Agricultural sensor network node dynamic access control and self-healing system

Abstract

The present application relates to the technical field of agricultural sensor network, and particularly relates to a dynamic access control and self-healing system of an agricultural sensor network node. A state monitoring unit acquires node identity recognition data and state detection signal data, and determines a node connection state distribution; a topology analysis unit constructs a network identity topology model, and determines a node partition parameter set in combination with the node connection state distribution; an access strategy unit calculates node access priority data according to the node partition parameter set, and determines a node access order sequence; a self-healing management unit monitors a node running state according to the node access order sequence, performs fault diagnosis and recovery operation, and generates a network self-healing data set; and a communication interface unit transmits the network self-healing data set to a central management platform through wireless communication technology. The system can realize dynamic access control of the node, enhance the network self-healing capability, and guarantee stable operation of the agricultural sensor network.

Description

Technical Field

[0001] This invention relates to the field of agricultural sensor network technology, and in particular to a dynamic access control and self-healing system for agricultural sensor network nodes. Background Technology

[0002] In modern agricultural production, sensor networks are increasingly widely used. They can collect key agricultural production data in real time, such as soil moisture, ambient temperature, and crop growth status, providing data support for precision planting and intelligent management. As the scale of agricultural production continues to expand, the coverage of sensor networks continues to extend, and the number of nodes in the network also increases significantly. The types of nodes are becoming increasingly diverse, including sensor nodes responsible for collecting environmental parameters and relay nodes that undertake data forwarding tasks. The functional positioning and operational requirements of different nodes in the network vary significantly.

[0003] Agricultural sensor networks often employ a relatively simple static access mechanism for node access control. This mechanism pre-sets the node access order and permissions during the initial network deployment, making it impossible to dynamically adjust based on changes in the actual status of nodes during network operation. For example, when some sensor nodes experience unstable connections due to environmental interference, or when new sensor nodes need to be added to the network to expand the monitoring range, the static access mechanism struggles to respond quickly to these changes. This can easily lead to problems such as node access delays and unreasonable allocation of network resources, thereby affecting the overall data acquisition efficiency and transmission stability of the network.

[0004] Existing agricultural sensor networks suffer from significant deficiencies in self-healing capabilities. Node failures are unavoidable during network operation, such as sensor nodes ceasing operation due to battery depletion or relay nodes failing to forward data due to hardware malfunctions. When these failures occur, existing networks typically require manual intervention for troubleshooting and repair, which is not only time-consuming and labor-intensive but also prone to data interruptions and monitoring blind spots if not handled promptly, severely impacting the real-time acquisition and analysis of critical data in agricultural production. Furthermore, some networks with some self-healing capabilities rely on limited fault diagnosis methods, only able to identify a few common node failure types. They struggle to accurately diagnose complex fault situations, resulting in poor fault recovery and an inability to effectively ensure the network's continuous and stable operation. Summary of the Invention

[0005] The main objective of this invention is to provide a dynamic access control and self-healing system for agricultural sensor network nodes, aiming to solve the technical problems in the prior art.

[0006] This invention proposes a dynamic access control and self-healing system for agricultural sensor network nodes, comprising:

[0007] The status monitoring unit is used to acquire node identification data and node status detection signal data in the agricultural sensor network, and to determine the node connection status distribution based on the node status detection signal data.

[0008] The topology analysis unit constructs a network identity topology model based on the node identity recognition data, and determines the node partition parameter set based on the network identity topology model and the node connection status distribution.

[0009] The access strategy unit calculates node access priority data based on the node partition parameter set and determines the node access order sequence based on the node access priority data.

[0010] The self-healing management unit monitors the node operating status according to the node access sequence, performs fault diagnosis and recovery operations according to the node operating status, and generates a network self-healing dataset.

[0011] The communication interface unit transmits the network self-healing dataset to the central management platform via wireless communication technology.

[0012] Preferably, the status monitoring unit is specifically:

[0013] Based on agricultural sensor equipment, acquire node identification data and node status detection signal data in agricultural sensor networks;

[0014] The node state detection signal data is subjected to noise suppression based on an adaptive filtering algorithm to obtain purified node state detection signal data. A signal mode analysis algorithm is introduced, and the initial values ​​of the parameters of the signal mode analysis algorithm are set.

[0015] The state detection signal data of the purification node is decomposed according to the signal mode analysis algorithm to obtain multiple signal mode components;

[0016] The signal modal components are transformed to construct an analytical signal for each signal modal component. The instantaneous amplitude and instantaneous frequency of each signal modal component are calculated based on the analytical signal. The instantaneous amplitude and instantaneous frequency are then used to construct a time-frequency feature matrix.

[0017] The energy distribution, mean amplitude, total energy, and peak amplitude of each frequency band received by the agricultural sensor device are calculated based on the time-frequency feature matrix to obtain the node state feature spectrum.

[0018] The measurement point location information of the node state feature spectrum is obtained based on the node state detection signal data. The node state feature spectrum of each measurement point is mapped based on the measurement point location information to construct a three-dimensional node state feature spectrum.

[0019] The node activity of each measurement point is determined based on the three-dimensional node state feature spectrum. A heat map of node connection state distribution and a node activity gradient field are constructed based on the node activity of each measurement point. The distribution of node connection state is determined based on the heat map of node connection state distribution and the node activity gradient field.

[0020] Preferably, the topology analysis unit is specifically:

[0021] The network structure of the node identity recognition data is analyzed, an identity feature tensor is constructed by processing node relationships, and node type recognition is performed on the signals of the node identity recognition data based on a classifier algorithm to generate a network identity topology model containing node identity, connection features and type labels.

[0022] The node connection status distribution is mapped to the network identity topology model to construct a three-dimensional spatial distribution model of node resources. The three-dimensional spatial distribution model of node resources is divided according to a preset grid size to construct multiple sub-space regions.

[0023] Based on the node connection status distribution, the node activity information of each subspace region is obtained. Based on the clustering algorithm, the subspace regions with similar node activity in the three-dimensional spatial distribution model of node resources are clustered to obtain the clustering results.

[0024] Based on the clustering results, the distribution probability of nodes in each subspace region is determined, and the corresponding location information of each subspace region in the agricultural sensor network is obtained. Based on the distribution probability and the corresponding location information, the node detection importance of each location in the agricultural sensor network is evaluated, and the detection importance score of each location is obtained.

[0025] The detection requirement information for each location is determined based on the detection importance score of each location. The detection requirement information includes whether to conduct detection and the detection accuracy requirement information.

[0026] Acquire signal acquisition clarity data of agricultural sensor equipment at nodes at different distances; determine the signal acquisition distance information of agricultural sensor equipment at each location based on the signal acquisition clarity data and detection requirement information; and determine the set of signal acquisition points of agricultural sensor equipment based on the signal acquisition distance information.

[0027] The node partition parameter set is determined based on the set of signal acquisition points.

[0028] Preferably, the access strategy unit is specifically:

[0029] Determine the access adjustment distinction value between any two adjacent sub-regions in the node partition parameter set, and construct the access weight graph of the node partition parameter set by using the sub-regions in the node partition parameter set as vertices and the access adjustment distinction value as the edge connecting the vertices.

[0030] Determine the minimum spanning tree of the access weight graph, and based on the minimum spanning tree, determine the access adjustment order of each sub-region in the node partition parameter set;

[0031] The node access operation is performed by sequentially processing each sub-region in the node partition parameter set according to the access adjustment order.

[0032] Preferably, determining the access adjustment distinction value between any two adjacent sub-regions in the node partition parameter set includes:

[0033] For the first and second sub-regions adjacent to each other in the node partition parameter set, the region types of the first and second sub-regions are identified respectively. The region type includes one of high-activity sub-region, medium-activity sub-region, and low-activity sub-region.

[0034] Based on the identified area type, determine the first access adjustment parameter for the first sub-area and the second access adjustment parameter for the second sub-area;

[0035] Calculate the difference between the first access adjustment parameter and the second access adjustment parameter, and determine the access adjustment distinction value between the first sub-region and the second sub-region based on the difference.

[0036] Preferably, determining the access adjustment distinction value between the first sub-region and the second sub-region based on the difference includes:

[0037] Identify the access adjustment parameters of each sub-region in the node partition parameter set, and determine the median of the identified access adjustment parameters;

[0038] Multiple difference intervals are determined based on the median, and a target difference interval is determined in which the calculated difference between the first access adjustment parameter and the second access adjustment parameter lies;

[0039] The preset differentiation value corresponding to the target difference interval is determined as the access adjustment differentiation value between the first sub-region and the second sub-region.

[0040] Preferably, the self-healing management unit is specifically:

[0041] Obtain standard state data of different types of nodes in an agricultural sensor network, and label the standard state data with node types to obtain labeled state data;

[0042] A node state recognition model is constructed based on a neural network, and the labeled state data is imported into the node state recognition model for training.

[0043] The node operation status data of the agricultural sensor network is obtained according to the node access sequence. The node operation status data is then imported into the trained node status recognition model for node type identification. The number of each node type is counted to obtain node status statistics.

[0044] Based on the node status statistics, fault diagnosis is performed, node connection parameters and recovery strategies are dynamically adjusted, and a network self-healing dataset is generated.

[0045] Preferably, the communication interface unit is specifically:

[0046] A hierarchical network communication protocol stack is constructed based on wireless communication technology, and the network self-healing dataset is modulated to construct a network self-healing transmission signal.

[0047] The network self-healing transmission signal is sent to the central management platform according to the layered network communication protocol stack, and the network self-healing transmission signal is decoded to obtain the node self-healing survey data of the agricultural sensor network.

[0048] Preferably, processing each sub-region in the node partition parameter set sequentially according to the access adjustment order includes:

[0049] Identify the control parameters of each control node in the access control sequence, and construct a residual parameter sequence based on the identified control parameters. The order of each parameter in the residual parameter sequence is consistent with the access control sequence.

[0050] The residual parameter sequence is input into the residual network to generate access control outputs for each of the control nodes.

[0051] Node access operations are performed according to the respective sub-regions corresponding to the generated access control output drivers.

[0052] Preferably, the fault diagnosis based on the node status statistics includes:

[0053] Monitor signal strength fluctuation and connection interruption in agricultural sensor networks, identify fluctuation ranges, remove node outliers, analyze the average fluctuation of data, and obtain signal and connection fluctuation data.

[0054] The impact of the signal and connection fluctuation range on the node communication rate is analyzed. Using the known node communication rate, the relationship between the signal and the connection is analyzed, and the communication rate under the different fluctuation range is calculated to obtain the communication rate impact data.

[0055] Based on the communication rate impact data, the node distribution path and signal connection ratio within the target range are dynamically adjusted. Adjustments are made based on the relationship between the communication rate impact data and the signal and connection fluctuation range. Node signal rates and connection control ranges are allocated, and a fault diagnosis parameter set is generated.

[0056] The beneficial effects of this invention are as follows:

[0057] This agricultural sensor network node dynamic access control and self-healing system, through the establishment of a status monitoring unit, can comprehensively acquire the identification data and status detection signal data of nodes in the network, and determine the distribution of node connection status accordingly. This provides comprehensive and accurate basic information for subsequent node access control and network management. Compared to traditional static access mechanisms, this real-time monitoring of node status can promptly grasp the operational dynamics of nodes, avoiding access decision errors caused by incomplete understanding of node status.

[0058] The topology analysis unit constructs a network identity topology model based on node identity recognition data and determines the node partitioning parameter set by combining node connection status distribution. This makes the overall network structure clearer and the classification and management of nodes more targeted. Through reasonable node partitioning, more practical access strategies can be formulated based on the functional characteristics and operational needs of nodes within different partitions. This avoids the disordered allocation of network resources globally, improves the utilization efficiency of network resources, and provides a scientific basis for subsequent calculation of node access priorities, ensuring that access decisions fully consider the overall network structure and the actual needs of nodes.

[0059] The access strategy unit calculates node access priority data based on the node partition parameter set and determines the node access sequence, achieving dynamic control of node access. This dynamic access method can rationally arrange the access order of nodes according to their priority, ensuring that important nodes can obtain network resources first, guaranteeing their stable access and efficient operation. For example, sensor nodes responsible for collecting critical crop growth data have higher access priority and can access the network first, ensuring the timely collection and transmission of critical data; while some non-critical auxiliary nodes access in an orderly manner when resources permit, avoiding access delays or failures for critical nodes due to resource contention, effectively improving the overall operating efficiency of the network.

[0060] The self-healing management unit monitors the operational status of nodes based on their access sequence, performs fault diagnosis and recovery operations, and generates a network self-healing dataset, significantly enhancing the network's self-healing capabilities. During network operation, this unit can track the operational status of nodes in real time, promptly detecting node faults. Whether it's a connection failure in a sensor node or a forwarding failure in a relay node, it can accurately identify the fault through effective fault diagnosis methods. After fault diagnosis, it can quickly initiate corresponding recovery operations without relying on manual intervention, significantly shortening fault handling time and reducing labor costs. Simultaneously, the generated network self-healing dataset records key information such as the time, type, and recovery process of the fault, providing valuable reference data for subsequent network optimization and fault prevention. This reduces network data interruptions and monitoring blind spots caused by node failures, ensuring the network can continuously and stably provide data support for agricultural production.

[0061] The communication interface unit transmits the network self-healing dataset to the central management platform via wireless communication technology, enabling administrators to monitor the network's self-healing status and overall operational status in real time. Administrators can access key information such as the progress of node failure handling and network resource allocation through the central management platform without needing to visit the site, facilitating remote monitoring and management and further enhancing the convenience and efficiency of network management. Overall, this system effectively addresses the problems of traditional agricultural sensor networks in node access control and self-healing capabilities, optimizes network operation, and better meets the needs of modern agricultural production for sensor networks. Attached Figure Description

[0062] Figure 1 This is a timing diagram of the dynamic access control and self-healing system for agricultural sensor network nodes described in this invention.

[0063] Figure 2 This is a schematic diagram of the working principle of the topology analysis unit.

[0064] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0065] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0066] like Figure 1 As shown, this application provides a dynamic access control and self-healing system for agricultural sensor network nodes, including: a status monitoring unit, a topology analysis unit, an access strategy unit, a self-healing management unit, and a communication interface unit.

[0067] The status monitoring unit is responsible for acquiring node identification data and node status detection signal data in the agricultural sensor network, and determining the node connection status distribution based on the node status detection signal data. The topology analysis unit uses the node identification data to construct a network identity topology model, and determines the node partition parameter set based on the node connection status distribution. The access strategy unit calculates node access priority data based on the node partition parameter set, and determines the node access sequence based on the node access priority data. The self-healing management unit monitors the node operating status according to the node access sequence, performs fault diagnosis and recovery operations, and generates a network self-healing dataset. The communication interface unit transmits the network self-healing dataset to the central management platform via wireless communication technology. Through the collaborative work of these units, the system realizes the dynamic node access and self-healing functions of the agricultural sensor network.

[0068] In one embodiment, Example 1: The status monitoring unit acquires node identification data and node status detection signal data in the agricultural sensor network based on agricultural sensor devices. Node identification data includes a unique identifier for each node, such as a MAC address or serial number, and a type classification, such as a temperature sensor or a humidity sensor. The node status detection signal data covers signal strength values, frequency response, and timing waveforms. This data is collected in real time by the wireless receiving module of the sensor device and stored in a local buffer for processing. An adaptive filtering algorithm is applied to the node status detection signal data for noise suppression. This algorithm uses the minimum mean square error principle to dynamically adjust the filter coefficients to minimize the impact of environmental interference such as electromagnetic noise or weather factors, thereby obtaining purified node status detection signal data. Its output is a smoothed signal sequence with reduced random fluctuations and abnormal peaks.

[0069] A signal modal analysis algorithm is incorporated into the processing flow. The initial values ​​of its parameters are preset based on signal characteristics; for example, the number of decomposition layers is set between 5 and 10, and the threshold level is adaptively determined based on the signal amplitude range, ensuring the algorithm can effectively handle diverse sensor signal patterns. The signal modal analysis algorithm decomposes the state detection signal data of the purification node. This algorithm extracts the intrinsic mode functions (EMFs) from the signal through an iterative filtering process. Each EMF represents a specific frequency component of the signal, resulting in multiple signal modal components, which correspond to low-frequency trends and high-frequency details, respectively. Signal transformation is then performed on the signal modal components. The Hilbert transform method is used to construct the analytical signal for each signal modal component. The analytical signal converts the real-number signal into a complex form to facilitate the analysis of phase and frequency characteristics. Then, based on the analytical signal, the instantaneous amplitude and instantaneous frequency of each signal modal component are calculated. The instantaneous amplitude represents the real-time change in signal amplitude, and the instantaneous frequency reflects the instantaneous shift in signal frequency. The calculated instantaneous amplitude and frequency are combined to construct a time-frequency feature matrix. Rows in this matrix correspond to time sampling points, and columns correspond to frequency indices. Each element stores combined amplitude and frequency information, forming a two-dimensional array capturing the dynamic characteristics of the signal overtime and frequency domains. Based on the time-frequency feature matrix, the energy distribution, mean amplitude, total energy, and peak amplitude of each frequency band received by the agricultural sensor device are calculated. The energy distribution is obtained by integrating the power spectral density of a specific frequency band; the mean amplitude is calculated by averaging the amplitudes at all time points; the total energy is obtained by summing the energy values ​​of all frequency bands; and the peak amplitude is determined by the maximum value in the identification matrix. This yields the node state feature spectrum, which summarizes the signal health status of the node in vector form.

[0070] The measurement point location information of the node state feature spectrum is obtained based on the node state detection signal data. This location information includes geographical coordinates such as latitude and longitude, or network topology coordinates such as node ID mapping. This information is extracted from the configuration data of the sensor equipment or the GPS module. Based on the measurement point location information, the node state feature spectrum of each measurement point is mapped to a location. Spatial interpolation techniques, such as inverse distance weighting or kriging, are used to assign feature values ​​to a continuous spatial grid, constructing a three-dimensional node state feature spectrum. The x and y dimensions represent horizontal position, and the z dimension represents feature spectrum values ​​such as energy or amplitude indices. The node activity of each measurement point is determined based on the three-dimensional node state feature spectrum. Node activity is quantified by calculating a weighted average of the feature spectrum values ​​or applying a threshold function; for example, high-energy areas correspond to high activity, and low-amplitude areas correspond to low activity, thus assigning an activity score to each location. Based on the node activity at each measurement point, a heatmap of node connectivity distribution and a gradient field of node activity are constructed. The heatmap uses color mapping to visualize the activity values, with warm colors representing high-activity areas and cool colors representing low-activity areas. The gradient field displays the rate of change by calculating the spatial derivative of the activity, using a vector field to represent direction and magnitude of change. The distribution of node connectivity states is determined based on the heatmap and gradient field. By analyzing the patterns in the heatmap and the flow direction in the gradient field, high-connectivity regions (such as densely populated node areas), low-connectivity regions (such as edge nodes), and potential fault points (such as areas with sudden drops in activity) are identified, thus completing the function of the state monitoring unit.

[0071] In one embodiment, Example 2: See Figure 2 The topology analysis unit performs network structure parsing on the node identification data. This data includes each node's unique identifier, type attributes, and connection relationships with other nodes. This data is typically stored as data packets or database records. The parsing process involves reading this data and extracting network topology features such as adjacency matrices or node degree distributions. An identity feature tensor is constructed through node relationship processing. This processing includes analyzing communication links, data flow paths, and physical location associations between nodes. The identity feature tensor is a multi-dimensional array structure whose dimensions correspond to the number of nodes, feature dimensions, and time series. Each element in the tensor stores a feature vector for a specific node, such as historical signal strength or connection status.

[0072] The system uses a classifier algorithm to identify the node type of the signal from the node identity recognition data. The classifier algorithm employs supervised learning methods such as Support Vector Machines or Random Forests. The training dataset contains signal features of known node types. The identification process involves extracting frequency domain features or statistical features from the signal and inputting them into the classifier. The output is a type label for each node, such as gateway node, relay node, or terminal sensor node. A network identity topology model containing node identity, connection features, and type labels is generated. This model is typically represented as a graph structure where vertices represent nodes and edges represent connections. Vertex attributes include identity information and type labels, while edge attributes include connection strength or latency indicators. The model may exist in the form of a network configuration file or an in-memory data structure. A location mapping operation is performed between the node connection state distribution and the network identity topology model. The node connection state distribution comes from the output of the state monitoring unit and includes the activity data of each node. The location mapping assigns the activity value to the corresponding node in the topology model through coordinate matching or node ID alignment. A three-dimensional spatial distribution model of node resources is constructed, representing node position (x, y) and resource indicators (z), such as activity or energy level, in a three-dimensional coordinate system. The 3D spatial distribution model of node resources is divided according to a preset grid size. The grid size is dynamically adjusted according to the network coverage area and node density. For example, each square meter is divided into a grid unit, generating multiple sub-space regions. Each region contains a set of nodes and their associated resource data.

[0073] The node activity information for each sub-region is obtained based on the node connection status distribution. This activity information is obtained by calculating the average activity or variance of nodes within the region. This data is extracted from the output of the status monitoring unit and associated with the corresponding grid coordinates. Sub-regions with similar node activity in the 3D spatial distribution model of node resources are clustered using clustering algorithms such as K-means or hierarchical clustering. These algorithms group regions based on Euclidean distance or similarity metrics of activity values, resulting in multiple clusters, each containing sub-regions with similar activity characteristics. The distribution probability of nodes in each sub-region is determined based on the clustering results. This probability is calculated by statistically analyzing the ratio of the number of nodes in each cluster to the total number of nodes, while also considering the weighted influence of activity values. The probability value represents the likelihood of a node appearing in that region. The corresponding location information of each sub-region in the agricultural sensor network is obtained. This location information includes geographic coordinates or network topology coordinates, which are directly extracted from the grid partitioning data or obtained through coordinate transformation. The importance of node detection in the agricultural sensor network is assessed based on the distribution probability and corresponding location information. The assessment model adopts a weighted scoring method, in which high distribution probability and strategic locations (such as the network center or boundary) receive higher scores, and the detection importance score of each location is obtained, with the score value ranging from 0 to 1.

[0074] Based on the detection importance score for each location, the detection requirement information for each location is determined. This information includes a Boolean flag indicating whether detection is needed, and an enumerated value representing the detection accuracy level, such as high or low accuracy. Locations with scores above a threshold typically require high-precision detection. Signal sharpness data for agricultural sensor equipment at different distances is acquired. This sharpness data is obtained through experimental measurements and stored as a distance-sharpness comparison table or fitted curve, representing the signal quality attenuation characteristics with distance. Based on the signal sharpness data and detection requirement information, the signal acquisition distance information for each location by the agricultural sensor equipment is determined. This distance information is selected using an optimization algorithm to maximize sharpness while meeting detection accuracy requirements; for example, high-precision detection requires a shorter acquisition distance. Based on the signal acquisition distance information, the set of signal acquisition points for the agricultural sensor equipment is determined. This set of acquisition points is generated using a spatial planning algorithm, listing the optimal coordinates for deploying the sensor equipment to ensure coverage of all important areas. Based on the set of signal acquisition points, a node partitioning parameter set is determined. This set includes sub-region boundary coordinates, a list of acquisition points, priority indicators, and configuration parameters. These parameters are output to subsequent units in the form of configuration files or data structures.

[0075] Taking a large, modern vineyard as an example, an IoT-based agricultural sensor network was deployed to monitor changes in soil moisture, light intensity, and temperature. The network comprises 328 sensor nodes, including 198 soil moisture sensors, 85 light sensors, and 45 temperature sensors. These nodes are distributed across a planting area of ​​approximately 20 hectares, with a spacing of about 15-20 meters between nodes, forming a mesh network topology via the ZigBee protocol. The topology analysis unit first receives node identification data from the status monitoring unit. This data is transmitted in packet form, containing each node's 16-bit MAC address, sensor type code, and installation coordinates. During network structure parsing of the node identification data, the system reads the payload of each packet, extracts the node's unique identifier and a list of neighboring nodes, and constructs the network's adjacency matrix by parsing these connections. This matrix is ​​328×328 in size, where a value of 1 indicates a direct communication link between nodes, and 0 indicates no direct connection.

[0076] The system constructs an identity feature tensor by processing node relationships. Each node's feature vector is defined as containing 12 dimensions of data, including mean signal strength, data upload frequency, and device runtime. All these feature vectors form a 328×12 feature matrix, which is then expanded to a 328×12×24 three-dimensional tensor by adding a time dimension, representing feature changes over 24 hours. A support vector machine (SVM) classifier algorithm is used to identify the node type of the signal data. The classifier uses a radial basis function kernel, and the training data comes from 5000 labeled samples from historical records. It classifies the real-time transmitted signal features, accurately distinguishing between soil moisture sensors (features include periodic data upload patterns), light sensors (features include daylight synchronization response), and temperature sensors (features include diurnal temperature difference response patterns). A network identity topology model containing node identity, connection features, and type labels is generated and stored in a graph database. Each node entity contains identity information (MAC address, coordinates), connection features (neighbor node list, link quality), and a type label (sensor type). Edge entities represent connection relationships and include signal strength and latency attributes. The system performs a location mapping operation between the node connection status distribution and the network identity topology model. The node connection status distribution comes from the activity heatmap output by the status monitoring unit. The system uses GPS coordinate matching to map the activity value of each node to the corresponding node in the topology model, forming three-dimensional point cloud data with resource indicators.

[0077] When constructing the 3D spatial distribution model of node resources, the system divides the entire vineyard area into a grid coordinate system with a precision of 1 meter. Kriging interpolation is used to transform discrete node resource data into a continuous 3D surface, where the X and Y axes represent geographic coordinates and the Z axis represents node activity values, forming a resource distribution model covering a 20-hectare area. This model is divided into approximately 2000 sub-regions using a preset grid size of 10 meters × 10 meters. Each region contains aggregated information of all nodes within its area. Node activity information for each sub-region is obtained based on the node connection status distribution. The system calculates the average activity value of nodes within each 10-meter grid, and records the variance and maximum value of the activity. These data are stored in a 2D array corresponding to the grid coordinates. Based on the DBSCAN clustering algorithm, sub-regions with similar node activity in the 3D spatial distribution model of node resources are clustered. The algorithm sets a neighborhood radius of 15 meters and a minimum sample size of 5 grid points, grouping regions with activity value differences within 10% into the same cluster, ultimately forming 12 main clustering regions.

[0078] Based on the clustering results, the distribution probability of nodes in each subspace region is determined. The system calculates the ratio of the number of nodes within each cluster to the total number of nodes, while also considering node density factors. For example, higher distribution probability values ​​(0.7-0.9) are assigned to the central region of the cluster, and lower probability values ​​(0.3-0.6) are assigned to the peripheral region. The corresponding location information of each subspace region in the agricultural sensor network is obtained directly from the grid coordinate system. Each grid has a unique coordinate identifier, and the elevation and terrain features of the grid are recorded. Based on the distribution probability and corresponding location information, the importance of node detection at each location in the vineyard sensor network is assessed. The system uses a weighted scoring model, where the distribution probability accounts for 60% and the location strategic value accounts for 40% (e.g., irrigation hub areas score higher). Finally, an importance score between 0 and 100 is generated for each grid. Based on the detection importance score of each location, detection requirements are determined. Thresholds are set: scores above 70 require high-precision detection (sampling interval < 1 minute), scores between 50 and 70 require medium-precision detection (sampling interval 1-5 minutes), and scores below 50 require low-precision detection (sampling interval > 5 minutes). The system acquires signal clarity data from agricultural sensor devices at different distances. This data, derived from field test records, shows that signal clarity remains above 90% within 50 meters, 70-90% between 50-100 meters, and below 70% beyond 100 meters. Based on this signal clarity data and detection requirements, the system determines the signal acquisition distance. For high-precision detection areas, the acquisition distance is limited to no more than 50 meters; for medium-precision areas, a distance of 50-100 meters is acceptable; and for low-precision areas, distances exceeding 100 meters are acceptable. The system then determines the set of signal acquisition points for the agricultural sensor devices. A greedy algorithm is used to select the optimal locations, minimizing the number of devices while meeting the detection requirements of all areas. Ultimately, 42 data acquisition nodes are required, with each node's location precisely calculated to ensure coverage of all important areas. Finally, the system determines a node partitioning parameter set based on the signal acquisition point set. This parameter set includes the coordinates of the 42 acquisition points, the coverage radius of each point, priority identifiers, and configuration parameters. These parameters are output in XML format to the system configuration file to guide subsequent node access operations.

[0079] In one embodiment, Example 3: The access policy unit determines the access adjustment discrimination value between any two adjacent sub-regions in the node partition parameter set. The sub-regions are derived from the partition parameter set output by the topology analysis unit, and their adjacency is defined based on spatial proximity or network topology connectivity, typically determined by calculating the Euclidean distance between the region centers or checking the existence of network links. For any two adjacent sub-regions in the node partition parameter set, the access adjustment discrimination value is calculated by comparing their regional characteristic differences. This value quantifies the degree of need for priority adjustment between regions. The calculation process involves obtaining attributes of each sub-region, such as activity level or resource density. Using the sub-regions in the node partition parameter set as vertices and the access adjustment discrimination value as the edge weight connecting the vertices, an access weight graph is constructed. This graph is represented using an undirected graph structure, where the vertex set corresponds to all sub-regions, the edge set represents adjacency relationships, and the weight matrix stores the discrimination value data. The minimum spanning tree of the access weight graph is determined by traversing the weighted graph using Prim's algorithm or Kruskal's algorithm, selecting the edge set connecting all vertices that minimizes the total weight. The spanning tree structure defines the most economical connection path in the network. The access regulation order of each sub-region is determined based on the minimum spanning tree in the node partition parameter set. This order is generated through depth-first or breadth-first traversal of the tree, forming a linear sequence to guide the processing order. Each sub-region in the node partition parameter set is processed sequentially according to the access regulation order to perform node access operations. Processing includes configuring network parameters, allocating bandwidth resources, or activating node communication modules. During processing, the regulation parameters of each regulating node in the access regulation order are identified. Each regulating node corresponds to a control entity in its sub-region. Regulation parameters include timing parameters such as latency, resource parameters such as power allocation, or protocol parameters such as retransmission count. A residual parameter sequence is constructed based on the identified regulation parameters. The residual parameters are calculated by the difference between actual measured values ​​and target reference values. The sequence elements are arranged according to the access regulation order to maintain spatiotemporal consistency. The residual parameter sequence is input into a residual network for processing. The residual network adopts a deep learning architecture with skip connections, and its mathematical expression is:

[0080]

[0081] in: This represents the output feature vector of the l-th layer. Represents the residual mapping function. This represents the set of learnable weight parameters for layer l. The network processes the input sequence through multi-layer nonlinear transformations. Access control outputs for each control node are generated via a residual network. These outputs include adjustment command vectors or parameter correction values, generated based on sequence pattern learning to adapt to dynamic network conditions. The generated access control outputs drive the corresponding sub-regions to perform node access operations. These operations are achieved by issuing control commands to the node devices within the region, enabling parameter adjustments and state switching, ultimately completing an orderly node access process. The entire process ensures efficient allocation of network resources and stability of the access process, and optimizes system performance through mathematical modeling and sequence processing.

[0082] In one embodiment, Example 4: When processing the node partition parameter set, the access strategy unit needs to determine the access regulation distinction value between any two adjacent sub-regions. This process is achieved by analyzing the type characteristics and parameter differences of the sub-regions. Assume an agricultural sensor network contains six sub-regions (A to F), whose region types and access regulation parameters are shown in the table below. These parameters are derived from the output of the topology analysis unit and reflect the activity level and resource allocation requirements of each sub-region, as shown in Table 1.

[0083] Table 1: Sub-region Types and Access Adjustment Parameters

[0084] subregion Region Type Access adjustment parameters A High-activity sub-regions 12 B Medium-activity sub-region 18 C Low-activity sub-region 8 D High-activity sub-regions 15 E Medium-activity sub-region 10 F Low-activity sub-region 6

[0085] For adjacent first and second sub-regions in the node partition parameter set, such as sub-regions A and B being adjacent regions, their region types are identified: sub-region A is a high-activity sub-region, and sub-region B is a medium-activity sub-region. The region type identification is based on the average signal activity threshold of the nodes within the sub-region. The threshold for high-activity sub-regions is typically set to an activity value greater than 20 units, medium-activity sub-regions are between 10 and 20 units, and low-activity sub-regions are below 10 units. This classification is directly derived from the node connection status distribution heatmap data generated by the status monitoring unit. Based on the identified region type, the first access adjustment parameter for the first sub-region and the second access adjustment parameter for the second sub-region are determined. The access adjustment parameter is a comprehensive value, calculated by weighting the sub-region's signal stability, energy level, and historical connection success rate. For example, the parameter for sub-region A is 12, and the parameter for sub-region B is 18. These parameters are stored in the configuration file of the partition parameter set. Calculate the difference between the first access adjustment parameter and the second access adjustment parameter. For example, the difference between sub-regions A and B is |12-18|=6. This difference reflects the degree of difference in access requirements between the two sub-regions. A positive value indicates that the second sub-region needs a higher access priority, while a negative value indicates that the first sub-region has a higher requirement.

[0086] The access control distinction between the first and second sub-regions is determined based on the difference. This process requires first identifying the access control parameters of each sub-region in the node partition parameter set and determining the median. The access control parameters for all sub-regions are extracted from the table: 12, 18, 8, 15, 10, 6, arranged in ascending order as 6, 8, 10, 12, 15, 18. Since the number of parameters is even, the median is the average of the two middle values, which is (10+12) / 2 = 11. Multiple difference intervals are determined based on the median of 11. For example, interval 1 is set as [0,5], interval 2 as (5,10], interval 3 as (10,15], and interval 4 as (15,20]. These interval ranges are defined based on the percentage offset of the median to ensure coverage of possible parameter difference ranges. The target difference interval in which the calculated difference falls is determined. For example, the difference of 6 between sub-regions A and B belongs to interval 2 (5,10]. This interval corresponds to a preset distinction value of 2 (the preset value is usually configured as an integer identifier and stored in the system parameter table). The preset distinction value corresponding to the target difference interval is determined as the access adjustment distinction value between the first and second sub-regions. For example, the distinction value between sub-regions A and B is 2. This value will be used to construct the access weight map to guide the subsequent node access sequence planning.

[0087] For other adjacent sub-region pairs, such as sub-regions B and C (parameters 18 and 8, difference 10, belonging to interval 3, discrimination value 3) and sub-regions C and D (parameters 8 and 15, difference 7, belonging to interval 2, discrimination value 2), the above process is repeated until all adjacent pairs have been processed, ultimately generating a complete set of discrimination values ​​for network optimization. The entire process ensures the objectivity of access adjustment through mathematical comparison and interval mapping, avoiding subjective biases from affecting network performance. Simultaneously, the calculation of discrimination values ​​relies on actual network data, dynamically adapting to the changing needs of different agricultural environments.

[0088] In one embodiment, Example 5: The self-healing management unit acquires standard status data of different types of nodes in the agricultural sensor network. This data comes from technical specifications documents and historical operation logs provided by equipment manufacturers. For example, the standard operating temperature range for temperature sensors is -10℃ to 50℃, the standard accuracy error for humidity sensors is ±3%RH, and the standard response time for light intensity sensors is less than 0.5 seconds. The standard status data is labeled with node types using a rule-based classification system, automatically labeling based on sensor model prefixes: T-series are labeled as temperature sensors, H-series as humidity sensors, and L-series as light intensity sensors, forming a labeled status data table containing device ID, type label, and standard parameter values. A node status recognition model is constructed based on a neural network. This model uses a one-dimensional convolutional neural network architecture, with an input layer of 256 neurons receiving time-series signal data, hidden layers containing 3 convolutional layers and 2 pooling layers, and an output layer using the Softmax activation function to generate a type probability distribution. The model is trained using the Adam optimizer with a learning rate of 0.001. The labeled state data was imported into the node state recognition model for training. The training process lasted for 500 epochs, with a batch size of 32. In each iteration, 1000 records were randomly selected from the labeled dataset as training samples, and the validation set accounted for 20%. After training, the model accuracy stabilized within an acceptable range.

[0089] The node operation status data of the agricultural sensor network is obtained based on the node access sequence, which is derived from the output of the access policy unit and specifies the priority order of data acquisition. The operation status data includes real-time signal strength values, data packet transmission success rate, and device power consumption indicators. This data is uploaded to the gateway device every 5 minutes via the ZigBee protocol. The node operation status data is then imported into a trained node status recognition model for node type identification. The model preprocesses the input data, including normalization and sliding window segmentation. Each window contains data from 60 consecutive time points, and outputs the type identification result and its confidence score for each window. The number of nodes of each type is statistically analyzed. The statistical process is based on the time dimension, calculating the online, offline, and abnormal numbers of each type of node at 1-hour intervals. Abnormal nodes are defined as devices whose output values ​​continuously exceed the standard range for more than 10 minutes. A node status statistics table containing timestamps, type statistics, and abnormal markers is generated.

[0090] Fault diagnosis is performed based on node status statistics. The diagnostic process employs a multi-indicator joint analysis method, simultaneously monitoring signal strength fluctuation and connection interruption rate in the agricultural sensor network. Signal strength fluctuation is obtained by calculating the standard deviation of the Received Signal Strength Indicator (RSSI) value per minute, while connection interruption rate is calculated by counting the number of TCP connection timeouts every 10 minutes. A dynamic threshold method is used to identify fluctuation ranges. Using the data from the most recent 24 hours as a baseline, the moving average and standard deviation of each indicator are calculated, and points whose current value exceeds twice the standard deviation of the mean are marked as abnormal fluctuation points. Outliers are removed using a box plot method. Interquartile ranges are calculated for the data of each sensor group, and data points exceeding 1.5 times the interquartile range are considered outliers and filtered out. The average volatility of the data is analyzed using a weighted average algorithm, with recent data given higher weights and historical data receiving decreasing weights. The final result is a signal and connection fluctuation data matrix containing three dimensions: time, space, and fluctuation indicators.

[0091] This study analyzes the impact of signal and connection fluctuation ranges on node communication rates. Communication rate indicators are derived from actual data transmission rates recorded in the network layer transmission logs. A rate-fluctuation relationship database is established using known node communication rate data, containing typical communication rate values ​​corresponding to different signal strength ranges. Correlation analysis is used to analyze the relationship between signals and connections, calculating the Pearson correlation coefficient matrix. The results show a positive correlation of over 0.7 between signal strength stability and communication rate, and a negative correlation of over 0.6 between connection interruption frequency and communication rate. Piecewise linear interpolation is used to calculate communication rates under differentiated fluctuation ranges, dividing the fluctuation range into 10 levels. Each level corresponds to an estimated communication rate value, generating a communication rate impact data table containing three columns: fluctuation level, expected rate, and confidence interval.

[0092] The node distribution path and signal connection ratio within the target range are dynamically adjusted based on the impact data of communication rate. The target range refers to the abnormal area identified by fault diagnosis. The node distribution path adjustment uses Dijkstra's algorithm to recalculate the optimal data transmission path, and the signal connection ratio adjustment is achieved by modifying the MAC layer protocol parameters, increasing the retransmission threshold and extending the response timeout. Adjustments are made based on the relationship between the impact data of communication rate and the fluctuation range of signal and connection. A multi-dimensional adjustment mapping table is established, quantifying the fluctuation range as an adjustment coefficient. The impact data of communication rate is used as a weighting factor to jointly determine the final parameter adjustment magnitude. A resource allocation algorithm is used to allocate node signal rates and connection control ranges, assigning dedicated time slot resources and frequency band resources to each node. The signal rate is set hierarchically according to node importance, with important monitoring points receiving higher bandwidth allocations. The connection control range is dynamically adjusted according to node density, reducing the connection radius in dense areas to reduce interference. A fault diagnosis parameter set is generated, containing the adjusted network parameter configuration table, resource allocation scheme, and anomaly handling instructions. These parameters are encapsulated in JSON format and transmitted to each network node for update operations, ultimately completing the self-healing management process.

[0093] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A dynamic access control and self-healing system for agricultural sensor network nodes, characterized in that, include: The status monitoring unit is used to acquire node identification data and node status detection signal data in the agricultural sensor network, and to determine the node connection status distribution based on the node status detection signal data. The topology analysis unit constructs a network identity topology model based on the node identity recognition data, and determines the node partition parameter set based on the network identity topology model and the node connection status distribution. The access strategy unit calculates node access priority data based on the node partition parameter set and determines the node access order sequence based on the node access priority data. The self-healing management unit monitors the node operating status according to the node access sequence, performs fault diagnosis and recovery operations according to the node operating status, and generates a network self-healing dataset. The communication interface unit transmits the network self-healing dataset to the central management platform via wireless communication technology. The access strategy unit is specifically: Determine the access adjustment distinction value between any two adjacent sub-regions in the node partition parameter set, and construct the access weight graph of the node partition parameter set by using the sub-regions in the node partition parameter set as vertices and the access adjustment distinction value as the edge connecting the vertices. Determine the minimum spanning tree of the access weight graph, and based on the minimum spanning tree, determine the access adjustment order of each sub-region in the node partition parameter set; The node access operation is performed by sequentially processing each sub-region in the node partition parameter set according to the access adjustment order. The determination of the access adjustment distinction value between any two adjacent sub-regions in the node partition parameter set includes: For the first and second sub-regions adjacent to each other in the node partition parameter set, the region types of the first and second sub-regions are identified respectively. The region type includes one of high-activity sub-region, medium-activity sub-region, and low-activity sub-region. Based on the identified area type, determine the first access adjustment parameter for the first sub-area and the second access adjustment parameter for the second sub-area; Calculate the difference between the first access regulation parameter and the second access regulation parameter, and determine the access regulation distinction value between the first sub-region and the second sub-region based on the difference; Determining the access adjustment distinction value between the first sub-region and the second sub-region based on the difference includes: Identify the access adjustment parameters of each sub-region in the node partition parameter set, and determine the median of the identified access adjustment parameters; Multiple difference intervals are determined based on the median, and a target difference interval is determined in which the calculated difference between the first access adjustment parameter and the second access adjustment parameter lies; The preset differentiation value corresponding to the target difference interval is determined as the access adjustment differentiation value between the first sub-region and the second sub-region.

2. The dynamic access control and self-healing system for agricultural sensor network nodes according to claim 1, characterized in that, The status monitoring unit is specifically: Based on agricultural sensor equipment, acquire node identification data and node status detection signal data in agricultural sensor networks; The node state detection signal data is subjected to noise suppression based on an adaptive filtering algorithm to obtain purified node state detection signal data. A signal mode analysis algorithm is introduced, and the initial values ​​of the parameters of the signal mode analysis algorithm are set. The state detection signal data of the purification node is decomposed according to the signal mode analysis algorithm to obtain multiple signal mode components; The signal modal components are transformed to construct an analytical signal for each signal modal component. The instantaneous amplitude and instantaneous frequency of each signal modal component are calculated based on the analytical signal. The instantaneous amplitude and instantaneous frequency are then used to construct a time-frequency feature matrix. The energy distribution, mean amplitude, total energy, and peak amplitude of each frequency band received by the agricultural sensor device are calculated based on the time-frequency feature matrix to obtain the node state feature spectrum. The measurement point location information of the node state feature spectrum is obtained based on the node state detection signal data. The node state feature spectrum of each measurement point is mapped based on the measurement point location information to construct a three-dimensional node state feature spectrum. The node activity of each measurement point is determined based on the three-dimensional node state feature spectrum. A heat map of node connection state distribution and a node activity gradient field are constructed based on the node activity of each measurement point. The distribution of node connection state is determined based on the heat map of node connection state distribution and the node activity gradient field.

3. The dynamic access control and self-healing system for agricultural sensor network nodes according to claim 2, characterized in that, The topology analysis unit is specifically: The network structure of the node identity recognition data is analyzed, an identity feature tensor is constructed by processing node relationships, and node type recognition is performed on the signals of the node identity recognition data based on a classifier algorithm to generate a network identity topology model containing node identity, connection features and type labels. The node connection status distribution is mapped to the network identity topology model to construct a three-dimensional spatial distribution model of node resources. The three-dimensional spatial distribution model of node resources is divided according to a preset grid size to construct multiple sub-space regions. Based on the node connection status distribution, the node activity information of each subspace region is obtained. Based on the clustering algorithm, the subspace regions with similar node activity in the three-dimensional spatial distribution model of node resources are clustered to obtain the clustering results. Based on the clustering results, the distribution probability of nodes in each subspace region is determined, and the corresponding location information of each subspace region in the agricultural sensor network is obtained. Based on the distribution probability and the corresponding location information, the node detection importance of each location in the agricultural sensor network is evaluated, and the detection importance score of each location is obtained. The detection requirement information for each location is determined based on the detection importance score of each location. The detection requirement information includes whether to conduct detection and the detection accuracy requirement information. Acquire signal acquisition clarity data of agricultural sensor equipment at nodes at different distances; determine the signal acquisition distance information of agricultural sensor equipment at each location based on the signal acquisition clarity data and detection requirement information; and determine the set of signal acquisition points of agricultural sensor equipment based on the signal acquisition distance information. The node partition parameter set is determined based on the set of signal acquisition points.

4. The dynamic access control and self-healing system for agricultural sensor network nodes according to claim 3, characterized in that, The self-healing management unit is specifically: Obtain standard state data of different types of nodes in an agricultural sensor network, and label the standard state data with node types to obtain labeled state data; A node state recognition model is constructed based on a neural network, and the labeled state data is imported into the node state recognition model for training. The node operation status data of the agricultural sensor network is obtained according to the node access sequence. The node operation status data is then imported into the trained node status recognition model for node type identification. The number of each node type is counted to obtain node status statistics. Based on the node status statistics, fault diagnosis is performed, node connection parameters and recovery strategies are dynamically adjusted, and a network self-healing dataset is generated.

5. The dynamic access control and self-healing system for agricultural sensor network nodes according to claim 4, characterized in that, The communication interface unit is specifically: A hierarchical network communication protocol stack is constructed based on wireless communication technology, and the network self-healing dataset is modulated to construct a network self-healing transmission signal. The network self-healing transmission signal is sent to the central management platform according to the layered network communication protocol stack, and the network self-healing transmission signal is decoded to obtain the node self-healing survey data of the agricultural sensor network.

6. The dynamic access control and self-healing system for agricultural sensor network nodes according to claim 5, characterized in that, The step of processing each sub-region in the node partition parameter set sequentially according to the access adjustment order includes: Identify the control parameters of each control node in the access control sequence, and construct a residual parameter sequence based on the identified control parameters. The order of each parameter in the residual parameter sequence is consistent with the access control sequence. The residual parameter sequence is input into the residual network to generate access control outputs for each of the control nodes. Node access operations are performed according to the respective sub-regions corresponding to the generated access control output drivers.

7. The dynamic access control and self-healing system for agricultural sensor network nodes according to claim 6, characterized in that, The fault diagnosis based on the node status statistics includes: Monitor signal strength fluctuation and connection interruption in agricultural sensor networks, identify fluctuation ranges, remove node outliers, analyze the average fluctuation of data, and obtain signal and connection fluctuation data. The impact of the signal and connection fluctuation range on the node communication rate is analyzed. Using the known node communication rate, the relationship between the signal and the connection is analyzed, and the communication rate under the different fluctuation range is calculated to obtain the communication rate impact data. Based on the communication rate impact data, the node distribution path and signal connection ratio within the target range are dynamically adjusted. Adjustments are made based on the relationship between the communication rate impact data and the signal and connection fluctuation range. Node signal rates and connection control ranges are allocated, and a fault diagnosis parameter set is generated.

Images