Intelligent parking lock communication and network fault prediction method and device

Abstract

The invention discloses an intelligent parking lock ad hoc network communication and network fault prediction method and device, and belongs to the technical field of intelligent parking. The method comprises the following steps that: each parking lock node periodically acquires signal intensity of a neighbor node and battery voltage data of the parking lock node and reports the signal intensity and the battery voltage data; the server constructs an adjacent matrix and a degree matrix of the parking lock network according to the reported data, and further calculates a Laplacian matrix; a node capable of being directly connected with a gateway is used as an initial cluster head, and multi-stage cluster-shaped routing is constructed based on electric quantity and signal intensity threshold value iteration; and the server periodically updates the Laplacian matrix and performs eigenvalue decomposition, the overall connectivity of the network is judged according to the second small eigenvalue lambda, a connectivity weak area is positioned according to the variation amplitude of elements in the corresponding eigenvector, and fault prediction is realized. The device comprises a parking lock node, a gateway and a server. According to the invention, long-distance and low-power-consumption wide-coverage communication is realized, network faults can be actively predicted, and the reliability and the operation and maintenance efficiency of the parking system are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent parking technology, and more specifically to a method and device for predicting communication and network faults in intelligent parking locks. Background Technology

[0002] With the development of the smart parking industry, parking locks, as the core terminal equipment for parking space management, directly determine the operational efficiency and user experience of the entire parking system through the stability and reliability of their communication networks.

[0003] Currently, the communication solutions for smart locks on the market mainly have the following limitations and drawbacks: First, in terms of communication distance and coverage, most parking locks use Bluetooth technology for data transmission. However, Bluetooth communication radius is usually short, mostly not exceeding 10 meters, which is difficult to meet the full coverage requirements of large parking lots (such as more than 50 parking spaces). When multiple devices are managed collaboratively, communication gaps are easily caused by distance or obstacles. Secondly, in terms of cost and power consumption, some solutions connect directly to the cellular network by inserting a 4G / 5G IoT card into the ground lock. Although this method solves the coverage problem, each device needs to pay continuous data traffic fees every month, which is costly in large-scale deployment scenarios. At the same time, the cellular communication module has high power consumption, with a standby current of around 10mA. Ground locks mostly rely on lithium batteries for power, and high power consumption leads to frequent charging or battery replacement, increasing maintenance costs. Furthermore, in terms of network management and operation and maintenance, existing solutions generally lack proactive prediction capabilities. The system only issues a passive alarm after a communication interruption occurs, and cannot identify in advance areas with weak network connectivity caused by factors such as signal obstruction and insufficient node power. This mode results in delayed fault handling, which seriously affects the user parking experience and system operation efficiency. Based on this, the present invention proposes a method and device for predicting communication and network faults in smart locks to solve the above problems. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and apparatus for predicting communication and network faults of smart locks, so as to solve the problems existing in the background art.

[0005] This invention provides the following technical solution: a method for predicting communication and network faults in intelligent parking locks, comprising the following steps: S1. Node Data Acquisition and Reporting: Each smart lock node in the parking lot periodically scans surrounding nodes to obtain the signal strength values ​​of the surrounding nodes and its own battery voltage data, and encapsulates the signal strength values ​​and battery voltage data into communication frames, which are then sent via LoRa wireless communication. S2. Network Matrix Construction: The server receives the data reported by all ground lock nodes, constructs the adjacency matrix A and degree matrix D of the ground lock network, calculates the Laplace matrix L=DA, and constructs the power matrix E that records the real-time battery voltage of each node. S3. Multi-hop clustered route construction: Using the ground lock nodes that can communicate directly with the gateway as the initial cluster head set, based on the preset battery voltage threshold and signal strength threshold, a multi-level clustered network structure is iteratively constructed to establish multi-hop routing paths to the gateway for each ground lock node; S4. Network Fault Prediction: The server periodically updates the Laplacian matrix L and performs eigenvalue decomposition on L to extract the second smallest eigenvalue λ2 and its corresponding eigenvector V; the overall connectivity of the network is judged based on the value of λ2, and weak connectivity areas in the network are located based on the change range of the element values ​​in the eigenvector V.

[0006] This solution achieves wide coverage and low power consumption operation of the ground lock communication network. Through mathematical analysis of the Laplace matrix, the network status is quantitatively evaluated, realizing the transformation from passive alarm to active prediction and significantly improving system reliability.

[0007] As a further aspect of the present invention: step S3, multi-hop clustered route construction, specifically includes: S3.1 Mark the ground lock nodes that can communicate directly with the gateway as the initial set A; S3.2 Select nodes from set A whose battery voltage is higher than the first voltage threshold as first-level cluster heads; S3.3 For a groundlock node not included in set A, if its signal strength value with any of the first-level cluster heads is higher than the first signal threshold, then it is added to the cluster where the first-level cluster head is located; S3.4 Iteratively execute the following process: Within the constructed K-level cluster, select nodes with battery voltages higher than the first voltage threshold as K+1-level cluster heads; for nodes that have not yet joined any cluster, if their signal strength value with any K+1-level cluster head is higher than the first signal threshold, then add them to the cluster where the K+1-level cluster head is located, until most nodes are incorporated into the clustered network structure.

[0008] This scheme constructs a stable and energy-balanced multi-hop network topology through iterative clustering based on power and signal strength. This avoids some nodes from running out of power prematurely due to excessive forwarding, extends the overall network lifespan, and ensures the reliability of data transmission paths.

[0009] As a further aspect of the present invention: the multi-hop cluster routing construction step S3 further includes: S3.5 Unassigned node processing: for isolated nodes that have not been added to any cluster after iterative construction, they are instructed to scan the surrounding network-connected nodes through LoRa, select the node with the largest signal strength value as a relay node, and transmit the data to the gateway through multi-hop forwarding.

[0010] This solution ensures high network coverage and reduces communication silos.

[0011] As a further aspect of the present invention: in step S4, network fault prediction, the step of judging the overall network connectivity strength based on the value of λ2 includes: If λ2 > 0, then the network is determined to be globally connected; A connectivity health threshold is preset. When the value of λ2 is lower than this health threshold, a network connectivity weakness warning is generated.

[0012] This solution provides clear and quantifiable metrics (λ2) to assess the health of the entire ground lock network, enabling maintenance personnel to quickly grasp the overall connectivity level and determine whether immediate intervention is needed.

[0013] As a further aspect of the present invention: the step of locating weak connectivity regions based on feature vector V includes: Calculate the difference between the corresponding element values ​​of adjacent indices in the feature vector V, and / or calculate the deviation of each element value from the average value of the vector; The ground lock nodes corresponding to elements whose difference or deviation exceeds a preset amplitude threshold are marked as nodes with weak connectivity. The physical regions where these weak nodes are located are the weak connectivity regions.

[0014] This solution can pinpoint the specific bottleneck node or micro-area causing the connectivity degradation from the global network, making maintenance actions (such as adding repeaters or replacing batteries) highly targeted, greatly improving maintenance efficiency and reducing the cost of blind troubleshooting.

[0015] As a further aspect of the present invention, it also includes an S5 network optimization step: Once a weak connectivity area is located, the server sends an early warning message to the maintenance terminal. The early warning message includes suggestions to add LoRa relay devices in the weak area or replace the batteries of low-power ground lock nodes.

[0016] By directly transforming the analysis results into actionable operation and maintenance instructions, the system is driven to achieve proactive and intelligent operation and maintenance, further reducing the complexity of human intervention.

[0017] A smart lock communication and network fault prediction device for implementing any of the methods described above, comprising: Multiple smart lock nodes, each node includes: The Lora communication module is used to enable wireless communication between nodes and between nodes and gateways. Microcontrollers are used to control data acquisition, communication frame encapsulation, and routing logic processing. The power detection module is used to detect the voltage of the lock's battery in real time. At least one gateway device, deployed in the parking lot, is used to aggregate data from lock nodes in various locations via LoRa and forward the data to the server via a wired or wireless network; A server is communicatively connected to the gateway and configured to perform the network matrix construction step, the multi-hop cluster route construction step, and the network fault prediction step.

[0018] As a further aspect of the present invention: the Lora communication module of the smart lock node uses a chip with a working frequency band of 470MHz-510MHz.

[0019] The use of the 470MHz-510MHz frequency band ensures the effectiveness and compliance of the communication solution.

[0020] As a further aspect of the present invention, the microcontroller is a low-power microcontroller.

[0021] This solution, combined with the low-power characteristics of the Lora module, further reduces the overall power consumption of the ground lock node.

[0022] As a further aspect of the present invention: the gateway device supports dual-mode communication between LoRa and Ethernet or 4G networks.

[0023] This solution improves the reliability of the uplink between the gateway and the server. Ethernet provides a stable and high-speed connection, while the 4G network serves as a backup to ensure the system's continuous online capability in the event of a backbone network failure, thereby enhancing the availability of the entire system.

[0024] Beneficial effects Wide coverage and low cost: Utilizing LoRa technology to achieve ultra-long-distance communication, combined with multi-hop routing, a single gateway can cover large parking lots, completely eliminating the monthly fee for IoT cards for each parking lock and greatly reducing the long-term communication costs of large-scale deployment.

[0025] Ultra-low power consumption and long battery life: The combination of LoRa modules and low-power MCUs allows the ground lock nodes to be in a very low-power sleep state most of the time, significantly extending battery life to more than a year and reducing the maintenance work of frequent battery replacements.

[0026] Proactive and intelligent operation and maintenance: This system has transformed from passive alarms to proactive prediction. It can identify potential fault points several days in advance, allowing maintenance personnel to intervene before users perceive problems, reducing communication failure rates by more than 80%, and greatly improving the service quality and reliability of the parking system.

[0027] Network dynamic adaptation: Both route construction and fault prediction models are based on periodically updated real-time data, enabling the network to automatically adapt to environmental changes such as node power decay and vehicle obstruction, possessing strong self-healing and optimization capabilities, and ensuring long-term stable operation. Attached Figure Description

[0028] The invention will now be further described with reference to the accompanying drawings.

[0029] Figure 1 This is a schematic diagram of the network topology based on multi-hop cluster routing in this invention.

[0030] Figure 2 This is a schematic diagram of the Laplace matrix eigenvalue analysis performed on the server side of this invention.

[0031] Figure 3 This is a flowchart of a method for predicting communication and network faults in smart locks based on LoRa according to the present invention. Detailed Implementation

[0032] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. These embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0033] This embodiment uses a large open-air parking lot as an example. The parking lot is rectangular, approximately 500 meters long and 300 meters wide, and plans to deploy 300 smart parking locks. A main gateway is deployed on a light pole in the central area of ​​the parking lot.

[0034] A method for predicting communication and network faults in smart locks is implemented through the following steps: S1. Node Data Acquisition and Reporting: Periodic setting: Each microcontroller (MCU) in the ground lock node is configured with a hardware timer that generates an interrupt every 30 seconds to wake up the Lora module and MCU in sleep mode.

[0035] Scanning and Acquisition: Upon wake-up, the MCU controls the LoRa module to enter receive mode and continuously scans for 1000 milliseconds. During this period, it records all parseable LoRa data frames from surrounding ground lock nodes and extracts the sending node ID and the Received Signal Strength Indication (RSSI) value of the frame. Simultaneously, the MCU activates its internal analog-to-digital converter (ADC) to acquire the voltage value of the voltage divider circuit connected to the positive terminal of the battery. Using a preset conversion formula (e.g., actual voltage = (ADC sample value / 4095) * 3.3 * voltage divider coefficient), it calculates the precise battery voltage with an accuracy controlled within ±0.01V.

[0036] Data Encapsulation and Transmission: The MCU encapsulates the local node ID, the calculated battery voltage (e.g., 3.82V), and the scanned neighbor node IDs and their corresponding RSSI values ​​(e.g., neighbor node ID: 123, RSSI: -95dBm) into a data frame according to a predefined frame format (format: frame header 0xAA0xBB + local node ID (2 bytes) + battery voltage (2 bytes) + number of neighbors (1 byte) + [neighbor 1 ID (2 bytes) + neighbor 1 RSSI (1 byte)] + ... + CRC16 checksum (2 bytes)). Subsequently, the MCU controls the LoRa module to transmit this data frame with a specific spreading factor, channel spacing, bandwidth, and frequency (e.g., SF=10, BW=125kHz, Freq=470MHz-510MHz).

[0037] Network matrix construction: Data reception: The cloud server continuously receives data frames forwarded from the gateway and originating from lock nodes in various locations via a long-lived Socket connection. The server-side data parsing service performs CRC checks on each data frame. After a successful check, the node ID, battery voltage, neighbor list, and RSSI information are parsed out.

[0038] Matrix Construction and Update: The server maintains an adjacency matrix A in memory, where the row and column indices correspond to the IDs of the LoRa modules. When processing the RSSI value reported by node i regarding neighbor j, the following assignment is performed: if RSSI ≥ -145dBm (the lower limit of the receiving sensitivity of the LoRa module used in this embodiment), the link is considered valid, and the matrix elements are updated accordingly. The value is set to the absolute value of the RSSI (e.g., 95); if RSSI < -145 dBm or not reported, then =0. Simultaneously, the server calculates the degree matrix D, which is a diagonal matrix whose diagonal elements... The value of is equal to the number of all non-zero elements in the i-th row of the adjacency matrix A. Subsequently, the server performs matrix subtraction to generate the Laplace matrix L=DA. Furthermore, the server maintains a one-dimensional array as the energy matrix E, where... The latest reported voltage of storage node i.

[0039] Multi-hop clustered route construction: The server runs a routing management algorithm. First, nodes that can successfully communicate directly with the gateway within the last 5 minutes are marked as the initial cluster head set.

[0040] The algorithm then iterates using this set as the root. For each node in the initial set, it checks whether the value in its electrical matrix E is higher than a preset first voltage threshold (6.5V). Nodes that meet the condition are appointed as first-level cluster heads. Subsequently, the server traverses other nodes in the network. If the RSSI value (RSSI) of a node corresponding to a first-level cluster head in the adjacency matrix A is stronger than a first signal threshold (-100dBm), then the node is assigned to the cluster of that cluster head. This process is repeated iteratively to construct a multi-level cluster structure.

[0041] Network Fault Prediction: The server triggers a fault prediction task every 5 minutes. This task calls a numerical computing library (such as the numpy.linalg.eig function in Python's NumPy library) to perform eigenvalue decomposition on the Laplacian matrix L at the current time step.

[0042] After the calculation is complete, all eigenvalues ​​are sorted, and the second smallest eigenvalue λ2 (i.e., algebraic connectivity) is extracted. At the same time, the eigenvector V corresponding to λ2 is obtained.

[0043] Determine overall connectivity: If λ2 > 0.5 (the preset health threshold), the network connectivity is considered good; if 0 < λ2 < 0.5, the network connectivity is considered weak, triggering an early warning.

[0044] Locating weak regions: Calculate the absolute value of the difference between each element in the feature vector V and the previous element, and find the top 3 index positions with the largest changes. The node IDs corresponding to these indices are the connectivity bottlenecks in the network.

[0045] Based on the above S3 step, the iterative construction is as follows: The initial cluster head set A is determined as follows: The gateway records the IDs of all groundlock nodes that directly establish a connection with it during the initial communication handshake phase, forming a set A={a1,a2,...,a15}.

[0046] First-level cluster construction: The server checks the battery level of each node ai in set A. .like If the voltage is greater than 6.5V, then ai is marked as a first-level cluster head Bi. Subsequently, the server scans the set of nodes X that do not belong to set A. For each node xj in X, it searches for its signal strength value A[xj][Bi] in the adjacency matrix A with all first-level cluster heads Bi. If there exists a Bi such that A[xj][Bi] > 100 (i.e., RSSI > -100dBm), then node xj is assigned to the cluster containing the Bi that maximizes its signal strength. After this process, multiple first-level clusters are formed, each centered around a first-level cluster head.

[0047] Secondary cluster construction: Within each established primary cluster, the server further filters nodes with a power level E > 6.5V and appoints them as secondary cluster heads Ck. Then, it searches for the remaining nodes that have not yet joined any clusters and assigns them to the secondary cluster head Ck with the strongest signal.

[0048] Iteration Termination: Repeat the above process of selecting the head within the cluster and merging the nodes outside the cluster until more than 95% of the nodes are included in the cluster structure, or the iteration level reaches the preset upper limit (such as level 5).

[0049] After the above iterative construction, a small number of nodes (e.g., 5) still failed to join any cluster due to extreme locations or severe signal obstruction. For these unassigned nodes, the server sends a command to put them into free-roaming mode. In this mode, the node increases its scanning frequency (e.g., once every 10 seconds), and instead of targeting the cluster head, continuously measures the RSSI of all audible already-joined nodes. The node selects the node with the strongest RSSI as its default next-hop relay. Data frames are sent to this relay node, which then forwards them according to its established routing path, eventually reaching the gateway.

[0050] In step S4 above, the specific implementation of the judgment based on the value of λ2 is as follows: The server has a pre-defined configurable threshold table: {Healthy: 0.5, Fair: 0.2, Vulnerable: 0.0}.

[0051] Each time λ2 is calculated, it is compared with the threshold table: if λ2>=0.5, the system status light is green: healthy; if 0.2<=λ2<0.5, it is yellow: normal, and a log is recorded; if 0<λ2<0.2, it is orange: weak, triggering a low-level warning and notifying maintenance personnel to pay attention; if λ2=0, it is red: disconnected, triggering a high-level alarm, indicating that the network has split.

[0052] The specific operation for locating areas with weak connectivity is as follows: Let the eigenvector be V = [v1, v2, ..., vn].

[0053] Calculate the magnitude of change: For each node index i (from 2 to n), calculate its magnitude of change Δi = vi − vi−1 At the same time, calculate the average value of the entire vector. The formula is: ; And calculate the absolute deviation of each element from the mean. .

[0054] Comprehensive sorting and filtering: sorting each node's... and A weighted sum (e.g., each weight is 0.5) yields a comprehensive volatility score. The formula is:

[0055] Sort all nodes in descending order of this score.

[0056] Identify weak nodes: Select the top 5% of nodes (or nodes with scores exceeding a threshold of 0.5). These node IDs are the weak connectivity nodes. The server highlights the locations of these nodes in the geographic information system, and the area formed by these locations is the weak area that needs optimization.

[0057] Once the vulnerable area is located, the server automatically performs the following network optimization recommendations: Generate an early warning message in the following format: [Network Optimization Suggestion] Time: <timestamp>; Vulnerable Area: Node <id1> , <id2>...;Main risks: <Low battery / signal obstruction>;Recommended measures: <Replace node> <idx>Battery > < at node <idy>and <idz>Add repeaters between them.

[0058] This structured message is pushed to the mobile app of the operations and maintenance personnel and the alarm list of the back-end management system through the preset HTTP API interface.

[0059] After performing the recommended actions, maintenance personnel can report the results on the app. The system then monitors whether the λ2 value in that area recovers, forming a closed loop.

[0060] A smart lock communication and network fault prediction device for implementing any of the above methods, comprising: Multiple smart lock nodes, each node includes: The Lora communication module is used to enable wireless communication between nodes and between nodes and gateways. Microcontrollers are used to control data acquisition, communication frame encapsulation, and routing logic processing. The power detection module is used to detect the voltage of the lock's battery in real time. At least one gateway device, deployed in the parking lot, is used to aggregate data from lock nodes in various locations via LoRa and forward the data to the server via a wired or wireless network; A server, which communicates with the gateway, is configured to perform network matrix construction steps, multi-hop cluster route construction steps, and network fault prediction steps.

[0061] The above-mentioned device is implemented through the following physical components: Smart lock nodes (300 units): Each lock housing houses a custom PCB board. The core of the board is an STM32L431RCT6 microcontroller (MCU), which connects to the SX1278IMLTRT Lora communication module via an SPI interface. The power detection circuit consists of a high-precision resistor divider network (such as 0805 surface-mount resistors with 1% accuracy) and the MCU's internal 12-bit ADC. The entire node is powered by a single 7.4V, 10000mAh lithium battery.

[0062] Gateway device (1 unit): A Hongdian H8921 industrial-grade LoRa gateway is used, deployed on a central light pole in the parking lot, and connected to the parking lot's fiber optic network via a network cable. The gateway is configured in transparent transmission mode, forwarding received LoRa data packets to the server IP address via the UDP protocol.

[0063] Server (1 set): A cloud server (such as Alibaba Cloud ECS, configured with 2 cores and 4GB) running the CentOS operating system. A background service program written in Python is deployed on it, responsible for implementing all the algorithm logic of claims 1 to 6. The server receives gateway data via Socket, persistently stores historical states using a MySQL database, and provides a RESTful API to the front-end management interface through the Spring Boot framework.

[0064] The hardware circuit design of the aforementioned SX1278IMLTRT Lora communication module strictly adheres to its datasheet's application guidelines for the 470MHz-510MHz frequency band. The RF section uses a 470MHz-510MHz bandpass filter (BPF) matched with a 470MHz-510MHz spring antenna. In terms of software configuration, when initializing the Lora module, the MCU sets its initial operating frequency to 470MHz, divides the channels, and performs frequency-hopping communication within the 470MHz-510MHz band.

[0065] The selected STM32L431RCT6 microcontroller is key to achieving low power consumption. At the software level, the program fully utilizes its low-power characteristics: during idle periods after data transmission, the MCU immediately switches from run mode (approximately 3mA) to stop mode, where its typical power consumption is only 1.7μA. Periodic switching between active and stop modes via timer interrupts reduces the average operating current of the node to the 10μA level, thus achieving ultra-long battery life.

[0066] The Hongdian H8921 gateway connects to two uplinks simultaneously: Main link (Ethernet): A stable IP address can be obtained by directly plugging the standard RJ45 network cable into the switch in the parking lot.

[0067] Backup link (4G): Insert a China Mobile 4G IoT card into the SIM card slot of the gateway.

[0068] The gateway device is configured with a link failure switching strategy: it prioritizes Ethernet uplink by default. The system checks the Ethernet connection status every minute, and if an Ethernet disconnection is detected, it automatically switches to the 4G network for data transmission within 30 seconds to ensure no data loss. When Ethernet is restored, it automatically switches back to the 4G network to save bandwidth costs.

[0069] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.< / idz> < / idy> < / idx> < / id1>

Claims

1. A method for intelligent parking lock self-organizing network communication and network fault prediction, characterized in that, Includes the following steps: S1. Node Data Acquisition and Reporting: Each smart lock node in the parking lot periodically scans surrounding nodes to obtain the signal strength values ​​of the surrounding nodes and its own battery voltage data, and encapsulates the signal strength values ​​and battery voltage data into communication frames, which are then sent via LoRa wireless communication. S2. Network Matrix Construction: The server receives the data reported by all ground lock nodes, constructs the adjacency matrix A and degree matrix D of the ground lock network, calculates the Laplace matrix L=DA, and constructs the power matrix E that records the real-time battery voltage of each node. S3. Multi-hop clustered route construction: Using the ground lock nodes that can communicate directly with the gateway as the initial cluster head set, based on the preset battery voltage threshold and signal strength threshold, a multi-level clustered network structure is iteratively constructed to establish multi-hop routing paths to the gateway for each ground lock node; S4. Network Fault Prediction: The server periodically updates the Laplacian matrix L and performs eigenvalue decomposition on L to extract the second smallest eigenvalue λ2 and its corresponding eigenvector V; the overall connectivity of the network is judged based on the value of λ2, and weak connectivity areas in the network are located based on the change range of the element values ​​in the eigenvector V.

2. The intelligent lock communication and network fault prediction method according to claim 1, characterized in that: The step S3, multi-hop cluster route construction, specifically includes: S3.1 Mark the ground lock nodes that can communicate directly with the gateway as the initial set A; S3.2 Select nodes from set A whose battery voltage is higher than the first voltage threshold as first-level cluster heads; S3.3 For a groundlock node not included in set A, if its signal strength value with any of the first-level cluster heads is higher than the first signal threshold, then it is added to the cluster where the first-level cluster head is located; S3.4 Iteratively execute the following process: Within the constructed K-level cluster, select nodes with battery voltages higher than the first voltage threshold as K+1-level cluster heads; for nodes that have not yet joined any cluster, if their signal strength value with any K+1-level cluster head is higher than the first signal threshold, then add them to the cluster where the K+1-level cluster head is located, until most nodes are incorporated into the clustered network structure.

3. The intelligent lock communication and network fault prediction method according to claim 2, characterized in that: The step S3 multi-hop cluster routing construction also includes: S3.5 unassigned node processing: for isolated nodes that have not been added to any cluster after iterative construction, they are instructed to scan the surrounding network-connected nodes through LoRa, select the node with the largest signal strength value as the relay node, and transmit the data to the gateway through multi-hop forwarding.

4. The intelligent lock communication and network fault prediction method according to claim 1, characterized in that: In step S4, network fault prediction, the step of judging the overall network connectivity strength based on the value of λ2 includes: If λ2 > 0, then the network is determined to be globally connected; A connectivity health threshold is preset. When the value of λ2 is lower than this health threshold, a network connectivity weakness warning is generated.

5. The intelligent lock communication and network fault prediction method according to claim 4, characterized in that: The process of locating weak connectivity regions based on feature vector V includes: Calculate the difference between the values ​​of adjacent indices in the feature vector V, and / or calculate the deviation of each element value from the average value of the vector; The ground lock nodes corresponding to elements whose difference or deviation exceeds a preset amplitude threshold are marked as nodes with weak connectivity. The physical regions where these weak nodes are located are the weak connectivity regions.

6. The intelligent lock communication and network fault prediction method according to claim 5, characterized in that: It also includes S5 network optimization steps: Once a weak connectivity area is located, the server sends an early warning message to the maintenance terminal. The early warning message includes suggestions to add LoRa relay devices in the weak area or replace the batteries of low-power ground lock nodes.

7. A smart lock communication and network fault prediction device for implementing the method of any one of claims 1 to 6, characterized in that, include: Multiple smart lock nodes, each node includes: The Lora communication module is used to enable wireless communication between nodes and between nodes and the gateway. Microcontrollers are used to control data acquisition, communication frame encapsulation, and routing logic processing. The power detection module is used to detect the voltage of the lock's battery in real time. At least one gateway device, deployed in the parking lot, is used to aggregate data from lock nodes in various locations via LoRa and forward the data to the server via a wired or wireless network; A server is communicatively connected to the gateway and configured to perform the network matrix construction step, the multi-hop cluster route construction step, and the network fault prediction step.

8. The intelligent ground lock communication and network fault prediction device according to claim 7, characterized in that: The Lora communication module of the smart lock node uses a chip with a working frequency band of 470MHz-510MHz.

9. The intelligent ground lock communication and network fault prediction device according to claim 7, characterized in that: The microcontroller is a low-power microcontroller.

10. The intelligent ground lock communication and network fault prediction device according to claim 7, characterized in that: The gateway device supports dual-mode communication between LoRa and Ethernet or 4G networks.

Images