Sensor network topological structure optimization method, electronic equipment and storage medium

Abstract

The invention relates to the technical field of network topology optimization, in particular to a sensor network topology structure optimization method, electronic equipment and a storage medium, and the method comprises the steps: constructing an initial connection graph according to the connection relation of a plurality of monitoring sensors and edge nodes, the method comprises the following steps: acquiring attribute characteristics and path characteristics of a monitoring sensor, generating an initial network topology structure according to the acquired attribute characteristics and path characteristics of the monitoring sensor, inputting the initial network topology structure into a GNN model, training the GNN model according to an attribute characteristic vector and a path characteristic vector to obtain a target network topology structure, and when a fault of a topological graph node in the target network topology structure is monitored, sending the fault to the GNN model. Obtaining a network topology prediction structure according to the actually measured parameter information of the plurality of topological graph nodes so as to optimize the network topology structure; according to the invention, the network topology structure can be dynamically adjusted according to the abnormal condition of the nodes of the topological graph, the flexibility and adaptive capacity of the network are greatly improved, and the continuous operation of the network and the stability of data transmission are further ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of network topology optimization, and in particular to a sensor network topology optimization method, electronic equipment and storage medium. Background Art

[0002] In the field of infrastructure monitoring, IoT sensor networks are used to monitor the health of infrastructure, such as monitoring the vibration, stress, temperature, etc. of building facilities or underground mines. However, most traditional sensor networks rely on predefined network topology structures, with low data transmission efficiency, and fixed communication paths between sensor nodes, poor flexibility and high communication energy consumption. When some nodes fail, it is difficult to cope with dynamic environmental changes, resulting in interruptions in data transmission. In current optimization algorithms, such as the ant colony algorithm, re-iteration and regression processing are required during optimization. Since the network topology of the facility monitoring system often involves a large number of nodes and a complex structure, the optimization process will cause long-term interruptions in data transmission, affecting the sustainability and reliability of sensor monitoring. Therefore, how to optimize the topology of the sensor network and solve the above problems is currently an urgent problem to be solved. Summary of the invention

[0003] In response to the above technical problems, the present invention provides a sensor network topology optimization method, an electronic device and a storage medium, which can dynamically adjust the network topology according to the abnormal conditions of the topology map nodes, greatly improving the flexibility and adaptability of the network, thereby ensuring the continuous operation of the network and the stability of data transmission.

[0004] According to a first aspect of the present invention, a method for optimizing a sensor network topology is provided, comprising the following steps: An initial connection graph is constructed based on the wireless communication relationship between a number of monitoring sensors installed in the target area and the communication relationship between each monitoring sensor and an edge node; the initial connection graph includes a number of sensor nodes and connection lines between the sensor nodes.

[0005] Based on the edge nodes, the attribute characteristics of the monitoring sensor corresponding to each sensor node and the path characteristics between the two monitoring sensors corresponding to each connecting line are obtained. The initial network topology is generated by combining the initial connection graph and input into the GNN model.

[0006] According to the attribute feature vector corresponding to each topology graph node and the path feature vector corresponding to each edge in the initial network topology structure, the GNN model is trained to optimize the initial network topology structure to the target network topology structure.

[0007] When a fault is detected in any topology map node, a network topology prediction structure is obtained based on measured parameter information of several topology map nodes in the target network topology structure except the faulty topology map node.

[0008] According to the network topology prediction structure, the connection mode between several topology map nodes in the target network topology structure is dynamically adjusted through edge nodes to optimize the target network topology structure.

[0009] According to a second aspect of the present invention, a non-transitory computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or the at least one program is loaded and executed by a processor to implement the above-mentioned sensor network topology optimization method.

[0010] According to a third aspect of the present invention, there is provided an electronic device comprising a processor and the above-mentioned non-transitory computer-readable storage medium.

[0011] The present invention has at least the following beneficial effects: The sensor network topology optimization method described in the embodiment of the present invention first constructs an initial connection graph according to the connection relationship between several monitoring sensors and the connection relationship with the edge nodes, and generates an initial network topology structure according to the acquired attribute characteristics and sensor path characteristics of the monitoring sensors to input into the GNN model, and trains the GNN model according to the acquired attribute feature vector corresponding to each sensor node and each path characteristic vector to obtain the target network topology structure. By training the GNN model, the real-time optimal data transmission path can be calculated, which greatly improves the flexibility and adaptability of the network, and at the same time improves the data transmission efficiency and reduces communication delay and consumption. When a topology map node in the target network topology structure is monitored to fail, the network topology prediction structure is obtained according to the measured parameter information of several topology map nodes to optimize the network topology structure. In this process, the GNN model can dynamically adjust the connection mode between nodes in real time, so that the edge node can quickly update the topology structure of the sensor network, avoid the transmission interruption of data tasks, and ensure the continuous operation of the network and the stability of data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1A flowchart of a sensor network topology optimization method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0015] The embodiment of the present invention provides a method for optimizing the topology of a sensor network. Figure 1 As shown, the method comprises the following steps: S100, construct an initial connection diagram based on the wireless communication relationship between several monitoring sensors installed in the target area, and the communication relationship between each monitoring sensor and the edge node; the initial connection diagram includes several sensor nodes and connection lines between sensor nodes; it can be understood that: the sensor node refers to the monitoring sensor ID, and the monitoring sensor ID is used as a unique identifier of the monitoring sensor.

[0016] Specifically, the target area refers to the area where the target facility is located; it can be understood that the monitoring sensors installed in the target area refer to the monitoring sensors installed in the target facility in the target area. For example, the target facility can be a high-rise building, an underground mine, etc.

[0017] Furthermore, the wireless communication relationship between the plurality of monitoring sensors refers to a relationship connected via a wireless communication protocol, wherein the wireless communication protocol may be WiFi, Zigbee and LoRa. In addition, the communication relationship between the monitoring sensor and the edge node may be wireless communication or wired communication.

[0018] As mentioned above, since the input and output of the GNN model are both graph structures, the preliminary construction of the initial connection graph is completed according to the wireless communication relationship between several monitoring sensors and the communication relationship between the monitoring sensors and the edge nodes, which can meet the input conditions of the GNN model and provide a basis for the subsequent generation of the network topology structure.

[0019] S200, based on the edge nodes, obtain the attribute characteristics of the monitoring sensor corresponding to each sensor node and the path characteristics between the two monitoring sensors corresponding to each connecting line, combine the initial connection graph to generate the initial network topology structure and input it into the GNN model; it can be understood as: taking the attribute characteristics of the monitoring sensor as the input characteristics of the corresponding sensor node, the path characteristics between the two monitoring sensors as the input characteristics of the edge, and combining the feature vector with the connection relationship between the sensor nodes to generate the initial network topology structure.

[0020] Among them, edge node refers to a business platform built on the edge of the network close to the user, providing storage, computing, network and other resources, sinking some key business applications to the edge of the access network to reduce the bandwidth and delay loss caused by network transmission and multi-level forwarding. It is a professional term known to those skilled in the art and will not be repeated here. In this application, it is used to receive raw data from monitoring sensors and perform preliminary data processing and network topology adjustment tasks.

[0021] Specifically, the attribute characteristics of the monitoring sensor include location information, task load, remaining power, sensor type, and wireless signal strength between the monitoring sensor and the edge node.

[0022] Furthermore, the location information is the physical location of the monitoring sensor, for example, the floor, longitude and latitude collected during installation, and the adjacent nodes can be determined.

[0023] Furthermore, the location information of the monitoring sensor is obtained through the following steps: S201, for any monitoring sensor, the position of the monitoring sensor collected in advance during installation and the center point position of the position reported by the monitoring sensor are used as the position information of the monitoring sensor. In a specific implementation, the installation position is generally based on the position information on the engineering drawing. When the actual installation position is inconsistent with the position marked on the engineering drawing, the actual installation position information shall prevail.

[0024] S202: If the location data reported by the monitoring sensor is not received, the distance between the monitoring sensor and the edge node and the distance between the monitoring sensor and the monitoring sensor corresponding to the neighbor node are calculated based on the signal strength between the corresponding sensor node and the edge node and the signal strength between the sensor node and the neighbor node.

[0025] Specifically, the distance D1 between the monitoring sensor and the edge node meets the following conditions: D1=10^((abs(RSSI 1 )-A) / (10×n)), where abs() is the absolute value function, RSSI 1 To monitor the signal strength between the sensor and the edge node, A is the signal strength when the transmitter and receiver are one meter apart, and n is the environmental attenuation factor.

[0026] Furthermore, the distance D2 between the monitoring sensor and the monitoring sensor corresponding to the neighboring node meets the following conditions: D2=10^((abs(RSSI 2 )-A) / (10×n)), where RSSI 2The signal strength between the monitoring sensor and the monitoring sensor corresponding to the neighboring node.

[0027] S203, according to the location information of the edge node, the distance between each monitoring sensor and the edge node, and the distance between each monitoring sensor and the monitoring sensor corresponding to the neighboring node, the location information of each monitoring sensor is obtained after finding the optimal solution.

[0028] As mentioned above, when the monitoring sensor does not have the function of reporting location information, the location information manually collected during installation has the disadvantage of strong subjectivity. The distance between each monitoring sensor and the edge node and the distance between the monitoring sensors are calculated according to the signal strength. After iterative processing, the location information of each monitoring sensor can be obtained. In this way, the collected location information is more accurate, the accuracy of the samples is increased, and the trained GNN model has higher reliability.

[0029] Specifically, the task load is the current workload of the monitoring sensor, which represents the amount of data being processed by the monitoring sensor or the frequency of monitoring tasks.

[0030] Furthermore, all sensor types include but are not limited to vibration sensors, stress sensors, temperature and humidity sensors, and displacement sensors.

[0031] To better understand the composition of attribute feature data, refer to the data shown in Table 1 below:

[0032] Table 1 shows the attribute feature data of sensor node AE Specifically, the path characteristics between the two monitoring sensors include the signal strength, physical distance and transmission delay between the two monitoring sensors.

[0033] Furthermore, after the edge node obtains the data of attribute features and path features, it performs data preprocessing on the obtained data; wherein the data preprocessing includes noise filtering, data format conversion and feature extraction.

[0034] As mentioned above, since the directed graph is composed of node features, edge features and topological structure, by obtaining the attribute features of several monitoring sensors and the path features between the monitoring sensors, they can be used as node features and edge features, and combined with the initial connection graph to generate an initial network topological structure for training the GNN model, and obtaining several attribute features and several path features corresponding to each node, which improves the feature dimension and makes the trained GNN model have better learning ability, thereby improving the optimization reliability of the network topological structure.

[0035] S300, training the GNN model according to the attribute feature vector corresponding to each topology map node in the initial network topology structure and the path feature vector corresponding to each edge, so as to optimize the initial network topology structure to the target network topology structure; it can be understood that: the topology map nodes in the initial network topology structure correspond one-to-one to the sensor nodes in the initial connection map.

[0036] Furthermore, the GNN model is trained through the following steps: S301, initialize the attribute characteristics of the monitoring sensor corresponding to each sensor node and the path characteristics between two monitoring sensors corresponding to each connecting line, and obtain the vector of each attribute feature after the initialization processing and the vector of each path feature after the initialization processing; it can be understood as: the vector of attribute characteristics is represented as the initial state of the corresponding topological graph node, and the vector of path characteristics is represented as the initial state of the corresponding edge.

[0037] The following is an example of the initialization process: As shown in Table 1, for location information, the building location is encoded with floor information such as the 1st floor and the 2nd floor. For example, the southeast corner of the 1st floor is 1, the southwest corner of the 1st floor is 2, the southeast corner of the 2nd floor is 3, the middle of the 2nd floor is 4, and the northwest corner of the 3rd floor is 5. For wireless signal strength, the wireless signal strength is normalized to the interval of 0-1, for example, -50dm is normalized to 0.75. For remaining power, the percentage is used, such as 85%. For task load, it can be expressed as frequency, such as 10 times / second. For sensor type, One-hot encoding can be used, vibration sensor is [1,0,0], stress sensor is [0,1,0], and temperature and humidity sensor is [0,0,1]. Therefore, the initialized attribute feature vector corresponding to node A is [1,0.75,85,10,1,0,0]. Since the signal strength between node A and node B is -50 dBm and -60 dBm, the average value is -55 dBm, which is normalized to 0.70. For the calculation of physical distance to improve location information, the physical distance between the monitoring sensor corresponding to the instance node A and the monitoring sensor corresponding to the instance node B is 15 meters. The transmission delay is calculated according to the physical distance and the communication protocol. The transmission delay calculated in the instance is 0.01 seconds. Therefore, the path feature vectors corresponding to the nodes A and B are [0.70, 15, 0.01].

[0038] S302, when any topology map node receives the current status information sent by its own neighbor node, the attribute feature vector corresponding to the topology map node itself and the path feature vector between the topology map node itself and the neighbor node are updated; it can be understood that: according to the message passing mechanism of the GNN model, the topology map nodes will transmit their own status information to each other.

[0039] S303, based on each update result of any attribute feature vector and path feature vector, adjust the target parameters of the GNN model to complete the training of the GNN model, wherein the target parameters are characterized as node embedding weights and edge weights. In the prior art, the GNN model includes functions such as node embedding, message passing, and message aggregation. Through the self-learning mechanism, the node embedding weights and edge weights can be updated according to the real-time data of the updated node features and edge features to calculate the optimal data transmission path.

[0040] As mentioned above, through the message passing mechanism of the GNN model, the node embedding weight and edge weight will be updated after each round of topology node message transmission, and the optimal data transmission path will be recalculated to optimize the network topology structure. Therefore, when the state of the topology node changes, such as power consumption decreases, task load increases, etc., the optimal data transmission path can be automatically adjusted to avoid nodes with overloaded signals or poor signals from bearing too many data transmission tasks, which greatly improves the flexibility and adaptability of the network, and at the same time improves the data transmission efficiency and reduces communication delays and consumption.

[0041] S400, when any topology map node is monitored to have a fault, a network topology prediction structure is obtained based on the measured parameter information of several topology map nodes in the target network topology structure except the faulty topology map node; it can be understood that: the measured parameter information of the topology map node refers to the data of the attribute characteristics corresponding to the topology map node.

[0042] Specifically, step S400 includes the following steps: S401, when it is detected that any topology map node fails, all topology map nodes to be connected are scanned according to all topology map nodes in the target network topology structure; it can be understood that: the topology map node to be connected refers to any topology map node that has not failed.

[0043] S402, based on the measured parameter information of the topological map nodes to be connected obtained in sequence, the corresponding relationship between the topological map nodes that have obtained the measured parameter information and the topological map nodes in the target network topological structure is identified through the GNN model, and the topological map nodes that have obtained the measured parameter information are connected in sequence according to the connection relationship of the topological map nodes in the target network topological structure to obtain the network topological structure to be repaired; it can be understood as: the edge node is used to receive real-time data from the monitoring sensor, and the corresponding relationship between the topological map nodes and the topological map nodes in the target network topological structure is obtained in sequence according to the sequence of the received topological map node data. In a specific implementation, when the corresponding IDs of two topological map nodes are consistent, the two topological map nodes are in a corresponding relationship.

[0044] S403, based on the topology map nodes newly connected to in the network topology structure to be repaired, the remaining topology map nodes to be connected are monitored to locally iterate the network topology structure to be repaired until no new topology map nodes to be connected can be monitored; it can be understood that: locally iterating the network topology structure to be repaired means that each time the measured parameter information of the topology map node is obtained, the topology map node with the measured parameter information is connected with the adjacent topology map nodes in the current network topology structure to be repaired according to the connection relationship of several topology map nodes in the target network topology structure, so as to update the network topology structure to be repaired.

[0045] Specifically, the remaining topology map nodes to be connected refer to any topology map nodes to be connected among all the topology map nodes to be connected obtained by scanning, except for the topology map nodes in the network topology structure to be repaired.

[0046] S404, according to the measured parameter information of all topology map nodes in the current network topology structure to be repaired, the network topology prediction structure is obtained through the GNN model; it can be understood as: according to the measured parameter information of all current topology map nodes, the topology of all topology map nodes is calculated again through the GNN model to obtain the network topology prediction structure.

[0047] As mentioned above, when a topology map node fails, the target network topology structure is destroyed, so the network needs to be reconstructed. During the reconstruction process, by querying the correspondence between the scanned topology map nodes and the topology map nodes in the target network topology structure, the scanned topology map nodes can be reconnected in turn until the final network topology prediction structure is obtained. In this process, the GNN model can dynamically adjust the connection mode between nodes in real time, so that the edge nodes can quickly update the topology structure of the sensor network, avoid the transmission interruption of data tasks, ensure the continuous operation of the network and the stability of data transmission, thereby improving the accuracy, reliability and sustainability of monitoring, and is suitable for complex and dynamic building and infrastructure monitoring scenarios.

[0048] In a specific embodiment, in step S400, the following steps are performed to detect that a node in the topology map has a fault: S410, for any topology map node, according to the message transmission mechanism of the GNN model, when the message transmission interruption of the topology map node is identified, it is considered that the topology map node is monitored to have a fault. For example, when the topology map node runs out of power, has a hardware failure, or has a signal failure, the communication between the topology map node and the neighboring node will be suddenly interrupted and the signal strength will drop significantly. The neighboring node can determine that the message transmission of the topology map node is interrupted and cannot receive the status update of the topology map node. At this time, the message transmission mechanism of the GNN model will identify the abnormal state of the topology map node.

[0049] S420, if the edge node does not receive the data information sent by the topology map node within the preset time period, or the signal strength between the topology map node and the edge node is lower than the preset signal strength threshold, it is considered that the topology map node is monitored to have a fault; it can be understood that: the edge node can regularly receive the status update information of the monitoring sensors corresponding to each topology map node.

[0050] Furthermore, the method further comprises the following steps: S41, when it is detected that any topology map node fails, the failed topology map node is isolated from the target network topology structure.

[0051] S42, sending the isolated topology map node information to other topology map nodes in the target network topology structure through the message passing mechanism of the GNN model, so that the neighbor nodes of the isolated topology map node stop sending data to the isolated topology map node; wherein, the other topology map nodes in the target network topology structure refer to any topology map node except the isolated topology map node.

[0052] In the above, by isolating the faulty topology map node and notifying other topology map nodes in time, other topology map nodes are prevented from continuing to send data to the faulty topology map node, and then the data transmission tasks are reallocated to ensure the continuous operation of the network topology and the stability of data transmission.

[0053] S500, according to the network topology prediction structure, dynamically adjusts the connection mode between several topology map nodes in the target network topology structure through edge nodes to optimize the target network topology structure; it can be understood as: the edge nodes connect to the corresponding child nodes in turn according to the network topology prediction structure, and the child nodes continue to connect to the corresponding grandchild nodes, and finally obtain the reconstructed new target network topology structure through the connection of layers of sensor nodes.

[0054] The sensor network topology optimization method described in the embodiment of the present invention first constructs an initial connection graph according to the connection relationship between several monitoring sensors and the connection relationship with the edge nodes, and generates an initial network topology structure according to the acquired attribute characteristics and sensor path characteristics of the monitoring sensors to input into the GNN model, and trains the GNN model according to the acquired attribute feature vector corresponding to each sensor node and each path characteristic vector to obtain the target network topology structure. By training the GNN model, the real-time optimal data transmission path can be calculated, which greatly improves the flexibility and adaptability of the network, and at the same time improves the data transmission efficiency and reduces communication delay and consumption. When a topology map node in the target network topology structure is monitored to fail, the network topology prediction structure is obtained according to the measured parameter information of several topology map nodes to optimize the network topology structure. In this process, the GNN model can dynamically adjust the connection mode between nodes in real time, so that the edge node can quickly update the topology structure of the sensor network, avoid the transmission interruption of data tasks, and ensure the continuous operation of the network and the stability of data transmission.

[0055] An embodiment of the present invention also provides a non-transitory computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to implementing a method in a method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement the sensor network topology optimization method provided in the above embodiment.

[0056] An embodiment of the present invention further provides an electronic device, comprising a processor and the aforementioned non-transitory computer-readable storage medium.

[0057] Although some specific embodiments of the present invention have been described in detail by way of example, it will be appreciated by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It will also be appreciated by those skilled in the art that various modifications may be made to the embodiments without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A sensor network topology optimization method, characterized in that: The method comprises the following steps: According to the wireless communication relationship between a plurality of monitoring sensors installed in the target area, and the communication relationship between each monitoring sensor and the edge node, an initial connection graph is constructed; the initial connection graph includes a plurality of sensor nodes and connection lines between the sensor nodes; Based on the edge nodes, the attribute characteristics of the monitoring sensor corresponding to each sensor node and the path characteristics between the two monitoring sensors corresponding to each connection line are obtained, and the initial network topology is generated by combining the initial connection graph and input into the GNN model; According to the attribute feature vector corresponding to each topology graph node and the path feature vector corresponding to each edge in the initial network topology structure, the GNN model is trained to optimize the initial network topology structure to the target network topology structure; When any topology map node is detected to have a fault, a network topology prediction structure is obtained based on measured parameter information of several topology map nodes in the target network topology structure except the topology map node that has a fault; According to the network topology prediction structure, the connection mode between several topology map nodes in the target network topology structure is dynamically adjusted through edge nodes to optimize the target network topology structure.

2. The sensor network topology optimization method according to claim 1, characterized in that: The attribute characteristics of the monitoring sensor include location information, task load, remaining power, sensor type and wireless signal strength between the monitoring sensor and the edge node.

3. The sensor network topology optimization method according to claim 2, characterized in that: Obtain the location information of the monitoring sensor through the following steps: For any monitoring sensor, the position of the monitoring sensor collected in advance during installation and the center point of the position reported by the monitoring sensor are used as the position information of the monitoring sensor; If the location data reported by the monitoring sensor is not received, the distance between the monitoring sensor and the edge node and the distance between the monitoring sensor and the monitoring sensor corresponding to the neighbor node are calculated according to the signal strength between the corresponding sensor node and the edge node and the signal strength between the sensor node and the neighbor node; According to the location information of the edge node, the distance between each monitoring sensor and the edge node, and the distance between each monitoring sensor and the monitoring sensor corresponding to the neighboring node, the location information of each monitoring sensor is obtained after finding the optimal solution.

4. The sensor network topology optimization method according to claim 1, characterized in that: The path characteristics between the two monitoring sensors include the signal strength, physical distance and transmission delay between the two monitoring sensors.

5. The sensor network topology optimization method according to claim 1, characterized in that: The GNN model is trained through the following steps: Initialize the attribute characteristics of the monitoring sensor corresponding to each sensor node and the path characteristics between two monitoring sensors corresponding to each connecting line, and obtain the vector of each attribute characteristic after the initialization processing and the vector of each path characteristic after the initialization processing; When any topology map node receives the current state information sent by its neighboring node, it updates the attribute feature vector corresponding to the topology map node itself and the path feature vector between the topology map node itself and its neighboring node; Based on the update result of any attribute feature vector and path feature vector each time, the target parameters of the GNN model are adjusted to complete the training of the GNN model, where the target parameters are represented by node embedding weights and edge weights.

6. The sensor network topology optimization method according to claim 1, characterized in that: When any topology map node is detected to have a fault, a network topology prediction structure is obtained according to measured parameter information of several topology map nodes in the target network topology structure except the topology map node having a fault, including the following steps: When any topology map node is detected to have a fault, all topology map nodes to be connected are scanned according to all topology map nodes in the target network topology structure; Based on the measured parameter information of the topology map nodes to be connected obtained in sequence, the corresponding relationship between the topology map nodes that have obtained the measured parameter information and the topology map nodes in the target network topology structure is identified through the GNN model, and the topology map nodes that have obtained the measured parameter information are connected in sequence according to the connection relationship of the topology map nodes in the target network topology structure to obtain the network topology structure to be repaired; Based on the topology map nodes newly connected to the topology map nodes in the network topology structure to be repaired, the remaining topology map nodes to be connected are monitored to locally iterate the network topology structure to be repaired until no new topology map nodes to be connected can be monitored; wherein the remaining topology map nodes to be connected refer to any topology map nodes to be connected except the topology map nodes in the network topology structure to be repaired among all the topology map nodes to be connected obtained by scanning; According to the measured parameter information of all topology nodes in the current network topology structure to be repaired, the network topology prediction structure is obtained through the GNN model.

7. The sensor network topology optimization method according to claim 1, characterized in that: The following steps are used to monitor the failure of the topology node: For any topology map node, according to the message transmission mechanism of the GNN model, when the message transmission interruption of the topology map node is identified, it is considered that the topology map node is detected to have a fault; If the edge node does not receive data information sent by the topology map node within a preset time period, or the signal strength between the topology map node and the edge node is lower than a preset signal strength threshold, it is considered that the topology map node is detected to be faulty.

8. The sensor network topology optimization method according to claim 1, characterized in that: The method further comprises the steps of: When any topology map node is detected to have a fault, the faulty topology map node is isolated from the target network topology structure; The isolated topology map node information is sent to other topology map nodes in the target network topology structure through the message passing mechanism of the GNN model, so that the neighbor nodes of the isolated topology map node stop sending data to the isolated topology map node; wherein, the other topology map nodes in the target network topology structure refer to any topology map node except the isolated topology map node.

9. A non-transitory computer-readable storage medium, wherein at least one instruction or at least one program is stored in the storage medium, characterized in that: The at least one instruction or the at least one program is loaded and executed by the processor to implement the sensor network topology optimization method according to any one of claims 1 to 8.

10. An electronic device, characterized in that: The invention comprises a processor and the non-transitory computer-readable storage medium as claimed in claim 9.

Images