A smart home device real-time monitoring method and device, a terminal device and a storage medium

Abstract

The application discloses a kind of intelligent home equipment real-time monitoring method, device, terminal equipment and storage medium, the method comprises: according to the traffic density of network throughput demand distribution map, set several electronic fence nodes;Router is mapped into root node, the electronic fence node is mapped into intermediate node, the intelligent home equipment is mapped into terminal node, and the whole house network topology is obtained in combination with the response message;The whole house network topology is input into the preset GCN network, and the real-time evaluation value of each node is obtained.The application realizes high-precision state evaluation by constructing graph convolution network deep fusion node feature and topological relationship, and can identify potential failure in advance.

Description

Technical Field

[0001] This invention relates to the field of electronic device monitoring, and more particularly to a method, apparatus, terminal device, and storage medium for real-time monitoring of smart home devices. Background Technology

[0002] Existing home network topologies typically use a star topology, with all devices directly connected to the router. This structure works well when the number of devices is small, but as the number of smart devices increases, problems become increasingly apparent: the router's processing capacity is limited, making it unable to handle the monitoring requests of a large number of devices simultaneously; the walls in different rooms cause varying degrees of attenuation of the wireless signal, leading to unstable monitoring in some areas; and the lack of intelligent allocation of network traffic means that high-bandwidth devices (such as cameras and smart TVs) consume a large amount of bandwidth, affecting the monitoring quality of other devices.

[0003] In network stability assessment, existing technologies primarily rely on simple signal strength indexes (RSSI) or connectivity status (online / offline). This approach is too coarse and fails to comprehensively reflect the actual operating status of devices. For example, a device may appear to be "online," but its data transmission latency is high, packet loss rate is large, and the actual service quality has severely degraded. Furthermore, existing monitoring systems lack the ability to dynamically analyze network topology, making it impossible to predict and prevent potential network failures. Summary of the Invention

[0004] This invention provides a method, apparatus, terminal device, and storage medium for real-time monitoring of smart home devices. By constructing a graph convolutional network to deeply fuse node features and topological relationships, it achieves high-precision status assessment and can identify potential faults in advance.

[0005] To achieve the above objectives, a first aspect of this application provides a method for real-time monitoring of smart home devices, comprising: Obtain floor plans of all rooms in the house and the distribution of each smart home device; Based on the number and type of smart home devices in each room, the floor plan of each room is mapped into a network throughput demand distribution map; Several electronic fence nodes are set up according to the traffic density of the network throughput demand distribution map; the electronic fence nodes periodically send handshake signals to other electronic fence nodes in adjacent rooms, other electronic fence nodes in the room, and each of the smart home devices in the room; each of the other electronic fence nodes and each of the smart home devices sends a response message after receiving the handshake signal. The router is mapped as the root node, the electronic fence node is mapped as the intermediate node, and the smart home device is mapped as the terminal node. The whole house network topology is obtained by combining the response message. Input the whole-house network topology diagram into the preset GCN network to obtain the real-time evaluation value of each node; If the real-time evaluation value of an intermediate node is lower than the first preset threshold, adjust the number or relative position of each electronic fence node in the room where the intermediate node is located; if the real-time evaluation value of a terminal node is lower than the second preset threshold, change the working status of the terminal node or replace the terminal node; if the real-time evaluation value of a root node is lower than the third preset threshold, change the working status of the root node or replace the root node.

[0006] In one possible implementation of the first aspect, obtaining the floor plan of each room in the house and the distribution of each smart home device specifically includes: By taking photos with a mobile phone and using image recognition, the room outline and the thickness of the interior walls are extracted to obtain the floor plan of each room. The type of each smart home device is identified by scanning the local area network.

[0007] In one possible implementation of the first aspect, mapping the floor plan of each of the rooms into a network throughput demand distribution map based on the number and type of smart home devices in each room specifically includes: Based on the type of each smart home device and the device fingerprint database, determine the throughput requirements of each smart home device; For each room, the corresponding floor plan is divided into square grid cells; the area of ​​each square grid cell is larger than the area of ​​the smallest smart home device. Based on the thickness of the interior wall, the throughput requirements of the smart home devices adjacent to the interior wall are amplified; adjacent to the interior wall means that the smart home devices are located within the area formed by the square grid units adjacent to the interior wall; For each room, the traffic density of each square grid cell is calculated based on the floor area of ​​each smart home device, the wall penetration loss in the room, and the throughput requirements of the smart home devices, to obtain a network throughput demand distribution map; the floor area is an integer multiple of the area of ​​the square grid cell.

[0008] In one possible implementation of the first aspect, setting up a plurality of electronic fence nodes according to the traffic density of the network throughput demand distribution map specifically includes: Based on the traffic density of the network throughput demand distribution map, identify all first-class connected regions whose traffic density exceeds a preset density threshold; the rated throughput of each electronic fence node is equal to the density threshold. A plurality of electronic fence nodes are evenly distributed within each of the first type of connected regions; the total rated throughput of all electronic fence nodes within each of the first type of connected regions is greater than 1.5 times the flow density of the first type of connected region. An electronic fence node is deployed for each of the second type of connected regions outside of the first type of connected regions.

[0009] In one possible implementation of the first aspect, mapping the router as a root node, the electronic fence node as an intermediate node, and the smart home device as a terminal node, combined with the response message to obtain a whole-house network topology map, specifically includes: The router is mapped as the root node, the electronic fence node is mapped as the intermediate node, and the smart home device is mapped as the terminal node. Based on all the response messages, determine the throughput, communication latency, handshake success rate, and data packet size between different nodes; the edges between different nodes are dashed lines, the thickness of which is set according to the handshake success rate between different nodes, the length of each sub-segment of which is set according to the data packet size between different nodes, the interval of each sub-segment of which is set according to the communication latency between different nodes, and the color of which is set according to the throughput between different nodes. Based on the edge and node characteristics between different nodes, a whole-house network topology diagram is obtained.

[0010] In one possible implementation of the first aspect, the step of inputting the whole-house network topology map into a preset GCN network to obtain real-time evaluation values ​​for each node specifically includes: Based on the whole-house network topology diagram, an adjacency matrix and a node feature matrix are obtained; the adjacency matrix is ​​an N×N matrix, where N is the total number of nodes; the node feature matrix is ​​an N×F matrix, where F is the feature dimension, and the feature dimension is equal to the maximum number of node features of a single node among all nodes. The adjacency matrix and the node feature matrix are input into a preset GCN network to obtain the real-time evaluation value of each node.

[0011] In one possible implementation of the first aspect, the preset GCN network specifically includes: an input layer, a feature preprocessing layer, a dual-channel graph convolutional layer, a node type-aware fusion layer, and an evaluation value generation layer; The input layer is used to input the adjacency matrix and the node feature matrix; The feature preprocessing layer is used to perform Z-score normalization on the root node, local normalization on the intermediate nodes based on deployment location, and group normalization on the terminal nodes based on device type. The dual-channel graph convolutional layer includes a local feature channel and a global feature channel. The local feature channel includes a first GCN network layer, a second GCN network layer, and a third GCN network layer. The first GCN network layer uses ReLU activation to aggregate 1-hop neighbor information. The second GCN network layer uses ReLU activation to aggregate 2-hop neighbor information. The third GCN network layer uses tanh activation to generate local feature representations. The node type-aware fusion layer is used to project the outputs of the local feature channels and the global feature channels onto a type-specific space for processing, respectively. The evaluation value generation layer includes a stability evaluation header, a performance evaluation header, and a performance evaluation header.

[0012] A second aspect of this application provides a real-time monitoring device for smart home devices, comprising: The acquisition module is used to acquire floor plans of each room in the house and the distribution of each smart home device; The first mapping module is used to map the floor plan of each room into a network throughput demand distribution map according to the number and type of smart home devices in each room. The setting module is used to set up a number of electronic fence nodes according to the traffic density of the network throughput demand distribution map; the electronic fence nodes periodically send handshake signals to other electronic fence nodes in adjacent rooms, other electronic fence nodes in the room, and each of the smart home devices in the room; each of the other electronic fence nodes and each of the smart home devices sends a response message after receiving the handshake signal. The second mapping module is used to map the router as the root node, the electronic fence node as the intermediate node, and the smart home device as the terminal node, and to obtain the whole-house network topology map by combining the response message. The evaluation module is used to input the whole-house network topology map into a preset GCN network to obtain the real-time evaluation value of each node. The decision module is used to adjust the number or relative position of each electronic fence node in the room where the intermediate node is located if the real-time evaluation value of an intermediate node is lower than the first preset threshold; to change the working status of a terminal node or replace the terminal node if the real-time evaluation value of a terminal node is lower than the second preset threshold; and to change the working status of a root node or replace the root node if the real-time evaluation value of a root node is lower than the third preset threshold.

[0013] A third aspect of this application provides a terminal device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement a real-time monitoring method for smart home devices as described above.

[0014] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a real-time monitoring method for smart home devices as described above.

[0015] Compared to existing technologies, this invention first obtains a room floor plan and equipment distribution through image recognition, calculates the throughput requirements of each area based on equipment type, and generates a network throughput requirement distribution map. Based on this, electronic fence nodes are dynamically deployed to form a hierarchical network architecture, in which fence nodes periodically send handshake signals to build the network topology. The topology map is input into a customized GCN network to comprehensively evaluate the stability, performance, and load status of the nodes. When the evaluation value is lower than the threshold, the layout of the fence nodes or the working status of the equipment is automatically adjusted.

[0016] The intelligent deployment of electronic fence nodes effectively solves the single-point bottleneck problem of traditional star networks. The wall penetration loss compensation mechanism significantly improves signal coverage quality, especially enhancing connection stability in corners and areas separated by walls. Graph convolutional networks deeply integrate node characteristics and topological relationships, achieving high-precision state assessment and enabling early identification of potential faults. The system supports large-scale concurrent monitoring of devices, significantly enhancing network stability, optimizing bandwidth resource allocation, and greatly improving fault prediction capabilities. This solution operates entirely at the edge, without relying on cloud services, achieving true localized closed-loop optimization. It provides a highly reliable and adaptive network foundation for smart homes, ensuring smooth operation of various smart applications and a qualitatively improved user experience. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a real-time monitoring method for smart home devices according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a real-time monitoring device for smart home devices provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] To resolve the above issues, please refer to [link / reference]. Figure 1 An embodiment of the present invention provides a method for real-time monitoring of smart home devices, comprising: S10. Obtain the floor plan of each room in the house and the distribution of each smart home device.

[0020] S11. Based on the number and type of smart home devices in each room, map the floor plan of each room into a network throughput demand distribution map.

[0021] S12. Set up several electronic fence nodes according to the traffic density of the network throughput demand distribution map; the electronic fence nodes periodically send handshake signals to other electronic fence nodes in adjacent rooms, other electronic fence nodes in the room, and each of the smart home devices in the room, and each of the other electronic fence nodes and each of the smart home devices sends a response message after receiving the handshake signal.

[0022] S13. Map the router as the root node, the electronic fence node as the intermediate node, and the smart home device as the terminal node, and combine the response message to obtain the whole house network topology map.

[0023] S14. Input the whole-house network topology map into the preset GCN network to obtain the real-time evaluation value of each node.

[0024] S15. If the real-time evaluation value of an intermediate node is lower than the first preset threshold, adjust the number or relative position of each electronic fence node in the room where the intermediate node is located; if the real-time evaluation value of a terminal node is lower than the second preset threshold, change the working status of the terminal node or replace the terminal node; if the real-time evaluation value of a root node is lower than the third preset threshold, change the working status of the root node or replace the root node.

[0025] S10-S11 is the spatial modeling and traffic demand analysis process: Room outlines and wall structures are extracted using mobile phone photography combined with computer vision algorithms. Simultaneously, local area network scanning technology automatically identifies the types and locations of devices in each room. Based on a device type database (such as cameras, sensors, smart speakers, etc.), the basic throughput requirements for each type of device are determined. The attenuation effect of wall materials on wireless signals is considered (concrete walls have a higher attenuation coefficient than wooden walls). A gridded calculation method is used to convert the floor plan into a traffic density distribution map. For example, a 4K camera deployed in the living room will create a high traffic density area due to video streaming requirements, while a temperature sensor in the bedroom will form a low-density area.

[0026] The above steps achieve a precise mapping between physical space and network requirements, automatically identifying signal blind spots and high-load areas in the home environment, providing a spatial data foundation for subsequent network optimization. The wall attenuation compensation mechanism effectively solves the problem of ignoring the influence of building structure in traditional network planning, making traffic demand distribution more consistent with actual usage scenarios, and significantly improving the accuracy of network coverage and the rationality of resource allocation.

[0027] S12 is the network architecture deployment process: Based on the traffic density distribution map, high-density areas (such as home theaters and office areas) and low-density areas (such as corridors and storage rooms) are divided. Multiple electronic fence nodes are deployed as needed in high-density areas to form local clusters, while single nodes are deployed in low-density areas. The fence nodes employ a hierarchical handshake mechanism: high-frequency communication with nodes in adjacent rooms (ensuring cross-area connectivity), medium-frequency synchronization with nodes in the same room (maintaining area coordination), and low-frequency interaction with terminal devices (reducing device power consumption). The handshake protocol includes multi-dimensional parameters such as signal strength, latency, and throughput, forming a dynamic network state-aware network.

[0028] The intelligent deployment of electronic fence nodes constructs a hierarchical network architecture, fundamentally solving the performance bottleneck problem caused by the traditional smart home approach of "all devices directly connecting to the router." The layered handshake mechanism ensures network stability while significantly reducing the energy consumption burden on terminal devices and extending the lifespan of battery-powered equipment. This architecture is particularly well-suited to complex house layouts, effectively penetrating multiple walls to achieve seamless coverage throughout the house, laying the network foundation for subsequent accurate assessments.

[0029] S13-S14 is the network topology evaluation process: mapping the physical network into a logical topology graph, where routers act as root nodes (responsible for global coordination), electronic fence nodes act as intermediate nodes (responsible for area management), and smart devices act as terminal nodes (providing specific services); the GCN network aggregates neighbor information through multi-layer graph convolution operations, adopts differentiated evaluation strategies for different node types, and comprehensively outputs scores in three dimensions: stability, performance, and load.

[0030] This assessment mechanism breaks through the limitations of traditional network monitoring, which only focuses on "online / offline" status, and achieves a comprehensive and in-depth assessment of network health. By integrating spatial location, device type, and real-time interactive data, it can accurately identify potential network risk points, such as signal attenuation areas, overloaded nodes, or devices about to fail. The introduction of the GCN network endows the system with "network intelligence," enabling it to understand the interrelationships between nodes, predict network performance trends in advance, and provide a scientific basis for adaptive optimization.

[0031] S15 is an adaptive optimization process that establishes a three-level threshold triggering mechanism. Different thresholds are set for different node types (the root node has the highest threshold, and the terminal node has the lowest). When the evaluation value of an intermediate node falls below the threshold, a gradient descent algorithm is used to recalculate the optimal location of the electronic fence node, or the number of nodes is increased in high-density areas. When the evaluation value of a terminal node is insufficient, the sampling frequency of non-critical equipment (such as environmental sensors) is automatically reduced, or it is recommended to replace critical equipment with insufficient performance (such as security cameras). When the performance of the root node degrades, a backup router is activated or the system switches to edge computing mode to maintain basic services. The optimization process adopts a gradual strategy to avoid network instability.

[0032] The adaptive optimization mechanism enables closed-loop management of network performance, dynamically adjusting resource allocation based on actual usage to ensure that critical applications (such as security monitoring and telemedicine) always receive priority. This mechanism significantly reduces the need for manual intervention, giving the smart home network self-diagnosis and self-repair capabilities. Through differentiated threshold settings, the system can balance performance and energy consumption, extending battery life while ensuring core functions, truly realizing the design concept of "smart network serving smart life" and providing users with a continuous, stable, and seamless smart home experience.

[0033] In summary, this embodiment first acquires the home space structure and device distribution through image recognition and device scanning, and generates an accurate traffic demand distribution map by combining device types and wall characteristics. Based on this distribution map, electronic fence nodes are intelligently deployed to form a hierarchical network architecture, establishing dynamic connections through periodic handshake signals. The physical network is mapped into a weighted topology graph, and a customized GCN network is used for in-depth analysis of node status, breaking through the limitations of traditional "online / offline" monitoring. Finally, node layout or device parameters are automatically adjusted based on the evaluation results to achieve continuous optimization of network performance. The entire system completes closed-loop operation at the edge, without relying on the cloud, and can proactively adapt to changes in the home environment, ensuring that critical applications such as security monitoring and environmental regulation always receive stable and reliable network support, truly realizing the self-sensing, self-evaluation, and self-optimization capabilities of the smart home network.

[0034] For example, obtaining the floor plan of each room in the house and the distribution of each smart home device specifically includes: By taking photos with a mobile phone and using image recognition, the room outline and the thickness of the interior walls are extracted to obtain the floor plan of each room.

[0035] The type of each smart home device is identified by scanning the local area network.

[0036] This step employs multimodal perception fusion technology. First, indoor environmental images are captured using a smartphone camera. A convolutional neural network (CNN) is then used for semantic segmentation to identify room boundaries, door and window locations, and wall structures. Wall thickness is estimated using an edge detection algorithm combined with a deep learning model. The wall edge strength parameter E(x,y) is calculated using the formula E(x,y) = |∇I(x,y)|·W, where I(x,y) is the image pixel intensity, ∇ is the gradient operator, and W is the wall material weighting coefficient (1.2 for concrete, 1.0 for brick, and 0.8 for wood). Simultaneously, the system detects devices within the local area network through ARP scanning, mDNS discovery, and the UPnP protocol. It then uses a device fingerprint database (containing MAC address prefixes, service port numbers, protocol characteristics, etc.) to identify device types. For example, smart cameras typically open port 554 (RTSP protocol), while smart speakers support port 1900 (SSDP protocol).

[0037] This technical solution achieves precise mapping between the physical space of the home and the digital network. It can automatically identify the impact of complex floor plans and wall materials on signal propagation, avoiding the errors and tediousness of traditional manual floor plan drawing. Simultaneously, through multi-protocol fusion scanning, it solves the compatibility issues of smart home devices from different brands, ensuring that all types of devices can be accurately identified and classified. This space-device joint perception mechanism provides precise input data for subsequent network optimization, enabling the system to customize the optimal network configuration for specific home environments. This significantly improves network coverage quality and device connection stability, providing users with a truly intelligent and seamless home network experience.

[0038] For example, mapping the floor plan of each room into a network throughput demand distribution map based on the number and type of smart home devices in each room specifically includes: Based on the type of each smart home device and the device fingerprint database, the throughput requirements of each smart home device are determined.

[0039] For each room, the corresponding floor plan is divided into square grid cells; the area of ​​each square grid cell is larger than the area of ​​the smallest smart home device.

[0040] Based on the thickness of the interior wall, the throughput requirements of the smart home devices adjacent to the interior wall are amplified; adjacent to the interior wall means that the smart home devices are located within the area formed by the square grid units adjacent to the interior wall.

[0041] For each room, the traffic density of each square grid cell is calculated based on the floor area of ​​each smart home device, the wall penetration loss in the room, and the throughput requirements of the smart home devices, to obtain a network throughput demand distribution map; the floor area is an integer multiple of the area of ​​the square grid cell.

[0042] This step constructs a precise mapping model between physical space and network requirements. First, the system matches the baseline throughput requirement B_d corresponding to each device type using a device fingerprint database. This fingerprint database contains unique characteristics such as device protocol features, service ports, and manufacturer identifiers. For example, 4K smart cameras have a higher B_d value due to their video streaming requirements, while temperature and humidity sensors have a lower B_d value. Second, the room floor plan is divided into standardized grid units. The grid side length L_grid must satisfy L_grid > √A_min, where A_min is the minimum device footprint, ensuring the rationality of spatial discretization. A wall effect compensation mechanism is implemented: for devices located adjacent to walls (defined as areas less than twice the grid unit distance from the wall), their throughput requirement is amplified using the formula B'_d = B_d × (1 + k·T_wall), where T_wall is the wall thickness (in meters), and k is the wall material attenuation coefficient (0.8 for concrete, 0.6 for brick, and 0.3 for wood). Finally, the flow density ρ(i,j) of each grid cell (i,j) is calculated by weighted superposition: ρ(i,j) = Σ [B'_d × w_distance × w_penetration], where w_distance is the distance attenuation weight from the device to the grid center, and w_penetration is the wall penetration loss weight, taking into account the device footprint, signal propagation characteristics and spatial distribution characteristics.

[0043] This mapping mechanism breaks through the limitations of the simple homogeneous assumptions in traditional network planning, enabling refined and spatial modeling of home network needs. The wall effect compensation mechanism effectively solves the coverage blind spot problem caused by signal attenuation. Especially for critical devices such as security cameras or smart locks near load-bearing walls, the system can predict and compensate for signal loss, avoiding connection instability caused by wall obstruction. The gridded calculation method transforms continuous spatial needs into discrete density distributions, providing a scientific basis for the precise deployment of subsequent electronic fence nodes. This physical environment-based traffic modeling method makes network resource allocation more closely aligned with actual usage scenarios, significantly improving service quality in weak signal areas while avoiding resource waste in strong signal areas. It achieves intelligent allocation and efficient utilization of network resources, laying a solid foundation for building a stable and reliable smart home network.

[0044] For example, setting up several electronic fence nodes based on the traffic density of the network throughput demand distribution map specifically includes: Based on the traffic density of the network throughput demand distribution map, identify all first-class connected regions whose traffic density exceeds a preset density threshold; the rated throughput of each electronic fence node is equal to the density threshold.

[0045] Several electronic fence nodes are evenly distributed within each of the first type of connected regions; the total rated throughput of all electronic fence nodes within each of the first type of connected regions is greater than 1.5 times the flow density of the first type of connected region.

[0046] An electronic fence node is deployed for each of the second type of connected regions outside of the first type of connected regions.

[0047] This step employs an adaptive deployment strategy based on traffic density. First, the system performs connectivity analysis on the network throughput demand distribution map and uses a threshold segmentation algorithm to identify traffic densities exceeding a preset density threshold. A continuous region (defined as a first-class connected region), where This represents the maximum throughput that a single electronic fence node can stably handle. Secondly, within each type I connected region, multiple electronic fence nodes are evenly distributed using the Poisson disk sampling algorithm, with the number of nodes... From the formula Confirmed, among which A factor of 1.5 is applied to the total traffic density of the area to ensure sufficient redundancy capacity for handling traffic peaks. Finally, for the remaining space not covered by the first type of area (the second type of connected area), a center-point positioning strategy is adopted, deploying a single electronic fence node at the geometric center of each area to form a basic coverage network. For example, in a home theater area (high traffic density), the system will deploy 3-4 fence nodes to form a cluster; while in a corridor or storage room (low traffic density), only 1 node will be deployed to provide basic coverage.

[0048] This deployment strategy achieves intelligent allocation and dynamic balancing of network resources. The multi-node cluster design in high-traffic areas effectively solves the performance bottleneck problem of traditional single-point coverage. The 1.5 times redundancy capacity mechanism ensures that the system remains stable and smooth even when users simultaneously use high-bandwidth applications such as 4K video and online games. Single-point coverage in low-traffic areas avoids resource waste and extends the lifespan of equipment. This layered deployment architecture is particularly suitable for the complex layout of modern homes, automatically adapting to the functional characteristics of different rooms (such as the high concurrency requirements of the living room and the low power consumption requirements of the bedroom), forming a network coverage that is both efficient and economical. The intelligent layout of electronic fence nodes not only optimizes signal coverage quality but also reduces the load on routers through regional management, giving the entire smart home network stronger anti-interference and self-healing capabilities, providing a solid and reliable network foundation for various smart applications.

[0049] For example, mapping the router as the root node, the electronic fence node as an intermediate node, and the smart home device as a terminal node, combined with the response message to obtain the whole-house network topology map, specifically includes: The router is mapped as the root node, the electronic fence node is mapped as the intermediate node, and the smart home device is mapped as the terminal node. Based on all the response messages, determine the throughput, communication latency, handshake success rate, and data packet size between different nodes; the edges between different nodes are dashed lines, the thickness of which is set according to the handshake success rate between different nodes, the length of each sub-segment of which is set according to the data packet size between different nodes, the interval of each sub-segment of which is set according to the communication latency between different nodes, and the color of which is set according to the throughput between different nodes. Based on the edge and node characteristics between different nodes, a whole-house network topology diagram is obtained.

[0050] First, the system maps network devices into a tree-like topology according to functional hierarchy: the home router acts as the root node (responsible for global coordination and external network connection), the electronic fence node acts as the intermediate node (responsible for area management and data aggregation), and various smart home devices (such as cameras, sensors, and smart appliances) act as terminal nodes (providing specific service functions). Second, based on the response messages of periodic handshake signals, the system extracts four key network parameters in real time: handshake success rate. (Ratio of successful responses to total requests), communication latency (Signal round-trip time), throughput (Amount of data transmitted per unit time), data packet size (Data volume per transmission). Topological edges are constructed using multi-dimensional visual coding technology: edge thickness. The success rate of handshakes increases monotonically. The thickest time, (Not visible at times); Sub-segment length Positively correlated with data packet size; sub-segment spacing Positively correlated with communication latency; edge color A gradient color mapping is used (blue indicates low throughput, and red indicates high throughput). For example, when a thick red dashed line appears between the smart refrigerator in the kitchen and the electronic fence node, but the sub-segments are spaced far apart, it visually indicates that the connection has high throughput but severe latency.

[0051] This topology construction mechanism enables multi-dimensional visualization of network status, breaking through the limitations of traditional network monitoring's single "connected / disconnected" state and transforming complex network performance parameters into an intuitive visual language. Maintenance personnel can quickly identify network bottlenecks without specialized tools: dense, thick red lines indicate high-load areas, sparse, thin blue lines suggest inefficient connections, and irregular sub-segment intervals reveal latency jitter issues. This deeply visualized topology map not only improves the efficiency of network fault location but also provides non-technical users with an intuitive interface to understand network health. The multi-parameter fusion visual encoding method enables the system to simultaneously monitor the interaction status of hundreds of nodes. Even in the context of increasingly complex home network environments (such as multi-layered walls and a surge in devices), it maintains accurate perception and rapid response capabilities for network behavior, providing a high-quality data foundation for subsequent intelligent optimization.

[0052] For example, the step of inputting the whole-house network topology map into a preset GCN network to obtain the real-time evaluation value of each node specifically includes: Based on the whole-house network topology diagram, an adjacency matrix and a node feature matrix are obtained; the adjacency matrix is ​​an N×N matrix, where N is the total number of nodes; the node feature matrix is ​​an N×F matrix, where F is the feature dimension, and the feature dimension is equal to the maximum number of node features of a single node among all nodes. The adjacency matrix and the node feature matrix are input into a preset GCN network to obtain the real-time evaluation value of each node.

[0053] Transform the visualized topology graph into a mathematically computable matrix representation: the adjacency matrix. Describes the strength of connections between nodes, where the element From the formula Calculations show that For nodes and The success rate of handshake between partners (0-1). Communication delay (milliseconds) For maximum tolerable delay, For throughput (Mbps). For maximum throughput, These are the weight coefficients (summing to 1). Node feature matrix. Integrating multi-dimensional attributes, among which For the largest feature dimension (e.g., the root node contains 16 features such as CPU load and memory usage, and the terminal node contains 12 features such as battery status and QoS level), missing features are filled using type mean. Subsequently, the customized GCN network extracts deep features through multi-layer graph convolution operations: the first... Hidden layer representation ,in For an adjacency matrix with self-loops ( (Using an identity matrix to ensure nodes retain their own characteristics) For degree matrix, For learnable weights, The activation function is non-linear. Finally, the node type-aware evaluation head outputs a comprehensive score from 0 to 100. For example, a living room camera node might receive 75 points due to high latency and fluctuating throughput, while a bedroom sensor might receive 92 points due to stable low load.

[0054] This evaluation mechanism overcomes the limitations of isolated node analysis in traditional network monitoring, achieving a global and interconnected understanding of network status. By deeply integrating topology and node characteristics, the system can accurately identify hidden problems that are difficult to detect using traditional methods. For example, a single electronic fence node may appear to be functioning normally, but the performance of its neighboring nodes may be generally degraded, indicating a potential fault. Similarly, moderate router load may be affected by the cumulative latency on the critical path, impacting user experience. This intelligent evaluation not only focuses on instantaneous status but also combines historical trends to predict performance changes, enabling the system to perform proactive maintenance. This provides a scientific basis for subsequent precise optimization, significantly improving the reliability of smart home networks and the continuity of user services, ensuring stable support for critical applications such as security monitoring and environmental control.

[0055] For example, the preset GCN network has the following specific structure: an input layer, a feature preprocessing layer, a dual-channel graph convolutional layer, a node type-aware fusion layer, and an evaluation value generation layer; The input layer is used to input the adjacency matrix and the node feature matrix; The feature preprocessing layer is used to perform Z-score normalization on the root node, local normalization on the intermediate nodes based on deployment location, and group normalization on the terminal nodes based on device type. The dual-channel graph convolutional layer includes a local feature channel and a global feature channel. The local feature channel includes a first GCN network layer, a second GCN network layer, and a third GCN network layer. The first GCN network layer uses ReLU activation to aggregate 1-hop neighbor information. The second GCN network layer uses ReLU activation to aggregate 2-hop neighbor information. The third GCN network layer uses tanh activation to generate local feature representations. The node type-aware fusion layer is used to project the outputs of the local feature channels and the global feature channels onto a type-specific space for processing, respectively. The evaluation value generation layer includes a stability evaluation header, a performance evaluation header, and a performance evaluation header.

[0056] This GCN network adopts a hierarchical heterogeneous architecture, specifically tailored for the complexity of smart home networks. The input layer receives an adjacency matrix A (describing node connection strength) and a node feature matrix X (integrating multi-dimensional device attributes). The feature preprocessing layer implements differentiated standardization: Z-score standardization is used for root nodes (routers) to eliminate dimensional differences in metrics such as CPU load and memory usage; location-aware standardization is used for intermediate nodes (electronic fences), establishing a local coordinate system centered on the deployment room to make node features comparable within the same area; and terminal nodes (smart devices) are grouped and standardized according to device type (e.g., cameras, sensors, appliances) to retain the behavioral commonalities of similar devices. Dual-channel graph convolutional layers process local and global information in parallel: the local channel aggregates neighbor information hierarchically through three GCN layers (the first layer aggregates directly connected devices, the second layer covers cross-room influences, and the third layer generates device group features); the global channel uses a graph attention mechanism to dynamically learn key connection weights. The node type-aware fusion layer projects the outputs of both channels into a type-specific space. The projection matrix W_type is dynamically selected based on the node category (W_root for root nodes focuses on global load, while W_terminal for terminal nodes emphasizes service quality). The evaluation value generation layer contains three dedicated evaluation headers that calculate stability (interference resistance), performance (response quality), and load (resource utilization) scores, which are then weighted and fused into a comprehensive evaluation value. For example, when the score of a bedroom camera drops, the system not only analyzes its own status but also evaluates the load of its connected fence nodes and routers, distinguishing between equipment failure and network congestion.

[0057] This architecture enables deep understanding and accurate assessment of smart home networks. A differentiated preprocessing mechanism effectively addresses the issue of varying feature scales among heterogeneous devices, allowing for collaborative analysis of high-dimensional system metrics from routers and simple status data from sensors within the same framework. The dual-channel design balances local details with global structure, capturing the collaborative behavior of device groups within a room while identifying network bottlenecks across regions. Type-aware fusion ensures targeted assessments, avoiding the bias of a "one-size-fits-all" approach. This multi-layered, multi-dimensional assessment mechanism allows the system to distinguish between apparent problems and root causes, such as identifying seemingly normal camera lag as actually stemming from router overload rather than device malfunction. The multi-assessment head structure provides fine-grained health status insights, offering precise guidance for subsequent optimization strategies, significantly improving network self-healing capabilities and user experience continuity. It ensures that critical applications such as security and health monitoring are always prioritized, giving smart home networks true intelligent capabilities of "understanding, judging, and making decisions."

[0058] Compared to existing technologies, this invention first obtains a room floor plan and equipment distribution through image recognition, calculates the throughput requirements of each area based on equipment type, and generates a network throughput requirement distribution map. Based on this, electronic fence nodes are dynamically deployed to form a hierarchical network architecture, in which fence nodes periodically send handshake signals to build the network topology. The topology map is input into a customized GCN network to comprehensively evaluate the stability, performance, and load status of the nodes. When the evaluation value is lower than the threshold, the layout of the fence nodes or the working status of the equipment is automatically adjusted.

[0059] The intelligent deployment of electronic fence nodes effectively solves the single-point bottleneck problem of traditional star networks. The wall penetration loss compensation mechanism significantly improves signal coverage quality, especially enhancing connection stability in corners and areas separated by walls. Graph convolutional networks deeply integrate node characteristics and topological relationships, achieving high-precision state assessment and enabling early identification of potential faults. The system supports large-scale concurrent monitoring of devices, significantly enhancing network stability, optimizing bandwidth resource allocation, and greatly improving fault prediction capabilities. This solution operates entirely at the edge, without relying on cloud services, achieving true localized closed-loop optimization. It provides a highly reliable and adaptive network foundation for smart homes, ensuring smooth operation of various smart applications and a qualitatively improved user experience.

[0060] Please see Figure 2 An embodiment of this application provides a real-time monitoring device for smart home devices, comprising: an acquisition module 20, a first mapping module 21, a setting module 22, a second mapping module 23, an evaluation module 24, and a decision module 25.

[0061] The acquisition module 20 is used to acquire floor plans of each room in the house and the distribution of each smart home device.

[0062] The first mapping module 21 is used to map the floor plan of each room into a network throughput demand distribution map according to the number and type of smart home devices in each room.

[0063] Setting module 22 is used to set up a number of electronic fence nodes according to the traffic density of the network throughput demand distribution map; the electronic fence nodes periodically send handshake signals to other electronic fence nodes in adjacent rooms, other electronic fence nodes in the room, and each of the smart home devices in the room, and each of the other electronic fence nodes and each of the smart home devices sends a response message after receiving the handshake signal.

[0064] The second mapping module 23 is used to map the router as the root node, the electronic fence node as the intermediate node, and the smart home device as the terminal node, and to obtain a whole-house network topology map by combining the response message.

[0065] Evaluation module 24 is used to input the whole-house network topology map into a preset GCN network to obtain the real-time evaluation value of each node.

[0066] The decision module 25 is used to adjust the number or relative position of each electronic fence node in the room where the intermediate node is located if the real-time evaluation value of an intermediate node is lower than the first preset threshold; to change the working status of a terminal node or replace the terminal node if the real-time evaluation value of a terminal node is lower than the second preset threshold; and to change the working status of a root node or replace the root node if the real-time evaluation value of a root node is lower than the third preset threshold.

[0067] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the drug storage management device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0068] Compared to existing technologies, this invention first obtains a room floor plan and equipment distribution through image recognition, calculates the throughput requirements of each area based on equipment type, and generates a network throughput requirement distribution map. Based on this, electronic fence nodes are dynamically deployed to form a hierarchical network architecture, in which fence nodes periodically send handshake signals to build the network topology. The topology map is input into a customized GCN network to comprehensively evaluate the stability, performance, and load status of the nodes. When the evaluation value is lower than the threshold, the layout of the fence nodes or the working status of the equipment is automatically adjusted.

[0069] The intelligent deployment of electronic fence nodes effectively solves the single-point bottleneck problem of traditional star networks. The wall penetration loss compensation mechanism significantly improves signal coverage quality, especially enhancing connection stability in corners and areas separated by walls. Graph convolutional networks deeply integrate node characteristics and topological relationships, achieving high-precision state assessment and enabling early identification of potential faults. The system supports large-scale concurrent monitoring of devices, significantly enhancing network stability, optimizing bandwidth resource allocation, and greatly improving fault prediction capabilities. This solution operates entirely at the edge, without relying on cloud services, achieving true localized closed-loop optimization. It provides a highly reliable and adaptive network foundation for smart homes, ensuring smooth operation of various smart applications and a qualitatively improved user experience.

[0070] One embodiment of this application provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the real-time monitoring method for smart home devices as described above.

[0071] One embodiment of this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the real-time monitoring method for smart home devices as described above.

[0072] The computer device may be a smartphone, tablet, desktop computer, or cloud server, among other computing devices. This computer device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the figures are merely examples of computer devices and do not constitute a limitation on the computer device. It may include more or fewer components than illustrated, or a combination of certain components, or different components, such as input / output devices, network access devices, etc.

[0073] The processor referred to can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, 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, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0074] In some embodiments, the memory may be an internal storage unit of the computer device, such as a hard drive or RAM. In other embodiments, the memory may be an external storage device of the computer device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory may include both internal and external storage units of the computer device. The memory is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory can also be used to temporarily store data that has been output or will be output.

[0075] This application provides a computer program product that, when run on a computer device, enables the computer device to execute the steps described in the various method embodiments above.

[0076] In the several embodiments provided in this application, it will be understood that each block in the flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the figures. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved.

[0077] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0078] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring of smart home devices, characterized in that, include: Obtain floor plans of all rooms in the house and the distribution of each smart home device; Based on the number and type of smart home devices in each room, the floor plan of each room is mapped into a network throughput demand distribution map; Several electronic fence nodes are set up according to the traffic density of the network throughput demand distribution map; the electronic fence nodes periodically send handshake signals to other electronic fence nodes in adjacent rooms, other electronic fence nodes in the room, and each of the smart home devices in the room; each of the other electronic fence nodes and each of the smart home devices sends a response message after receiving the handshake signal. The process involves mapping routers as root nodes, electronic fence nodes as intermediate nodes, and smart home devices as terminal nodes, and combining this with the response messages to obtain a whole-house network topology. Specifically, this includes: mapping routers as root nodes, electronic fence nodes as intermediate nodes, and smart home devices as terminal nodes; determining the throughput, communication latency, handshake success rate, and data packet size between different nodes based on all the response messages; using dashed lines as the edges between different nodes, where the thickness of the dashed line is set according to the handshake success rate, the length of each sub-segment is set according to the data packet size, the interval between each sub-segment is set according to the communication latency, and the color of the dashed line is set according to the throughput; and obtaining the whole-house network topology based on the edges and node characteristics. Input the whole-house network topology diagram into the preset GCN network to obtain the real-time evaluation value of each node; If the real-time evaluation value of an intermediate node is lower than the first preset threshold, adjust the number or relative position of each electronic fence node in the room where the intermediate node is located; if the real-time evaluation value of a terminal node is lower than the second preset threshold, change the working status of the terminal node or replace the terminal node; if the real-time evaluation value of a root node is lower than the third preset threshold, change the working status of the root node or replace the root node.

2. The real-time monitoring method for smart home devices as described in claim 1, characterized in that, The acquisition of floor plans of each room in the house and the distribution of each smart home device specifically includes: By taking photos with a mobile phone and using image recognition, the room outline and the thickness of the interior walls are extracted to obtain the floor plan of each room. The type of each smart home device is identified by scanning the local area network.

3. The real-time monitoring method for smart home devices as described in claim 2, characterized in that, The step of mapping the floor plan of each room into a network throughput demand distribution map based on the number and type of smart home devices in each room specifically includes: Based on the type of each smart home device and the device fingerprint database, determine the throughput requirements of each smart home device; For each room, the corresponding floor plan is divided into square grid cells; the area of ​​each square grid cell is larger than the area of ​​the smallest smart home device. Based on the thickness of the interior wall, the throughput requirements of the smart home devices adjacent to the interior wall are amplified; adjacent to the interior wall means that the smart home devices are located within the area formed by the square grid units adjacent to the interior wall; For each room, the traffic density of each square grid cell is calculated based on the floor area of ​​each smart home device, the wall penetration loss in the room, and the throughput requirements of the smart home devices, to obtain a network throughput demand distribution map; the floor area is an integer multiple of the area of ​​the square grid cell.

4. The real-time monitoring method for smart home devices as described in claim 1, characterized in that, The step of setting up several electronic fence nodes based on the traffic density of the network throughput demand distribution map specifically includes: Based on the traffic density of the network throughput demand distribution map, identify all first-class connected regions whose traffic density exceeds a preset density threshold; the rated throughput of each electronic fence node is equal to the density threshold. A plurality of electronic fence nodes are evenly distributed within each of the first type of connected regions; the total rated throughput of all electronic fence nodes within each of the first type of connected regions is greater than 1.5 times the flow density of the first type of connected region. An electronic fence node is deployed for each of the second type of connected regions outside of the first type of connected regions.

5. A real-time monitoring method for smart home devices as described in claim 1, characterized in that, The step of inputting the whole-house network topology map into a preset GCN network to obtain the real-time evaluation value of each node specifically includes: Based on the whole-house network topology diagram, an adjacency matrix and a node feature matrix are obtained; the adjacency matrix is ​​an N×N matrix, where N is the total number of nodes; the node feature matrix is ​​an N×F matrix, where F is the feature dimension, and the feature dimension is equal to the maximum number of node features of a single node among all nodes. The adjacency matrix and the node feature matrix are input into a preset GCN network to obtain the real-time evaluation value of each node.

6. The real-time monitoring method for smart home devices as described in claim 5, characterized in that, The preset GCN network has the following specific structure: an input layer, a feature preprocessing layer, a dual-channel graph convolutional layer, a node type-aware fusion layer, and an evaluation value generation layer. The input layer is used to input the adjacency matrix and the node feature matrix; The feature preprocessing layer is used to perform Z-score normalization on the root node, local normalization on the intermediate nodes based on deployment location, and group normalization on the terminal nodes based on device type. The dual-channel graph convolutional layer includes a local feature channel and a global feature channel. The local feature channel includes a first GCN network layer, a second GCN network layer, and a third GCN network layer. The first GCN network layer uses ReLU activation to aggregate 1-hop neighbor information. The second GCN network layer uses ReLU activation to aggregate 2-hop neighbor information. The third GCN network layer uses tanh activation to generate local feature representations. The node type-aware fusion layer is used to project the outputs of the local feature channels and the global feature channels onto a type-specific space for processing, respectively. The evaluation value generation layer includes a stability evaluation header, a performance evaluation header, and a performance evaluation header.

7. A real-time monitoring device for smart home devices, characterized in that, include: The acquisition module is used to acquire floor plans of each room in the house and the distribution of each smart home device; The first mapping module is used to map the floor plan of each room into a network throughput demand distribution map according to the number and type of smart home devices in each room. The setting module is used to set up a number of electronic fence nodes according to the traffic density of the network throughput demand distribution map; the electronic fence nodes periodically send handshake signals to other electronic fence nodes in adjacent rooms, other electronic fence nodes in the room, and each of the smart home devices in the room; each of the other electronic fence nodes and each of the smart home devices sends a response message after receiving the handshake signal. The second mapping module is used to map the router as the root node, the electronic fence node as the intermediate node, and the smart home device as the terminal node, and to obtain a whole-house network topology map by combining the response messages. Specifically, this includes: mapping the router as the root node, the electronic fence node as the intermediate node, and the smart home device as the terminal node; determining the throughput, communication latency, handshake success rate, and data packet size between different nodes based on all the response messages; the edges between different nodes are represented by dashed lines, the thickness of which is set according to the handshake success rate, the length of each sub-segment of which is set according to the data packet size, the interval between each sub-segment of which is set according to the communication latency, and the color of which is set according to the throughput; and obtaining the whole-house network topology map based on the edges between different nodes and the node characteristics. The evaluation module is used to input the whole-house network topology map into a preset GCN network to obtain the real-time evaluation value of each node. The decision module is used to adjust the number or relative position of each electronic fence node in the room where the intermediate node is located if the real-time evaluation value of an intermediate node is lower than the first preset threshold; to change the working status of a terminal node or replace the terminal node if the real-time evaluation value of a terminal node is lower than the second preset threshold; and to change the working status of a root node or replace the root node if the real-time evaluation value of a root node is lower than the third preset threshold.

8. A terminal device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement a real-time monitoring method for a smart home device as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements a real-time monitoring method for smart home devices as described in any one of claims 1 to 6.

Images