Wireless network time synchronization method and system, and storage medium

Abstract

The invention discloses a wireless network time synchronization method and system, and a storage medium, and the method comprises the steps: collecting network node time data, forming an original sampling sequence, and generating a whole network clock observation matrix; based on the connection strength between each pair of network nodes, constructing a Laplacian matrix; on the basis of an objective function of a graph regularization robust low-rank decomposition model, carrying out joint optimization solution on the whole-network clock observation matrix, and alternately updating a low-rank matrix and a sparse matrix; after convergence conditions are met, drift and offset correction values are calculated, and correction parameters are generated; and performing wireless network time synchronization according to the correction parameter. The problem that an existing synchronization algorithm is unstable in precision under the conditions of strong interference, packet loss and dynamic topology is solved, and unification of high precision and high robustness of clocks of the whole network is achieved.

Description

Technical Field

[0001] This invention relates to the field of communication network technology, and in particular to a wireless network time synchronization method, system, and storage medium. Background Technology

[0002] Time synchronization is a core foundation of wireless communication systems. Whether in distributed cooperative communication, precise ranging and positioning, power control, multi-antenna coordination, or time-triggered testing and control systems, accurate time alignment is a prerequisite for system performance optimization and correct protocol execution.

[0003] Commonly used synchronization schemes mainly include bidirectional message synchronization mechanisms based on the IEEE 1588 Precise Time Protocol (PTP) and Network Time Protocol (NTP), distributed consensus synchronization methods based on node adjacency, and clock recovery algorithms based on matrix completion or low-rank approximation. Among them, PTP and NTP methods rely on the assumption of symmetric link delay, and the synchronization accuracy decreases when there is fading, random delay fluctuations, and packet loss in the wireless channel. Although consensus synchronization methods can achieve iterative alignment under distributed conditions, they lack robustness in the presence of non-Gaussian interference, abnormal nodes, or link failures. Although synchronization algorithms based on low-rank modeling can recover the global clock state when some timestamps are missing, they do not take into account the smoothing constraints of network topology and time series, and the synchronization error is easily affected by noise, resulting in unstable fluctuations.

[0004] As wireless networks continue to expand in scale and application scenarios become increasingly complex, such as in low-altitude communication, industrial field networking, distributed telemetry and control systems, and satellite networking, the time data between nodes typically exhibits a mixed structural characteristic of low-rank properties, sparse and anomalous interference, and random noise. Existing synchronization algorithms struggle to uniformly model these multiple structural characteristics and fail to effectively integrate network space topology information with the smoothness constraints of the time dimension, resulting in significant deficiencies in synchronization accuracy and system robustness under complex environments. Summary of the Invention

[0005] Based on the shortcomings of the existing technology, the present invention provides a wireless network time synchronization method, system, and storage medium, which solves the problem of unstable accuracy of existing synchronization algorithms under strong interference, packet loss, and dynamic topology conditions, and achieves a balance between high accuracy and high robustness of the entire network clock.

[0006] To address the aforementioned technical problems, the first aspect of this invention discloses a wireless network time synchronization method, the method comprising: Collect time data from network nodes to form raw sampling sequences and generate a network-wide clock observation matrix; Construct a graph Laplacian matrix based on the connection strength between each pair of network nodes; Based on the objective function of the graph regularized robust low-rank decomposition model, the entire network clock observation matrix is ​​jointly optimized and solved, and the low-rank matrix and sparse matrix are alternately updated. After the convergence condition is met, the drift and offset correction amounts are calculated, and the correction parameters are generated. Wireless network time synchronization is performed based on calibration parameters.

[0007] In some implementations, a graph Laplacian matrix is ​​constructed based on the connection strength between each pair of network nodes, including: Assign a position to each network node in a two-dimensional space, and calculate the pairwise distances between all network nodes; For each network node, identify a preset number of nearest neighbors, establish connections between the network node and the nearest neighbors, and determine the connection weights; Construct a weight matrix based on connection weights; Accumulate the weights connected to each network node to construct a degree matrix; Based on the degree matrix and weight matrix, construct the graph Laplacian matrix.

[0008] 3. The wireless network time synchronization method according to claim 2, characterized in that updating the low-rank matrix includes: Based on the current iteration results, gradient updates are performed on the Huber robust data term and the graph Laplacian regularization term to obtain the intermediate matrix; The intermediate matrix is ​​subjected to singular value soft thresholding. Based on the time series characteristics of low-rank components, a one-dimensional fusion Lasso operator is used for smoothing and refinement.

[0009] In some implementations, updating the sparse matrix includes: Based on the current low-rank matrix, robust scaling estimates are calculated for the sparse matrix column by column, and the scaling estimates are used as the soft threshold shrinkage degree of the corresponding column. Based on the soft thresholds in the above columns, perform L1 soft threshold update on the sparse matrix to obtain the updated sparse components.

[0010] In some implementations, the drift and offset correction amounts are calculated to generate correction parameters, including: Extract the clean recovery signal of the network node and construct a time index vector; A robust linear model is fitted based on the time index vector, and correction parameters are set; the correction parameters include frequency drift correction and clock offset correction.

[0011] In some implementations, the convergence condition is that the difference between two adjacent target values ​​is less than a first threshold, or the number of iterations is not less than a second threshold.

[0012] In some implementations, wireless network time synchronization is performed based on calibration parameters, including: A synchronization signal, including correction parameters, is sent to each network node; the synchronization signal is used by each network node to update its local clock according to the correction parameters.

[0013] In some implementations, the objective function is:

[0014] in, Huber robust loss function; For nuclear norm; For the graph Laplacian regularization term; For time smoothing term, Let L be the i-th row of a low-rank matrix L, where each row corresponds to the time-series clock error of a specific node; For L1 sparse constraints; This is the robust residual loss term; The nuclear norm regularization coefficient; The graph regularization coefficient; The sparse regularization coefficient; is the time smoothing regularization coefficient.

[0015] Secondly, a wireless network time synchronization system is disclosed, comprising: The data acquisition module is used to collect local clock data of network nodes and timestamps of communication between nodes; The time synchronization algorithm processing module is used to collect time data from network nodes, form raw sampling sequences, and generate a network-wide clock observation matrix. Construct a graph Laplacian matrix based on the connection strength between each pair of network nodes; Based on the objective function of the graph regularized robust low-rank decomposition model, the entire network clock observation matrix is ​​jointly optimized and solved, and the low-rank matrix and sparse matrix are alternately updated. After the convergence condition is met, the drift and offset correction amounts are calculated to generate correction parameters; wireless network time synchronization is performed based on the correction parameters to achieve time alignment across the entire network.

[0016] Thirdly, a computer storage medium is disclosed, on which a computer program is stored, which, when executed by a processor, implements a wireless network time synchronization method as described in any of the above.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a wireless network time synchronization method, system, and storage medium. Based on this, a robust low-rank decomposition model with graph regularization is used to jointly optimize and solve the clock observation matrix of the entire network. The low-rank matrix and sparse matrix are updated alternately. Under conditions of noise, interference, and partial data loss, the main clock variation trend is robustly recovered and anomalous components are separated, thereby improving the accuracy of clock information extraction. After the iteration meets the convergence condition, this invention calculates the drift and offset correction amounts of each node based on the low-rank components and generates corresponding correction parameters, thereby adjusting the local clock of the network nodes. This invention achieves highly consistent time synchronization in multi-node wireless networks, contributing to improved overall system synchronization stability and anti-interference capability. Attached Figure Description

[0018] Figure 1 A schematic diagram of the system used in the wireless network time synchronization method provided by the present invention; Figure 2 A schematic diagram of a wireless network time synchronization method provided by the present invention; Figure 3 This is a flowchart illustrating step S2 in a wireless network time synchronization method provided by the present invention. Figure 4 This is a flowchart illustrating step S3 in a wireless network time synchronization method provided by the present invention. Figure 5 This is a flowchart illustrating step S5 in a wireless network time synchronization method provided by the present invention. Figure 6 This diagram illustrates a comparison of the root mean square error (RMSE) between the wireless network time synchronization method provided by this invention and different existing algorithms. Figure 7 This is a comparative schematic diagram of the synchronization error distribution (IQR) in a wireless network time synchronization method provided by the present invention. Detailed Implementation

[0019] To better understand and implement this invention, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] The terms “comprising” and “having” and any variations thereof in this invention are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products or devices.

[0021] The present invention discloses a wireless network time synchronization method, which solves the problem of unstable accuracy of existing synchronization algorithms under strong interference, packet loss and dynamic topology conditions, and achieves a balance between high accuracy and high robustness of the network clock.

[0022] Figure 1 This diagram illustrates the system model to which this method applies. It includes multiple network nodes, a time synchronization algorithm processing module, and a unified time base module for the entire network. These components interact wirelessly to achieve precise clock synchronization across the entire network. Figure 2 As shown, this method includes: Step S1: Collect network node time data to form the original sampling sequence and generate the network-wide clock observation matrix.

[0023] like Figure 1 As shown, network nodes may include network node A, network node B, network node C, network node D, etc., and the number is not limited. Each node contains a local clock to record the current time data of that node. In some embodiments, the current time data includes parameters representing the clock state, such as the transmission timestamp, the reception timestamp, and the node's current local time.

[0024] After receiving reports from all nodes, the time observation data from multiple network nodes spanning multiple sampling times are arranged according to node number and time order, forming a two-dimensional structured representation of multiple nodes and multiple times. This representation is organized into a network-wide clock observation matrix M, where the rows of the matrix correspond to different nodes in the network, the columns correspond to discrete sampling times, and the matrix elements are the local time observation values ​​of each node at the corresponding time. Due to potential packet loss, sudden delays, or temporary unavailability of some nodes in the wireless link, some observation values ​​are missing. In this step, missing locations are identified using a mask W. By constructing a mask matrix with the same dimension as the observation matrix, valid observations and invalid locations are distinguished, facilitating the robust low-rank decomposition to effectively recover the global time relationship between nodes even with partial data loss.

[0025] Step S2: Construct the graph Laplacian matrix based on the connection strength between each pair of network nodes; After obtaining the network-wide clock observation matrix, the topological relationships and spatial correlations between nodes in the network are described. A graph structure is constructed according to the physical location and proximity of network nodes, and a graph Laplacian matrix is ​​generated for subsequent regularization constraints. Specifically, such as... Figure 3 The following are included: Step S21: Assign the position of each network node in the two-dimensional space and calculate the pairwise distances between all network nodes.

[0026] Based on the network deployment, assign the location coordinates of each network node in two-dimensional space. Based on these locations, calculate the pairwise distances between all network nodes, for example, using Euclidean distance. Calculate the spatial distance for each pair of nodes for subsequent neighbor identification and connectivity strength assessment.

[0027] Step S22: For each network node, identify a preset number of nearest neighbors, establish connections between the network node and the nearest neighbors, determine the connection weights, and construct a weight matrix based on the connection weights; For each network node, a predetermined number of nearest neighbor nodes are identified based on the pairwise distance, and a connection is established between that node and its nearest neighbors. The connection weights are determined based on spatial proximity; closer nodes are assigned greater connection strength, and farther nodes are assigned less. Node pairs not included in the nearest neighbor set have a connection weight of zero. To ensure consistency, a weight matrix... Maintaining symmetry, that is .

[0028] Step S23: Accumulate the weights connected to each network node and construct the degree matrix.

[0029] For each network node, the weights of all connections connected to it are summed to form the node's degree. The degree values ​​of all nodes are then placed on the diagonal of the matrix to obtain the degree matrix. The degree matrix is ​​a diagonal matrix, with off-diagonal elements being zero.

[0030] Step S24: Construct the graph Laplacian matrix based on the degree matrix and weight matrix.

[0031] according to Calculate the graph Laplacian matrix. This Laplacian matrix characterizes the structural features of the node network: nodes that are spatially close appear in the weight matrix. This will create stronger connections, and when Laplace regularization is applied, the behavior coordination between adjacent nodes will be better.

[0032] Step S3: Based on the objective function of the graph regularized robust low-rank decomposition model, perform joint optimization to solve the entire network clock observation matrix, and alternately update the low-rank matrix and the sparse matrix.

[0033] The objective function is:

[0034] in, Huber robust loss function, used to suppress non-Gaussian impulse noise; : Nuclear norm, ensuring the low-rank structure of the clock observation matrix; The Laplacian regularization term in the graph describes the spatial topological relationships between nodes. : Time smoothing term, which limits abrupt shifts between adjacent sampling times. Let L represent the i-th row of the low-rank matrix L, where each row corresponds to the time-series clock error of a specific node. The low-rank matrix L is estimated during the optimization process. Capture the dynamic clock error of node i over time; L1 sparse constraints are used to identify and eliminate outlier observations. Robust residual loss term; : Nuclear norm regularization coefficient, used to control the strength of the low-rank constraint on the low-rank matrix L to ensure that the low-rank matrix L maintains its low rank; : Graph regularization coefficient, used to control the smoothness of the low-rank matrix L between nodes, and to force spatial smoothness through graph structure; : Sparse regularization coefficient, used to control the temporal smoothness of clock errors in each network node, penalizing large jumps in the time scale; : Time smoothing regularization coefficient, used to adjust the sparsity of the disturbance matrix S and shrink outliers.

[0035] The nuclear norm regularization coefficient, graph regularization coefficient, sparsity regularization coefficient, and time smoothing regularization coefficient are hyperparameters used to control low rank, local TV smoothing, global sparse noise, and graph structure smoothing. They are kept constant during optimization and their values ​​are selected before optimization through a small-scale grid search to minimize the root mean square error (RMSE) on the validation set. These coefficients strike a balance between data fidelity, low-rank structure, smoothness, and sparsity, thereby optimizing clock synchronization between network nodes.

[0036] Using the above objective function as the optimization objective, the low-rank matrix and sparse matrix are solved based on the alternating direction multiplier method (ADMM), as follows: Figure 4 As shown, it includes: Step S31: Update the low-rank matrix, including: Based on the current iteration results, gradient updates are performed on the Huber robust data term and the graph Laplacian regularization term to obtain the intermediate matrix; The intermediate matrix is ​​subjected to singular value soft thresholding. Based on the time series characteristics of low-rank components, a one-dimensional fusion Lasso operator is used for smoothing and refinement.

[0037] In this step, for the low-rank matrix L and sparse matrix S at the current iteration time, a gradient update is performed based on the structure of the Huber robust data term and the graph Laplacian regularization term to obtain an intermediate matrix representing the trend of the low-rank components. This intermediate matrix simultaneously reflects the original observation error, the suppression effect of robust loss on outliers, and the constraint of the graph structure on the coordination of clock behavior between adjacent nodes. After obtaining the intermediate matrix, it is processed by a singular value soft thresholding operation to remove high-order singular value components and extract the stable and smooth low-rank trend part.

[0038] Based on the physical characteristic that network node clocks exhibit piecewise linear drift over time, a one-dimensional fused Lasso operator is used to further refine each row of the low-rank matrix, resulting in a piecewise smooth change on the time axis that more closely resembles the behavior of actual clocks. The final low-rank matrix update is obtained by combining the results of the nuclear norm proximal update and the fused Lasso update.

[0039] Step S32: Update the sparse matrix, including: Based on the current low-rank matrix, robust scaling estimates are calculated for the sparse matrix column by column, and the scaling estimates are used as the soft threshold shrinkage degree of the corresponding column. Based on the soft thresholds of the columns mentioned above, L1 soft thresholding is performed on the sparse matrix to obtain the updated sparse components. After determining the column thresholds, L1 soft thresholding is performed on the sparse matrix to remove instantaneous abnormal disturbances or link jitter terms that are inconsistent with the low-rank model, automatically adapting to the time-varying intensity of abnormal interference and improving the accuracy of anomaly localization.

[0040] Based on the low-rank matrix L updated in step S31, adaptive soft thresholding is performed on the sparse matrix S column by column. For each time column, a robust scaling estimate is calculated, for example, using the median absolute deviation (MAD) to characterize the overall volatility level of the observations in that column. The calculated scaling estimate serves as the degree of soft threshold contraction for that column, allowing outliers from different time periods to be adaptively processed according to the magnitude of volatility.

[0041] Step S33: Update the multipliers. After updating the low-rank matrix and the sparse matrix, update the corresponding multipliers to ensure that the constraints of the alternating direction multiplier method gradually converge, so that the overall optimization process stably approaches the solution of the objective function.

[0042] Step S4: After the convergence condition is met, calculate the drift and offset correction amounts and generate correction parameters.

[0043] The convergence condition is that the difference between two consecutive target values ​​is less than a first threshold, or the number of iterations is not less than a second threshold K. The first threshold... A minimum relative change is set for the continuous iterations of the low-rank component L to determine whether the iterations have converged. The choice of this threshold depends on the balance between computational cost and numerical accuracy: a smaller threshold yields a more accurate solution but typically requires more iterations; a larger threshold can speed up the computation process at the expense of some accuracy. In common applications of matrix reconstruction and robust PCA, the tolerance is generally taken as... to Within the range. This method selects approximately... The tolerance value ensures stable convergence of the iterative process and maintains high computational efficiency in scenarios with time synchronization of multiple network nodes.

[0044] If the convergence condition is not met, repeat step S3 to continue iterative updates. After the convergence condition is met, calculate the drift and offset correction amounts, and generate correction parameters, such as... Figure 5 As shown, it includes: Step S51: Extract the clean recovery signal of the network node and construct a time index vector; Step S52: Fit a robust linear model based on the time index vector and set correction parameters; the correction parameters include frequency drift correction and clock offset correction.

[0045] After completing the low-rank and sparse decomposition and obtaining the low-rank matrix L, the frequency drift and static clock offset of each node are estimated using the clean time series provided by the low-rank matrix L. Each row of the low-rank matrix L corresponds to the recovered signal of a single node at multiple sampling times, which has eliminated the transient effects caused by burst noise, packet loss anomalies, and non-stationary interference, serving as a reliable input for subsequent drift estimation.

[0046] For network node i, extract its corresponding row vector as the clean recovery signal. A time index vector t = 1, 2, ..., T is constructed, and a robust linear model is fitted based on the recovered signal and the time index to estimate the frequency drift correction and clock offset correction. The regression form used is a simple linear model:

[0047] in This represents the frequency drift correction amount for network node i. This represents the static offset correction for network node i. To reduce the impact of residual outliers on the estimation, this embodiment employs robust least squares fitting to ensure that the obtained correction parameters better reflect the actual clock behavior of the node. The slope term in the fitting result is used as the frequency drift correction, and the intercept term is used as the clock offset correction. This process is performed independently for each node to generate its own frequency drift and clock offset correction parameters.

[0048] Decomposing the low-rank matrix L and the sparse matrix S before drift estimation is an important preprocessing step. The raw measurement data contains burst noise, outliers, link anomalies, and various non-stationary disturbances. Directly performing regression based on the raw data will lead to significant biases in drift estimation. The low-rank component L can capture the smooth trend of clock evolution over time, while the sparse component effectively separates anomalous disturbances, thus ensuring that the final drift and offset parameters reflect the true clock behavior of the nodes.

[0049] Step S5: Synchronize the wireless network time according to the calibration parameters.

[0050] After estimating the frequency drift correction and clock offset correction for each network node, a synchronization signal containing the time synchronization information of each network node is constructed based on the obtained correction parameters. This synchronization signal is then transmitted to all network nodes in the network via a wireless link.

[0051] Upon receiving the synchronization signal, each network node extracts the corresponding frequency drift correction and clock offset correction, and corrects them according to the current state of its local clock to align it with the unified time reference. After the local clock update is completed, the time of all nodes in the network is consistent, forming a unified reference time for the entire network, thus completing time synchronization.

[0052] The solution framework used in this application differs significantly from the traditional PRCA-ADMM method. The multi-node wireless clock synchronization scenario applied in this application suffers from heavy-tailed interference distribution, correlations between different nodes, time-varying noise levels, and a progressively smooth evolution of clock drift. Furthermore, actual observation data is often missing. To address these characteristics, this method introduces the following mechanism during the optimization process: (1) Use Huber robust data terms to adapt to sudden non-Gaussian peak disturbances; (2) Introduce graph Laplacian smoothness constraints to reflect the correlation behavior between spatially adjacent nodes; (3) Apply a total time variation penalty to each row of the low-rank matrix so that it can characterize the smoothness of the real clock over time; (4) Use a threshold that is dynamically adjusted with iteration in the sparse matrix update to adapt to the changes in the intensity of disturbances in different time periods; (5) Use a near-end gradient-ADMM hybrid structure so that the optimization process can simultaneously handle missing data and combinations of multiple regularization terms.

[0053] This method is more suitable for multi-node clock synchronization tasks in wireless networks, and can achieve stable and reliable network-wide time alignment in environments with complex interference and incomplete data.

[0054] To verify the time synchronization performance of the graph-regularized robust low-rank (GR-RLR) algorithm of this invention, this embodiment selects root mean square error (RMSE) and interquartile range (IQR) as core evaluation metrics.

[0055] RMSE measures the overall accuracy of time synchronization across the entire network. A smaller RMSE value indicates a more accurate estimation of clock skew and frequency drift by the algorithm. IQR reflects the dispersion of the synchronization error distribution; it is calculated as IQR = Q³. Q1, where Q3 and Q1 are the third and first quartiles of the error sample, respectively. The smaller the IQR value, the more concentrated the synchronization error is, the less the system is affected by abnormal disturbances, and the more robust the algorithm is in complex environments.

[0056] like Figure 6 As shown, the synchronization method based on GR-RLR proposed in this invention exhibits significant advantages under conditions of interference and packet loss. Compared with traditional consensus synchronization algorithms and conventional low-rank methods, GR-RLR reduces the time synchronization RMSE by approximately 25%, and the distribution of synchronization errors is more concentrated. This method demonstrates higher robustness and stability in complex wireless network scenarios.

[0057] Figure 7This paper compares the synchronization error distribution of our proposed method with other typical time synchronization algorithms in a wireless network environment. Box plots are used to illustrate the error statistics of various algorithms under the same experimental conditions, reflecting the differences in stability and robustness among the algorithms. The horizontal axis lists the Baseline, Fill+SVD, Soft-Impute, Go-Decomposition, and the GR-RLR method proposed in this invention, in that order; the vertical axis represents the time synchronization error (seconds), ranging from 0 to 1. Each box plot includes statistical elements such as the interquartile range (IQR), median, upper and lower whiskers, and outliers. The height of the box represents the dispersion of the error, the median reflects the central trend of the error, the upper and lower whiskers show the overall range of error expansion, and outliers represent a small number of abnormal node errors deviating from the main distribution, providing a more comprehensive evaluation of the stability performance of the synchronization algorithm under complex interference conditions.

[0058] from Figure 7 As can be observed, the Baseline method has the tallest and widest bin, indicating the most dispersed error distribution and significantly insufficient algorithm stability. The Fill+SVD and Soft-Impute methods show improvement under some noise conditions, but still exhibit numerous outliers. The Go-Decomposition algorithm has a narrower bin, indicating some convergence capability under packet loss conditions. Compared to these methods, the GR-RLR algorithm proposed in this invention has the narrowest bin, the lowest median, and the fewest outliers, demonstrating superior error suppression and robustness compared to other methods. Figure 7 As shown in the statistics in the upper right corner, the root mean square error (RMSE) of this method is reduced by about 25% compared with the traditional method, and the average synchronization error is significantly reduced. At the same time, its interquartile range (IQR) is reduced by about 46%, and the fluctuation range of the synchronization error is greatly reduced, indicating that it can still maintain a high degree of consistent synchronization performance in the presence of interference, packet loss and unstable link environments.

[0059] In summary, the time synchronization method based on graph regularized robust low-rank decomposition (GR-RLR) proposed in this invention exhibits higher accuracy and stability in complex wireless networks. This method reduces the mean squared error (RMSE) while effectively converging the individual quality rate (IQR), and possesses the ability to maintain synchronization consistency under scenarios of strong interference, frequent packet loss, and dynamic changes in network topology.

[0060] Combination Figure 6 and Figure 7 Experimental results show that this method outperforms existing technologies in both overall time synchronization accuracy and error distribution stability. In multi-node wireless environments, this invention achieves approximately 25% improvement in synchronization accuracy and approximately 46% improvement in stability.

[0061] Based on the same inventive concept, this invention also provides a schematic diagram of a system model to which this method is applicable, such as... Figure 1 As shown, it consists of multiple network nodes, a time synchronization algorithm processing module, and a unified time base module for the entire network. These components interact wirelessly to achieve precise clock synchronization across the entire network. Network nodes are interconnected via wireless signals and can periodically exchange timestamp data. Simultaneously, all network nodes upload their local clock observations to the central processing unit, i.e., the time synchronization algorithm processing module.

[0062] After receiving time observation data from all nodes in the network, the time synchronization algorithm processing module first constructs a network-wide clock observation matrix and a corresponding mask matrix based on the valid data to identify valid and missing observation locations. Then, it constructs the network topology based on signal strength, physical distance, or link quality between nodes and calculates the graph Laplacian matrix to represent the spatial correlation between nodes.

[0063] After the topology is established, the system uses a robust low-rank decomposition method to jointly optimize and solve the observation matrix, separating the low-rank component reflecting the global clock drift trend and the sparse component characterizing anomalous interference. Based on the recovered low-rank matrix, the frequency drift and static offset compensation of each node can be further calculated to form a parameter set for clock correction.

[0064] The time synchronization algorithm processing module sends a synchronization signal containing correction parameters to all wireless nodes. Upon receiving the corresponding frequency drift correction and clock offset correction, each node updates its local clock, gradually aligning all nodes in the network to a unified time base. After this synchronization task is completed, the unified time base module outputs the corrected network reference time. This reference time can be broadcast by the central node or propagated hierarchically within the distributed system.

[0065] The system architecture of this invention can be deployed on a network master node, wireless measurement equipment, or edge computing platform, and can support online operation and offline data processing. It is suitable for various application scenarios such as wireless communication, industrial IoT, low-altitude networks, and distributed measurement and control.

[0066] Based on the same inventive concept, the present invention also provides a computer device, comprising: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the steps of the wireless network time synchronization method described above.

[0067] The processing methods for computer devices can be referred to the description of the methods above, and will not be repeated here.

[0068] This application also provides a non-transitory machine-readable storage medium storing an executable program, which, when run by a microprocessor, causes the processor to execute the method provided in the above embodiments.

[0069] This invention discloses a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform the described methods.

[0070] This invention discloses a computer program product including a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform the described method.

[0071] The embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0072] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0073] Finally, it should be noted that the embodiments disclosed in this invention are merely preferred embodiments of this invention and are only used to illustrate the technical solutions of this invention, not to limit it. Although this invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention.

Claims

1. A wireless network time synchronization method, characterized in that, The method includes: Collect time data from network nodes to form raw sampling sequences and generate a network-wide clock observation matrix; Construct a graph Laplacian matrix based on the connection strength between each pair of network nodes; Based on the objective function of the graph regularized robust low-rank decomposition model, the entire network clock observation matrix is ​​jointly optimized and solved, and the low-rank matrix and sparse matrix are alternately updated. After the convergence condition is met, the drift and offset correction amounts are calculated, and the correction parameters are generated. Wireless network time synchronization is performed based on calibration parameters.

2. The wireless network time synchronization method according to claim 1, characterized in that, Based on the connection strength between each pair of network nodes, construct the graph Laplacian matrix, including: Assign a position to each network node in a two-dimensional space, and calculate the pairwise distances between all network nodes; For each network node, identify a preset number of nearest neighbors, establish connections between the network node and the nearest neighbors, and determine the connection weights; Construct a weight matrix based on connection weights; Accumulate the weights connected to each network node to construct a degree matrix; Based on the degree matrix and weight matrix, construct the graph Laplacian matrix.

3. The wireless network time synchronization method according to claim 2, characterized in that, Update the low-rank matrix, including: Based on the current iteration results, gradient updates are performed on the Huber robust data term and the graph Laplacian regularization term to obtain the intermediate matrix; The intermediate matrix is ​​subjected to singular value soft thresholding. Based on the time series characteristics of low-rank components, a one-dimensional fusion Lasso operator is used for smoothing and refinement.

4. The wireless network time synchronization method according to claim 3, characterized in that, Updating the sparse matrix includes: Based on the current low-rank matrix, robust scaling estimates are calculated for the sparse matrix column by column, and the scaling estimates are used as the soft threshold shrinkage degree of the corresponding column. The sparse matrix is ​​updated using L1 soft thresholding based on the column soft thresholds to obtain the updated sparse components.

5. A wireless network time synchronization method according to claim 2, characterized in that, Calculate drift and offset corrections, and generate correction parameters, including: Extract the clean recovery signal of the network node and construct a time index vector; A robust linear model is fitted based on the time index vector, and correction parameters are set; the correction parameters include frequency drift correction and clock offset correction.

6. The wireless network time synchronization method according to claim 3, characterized in that, The convergence condition is that the difference between two adjacent target values ​​is less than a first threshold, or the number of iterations is not less than a second threshold.

7. A wireless network time synchronization method according to claim 6, characterized in that, Wireless network time synchronization is performed based on calibration parameters, including: A synchronization signal, including correction parameters, is sent to each network node; the synchronization signal is used by each network node to update its local clock according to the correction parameters.

8. A wireless network time synchronization method according to claim 7, characterized in that, The objective function is: in, Huber robust loss function; For nuclear norm; For the graph Laplacian regularization term; For time smoothing term, Let L be the i-th row of a low-rank matrix L, where each row corresponds to the time series clock error of a specific node; For L1 sparse constraints; This is the robust residual loss term; The nuclear norm regularization coefficient; The graph regularization coefficient; The sparse regularization coefficient; is the time smoothing regularization coefficient.

9. A wireless network time synchronization system, characterized in that, include: The data acquisition module is used to collect local clock data of network nodes and timestamps of communication between nodes; The time synchronization algorithm processing module is used to collect time data from network nodes, form raw sampling sequences, and generate a network-wide clock observation matrix. Construct a graph Laplacian matrix based on the connection strength between each pair of network nodes; Based on the objective function of the graph regularized robust low-rank decomposition model, the entire network clock observation matrix is ​​jointly optimized and solved, and the low-rank matrix and sparse matrix are alternately updated. After the convergence condition is met, the drift and offset correction amounts are calculated to generate correction parameters; wireless network time synchronization is performed based on the correction parameters to achieve time alignment across the entire network.

10. A computer storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the steps of a wireless network time synchronization method as claimed in any one of claims 1-8.

Images