

[0152] In one possible application scenario, during a search and rescue mission, each drone might only cover a small area. By aggregating the search results from all drones, the ground user can generate a complete search map to pinpoint the target's location.

[0153] After perception, reasoning, and identification, each drone in the drone swarm sends the reasoning results of its respective sub-model to the ground user. The ground user then aggregates these results to obtain detailed and accurate reasoning results. Since only the reasoning results are sent to the ground user, the transmission of a large amount of perception data is avoided, reducing communication traffic and lowering communication latency.

[0154] S414. Obtain the updated inference model transmitted by the model enable server and return to execute S408.

[0155] The updated inference model is determined by the updated configuration information of the drone swarm received by the model enabling server.

[0156] During the execution of a drone swarm mission, the drones report their configuration information to the model server. If the configuration information is updated, the model enabling server adjusts the model parameters of the inference model based on the updated configuration information, resulting in an updated inference model. The model enabling server then sends the updated inference model to the core network elements. The core network elements receive the updated inference model from the model enabling server and divide it into multiple sub-models, sending them to their respective drones so that each drone can use the received sub-models to perform its own inference task.

[0157] In this embodiment, during the drone swarm's task execution, the inference model can be updated based on the drone swarm's configuration information, thereby updating the sub-models of each drone. When a drone leaves the drone swarm or a new drone joins, the strategy can be quickly adjusted by updating the inference model, ensuring the stable operation of the entire drone swarm and improving inference efficiency.

[0158] Based on the above embodiments, this embodiment provides a method for registering drone swarms, such as... Figure 5 As shown, the drone swarm registration method provided in this embodiment includes the following steps.

[0159] S502, Receive a drone group registration request message sent by a drone, wherein the drone group registration request message includes: task requirement information and drone group configuration information.

[0160] A registration request message can be understood as a message sent by a swarm of drones to the model enablement server to register or apply to participate in a specific task. Task requirement information can be understood as the specific information about the task to be completed, including but not limited to: task objectives, expected results, and execution time range.

[0161] In one possible implementation, the drone swarm initializes after startup, and the drones share information such as location, hardware resources, and remaining energy. The drone with the highest remaining energy communicates with the core network elements and registers on the model enablement server platform.

[0162] In one possible implementation, the drone sends a drone group registration request message to the NWDAF in the core network via a base station. The drone group registration request message includes the location of the drone group, the hardware resources of each drone, the remaining energy of each drone, the user information to which the drone group belongs, and the mission requirement information.

[0163] S504. Send a drone swarm registration request message to the model enablement server.

[0164] In some exemplary embodiments of this disclosure, sending a drone swarm registration request message to a model enablement server includes the following steps.

[0165] S5042. Determine the model enabling server that meets the requirements based on the user information to which the drone group belongs.

[0166] A qualified model-enabled server can be understood as the model-enabled server closest to the drone swarm's location, or the model-enabled server closest to the ground user terminal.

[0167] In one possible implementation, NWDAF searches for model enable servers based on the various information included in the drone swarm registration request message, and finds the deployed model enable server that is closest to the ground user terminal or the drone swarm.

[0168] S5044. Send a drone swarm registration request message to the model enablement server that meets the requirements.

[0169] Based on the search results in step S5022, a drone swarm registration request message is sent to the model enabling server closest to the ground user terminal or the drone swarm. The drone swarm registration request message includes the location of the drone swarm, the hardware resources of each drone, the remaining energy of each drone, the user information to which the drone swarm belongs, and the mission requirement information.

[0170] In this embodiment, by selecting the model enabling server that is closest to the ground user terminal or the drone swarm, data transmission latency can be reduced and the response speed of the drone swarm can be improved.

[0171] S506. After receiving the drone group registration reply message sent by the model enabling server, determine the model identifier of the inference model that matches the drone group based on at least one of the following: task requirement information, drone group configuration information, and model enabling server configuration information.

[0172] After receiving the drone group registration request message, the model enabling server stores the drone group configuration information and user information locally, generates the corresponding drone group number information, and then sends a drone group registration reply message to the core network element. The drone group registration reply message contains the drone group number information and registration success information.

[0173] After receiving the drone swarm registration response message, the core network element performs model matching based on the drone swarm's configuration information, the model enablement server's configuration information, and the task requirement information to determine the inference model that matches the drone swarm.

[0174] Send a model selection signaling message to the model enablement server, informing the model enablement server of the model identifier of the selected inference model, so that the model enablement service can send the inference model corresponding to the model identifier to the core network elements.

[0175] Based on at least one of the following information—task requirements, drone swarm configuration, and model enablement server configuration—a model identifier is determined for the inference model matching the drone swarm. This includes: determining a first set of inference models corresponding to the drone swarm based on the task requirements. Models such as linear regression, convolutional models, and large language models are selected based on the task requirements. For example, if the drone swarm needs to perform large-scale farmland monitoring, an inference model that processes large amounts of image data and identifies crop health status is selected as the matching inference model. Then, a second set of inference models is selected from the first set based on the model enablement server configuration. For example, the second set includes models whose model enablement server configuration meets the hardware requirements of each inference model in the second set, or whose location is within the service range of the core network element. Finally, an inference model matching the drone swarm is selected from the second set based on the drone swarm configuration. For example, an inference model matching the drone's hardware resources is selected from the second set as the matching inference model for the drone swarm.

[0176] In this embodiment, the optimal inference model is selected by comprehensively considering task requirements, model enablement server configuration, and drone swarm configuration, thereby improving task execution efficiency and system adaptability, and enhancing user experience.

[0177] S508: Send a drone group registration reply message to the drone group.

[0178] The core network element broadcasts a registration response message from the base station to the drones in the drone swarm, indicating that the drones have successfully registered and informing them of the model enablement server and the selected model number.

[0179] In this embodiment, a process is provided for drone swarms to register with the model enabling server through the core network, thereby enabling centralized management of drone swarms, effective resource allocation and task scheduling, and ensuring smoother coordination and cooperation between different tasks.

[0180] Figure 6 This illustration shows a flowchart of a data processing method according to an embodiment of the present disclosure. This embodiment provides a data processing method that can be executed by any unmanned aerial vehicle (UAV) with computing capabilities. Figure 6 As shown, the data processing method provided in this embodiment includes the following steps.

[0181] S602. Receive the sub-model sent by the core network element. The sub-model is obtained by the core network element from the segmentation of the inference model. The inference model is the model that the core network element obtains from the model enablement server and matches the drone group.

[0182] In this embodiment, the core network element divides the inference model into multiple sub-models, each sub-model corresponding to a UAV. The core network element sends its corresponding sub-model to each UAV, and the UAV receives the sub-model sent by the core network element.

[0183] The process by which core network elements divide the inference model into multiple sub-models can be referred to the description in the above embodiments.

[0184] S604. Acquire sensor data.

[0185] Sensing data can be understood as information about the environment, objects, or events collected by sensors or other devices deployed on a drone. Sensing data includes, but is not limited to: images, videos, sounds, temperature, humidity, air pressure, and location information. The aforementioned sensors include, but are not limited to: cameras, infrared sensors, lidar, ultrasonic sensors, inertial measurement units, and barometers.

[0186] S606. Based on the sub-model and the perceived data, reasoning is performed to obtain the reasoning result of the sub-model.

[0187] The perceptual data collected in step S604 is input into the sub-model. The sub-model processes and analyzes the perceptual data based on its internal algorithm or learned knowledge, and finally generates an inference result based on the perceptual data. This inference result may include: classification labels, predicted values, anomaly detection tags, etc.

[0188] S608. Send the inference results of the sub-models to the ground user terminal so that the ground user terminal can collect and aggregate the inference results of each sub-model.

[0189] Aggregation refers to integrating, analyzing, and processing inference results from multiple drones to generate a global inference result. Aggregation methods include, but are not limited to: fusion of inference results, fusion of inference parameters, and fusion of multi-dimensional inference parameters and inference results.

[0190] In one possible application scenario, during a search and rescue mission, each drone might only cover a small area. By aggregating the search results from all drones, the ground user can generate a complete search map to pinpoint the target's location.

[0191] Each drone in the drone swarm sends its inference results to the ground user after perception, inference, and identification. The ground user then aggregates these results to obtain a detailed and accurate inference result. By sending only the inference results to the ground user, the transmission of large amounts of perception data is avoided, reducing communication traffic and lowering communication latency.

[0192] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 7 As shown, the optimized data processing method includes the following steps.

[0193] S702, Obtain shared information from each drone in the drone group.

[0194] The shared information includes, but is not limited to, location data, drone identification, and remaining energy.

[0195] In one exemplary embodiment of this disclosure, the drones in a drone swarm are networked and share information. A drone in the swarm receives shared information from the other drones, sorts them according to their remaining energy from highest to lowest, obtains their priorities, and sends these priorities to the core network element. For example, the drone with the highest priority sends the priorities of all drones to the core network element.

[0196] This is achieved through wireless communication technologies, such as Wi-Fi, radio waves, or other proprietary protocols, to obtain shared information from individual drones within a drone swarm.

[0197] S704. Determine the configuration information of the drone group based on the shared information of each drone.

[0198] By integrating the shared information from each drone, the configuration information of the drone group can be obtained. It is also possible to obtain relevant information about the users to whom the drone group belongs, as well as their task requirements. The configuration information of the drone group can then be obtained by integrating the shared information from each drone, the relevant information from the users to whom the drone group belongs, and the task requirements.

[0199] S706. Send a drone group registration request message to the core network element. The drone group registration request message is used to instruct the core network element to forward the drone group registration request message to the model enabling server, so that after receiving the drone group registration reply message sent by the model enabling server, the core network element can obtain the model identifier of the inference model that matches the drone group based on at least one of the following: task requirement information, drone group configuration information, and model enabling server configuration information.

[0200] A registration request message can be understood as a message sent by a swarm of drones to the model enablement server to register or apply to participate in a specific task. Task requirement information can be understood as the specific information about the task to be completed, including but not limited to: task objectives, expected results, and execution time range.

[0201] In one possible implementation, the drone swarm initializes after startup, and the drones share information such as location, hardware resources, and remaining energy. The drone with the highest remaining energy communicates with the core network elements and registers on the model enablement server platform.

[0202] In one possible implementation, the drone sends a drone group registration request message to the NWDAF in the core network via a base station. The drone group registration request message includes the location of the drone group, the hardware resources of each drone, the remaining energy of each drone, the user information to which the drone group belongs, and the mission requirement information.

[0203] In one possible implementation, NWDAF searches for model enablement servers based on the various information included in the drone swarm registration request message, identifying the nearest deployed model enablement server to the ground user or the drone swarm. It then sends the drone swarm registration request message to this nearest server. The drone swarm registration request message includes the drone swarm's location, the hardware resources of each drone, the remaining energy of each drone, user information of the drone swarm, and mission requirements.

[0204] After receiving the drone group registration request message, the model enabling server stores the drone group configuration information and user information locally, generates the corresponding drone group number information, and then sends a drone group registration reply message to the core network element. The drone group registration reply message contains the drone group number information and registration success information.

[0205] After receiving the drone swarm registration response message, the core network element matches the drone swarm's configuration information, the model enablement server's configuration information, and the task requirements information to determine the inference model that matches the drone swarm. It then sends a model selection signaling message to the model enablement server, informing it of the model identifier of the selected inference model, so that the model enablement service can send the inference model corresponding to the model identifier to the core network element.

[0206] S708: Receives drone group registration reply messages sent by core network elements.

[0207] After receiving the drone group registration reply message, the core network element broadcasts the drone group registration reply message to the drones in the drone group through the base station, indicating that the drone registration was successful, and informing them of the registered model enable server and the selected model number.

[0208] In this embodiment, a process is provided for drone swarms to register with the model enabling server through the core network, thereby enabling centralized management of drone swarms, effective resource allocation and task scheduling, and ensuring smoother coordination and cooperation between different tasks.

[0209] S710, the correspondence between the sub-model identifier and the UAV identifier of the core network element broadcast.

[0210] After the model enabling service sends the inference model corresponding to the model identifier to the core network elements, the core network elements divide the inference model into multiple sub-models and determine the correspondence between the sub-model identifiers and the UAV identifiers. The correspondence is then broadcast to the UAV group, allowing each UAV to receive the correspondence and determine its corresponding sub-model identifier.

[0211] S712. Determine the sub-model identifier corresponding to the UAV based on the correspondence.

[0212] After receiving the corresponding relationship, each drone parses the message content and extracts the sub-model identifier corresponding to its own drone identifier.

[0213] S714. Send a sub-model request message to the core network element, wherein the sub-model request message includes a sub-model identifier.

[0214] After receiving the broadcast correspondence, each drone generates a sub-model request message containing the following information: sub-model identifier and drone identifier. The sub-model identifier indicates the sub-model that the drone needs to load, and the drone identifier identifies the drone that sent the request.

[0215] The drones send sub-model request messages directly to the core network elements, and the core network elements receive the sub-model request messages sent by each drone.

[0216] S716, Receive the sub-model corresponding to the sub-model identifier sent by the core network element.

[0217] The core network element retrieves the corresponding sub-model data from its local storage based on the sub-model identifier in the sub-model request message. The sub-model data includes, but is not limited to, necessary information such as model structure, weight parameters, and configuration files to ensure that the UAV can correctly load and run the sub-model.

[0218] After locating the corresponding sub-model data, the core network element sends the sub-model data to the drone. Upon receiving the sub-model data from the core network element, the drone performs an integrity check. If the verification passes, the drone saves the sub-model data to its local storage and prepares to load the sub-model to begin task execution. If the verification fails, the drone notifies the core network element, requesting that the problematic portion of the data be resent.

[0219] In this embodiment, by broadcasting the correspondence, the drone requests the sub-model it needs, avoiding the indiscriminate acquisition of the entire inference model's data, reducing the amount of data transmitted over the network, and lowering the risk of network congestion.

[0220] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 8 As shown, the optimized data processing method includes the following steps.

[0221] S802: Receive updated configuration information for the drone swarm.

[0222] During a drone swarm's mission, the drones share location and online status information. When refreshed using a ground-based user terminal, the drones can detect newly added or offline drones within the swarm. Based on the information of these newly added or offline drones, the updated configuration information of the drone swarm can be obtained.

[0223] S804. Send the updated configuration information to the model enable server so that the model enable server can determine the updated inference model based on the updated configuration information of the UAV group and send the updated inference model to the core network.

[0224] The drones report their updated configuration information to the model server. The model enabling server adjusts the inference model parameters based on this updated configuration information, resulting in an updated inference model. The model enabling server then sends the updated inference model to the core network elements. The core network elements receive the updated inference model from the model enabling server and divide it into multiple sub-models, sending each sub-model to its corresponding drone. This allows each drone to execute its respective inference task using the received sub-model.

[0225] In this embodiment, the configuration information of the drone group is reported by the drone. The model enablement server can quickly adjust the inference model through the updated configuration information to ensure that the sub-models sent to the drones are always consistent with the current state of the drone group, thereby improving the inference accuracy.

[0226] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 9 As shown, the optimized data processing method includes the following steps.

[0227] S902 periodically receives configuration information from drone swarms.

[0228] During the mission, the drones in the drone swarm share location and online information, and the drones obtain the configuration information of the drone swarm in real time or periodically.

[0229] S904. Periodically send the configuration information of the UAV group to the model enable server so that the model enable server can determine whether the configuration information of the UAV group has been updated. When the configuration information of the UAV group has been updated, determine the updated inference model based on the updated configuration information of the UAV group and send the updated inference model to the core network elements.

[0230] The drones periodically report their configuration information to the model server. If the drone group's configuration information is updated, the model enabling server adjusts the inference model parameters based on the updated configuration information, resulting in an updated inference model. The model enabling server then sends the updated inference model to the core network elements. The core network elements receive the updated inference model from the model enabling server and divide it into multiple sub-models, sending them to their respective drones so that each drone can use the received sub-models to perform its own inference task.

[0231] In this embodiment, the configuration information of the drone group is reported by the drone. The model enable server checks whether the configuration information of the drone group has been updated. If it has been updated, the inference model can be quickly adjusted using the updated configuration information to ensure that the sub-models sent to the drones are always consistent with the current state of the drone group, thereby improving the inference accuracy.

[0232] Figure 10 This illustration shows a flowchart of a data processing method according to an embodiment of the present disclosure. This embodiment provides a data processing method that can be executed by any model-enabled server with computational processing capabilities. Figure 10 As shown, the data processing method provided in this embodiment includes the following steps.

[0233] S1002. Determine the inference model that matches the drone swarm.

[0234] In one exemplary embodiment of this disclosure, the core network element embeds the model identifier into the model selection signaling and sends the model selection signaling to the model enablement server via a network or other communication technology. Upon receiving the model selection signaling, the model enablement server parses the signaling to obtain the model identifier and acquires the inference model corresponding to that identifier.

[0235] In one exemplary embodiment of this disclosure, the drone reports the updated configuration information of the drone group to the model server. The model enabling server adjusts the model parameters of the inference model according to the updated configuration information of the drone group to obtain the updated inference model.

[0236] In one exemplary embodiment of this disclosure, the drone periodically reports the configuration information of the drone group to the model server. If the configuration information of the drone group is updated, the model enabling server adjusts the model parameters of the inference model according to the updated configuration information of the drone group to obtain the updated inference model.

[0237] S1004. Transmit the inference model to the core network element so that the core network element divides the inference model into multiple sub-models and sends the corresponding sub-models to each drone in the drone group so that the drones can use the received sub-models for data processing.

[0238] The core network element sends a model selection signaling message to the model enablement server. This message includes a model identifier for the inference model and is used to request the inference model corresponding to that identifier. Upon receiving the model selection signaling message, the model enablement server locates the corresponding inference model based on the model identifier and then transmits the inference model to the core network element via the network or other communication methods. The transmission of the inference model by the model enablement server includes: the model architecture, model parameters, model identifier, relevant information about the UAV swarm, and relevant information about the ground user terminal corresponding to the UAV swarm.

[0239] After receiving the inference model, the core network element divides the inference model into multiple sub-models and sends the sub-models to the corresponding UAVs. The UAVs use the sub-models to infer the perception data and obtain the inference results of the sub-models.

[0240] In this solution, the core network element divides the inference model into multiple sub-models and sends them to the corresponding UAVs. The UAVs use the sub-models to infer the perception data, avoiding the UAVs sending a large amount of perception data to the ground user terminal, reducing the latency of data inference, and lowering the latency of UAV perception services.

[0241] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 11 As shown, the optimized data processing method includes the following steps.

[0242] S1102, Receive the drone group registration request message sent by the core network element.

[0243] In one exemplary embodiment of this disclosure, the drones in a drone swarm are networked and share information. A drone in the swarm receives shared information from other drones and sorts them according to their remaining energy from highest to lowest to determine their priorities. The shared information from each drone, along with information about the user to whom the drone swarm belongs, is integrated to obtain the drone swarm's configuration information. Based on the drone swarm's configuration information, the user's information, and task requirements, a drone swarm registration request message is generated.

[0244] The drone sends a drone group registration request message to the NWDAF in the core network via the base station. This message includes the drone group's location, the hardware resources of each drone, the remaining energy of each drone, user information of the drone group, and mission requirements. Based on the information in the drone group registration request message, the NWDAF searches for the model enabler server, identifying the nearest deployed model enabler server to the ground user or the drone group. It then sends the drone group registration request message to the nearest model enabler server.

[0245] S1104. The drone group registration reply message sent to the core network element enables the core network element to obtain the model identifier of the inference model matching the drone group based on at least one of the following information: task requirement information, drone group configuration information, and model enable server configuration information, and forward the drone group registration reply message to the drone group.

[0246] The model enabling server stores the configuration and user information of the drone group locally and generates the corresponding drone group number information. Then, it sends a drone group registration reply message to the core network element. The drone group registration reply message contains the drone group number information and registration success information.

[0247] After receiving the drone swarm registration response message, the core network element performs model matching based on the drone swarm's configuration information, the model enablement server's configuration information, and the task requirement information to determine the inference model that matches the drone swarm.

[0248] In this embodiment, a process is provided for drone swarms to register with the model enabling server through the core network, thereby enabling centralized management of drone swarms, effective resource allocation and task scheduling, and ensuring smoother coordination and cooperation between different tasks.

[0249] S1106. Receive model selection signaling sent by the core network element, wherein the model selection signaling includes the model identifier of the inference model matched with the UAV group.

[0250] After receiving the drone swarm registration response message, the core network element matches the drone swarm's configuration information, the model enablement server's configuration information, and the task requirements information to determine the inference model that matches the drone swarm. It then sends a model selection signaling message to the model enablement server, informing it of the model identifier of the selected inference model.

[0251] S1108. Determine the inference model corresponding to the model identifier.

[0252] The model enabling service queries the corresponding inference model based on the model identifier and sends the inference model corresponding to the model identifier to the core network elements.

[0253] In this embodiment, by sending model selection signaling to the model enable server, all inference models are uniformly managed and distributed by the server, reducing the need for local storage and maintenance of models by the UAV and reducing the complexity of the system.

[0254] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 12 As shown, the optimized data processing method includes the following steps.

[0255] S1202, Receive updated configuration information of the drone swarm.

[0256] During a drone swarm's mission, the drones share location and online status information. When refreshed using a ground-based user terminal, the drones can detect newly added or offline drones within the swarm. Based on the information of these newly added or offline drones, the updated configuration information of the drone swarm can be obtained.

[0257] S1204. Determine the updated inference model based on the updated configuration information of the drone swarm.

[0258] The drones report their updated configuration information to the model server. The model enabling server adjusts the inference model parameters based on this updated configuration information, resulting in an updated inference model. The model enabling server then sends the updated inference model to the core network elements. The core network elements receive the updated inference model from the model enabling server and divide it into multiple sub-models, sending each sub-model to its corresponding drone. This allows each drone to execute its respective inference task using the received sub-model.

[0259] In this embodiment, the configuration information of the drone group is reported by the drone. The model enablement server can quickly adjust the inference model through the updated configuration information to ensure that the sub-models sent to the drones are always consistent with the current state of the drone group, thereby improving the inference accuracy.

[0260] Based on the above embodiments, this embodiment optimizes the data processing method, such as... Figure 13 As shown, the optimized data processing method includes the following steps.

[0261] S1302. Periodically receive configuration information of drone groups sent by drones.

[0262] During the mission execution by the drone swarm, the individual drones in the swarm share location and online information. The drones acquire the swarm's configuration information in real time or periodically, and periodically send this configuration information to the model enablement server.

[0263] S1304. When the configuration information of the drone swarm is updated, determine the updated inference model based on the updated configuration information of the drone swarm.

[0264] If the configuration information of the drone swarm is updated, the model enabling server adjusts the model parameters of the inference model according to the updated configuration information, resulting in an updated inference model. The model enabling server sends the updated inference model to the core network elements. The core network elements receive the updated inference model from the model enabling server, divide the updated inference model into multiple sub-models, and send them to their respective drones, so that each drone can use the received sub-models to perform its own inference task.

[0265] In this embodiment, the configuration information of the drone group is reported by the drone. The model enable server checks whether the configuration information of the drone group has been updated. If it has been updated, the inference model can be quickly adjusted using the updated configuration information to ensure that the sub-models sent to the drones are always consistent with the current state of the drone group, thereby improving the inference accuracy.

[0266] Based on the above embodiments, this embodiment provides a data processing system, such as... Figure 14 As shown, the data processing system includes: core network elements, a model enabling server, drones, and ground user terminals. Figure 14 As shown, the interactive method for data processing includes the following steps.

[0267] S1402, The model enabling server determines the inference model that matches the drone swarm.

[0268] S1404, Core network element acquisition model enables the inference model sent by the server.

[0269] S1406, the core network elements divide the inference model into multiple sub-models.

[0270] S1408, the core network element sends the corresponding sub-model to each drone in the drone swarm.

[0271] S1410 The UAV uses the received sub-model to process data and obtain the inference results of the sub-model.

[0272] S1412. The UAV sends the inference results of the sub-model to the ground user terminal.

[0273] S1414. The ground user terminal receives and aggregates the inference results of each sub-model.

[0274] The data processing system interaction method provided in the embodiments of this disclosure includes: a core network element acquiring an inference model sent by a model enablement server, then dividing the inference model into multiple sub-models; and sending the corresponding sub-models to each drone in the drone group; the drones using the received sub-models to process data and obtain the inference results of each sub-model, and sending them to the ground user terminal; the ground user terminal receiving and aggregating the inference results of each sub-model. In this scheme, the core network element divides the inference model into multiple sub-models and sends them to the corresponding drones. The drones use the sub-models to infer the perception data, avoiding the drones sending a large amount of perception data to the ground user terminal, reducing the latency of data inference, and lowering the latency of drone perception services.

[0275] Based on the above embodiments, this embodiment provides a data processing system, such as... Figure 15 As shown, the data processing system includes: a drone 1510, a core network 1520, a model enabling server 1530, and a ground user terminal 1540. The drone 1510 includes a model enabling client 1511 and a perception client 1512, and the core network 1520 includes an NWDAF unit 1521 and a UPF unit 1522. The model enabling server 1530 deploys an inference model.

[0276] In one possible implementation, the perception client 1512 establishes a connection with the model-enabled server 1530 through the UPF unit 1522. The model-enabled client 1511 collects user perception data processing requirements and requests the necessary inference model from the model-enabled server 1530 through the NWDAF unit 1521.

[0277] The perception client 1512 locally perceives multimodal environmental information, including images, videos, and audio, and transmits it to the inference model deployed on the model enablement server 1530 via the UPF unit 1522 in the core network. The data perceived by the perception client 1512 is compressed and encrypted locally on the device, and compressed data and corresponding labels are provided to the inference model for fine-tuning training. After matching, the model enablement server 1530 distributes the corresponding model to the perception client 1512 and the ground user terminal 1540 via the data channel of the UPF unit 1522. The perception client 1512 transmits the inference results to the ground user terminal 1540, which performs aggregation processing to obtain the required recognition inference results.

[0278] In one possible implementation scenario, a drone swarm registration process is provided, such as... Figure 16 As shown, the registration process for drone swarms includes the following steps.

[0279] S1602, The UAV sends a UAV group registration request message to the NWDAF unit.

[0280] After startup, the drone swarm initializes, with each drone sharing information such as location, hardware, and battery level. The drone with the highest remaining battery level communicates with the base station and registers with the model enablement server. This drone then sends a drone swarm registration request message to the NWDAF in the core network via the base station. The request includes information such as the drone swarm's location, hardware, battery level, and the user information to which the drone swarm belongs.

[0281] S1604, NWDAF enables the server by finding the nearest model based on the user's location.

[0282] NWDAF retrieves the model enable server based on the information in step S1602, and finds the deployed model enable server that is closest to the user.

[0283] S1606 and NWDAF send a drone swarm registration request message to the model enablement server.

[0284] Based on the search results in step S1604, NWDAF sends a drone group registration request message to the selected model enabling server, which includes the drone group information and the user information from step S1602.

[0285] S1608, The model enablement server sends a drone swarm registration reply message to NWDAF.

[0286] After receiving the drone group registration request message from step S1606, the model enabling server stores the drone group information and user information locally, generates the corresponding drone group number information, and then sends a drone group registration reply message to NWDAF, which includes the drone group number information and registration success information.

[0287] S1610, NWDAF determines the model identifier of the inference model that matches the drone group based on at least one of the task requirement information, the configuration information of the drone group, and the configuration information of the model enabling server.

[0288] S1612, NWDAF sends a model selection signaling message to the model enable server.

[0289] S1614 and NWDAF send drone group registration reply messages to the drones.

[0290] NWDAF broadcasts a drone group registration reply message to each drone in the drone group via the base station, indicating that the drone has successfully registered and informing the registered model enable server and the selected model number.

[0291] In one possible implementation scenario, a drone swarm management process is provided, such as... Figure 17 As shown, the registration process for drone swarms includes the following steps.

[0292] S1702. The drone sends the updated configuration information of the drone group to the model enablement server.

[0293] The drone swarm shares location and online information. When the user refreshes the app, the drone with the first number detects the appearance of a new drone or the offline status of an existing drone, and then sends a message to the model enablement server platform to report the update information of the drone swarm.

[0294] Based on shared location and online information, the drones periodically report drone swarm information to the model enablement server via base stations.

[0295] S1704. The model enablement server updates the local storage information based on the updated configuration information of the drone swarm.

[0296] The model enabling server updates its locally stored information based on the information reported by the drones, including the location and online information of the drone group in step S1702.

[0297] S1706, The model enablement server sends a group information update message to the drone.

[0298] The model enable server sends a group information update message to the drone, indicating that the information stored locally on the model enable server has been updated.

[0299] S1708. The model enabling server updates and distributes the inference model based on the updated configuration information.

[0300] In one possible implementation scenario, users use drones to scan the environment and collect information from the air, including data from photography, infrared cameras, and radio signal acquisition. Users and drone groups register with a model enabling server via the network, searching for and matching the required inference models according to their needs. Multiple drones within a cellular area perform tasks such as image recognition and identify each other. The drones with the highest priority recognize the models sent from the cloud and save the smallest granular model and parameters from the original model. Other drones save the remaining models and parameters in order of priority. The drones perform perception, recognition, and inference, and then send the recognition results to the ground user terminal for aggregation to obtain detailed and accurate results.

[0301] In this embodiment, the airborne distributed inference architecture enables task offloading at different granularities within the drone swarm, effectively reducing the latency of integrated drone communication, perception, and inference, optimizing user experience, and reducing the energy consumption of drone services to a certain extent.

[0302] It should be noted that the acquisition, storage, use, and processing of data in this disclosed technical solution comply with the relevant provisions of national laws and regulations. The various types of data, such as personal identity data, operational data, and behavioral data related to individuals, customers, and groups, obtained in the embodiments of this disclosure have all been authorized.

[0303] Based on the same inventive concept, this disclosure also provides a data processing apparatus, as described in the following embodiments. Since the principle by which this apparatus solves the problem is similar to that of the method embodiments described above, the implementation of this apparatus embodiment can refer to the implementation of the method embodiments described above, and repeated details will not be repeated.

[0304] Figure 18 This diagram illustrates a data processing apparatus according to an embodiment of the present disclosure. The data processing apparatus is configured within a core network element, such as... Figure 18 As shown, the device includes: a model acquisition module 1810, a model segmentation module 1820, and a sub-model sending module 1830.

[0305] The model acquisition module 1810 is used to acquire the inference model sent by the model enablement server; the model segmentation module 1820 is used to divide the inference model into multiple sub-models; and the sub-model sending module 1830 is used to send the corresponding sub-model to each drone in the drone group so that the drone can use the received sub-model for data processing.

[0306] In an exemplary embodiment of this disclosure, the model segmentation module 1820 is specifically used to divide the inference model into multiple sub-models according to at least one of the configuration information of the UAV swarm, the configuration information of the inference model, and the task requirement information.

[0307] In an exemplary embodiment of this disclosure, the model segmentation module 1820 is specifically used to divide the inference model into multiple sub-models and determine the correspondence between sub-model identifiers and drone identifiers; broadcast the correspondence to the drone group so that the drones can determine their corresponding sub-model identifiers; the sub-model sending module 1830 is specifically used to receive sub-model request messages sent by drones for each drone in the drone group, wherein the sub-model request message includes a sub-model identifier; and send the sub-model corresponding to the sub-model identifier to the drone.

[0308] In an exemplary embodiment of this disclosure, the configuration information of the drone group includes the priority of each drone in the drone group; the model segmentation module 1820 is specifically used to determine the correspondence between the sub-model identifier and the drone identifier based on the granularity of the sub-model and the priority of the drone.

[0309] In one exemplary embodiment of this disclosure, the priority of a drone is determined by the drone's remaining energy and / or the distance between the drone and core network elements.

[0310] In one exemplary embodiment of this disclosure, the granularity of the sub-model is inversely proportional to the priority of the drone.

[0311] In an exemplary embodiment of this disclosure, the model acquisition module 1810 is specifically used to determine the model identifier of the inference model matching the drone swarm; send a model selection signaling to the model enabling server, wherein the model selection signaling includes the model identifier of the inference model; and acquire the inference model corresponding to the model identifier sent by the model enabling server.

[0312] In an exemplary embodiment of this disclosure, the model acquisition module 1810 is specifically used to acquire the updated inference model sent by the model enabling server, wherein the updated inference model is determined by the updated configuration information of the drone group received by the model enabling server.

[0313] In an exemplary embodiment of this disclosure, the model acquisition module 1810 is specifically used to determine the model identifier of the inference model that matches the drone group based on at least one of the task requirement information, the configuration information of the drone group, and the configuration information of the model enabling server.

[0314] In an exemplary embodiment of this disclosure, the model acquisition module 1810 is specifically configured to receive a drone group registration request message sent by a drone, wherein the drone group registration request message includes: task requirement information and drone group configuration information; send the drone group registration request message to the model enabling server; after receiving the drone group registration reply message sent by the model enabling server, determine the model identifier of the inference model matching the drone group based on at least one of the task requirement information, drone group configuration information and model enabling server configuration information; and send the drone group registration reply message to the drone group.

[0315] In an exemplary embodiment of this disclosure, the configuration information of the drone swarm includes user information to which the drone swarm belongs; the model acquisition module 1810 is specifically used to determine a model enabling server that meets the requirements based on the user information to which the drone swarm belongs; and to send a drone swarm registration request message to the model enabling server that meets the requirements.

[0316] In one exemplary embodiment of this disclosure, it further includes: an inference model sending module, used to send the inference model to a ground user terminal corresponding to the drone group, so that the ground user terminal receives and aggregates the inference results sent by each drone.

[0317] Figure 19 This diagram illustrates a data processing apparatus according to an embodiment of the present disclosure, which is configured in a drone, such as... Figure 19 As shown, the device includes: a sub-model receiving module 1910, a perception data acquisition module 1920, a data reasoning module 1930, and a reasoning result sending module 1940.

[0318] The sub-model receiving module 1910 is used to receive sub-models sent by core network elements. The sub-models are obtained by the core network elements from the segmentation of the inference model. The inference model is a model that matches the UAV group obtained by the core network elements from the model enabling server. The perception data acquisition module 1920 is used to acquire perception data. The data inference module 1930 is used to perform inference based on the sub-models and perception data to obtain the inference results of the sub-models. The inference result sending module 1940 is used to send the inference results of the sub-models to the ground user terminal so that the ground user terminal can collect and aggregate the inference results of each sub-model.

[0319] In some exemplary embodiments of this disclosure, the sub-model receiving module 1910 is specifically used to receive the correspondence between the sub-model identifier and the UAV identifier broadcast by the core network element; determine the sub-model identifier corresponding to the UAV according to the correspondence; send a sub-model request message to the core network element, wherein the sub-model request message includes the sub-model identifier; and receive the sub-model corresponding to the sub-model identifier sent by the core network element.

[0320] In some exemplary embodiments of this disclosure, a registration module is configured to: obtain shared information of each drone in a drone swarm; determine configuration information of the drone swarm based on the shared information of each drone; send a drone swarm registration request message to a core network element, wherein the drone swarm registration request message instructs the core network element to forward the drone swarm registration request message to a model enabling server, so that after receiving a drone swarm registration reply message from the model enabling server, the core network element obtains the model identifier of the inference model matching the drone swarm based on at least one of the task requirement information, the drone swarm configuration information, and the model enabling server configuration information; and receive a drone swarm registration reply message from the core network element.

[0321] In some exemplary embodiments of this disclosure, the system further includes: a configuration information update module, configured to receive updated configuration information of the drone swarm; send the updated configuration information to a model enable server, so that the model enable server determines the updated inference model based on the updated configuration information of the drone swarm, and sends the updated inference model to the core network.

[0322] In some exemplary embodiments of this disclosure, the configuration information update module is further configured to periodically send configuration information of the drone group to the model enablement server, so that the model enablement server can determine whether the configuration information of the drone group has been updated. When the configuration information of the drone group is updated, the updated inference model is determined based on the updated configuration information of the drone group, and the updated inference model is sent to the core network.

[0323] Figure 20 This diagram illustrates a data processing apparatus according to an embodiment of the present disclosure. The data processing apparatus is configured in a model enable server, such as... Figure 20 As shown, the device includes: an inference model determination module 2010 and an inference model transmission module 2020.

[0324] The inference model determination module 2010 is used to determine the inference model that matches the UAV group; the inference model sending module 2020 is used to transmit the inference model to the core network element, so that the core network element divides the inference model into multiple sub-models and sends the corresponding sub-models to each UAV in the UAV group, so that the UAV can use the received sub-models for data processing.

[0325] In some exemplary embodiments of this disclosure, the inference model determination module 2010 is specifically used to receive model selection signaling sent by a core network element, wherein the model selection signaling includes a model identifier of an inference model matching the UAV group; and to determine the inference model corresponding to the model identifier.

[0326] In some exemplary embodiments of this disclosure, the inference model determination module 2010 is specifically used to receive updated configuration information of the drone swarm and determine an updated inference model based on the updated configuration information of the drone swarm.

[0327] In some exemplary embodiments of this disclosure, the inference model determination module 2010 is specifically used to periodically receive configuration information of a drone group sent by a drone; when the configuration information of the drone group is updated, it determines an updated inference model based on the updated configuration information of the drone group.

[0328] In some exemplary embodiments of this disclosure, the system further includes: a registration module, configured to receive a drone group registration request message sent by a core network element; and send a drone group registration reply message to the core network element, so that after receiving the drone group registration reply message sent by the model enablement server, the core network element obtains the model identifier of the inference model matching the drone group based on at least one of the task requirement information, the configuration information of the drone group, and the configuration information of the model enablement server, and forwards the drone group registration reply message to the drone group.

[0329] It should be noted that the examples and application scenarios implemented by the modules in the above device embodiments and the corresponding steps in the method embodiments are the same, but are not limited to the content disclosed in the above method embodiments. It should also be noted that the above modules, as part of the device, can be executed in a computer system such as a set of computer-executable instructions.

[0330] Those skilled in the art will understand that various aspects of this disclosure can be implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which can be collectively referred to herein as a "circuit", "module" or "system".

[0331] According to the same inventive concept, this disclosure also provides an electronic device, which includes: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the data processing method described above by executing the executable instructions. Since the principle by which this electronic device solves the problem is similar to that of the above method embodiments, the implementation of this electronic device embodiment can refer to the implementation of the above method embodiments, and repeated details will not be described again.

[0332] The following reference Figure 21 To describe an electronic device 2100 according to such an embodiment of the present disclosure. Figure 21 The electronic device 2100 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments disclosed herein. The electronic device 2100 may include any one of a core network element, a drone, or a model enabling server.

[0333] like Figure 21 As shown, the electronic device 2100 is manifested in the form of a general-purpose computing device. The components of the electronic device 2100 may include, but are not limited to: the aforementioned processing unit 2110, the aforementioned storage unit 2120, and the bus 2130 connecting different system components (including the storage unit 2120 and the processing unit 2110).

[0334] The storage unit stores program code that can be executed by the processing unit 2110, causing the processing unit 2110 to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure. For example, the processing unit 2110 can perform any data processing method in the above embodiments.

[0335] Storage unit 2120 may include readable media in the form of volatile storage units, such as random access memory (RAM) 21201 and / or cache memory 21202, and may further include read-only memory (ROM) 21203.

[0336] Storage unit 2120 may also include a program / utility 21204 having a set (at least one) program module 21205, such program module 21205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0337] Bus 2130 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0338] Electronic device 2100 can also communicate with one or more external devices 2140 (e.g., keyboard, pointing device, Bluetooth device, etc.), one or more devices that enable a user to interact with electronic device 2100, and / or any device that enables electronic device 2100 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 2150. Furthermore, electronic device 2100 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 2160. As shown, network adapter 2160 communicates with other modules of electronic device 2100 via bus 2130. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 2100, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0339] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0340] Based on the same inventive concept, this disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements any of the above-described data processing methods. Since the principle by which this computer-readable storage medium embodiment solves the problem is similar to that of the above-described method embodiments, the implementation of this computer-readable storage medium embodiment can refer to the implementation of the above-described method embodiments, and repeated details will not be elaborated further.

[0341] More specific examples of computer-readable storage media in this disclosure may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0342] In this disclosure, a computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting a program for use by or in connection with an instruction execution system, apparatus, or device.

[0343] Optionally, the program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0344] In practical implementation, program code for performing the operations of this disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0345] Based on the same inventive concept, this disclosure also provides a computer program product, comprising: a computer program or instructions, wherein the computer program or instructions, when executed by a processor, implement any one of the data processing methods described in the above method embodiments. Since the principle by which this computer program product embodiment solves the problem is similar to that of the above method embodiments, the implementation of this computer program product embodiment can refer to the implementation of the above method embodiments, and repeated details will not be elaborated further.

[0346] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0347] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0348] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, mobile terminal, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0349] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

Claims

1. A data processing method, characterized in that, The method is applied to core network elements, including: Obtain the inference model sent by the model enablement server; The inference model is divided into multiple sub-models; dividing the inference model into multiple sub-models includes: dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and UAV identifiers; broadcasting the correspondence to each UAV in the UAV group, so that each UAV determines its corresponding sub-model identifier; the correspondence between the sub-model identifier and the UAV identifier is determined according to the granularity of the sub-model and the priority of the UAV; the priority of the UAV is determined by the remaining energy of the UAV and / or the distance between the UAV and the core network element; the granularity of the sub-model is inversely proportional to the priority of the UAV; Sending a corresponding sub-model to each drone in the drone swarm, so that the drones can use the received sub-models for perception data inference; sending the corresponding sub-model to each drone in the drone swarm includes: receiving a sub-model request message sent by each drone in the drone swarm, wherein the sub-model request message includes a sub-model identifier; and sending the sub-model corresponding to the sub-model identifier to the drone.

2. The data processing method according to claim 1, characterized in that, The step of dividing the inference model into multiple sub-models includes: The inference model is divided into multiple sub-models based on at least one of the following: the configuration information of the drone swarm, the configuration information of the inference model, and the task requirement information.

3. The data processing method according to claim 2, characterized in that, The configuration information of the drone swarm includes the priority of each drone in the drone swarm.

4. The data processing method according to claim 2, characterized in that, The inference model sent by the model enabling server includes: Determine the model identifier for the inference model that matches the drone swarm; Send a model selection signaling message to the model enable server, wherein the model selection signaling message includes the model identifier of the inference model; Obtain the inference model corresponding to the model identifier sent by the model enable server.

5. The data processing method according to claim 1, characterized in that, The inference model sent by the model enabling server includes: Obtain the updated inference model sent by the model enabling server, wherein the updated inference model is determined by the updated configuration information of the drone group received by the model enabling server.

6. The data processing method according to claim 4, characterized in that, The model identifier for determining the inference model that matches the drone swarm includes: Based on at least one of the task requirement information, the configuration information of the drone swarm, and the configuration information of the model enabling server, determine the model identifier of the inference model that matches the drone swarm.

7. The data processing method according to claim 6, characterized in that, The step of determining the model identifier of the inference model matching the drone swarm based on at least one of the task requirement information, the configuration information of the drone swarm, and the configuration information of the model enabling server includes: Receive a drone group registration request message sent by the drone, wherein the drone group registration request message includes: the task requirement information and the configuration information of the drone group; Send the drone group registration request message to the model enabling server; After receiving the drone group registration response message sent by the model enabling server, the model identifier of the inference model matching the drone group is determined based on at least one of the task requirement information, the configuration information of the drone group, and the configuration information of the model enabling server. Send the drone group registration reply message to the drone group.

8. The data processing method according to claim 7, characterized in that, The configuration information of the drone group includes the user information to which the drone group belongs; Sending a drone swarm registration request message to the model enabling server includes: The model enabling server that meets the requirements is determined based on the user information to which the drone group belongs; Send the drone group registration request message to the model enabling server that meets the requirements.

9. The data processing method according to claim 1, characterized in that, Also includes: The inference model is sent to the ground user terminal corresponding to the drone group, so that the ground user terminal receives and aggregates the inference results sent by each drone.

10. A data processing method, characterized in that, The method is applied to drones and includes: The process involves receiving a sub-model sent by a core network element, wherein the sub-model is obtained by the core network element from segmenting an inference model, and the inference model is a model matched with a drone swarm obtained by the core network element from a model enabling server; the segmentation of the inference model by the core network element includes: the core network element dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and drone identifiers; broadcasting the correspondence to each drone in the drone swarm, so that each drone determines its corresponding sub-model identifier; and receiving the sub-model sent by the core network element includes receiving the sub-model identifier broadcast by the core network element and... The process involves: establishing a mapping relationship between drone identifiers; determining the sub-model identifier corresponding to the drone based on the mapping relationship; sending a sub-model request message to the core network element, wherein the sub-model request message includes the sub-model identifier; receiving the sub-model corresponding to the sub-model identifier sent by the core network element; the mapping relationship between the sub-model identifier and the drone identifier is determined based on the granularity of the sub-model and the priority of the drone; the priority of the drone is determined by the remaining energy of the drone and / or the distance between the drone and the core network element; the granularity of the sub-model is inversely proportional to the priority of the drone. Acquire sensor data; The reasoning result of the sub-model is obtained by reasoning based on the sub-model and the perceived data; The inference results of the sub-models are sent to the ground user terminal so that the ground user terminal can aggregate the inference results of each sub-model.

11. The data processing method according to claim 10, characterized in that, Also includes: Obtain shared information from each drone in the drone group; The configuration information of the drone group is determined based on the shared information of each drone. A drone swarm registration request message is sent to the core network element, wherein the drone swarm registration request message is used to instruct the core network element to forward the drone swarm registration request message to the model enabling server, so that after receiving the drone swarm registration reply message sent by the model enabling server, the core network element obtains an inference model matching the drone swarm based on at least one of the following: task requirement information, configuration information of the drone swarm, and configuration information of the model enabling server. Receive the drone group registration reply message sent by the core network element.

12. The data processing method according to claim 10, characterized in that, Also includes: Receive the updated configuration information of the drone swarm; The updated configuration information is sent to the model enabling server so that the model enabling server determines the updated inference model based on the updated configuration information of the UAV group and sends the updated inference model to the core network element.

13. The data processing method according to claim 10, characterized in that, Also includes: The configuration information of the UAV group is periodically sent to the model enabling server so that the model enabling server can determine whether the configuration information of the UAV group has been updated. When the configuration information of the UAV group is updated, the updated inference model is determined based on the updated configuration information of the UAV group and the updated inference model is sent to the core network element.

14. A data processing method, characterized in that, The method is applied to a model enabling server and includes: Determine the inference model that matches the drone swarm; The inference model is sent to a core network element, which then divides the inference model into multiple sub-models and sends the corresponding sub-models to each drone in the drone swarm, enabling the drones to perform perception data inference using the received sub-models. The process of the core network element dividing the inference model into multiple sub-models and sending the corresponding sub-models to each drone in the drone swarm includes: dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and drone identifiers; broadcasting the correspondence to each drone in the drone swarm, enabling each drone to determine its corresponding sub-model identifier; receiving a sub-model request message from each drone in the drone swarm, wherein the sub-model request message includes a sub-model identifier; and sending the sub-model corresponding to the sub-model identifier to the drone. The correspondence between the sub-model identifier and the drone identifier is determined based on the granularity of the sub-model and the priority of the drone. The priority of the drone is determined by the drone's remaining energy and / or the distance between the drone and the core network element. The granularity of the sub-model is inversely proportional to the priority of the drone.

15. The data processing method according to claim 14, characterized in that, The inference model for determining the match with the drone swarm includes: The system receives model selection signaling sent by the core network element, wherein the model selection signaling includes a model identifier of the inference model matched with the UAV group. Determine the inference model corresponding to the model identifier.

16. The data processing method according to claim 14, characterized in that, The inference model for determining the match with the drone swarm includes: Receive the updated configuration information of the drone swarm; The updated inference model is determined based on the updated configuration information of the drone swarm.

17. The data processing method according to claim 14, characterized in that, The inference model for determining the match with the drone swarm includes: Periodically receive configuration information of the drone group sent by the drone; When the configuration information of the drone swarm is updated, the updated inference model is determined based on the updated configuration information of the drone swarm.

18. The data processing method according to claim 15, characterized in that, Also includes: Receive the drone group registration request message sent by the core network element; The drone group registration response message sent to the core network element enables the core network element to determine the model identifier of the inference model matching the drone group based on at least one of the following information: task requirement information, configuration information of the drone group, and configuration information of the model enabling server, and then forward the drone group registration response message to the drone group.

19. A data processing system, characterized in that, The system includes core network elements, a model enabling server, drones, and ground user terminals; The model-enabled server is used to determine the inference model that matches the drone swarm. The core network element is used to acquire the inference model sent by the model enabling server; divide the inference model into multiple sub-models; and send the corresponding sub-models to each drone in the drone group. Dividing the inference model into multiple sub-models and sending the corresponding sub-models to each drone in the drone swarm includes: dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and drone identifiers; broadcasting the correspondence to each drone in the drone swarm, so that each drone determines its corresponding sub-model identifier; receiving a sub-model request message from each drone in the drone swarm, wherein the sub-model request message includes a sub-model identifier; sending the sub-model corresponding to the sub-model identifier to the drone; the correspondence between the sub-model identifier and the drone identifier is determined based on the granularity of the sub-model and the priority of the drone; the priority of the drone is determined by the drone's remaining energy and / or the distance between the drone and the core network element; the granularity of the sub-model is inversely proportional to the priority of the drone; The drone is used to process the received sub-models to obtain the reasoning results of each sub-model. The ground user terminal is used to receive and aggregate the inference results of each of the sub-models.

20. A data processing apparatus, characterized in that, The device is configured in a core network element and includes: The model acquisition module is used to acquire the inference model sent by the model enablement server; A model segmentation module is used to divide the inference model into multiple sub-models. Dividing the inference model into multiple sub-models includes: dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and UAV identifiers; broadcasting the correspondence to each UAV in the UAV group, so that each UAV determines its corresponding sub-model identifier; the correspondence between the sub-model identifier and the UAV identifier is determined based on the granularity of the sub-model and the priority of the UAV; the priority of the UAV is determined by the remaining energy of the UAV and / or the distance between the UAV and the core network element; the granularity of the sub-model is inversely proportional to the priority of the UAV. The sub-model sending module is used to send corresponding sub-models to each drone in the drone group, so that the drones can use the received sub-models to perform perception data inference; sending corresponding sub-models to each drone in the drone group includes: receiving a sub-model request message sent by each drone in the drone group, wherein the sub-model request message includes a sub-model identifier; and sending the sub-model corresponding to the sub-model identifier to the drone.

21. A data processing apparatus, characterized in that, The device is configured on a drone and includes: A sub-model receiving module is used to receive sub-models sent by core network elements. The sub-models are obtained by the core network elements from segmenting an inference model. The inference model is a model obtained by the core network elements from a model enabling server that matches the drone group. The segmentation of the inference model by the core network elements includes: dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and drone identifiers; broadcasting the correspondence to each drone in the drone group, so that each drone can determine its corresponding sub-model identifier. Receiving the sub-models sent by the core network elements includes receiving the sub-models broadcast by the core network elements. The system establishes a correspondence between model identifiers and UAV identifiers; determines the sub-model identifier corresponding to the UAV based on the correspondence; sends a sub-model request message to the core network element, wherein the sub-model request message includes the sub-model identifier; receives the sub-model corresponding to the sub-model identifier sent by the core network element; the correspondence between the sub-model identifier and the UAV identifier is determined based on the granularity of the sub-model and the priority of the UAV; the priority of the UAV is determined by the remaining energy of the UAV and / or the distance between the UAV and the core network element; the granularity of the sub-model is inversely proportional to the priority of the UAV. The sensing data acquisition module is used to acquire sensing data; The data reasoning module is used to perform reasoning based on the sub-model and the perceived data to obtain the reasoning result of the sub-model; The inference result sending module is used to send the inference results of the sub-models to the ground user terminal, so that the ground user terminal can aggregate the inference results of each sub-model.

22. A data processing apparatus, characterized in that, The device is configured in the model enabling server and includes: The inference model determination module is used to determine the inference model that matches the drone swarm. An inference model sending module is used to transmit the inference model to a core network element, so that the core network element divides the inference model into multiple sub-models and sends the corresponding sub-models to each drone in the drone group, enabling the drones to perform perception data inference using the received sub-models. The core network element dividing the inference model into multiple sub-models and sending the corresponding sub-models to each drone in the drone group includes: dividing the inference model into multiple sub-models and determining the correspondence between sub-model identifiers and drone identifiers; broadcasting the correspondence to each drone in the drone group, so that each drone determines its corresponding sub-model identifier; receiving a sub-model request message from each drone in the drone group, wherein the sub-model request message includes a sub-model identifier; and sending the sub-model corresponding to the sub-model identifier to the drone. The correspondence between the sub-model identifier and the drone identifier is determined based on the granularity of the sub-model and the priority of the drone. The priority of the drone is determined by the drone's remaining energy and / or the distance between the drone and the core network element. The granularity of the sub-model is inversely proportional to the priority of the drone.

23. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to perform the data processing method of any one of claims 1 to 18 by executing the executable instructions.

24. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the data processing method according to any one of claims 1 to 18.

25. A computer program product comprising: A computer program or instruction, characterized in that, when executed by a processor, the computer program or instruction implements the data processing method according to any one of claims 1 to 18.

Images