Sensor data processing method and apparatus based on cloud computing center

Abstract

The application discloses a kind of sensor data processing method and device based on cloud computing center, it is related to sensor technical field.The method comprises: in cloud computing center, using artificial intelligence algorithm, construct sensor data processing engine, and based on layered mechanism, construct distributed storage network and data processing network;In edge gateway, the pre-processing of the sensor data collected is carried out, and the sensor data after pre-processing is uploaded to cloud computing center by encryption channel;In cloud computing center, using sensor data processing engine, based on data processing network, the pre-processing of the sensor data is carried out, and based on distributed storage network, the sensor data after processing obtained is distributedly stored.The problems that the existing technology exists, such as centralized processing bottleneck, poor processing model generalization ability, low efficiency of task scheduling and resource allocation, and insufficient data security and reliability are solved.

Description

Technical Field

[0001] This invention relates to the field of sensor technology, and in particular to a sensor data processing method and apparatus based on a cloud computing center. Background Technology

[0002] With the rapid development of sensor technology, sensors have been widely used in various fields such as smart cities, industrial internet, and intelligent transportation, generating massive amounts of multi-source, heterogeneous sensor data. How to efficiently, intelligently, and securely process this data and extract valuable information from it is a key challenge we currently face.

[0003] Traditional sensor data processing methods typically have the following shortcomings:

[0004] 1) Centralized processing bottleneck: Transmitting all raw data to a single central server for processing can easily lead to network bandwidth congestion and excessive computing load on the central server, creating a performance bottleneck and making it difficult to meet the real-time processing needs of massive amounts of data.

[0005] 2) Poor generalization ability of processing models: Traditional data processing models are usually designed for specific scenarios. When the application scenario changes (such as switching from industrial monitoring to environmental monitoring), the model needs to be retrained and deployed, lacking cross-scenario adaptability and generalization ability.

[0006] 3) Inefficient task scheduling and resource allocation: In a cloud computing environment, static or simple task scheduling and resource allocation strategies cannot optimize the allocation of computing resources in the face of dynamically changing task loads, resulting in low resource utilization and large task processing delays.

[0007] 4) Insufficient data security and reliability: Sensor data faces the risk of being stolen or tampered with during transmission and storage. Furthermore, centralized storage methods have a single point of failure risk; if the central storage node fails, data may be permanently lost. Summary of the Invention

[0008] This invention provides a sensor data processing method and apparatus based on a cloud computing center, which solves the problems of centralized processing bottleneck, poor generalization ability of processing models, low efficiency of task scheduling and resource allocation, and insufficient data security and reliability in existing technologies.

[0009] In a first aspect, embodiments of the present invention provide a sensor data processing method based on a cloud computing center, the method comprising:

[0010] In the cloud computing center, artificial intelligence algorithms are used to build a sensor data processing engine, and a distributed storage network and data processing network are built based on a hierarchical mechanism.

[0011] At the edge gateway, the collected sensor data is preprocessed and then uploaded to the cloud computing center through an encrypted channel.

[0012] In the cloud computing center, a sensor data processing engine is used to process the pre-processed sensor data based on a data processing network, and the processed sensor data is distributed and stored based on a distributed storage network.

[0013] The technical solution provided in this application has at least the following beneficial effects:

[0014] Data preprocessing via edge gateways effectively filters invalid data, reducing network transmission and computing pressure on the cloud computing center. A layered data processing network enables parallel task processing, significantly improving data processing efficiency. A federated learning mechanism is used to train a sensor data processing engine integrating multiple advanced AI models across different scenarios, giving the engine powerful generalization and adaptive capabilities to flexibly meet the data processing needs of various application scenarios. A dynamic task scheduling model intelligently decides task allocation based on real-time node status and task characteristics, while a cloud computing resource allocation model performs multi-objective optimization of cloud computing resources. These two mechanisms work together to maximize resource utilization and minimize task processing latency. A layered distributed storage network and a replication mechanism ensure data redundancy and high availability, avoiding single points of failure. Data transmission at the edge and in the cloud uses encrypted channels, guaranteeing data transmission security. Distributed ledger technology ensures the immutability and traceability of storage address information.

[0015] In one alternative implementation, an artificial intelligence algorithm is used to build a sensor data processing engine in a cloud computing center, and a hierarchical mechanism is used to build a distributed storage network and a data processing network, including:

[0016] In the cloud computing center, based on cloud datasets, we use deep learning algorithms to build initial data analysis and data processing models, reinforcement learning algorithms to build initial dynamic task scheduling models, and optimization algorithms to build initial cloud computing resource allocation models.

[0017] By combining the initial data analysis model, the initial data processing model, the initial dynamic task scheduling, and the initial cloud computing resource allocation model, an initial sensor data processing engine is obtained.

[0018] Based on the federated learning mechanism, the initial sensor data processing engine is trained in a federated learning architecture consisting of several edge servers in different scenarios to obtain the final sensor data processing engine.

[0019] Based on a hierarchical mechanism, several nodes with better storage performance are distributed and connected as storage nodes to obtain a distributed storage network, and several nodes with better cloud computing performance are distributed as cloud computing nodes to obtain a data processing network.

[0020] In one alternative implementation, the data analysis model is constructed based on the LSTM-CNN-Attention-MLP algorithm, and the data analysis model includes a time-series feature extraction module constructed based on the LSTM algorithm and an image feature extraction module constructed based on the CNN algorithm in parallel, as well as a first weighted fusion module constructed based on the Attention mechanism and a data analysis module constructed based on the MLP algorithm connected in sequence.

[0021] The data processing model is built on the Attention-DBN-AE algorithm and includes a second weighted fusion module built on the Attention mechanism, as well as a data prediction module built on the DBN algorithm and a data aggregation module built on the AE algorithm in parallel.

[0022] The dynamic task scheduling model is built based on the MPO-PPO algorithm, and includes an agent adjustment module built based on the MPO algorithm and a dynamic task scheduling module built based on the PPO algorithm. The dynamic task scheduling module includes an agent and an experience replay pool.

[0023] The cloud computing resource allocation model is built based on the ISC algorithm.

[0024] In one optional implementation, based on a federated learning mechanism, the initial sensor data processing engine is trained on various scenarios within a federated learning architecture consisting of several edge servers from different scenarios to obtain the final sensor data processing engine, including:

[0025] Extract the metadata of the initial sensor data processing engine and send it to several edge servers in different scenarios in the federated learning architecture through an encrypted channel. On each edge server, the initial sensor data processing engine is reconstructed based on the metadata.

[0026] Based on the scene dataset of each edge server, the sensor data processing engine is trained locally, and the model update amount and scene data amount of the locally trained sensor data processing engine are extracted and uploaded to the cloud computing center.

[0027] In the cloud computing center, based on the federated learning mechanism, the initial sensor data processing engine is adjusted according to the model update volume and scene data volume of several edge servers in different scenarios to obtain the final sensor data processing engine.

[0028] In one alternative implementation, at a cloud computing center, a sensor data processing engine is used to process the preprocessed sensor data based on a data processing network, and the processed sensor data is distributed and stored based on a distributed storage network, including:

[0029] In the cloud computing center, the data analysis model of the sensor data processing engine is used to perform data analysis on the preprocessed sensor data to obtain data analysis results.

[0030] If the data analysis results show anomalies, the preprocessed sensor data is marked and isolated, and data processing ends; otherwise, proceed to the next step.

[0031] The dynamic task scheduling model of the sensor data processing engine is used to retrieve the node running status of the data processing network, and dynamic task scheduling is performed based on the node running status and data analysis results to obtain the dynamic task scheduling strategy.

[0032] Based on the dynamic task scheduling strategy, cloud computing resource allocation model is used to allocate cloud computing resources and obtain a cloud computing resource allocation scheme.

[0033] According to the dynamic task scheduling strategy, several target cloud computing nodes are scheduled in the data processing network, and corresponding cloud computing resources are allocated to several target cloud computing nodes according to the cloud computing resource allocation scheme.

[0034] Based on several target cloud computing nodes with cloud computing resources, the data processing model of the sensor data processing engine is called to process the preprocessed sensor data and obtain the processed sensor data.

[0035] Based on the replication mechanism, the processed sensor data is divided into several data fragments including replicas, and the data fragments are sent to the distributed storage network.

[0036] Each storage node in the distributed storage network stores several data fragments locally and updates the storage addresses to the distributed ledger of the distributed storage network.

[0037] In one alternative implementation, at a cloud computing center, a data analysis model using a sensor data processing engine is used to perform data analysis on the preprocessed sensor data, obtaining data analysis results, including:

[0038] In the cloud computing center, the preprocessed sensor data is input into the data analysis model of the sensor data processing engine;

[0039] The temporal feature extraction module of the data analysis model is used to extract the temporal features of the temporal format data in the preprocessed sensor data;

[0040] The image feature extraction module of the data analysis model is used to extract image features from the image format data in the preprocessed sensor data.

[0041] Based on the dynamically generated first attention weight, the first weighted fusion module of the data analysis model is used to perform weighted fusion of temporal features and image features to obtain the first weighted fusion feature.

[0042] Based on the first weighted fusion feature, the data analysis module of the data analysis model is used to perform data analysis and obtain data analysis results. The data analysis results include scene type, anomaly detection results, tasks to be processed, and task priorities.

[0043] In one optional implementation, the dynamic task scheduling model of the sensor data processing engine is used to retrieve the node running status of the data processing network, and dynamic task scheduling is performed based on the node running status and data analysis results to obtain a dynamic task scheduling strategy, including:

[0044] Based on the scenario type in the data analysis results, the agent adjustment module of the dynamic task scheduling model of the sensor data processing engine is used to adjust the policy network of the agent in the dynamic task scheduling module to obtain an updated agent.

[0045] Retrieve the node running status of the data processing network and write the node running status and data analysis results into the updated state space of the intelligent agent in the dynamic task scheduling module;

[0046] Several historical experiences are randomly selected from the experience replay pool, and the action sets of these historical experiences are integrated and written into the action space of the updated agent in the dynamic task scheduling module.

[0047] Based on the state space, the updated agent selects the optimal action in the action space to obtain a dynamic task scheduling strategy; the dynamic task scheduling strategy includes the decision of task allocation and the decision of task priority adjustment.

[0048] We collect real-time experience in dynamic task scheduling, and based on this real-time experience and some historical experience, we continuously train the dynamic task scheduling model to obtain an updated dynamic task scheduling model.

[0049] In one optional implementation, cloud computing resources are allocated using a cloud computing resource allocation model based on a dynamic task scheduling strategy, resulting in a cloud computing resource allocation scheme, including:

[0050] To minimize task completion time and maximize resource utilization, a multi-objective optimization function is set, and this multi-objective optimization function is used as the fitness function.

[0051] Based on the dynamic task scheduling strategy, determine the target cloud computing node number, its task priority, and the task node mapping relationship; set the cloud computing resource allocation scheme format; and define the individual vector format based on the cloud computing resource allocation scheme format.

[0052] Set the ISC population parameters and maximum number of iterations for the ISC algorithm, and initialize the initial ISC population using the Tent chaotic mapping sequence based on the ISC population parameters and individual vector format.

[0053] Using the fitness function, the fitness value of each initial ISC individual in the initial ISC population is calculated, and the initial ISC individual with the best fitness value is taken as the optimal solution.

[0054] Based on the fitness value, roles are assigned to the initial ISC population, resulting in the initial rooster population, the initial hen population, and the initial chick population.

[0055] The initial rooster population, the initial hen population, and the initial chick population are iteratively updated to obtain the updated ISC population.

[0056] Using the fitness function, calculate the fitness value of each updated ISC individual in the updated ISC population, and update the updated ISC individual with the best fitness value as the optimal solution.

[0057] When the number of iterations reaches the maximum number of iterations or the fitness value of the optimal solution meets the requirements, the iteration update is terminated and the optimal solution of the current iteration is output.

[0058] The individual vectors of the ISC individuals corresponding to the optimal solution are decoded to obtain the optimal cloud computing resource allocation scheme; the cloud computing resource allocation scheme includes the resource node mapping relationship of each target cloud computing node and the cloud computing resource quota.

[0059] In one optional implementation, based on cloud computing resources, several target cloud computing nodes invoke the data processing model of the sensor data processing engine to process the preprocessed sensor data, obtaining processed sensor data, including:

[0060] Based on the dynamically generated second attention weight, the second weighted fusion module of the data processing model is used to perform weighted fusion of the temporal features and image features corresponding to the preprocessed sensor data to obtain the second weighted fusion features;

[0061] Based on the second weighted fusion feature, the data aggregation module of the data processing model is used to aggregate the preprocessed sensor data to obtain aggregated sensor data.

[0062] Based on the second weighted fusion feature, the data prediction module of the data processing model is used to perform data prediction and obtain the aggregated sensor data prediction value for future time.

[0063] The aggregated sensor data and the predicted values ​​of the aggregated sensor data are integrated to obtain the processed sensor data.

[0064] Secondly, embodiments of the present invention provide a sensor data processing device based on a cloud computing center, used to implement a sensor data processing method, the device comprising:

[0065] The initialization unit is used in the cloud computing center to build a sensor data processing engine using artificial intelligence algorithms, and to build a distributed storage network and data processing network based on a hierarchical mechanism.

[0066] The preprocessing unit is used to preprocess the collected sensor data at the edge gateway and upload the preprocessed sensor data to the cloud computing center through an encrypted channel.

[0067] The data processing unit is used in the cloud computing center to process the preprocessed sensor data using a sensor data processing engine based on a data processing network, and to distribute and store the processed sensor data based on a distributed storage network.

[0068] A third aspect of this invention provides an electronic device, which includes:

[0069] At least one processor; and a memory communicatively connected to the at least one processor; wherein,

[0070] The memory stores instructions that can be executed by at least one processor, such that the at least one processor can perform the method proposed in the first aspect of the present invention.

[0071] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention. Attached Figure Description

[0072] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention;

[0073] Figure 2 This is a flowchart of the steps of a sensor data processing method based on a cloud computing center provided in an embodiment of the present invention;

[0074] Figure 3 This is a functional unit diagram of a sensor data processing device based on a cloud computing center provided in an embodiment of the present invention. Detailed Implementation

[0075] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0076] The present invention will be further described below with reference to the accompanying drawings.

[0077] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention.

[0078] like Figure 1 As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0079] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0080] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating device, a data storage module, a network communication module, a user interface module, and electronic programs.

[0081] exist Figure 1In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device. The electronic device calls the sensor data processing device based on the cloud computing center stored in the memory 1005 through the processor 1001 and executes the sensor data processing method based on the cloud computing center provided in the embodiment of the present invention.

[0082] Reference Figure 2 The present invention provides a sensor data processing method based on a cloud computing center, the method comprising:

[0083] S201: In the cloud computing center, artificial intelligence algorithms are used to build a sensor data processing engine, and a distributed storage network and data processing network are built based on a hierarchical mechanism;

[0084] S202: At the edge gateway, the collected sensor data is preprocessed and uploaded to the cloud computing center through an encrypted channel;

[0085] S203: In the cloud computing center, a sensor data processing engine is used to process the preprocessed sensor data based on the data processing network, and the processed sensor data is distributed and stored based on the distributed storage network.

[0086] The technical solution provided in this application has at least the following beneficial effects:

[0087] Data preprocessing via edge gateways effectively filters invalid data, reducing network transmission and computing pressure on the cloud computing center. A layered data processing network enables parallel task processing, significantly improving data processing efficiency. A federated learning mechanism is used to train a sensor data processing engine integrating multiple advanced AI models across different scenarios, giving the engine powerful generalization and adaptive capabilities to flexibly meet the data processing needs of various application scenarios. A dynamic task scheduling model intelligently decides task allocation based on real-time node status and task characteristics, while a cloud computing resource allocation model performs multi-objective optimization of cloud computing resources. These two mechanisms work together to maximize resource utilization and minimize task processing latency. A layered distributed storage network and a replication mechanism ensure data redundancy and high availability, avoiding single points of failure. Data transmission at the edge and in the cloud uses encrypted channels, guaranteeing data transmission security. Distributed ledger technology ensures the immutability and traceability of storage address information.

[0088] In one alternative implementation, an artificial intelligence algorithm is used to build a sensor data processing engine in a cloud computing center, and a hierarchical mechanism is used to build a distributed storage network and a data processing network, including:

[0089] S2011: In the cloud computing center, based on large-scale and diverse cloud datasets, use deep learning algorithms to build initial data analysis models and initial data processing models, use reinforcement learning algorithms to build initial dynamic task scheduling models, and use optimization algorithms to build initial cloud computing resource allocation models.

[0090] S2012: Combining the initial data analysis model, the initial data processing model, the initial dynamic task scheduling, and the initial cloud computing resource allocation model, the initial sensor data processing engine is obtained.

[0091] S2013: Based on the federated learning mechanism, the initial sensor data processing engine is trained in a federated learning architecture consisting of several edge servers in different scenarios to obtain the final sensor data processing engine.

[0092] In this embodiment, the federated learning process of distribution-training-aggregation is repeated until the engine's performance converges, resulting in a final sensor data processing engine with strong generalization capabilities.

[0093] S2014: Based on a hierarchical mechanism, several nodes with better storage performance are distributed and connected as storage nodes to obtain a distributed storage network, and several nodes with better cloud computing performance are distributed as cloud computing nodes to obtain a data processing network.

[0094] In this embodiment, the distributed storage network is based on a hierarchical mechanism. All physical nodes in the cloud computing center are evaluated, and nodes with high I / O performance, large storage capacity, and low network latency are selected as storage nodes. These storage nodes are then distributed and connected through a high-speed network to form an independent, highly available storage layer.

[0095] Data processing network: Also based on a hierarchical mechanism, nodes with strong CPU / GPU computing power and large memory are selected as cloud computing nodes. These nodes constitute a processing layer specifically for performing computing tasks.

[0096] In one alternative implementation, the data analysis model is constructed based on the Long Short-Term Memory (LSTM)-Convolutional Neural Network (CNN)-Attention-Multi-Layer Perceptron (MLP) algorithm. The data analysis model includes a time-series feature extraction module constructed based on the LSTM algorithm and an image feature extraction module constructed based on the CNN algorithm, which are connected in parallel, as well as a first weighted fusion module constructed based on the Attention mechanism and a data analysis module constructed based on the MLP algorithm, which are connected in sequence.

[0097] In this embodiment, the time-series feature extraction module is used to process time-series data such as voltage, current, temperature, and humidity, and the image feature extraction module is used to process image data such as those from cameras and infrared sensors. The outputs of the two are fused by the first weighted fusion module, which dynamically learns the weights of different features. Finally, the data analysis module performs comprehensive analysis and outputs data analysis results, including scene type (such as industrial, transportation, and environmental), anomaly detection results (normal / abnormal), tasks to be processed (such as data aggregation and prediction), and task priorities.

[0098] The data processing model is built on the Attention-(Deep Belief Network, DBN)-Autoencoder (AE) algorithm, and includes a second weighted fusion module built on the Attention mechanism, as well as a data prediction module built on the DBN algorithm and a data aggregation module built on the AE algorithm in parallel.

[0099] In this embodiment, the data prediction module is used to predict data values ​​at future times; the data aggregation module is used to perform dimensionality reduction and clustering data aggregation processing on high-dimensional data to extract core information.

[0100] The dynamic task scheduling model is built on the Meta-Policy Optimization (MPO)-Proximal Policy Optimization (PPO) algorithm. The dynamic task scheduling model includes an agent adjustment module built on the MPO algorithm and a dynamic task scheduling module built on the PPO algorithm. The dynamic task scheduling module includes an agent and an experience replay pool.

[0101] The formula for the reward function of the agent is:

[0102]

[0103] In the formula, This is the reward value; The task completion time; For resource utilization rate; Overload penalty; Adjustable reward weights; positive rewards: short task completion time and high resource utilization; negative rewards: node overload, task timeout, and task failure.

[0104] In this embodiment, the agent adjustment module is used to fine-tune the policy network parameters of the agent in the subsequent PPO module according to the "scene type" in the data analysis results, so that its policy is more adapted to the current scene. The core dynamic task scheduling module includes an agent and an experience replay pool. The agent learns the optimal task scheduling policy by interacting with the environment (i.e., the data processing network).

[0105] The cloud computing resource allocation model is built on the Improved Chicken Swarm (ISC) algorithm;

[0106] In this embodiment, the cloud computing resource allocation model is used to find the optimal solution for resource allocation under complex constraints.

[0107] In one optional implementation, based on a federated learning mechanism, the initial sensor data processing engine is trained on various scenarios within a federated learning architecture consisting of several edge servers from different scenarios to obtain the final sensor data processing engine, including:

[0108] S20131: Extract the metadata of the initial sensor data processing engine, send it to several edge servers in different scenarios in the federated learning architecture through an encrypted channel, and reconstruct the initial sensor data processing engine on each edge server based on the metadata.

[0109] In this embodiment, the cloud computing center sends the initial engine's model structure, parameters, and other metadata to multiple edge servers in the federated learning architecture through an encrypted channel. These edge servers are deployed in different physical scenarios (such as factory A, highway B, and park C).

[0110] S20132: Based on the scene dataset of each edge server, train the sensor data processing engine locally, extract the model update amount and scene data amount of the locally trained sensor data processing engine, and upload them to the cloud computing center.

[0111] In this embodiment, each edge server reconstructs the initial engine based on metadata and performs multiple rounds of local training using local, private scene datasets. After training, the original data is not uploaded. Instead, the model update amount (such as gradient or weight difference) and the amount of scene data used in this training are extracted and uploaded to the cloud computing center.

[0112] S20133: In the cloud computing center, based on the federated learning mechanism, the initial sensor data processing engine is adjusted according to the model update volume and scene data volume of several edge servers in different scenarios to obtain the final sensor data processing engine.

[0113] In this embodiment, after receiving updates from all edge servers, the cloud computing center adopts a weighted aggregation strategy to aggregate the model update amount based on the data volume of each edge server, generate a global model update amount, and use it to update the initial sensor data processing engine.

[0114] The formula is:

[0115]

[0116] In the formula, The global model parameters for iteration number t+1, t; For edge servers The model update amount; t is the current iteration number; For edge servers at iteration number t Dynamic federated weights; This represents the total number of edge servers. For edge server indication; This represents the global model update amount.

[0117]

[0118] In the formula, For edge servers at iteration number t Dynamic federated weights; For edge servers with iteration number t ,j is the amount of data; j is the indicator value for the edge server.

[0119] In one optional implementation, the collected sensor data is preprocessed at the edge gateway, and the preprocessed sensor data is uploaded to the cloud computing center via an encrypted channel, including:

[0120] S2011: Use various sensors to collect raw data and transmit it to the edge gateway;

[0121] In this embodiment, the sensor node collects analog or digital signals according to a preset sampling frequency;

[0122] Preprocessing of the original signal, such as analog-to-digital conversion, filtering, and amplification;

[0123] The data packet is encapsulated and includes the following fields:

[0124] Sensor ID (unique identifier);

[0125] Data types (such as voltage, current, power, temperature, images);

[0126] Timestamp (Coordinated Universal Time);

[0127] Data value (e.g., 25.6℃);

[0128] Optional: Status information such as geographic location and battery level;

[0129] S2012: At the edge gateway, raw sensor data from various sensors is collected, and the sensor data is cleaned (noise removal, missing value filling), formatted (data format unification), downsampled, etc. to reduce the amount of data and improve the data quality, resulting in preprocessed sensor data.

[0130] Data reception:

[0131] Receive sensor data via API gateway or message broker (such as Kafka, RabbitMQ);

[0132] Data validation: Checking the data format, integrity, and legality;

[0133] Discard or log illegal data;

[0134] Data cleaning of sensor data:

[0135] Remove duplicate data;

[0136] Impute missing values ​​(using the mean, interpolation, or prediction model);

[0137] Smoothing noise (using moving average, wavelet transform, etc.);

[0138] S2013: Upload the preprocessed sensor data to the cloud computing center;

[0139] Data Upload:

[0140] Transmission protocol:

[0141] It adopts lightweight protocols such as MQTT and CoAP, making it suitable for low-power, low-bandwidth scenarios;

[0142] For high bandwidth and high real-time requirements, HTTP / 2 or gRPC can be used;

[0143] Security mechanisms:

[0144] Data transmission uses TLS / DTLS encryption;

[0145] Device authentication uses tokens, certificates, or two-way authentication.

[0146] Exception handling:

[0147] Data is cached locally when the network is interrupted and retransmitted after the network is restored.

[0148] If no response is received within a timeout period, the system will automatically retry or switch to a backup server.

[0149] In one alternative implementation, at a cloud computing center, a sensor data processing engine is used to process the preprocessed sensor data based on a data processing network, and the processed sensor data is distributed and stored based on a distributed storage network, including:

[0150] S2031: In the cloud computing center, the data analysis model of the sensor data processing engine is used to perform data analysis on the preprocessed sensor data to obtain data analysis results;

[0151] S2032: If the data analysis results include anomalies (such as sensor failure or data mutation), then mark (label) and isolate (store in a special abnormal data area) the preprocessed sensor data, and end the data processing; otherwise, proceed to the next step.

[0152] S2033: Using the dynamic task scheduling model of the sensor data processing engine, the running status of the nodes in the data processing network is retrieved, and dynamic task scheduling is performed based on the running status of the nodes and the data analysis results to obtain a dynamic task scheduling strategy.

[0153] S2034: Based on the dynamic task scheduling strategy, use the cloud computing resource allocation model to allocate cloud computing resources and obtain a cloud computing resource allocation scheme;

[0154] S2035: According to the dynamic task scheduling strategy, schedule several target cloud computing nodes in the data processing network, and allocate corresponding cloud computing resources to several target cloud computing nodes according to the cloud computing resource allocation scheme.

[0155] S2036: Based on cloud computing resources, several target cloud computing nodes call the data processing model of the sensor data processing engine to process the preprocessed sensor data and obtain the processed sensor data.

[0156] S2037: Based on the replication mechanism, the processed sensor data is divided into several data fragments including replicas, and the data fragments are sent to the distributed storage network.

[0157] In this embodiment, based on a replication mechanism (such as a 3-replica strategy), the complete processed sensor data is divided into several data fragments, and multiple identical replicas are created for each fragment. These data fragments and their replicas are distributed to different storage nodes through a distributed storage network. The distribution strategy can ensure that multiple replicas of the same data are not stored in the same physical rack or data center to resist local failures.

[0158] S2038: Use each storage node of the distributed storage network to store several data fragments received locally and update the storage address to the distributed ledger of the distributed storage network;

[0159] In this embodiment, after each storage node receives a data shard, it writes it to its local disk. At the same time, it records the metadata of this storage operation (such as data ID, shard ID, storage node IP, and timestamp) in a distributed ledger shared by the entire distributed storage network (such as a log system based on blockchain technology or a similar consensus mechanism). The decentralized and tamper-proof nature of this ledger ensures the transparency, traceability, and security of the stored information.

[0160] In one alternative implementation, at a cloud computing center, a data analysis model using a sensor data processing engine is used to perform data analysis on the preprocessed sensor data, obtaining data analysis results, including:

[0161] S20311: In the cloud computing center, the preprocessed sensor data is input into the data analysis model of the sensor data processing engine;

[0162] S20312: A time-series feature extraction module using a data analysis model to extract time-series features from preprocessed sensor data in time-series format.

[0163] S20313: Image feature extraction module using data analysis model to extract image features from image format data in preprocessed sensor data;

[0164] S20314: Based on the dynamically generated first attention weight, the first weighted fusion module of the data analysis model is used to perform weighted fusion of temporal features and image features to obtain the first weighted fusion feature;

[0165] S20315: Based on the first weighted fusion feature, the data analysis module of the data analysis model is used to perform data analysis and obtain data analysis results; the data analysis results include scene type, anomaly detection results, tasks to be processed, and task priorities.

[0166] In one optional implementation, the dynamic task scheduling model of the sensor data processing engine is used to retrieve the node running status of the data processing network, and dynamic task scheduling is performed based on the node running status and data analysis results to obtain a dynamic task scheduling strategy, including:

[0167] S20331: Based on the scenario type in the data analysis results, the agent adjustment module of the dynamic task scheduling model of the sensor data processing engine is used to adjust the policy network of the agent in the dynamic task scheduling module to obtain an updated agent that is more suitable for the current scenario.

[0168] S20332: Retrieve the node running status of the data processing network (such as CPU utilization, memory usage, task queue length, and network bandwidth of cloud computing nodes), and write the node running status and data analysis results (current tasks to be processed, quantity, and priority) into the updated state space of the intelligent agent in the dynamic task scheduling module.

[0169] S20333: Randomly extract several historical experiences from the experience replay pool, integrate the action sets of several historical experiences (such as assigning task A to nodes 1, 2, and 3, increasing the priority of task B, discarding low-priority task C, etc.), and write them into the updated action space of the intelligent agent in the dynamic task scheduling module.

[0170] S20334: Based on the state space, the updated agent selects the optimal action with the highest expected reward in the action space to obtain a dynamic task scheduling strategy; the dynamic task scheduling strategy includes the decision to allocate tasks to be processed and the decision to adjust task priorities.

[0171] S20335: Collect real-time experience (state, action, reward, new state) of dynamic task scheduling, and continuously train the dynamic task scheduling model based on real-time experience and some historical experience to obtain an updated dynamic task scheduling model, so that its scheduling capability can be continuously evolved.

[0172] In one optional implementation, cloud computing resources are allocated using a cloud computing resource allocation model based on a dynamic task scheduling strategy, resulting in a cloud computing resource allocation scheme, including:

[0173] S20341: Set a multi-objective optimization function to minimize task completion time and maximize resource utilization, and use the multi-objective optimization function as the fitness function;

[0174]

[0175] In the formula, Let X be the fitness value of individual X in ISC. The task completion time for the cloud computing resource allocation scheme corresponding to ISC individual X; Resource utilization rate of the cloud computing resource allocation scheme corresponding to ISC individual X; Overload penalty for the cloud computing resource allocation scheme corresponding to ISC individual X; The fitness weighting coefficient can be adjusted according to actual needs; X is the ISC individual reference parameter; It is a very small constant;

[0176] S20342: Based on the dynamic task scheduling strategy, determine the target cloud computing node number, its task priority, and the task node mapping relationship; set the cloud computing resource allocation scheme format; and define the individual vector format based on the cloud computing resource allocation scheme format.

[0177] Constraints of the individual vector format:

[0178] Node resource limit;

[0179] Task deadline;

[0180] Network bandwidth limitations;

[0181] S20343: Set the ISC population parameters and maximum number of iterations for the ISC algorithm, and initialize the ISC population using the Tent chaotic mapping sequence based on the ISC population parameters and individual vector format to obtain the initial ISC population; each ISC individual in the ISC population corresponds to an alternative cloud computing resource allocation scheme.

[0182] The formula is:

[0183]

[0184] In the formula, The i-th initial ISC individual in the initial ISC population; Let i be the i-th chaotic variable; Let represent the upper and lower bounds of the search space; i represents the ISC individual indicator.

[0185]

[0186] In the formula, Let i be the (i-1)th chaotic variable; compared with random initialization, chaotic initialization can ensure that the population is evenly distributed in the solution space, thus enhancing diversity.

[0187] S20344: Using the fitness function, calculate the fitness value of each initial ISC individual in the initial ISC population, and take the initial ISC individual with the best fitness value as the optimal solution.

[0188] S20345: Based on the fitness value, assign roles to the initial ISC population to obtain the initial rooster population, the initial hen population, and the initial chick population.

[0189] Rooster individuals: fewer in number, representing the optimal solution, and possessing strong search capabilities;

[0190] Individual hens: They are the most numerous, follow roosters to forage, have local search capabilities, and randomly select roosters to form "families".

[0191] Chicks: follow the mother hen and represent a poor solution. They are follower-oriented. Each chick is randomly assigned a mother hen.

[0192] S20346: Iteratively update the initial rooster population, the initial hen population, and the initial chick population to obtain the updated ISC population;

[0193] The formula is:

[0194]

[0195] In the formula, For the i-th updated rooster individual in iteration t+1, during the first iteration update... This refers to the initial rooster individual; The convergence factor; is a Levy distribution random number; b is the Levy step size, and b∈[1,2];

[0196]

[0197] In the formula, These are the maximum and minimum values ​​of the convergence factor; t represents the maximum number of iterations; t represents the current number of iterations. , To adjust the parameters; It is the hyperbolic tangent function;

[0198]

[0199] In the formula, For the i-th updated hen in iteration t+1, during the first iteration update... This refers to the initial individual hen; A random number between [0,1]; For individual hens The rooster belonging to the family; For random individual roosters; For learning factors;

[0200]

[0201] In the formula, For the i-th updated chick in iteration t+1, during the first iteration update... The initial chick individual; The following factor is [0.5, 2], representing the degree to which chicks follow their mothers; Individual chicks The mother hen;

[0202] S20347: Using the fitness function, calculate the fitness value of each updated ISC individual in the updated ISC population, and update the updated ISC individual with the best fitness value as the optimal solution;

[0203] S20348: When the number of iterations reaches the maximum number of iterations or the fitness value of the optimal solution meets the requirements, terminate the iteration update and output the optimal solution of the current iteration;

[0204] S20349: Decode the individual vector of the ISC individual corresponding to the optimal solution to obtain the optimal cloud computing resource allocation scheme; the cloud computing resource allocation scheme includes the resource node mapping relationship of each target cloud computing node and the cloud computing resource quota (such as the CPU, memory and bandwidth quota of each cloud computing node).

[0205] In one optional implementation, based on cloud computing resources, several target cloud computing nodes invoke the data processing model of the sensor data processing engine to process the preprocessed sensor data, obtaining processed sensor data, including:

[0206] S20361: Based on the dynamically generated second attention weight, the second weighted fusion module of the data processing model is used to perform weighted fusion of the temporal features and image features corresponding to the preprocessed sensor data to obtain the second weighted fusion features.

[0207] S20362: Based on the second weighted fusion feature, the data aggregation module of the data processing model is used to aggregate the preprocessed sensor data to obtain aggregated sensor data.

[0208] In this embodiment, the second fusion feature is encoded, dimensionality reduced, and clustered to obtain structured aggregated sensor data. For example, thousands of voltage readings are aggregated into "regional average voltage" and "voltage change trend".

[0209] S20363: Based on the second weighted fusion feature, the data prediction module of the data processing model is used to perform data prediction to obtain the aggregated sensor data prediction value for a future time (e.g., 1 hour in the future).

[0210] S20364: Integrate the aggregated sensor data and the predicted values ​​of the aggregated sensor data to obtain processed sensor data, which includes both current status analysis and future trends.

[0211] This invention also provides a sensor data processing device based on a cloud computing center, as described in the embodiments of the present invention. Figure 3 The diagram shows a functional unit diagram of a sensor data processing device 300 based on a cloud computing center according to the present invention. The device may include the following units:

[0212] Initialization unit 301 is used to build a sensor data processing engine in the cloud computing center using artificial intelligence algorithms, and to build a distributed storage network and data processing network based on the hierarchical mechanism 302.

[0213] The preprocessing unit 303 is used to preprocess the collected sensor data at the edge gateway and upload the preprocessed sensor data to the cloud computing center through an encrypted channel.

[0214] The data processing unit 304 is used in a cloud computing center to process preprocessed sensor data using a sensor data processing engine based on a data processing network, and to distribute and store the processed sensor data based on a distributed storage network.

[0215] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus.

[0216] Memory, used to store computer programs;

[0217] The processor, when executing a program stored in memory, implements the sensor data processing method based on a cloud computing center according to the present invention.

[0218] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned terminal and other devices. The memory can include Random Access Memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0219] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0220] Furthermore, to achieve the above objectives, embodiments of the present invention also propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the sensor data processing method based on a cloud computing center according to embodiments of the present invention.

[0221] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable vehicles (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0222] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0223] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0224] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0225] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "" and / or "" indicate that either one or both can be selected. Furthermore, the terms "includes," "contains," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the statement "includes a..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0226] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A sensor data processing method based on a cloud computing center, characterized in that, The method includes: In the cloud computing center, artificial intelligence algorithms are used to build a sensor data processing engine, and a distributed storage network and data processing network are constructed based on a hierarchical mechanism, including: In the cloud computing center, based on cloud datasets, we use deep learning algorithms to build initial data analysis and data processing models, reinforcement learning algorithms to build initial dynamic task scheduling models, and optimization algorithms to build initial cloud computing resource allocation models. By combining the initial data analysis model, the initial data processing model, the initial dynamic task scheduling, and the initial cloud computing resource allocation model, an initial sensor data processing engine is obtained. The cloud computing resource allocation model is constructed based on the ISC algorithm; Based on the federated learning mechanism, the initial sensor data processing engine is trained on various scenarios within a federated learning architecture consisting of several edge servers for different scenarios, resulting in the final sensor data processing engine, including: Extract the metadata of the initial sensor data processing engine and send it to several edge servers in different scenarios in the federated learning architecture through an encrypted channel. On each edge server, the initial sensor data processing engine is reconstructed based on the metadata. Based on the scene dataset of each edge server, the sensor data processing engine is trained locally, and the model update amount and scene data amount of the locally trained sensor data processing engine are extracted and uploaded to the cloud computing center. In the cloud computing center, based on the federated learning mechanism, the initial sensor data processing engine is adjusted according to the model update volume and scene data volume of several edge servers in different scenarios to obtain the final sensor data processing engine. After receiving updates from all edge servers, the cloud computing center uses a weighted aggregation strategy to aggregate the model update volume based on the data volume of each edge server, generating a global model update volume, and uses it to update the initial sensor data processing engine. The formula is: In the formula, Number of iterations t+ 1, t Global model parameters; For edge servers Model update volume; t This represents the current iteration number; To the number of iterations t edge servers Dynamic federated weights; This represents the total number of edge servers. For edge server indication; This represents the global model update amount. In the formula, To the number of iterations t edge servers Dynamic federated weights; Number of iterations t edge servers ,j The amount of data; j For edge server indication; Based on a hierarchical mechanism, several nodes with better storage performance are distributed and connected as storage nodes to obtain a distributed storage network, and several nodes with better cloud computing performance are distributed as cloud computing nodes to obtain a data processing network. At the edge gateway, the collected sensor data is preprocessed and then uploaded to the cloud computing center through an encrypted channel. In the cloud computing center, a sensor data processing engine is used to process the pre-processed sensor data based on a data processing network, and the processed sensor data is then distributed and stored based on a distributed storage network, including: In the cloud computing center, the data analysis model of the sensor data processing engine is used to perform data analysis on the preprocessed sensor data to obtain data analysis results. If the data analysis results show anomalies, the preprocessed sensor data is marked and isolated, and data processing ends; otherwise, proceed to the next step. The dynamic task scheduling model of the sensor data processing engine is used to retrieve the node running status of the data processing network, and dynamic task scheduling is performed based on the node running status and data analysis results to obtain the dynamic task scheduling strategy. Based on the dynamic task scheduling strategy, a cloud computing resource allocation model is used to allocate cloud computing resources, resulting in a cloud computing resource allocation scheme, including: To minimize task completion time and maximize resource utilization, a multi-objective optimization function is set, and this function is used as the fitness function. The formula is as follows: In the formula, For ISC individuals X fitness value; For ISC individuals X The corresponding task completion time for the cloud computing resource allocation scheme; For ISC individuals X Resource utilization rate of the corresponding cloud computing resource allocation scheme; For ISC individuals X The corresponding overload penalty for cloud computing resource allocation schemes; The fitness weighting coefficient can be adjusted according to actual needs; X Parameters for ISC individuals; It is a very small constant; Based on the dynamic task scheduling strategy, determine the target cloud computing node number, its task priority, and the task node mapping relationship; set the cloud computing resource allocation scheme format; and define the individual vector format based on the cloud computing resource allocation scheme format. Set the ISC population parameters and maximum number of iterations for the ISC algorithm, and initialize the initial ISC population using the Tent chaotic mapping sequence based on the ISC population parameters and individual vector format. Using the fitness function, the fitness value of each initial ISC individual in the initial ISC population is calculated, and the initial ISC individual with the best fitness value is taken as the optimal solution. Based on the fitness value, roles are assigned to the initial ISC population, resulting in the initial rooster population, the initial hen population, and the initial chick population. The initial rooster population, initial hen population, and initial chick population are iteratively updated to obtain the updated ISC population, using the following formula: In the formula, Number of iterations t+ 1, t The i A newer rooster individual, during the first iteration update. This refers to the initial rooster individual; The convergence factor; for Levy Distribute random numbers; b for Levy Step length, and b ∈[1,2]; In the formula, These are the maximum and minimum values ​​of the convergence factor; This represents the maximum number of iterations. t This represents the current iteration number; , To adjust the parameters; It is the hyperbolic tangent function; In the formula, Number of iterations t+ 1, t The i A newer individual hen, during the first iteration update. This refers to the initial individual hen; A random number between [0,1]; For individual hens The rooster belonging to the family; For random individual roosters; For learning factors; In the formula, Number of iterations t+ 1, t The i A newer chick individual, during the first iteration update The initial chick individual; The following factor is [0.5, 2]. Individual chicks The mother hen; Using the fitness function, calculate the fitness value of each updated ISC individual in the updated ISC population, and update the updated ISC individual with the best fitness value as the optimal solution. When the number of iterations reaches the maximum number of iterations or the fitness value of the optimal solution meets the requirements, the iteration update is terminated and the optimal solution of the current iteration is output. The individual vectors of the ISC individuals corresponding to the optimal solution are decoded to obtain the optimal cloud computing resource allocation scheme; the cloud computing resource allocation scheme includes the resource node mapping relationship and cloud computing resource quota of each target cloud computing node; According to the dynamic task scheduling strategy, several target cloud computing nodes are scheduled in the data processing network, and corresponding cloud computing resources are allocated to several target cloud computing nodes according to the cloud computing resource allocation scheme. Based on several target cloud computing nodes with cloud computing resources, the data processing model of the sensor data processing engine is called to process the preprocessed sensor data and obtain the processed sensor data. Based on the replication mechanism, the processed sensor data is divided into several data fragments including replicas, and the data fragments are sent to the distributed storage network. Each storage node in the distributed storage network stores several data fragments locally and updates the storage addresses to the distributed ledger of the distributed storage network.

2. The sensor data processing method based on a cloud computing center according to claim 1, characterized in that, The data analysis model is constructed based on the LSTM-CNN-Attention-MLP algorithm, and includes a time-series feature extraction module and an image feature extraction module constructed based on the CNN algorithm in parallel, as well as a first weighted fusion module constructed based on the Attention mechanism and a data analysis module constructed based on the MLP algorithm connected in sequence. The data processing model is built based on the Attention-DBN-AE algorithm, and includes a second weighted fusion module built based on the Attention mechanism, as well as a data prediction module built based on the DBN algorithm and a data aggregation module built based on the AE algorithm in parallel. The dynamic task scheduling model is constructed based on the MPO-PPO algorithm, and includes an agent adjustment module constructed based on the MPO algorithm and a dynamic task scheduling module constructed based on the PPO algorithm. The dynamic task scheduling module includes an agent and an experience replay pool.

3. The sensor data processing method based on a cloud computing center according to claim 2, characterized in that, In the cloud computing center, the data analysis model of the sensor data processing engine is used to analyze the preprocessed sensor data and obtain the data analysis results, including: In the cloud computing center, the preprocessed sensor data is input into the data analysis model of the sensor data processing engine; The temporal feature extraction module of the data analysis model is used to extract the temporal features of the temporal format data in the preprocessed sensor data; The image feature extraction module of the data analysis model is used to extract image features from the image format data in the preprocessed sensor data. Based on the dynamically generated first attention weight, the first weighted fusion module of the data analysis model is used to perform weighted fusion of temporal features and image features to obtain the first weighted fusion feature. Based on the first weighted fusion feature, the data analysis module of the data analysis model is used to perform data analysis and obtain data analysis results; the data analysis results include scene type, anomaly detection results, tasks to be processed, and task priorities.

4. The sensor data processing method based on a cloud computing center according to claim 3, characterized in that, Using the dynamic task scheduling model of the sensor data processing engine, the running status of the nodes in the data processing network is retrieved, and dynamic task scheduling is performed based on the node running status and data analysis results to obtain a dynamic task scheduling strategy, including: Based on the scenario type in the data analysis results, the agent adjustment module of the dynamic task scheduling model of the sensor data processing engine is used to adjust the policy network of the agent in the dynamic task scheduling module to obtain an updated agent. Retrieve the node running status of the data processing network and write the node running status and data analysis results into the updated state space of the intelligent agent in the dynamic task scheduling module; Several historical experiences are randomly selected from the experience replay pool, and the action sets of these historical experiences are integrated and written into the action space of the updated agent in the dynamic task scheduling module. Based on the state space, the updated agent selects the optimal action in the action space to obtain a dynamic task scheduling strategy; the dynamic task scheduling strategy includes task allocation decisions and task priority adjustment decisions. We collect real-time experience in dynamic task scheduling, and based on this real-time experience and some historical experience, we continuously train the dynamic task scheduling model to obtain an updated dynamic task scheduling model.

5. The sensor data processing method based on a cloud computing center according to claim 4, characterized in that, Based on several target cloud computing nodes with cloud computing resources, the data processing model of the sensor data processing engine is invoked to process the preprocessed sensor data, resulting in processed sensor data, including: Based on the dynamically generated second attention weight, the second weighted fusion module of the data processing model is used to perform weighted fusion of the temporal features and image features corresponding to the preprocessed sensor data to obtain the second weighted fusion features; Based on the second weighted fusion feature, the data aggregation module of the data processing model is used to aggregate the preprocessed sensor data to obtain aggregated sensor data. Based on the second weighted fusion feature, the data prediction module of the data processing model is used to perform data prediction and obtain the aggregated sensor data prediction value for future time. The aggregated sensor data and the predicted values ​​of the aggregated sensor data are integrated to obtain the processed sensor data.

6. A sensor data processing device based on a cloud computing center, used to implement the sensor data processing method as described in any one of claims 1-5, characterized in that, The device includes: The initialization unit is used in the cloud computing center to build a sensor data processing engine using artificial intelligence algorithms, and to build a distributed storage network and data processing network based on a hierarchical mechanism. The preprocessing unit is used to preprocess the collected sensor data at the edge gateway and upload the preprocessed sensor data to the cloud computing center through an encrypted channel. The data processing unit is used in the cloud computing center to process the preprocessed sensor data using a sensor data processing engine based on a data processing network, and to distribute and store the processed sensor data based on a distributed storage network.

Images