149 results about "Distributed collaboration" patented technology

A composite AI system and method for advanced reasoning and automation that integrates symbolic knowledge graphs and algorithms with non-symbolic, or connectionist, models such as neural embeddings. A hierarchical architecture enables dynamically distributed, cooperative reasoning through layperson and expert-led challenge-based verification, model blending, model fitness and retraining and selection, comprehensive feedback loops at individual model or model blend or process flow with or without supervision, and specialized routing of processing to account for various operational risk, regulatory, legal, privacy, or economic considerations. Models, datasets, knowledge bases, simulations and simulation components, and embeddings are iteratively refined using knowledge graph elements and model, process, simulation or flow / process optimal hyperparameters which are recorded and tracked. Extraction of symbolic representations from connectionist models links them to curated ontologies of facts and principles.

The invention provides a medical data privacy protection method and system based on artificial intelligence, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the feature analysis of medical data, building a scoring matrix, generating a privacy risk assessment result, dividing a security domain, and building a distributed cooperation mechanism through federal learning; preprocessing data at a local node, constructing a deep neural network model, encrypting gradient information by using a dynamic key matrix, and performing distributed aggregation operation and verification; forming a data privacy protection scheme; and performing identity verification and anomaly detection on the access request, and dynamically adjusting the access granularity. According to the method, the security and the flexibility of medical data privacy protection are improved, and meanwhile, the availability of the data is ensured.

The invention discloses a virtual power plant system design method based on multilevel federated learning, and relates to the technical field of power system planning, and the design method comprises the steps: designing and constructing a virtual power plant system architecture comprising an energy resource module, a local controller, a regional coordinator, a central coordinator and a regional block chain network. Intelligent prediction and optimal control of distributed energy resources are realized by using a deep learning framework and a model training algorithm, the intelligent level of the system is improved, automatic management of energy transaction and execution of an intelligent contract are realized by using a block chain technology, transparency and security of the transaction are ensured, and multi-level federated learning is adopted, so that the efficiency of the system is improved. According to the invention, distributed cooperation of data distributed processing and model training is realized, risks of data leakage and privacy invasion are effectively avoided, cooperative work of a local controller and a regional coordinator is cooperated, intelligent management and optimal scheduling of distributed energy resources are realized, and manual intervention and management cost are reduced.

The invention discloses a remote centralized control distributed data processing and analysis method based on a JMS, and the method comprises the steps: achieving the message routing and resource scheduling through the real-time monitoring of the state of a system, reinforcement of learning decision, deep learning optimization and multi-dimensional message distribution; dynamically adjusting message flow rate and queue length based on real-time network conditions, node load and flow fluctuation; through edge node intelligent task scheduling, distributed collaboration of calculation and storage is realized; adaptive compression algorithm selection is carried out by using deep learning, and the consumption of message transmission bandwidth is reduced; non-tampering storage of messages is realized through a block chain technology; based on a container arrangement technology, dynamic expansion and contraction of computing resources are realized; and dynamic resource scheduling and optimization decision are realized by utilizing reinforcement learning and multi-dimensional monitoring data. According to the invention, the bottleneck problem of the traditional message transmission system is overcome, and the system can efficiently and reliably process mass data in a high-concurrency and high-throughput scene.

A method and system for resolving a multi-operator distributed collaborative conflict of an unmanned air vehicle is disclosed, including: resolving a conflict in a strategic stage: generating multidimensional data based on time and spatial elements, rasterizing the multidimensional data, performing preliminary conflict resolution according to a preset airspace and an airspace requirement after rasterizing, and generating strategy approval information; and completing strategy conflict resolution by checking the strategy information; resolving a conflict in a pre-tactical stage: acquiring flight plan approval information, designating checking parties, performing multiple checks on the flight plan information, and performing judgment in combination with checking results of a plurality of checking parties to generate a conflict checking result; and resolving a conflict in a tactical stage: identifying a trajectory deviation according to real-time flight data, and sending out a warning according to a deviation category; and optimizing a path according to the deviated trajectory.

The invention discloses a large-span bridge structure health monitoring abnormity identification method based on image classification, and relates to the technical field of bridge structure health monitoring, and the method comprises the following steps: 1, generating abnormal samples through employing a generative adversarial network (GAN), and the abnormal samples comprise cracks, peeling, and corrosion; according to the large-span bridge structure health monitoring abnormity identification method based on image classification, abnormal samples are generated and optimized through the generative adversarial network, and the data scarcity condition is relieved; a dynamic generation-learning-optimization closed-loop system and a distributed cooperation platform reduce the labeling cost; by means of weak supervised learning, a knowledge graph and an automatic quality inspection tool, the labeling quality stability is improved; the utilization rate of label-free data is improved during multi-modal data fusion by utilizing a self-supervised learning technology, the problems in the aspects of data annotation and quality are solved comprehensively, and powerful support is provided for health monitoring of a bridge structure.

A reliable reasoning scheduling method based on an edge hybrid expert large model relates to the field of computers, and comprises the following steps: deploying a MoE large model to an edge network in a distributed edge manner; malicious node behavior modeling is carried out, distributed cooperative reasoning is simulated, and an auto-encoder input vector space set V and a reasoning task state set S are generated respectively; performing abnormal node detection neural network training to obtain calculation reliability of each edge node; performing deep reinforcement learning neural network training to obtain an expert selection strategy; and performing reliable reasoning scheduling of the edge MoE large model based on the trained deep reinforcement learning neural network. According to the method, the service quality and privacy protection advantages of distributed edge deployment are utilized, according to an original routing result of a large model gating function to an expert network, the node reliability and real-time load in an actual dynamic deployment environment are combined, the reasoning process is effectively optimized, and the reasoning efficiency is improved. Therefore, the cooperative reasoning efficiency and reliability of the MoE large model under distributed edge deployment are improved.

The invention relates to the technical field of lithium battery safety management, and discloses a lithium battery fault diagnosis method based on a BMS, and the method comprises the steps: obtaining multi-dimensional battery operation data and real-time data, firstly extracting a local feature vector, generating a preliminary fault signal indication, then extracting an abnormal feature vector according to the preliminary fault signal indication, and carrying out the fault diagnosis according to the abnormal feature vector; and if the preset threshold is exceeded, compressing and transmitting to the adjacent management unit to form a shared data packet. According to the local feature vector and the shared data packet, updating a diagnosis model parameter to obtain an optimized fault recognition model for analyzing real-time data and calculating a fault matching degree, and determining a potential fault type if a threshold value is exceeded; and generating a collaborative query request to the distributed network to obtain a historical fault empirical data set, integrating the historical fault empirical data set, refining parameters to obtain an accurate fault probability, activating an alarm and recording a log if a warning threshold is exceeded, and finally updating the global shared knowledge base. According to the method, the problem of insufficient lithium battery fault diagnosis accuracy in a distributed scene is solved, and collaborative optimization of diagnosis accuracy and distributed collaboration is realized.

The invention discloses a power equipment monitoring and early warning system based on intelligent patrol, and relates to the technical field of power system intelligent monitoring and safety early warning, the system fuses visible light, infrared, ultraviolet, acoustics, vibration and other multi-source signals, innovatively encodes multi-modal features in a complex manner, realizes high-speed and high-dimensional feature extraction through an optical convolution chip, and realizes high-precision and high-dimensional feature extraction. And the early perception capability of hidden dangers is greatly improved. The adaptive complex neural network is introduced, the model structure and parameter dynamic optimization, self-learning and self-evolution functions are achieved, the fault recognition accuracy is remarkably improved, and false alarm and missing alarm are reduced. The system supports cloud side end distributed collaboration, data real-time convergence and multi-scene visualization, has hierarchical deployment and online adaptive capabilities, and adapts to intelligent operation and maintenance requirements of multiple types of power stations. The system has the advantages of high performance, high intelligence, easiness in integration, expandability and the like, is prominent in innovativeness, and can be widely applied to intelligent monitoring and operation and maintenance in electric power and related fields.

The invention discloses an efficient task scheduling method based on distributed collaboration, and belongs to the technical field of distributed computing. Aiming at the problems of non-uniform resource allocation, insufficient task type difference adaptation, lack of multi-task dependence global optimization and the like existing in the existing scheduling strategy, the invention provides the following technical scheme: firstly, classifying tasks based on calculation characteristics and establishing a multi-dimensional resource index; secondly, performing dynamic weight evaluation on the heterogeneous resources by adopting an optimal worst weighting method (BWM), and constructing a task-resource matching matrix; a task dependency relationship is modeled through a directed acyclic graph (DAG), and a global priority sequence is generated in combination with a list scheduling algorithm; and finally, single-task optimal node matching is realized by adopting an elimination selection method, and cooperative scheduling is performed on multiple tasks by applying an arithmetic optimization algorithm (AOA). According to the method, accurate resource matching of a calculation-intensive task and a data-intensive task is realized through three technical dimensions of task feature perception, resource dynamic adaptation and dependency relationship collaborative optimization.

The invention provides a commodity service recommendation method and system, and belongs to the technical field of artificial intelligence, and the method comprises the steps: receiving a global recommendation model issued by a service terminal; processing the historical interaction behavior data set of the user in the client based on the global recommendation model to obtain an initial candidate recommendation set; a preset cue word template is filled with the initial candidate recommendation set and the historical interaction behavior data set, and recommended cue words are generated; inputting the recommendation prompt words into the large language model to obtain a reordering recommendation candidate set output by the large language model; and performing similarity verification on the initial candidate recommendation set and the reordering recommendation candidate set of each user, screening out unmatched recommendation items in the initial candidate recommendation set, and outputting a service recommendation result. According to the method, distributed collaboration of federal learning and semantic comprehension and generation capability of a large language model are combined, so that accurate commodity service recommendation is realized on the premise of protecting data privacy.

The invention discloses a multi-time-slot wireless resource allocation method based on multi-cell distributed cooperation, relates to the technical field of communication, and aims to solve the problem of cache dynamic reduction caused by the actual data service scale of a multi-cell cellular system in multi-time-slot transmission and continuous transmission in each scheduling period. The resource management problem of batch task multi-time-slot downloading under dense cellular network deployment is researched, the defect that an existing scheme only pays attention to a single time slot and an infinite cache hypothesis is overcome, and the characteristic that the amount of data to be transmitted is limited in actual application is considered; an original problem is converted into a series of online sub-problems with the maximum rate to decouple the multi-dimensional time coupling amount, then a multi-base-station parallel distributed algorithm is proposed to decompose a global calculation task to respective base stations, and the designed low-complexity algorithm can effectively reduce the transmission delay of the system. Therefore, the efficiency and the reliability of batch downloading of multi-user services are improved.

The invention relates to a privacy protection heterogeneous federal learning method based on split learning. According to the method, federated learning and split learning are combined, privacy protection on client data in distributed collaborative learning is enhanced by adopting a proper model decomposition strategy, and negative effects of data isomerism are reduced by combining a private group normalization layer into a convolutional neural network. The method comprises the following stages: an initialization stage, a model broadcasting stage, a model training updating stage, a model aggregation stage and a training stop judging stage. Compared with the prior art, the method has the advantages that high model prediction accuracy can be kept under the scene of high client data isomerism in the Internet of Things, the data leakage risk is reduced, and the calculation and communication cost of the client is reduced.

A composite AI system and method for advanced reasoning and automation that integrates symbolic knowledge graphs and algorithms with non-symbolic, or connectionist, models such as neural embeddings. A hierarchical architecture enables dynamically distributed, cooperative reasoning through layperson and expert-led challenge-based verification, model blending, model fitness and retraining and selection, comprehensive feedback loops at individual model or model blend or process flow with or without supervision, and specialized routing of processing to account for various operational risk, regulatory, legal, privacy, or economic considerations. Models, datasets, knowledge bases, simulations and simulation components, and embeddings are iteratively refined using knowledge graph elements and model, process, simulation or flow / process optimal hyperparameters which are recorded and tracked. Extraction of symbolic representations from connectionist models links them to curated ontologies of facts and principles.

The invention provides a new energy automobile in-vehicle network real-time intrusion detection system based on edge computing, which relates to the field of new energy automobile network security and intrusion detection and comprises an in-vehicle edge computing node, a vehicle-mounted communication interface and a regional collaboration module. The in-vehicle edge computing node, the vehicle-mounted communication interface and the regional collaboration module form a distributed collaboration architecture to realize in-vehicle network real-time intrusion detection and dynamic defense, and the in-vehicle edge computing node is deployed in a vehicle electronic control unit and a gateway, collects and analyzes network traffic and sensor data, executes a local defense decision, and sends the local defense decision to the regional collaboration module. The system monitors the controller local area network flow, the electronic control unit state and the sensor signal of the in-vehicle network in real time through edge calculation, network abnormity and physical attack signs can be found in time, the safety risk caused by delay detection is effectively avoided, and the efficiency and timeliness of in-vehicle network monitoring are improved. The system adopts intelligent analysis methods such as a topological immune reconstruction algorithm and a photon trajectory aggregation algorithm.

The invention discloses a network security early warning method and system based on distributed cooperation, and relates to the technical field of security early warning, and the method comprises the steps: deploying distributed collection nodes, collecting node data, carrying out the preprocessing, and extracting the static features and dynamic features of the data; and performing dimension reduction on the static features of the nodes, performing static risk assessment based on the static features after dimension reduction to obtain a static risk score, and generating a dynamic risk score based on the dynamic features. According to the method, accurate classification and feature extraction of static data and dynamic data are realized, optimization modeling is carried out for the two types of features, the problems of insufficient real-time performance and data islands in a traditional network security method are solved, comprehensive global risk assessment can be realized through weighted fusion of static risk scores and dynamic risk scores, and the risk assessment efficiency is improved. And in combination with an automatic early warning response mechanism, the efficiency and accuracy of network security monitoring are remarkably improved, so that complex security threats in a large-scale distributed network environment can be more effectively coped with.

The invention provides a load balancing algorithm system of a server cluster, which relates to the field of computer server clusters and comprises a priority management module, a resource monitoring module, a load distribution module, a cooperative communication module, a cost evaluation module and a large flow algorithm splitting and sequential execution module. The large-scale process algorithm splitting and sequential execution module comprises a process splitting unit, a sequential arrangement unit and a load balancing execution unit. The method has the beneficial effects that a large-scale process algorithm is split and sequenced, so that the method runs according to a specified sequence under the balance of a plurality of loads, and the complex service processing capability of the system is improved; and meanwhile, algorithms such as service priority, resource adaptation, distributed cooperation and cost benefit optimization are comprehensively applied, the resource utilization rate, the system performance and the cost benefit are improved, the load balancing effect of the server cluster is enhanced, a distributed cooperation mechanism avoids a single-point fault, local load changes can be quickly responded, and the stability and availability of the cluster are improved.

The present invention relates to the field of avionics equipment. Provided is an avionics fault knowledge fusion method oriented to distributed collaboration. The method comprises the following steps: S1, acquiring multi-source fault knowledge data of avionics equipment, and performing analysis processing and screening on the acquired multi-source fault knowledge data of the avionics equipment; S2, for the problem of knowledge conflicts across different timelines, comprehensively calculating machining fault knowledge on the basis of attribute similarity, in order to complete fault knowledge fusion; S3, for the problem of fault knowledge conflicts from different sources, completing fault knowledge fusion by means of fault element annotation and identification; and S4, constructing a fault knowledge graph, and storing fused fault knowledge. The present invention can mitigate knowledge conflicts during both the initial construction of a fault knowledge model and the expansion of existing fault knowledge models, thereby improving the accuracy, reliability and integrity of knowledge model construction for avionics information equipment, and thus providing better support for fault localization and maintenance of the avionics equipment.

The invention relates to the technical field of gateway performance optimization, and discloses a multi-terminal gateway connection performance optimization method and system.The method comprises the steps that SPI bus parameters of gateway hardware are set, SPI high-speed communication driving configuration is obtained, multi-channel frequency resource allocation is executed, and a plurality of channel configuration items are obtained; performing parallel spreading factor monitoring on the plurality of channel configuration items to obtain spreading factor dynamic adjustment data; monitoring the gateway network connection state according to the spreading factor dynamic adjustment data to obtain a network state monitoring result; and inputting the network state monitoring result into a distributed cooperation system for gateway cooperation group division to obtain a gateway resource allocation scheme, thereby realizing dynamic optimal allocation of resources between gateways, avoiding overload operation of a single gateway, and improving the signal coverage range and the transmission quality.

The invention discloses a digital twin control platform. The platform comprises a multi-protocol adaptive edge intelligent gateway; the dynamic twin modeling layer comprises an online transfer learning engine and a federal learning coordinator; the intelligent decision control layer is integrated with an MPC-DRL hybrid algorithm and an elastic resource scheduler; and the security interaction layer supports RLWE homomorphic encryption and block chain traceability. According to the edge intelligent gateway, through the wavelet denoising algorithm and the improved Huffman encoder, the 5G and TSN networks are combined, the multi-modal data acquisition and transmission efficiency is remarkably improved, the problem of processing lag of a traditional platform is solved, and the industrial real-time requirement is met. The dynamic twin modeling layer adopts a physical constraint loss function of an online transfer learning engine and a differential privacy mechanism of a federated learning coordinator to realize real-time updating of the model and distributed cooperative privacy guarantee, and overcomes the defects of static modeling and data leakage.

The invention discloses a multi-agent cooperative electric power inspection method and system, which realizes intelligent and efficient inspection task distribution and operation inspection task processing by introducing a task coordinator middleware to optimize a task distribution process of a contract network protocol and combining a priority task queue and a dynamic task scheduling strategy. The method provides help for distributed cooperation and efficient operation inspection application of a panoramic inspection intelligent agent and an equipment operation inspection intelligent agent of an electric power station, and solves the problems of low inspection efficiency of an existing electric power industrial station and low task allocation and resource utilization rate in a multi-agent electric power inspection method.

The invention discloses a heterogeneous federated learning method based on attention guidance aggregation and prototype enhancement, and relates to a heterogeneous federated learning method. The invention aims to solve the problem that the existing heterogeneous federal learning method cannot improve the granularity of client feature representation learning and cannot solve class imbalance in global head training. The method comprises the following steps: step 1, constructing a feature aggregation mechanism based on attention guidance; 2, designing a contrast learning objective function, and explicitly enhancing inter-class separability and intra-class compactness of feature representation in a local training process; 3, designing a self-adaptive prototype enhancement strategy, and relieving a class imbalance problem; and 4, constructing a collaborative optimization framework of the global model and the local model. The invention belongs to the technical field of distributed collaborative learning.

The invention belongs to the technical field of urban power grid safety control, and particularly relates to a source-load-storage modeling method and a power grid voltage control method and system. The method comprises the following steps: carrying out fine modeling of photovoltaic, energy storage and classified loads, collecting equipment operation state data and identifying parameters; building a centralized coordination-distributed coordination double-layer control architecture, deploying a centralized coordinator and a distributed coordinator, and defining functions; measuring topology, source load real-time and historical data of the power distribution network under the control architecture, and executing dynamic cluster division; based on initialization of a historical optimal solution database, an ADMM distributed optimization algorithm is adopted to calculate the cluster total power adjustment amount of each cluster; and based on the cluster total power adjustment amount, the voltage deviation factor and the resource margin factor, power distribution of the photovoltaic reactive adjustment amount and the energy storage active adjustment amount is carried out on equipment connected with the nodes in combination with the model and the model parameters. And due to refined modeling, a basis is provided for voltage control, and the accuracy and reliability of control can be improved.

The invention discloses a distributed collaborative waveform switching method and system oriented to networked computing, and mainly solves the problems of overlarge time delay and overhigh complexity when distributed collaborative waveform switching achieves the consistency of the whole network in a multi-hop ad hoc network environment in the prior art. The implementation scheme comprises a distributed collaboration stage and a distributed decision consistency stage. Wherein in the distributed collaboration stage, each node periodically exchanges a state packet and a collaboration data packet, and in the exchange process, distributed waveform decision making and time delay information updating are respectively executed in the nodes; in the distributed decision consistency stage, a unique central node is selected through the collaborative data packet stored in the distributed collaborative stage, and the central node performs majority decision through the internally stored waveform decision information of each node to make an optimal waveform switching decision. According to the method, the calculation complexity can be reduced, the decision timeliness and accuracy can be improved, the distributed consistency of the whole network can be ensured, the communication time delay can be remarkably reduced, the system performance can be improved, and the method can be used for networked calculation in a wireless ad hoc network.

The present invention relates to an active-standby switching method for clock asynchronization and a weak-network environment, and relates to the technical field of distributed systems. The method comprises: providing a distributed lock service of FIFO, and in each process, determining, on the basis of the distributed lock service, whether a lock has been obtained; after startup, a management node first waiting, by means of the distributed lock service, to obtain the lock, creating a local lease after obtaining the lock, loading lease information from a parent node of a temporary node in a distributed cooperative system, and writing the lease information into the parent node of the temporary node after waiting for a preset loading time interval, wherein after the lease information is successfully written, the management node is considered to become an active node; and before data is written, the management node first determining, by means of a data writing fence, whether a data writing condition is met, and generating a denial of service anomaly if the data writing condition is not met. The problem of active-standby switching only being completed depending on clock synchronization and a smooth network is solved.

To provide a data compression system that combines large-scale language models (LLMs) with distributed computing to solve the problems that conventional compression techniques only consider statistical redundancy and are unable to exploit semantic structure or contextual information, and furthermore, large-scale model-based compression requires high resources and fails to fully exploit the interrelationships in multimodal data.SOLUTION: The present invention achieves highly efficient compression even in resource-constrained environments, by utilizing the interrelationships between different data formats using a multimodal understanding module, selecting an optimal model according to the data characteristics and communication environment using an adaptive model selection module, utilizing shared knowledge between terminals via a distributed collaborative compression module, transferring the compression capabilities of large-scale models using a knowledge distillation module to lightweight models, and integrating efficient coding using an adaptive arithmetic coding module.SELECTED DRAWING: None

The invention discloses an end network information fusion detection method based on AI assistance. The end network information fusion detection method comprises the steps that S01, terminal data are collected, and equipment fingerprints are constructed; S02, a unique equipment ID is generated through a Hash algorithm, and ID preposition marks are used for distinguishing terminal types and establishing virtual identities for equipment; S03, multi-source information fusion is carried out; s04, carrying out abnormal feature evaluation through a multi-source data cross validation model, and constructing a dynamic defense network through a bee colony distributed cooperation mechanism, S06, carrying out information integration traceability tracking, and S07, carrying out causal chain reverse deduction; the method has the advantages that the updated multi-source data cross validation model is higher in application list, behavior log and flow feature anomaly detection capacity and wider in range, the updated multi-source data cross validation model is further applied to dynamic behavior modeling, known dangerous behavior data can be removed, and the method is suitable for dynamic behavior modeling. And repeated data accumulation is avoided, the data stock is reduced, and the problem of high maintenance and updating cost is solved.

The invention discloses an intelligent MIMO signal transmission method and system based on distributed cooperation, and the detection method comprises the steps: continuously employing different distributed units in an MIMO system to execute preset detection steps until the number of iterations reaches a preset number-of-iterations threshold L; sending the estimation vector subjected to the L times of iteration to a central processing unit, and quantizing the estimation vector by the central processing unit to obtain a final detection signal; the preset detection step comprises the following steps: processing a received radio frequency signal into a baseband signal; performing channel estimation by using the pilot signal to obtain a local channel matrix; based on a multi-step conditional random block Kaczmarz algorithm, updating an estimation vector by using a baseband signal and a local channel matrix, and selecting a next distributed unit based on a multi-step conditional sampling probability; and sending the updated estimation vector to the selected next distributed unit based on the topological structure of the MIMO system. According to the invention, not only can higher performance and lower complexity be realized, but also the required data transmission bandwidth is smaller.

This invention provides a method for reconstructing actuator faults in a heterogeneous multi-agent integrated energy system, relating to the field of fault diagnosis. The invention establishes a unified dynamic model for the agents; defines intermediate variables and constructs their dynamic equations; builds a global error dynamic system for distributed intermediate observers and agents; derives the stability criterion for the global error system based on Lyapunov stability theory, obtaining the state observer gain matrix for state estimation; loads the state observer gain matrix and the gain matrix for state estimation into each agent; and during the real-time operation of the agents, obtains the current state vector estimate and intermediate variable estimate, and calculates the reconstructed actuator fault value. This invention utilizes an output space coupling mechanism, enabling distributed collaboration between different energy networks without requiring a unified model dimension, thus breaking down the physical barriers to actuator fault reconstruction in integrated energy systems.

The application relates to a communication-optimized distributed cooperation method, and belongs to the field of machine learning. Firstly, the relative entropy is calculated based on the difference between the global model of the last round and the local model of the current round, and whether the current model layer is transmitted is judged according to the result of the relative entropy; then, different quantization levels are specified by the server under different rounds to realize time-adaptive quantization, and different quantization levels are continuously specified for different clients under the same round to realize client-adaptive quantization, so that the problem of excessively high communication cost in a bandwidth-limited environment is solved. In addition, on the basis of normal local training of the client, the situation of frequent accidental interruption in the transmission process is considered, and when the geographical position of the client continuously changes, the federated learning can still be normally trained.