Cloud Service-Based Software-Defined Internet of Things Control Method and System for Urban Components

Abstract

The present invention relates to the technical field of urban intelligent management, and discloses a software-defined urban component Internet of Things control method and system based on cloud services. The method constructs a digital twin model of urban components, fuses multi-source data to generate a state feature vector; uses a graph neural network prediction model to predict the resource demand distribution in combination with environmental and user behavior data; adopts a reinforcement learning algorithm to optimize the network strategy, considering cross-layer protocol adaptation and latency constraints; deploys an edge-cloud collaborative control mechanism to ensure state synchronization; constructs a protocol conversion middleware to achieve heterogeneous device protocol adaptation. The invention can achieve accurate state perception of urban components, reasonable resource allocation, efficient network operation, edge-cloud collaboration, and heterogeneous device compatibility, effectively improve the intelligent management level of urban components, improve urban operation efficiency and resource utilization rate, and has good application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of urban intelligent management, and specifically to a software-defined urban component Internet of Things control method and system based on cloud services. Background Art

[0002] With the acceleration of the urbanization process, the scale of cities is constantly expanding, the number and types of urban components are increasing rapidly, and the intelligent control and management of urban components have become a key requirement for the efficient operation and sustainable development of cities.

[0003] There are many limitations in the traditional urban component management methods. In terms of data processing, multi-source data such as sensor data, historical operation data, and geographic information system data involved in urban components have problems of inconsistent formats and scattered sources, making it difficult to conduct effective fusion and analysis. For example, sensors produced by different manufacturers have different data formats and communication protocols, resulting in difficult data integration and unable to form comprehensive and accurate urban component status information, thus affecting the overall control and decision-making of urban components.

[0004] From the perspective of resource allocation, there is a lack of accurate resource demand prediction means. In the past, resource allocation mostly relied on experience or simple statistical methods, and it was impossible to accurately predict the resource demand distribution of each urban component in the future period. When the traffic flow is large, the energy supply and network bandwidth of traffic lights cannot be adjusted in time, resulting in traffic congestion and resource waste; when the weather conditions change, the urban lighting system cannot reasonably allocate energy according to actual needs, leading to low energy utilization efficiency.

[0005] In terms of network strategies, the existing network bandwidth allocation strategies and task scheduling priorities lack the ability of dynamic optimization. Facing the complex and changeable urban environment and constantly changing business requirements, it is difficult to ensure the efficient operation of the network and the timely processing of tasks. When multiple urban components request network resources at the same time, network congestion is likely to occur, affecting the timeliness and stability of data transmission.

[0006] In heterogeneous device communication, there are a large number of heterogeneous devices in urban components, and their private communication protocols are different, which poses a huge challenge to the interconnection and collaborative work between devices. Different brands and models of Internet of Things devices cannot directly interact with each other due to different communication protocols, increasing the difficulty of system integration and management.

[0007] In addition, there are also deficiencies in the collaboration between edge computing and cloud computing. The processing capacity of edge nodes is limited, and they cannot fully utilize the powerful computing resources and storage resources of cloud computing. Moreover, it is difficult to ensure data synchronization and state consistency between the two. When an edge node fails, tasks cannot be transferred to the cloud for processing in a timely manner, resulting in interruptions in the control and management of urban components. These problems severely restrict the intelligent development of cities and urgently require a new technology to solve these problems and achieve efficient and intelligent management of urban components. Summary of the Invention

[0008] The purpose of the present invention is to provide a software-defined Internet of Things control method and system for urban components based on cloud services to solve the problems raised in the above background technology.

[0009] To achieve the above purpose, the present invention provides the following technical solutions: A software-defined Internet of Things control method for urban components based on cloud services, the method includes:

[0010] Construct a digital twin model of urban components, including integrating real-time sensor data, historical operation data, and urban geographic information system data through a cloud platform to generate a dynamic virtual mapping model; the digital twin model supports multi-source heterogeneous data fusion and generates state feature vectors of urban components through spatio-temporal correlation analysis;

[0011] The urban components include Internet of Things-enabled infrastructure devices deployed in urban public spaces, specifically including traffic management devices, public facility devices, energy management devices, and environmental monitoring devices; among them, the traffic management devices include traffic lights, vehicle detection sensors, and electronic signs; the public facility devices include smart street lights, smart manhole covers, and public charging piles; the energy management devices include smart electricity meters, distribution monitoring terminals, and photovoltaic inverters; the environmental monitoring devices include air quality sensors, noise monitors, and water quality monitoring terminals; the urban components are connected to the cloud platform through wired or wireless communication protocols and have real-time status collection, remote control, and data interaction functions.

[0012] Execute dynamic resource demand prediction, including a prediction model based on a graph neural network, inputting the state feature vector, real-time environmental data, and user behavior data, and outputting the resource demand distribution of each urban component in the future time period; the real-time environmental data includes traffic flow, meteorological conditions, and energy consumption data;

[0013] Generate software-defined network policies, including dynamically optimizing the network bandwidth allocation policy and task scheduling priority according to the resource demand distribution by using a reinforcement learning algorithm; the policy generation process includes cross-layer protocol adaptation and end-to-end delay constraint modeling;

[0014] Deploy an edge-cloud collaborative control mechanism, including real-time processing of highly time-sensitive control instructions through edge nodes and global resource coordination through a cloud platform; the collaborative mechanism uses a distributed consensus algorithm to ensure the synchronization of edge and cloud states;

[0015] Implement heterogeneous device protocol adaptation, including building a lightweight protocol conversion middleware to uniformly map the private communication protocols of urban components to standard Internet of Things protocols; the middleware supports dynamic protocol plug-and-play and adaptive data encapsulation.

[0016] Preferably, the construction of the digital twin model includes: using an open-source 3D engine to render the physical topology of urban components, and realizing real-time data stream access in combination with Apache Kafka; using a time-series database to store historical operation data, and accelerating the generation of feature vectors through a spatio-temporal indexing algorithm.

[0017] Preferably, the structure of the graph neural network prediction model includes:

[0018] Input layer: Receive the state feature vector, environmental data, and user behavior data;

[0019] Graph construction layer: Generate a dynamic graph structure based on the geographical location and functional dependencies of urban components;

[0020] Graph convolution layer: Extract spatial dependence features through multi-layer graph convolution operations;

[0021] Temporal fusion layer: Introduce a gated recurrent unit to fuse temporal dynamic features;

[0022] Output layer: Generate a confidence matrix of resource demand distribution.

[0023] Preferably, the optimization objective function of the reinforcement learning algorithm is:

[0024]

[0025] Where, is the immediate reward function, including network load balance, task completion rate, and energy consumption efficiency; is the discount factor; and represent the state space and action space respectively; represents the next state, the state to which the system transfers after executing the action and is the optimal action available in the state

[0026] Preferably, the distributed consensus algorithm uses the Raft protocol to achieve data synchronization between edge nodes and the cloud platform, and detects node abnormal states through a heartbeat mechanism to trigger a fault tolerance control strategy.​

[0027] Preferably, the implementation of the protocol conversion middleware includes:

[0028] Protocol parsing module: Parsing private protocol data frames based on a finite state machine;

[0029] Field mapping module: Dynamically binding private protocol fields to standard Internet of Things protocol fields;

[0030] Data encapsulation module: Implementing cross - protocol data encapsulation in TLV format.

[0031] Preferably, the formula for the end - to - end delay constraint modeling is:

[0032]

[0033] Where represents the total end - to - end delay, is the processing delay of the th task, is the transmission delay, is the maximum tolerable delay of the system.

[0034] Preferably, the spatio - temporal correlation analysis includes: Dividing the functional areas of urban components based on Voronoi diagrams and predicting the state transition probabilities between regions through hidden Markov models.

[0035] Preferably, the present invention further includes a software - defined urban component Internet of Things control system based on cloud services, and the system includes:

[0036] Urban component digital twin modeling module: Used to integrate real - time sensor data, historical operation data, and urban geographic information system data through a cloud platform to generate a dynamic virtual mapping model, support multi - source heterogeneous data fusion, and generate urban component state feature vectors through spatio - temporal correlation analysis;

[0037] Dynamic resource demand prediction module: Based on a prediction model of graph neural networks, receiving urban component state feature vectors, real - time environmental data, and user behavior data, and outputting the resource demand distribution of each urban component in the future time period, where the real - time environmental data includes traffic flow, meteorological conditions, and energy consumption data;

[0038] Software - defined network policy generation module: According to the resource demand distribution, using reinforcement learning algorithms to dynamically optimize network bandwidth allocation policies and task scheduling priorities, and the policy generation process covers cross - layer protocol adaptation and end - to - end delay constraint modeling;

[0039] Edge Cloud Collaborative Control Mechanism Deployment Module: Leverage edge nodes to process highly time-sensitive control instructions in real-time, utilize the cloud platform for global resource coordination, and adopt a distributed consensus algorithm to ensure the synchronization of edge and cloud states;

[0040] Heterogeneous Device Protocol Adaptation Module: Build a lightweight protocol conversion middleware to uniformly map the private communication protocols of urban components into standard Internet of Things protocols. This middleware supports dynamic protocol plug-and-play and adaptive data encapsulation.

[0041] The present invention also includes a computer-readable storage medium, on which a computer program is stored. The program, when executed by a processor, implements the above method for software-defined Internet of Things control of urban components based on cloud services.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] By constructing a digital twin model of urban components, integrating real-time sensor data, historical operation data, and urban geographic information system data, supporting the fusion of multi-source heterogeneous data, and performing spatio-temporal correlation analysis to generate state feature vectors of urban components. This enables urban managers to comprehensively, accurately, and in real-time grasp the operating status of urban components. Taking intelligent street lamp management as an example, by combining light sensor data, past energy consumption data, and the geographical location information of street lamps, not only can the working status of each street lamp be understood in real-time, but also its failure risk can be predicted, and maintenance can be carried out in advance, greatly improving the reliability and operation and maintenance efficiency of the street lamp system, and reducing the adverse impacts on the lives of urban residents and traffic safety caused by street lamp failures. Based on the prediction model of graph neural network, by inputting state feature vectors, real-time environmental data, and user behavior data, the resource demand distribution of each urban component in the future period can be output. This provides a scientific basis for the reasonable allocation of resources, avoiding waste and shortage of resources. In traffic management, according to data such as traffic flow and meteorological conditions, predict the resource requirements of traffic lights on different sections, adjust the signal timing plan and network bandwidth in advance, relieve traffic congestion, and improve road traffic efficiency; in energy management, according to energy consumption data and user behavior, predict the energy requirements of each region, optimize energy distribution, reduce energy losses, and achieve the goal of energy conservation and emission reduction.

[0044] Utilize reinforcement learning algorithms to dynamically optimize network bandwidth allocation strategies and task scheduling priorities, and perform cross-layer protocol adaptation and end-to-end delay constraint modeling. This ensures that the network can operate efficiently and tasks can be processed in a timely manner in a complex urban environment. In the urban emergency response scenario, when an emergency occurs, control instructions of relevant devices can be transmitted quickly and accurately, giving priority to ensuring the execution of critical tasks, improving the city's ability to respond to emergencies, and reducing losses and social impacts.

[0045] Deploy an edge-cloud collaborative control mechanism. The edge nodes process high-timeliness control instructions in real time, the cloud platform coordinates global resources, and a distributed consistency algorithm is used to ensure the synchronization of the edge and cloud states. This collaborative mechanism improves the response speed and reliability of the system. When abnormalities occur in urban components near the edge nodes, the edge nodes can quickly respond and simultaneously synchronize the information to the cloud. The cloud can then perform global scheduling and decision-making to ensure the stable operation of the entire system. Even if some edge nodes fail, the cloud platform can promptly take over the tasks to ensure the normal operation of urban components.

[0046] Implement heterogeneous device protocol adaptation, build a lightweight protocol conversion middleware, uniformly map the private communication protocols of urban components to standard Internet of Things protocols, and support dynamic protocol plug-and-play and adaptive data encapsulation. This enables devices from different manufacturers and of different types to seamlessly access the system and achieve interconnection and interoperability. In the construction of urban Internet of Things, there is no need to re-develop complex interfaces and communication programs for newly connected devices, reducing the system integration cost, improving the scalability and compatibility of the system, facilitating the continuous upgrade and function expansion of urban Internet of Things, and meeting the growing demand for intelligent management in future urban development. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is the working principle diagram of the software-defined urban component Internet of Things control method based on cloud services according to the present invention;

[0048] Figure 2 is the working flow chart of the graph neural network prediction model;

[0049] Figure 3 is the working flow chart of the edge-cloud collaborative control mechanism.

[0050] Figure 4 is the working flow chart of the protocol conversion middleware. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] Please refer to Figures 1-4 , the present invention provides a technical solution: a software-defined urban component Internet of Things control method based on cloud services, the method includes:

[0053] Build a digital twin model of urban components, integrate real-time sensor data, historical operation data, and urban geographic information system data to generate a dynamic virtual mapping model, support multi-source heterogeneous data fusion, and generate state feature vectors of urban components through spatio-temporal correlation analysis.

[0054] Execute dynamic resource demand prediction. Use a prediction model based on graph neural networks, input the state feature vector, real-time environmental data, and user behavior data, and output the resource demand distribution of each urban component in the future time period.

[0055] Generate software-defined network policies. According to the resource demand distribution, use reinforcement learning algorithms to dynamically optimize network bandwidth allocation policies and task scheduling priorities. The process involves cross-layer protocol adaptation and end-to-end delay constraint modeling.

[0056] Deploy an edge-cloud collaborative control mechanism. Real-time process high-timeliness control instructions through edge nodes, coordinate global resources with the help of the cloud platform, and use a distributed consistency algorithm to ensure the state synchronization between the edge and the cloud.

[0057] Implement heterogeneous device protocol adaptation. Build a lightweight protocol conversion middleware to uniformly map the private communication protocols of urban components to standard Internet of Things protocols. The middleware supports dynamic protocol plug-and-play and adaptive data encapsulation.

[0058] The present invention will be further described below in conjunction with Embodiments 1 to 5:

[0059] Embodiment 1:

[0060] When building a digital twin model of urban components, an open-source 3D engine (such as Unity) is selected to render the physical topology of urban components. Unity has powerful graphics rendering capabilities and rich plugin resources, and can realistically present the appearance, position, and spatial relationship of urban components. Using its 3D modeling function, various urban components, such as street lights, manhole covers, traffic lights, etc., are modeled according to actual sizes and positions to build a virtual scene consistent with the layout of real urban components.

[0061] Integrate real-time data stream access in combination with Apache Kafka. Apache Kafka is a high-throughput distributed messaging system, which can play a key role in the case of numerous urban components and high data generation frequencies. Various sensors installed on urban components, such as temperature sensors, humidity sensors, position sensors, etc., will collect a large amount of data in real time. These data are sent to the Kafka cluster through the producer client of Kafka, and then the consumer client obtains the data from the cluster and transmits it to the cloud platform. The partition and replication mechanisms of Kafka ensure the reliability and efficiency of data transmission. Even if some nodes fail, the data will not be lost or the transmission will be interrupted.

[0062] Historical operation data is stored in a time series database (such as InfluxDB). InfluxDB is optimized for time series data and can quickly store and query data that changes over time. The historical operation data of urban components, such as the on / off time of street lights and the fault records of devices, is stored in InfluxDB in chronological order. Through the spatio-temporal indexing algorithm, with the geographical location and time of urban components as the indexing dimensions, the speed of querying and extracting data is greatly improved. When generating the state feature vector of urban components, relevant historical data can be quickly located, providing a basis for accurately analyzing the component state.

[0063] In terms of spatio-temporal correlation analysis, the functional areas of urban components are divided based on Voronoi diagrams. A Voronoi diagram divides space according to a known set of points in space, and each region consists of the spatial points closest to a specific point. In the scenario of urban components, each urban component is regarded as a point, and a Voronoi diagram is constructed based on its geographical location. In this way, urban components with similar or related functions will be divided into the same region. For example, multiple street lights in a certain area can be divided into an illumination functional area, and traffic lights and traffic flow monitoring devices in the area can be divided into a traffic management functional area.

[0064] The hidden Markov model is used to predict the state transition probability between regions. The hidden Markov model is a statistical model that assumes that the state of the system is not directly observable (hidden state), but the change of the state can be inferred through the observed output data. In the Internet of Things control of urban components, different functional areas are regarded as the states of the hidden Markov model, and the state data of urban components within the region are used as observations. Through learning historical data, the model can predict how the state change of one region affects the state transition probability of adjacent regions at different time points. For example, when the traffic flow monitoring area detects a sudden increase in traffic flow, the model can predict the probability that the adjacent traffic light area adjusts the signal light duration, so as to make a response in advance and optimize the traffic control strategy.

[0065] Embodiment 2:

[0066] The input layer of the graph neural network prediction model is responsible for receiving various data. Among them, the state feature vector is generated by the digital twin model and contains information about the current operating state and performance indicators of urban components.

[0067] The operating state includes device working parameters, fault states, and communication connection states; the working parameters include current, voltage, temperature, switch quantity state, and device-specific operating parameters, the fault states include fault codes, abnormal threshold trigger states, and device self-check results, and the communication connection states include signal strength, packet loss rate, and network throughput;

[0068] The state feature vector consists of a timestamp, a geographical location identifier, and multi-dimensional state data, where the multi-dimensional state data is structured and encoded according to device types, including:

[0069] Encoding the phase state, countdown information, and signal switching delay time of traffic lights;

[0070] Encoding the brightness level, light intensity compliance rate, and lamp life attenuation coefficient of intelligent street lights;

[0071] Encoding the tilt angle, vibration frequency, and illegal opening alarm status of intelligent manhole covers.

[0072] The performance indicators include efficiency indicators, energy consumption indicators, and reliability indicators; the efficiency indicators include task completion time, response delay time, and resource utilization rate, the energy consumption indicators include real-time power value, cumulative energy consumption value, and energy efficiency ratio, and the reliability indicators include mean time between failures, repair response rate, and false alarm rate.

[0073] The traffic flow data in the real-time environmental data can be obtained from the traffic monitoring system and reflects the flow of vehicles on the road; the meteorological condition data, such as temperature, humidity, wind speed, etc., can be obtained from the interfaces of the meteorological department; the energy consumption data comes from the energy consumption monitoring module of the urban components themselves. The user behavior data can be obtained through the interaction records between users and urban components, such as the usage time of intelligent charging piles by users, the brightness adjustment operations on the intelligent lighting system, etc. These multi-source data are integrated and then input into the model.

[0074] The graph construction layer generates a dynamic graph structure based on the geographical location and functional dependencies of urban components. The geographical location of urban components is determined using Geographic Information System (GIS) data, and the functional dependencies are determined by analyzing the data interaction and business associations between components. For example, there is a data sharing relationship between intelligent street lights and nearby environmental monitoring devices, and intelligent traffic lights cooperate with surrounding vehicle detection sensors. Based on these relationships, urban components are abstracted as nodes in the graph, and the connection relationships between components are abstracted as edges, thus constructing a dynamic graph structure. As the states and relationships of urban components change, the graph structure will also be updated in real time.

[0075] The graph convolution layer extracts spatial dependency features through multiple layers of graph convolution operations. The graph convolution operation can aggregate and update node features on the graph structure. In this model, through multiple layers of graph convolution, the model can learn the spatial association patterns between different urban components. For example, through graph convolution, it can be found that the brightness adjustment of multiple intelligent street lights in a certain area is related to the surrounding environmental light intensity, and the impact of this relationship on energy consumption. Each layer of graph convolution extracts spatial features at a more abstract level, enabling the model to have a deeper understanding of spatial dependency relationships.

[0076] The temporal fusion layer introduces the Gated Recurrent Unit (GRU) to fuse temporal dynamic features. GRU is a recurrent neural network structure that can effectively process time series data. The states and resource requirements of urban components change over time, showing obvious temporal characteristics. GRU performs temporal analysis on the data processed by the graph convolutional layer and controls the transmission and forgetting of information through a gating mechanism. It can remember long-term time series information and dynamically adjust the dependence on historical information according to the current input. For example, when analyzing the resource requirements of intelligent charging piles, GRU can combine the charging demand change trends in the past period to predict future demands and improve the prediction accuracy.

[0077] The output layer generates a confidence matrix of the resource demand distribution. After the processing of the previous layers, the model has a comprehensive understanding of the states of urban components, environmental factors, and user behaviors. The output layer calculates the likelihood of each urban component's demand for different resources (such as electricity, network bandwidth, etc.) in the future time period based on this information and outputs it in the form of a confidence matrix. This matrix provides key data support for the subsequent software-defined network policy generation and helps the system plan resource allocation in advance.

[0078] Embodiment 3:

[0079] This embodiment elaborates in detail the application of the reinforcement learning algorithm in software-defined network policy generation and the specific process of end-to-end delay constraint modeling, aiming to improve network performance and task execution efficiency by optimizing the objective function and controlling the delay.

[0080] When generating software-defined network policies, the reinforcement learning algorithm is used to dynamically optimize the network bandwidth allocation policy and task scheduling priority. The optimization objective function of the reinforcement learning algorithm is:

[0081] where, is the immediate reward function, including network load balance, task completion rate, and energy consumption efficiency; is the discount factor; and represent the state space and action space respectively; represents the next state, the state that the system transfers to after executing the action , the optimal action available in the state The network load balance is measured by monitoring the load conditions of each node in the network. A load-balanced network can make more reasonable use of resources. The task completion rate reflects the proportion of tasks successfully completed by the system within a certain time. Improving the task completion rate is crucial for ensuring the normal operation of urban components. The energy consumption efficiency reflects the energy utilization situation of the system during operation. Optimizing the energy consumption efficiency helps reduce operating costs.

[0082] is the discount factor, with a value between 0 and 1. It is used to balance the importance of immediate rewards and future rewards. When is close to 1, the algorithm pays more attention to long-term benefits and will consider the impact of the current decision on future states; when is close to 0, the algorithm focuses more on immediate rewards. In practical applications, adjust the value of according to the characteristics of the urban component IoT control scenario. For example, in the intelligent transportation control scenario with extremely high real-time requirements, appropriately reduce the value of to prioritize the rapid completion of the current task; in the urban energy management scenario that attaches more importance to long-term resource optimization, increase the value of .

[0083] and represent the state space and action space respectively, represents the state to which the system transfers after executing the action , is the optimal action available in state . In the urban component IoT control scenario, the state space includes information such as the resource demand distribution of urban components, the current network bandwidth usage, and the task queue status; the action space includes different network bandwidth allocation strategies and task scheduling priority settings. The reinforcement learning algorithm gradually learns the optimal strategy by continuously trying different actions and calculating the reward value according to the optimization objective function.

[0084] The end-to-end delay constraint modeling formula is:

[0085] where, represents the total end-to-end delay, is the processing delay of the th task, is the transmission delay, is the maximum tolerable delay of the system. In actual operation, accurately calculate the processing delay and transmission delay of each task. For example, for the data transmission task of an intelligent security camera, the processing delay includes the encoding processing time of the camera for image data, and the transmission delay is the time for the data to be transmitted from the camera to the cloud platform in the network. By monitoring and analyzing these delay data, when it is found that the total delay is close to or exceeds , adjust the network bandwidth allocation strategy in a timely manner to prioritize tasks sensitive to delay and ensure the normal operation of the system.

[0086] Example 4:

[0087] This embodiment specifically introduces the operation mode of the edge-cloud collaborative control mechanism and the implementation details of the distributed consistency algorithm, ensuring efficient collaboration between edge nodes and the cloud platform, accurate and reliable data synchronization, and improving the overall stability and response speed of the system.

[0088] When deploying the edge-cloud collaborative control mechanism, edge nodes are responsible for real-time processing of highly time-sensitive control instructions. Edge nodes are usually deployed close to urban components, with low network latency. For example, in the intelligent transportation scenario, edge nodes can receive the status information of traffic lights and data from vehicle detection sensors in real time, and adjust traffic lights in real time according to preset rules. This can quickly respond to changes in traffic conditions and avoid traffic congestion.

[0089] The cloud platform is responsible for global resource coordination. The cloud platform integrates a large amount of computing, storage, and network resources, and can conduct macroscopic management of the entire urban component IoT system. It collects data from each edge node, analyzes the overall operating status and resource requirements of urban components, and then allocates and schedules resources based on this information. For example, when the network bandwidth demand of intelligent devices in a certain area suddenly increases, the cloud platform can allocate bandwidth resources from other idle areas to ensure the normal operation of the devices.

[0090] The Raft protocol is used to achieve data synchronization between edge nodes and the cloud platform. The Raft protocol ensures data consistency in a distributed system through mechanisms such as leader election and log replication. In this system, the cloud platform acts as the leader node, and edge nodes act as follower nodes. The leader receives requests from clients, records data change operations in the log, and then replicates the log to each edge node. Edge nodes execute corresponding operations in the order of the log to ensure data consistency with the cloud platform.

[0091] The heartbeat mechanism is used to detect node abnormal states. The cloud platform periodically sends heartbeat messages to edge nodes, and edge nodes immediately reply after receiving the heartbeat messages. If the cloud platform does not receive a reply from a certain edge node within a certain period of time, it is determined that the edge node is abnormal. At this time, the system will trigger a fault tolerance control strategy. For example, temporarily isolate the abnormal edge node to prevent it from affecting the operation of the entire system, and try to reconnect to the node. If the node failure cannot be recovered, reallocate the tasks it undertakes to ensure the stability and reliability of the system.

[0092] Example 5:

[0093] This embodiment details the specific methods of heterogeneous device protocol adaptation and the working principles of each module of the protocol conversion middleware, solves the problem of incompatible communication protocols of different devices in urban components, and realizes seamless communication and data interaction between devices.

[0094] When implementing heterogeneous device protocol adaptation, a lightweight protocol conversion middleware is constructed. This middleware mainly includes a protocol parsing module, a field mapping module, and a data encapsulation module.

[0095] The protocol parsing module parses the private protocol data frame based on a finite state machine. Different urban components may use their own private communication protocols. The finite state machine parses the data frame bit by bit according to the syntax and semantic rules of the protocol. For example, for a specific private protocol, the finite state machine waits for the start flag of the data frame in the initial state. When the start flag is detected, it enters the data length parsing state and parses the length information of the data frame according to the format specified by the protocol. Then, according to the length information, it enters the data field parsing state and sequentially parses the content of each data field. In this way of state transition, the valid information in the private protocol data frame is accurately extracted.

[0096] The field mapping module dynamically binds the private protocol fields to the standard Internet of Things protocol fields. Since there are differences in the field definitions and formats between the private protocol and the standard Internet of Things protocol, the field mapping module maps the parsed private protocol fields to the standard Internet of Things protocol fields one by one according to the pre-set mapping rules. For example, the field used to represent the device temperature in the private protocol needs to be mapped to the temperature measurement field specified in the standard Internet of Things protocol, and data type and format conversions are performed. This dynamic binding mechanism can adapt to a variety of different private protocols and improve the generality of the protocol conversion middleware.

[0097] The data encapsulation module uses the TLV (Type - Length - Value) format to implement cross - protocol data encapsulation. The TLV format divides the data into three parts: type, length, and value. In the data encapsulation module, the standard Internet of Things protocol data after field mapping is encapsulated in the TLV format. For example, for the water volume data collected by a smart water meter, the type field identifies that the data is water volume data, the length field represents the number of bytes of the data, and the value field stores the actual water volume measurement value. The encapsulated data packet can be transmitted in different network environments and protocol stacks, ensuring the integrity and correctness of the data. At the same time, this middleware supports dynamic protocol plug - and - play and adaptive data encapsulation. When a new private protocol device is connected to the system, the corresponding protocol parsing and mapping rules can be easily added to support the new device.

[0098] The present invention also includes a software - defined urban component Internet of Things control system based on cloud services. The system includes:

[0099] Digital Twin Modeling Module for Urban Components: It is used to integrate real-time sensor data, historical operation data, and urban geographic information system data through a cloud platform to generate a dynamic virtual mapping model, support the fusion of multi-source heterogeneous data, and generate state feature vectors of urban components through spatio-temporal correlation analysis;

[0100] Dynamic Resource Demand Prediction Module: Based on a prediction model of a graph neural network, it receives state feature vectors of urban components, real-time environmental data, and user behavior data, and outputs the resource demand distribution of each urban component in the future time period. The real-time environmental data includes traffic flow, meteorological conditions, and energy consumption data;

[0101] Software-Defined Network Policy Generation Module: According to the resource demand distribution, it dynamically optimizes the network bandwidth allocation policy and task scheduling priority using a reinforcement learning algorithm, and the policy generation process covers cross-layer protocol adaptation and end-to-end delay constraint modeling;

[0102] Edge-Cloud Cooperative Control Mechanism Deployment Module: It uses edge nodes to process high-timeliness control instructions in real time, and uses the cloud platform for global resource coordination, and adopts a distributed consistency algorithm to ensure the state synchronization between the edge and the cloud;

[0103] Heterogeneous Device Protocol Adaptation Module: It constructs a lightweight protocol conversion middleware to uniformly map the private communication protocols of urban components to standard Internet of Things protocols. This middleware supports dynamic protocol plug-and-play and adaptive data encapsulation.

[0104] The implementation manner of this system refers to the above embodiments and will not be elaborated in the specification.

[0105] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0106] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A cloud service-based software-defined city component IoT control method, characterized in that: include: Constructing a digital twin model of urban components, including integrating real-time sensor data, historical operation data and urban geographic information system data through a cloud platform to generate a dynamic virtual mapping model; the digital twin model supports multi-source heterogeneous data fusion and generates a characteristic vector of the state of urban components through spatiotemporal correlation analysis; Perform dynamic resource demand forecasting, including a forecasting model based on a graph neural network, inputting the state feature vector, real-time environmental data and user behavior data, and outputting resource demand distribution of each city component in a future period; the real-time environmental data includes traffic flow, meteorological conditions and energy consumption data; Generating a software-defined network strategy, including dynamically optimizing a network bandwidth allocation strategy and a task scheduling priority using a reinforcement learning algorithm according to the resource demand distribution; the strategy generation process includes cross-layer protocol adaptation and end-to-end delay constraint modeling; Deploy edge-cloud collaborative control mechanisms, including real-time processing of time-sensitive control instructions by edge nodes and global resource coordination through the cloud platform; the collaborative mechanism uses a distributed consistency algorithm to ensure synchronization of edge and cloud states; Implement heterogeneous device protocol adaptation, including building lightweight protocol conversion middleware to uniformly map the private communication protocols of city components to standard IoT protocols; The middleware supports dynamic protocol plug-in and self-adaptive data encapsulation; The construction of the digital twin model includes: using an open source 3D engine to render the physical topology of urban components, combining Apache Kafka to achieve real-time data stream access; using a time series database to store historical operation data, and accelerating feature vector generation through a spatiotemporal indexing algorithm; The structure of the graph neural network prediction model includes: Input layer: receiving the state feature vector, environmental data and user behavior data; Graph construction layer: Generate dynamic graph structure based on the geographical location and functional dependencies of city components; Graph convolution layer: extracts spatial dependency features through multi-layer graph convolution operations; Time series fusion layer: introduce gated recurrent units to fuse time series dynamic features; Output layer: Generates the confidence matrix of resource demand distribution; The spatiotemporal correlation analysis includes: dividing the functional areas of urban components based on the Voronoi diagram, and predicting the state transition probability between areas through a hidden Markov model.

2. The method for controlling the Internet of Things of software-defined city components based on cloud services according to claim 1, characterized in that: The optimization objective function of the reinforcement learning algorithm is: , in, is the immediate reward function, including network load balance, task completion rate and energy efficiency; is the discount factor; and Represent the state space and action space respectively; Indicates the next state and performs actions After that the system transfers to the state, In Status The best possible action to choose from.

3. The method for controlling the Internet of Things of software-defined city components based on cloud services according to claim 1, characterized in that: The distributed consensus algorithm uses the Raft protocol to achieve data synchronization between edge nodes and the cloud platform, and detects node abnormalities through a heartbeat mechanism to trigger a fault-tolerant control strategy.

4. The method for controlling the Internet of Things of software-defined city components based on cloud services according to claim 1, characterized in that: The implementation of the protocol conversion middleware includes: Protocol parsing module: parses private protocol data frames based on finite state machines; Field mapping module: dynamically binds private protocol fields with standard IoT protocol fields; Data encapsulation module: uses TLV format to implement cross-protocol data encapsulation.

5. The method for controlling the Internet of Things of software-defined city components based on cloud services according to claim 1, characterized in that: The formula for modeling the end-to-end delay constraint is: , in, represents the total end-to-end delay, For the The processing delay of a task, is the transmission delay, is the maximum tolerable delay of the system.

6. A cloud service-based software-defined city component IoT control system, characterized in that: include: Urban component digital twin modeling module: used to integrate real-time sensor data, historical operation data and urban geographic information system data through the cloud platform, generate dynamic virtual mapping models, support multi-source heterogeneous data fusion, and generate urban component status feature vectors through spatiotemporal correlation analysis; specifically includes: Use open source 3D engines to render the physical topology of city components, and combine with Apache Kafka to access real-time data streams. Use time series databases to store historical operation data, and use spatiotemporal indexing algorithms to accelerate feature vector generation. The spatiotemporal correlation analysis includes: dividing the functional areas of urban components based on the Voronoi diagram, and predicting the state transition probability between areas through a hidden Markov model; Dynamic resource demand prediction module: Based on the prediction model of graph neural network, it receives the state feature vector of urban components, real-time environmental data and user behavior data, and outputs the resource demand distribution of each urban component in the future period. The real-time environmental data includes traffic flow, meteorological conditions and energy consumption data; The structure of the graph neural network prediction model includes: Input layer: receiving the state feature vector, environmental data and user behavior data; Graph construction layer: Generate dynamic graph structure based on the geographical location and functional dependencies of city components; Graph convolution layer: extracts spatial dependency features through multi-layer graph convolution operations; Time series fusion layer: introduce gated recurrent units to fuse time series dynamic features; Output layer: Generates the confidence matrix of resource demand distribution; Software-defined network policy generation module: Based on the resource demand distribution, the reinforcement learning algorithm is used to dynamically optimize the network bandwidth allocation strategy and task scheduling priority, and the policy generation process covers cross-layer protocol adaptation and end-to-end delay constraint modeling; Edge-cloud collaborative control mechanism deployment module: Use edge nodes to process high-efficiency control instructions in real time, use the cloud platform for global resource coordination, and use a distributed consistency algorithm to ensure the synchronization of edge and cloud status; Heterogeneous device protocol adaptation module: Build lightweight protocol conversion middleware to uniformly map the private communication protocols of urban components to standard IoT protocols. The middleware supports dynamic protocol plug-in and adaptive data encapsulation.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, the cloud service-based software-defined city component Internet of Things control method as described in any one of claims 1-5 is implemented.

Images