Precise clock synchronization parameter tracking method based on augmented kalman neural network

Abstract

The present invention relates to the field of time synchronization, and relates to a precise clock synchronization parameter tracking method based on an augmented Kalman neural network. In the method, for a multi-hop network scenario where an ambient temperature changes and a link transmission delay is asymmetric, taking into account the impact of an accumulated asymmetric link transmission delay in a multi-hop network on clock synchronization accuracy, a multi-hop network clock observation equation containing an accumulated asymmetric link delay is derived, and on the basis of a quadratic polynomial model of a temperature and a clock frequency offset, a recursive state equation of the clock frequency offset and a temperature change is established; the evolution processes of an accumulated asymmetric delay and an unknown temperature coefficient are modeled as a first-order linear difference equation, and the accumulated asymmetric delay and the unknown temperature coefficient are augmented to a clock parameter tracking state space model; and an augmented Kalman neural network algorithm is used to realize joint tracking of a clock phase offset and a clock frequency offset of a slave node. This method employs a hybrid data-model-driven approach to perform joint tracking of the clock phase offset and frequency offset, thereby improving accuracy and robustness of clock parameter tracking.

Description

A precise clock synchronization parameter tracking method based on augmented Kalman neural network Technical Field

[0001] This invention belongs to the field of time synchronization and relates to a method for tracking accurate clock synchronization parameters based on an augmented Kalman neural network. Background Technology

[0002] Time synchronization provides a unified time reference for many time-critical applications in networks, such as data fusion, resource scheduling, energy management, and collaborative measurement, and is a fundamental requirement for ensuring the normal operation of the network. Traditional time synchronization protocols include: Global Positioning System (GPS), Network Time Protocol (NTP), and Precise Time Protocol (PTP). Precise Time Protocol has become one of the most commonly used protocols in distributed networks due to its fewer limitations, high cost-effectiveness, and ability to ensure efficient distributed interaction.

[0003] The key to accurate clock synchronization is eliminating the clock phase offset from the reference clock, which is corrected by collecting hardware timestamps. However, due to the accumulation of clock frequency offset, the calibrated clock may still experience unbounded clock phase offset, leading to frequent resynchronization. Therefore, to extend the synchronization interval, it is necessary to track both clock phase offset and clock frequency offset simultaneously. Existing studies on the joint tracking of clock phase offset and clock frequency offset for accurate clock synchronization in multi-hop networks mostly assume symmetrical uplink and downlink delays, thus employing traditional Kalman filters for joint tracking of clock phase offset and frequency offset. However, these studies have the following shortcomings: 1) In actual multi-hop networks, due to routing asymmetry, line speed asymmetry, and variable processing delays of intermediate devices, the uplink and downlink delays of the master clock and the clock to be synchronized are asymmetric. 2) Affected by the variable temperature of the network deployment environment, temperature changes will affect the inherent characteristics of the clock crystal, causing changes in the clock crystal output frequency, and thus generating unbounded clock phase offset. 3) In practical multi-hop networks, the observation noise covariance is difficult to be given empirically or the given observation noise covariance is not accurate. Traditional Kalman filtering is very sensitive to the observation noise covariance. Unknown observation noise covariance will lead to a decrease in the tracking performance of clock parameters of Kalman filtering.

[0004] Based on the above-mentioned shortcomings, this invention urgently needs to solve the problem of joint reliable tracking of accurate clock phase offset and frequency offset in multi-hop networks under the influence of accumulated time delay asymmetry and temperature changes and unknown observation noise covariance.

[0005] Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a precise clock synchronization parameter tracking method based on an augmented Kalman neural network. In a multi-hop network consisting of a source node, a target node, and multiple intermediate nodes, the impact of cumulative asymmetric link transmission delay on clock synchronization accuracy is considered. A clock observation equation for the multi-hop network containing cumulative asymmetric link delay is derived. A recursive state equation for clock frequency offset and temperature change is established based on a quadratic polynomial model of temperature and clock frequency offset. The evolution process of cumulative asymmetric delay and unknown temperature coefficient is modeled as a first-order linear difference equation and augmented to the clock parameter tracking state space model. The augmented Kalman neural network algorithm is used to achieve joint tracking of clock phase offset and clock frequency offset of slave nodes. This invention considers the impact of cumulative asymmetric link delay and temperature change in the multi-hop network. By using a quadratic augmented Kalman filter, the cumulative asymmetric link delay and unknown temperature coefficient are augmented to the state space vector, compensating for the impact of asymmetric delay and temperature change on the tracking clock phase offset and frequency offset. Furthermore, this invention introduces a deep learning method, which learns the Kalman gain through a neural network, overcoming the problem of unknown observation noise covariance in multi-hop networks, and significantly improving the accuracy and robustness of clock parameter tracking.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for accurate clock synchronization parameter tracking based on an augmented Kalman neural network, the method specifically includes the following steps:

[0009] S1: According to the precise clock protocol in the multi-hop network, synchronization messages are sent in a point-to-point mode. Considering the cumulative asymmetric delay, the target node establishes a clock observation equation based on the received synchronization messages. Considering the influence of temperature on the clock crystal oscillator, a recursive model of node clock frequency offset related to temperature is established. Based on the two-state clock model, a clock state equation is established.

[0010] S2: Combine the clock observation equation and the clock state equation to establish a clock state space model, model the unknown temperature coefficient and cumulative asymmetric delay as a first-order linear difference equation, and augment the unknown temperature coefficient and cumulative asymmetric delay to the state space vector.

[0011] S3: The clock parameters are tracked using an augmented Kalman neural network. The observed and true values ​​of the clock state parameters are used as the training set, and a recurrent neural network is used to train the clock parameters to obtain a prediction model. The prediction gain output by the prediction model is obtained by inputting the posterior estimate of the previous time step and the observed value of the current time step into the prediction model. This prediction gain is used to weigh the confidence of the observed values ​​and the predicted values.

[0012] Furthermore, in S1, the observation equations and state equations for accurate clock synchronization of the multi-hop network under the influence of cumulative asymmetric time delay and ambient temperature are established. The specific steps are as follows:

[0013] S11: Intermediate node i calculates the clock offset from the master clock using the received timestamp. Represented as:

[0014] in, Indicates the time when the master clock sends the Sync message. Indicates downstream transparent clock TC i The time it takes to receive the Sync message; Represented as:

[0015] in, This represents the link delay between transparent clock i and node i-1. This indicates the residence time of the synchronization message on node i-1. This represents the total delay from the previous node i to the master clock, with an initial value of [value missing]. The dwell time of node i-1 is zero; the dwell time of node i-1 can be calculated as:

[0016] Where ε (i-1,i-1) It is a random variable, usually assumed to be Gaussian white noise, representing the contribution of timestamps to different clock readings;

[0017] S12: Synchronize in the k-th round, assuming the clock offset between node i and node i-1 is... Furthermore, assuming that the clock phase offset remains constant over a sufficiently short time interval, then:

[0018] in, and These represent the uplink and downlink fixed delays between nodes i and i-1, respectively, and their relationship is as follows: in This represents a fixed delay asymmetric measure between node i and node i-1 during the k-th round of synchronization;

[0019] S13: Assume the average link delay is... but It can be represented as:

[0020] S14: The formula for measuring the clock offset between the master clock and the transparent clock i is:

[0021] in, The measured value representing clock offset. and These correspond to the accumulated dwell time and accumulated unknown asymmetric delay of the synchronization information between the master clock and the transparent clock i, respectively. The total observation noise is represented as:

[0022] in This represents the measurement error of the i-th node during the k-th round of synchronization; in a multi-hop network, It is difficult to acquire through experience, let's assume The signal is a Gaussian noise signal with zero mean and unknown covariance;

[0023] S15: Clock frequency offset measurement between the master clock and transparent clock i The value can be measured by clock phase offset. Given:

[0024] S16: Establish a parabolic model of the unknown coefficients related to clock frequency offset and temperature change, and treat interference as noise input. If the sampling interval is small enough, the recursive equation for clock frequency offset is modeled as follows:

[0025] in, It is the unknown relationship between the current clock frequency offset of intermediate node i and temperature change;

[0026] S17: For curves with unknown parabolic coefficients, The change is related to the central axis of the parabola, which is the reference temperature. Celsius, approximately:

[0027] in, The unknown temperature coefficient of node i at the synchronization point in the kth round represents an unknown value that fluctuates within a small range.

[0028] The recursive form of the clock phase offset can then be obtained as:

[0029] S18: When designing the state equations, a two-state clock model is used to determine the clock phase offset to illustrate the clock's behavior at consecutive synchronization moments; the clock recursive state equations are shown below:

[0030] Furthermore, in step S2, a clock state-space model is established, and the unknown temperature coefficient and cumulative asymmetric delay are modeled as first-order linear difference equations. The unknown temperature coefficient and cumulative asymmetric delay are then augmented to a state-space vector. The specific steps are as follows:

[0031] S21: Combining the clock state equation and clock observation equation of intermediate node i, the clock state space equation is:

[0032] in B = [1 0] T ,

[0033] S22: The cumulative asymmetric delay between the master clock and the transparent clock i is expressed as:

[0034] The asymmetric delay between two adjacent transparent clocks is modeled as a first-order difference equation:

[0035] This represents Gaussian noise with zero mean;

[0036] S23: The recursive form of the cumulative asymmetric delay is obtained as follows:

[0037] The covariance is difficult to obtain accurately due to the accumulation of multiple zero-mean Gaussian noises.

[0038] S24: Similarly, the unknown temperature coefficient can be modeled as a first-order difference equation:

[0039] S25: Augmenting the accumulated asymmetric delay and the unknown temperature coefficient to the state space vector, we obtain the augmented clock state space model as follows:

[0040] in,

[0041] Furthermore, in step S3, an augmented Kalman neural network is used to track clock parameters. The specific steps are as follows:

[0042] S31: Calculate the prior state estimate of transparent node i based on the augmented state space model. The calculation formula is:

[0043] in It is the posterior estimate of node i at the previous time step;

[0044] S32: Use the measurement at node i at time k. and prior estimates residual Calculated as:

[0045] S33: The method based on the augmented Kalman neural network uses a neural network to learn the Kalman gain Through the learned Perform the execution state update, expressed as:

[0046] S34: Kalman gain The calculation method is to optimize the mean square error loss function of parameter Θ through end-to-end training; the data set consists of N sequences with length m, expressed as where is the observed vector sequence of trajectory j, is the true value sequence of trajectory j; extend the recursion to m time steps to obtain a sequence-to-sequence supervised training scheme; for each trajectory j, the empirical minimum mean square error loss function metric is:

[0047] where, Ψ Θ (.) is the output of the recurrent neural network, Θ is its trainable parameter, and α is the regularization coefficient;

[0048] S35: To optimize parameter Θ, an adaptive momentum stochastic optimization method is adopted, which combines second-order momentum with stochastic gradient descent. For each batch of trajectories indexed by r, randomly select L < N training indices and calculate the batch loss ξ r (Θ) is:

[0049] The beneficial effects of the present invention are as follows:

[0050] (1) The present invention fully considers the influence of the cumulative uplink and downlink transmission delay asymmetry on the clock phase offset and the influence of temperature change on the clock frequency offset in the actual multi-hop network, and uses the augmented Kalman neural network algorithm to jointly track the clock phase offset and frequency offset in the multi-hop network, significantly reducing the influence of temperature change and cumulative asymmetric delay on the clock parameter tracking accuracy in the multi-hop network.

[0051] (2) The present invention uses a neural network to learn the Kalman gain, and weighs the trustworthiness of the estimated value and the measured value through the learned gain, overcoming the problem of the decline in the joint tracking accuracy of the clock phase offset and frequency offset caused by the unknown observation noise covariance, and significantly improving the accuracy and robustness of the clock parameter tracking.

[0052] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0053] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0054] Figure 1 is a schematic diagram of the multi-hop network clock synchronization information interaction in this invention;

[0055] Figure 2 shows the flowchart of the precise clock synchronization parameter tracking method based on augmented Kalman neural network in this invention;

[0056] Figure 3 is a performance comparison chart of the accurate clock phase offset tracking method based on augmented Kalman neural network in this invention;

[0057] Figure 4 is a performance comparison chart of the accurate clock frequency offset tracking method based on augmented Kalman neural network in this invention. Detailed Implementation

[0058] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0059] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0060] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0061] Please refer to Figures 1 to 4. Figure 1 is a schematic diagram of time message interaction in multi-hop network precise clock synchronization provided by the present invention. As shown in Figure 1, the master clock and the slave clock exchange dedicated synchronization messages for precise clock synchronization through multiple intermediate transparent clock nodes. The specific process is as follows:

[0062] 1) In the k-th round of synchronization messages, the master clock is in A Sync message is sent continuously and forwarded by an intermediate transparent clock to the slave clock. The slave clock records the moment the message is received. 2) After the Sync message is sent, the master clock will send a Follow_Up message. Transparent clock TC is forwarded from the intermediate n-1 cascaded transparent clocks to the slave clocks; i-1 After receiving the Sync message, record the time of receipt. The message is then forwarded from other active ports to the downstream transparent clock TC. i TC i Record its receiving time And record the time of its forwarding. Meanwhile, transparent clock TC i-1 Calculate the dwell time of the Sync message as follows Transparent clock TC i-1 The processing delay from receiving the Sync message to forwarding it is accumulated and recorded in the correctionField field of the Follow_Up message; 4) Transparent clock TC i Synchronization messages are obtained in TC using a point-to-point delay measurement method. i-1 With TC i The average link transmission delay between them. Specifically, TC i Send a Pdelay_Req message to the upstream clock and record its sending time as... After receiving the message, the upstream clock records the reception time as follows: And in At any time, towards TC i Reply to the Pdelay_Res message. TC i After receiving the message, record the time of receipt. Subsequently, the upstream clock will send a Pdelay_Res_Follow_Up message. Send to TC i At this time, the transparent clock TC i Collected Four timestamps are used, and TC is calculated under the assumption of symmetric link delay. i-1 With TC i The average link transmission delay between them. This is also recorded in the `correctionField` field of the `Follow_Up` message;

[0063] Intermediate node i calculates the clock offset from the master clock using the received timestamp. It can be represented as:

[0064] in This represents the total path delay from the master clock to the intermediate transparent clock node i. It can be represented as:

[0065] in, This represents the link delay between transparent clock i and node i-1. This indicates the residence time of the synchronization message on node i-1. This represents the total delay from the previous node i to the master clock, with an initial value of [value missing]. The dwell time of node i-1 is zero.

[0066] Where ε (i-1,i-1) It is a random variable, usually assumed to be Gaussian white noise, representing the contribution of timestamps to different clock readings.

[0067] In the k-th round of synchronization, node i collects data through the correctionField field. A timestamp can be used to calculate the clock offset between node i and node i-1. for

[0068] in, and These represent the uplink and downlink fixed delays between nodes i and i-1, respectively, and their relationship is as follows: in This represents the fixed delay asymmetry measure between node i and node i-1 during the k-th round of synchronization.

[0069] Assume the average link delay is set to Subtracting the two formulas in equation (4), and further assuming that the clock phase offset remains constant over a sufficiently short time interval, the average link delay can be calculated as follows:

[0070] Combining formulas (1), (2), (3), and (5), the formula for the clock phase offset between the master clock and the transparent clock i can be obtained.

[0071] in, The measured value representing clock offset. and These correspond to the cumulative dwell time and cumulative asymmetric link delay of the synchronization information between the master clock and the transparent clock i, respectively. The total observation noise can be represented as:

[0072] in This represents the measurement error of the i-th node during the k-th round of synchronization. It is difficult to obtain through experience, therefore it is assumed It is an unknown noise signal with zero mean and covariance.

[0073] Clock frequency offset measurement between master clock and transparent clock i The value can be measured by clock phase offset. The calculation shows that:

[0074] Furthermore, a recursive form for the clock frequency offset in the multi-hop model is established, with the following specific steps:

[0075] A parabolic model of the unknown coefficients related to clock frequency offset and temperature change is established, and interference is treated as noise input. If the sampling interval is small enough, the recursive equation for clock frequency offset can be modeled as follows:

[0076] in, It is the unknown relationship between the current clock frequency offset of intermediate node i and temperature change.

[0077] For a curve with unknown parabolic coefficients, The change is related to the central axis of the parabola (i.e., the reference temperature). It is related to (degrees Celsius) and can be approximated as:

[0078] in, The unknown temperature coefficient of node i at the synchronization point in the kth round represents an unknown value that fluctuates within a small range.

[0079] Combining equations (9) and (10), the recursive form of the clock phase offset can be obtained as follows:

[0080] When designing the state equations, a two-state clock model is used to determine the clock phase offset to explain the behavior of the clock at continuous synchronization moments. Combining equation (11), the clock recursive state equations are as follows:

[0081] Furthermore, a clock state-space model is established, modeling the unknown temperature coefficient and cumulative asymmetric delay as first-order linear difference equations. Simultaneously, the unknown temperature coefficient and cumulative asymmetric delay are augmented to the state-space vector. The specific steps are as follows:

[0082] Combining equations (6), (8), and (12), the clock state space equation for intermediate node i is:

[0083] in

[0084] B = [1 0] T ,

[0085] Subsequently, the cumulative asymmetric link delay and the unknown temperature coefficient will be augmented. First, the cumulative asymmetric delay between the master clock and the transparent clock i is expressed as:

[0086] The asymmetric delay between two adjacent transparent clocks can be modeled as a first-order difference equation:

[0087] This represents Gaussian noise with zero mean.

[0088] Combining equations (14) and (15), the recursive form of the cumulative asymmetric delay can be obtained as follows:

[0089] The covariance is difficult to obtain accurately due to the accumulation of multiple zero-mean Gaussian noises.

[0090] Similarly, the unknown temperature coefficient can be modeled as a first-order difference equation:

[0091] Augmenting the accumulated asymmetric delay and the unknown temperature coefficient to the state space vector, we obtain the augmented clock state space model as follows:

[0092] Among them,

[0093] Furthermore, an accurate clock synchronization parameter tracking method based on the augmented Kalman neural network is adopted. The specific steps are as follows:

[0094] Calculate the prior state estimate value of node i according to the augmented state space model The calculation formula is:

[0095] Among them is the posterior estimate value of node i at the previous moment.

[0096] Use the measurement value of node i at time k and the prior estimate value The residual can be calculated:

[0097] The method based on the augmented Kalman neural network uses the Kalman gain learned by the neural network to perform state update, expressed as:

[0098] Kalman gain The calculation method is to optimize the mean square error loss function of parameter Θ through end-to-end training. The dataset consists of N sequences with length m, expressed as Among them is the observed vector sequence of trajectory j, is the true value sequence of trajectory j. Extend the recursion to m time steps, thus obtaining a sequence-to-sequence supervised training scheme. For each trajectory j, the empirical minimum mean square error loss function metric is:

[0099] Among them, Ψ Θ (.) is the output of the recurrent neural network, Θ is its trainable parameter, and α is the regularization coefficient.

[0100] To optimize parameter Θ, an adaptive momentum stochastic optimization method is adopted. This method combines second-order momentum with stochastic gradient descent and has the advantages of simple implementation, high computational efficiency, and small memory overhead. For each batch of trajectories indexed by r, randomly select L < N training indices and calculate the batch loss ξ r (Θ) is

[0101] Figure 2 is a flowchart of the accurate clock synchronization parameter tracking method based on augmented Kalman neural network in this embodiment. This embodiment provides a joint tracking method for clock phase offset and frequency offset suitable for multi-hop networks under the influence of temperature changes and link delay asymmetry, considering the unknown observation noise covariance. As shown in Figure 2, it specifically includes the following steps:

[0102] P1: The process of tracking precise clock synchronization parameters based on augmented Kalman neural networks begins;

[0103] P2: Initialization based on augmented Kalman neural network method;

[0104] P3: The slave clock performs precise clock synchronization with the master clock through an intermediate transparent clock, and records relevant timestamps.

[0105] P4: Based on the recorded timestamps, obtain the observed values ​​of clock phase offset and frequency offset;

[0106] P5: Obtain the temperature at time K using a temperature sensor and establish a clock state equation;

[0107] P6: Predict the prior state estimates of clock phase offset and frequency offset at time K using a clock state model;

[0108] P7: Calculate the filter residuals based on the observed and predicted values;

[0109] P8: Input the current observation value and the posterior estimate value of the previous time step, and use a neural network to calculate the Kalman gain;

[0110] P9: The gain calculated by the neural network is used to weight the predicted value and the observed value, and the posterior estimate is updated.

[0111] P10~P12: Determine whether the synchronization cycle has reached the preset value. If it has, the clock phase offset tracking process ends; otherwise, the synchronization cycle is incremented by 1, and the process continues in P3 to continue the clock parameter tracking process.

[0112] Figures 3 and 4 show a performance comparison of the accurate clock synchronization parameter tracking method based on augmented Kalman neural networks provided in this embodiment. As can be seen from Figures 3 and 4, the augmented Kalman neural network-based method effectively eliminates the impact of accumulated asymmetric link delay and compensates for rapid environmental temperature changes on the accurate tracking of clock phase and frequency offsets, thus tracking the true values ​​of clock phase and frequency offsets more accurately and reliably. This demonstrates the practicality of the accurate clock synchronization parameter tracking method based on augmented Kalman neural networks provided in this invention.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for accurate clock synchronization parameter tracking based on an augmented Kalman neural network, characterized in that: The method specifically includes the following steps: S1: According to the precise clock protocol in the multi-hop network, synchronization messages are sent in a point-to-point mode. Considering the cumulative asymmetric delay, the target node establishes a clock observation equation based on the received synchronization messages. Considering the influence of temperature on the clock crystal oscillator, a recursive model of node clock frequency offset related to temperature is established. Based on the two-state clock model, a clock state equation is established. S2: Combine the clock observation equation and the clock state equation to establish a clock state space model, model the unknown temperature coefficient and cumulative asymmetric delay as a first-order linear difference equation, and augment the unknown temperature coefficient and cumulative asymmetric delay to the state space vector. S3: The clock parameters are tracked using an augmented Kalman neural network. The observed and true values ​​of the clock state parameters are used as the training set, and a recurrent neural network is used to train the clock parameters to obtain a prediction model. The prediction gain output by the prediction model is obtained by inputting the posterior estimate of the previous time step and the observed value of the current time step into the prediction model. This prediction gain is used to weigh the confidence of the observed values ​​and the predicted values.

2. The precise clock synchronization parameter tracking method based on augmented Kalman neural network according to claim 1, characterized in that: In step S1, the precise clock synchronization observation equation and state equation of the multi-hop network under the influence of cumulative asymmetric time delay and ambient temperature are established. The specific steps are as follows: S11: Intermediate node i calculates the clock offset from the master clock using the received timestamp. Represented as: in, Indicates the time when the master clock sends the Sync message. Indicates downstream transparent clock TC i The time to receive the Sync message, and This represents the total path delay from the master clock to the intermediate transparent clock node i; Represented as: in, This represents the link delay between transparent clock i and node i-1. This indicates the residence time of the synchronization message on node i-1. This represents the total delay from the previous node i to the master clock, with an initial value of [value missing]. The dwell time of node i-1 is zero; the dwell time of node i-1 is calculated as follows: Where ε (i-1,i-1) It is a random variable, assumed to be Gaussian white noise, representing the contribution of timestamps to different clock readings; S12: Synchronize in the k-th round, let the clock offset between node i and node i-1 be . Assuming the clock phase offset remains constant over a sufficiently short time interval, then: in, and These represent the uplink and downlink fixed delays between nodes i and i-1, respectively, and their relationship is as follows: in This represents a fixed delay asymmetric measure between node i and node i-1 during the k-th round of synchronization; S13: Assume the average link delay is... but Represented as: S14: The formula for measuring the clock offset between the master clock and the transparent clock i is: in, The measured value representing clock offset. and These correspond to the accumulated dwell time and accumulated unknown asymmetric delay of the synchronization information between the master clock and the transparent clock i, respectively. The total observation noise is represented as: in Let represent the measurement error of the i-th node under the k-th round of synchronization; in a multi-hop network, let The signal is a Gaussian noise signal with zero mean and unknown covariance; S15: Clock frequency offset measurement between the master clock and transparent clock i Measurements based on clock phase offset Given: S16: Establish a parabolic model of the unknown coefficients related to clock frequency offset and temperature change, and treat interference as noise input. If the sampling interval is small enough, the recursive equation for clock frequency offset is modeled as follows: in, It is the unknown relationship between the current clock frequency offset of intermediate node i and temperature change; S17: For curves with unknown parabolic coefficients, The change is related to the central axis of the parabola, which is the reference temperature. Celsius, approximately: in, The unknown temperature coefficient of node i at the synchronization point in the kth round represents an unknown value that fluctuates within a small range. The recursive form of the clock phase offset is: S18: When designing the state equations, a two-state clock model is used to determine the clock phase offset to illustrate the clock's behavior during continuous synchronization time; the clock recursive state equations are shown below: Where τ represents the sampling interval.

3. The precise clock synchronization parameter tracking method based on augmented Kalman neural network according to claim 2, characterized in that: In step S2, a clock state-space model is established, and the unknown temperature coefficient and cumulative asymmetric delay are modeled as first-order linear difference equations. The unknown temperature coefficient and cumulative asymmetric delay are then augmented to a state-space vector. The specific steps are as follows: S21: Combining the clock state equation and clock observation equation of intermediate node i, the clock state space equation is: in B＝[1 0] T ， S22: The cumulative asymmetric delay between the master clock and the transparent clock i is expressed as: The asymmetric delay between two adjacent transparent clocks is modeled as a first-order difference equation: This represents Gaussian noise with zero mean; S23: The recursive form of the cumulative asymmetric delay is obtained as follows: The covariance is difficult to obtain accurately due to the accumulation of multiple zero-mean Gaussian noises. S24: Similarly, the unknown temperature coefficient can be modeled as a first-order difference equation: S25: Augmenting the accumulated asymmetric delay and the unknown temperature coefficient to the state space vector, we obtain the augmented clock state space model as follows: in, 4. The precise clock synchronization parameter tracking method based on augmented Kalman neural network according to claim 3, characterized in that: In step S3, an augmented Kalman neural network is used for clock parameter tracking. The specific steps are as follows: S31: Calculate the prior state estimate of transparent node i based on the augmented state space model. The calculation formula is: in It is the posterior estimate of node i at the previous time step; S32: Use the measurement at node i at time k. and prior estimates residual Calculated as: S33: The method based on augmented Kalman neural networks uses neural networks to learn Kalman gain. Through learning To perform an execution status update, it is represented as: S34: Kalman Gain The calculation method involves optimizing the mean squared error loss function of parameter Θ through end-to-end training; the dataset consists of N sequences of length m, represented as... in It is the sequence of observation vectors for trajectory j. It is the truth sequence of trajectory j; Extending the recursion to m time steps yields a sequence-to-sequence supervised training scheme; for each trajectory j, the empirical minimum mean square error loss function is: Among them, Ψ Θ (.) is the output of the recurrent neural network, Θ is its trainable parameter, and α is the regularization coefficient; S35: To optimize parameter Θ, an adaptive momentum stochastic optimization method is adopted, combining second-order momentum with stochastic gradient descent. For each batch of trajectories indexed by \(r\), randomly select \(L < N\) training indices and compute the batch loss \(\xi\) r for \((\Theta)\) is:

Images