Relay terminal based unmanned aerial vehicle control method and system

Abstract

The application is suitable for the technical field of unmanned aerial vehicle control, and particularly relates to a method and system for controlling unmanned aerial vehicles based on a relay end, which comprises the following steps: obtaining a task of an unmanned aerial vehicle, and determining a data transmission link between the unmanned aerial vehicle and an edge device; building an edge layer by using data in the unmanned aerial vehicle, building a fog layer by using data in the edge device, constructing a distributed computing architecture, generating a delimiter by using the edge layer, dividing the task, containerizing the divided slices, generating a container snapshot, and caching the container snapshot in the fog layer; and opening the control authority of the unmanned aerial vehicle, and establishing a control relationship between the distributed computing architecture and the unmanned aerial vehicle. The application can greatly shorten the response time of the unmanned aerial vehicle by generating a container snapshot and restoring the unmanned aerial vehicle in case of failure, and can also quickly start a backup unmanned aerial vehicle to continue executing the task, so as to realize seamless replacement, avoid data loss, and ensure the normal execution of the task.

Description

Technical Field

[0001] The present invention relates to the field of drone control technology, and in particular to a drone control method and system based on a relay terminal. Background Art

[0002] Usually, the transmission distance of wireless network signals is limited. When used on a large scale, a large number of base stations need to be set up along the transmission path of the wireless network as nodes in the network transmission process, so as to achieve wireless network signal coverage to a farther range; however, if the drone needs to temporarily access the wireless network in an area outside the coverage range of the wireless network, the method of setting up base stations in the existing technology is obviously not applicable, which requires the use of relay terminals. Relay terminals play a key role in drone systems, wireless communications and other remote sensing networks, especially in the process of performing complex tasks and extending the operating range. The main function of the relay terminal is to enhance signal transmission, extend the communication distance, and improve the flexibility and reliability of the system.

[0003] However, when executing a mission, the drone needs to cooperate with the cloud server, which requires a large amount of data transmission, which can easily lead to transmission delays. In addition, this will also increase the difficulty of data processing. When a drone fails, a longer response time is required, affecting the normal execution of the mission. Therefore, "how to respond quickly when a drone fails" is the technical problem that the present invention needs to solve. Summary of the Invention

[0004] The purpose of the present invention is to provide a drone control method and system based on a relay terminal to solve the problem of "how to quickly respond when a drone fails" raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for controlling a drone based on a relay terminal, the method comprising:

[0007] Obtain the drone's mission and determine the data transmission link between the drone and the edge device;

[0008] The edge layer is built using data from drones, and the fog layer is built using data from edge devices, thereby constructing a distributed computing architecture. The edge layer is used to generate delimiters, and the tasks are segmented. The segmented slices are containerized, and container snapshots are generated. The container snapshots are then cached in the fog layer.

[0009] Open the control authority of the drone and establish the control relationship between the distributed computing architecture and the drone. When receiving the fault data sent by the drone, extract the container snapshot in the fog layer and restore the snapshot in the pre-configured backup drone. Use the restored data to redeploy the backup drone, use the backup drone as the controlled object, and migrate the control relationship.

[0010] Furthermore, the steps of obtaining the mission of the drone and determining the data transmission link between the drone and the edge device include:

[0011] collecting location data of the area where the task is located, and clustering the task into a flight part and an execution part;

[0012] Establish the corresponding relationship between the flight part, execution part and position data respectively;

[0013] Define a data transmission protocol, debug the data transmission link, and determine an optimization method based on the debugging results.

[0014] Furthermore, the steps of building an edge layer using data from drones and a fog layer using data from edge devices to construct a distributed computing architecture include:

[0015] Perform data preprocessing in the edge layer to obtain valid data, and superimpose the created local decision mechanism into the edge layer;

[0016] The number of edge devices in the fog layer is determined, the number being at least one, and a division of labor is set for all edge devices.

[0017] Furthermore, the steps of generating delimiters using the edge layer, segmenting the task, containerizing the segmented slices, generating container snapshots, and caching the container snapshots in the fog layer include:

[0018] Generate a separator based on the progress of the task, and divide the task into unfinished task slices and completed task slices;

[0019] Containerize the unfinished task slices, obtain container snapshots, and cache the container snapshots in the fog layer;

[0020] An incremental snapshot is generated using the progress, and the container snapshot is revised.

[0021] Furthermore, the control authority of the drone is opened, and the control relationship between the distributed computing architecture and the drone is established. When the fault data sent by the drone is received, the container snapshot in the fog layer is extracted and restored in the pre-configured backup drone. The backup drone is redeployed using the restored data, and the backup drone is used as the controlled object. The steps of migrating the control relationship include:

[0022] Loading the container snapshot and restoring it, and redeploying the restored data to the backup drone through the control relationship;

[0023] Develop and execute flight plans, and transmit data acquired by the backup drone during flight to the edge layer.

[0024] Furthermore, the method further comprises:

[0025] Initialize the properties of the edge device and configure the correspondence between tasks and edge devices;

[0026] By traversing the created localization decision set, the data to be uploaded is selected from the edge layer;

[0027] By utilizing the upload channel between the edge device and the cloud, the data to be uploaded is transferred to the processing model in the cloud;

[0028] Establishing a federated architecture for edge devices, backing up the processing model to the edge devices, and training the processing model using the federated architecture;

[0029] The parameters obtained during training are uploaded to the cloud and aggregated, and the processing model is updated using the aggregated data.

[0030] Furthermore, the method further comprises:

[0031] Inserting time scheduling windows into the federated architecture and formulating asynchronous update strategies;

[0032] Insert version tags into the processing models of the edge device and the cloud, and determine the rollback steps of the processing models based on a preset rollback mechanism.

[0033] Furthermore, the system includes:

[0034] The determination module is used to obtain the UAV's mission and determine the data transmission link between the UAV and the edge device;

[0035] A slicing module is used to build an edge layer using data from drones and a fog layer using data from edge devices, thereby constructing a distributed computing architecture. The module generates delimiters using the edge layer, slices the tasks, containerizes the slices obtained after slicing, generates container snapshots, and caches the container snapshots in the fog layer.

[0036] The recovery module is used to open the control authority of the drone and establish the control relationship between the distributed computing architecture and the drone. When the fault data sent by the drone is received, the container snapshot in the fog layer is extracted and the snapshot is restored in the pre-configured backup drone. The backup drone is redeployed using the restored data, and the control relationship is migrated with the backup drone as the controlled object.

[0037] Furthermore, the determining module includes:

[0038] a clustering unit, configured to collect location data of an area where the mission is located and cluster the mission into a flight portion and an execution portion;

[0039] A corresponding unit, used to establish corresponding relationships between the flight part, the execution part and the position data respectively;

[0040] The debugging unit is used to define the data transmission protocol, debug the data transmission link, and determine the optimization method according to the debugging result.

[0041] Furthermore, the segmentation module includes:

[0042] An acquisition unit, configured to perform data preprocessing in the edge layer, acquire valid data, and superimpose the created local decision mechanism into the edge layer;

[0043] The statistical unit is used to determine the number of edge devices in the fog layer, where the number is at least one, and set a division of labor for all edge devices.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. By determining the data transmission link, the data in the drone can be transmitted to the edge device for processing, shortening the data transmission distance, reducing latency, and greatly improving the interactivity of the drone. By building a distributed computing architecture, the bandwidth requirements of the drone are reduced, greatly improving the data processing efficiency. By splitting the tasks, the workload of data transmission is further reduced. By generating container snapshots and restoring them when the drone fails, the response time of the drone can be greatly shortened. At the same time, when a drone fails, the backup drone can be quickly started to continue the mission, thereby achieving seamless replacement, avoiding data loss, and ensuring the normal progress of the mission.

[0046] 2. By traversing the localized decision set, decisions can be made in edge devices, greatly improving the decision-making efficiency of drones. At the same time, by building a federated architecture, processing models in the cloud can be trained locally, which protects data privacy while greatly reducing data transmission costs. By determining the rollback steps, the processing model can be restored to the most stable version, thereby improving the accuracy of the processing model. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention.

[0048] Figure 1 A flowchart of a relay-based drone control method according to an embodiment of the present invention;

[0049] Figure 2 A block diagram of the first sub-process of the relay-based drone control method provided in an embodiment of the present invention;

[0050] Figure 3 A second sub-flow chart of the relay-based drone control method provided in an embodiment of the present invention;

[0051] Figure 4 A block diagram of the third sub-flow of the relay-based drone control method provided in an embodiment of the present invention;

[0052] Figure 5 A block diagram of the relay-based drone control system provided in an embodiment of the present invention;

[0053] Figure 6 A block diagram of the composition of the determination module in the relay-based UAV control system provided by an embodiment of the present invention;

[0054] Figure 7 A block diagram of the composition of the segmentation module in the relay-based UAV control system provided by an embodiment of the present invention;

[0055] Figure 8 A block diagram of the recovery module in a relay-based drone control system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0057] In Example 1, Figure 1The following is a flowchart of the relay-based drone control method according to an embodiment of the present invention, which is described in detail as follows:

[0058] S100: Obtain the mission of the drone and determine the data transmission link between the drone and the edge device.

[0059] Obtain the tasks that the drone needs to perform and determine the data transmission link between the drone and the edge device. The edge device here refers to a device capable of processing drone data. It can be a ground device or a drone with integrated edge computing capabilities. When the edge device needs to be used in conjunction with an edge gateway, control device, etc., after determining the data transmission link, testing is required to ensure the stability and reliability of the data transmission link.

[0060] S200: Build an edge layer using data from drones and a fog layer using data from edge devices to construct a distributed computing architecture. Generate delimiters using the edge layer and segment the tasks. Containerize the segmented slices to generate container snapshots, and cache the container snapshots in the fog layer.

[0061] The edge layer is built using data from drones. The data in the drones should be divided into two parts: one is the flight data needed by the drone during flight, and the other is the data generated when the drone collects or performs tasks. These data are transferred to the edge layer, and a fog layer is built in the edge device to build a distributed computing structure. The edge layer is located at the outermost layer, and the fog layer is located between the edge layer and the cloud. The difference between the two is that the edge layer and the fog layer are used to process different data. For some data with higher latency requirements, they should be processed in the edge layer, while for some data with lower latency requirements and larger computational load, they need to be processed in the fog layer or the cloud.

[0062] Based on the data generated during the drone flight, a delimiter is generated, and the task is divided into a completed part and an unfinished part using the delimiter. The unfinished part and the corresponding data are containerized, and the generated container snapshot is cached in the fog layer.

[0063] S300: Open the control authority of the drone and establish the control relationship between the distributed computing architecture and the drone. After receiving the fault data sent by the drone, extract the container snapshot in the fog layer and restore the snapshot in the pre-configured backup drone. Use the data obtained after recovery to redeploy the backup drone, use the backup drone as the controlled object, and migrate the control relationship.

[0064] The control authority of the unmanned aerial vehicle is opened to the edge device in the distributed computing architecture. The unmanned aerial vehicle sends data to the edge device during the execution of a task. When the edge device receives fault data, a container snapshot in the edge device is extracted and snapshot recovery is performed. It should be noted that the backup unmanned aerial vehicle needs to be pre-configured, so that the unmanned aerial vehicle can take off quickly and execute the task. The backup unmanned aerial vehicle is re-deployed by using the recovered data, and the backup unmanned aerial vehicle is taken as a controlled object to migrate and establish a control relationship between the backup unmanned aerial vehicle and the edge device.

[0065] In embodiment 2, Figure 2 The implementation flow of the unmanned aerial vehicle control method based on the relay end is shown. The steps of obtaining the task of the unmanned aerial vehicle and determining the data transmission link between the unmanned aerial vehicle and the edge device are described as follows.

[0066] S101: Collect position data of the area where the task is located, and cluster the task into a flight part and an execution part.

[0067] The area where the unmanned aerial vehicle executes the task is determined according to the task, and the position data of the area is determined. The task is divided into a flight part and an execution part. The flight part is the task of controlling the unmanned aerial vehicle, and the execution part is the action required to execute the task.

[0068] For example, a flight task of an unmanned aerial vehicle is to reach point A and capture high-resolution images of farmland by using a multispectral camera carried by the unmanned aerial vehicle. The reaching of point A and the adjustment of the specific flight attitude during the image capturing process are the tasks of the flight part, and how to start and adjust the multispectral camera to capture images is the task of the execution part.

[0069] S102: Corresponding relationships between the flight part, the execution part and the position data are respectively established.

[0070] The corresponding relationships between the flight part and the position data and between the execution part and the position data are established, so as to determine the action required by the unmanned aerial vehicle at each position.

[0071] S103: Define a data transmission protocol, debug the data transmission link, and determine an optimization mode according to the debugging result.

[0072] The data transmission protocols of the unmanned aerial vehicle and the edge device and the edge device and the cloud are determined. The data transmission link is debugged, and the data transmission link is optimized according to the debugging result to ensure the stability and reliability of the data transmission link. The specific optimization mode can be data compression, batch transmission and the like.

[0073] In embodiment 3, Figure 3The implementation process of the relay-based drone control method provided by an embodiment of the present invention is shown. The following details the steps of using the data in the drone to build the edge layer, the data in the edge device to build the fog layer, and the distributed computing architecture.

[0074] S201: Perform data preprocessing in the edge layer to obtain valid data, and superimpose the created local decision mechanism into the edge layer.

[0075] The data in the edge layer is preprocessed to improve the accuracy of data processing and determine the useful data. Some of the data in the edge layer needs to be processed in the drone, while some needs to be sent to the edge device for processing. Here, the data that needs to be sent to the edge device for processing is defined as valid data, and the valid data is classified into the fog layer, and then the local decision-making mechanism is superimposed on the edge layer. When the data is processed in the edge layer, a decision can be made by integrating the local decision-making mechanism. The advantage of this is that it can improve the decision-making efficiency of the edge layer and reduce data transmission.

[0076] S202: Determine the number of edge devices in the fog layer, which number is at least one, and set a division of labor for all edge devices.

[0077] Determine the number of edge devices in the fog layer and set up division of labor for all edge devices. Different edge devices are responsible for processing different valid data. The advantage of this is that valid data can be processed in parallel, thereby improving the processing efficiency of valid data.

[0078] In Example 4, Figure 3 The implementation process of the relay-based drone control method provided by an embodiment of the present invention is shown. The following details the steps of generating delimiters using the edge layer, segmenting the task, containerizing the resulting slices, generating container snapshots, and caching the container snapshots in the fog layer.

[0079] S203: Generate a separator according to the progress of the task, and divide the task into unfinished task slices and completed task slices.

[0080] The progress of the task is determined based on the data in the edge layer, and a separator is generated. The separator is mainly used to split the task into unfinished task slices and completed task slices. After the task is split, it can be processed in different edge devices to improve data processing efficiency.

[0081] S204: Containerize the unfinished task slices, obtain a container snapshot, and cache the container snapshot in the fog layer.

[0082] The unfinished task slices are containerized. The advantage of containerization is that they can be quickly deployed in different application environments. At the same time, they can also update container snapshots while using fewer transmission resources to ensure the accuracy of container snapshots.

[0083] S205: Generate an incremental snapshot using the progress, and modify the container snapshot.

[0084] By using the data in the progress and edge layers, incremental snapshots are generated and used to correct the container snapshots. By using incremental snapshots, the consumption of transmission resources can be further reduced.

[0085] In Example 5, Figure 4 The implementation process of the relay-based drone control method provided by an embodiment of the present invention is shown. The following details the steps of opening the control authority of the drone, establishing the control relationship between the distributed computing architecture and the drone, extracting the container snapshot in the fog layer after receiving the fault data sent by the drone, and restoring the snapshot in a pre-configured backup drone. The backup drone is redeployed using the restored data, and the backup drone is used as the controlled object to migrate the control relationship.

[0086] S301: Load the container snapshot and restore it, and redeploy the restored data to the backup drone through the control relationship.

[0087] Determine the control relationship between the distributed computing architecture and the drone, extract the container snapshot from the fog layer, and restore the snapshot. Using the control relationship, redeploy the restored data to the backup drone, and migrate the original control relationship established with the drone to the backup drone.

[0088] S302: Formulate and execute a flight plan, and transmit the data acquired by the backup UAV during the flight to the edge layer.

[0089] A flight plan for the backup drone is formulated in the distributed computing architecture, and the flight plan is sent to the backup drone. At the same time, the data generated by the backup drone during the flight is transmitted to the edge layer.

[0090] In Example 6, different from Example 1, in this embodiment of the present invention, the method further includes:

[0091] Initialize the properties of the edge device and configure the correspondence between tasks and edge devices;

[0092] By traversing the created localization decision set, the data to be uploaded is selected from the edge layer;

[0093] By utilizing the upload channel between the edge device and the cloud, the data to be uploaded is transferred to the processing model in the cloud;

[0094] Establishing a federated architecture for edge devices, backing up the processing model to the edge devices, and training the processing model using the federated architecture;

[0095] The parameters obtained during training are uploaded to the cloud and aggregated, and the processing model is updated using the aggregated data.

[0096] Determine the attributes of edge devices, namely their data processing capabilities and types, to establish a correspondence between unfinished task slices, completed task slices, and edge devices. Use edge devices to process both unfinished and completed task slices simultaneously. Based on the processing results, traverse the localized decision set, select the corresponding decision, and complete the processing of data in the fog layer. In reality, completed task slices also need to be processed to mine useful data.

[0097] The data that needs to be uploaded to the cloud for processing, that is, the data to be uploaded, is selected from the fog layer, and the data to be uploaded is transferred to the processing model on the cloud through the upload channel between the edge device and the cloud, and the uploaded data is analyzed and processed; a federal architecture is established through edge devices, the processing model is backed up to all edge devices, the processing model is trained using the data in the fog layer, and the variables of the parameters in the trained processing model are uploaded to the cloud. In the cloud, the parameters of the variables and the processing model in the cloud are aggregated, and the processing model in the cloud is updated based on the aggregation results.

[0098] In Example 7, different from Example 1, in this embodiment of the present invention, the method further includes:

[0099] Inserting time scheduling windows into the federated architecture and formulating asynchronous update strategies;

[0100] Insert version tags into the processing models on the edge device and the cloud, and determine the rollback steps of the processing models based on the preset rollback mechanism

[0101] Insert a time scheduling window into the federated architecture. The time scheduling window is mainly used to determine the training time of the processing model. The asynchronous update strategy allows different edge devices to train the processing model in different time periods. At the same time, a version tag is inserted into the trained processing model. That is to say, after each training of the processing model, a version tag is inserted and a rollback mechanism is set to roll back the processing model, so as to facilitate the restoration of the processing model and find the most stable processing model. The preset rollback mechanism is the method for restoring the processing model.

[0102] Figure 5 The figure shows a structural block diagram of a relay-based drone control system according to an embodiment of the present invention. The relay-based drone control system 1 includes:

[0103] Determination module 11, used to obtain the mission of the drone and determine the data transmission link between the drone and the edge device;

[0104] The slicing module 12 is used to build an edge layer using data from drones and a fog layer using data from edge devices, thereby constructing a distributed computing architecture. The slicing module 12 generates delimiters using the edge layer, slices the tasks, containerizes the slices obtained after slicing, generates container snapshots, and caches the container snapshots in the fog layer.

[0105] The recovery module 13 is used to open the control authority of the drone and establish the control relationship between the distributed computing architecture and the drone. After receiving the fault data sent by the drone, it extracts the container snapshot in the fog layer and performs snapshot recovery in the pre-configured backup drone. The backup drone is redeployed using the data obtained after recovery, and the control relationship is migrated with the backup drone as the controlled object.

[0106] Figure 6 The following is a structural block diagram of a relay-based drone control system according to an embodiment of the present invention. The determining module 11 includes:

[0107] A clustering unit 111 is used to collect location data of the area where the task is located and cluster the task into a flight part and an execution part;

[0108] A corresponding unit 112 is used to establish corresponding relationships between the flight part, the execution part and the position data respectively;

[0109] The debugging unit 113 is used to define a data transmission protocol, debug the data transmission link, and determine an optimization method based on the debugging result.

[0110] Figure 7 The structure block diagram of the relay-based drone control system provided by an embodiment of the present invention is shown. The segmentation module 12 includes:

[0111] An acquisition unit 121 is configured to perform data preprocessing in the edge layer, acquire valid data, and superimpose the created local decision mechanism into the edge layer;

[0112] The statistical unit 122 is configured to determine the number of edge devices in the fog layer, which number is at least one, and set a division of labor for all edge devices.

[0113] The separation unit 123 is used to generate a separator according to the progress of the task and divide the task into unfinished task slices and completed task slices;

[0114] The cache unit 124 is used to containerize the unfinished task slices, obtain container snapshots, and cache the container snapshots in the fog layer;

[0115] The correction unit 125 is configured to generate an incremental snapshot using the progress and to correct the container snapshot.

[0116] Figure 8 The figure shows a structural block diagram of a relay-based drone control system according to an embodiment of the present invention. The recovery module 13 includes:

[0117] A deployment unit 131 is configured to load the container snapshot, restore it, and redeploy the restored data to the backup drone through the control relationship;

[0118] The backup unit 132 is used to formulate and execute a flight plan and transmit the data of the backup UAV during flight to the edge layer.

[0119] The determination module 11 is used to complete step S100, the segmentation module 12 is used to complete step S200, and the recovery module 13 is used to complete step S300;

[0120] The clustering unit 111 is used to complete step S101, the corresponding unit 112 is used to complete step S102, and the debugging unit 113 is used to complete step S103;

[0121] The acquisition unit 121 is used to complete step S201, the counting unit 122 is used to complete step S202, the separation unit 123 is used to complete step S203, the cache unit 124 is used to complete step S204, and the correction unit 125 is used to complete step S205;

[0122] The deployment unit 131 is used to complete step S301 , and the backup unit 132 is used to complete step S302 .

[0123] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0124] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A drone control method based on a relay terminal, characterized in that: The method comprises: Obtain the drone's mission and determine the data transmission link between the drone and the edge device; The edge layer is built using data from drones, and the fog layer is built using data from edge devices, thereby constructing a distributed computing architecture. The edge layer is used to generate delimiters, and the tasks are segmented. The segmented slices are containerized, and container snapshots are generated. The container snapshots are then cached in the fog layer. The delimiter is used to split the task into unfinished task slices and completed task slices. The unfinished task slices and corresponding data are containerized, and the generated container snapshots are cached in the fog layer. Open up the control permissions of the drones and establish the control relationship between the distributed computing architecture and the drones. After receiving the fault data sent by the drones, extract the container snapshot in the fog layer and restore it in a pre-configured backup drone. Use the restored data to redeploy the backup drone, using it as the controlled object and migrating the control relationship. Formulate and execute flight plans, and transmit data acquired by backup drones during flight to the edge layer; The steps of generating delimiters by using the edge layer, segmenting the task, containerizing the segmented slices, generating container snapshots, and caching the container snapshots in the fog layer include: Generate a separator according to the progress of the task, and divide the task into unfinished task slices and completed task slices; Containerizing the unfinished task slices, obtaining container snapshots, and caching the container snapshots in the fog layer; Using the progress, generating an incremental snapshot, and revising the container snapshot; The method further comprises: Initialize the properties of the edge device and configure the corresponding relationship between the task and the edge device; By traversing the created localization decision set, the data to be uploaded is selected from the edge layer; By utilizing the upload channel between the edge device and the cloud, the data to be uploaded is transferred to the processing model in the cloud; Establishing a federated architecture for edge devices, backing up the processing model to the edge devices, and training the processing model using the federated architecture; The parameters obtained during training are uploaded to the cloud and aggregated, and the processing model is updated using the aggregated data.

Images