CN115376031A - Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning - Google Patents

Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning Download PDF

Info

Publication number
CN115376031A
CN115376031A CN202211299471.8A CN202211299471A CN115376031A CN 115376031 A CN115376031 A CN 115376031A CN 202211299471 A CN202211299471 A CN 202211299471A CN 115376031 A CN115376031 A CN 115376031A
Authority
CN
China
Prior art keywords
learning model
federal learning
model parameters
federal
global
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202211299471.8A
Other languages
Chinese (zh)
Other versions
CN115376031B (en
Inventor
张安安
古和今
张苗辉
王志成
邓芳明
韦宝泉
曾晗
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
ENERGY RESEARCH INSTITUTE OF JIANGXI ACADEMY OF SCIENCES
Original Assignee
ENERGY RESEARCH INSTITUTE OF JIANGXI ACADEMY OF SCIENCES
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ENERGY RESEARCH INSTITUTE OF JIANGXI ACADEMY OF SCIENCES filed Critical ENERGY RESEARCH INSTITUTE OF JIANGXI ACADEMY OF SCIENCES
Priority to CN202211299471.8A priority Critical patent/CN115376031B/en
Publication of CN115376031A publication Critical patent/CN115376031A/en
Application granted granted Critical
Publication of CN115376031B publication Critical patent/CN115376031B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/17—Terrestrial scenes taken from planes or by drones
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Multimedia (AREA)
  • Remote Sensing (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a method for processing routing inspection data of a road unmanned aerial vehicle based on federal adaptive learning, wherein an edge server performs federal learning model training locally after receiving initial federal learning model parameters and road routing inspection images, and adaptively adjusts the iteration times of next round of training through an adaptive algorithm before each round of iteration starts in the training process.

Description

Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning
Technical Field
The invention relates to the technical field of information transmission, in particular to a highway unmanned aerial vehicle routing inspection data processing method based on federal self-adaptive learning.
Background
When the highway is patrolled and examined and maintained, artifical tour is easily influenced by bad weather or geographical environment, great operation degree of difficulty and risk have, artifical maintenance has the feedback untimely simultaneously, efficiency is lower, topography instrument restriction easily forms shortcomings such as patrol blind spot, along with the development of unmanned aerial vehicle technique, unmanned aerial vehicle has used in highway patrol and examine and maintain the work widely, utilize machine learning and deep learning technique to analyze the image, autonomic vision navigation tracks key information such as highway marking, make and patrol and examine efficiency and obtain huge promotion. However, with the development of the internet of things and cloud computing, the cloud storage data is huge, the storage capacity and the processing capacity of the cloud face huge challenges, and the cloud data also faces the risk of being attacked.
Federal Learning (Federal Learning) is used as a novel distributed machine Learning technology, and can be used for combining a plurality of scattered edge devices to cooperatively train a global model under the condition that each local server does not upload data, so that the model training efficiency can be effectively improved, the communication cost is reduced, and the data security is improved. However, in the prior art, gradient and model parameters need to be updated continuously in the federal learning process, cloud-edge communication is frequently performed, and meanwhile, the unmanned aerial vehicle and the edge server also have a communication process, so that huge burden is brought to communication overhead in the whole process, and the communication efficiency of the system is affected.
Disclosure of Invention
Therefore, the invention aims to provide a highway unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning so as to improve the communication efficiency.
A highway unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning comprises the following steps:
step 1, acquiring data and preprocessing the data by an unmanned aerial vehicle to obtain a road inspection image, and uploading the road inspection image to an edge server located at a ground station of the unmanned aerial vehicle;
step 2, the cloud server sends the initial federated learning model parameters to an edge server;
step 3, after receiving initial federal learning model parameters and road patrol images, the edge server performs federal learning model training locally, before each iteration starts in the training process, the edge server adaptively adjusts the iteration times of the next round of training through a self-adaptive algorithm, obtains federal learning model parameters after the training is finished, and uploads the federal learning model parameters to a cloud server;
step 4, the cloud server receives the Federal learning model parameters uploaded by the edge servers, and the Federal learning algorithm is used for aggregating the obtained Federal learning model parameters to obtain global Federal learning model parameters;
step 5, the cloud server judges whether the model precision of the global federal learning model parameters meets the preset requirement;
step 6, if the model precision of the global federal learning model parameters reaches the preset requirement, the cloud server issues the global federal learning model parameters to each edge server, and the process is ended;
and 7, if the model precision of the global federal learning model parameters does not meet the preset requirements, the cloud server issues the global federal learning model parameters to each edge server, the edge servers perform local federal learning model training again according to the received global federal learning model parameters and road patrol images to obtain updated federal learning model parameters, the updated federal learning model parameters are uploaded to the cloud server, the cloud server receives the updated federal learning model parameters uploaded by the edge servers, the updated federal learning model parameters are aggregated by using a federal learning algorithm to obtain updated global federal learning model parameters, the cloud server judges whether the model precision of the updated global federal learning model parameters meets the preset requirements or not, if so, the updated global federal learning model parameters are issued to each edge server, and the process is ended.
According to the highway unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning, in step 3, when the edge server conducts federal learning model training locally, the parameters of the federal learning model need to be updated, and an updating formula is represented as follows:
Figure 613692DEST_PATH_IMAGE001
wherein the content of the first and second substances,
Figure 649781DEST_PATH_IMAGE002
represents the Federal learning model parameters, W, of the kth edge Server at the t-th iteration t-1 Representing the global federated learning model parameters at the t-1 iteration,
Figure 807093DEST_PATH_IMAGE003
it is indicated that the learning rate is,
Figure 600736DEST_PATH_IMAGE004
indicating the gradient descent update amount of the k-th edge server.
According to the highway unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning, in step 4, the cloud server adopts the following formula to aggregate the acquired federal learning model parameters:
Figure 704959DEST_PATH_IMAGE005
wherein, W t Representing the global Federal learning model parameter, n, at the t-th iteration k Denotes the amount of data of the K-th edge server, N denotes the total amount of data of all edge servers, and K denotes the total number of edge servers.
The road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning comprises the following steps of (1) determining the iteration times of next round of training through the following formula:
Figure 392292DEST_PATH_IMAGE006
Figure 595871DEST_PATH_IMAGE007
wherein, epoch t+1 Represents the number of iterations, epoch, of the t +1 th round of training t Represents the iteration number of the t-th round training t+1 Indicating the number of iterations that the t +1 th round is increased than the t th round, Δ e t Representing the number of iterations increased for round t over round t-1, E representing the energy consumption of the edge server,
Figure 939128DEST_PATH_IMAGE008
、
Figure 909358DEST_PATH_IMAGE009
indicating the set adaptive parameters of the mobile station,
Figure 654460DEST_PATH_IMAGE010
representing the increased model accuracy of the global federal learning model parameters for round t over round t-1,
Figure 294520DEST_PATH_IMAGE011
model accuracy of the global federal learned model parameters representing an increase in the t-1 th run over the t-2 th run.
The road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning comprises the following steps of:
Figure 125073DEST_PATH_IMAGE012
Figure 633415DEST_PATH_IMAGE013
wherein E is n Represents the nth edge server energy consumption, γ n Effective switched capacitor, C, representing the nth edge server n Representing the number of cycles of CPU operation of the local computation task of each round of training of the nth edge server, B n Representing the amount of task data calculated locally by the nth edge server, f n And the CPU main frequency of the nth edge server is shown.
According to the method for processing the inspection data of the unmanned highway vehicle based on the federal adaptive learning, after an edge server receives initial federal learning model parameters and highway inspection images, federal learning model training is carried out locally, in the training process, before each iteration is started, the iteration times of the next round of training are adjusted adaptively through an adaptive algorithm, if the model precision of the global federal learning model parameters does not meet the preset requirement, a cloud server issues the global federal learning model parameters to each edge server, the edge server carries out the federal learning model training again to obtain updated federal learning model parameters, and through the process, more calculation tasks can be distributed to the edge server, so that a better balance state between the local training cost and the global communication efficiency is achieved, the number of global communication rounds is reduced, and the communication efficiency is improved effectively.
Drawings
Fig. 1 is a flowchart of a method for processing data of routing inspection by a road unmanned aerial vehicle based on federal adaptive learning according to an embodiment of the present invention;
fig. 2 is a comparison chart of training results of the road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning and the traditional federal average algorithm provided by the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, a method for processing road unmanned aerial vehicle routing inspection data based on federal adaptive learning according to an embodiment of the present invention includes the following steps 1 to 7:
step 1, data acquisition and data preprocessing are carried out through an unmanned aerial vehicle, a road inspection image is obtained, and the road inspection image is uploaded to an edge server located at a ground station of the unmanned aerial vehicle.
And 2, the cloud server sends the initial federal learning model parameters to the edge server.
And 3, after receiving the initial federal learning model parameters and the road patrol images, the edge server locally performs the federal learning model training, adaptively adjusts the iteration times of the next round of training through a self-adaptive algorithm before each round of iteration starts in the training process, obtains the federal learning model parameters after the training is finished, and uploads the federal learning model parameters to the cloud server.
When the edge server carries out the federal learning model training locally, the parameters of the federal learning model need to be updated, and the updating formula is expressed as follows:
Figure 902197DEST_PATH_IMAGE014
wherein the content of the first and second substances,
Figure 775475DEST_PATH_IMAGE002
represents the Federal learning model parameters, W, of the kth edge Server at the t-th iteration t-1 Representing the global federated learning model parameters at the t-1 th iteration,
Figure 155641DEST_PATH_IMAGE003
it is indicated that the learning rate is,
Figure 139778DEST_PATH_IMAGE015
indicating the gradient descent update amount of the k-th edge server.
In this embodiment, the iteration number of the next round of training is determined by the following formula:
Figure 531576DEST_PATH_IMAGE016
Figure 638072DEST_PATH_IMAGE007
wherein, epoch t+1 Represents the number of iterations, epoch, of the t +1 th round of training t Shows the number of iterations of the t-round training t+1 Indicating the number of iterations that the t +1 th round is increased than the t th round, Δ e t Representing the number of iterations increased for round t over round t-1, E representing the energy consumption of the edge server,
Figure 443217DEST_PATH_IMAGE008
、
Figure 168728DEST_PATH_IMAGE009
indicating the set adaptive parameters of the mobile station,
Figure 742928DEST_PATH_IMAGE017
representing the increased model accuracy of the global federal learning model parameters for round t over round t-1,
Figure 754747DEST_PATH_IMAGE011
model accuracy of the global federal learned model parameters representing an increase in the t-1 th run over the t-2 th run.
In the process of federal learning training, the energy consumption of each training round is different due to the difference of data volume and CPU of the edge servers participating in the training, so the total energy consumption E calculated locally by all the edge servers participating in the training is used as an index for measuring self-adaptation. Specifically, the calculation formula of the total energy consumption E of the edge server is as follows:
Figure 719292DEST_PATH_IMAGE012
Figure 310810DEST_PATH_IMAGE018
wherein E is n Represents the nth edge server energy consumption, γ n Effective switched capacitor, C, representing the nth edge server n Representing the number of cycles that the CPU of the local computation task of the nth edge server runs for each round of training (i.e. the complexity of the computation task of the nth edge server), B n Representing the amount of task data calculated locally by the nth edge server, f n Representing the CPU dominant frequency of the nth edge server (i.e., the computing power of the nth edge server).
Increasing the computational load of the local edge servers is affected by two factors: and the promotion of the promotion amplitude of the model precision to the increment of the calculated quantity and the negative feedback of the energy consumption of the edge server to the increment of the calculated quantity. By these two factors, it can be determined how many computing tasks each edge server should be assigned in the next round during each round of global iteration.
And 4, the cloud server receives the federal learning model parameters uploaded by the edge servers, and the acquired federal learning model parameters are aggregated by using a federal learning algorithm to acquire global federal learning model parameters.
The cloud server adopts the following formula to aggregate the acquired Federal learning model parameters:
Figure 801834DEST_PATH_IMAGE019
wherein, W t Representing the global Federal learning model parameter, n, at the t-th iteration k Denotes the amount of data of the K-th edge server, N denotes the total amount of data of all edge servers, and K denotes the total number of edge servers.
And 5, judging whether the model precision of the global federated learning model parameters meets the preset requirement or not by the cloud server.
And 6, if the model precision of the global federal learning model parameters meets the preset requirement, the cloud server issues the global federal learning model parameters to each edge server, and the process is ended.
And 7, if the model precision of the global federal learning model parameters does not meet the preset requirements, the cloud server issues the global federal learning model parameters to each edge server, the edge servers perform federal learning model training again locally according to the received global federal learning model parameters and the road inspection images to obtain updated federal learning model parameters, the updated federal learning model parameters are uploaded to the cloud server, the cloud server receives the updated federal learning model parameters uploaded by the edge servers and aggregates the obtained updated federal learning model parameters by using a federal learning algorithm to obtain the updated global federal learning model parameters, the cloud server judges whether the model precision of the updated global federal learning model parameters meets the preset requirements, if so, the updated global federal learning model parameters are issued to each edge server, and the process is ended. It can be understood that, if the cloud server determines that the model precision of the updated global federal learning model parameter does not meet the preset requirement, the above process is repeated until the model precision of the global federal learning model parameter meets the preset requirement.
Referring to fig. 2, the method for processing the inspection data of the unmanned road vehicle based on federal adaptive learning according to the present invention is compared with the training result of the conventional federal average algorithm.
As can be seen from FIG. 2, the final model accuracy of both methods stays at about 88%, and the model training efficiency is mainly reflected in the time of each training round and the number of iterations required to reach the target accuracy.
Using the traditional federal averaging algorithm requires 25 global iterations to achieve a model accuracy of nearly 80%, whereas using the invention of the present invention only requires 15 rounds to achieve the same effect for the global model. Similarly, the present invention can automatically adjust the sample size before each training cycle, and thus the training time is shortened by 10%. In contrast, the method of the present invention improves model training efficiency by about 40% over the traditional federal averaging algorithm.
In summary, according to the method for processing the inspection data of the unmanned highway vehicle based on the federal adaptive learning provided by the invention, after the edge server receives the initial federal learning model parameters and the highway inspection image, the federal learning model training is performed locally, in the training process, before each iteration starts, the iteration times of the next round of training are adjusted adaptively through an adaptive algorithm, and if the model precision of the global federal learning model parameters does not meet the preset requirement, the cloud server issues the global federal learning model parameters to each edge server, the edge server performs the federal learning model training again to obtain updated federal learning model parameters, through the process, more calculation tasks can be distributed to the edge server, so that a better balance state between the local training cost and the global communication efficiency is achieved, the number of global communication rounds is reduced, and the communication efficiency is improved effectively.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following technologies, which are well known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims (5)

1. A highway unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning is characterized by comprising the following steps:
step 1, acquiring data and preprocessing the data by an unmanned aerial vehicle to obtain a road inspection image, and uploading the road inspection image to an edge server located at a ground station of the unmanned aerial vehicle;
step 2, the cloud server sends the initial federal learning model parameters to an edge server;
step 3, after receiving initial federal learning model parameters and road patrol images, the edge server locally performs federal learning model training, adaptively adjusts the iteration times of the next training round through a self-adaptive algorithm before each iteration round starts in the training process, obtains the federal learning model parameters after the training is finished, and uploads the federal learning model parameters to the cloud server;
step 4, the cloud server receives the federal learning model parameters uploaded by the edge servers, and the acquired federal learning model parameters are aggregated by using a federal learning algorithm to acquire global federal learning model parameters;
step 5, the cloud server judges whether the model precision of the global federal learning model parameters meets the preset requirement;
step 6, if the model precision of the global federal learning model parameters reaches the preset requirement, the cloud server issues the global federal learning model parameters to each edge server, and the process is ended;
and 7, if the model precision of the global federal learning model parameters does not meet the preset requirements, the cloud server issues the global federal learning model parameters to each edge server, the edge servers perform local federal learning model training again according to the received global federal learning model parameters and road patrol images to obtain updated federal learning model parameters, the updated federal learning model parameters are uploaded to the cloud server, the cloud server receives the updated federal learning model parameters uploaded by the edge servers, the updated federal learning model parameters are aggregated by using a federal learning algorithm to obtain updated global federal learning model parameters, the cloud server judges whether the model precision of the updated global federal learning model parameters meets the preset requirements or not, if so, the updated global federal learning model parameters are issued to each edge server, and the process is ended.
2. The method for processing the patrol data of the unmanned aerial vehicle on the highway based on the federal adaptive learning of claim 1, wherein in the step 3, when the edge server carries out the federal learning model training locally, the parameters of the federal learning model need to be updated, and the updating formula is represented as follows:
Figure 359208DEST_PATH_IMAGE001
wherein, the first and the second end of the pipe are connected with each other,
Figure 45404DEST_PATH_IMAGE002
representing the Federal learning model parameters, W, of the k-th edge server at the t-th iteration t-1 Representing the global federated learning model parameters at the t-1 th iteration,
Figure 783553DEST_PATH_IMAGE003
it is indicated that the learning rate is,
Figure 162582DEST_PATH_IMAGE004
indicating the gradient descent update amount of the k-th edge server.
3. The method for processing the inspection data of the unmanned aerial vehicle on the highway based on the federal adaptive learning of claim 2, wherein in the step 4, the cloud server adopts the following formula to aggregate the acquired federal learning model parameters:
Figure 228758DEST_PATH_IMAGE005
wherein, W t Representing the global Federal learning model parameter, n, at the t-th iteration k Represents the amount of data of the K-th edge server, N represents the total amount of data of all edge servers, and K represents the total number of edge servers.
4. A road unmanned aerial vehicle inspection data processing method based on federal adaptive learning according to claim 3, wherein in step 3, the iteration number of the next round of training is determined by the following formula:
Figure 718645DEST_PATH_IMAGE006
Figure 311300DEST_PATH_IMAGE007
wherein, epoch t+1 Represents the number of iterations, epoch, of the t +1 th round of training t Represents the iteration number of the t-th round training t+1 Indicating the number of iterations that the t +1 th round is increased than the t th round, Δ e t Indicating the number of iterations increased for round t over round t-1, E indicating the total energy consumption of the edge server,
Figure 736597DEST_PATH_IMAGE008
、
Figure 414703DEST_PATH_IMAGE009
the adaptive parameters that are indicative of the settings are,
Figure 708281DEST_PATH_IMAGE010
representing the increased model accuracy of the global federal learning model parameters for round t over round t-1,
Figure 155443DEST_PATH_IMAGE011
representing the increased model accuracy of the global federal learned model parameters for round t-1 over round t-2.
5. The road unmanned aerial vehicle inspection data processing method based on federal adaptive learning of claim 4, wherein the calculation formula of the total energy consumption E of the edge server is as follows:
Figure 486061DEST_PATH_IMAGE012
Figure 651463DEST_PATH_IMAGE013
wherein E is n Represents the nth edge server energy consumption, γ n Effective switched capacitor, C, representing the nth edge server n Representing the number of cycles of CPU operation of the local computation task of each round of training of the nth edge server, B n Representing the amount of task data calculated locally by the nth edge server, f n Representing the CPU master frequency of the nth edge server.
CN202211299471.8A 2022-10-24 2022-10-24 Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning Active CN115376031B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211299471.8A CN115376031B (en) 2022-10-24 2022-10-24 Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211299471.8A CN115376031B (en) 2022-10-24 2022-10-24 Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning

Publications (2)

Publication Number Publication Date
CN115376031A true CN115376031A (en) 2022-11-22
CN115376031B CN115376031B (en) 2023-03-24

Family

ID=84073451

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211299471.8A Active CN115376031B (en) 2022-10-24 2022-10-24 Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning

Country Status (1)

Country Link
CN (1) CN115376031B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116506469A (en) * 2023-06-25 2023-07-28 华东交通大学 Bridge state monitoring method based on vibration energy-taking RFID sensor and federal learning
CN116582871A (en) * 2023-07-07 2023-08-11 南京邮电大学 Unmanned aerial vehicle cluster federal learning model optimization method based on topology optimization
CN116778363A (en) * 2023-06-25 2023-09-19 河海大学 Low-traffic reservoir area water environment risk identification method based on federal learning

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200027033A1 (en) * 2018-07-19 2020-01-23 Adobe Inc. Updating Machine Learning Models On Edge Servers
CN112181666A (en) * 2020-10-26 2021-01-05 华侨大学 Method, system, equipment and readable storage medium for equipment evaluation and federal learning importance aggregation based on edge intelligence
CN112532451A (en) * 2020-11-30 2021-03-19 安徽工业大学 Layered federal learning method and device based on asynchronous communication, terminal equipment and storage medium
CN112836828A (en) * 2020-12-16 2021-05-25 中山大学 Self-organizing federal learning method based on game theory
CN113052331A (en) * 2021-02-19 2021-06-29 北京航空航天大学 Block chain-based Internet of things personalized federal learning method
CN114116198A (en) * 2021-10-21 2022-03-01 西安电子科技大学 Asynchronous federal learning method, system, equipment and terminal for mobile vehicle
CN114880123A (en) * 2022-05-20 2022-08-09 苏州中桦科技有限公司 Federal learning method and system for industrial Internet of things communication control
CN115049071A (en) * 2022-06-10 2022-09-13 重庆邮电大学 Federal learning method based on DAG block chain

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200027033A1 (en) * 2018-07-19 2020-01-23 Adobe Inc. Updating Machine Learning Models On Edge Servers
CN112181666A (en) * 2020-10-26 2021-01-05 华侨大学 Method, system, equipment and readable storage medium for equipment evaluation and federal learning importance aggregation based on edge intelligence
CN112532451A (en) * 2020-11-30 2021-03-19 安徽工业大学 Layered federal learning method and device based on asynchronous communication, terminal equipment and storage medium
CN112836828A (en) * 2020-12-16 2021-05-25 中山大学 Self-organizing federal learning method based on game theory
CN113052331A (en) * 2021-02-19 2021-06-29 北京航空航天大学 Block chain-based Internet of things personalized federal learning method
CN114116198A (en) * 2021-10-21 2022-03-01 西安电子科技大学 Asynchronous federal learning method, system, equipment and terminal for mobile vehicle
CN114880123A (en) * 2022-05-20 2022-08-09 苏州中桦科技有限公司 Federal learning method and system for industrial Internet of things communication control
CN115049071A (en) * 2022-06-10 2022-09-13 重庆邮电大学 Federal learning method based on DAG block chain

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
YASAMAN AMANNEJAD: "Building and Evaluating Federated Models for Edge Computing", 《2020 16TH INTERNATIONAL CONFERENCE ON NETWORK AND SERVICE MANAGEMENT (CNSM)》 *
董业等: "基于秘密分享和梯度选择的高效安全联邦学习", 《计算机研究与发展》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116506469A (en) * 2023-06-25 2023-07-28 华东交通大学 Bridge state monitoring method based on vibration energy-taking RFID sensor and federal learning
CN116506469B (en) * 2023-06-25 2023-09-05 华东交通大学 Bridge state monitoring method based on vibration energy-taking RFID sensor and federal learning
CN116778363A (en) * 2023-06-25 2023-09-19 河海大学 Low-traffic reservoir area water environment risk identification method based on federal learning
CN116778363B (en) * 2023-06-25 2024-04-30 河海大学 Low-traffic reservoir area water environment risk identification method based on federal learning
CN116582871A (en) * 2023-07-07 2023-08-11 南京邮电大学 Unmanned aerial vehicle cluster federal learning model optimization method based on topology optimization
CN116582871B (en) * 2023-07-07 2023-10-13 南京邮电大学 Unmanned aerial vehicle cluster federal learning model optimization method based on topology optimization

Also Published As

Publication number Publication date
CN115376031B (en) 2023-03-24

Similar Documents

Publication Publication Date Title
CN115376031B (en) Road unmanned aerial vehicle routing inspection data processing method based on federal adaptive learning
CN113780395B (en) Mass high-dimensional AIS trajectory data clustering method
CN112181666A (en) Method, system, equipment and readable storage medium for equipment evaluation and federal learning importance aggregation based on edge intelligence
CN110807781B (en) Point cloud simplifying method for retaining details and boundary characteristics
CN113344956B (en) Ground feature contour extraction and classification method based on unmanned aerial vehicle aerial photography three-dimensional modeling
CN114332291A (en) Oblique photography model building outer contour rule extraction method
CN105139374A (en) Power transmission line tracking method and system
CN110879927A (en) Sea clutter amplitude statistical distribution field modeling method for sea target detection
CN111707997B (en) Radar target tracking method and device, electronic equipment and storage medium
CN116112981B (en) Unmanned aerial vehicle task unloading method based on edge calculation
CN108537322A (en) Neural network interlayer activation value quantization method and device
CN112419345A (en) Patrol car high-precision tracking method based on echo state network
CN109934151B (en) Face detection method based on movidius computing chip and Yolo face
CN115619953A (en) Rugged terrain-oriented mobile robot terrain mapping method and system
CN112762824B (en) Unmanned vehicle positioning method and system
Gu et al. Dynamic Allan variance analysis method with time-variant window length based on fuzzy control
CN114070438A (en) Detection method and system for 5G base station electromagnetic signals
CN114741806A (en) Suspension optimization method, system, device, equipment and medium
CN112229412B (en) Lane positioning method and device, storage medium and server
CN114281808A (en) Traffic big data cleaning method, device, equipment and readable storage medium
CN111768493A (en) Point cloud processing method based on distribution parameter coding
CN117291443B (en) Intelligent paying-off system based on multidimensional sensing technology
Yang et al. Real-time terrain estimation based on multi-scale radial basis function for unmanned ground vehicle
CN111627056B (en) Driving visibility determination method and device based on depth estimation
CN115688004B (en) Target attribute determining method, medium and device based on Hilbert coding

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant