An adaptive bidirectional forwarding detection method

Abstract

The application discloses a kind of self-adapting bidirectional forwarding detection method, belong to space communication and network technical field.This method is based on the change rule of link propagation characteristics to construct delay prediction and detection parameter adaptive adjustment mechanism: first, satellite link historical round-trip delay sequence is acquired, parameterized modeling and compression are carried out to the propagation round-trip delay in window, and light detection parameter configuration file is generated;Satellite-borne node recovers delay prediction model according to the parameter configuration file, and real-time obtains link round-trip delay prediction result, and dynamically adjusts bidirectional forwarding detection control message sending period and fault detection timeout threshold according to this;Further, the timeout determination mechanism is established by combining the delay change rule in the transmission process of control message, to reduce the influence of space link dynamic propagation delay on fault detection accuracy.

Description

Technical Field

[0001] This invention belongs to the field of space communication and network technology, and specifically relates to an adaptive bidirectional forwarding detection method. Background Technology

[0002] Bidirectional Forwarding Detection (BFD) for satellite links, as the de facto standard protocol for link status monitoring, was initially designed primarily for relatively static terrestrial network architectures. It determines link status by having communicating parties exchange probe messages at a fixed frequency and wait for a response from the other end within a timeout threshold. However, when directly deployed in highly dynamic space network topologies, it reveals significant limitations. Taking the BeiDou Navigation Satellite System (BDS-3) hybrid constellation as an example, its inter-satellite link network is interwoven with Inclined Geosynchronous Orbit (IGSO) and Medium Earth Orbit (MEO) satellites. Due to the significant relative motion between satellites at different orbital altitudes, the inter-satellite round-trip time (RTT) can vary drastically from tens to hundreds of milliseconds within a typical link visibility window. Simultaneously, queuing jitter, random packet loss, and Doppler shift in the space channel further exacerbate the difficulty of link status monitoring.

[0003] Traditional static BFD protocols face the challenge of simultaneously optimizing the detection period and false positive probability in dynamic environments: If a short transmission period is configured to pursue state detection sensitivity, a large number of control messages will accumulate in the channel when the distance between satellites increases significantly, leading to excessive bandwidth consumption; simultaneously, the high-frequency detection mechanism causes the timeout threshold to be frequently breached by sudden jitter, greatly increasing the probability of false link failures. Conversely, if a longer transmission period is configured, the timeliness of state monitoring is lost when satellites are close together. Existing research on fixed-parameter BFD shows that any static parameter configuration can only achieve local optima within a specific link period and cannot cover the dynamic changes throughout the entire visible window of the link.

[0004] Furthermore, some studies have explored applying machine learning methods to satellite link status prediction. However, onboard processors are unlikely to support real-time inference of complex deep learning models in the short term, and model training relies on a large amount of historical telemetry data, making on-orbit updates significantly difficult. Therefore, how to overcome the performance bottleneck of traditional static BFD protocols in dynamic inter-satellite links under the dual constraints of onboard computing power and inter-satellite bandwidth has become a key problem that urgently needs to be solved in the field of space networks. Summary of the Invention

[0005] The purpose of this invention is to address the problem that the traditional BFD protocol, which uses a fixed detection period and a static timeout threshold, is not adaptable to changes in link state under highly dynamic inter-satellite link environments. An adaptive bidirectional forwarding detection method is proposed, which achieves adaptive fault detection of highly dynamic satellite links by constructing a lightweight propagation delay prediction model and combining it with a dynamic transmission period and a dynamic timeout detection threshold.

[0006] In highly dynamic space networks, satellite link propagation delays continuously change with the relative motion of satellites. Traditional BFD methods typically employ fixed transmission periods and fixed detection timeouts, making it difficult to simultaneously meet the requirements of low control overhead and high-reliability fault detection. This invention achieves low-complexity prediction of link propagation delay status by parametrically modeling the round-trip delay sequence of inter-satellite links, compressing the continuous delay variation into lightweight model parameters, and recovering the propagation delay prediction model at the onboard node side.

[0007] Meanwhile, the present invention dynamically adjusts the BFD control message transmission period based on the predicted link propagation delay status, and constructs a dynamic timeout determination mechanism based on the timing relationship in the continuous BFD control message transmission process; in an optional embodiment, corresponding detection parameters can also be configured in combination with the link status auxiliary parameters, so that the BFD detection parameters can be adaptively adjusted with the changes in the link status, thereby reducing the control message overhead and improving the accuracy and stability of fault detection in a highly dynamic inter-satellite link environment.

[0008] To achieve the above objectives, the present invention provides an adaptive bidirectional forwarding detection method using the following technical solution:

[0009] S1. Obtain the round-trip delay sequence corresponding to the satellite link, perform parameterized modeling on the round-trip delay sequence to obtain a propagation delay prediction model for describing the link delay variation law, and generate a detection parameter configuration file; the propagation delay prediction model includes a trend term for describing the round-trip delay variation trend and a periodic term for describing the periodic variation law;

[0010] S1.1 Extract time-domain features from the round-trip delay sequence within the historical visible link window to obtain a trend term to characterize the delay change trend and a periodic term to characterize the periodic change pattern.

[0011] S1.2 Construct a propagation delay prediction model based on the trend term and periodic term, and optimize the model parameters by fitting.

[0012] S1.3. Encapsulate the optimized model parameters into a parsable detection parameter configuration file for the onboard node;

[0013] S2. The onboard node obtains the model parameters corresponding to the propagation delay prediction model;

[0014] S2.1. The onboard node parses the detection parameter configuration file and restores the propagation delay prediction model described in S1 based on the model parameters;

[0015] S2.2. Input the propagation delay prediction model based on the current time information to obtain the predicted round-trip delay of the link at the current time.

[0016] S3. Dynamically adjust the transmission period of BFD control messages according to the predicted round-trip delay described in S2;

[0017] The candidate transmission period is calculated based on the predicted round-trip time and adaptive scaling factor at the current moment. The adaptive scaling factor is configured according to the link propagation characteristics and service latency requirements so that the BFD control message transmission period is adapted to the link propagation characteristics and service requirements.

[0018] When the candidate transmission period is less than the preset minimum transmission period, the preset minimum transmission period is used as the actual BFD transmission period; otherwise, the candidate transmission period is used as the actual BFD transmission period.

[0019] S4. Generate a dynamic timeout detection threshold according to the BFD transmission cycle described in S3;

[0020] Based on the timing relationship in the continuous BFD control message transmission process, the timeout judgment base time is determined according to the actual BFD transmission cycle, and then dynamically adjusted by combining the preset detection multiplier and the delay jitter margin to obtain the dynamic timeout detection threshold.

[0021] The preset detection multiplier is a detection multiplier parameter defined by the BFD protocol. Its specific value is pre-configured according to the service type and fault detection requirements, and is used to determine the number of times the peer response message is not received consecutively during the link fault determination process.

[0022] S5. Send BFD control messages according to the BFD sending cycle described in S3, and determine the link status based on the dynamic timeout detection threshold;

[0023] If a response message is received from the peer within the dynamic timeout detection threshold, the detection status is updated; otherwise, the link is determined to be faulty.

[0024] Furthermore, the propagation delay prediction model includes a trend term to describe the trend of round-trip delay changes and a periodic term to describe the periodic change pattern. The trend term is used to characterize the long-term evolution of satellite link propagation delay over time, and the periodic term is used to characterize the periodic change pattern of link propagation delay caused by satellite motion.

[0025] Furthermore, the detection parameter configuration file may also include link state auxiliary parameters for configuring the BFD control message sending period and dynamic timeout detection threshold.

[0026] Compared with existing fixed-parameter BFD methods, the present invention has the following advantages:

[0027] 1. This invention addresses the problem of continuously changing propagation delay in high-dynamic satellite links due to the relative motion of the satellite. It parametrically models the historical RTT (Real-Time Tolerance) sequences of satellite links, converting the continuous delay variation process into a lightweight model parameter representation. The propagation delay prediction model is then reconstructed by the onboard nodes to obtain the predicted RTT for the current moment. Compared to methods that directly store or transmit the complete delay sequence, this reduces the amount of link state information transmitted and the onboard storage requirements. Furthermore, by deploying the complex propagation delay modeling and parameter optimization process outside the onboard nodes, it reduces the onboard computational burden and improves deployment efficiency on resource-constrained satellite platforms.

[0028] 2. This invention dynamically adjusts the BFD control message sending period based on the predicted RTT, and configures an adaptive scaling factor in combination with the link propagation characteristics and service requirements, so that the BFD control message sending period can be adapted to the link propagation characteristics and service requirements, thereby reducing control overhead and taking into account the real-time performance of link fault detection.

[0029] 3. This invention is based on the timing relationship in the continuous BFD control message transmission process, and combines the actual BFD transmission period, detection multiplier and delay jitter margin to generate a dynamic timeout detection threshold. This enables the timeout judgment to adapt to the changes and fluctuations in satellite link propagation delay, reduces the probability of link misjudgment caused by fixed timeout thresholds, and improves the accuracy and stability of link fault detection.

[0030] 4. This invention can also uniformly manage the propagation delay prediction model parameters and link status auxiliary parameters through the detection parameter configuration file, providing a configuration basis for propagation delay prediction, adaptive transmission period adjustment and dynamic timeout detection threshold generation, and improving the environmental adaptability of the link fault detection method in complex spatial environments. Attached Figure Description

[0031] Figure 1 This is an overview diagram of the system architecture of the adaptive bidirectional forwarding detection method of the present invention.

[0032] Figure 2 This is a schematic diagram illustrating the satellite link visibility determination method of the present invention.

[0033] Figure 3 This is a flowchart of the RTT sequence parameterization modeling process of the present invention.

[0034] Figure 4 This is a flowchart of the adaptive BFD execution and fault determination process for the onboard node of this invention.

[0035] Figure 5 This is a schematic diagram illustrating the satellite link RTT evolution and prediction fitting effect of the present invention. Figure 5 In the figure, (a) shows the fitting results of the actual round-trip delay and the predicted round-trip delay of the satellite link, and (b) shows the absolute residual of the predicted round-trip delay relative to the actual round-trip delay and the preset error threshold.

[0036] Figure 6 This is a schematic diagram illustrating the adaptive adjustment relationship between the transmission period and the timeout threshold of the present invention.

[0037] Figure 7 This is a performance comparison chart between the present invention and the traditional static BFD scheme. Figure 7 In the table, (a) shows the comparison results of the average number of false status judgments for different detection strategies, (b) shows the comparison results of the average fault detection latency for different detection strategies, (c) shows the comparison results of the control message overhead for different detection strategies, and (d) shows the comparison results of the fault detection success rate for different detection strategies. Detailed Implementation

[0038] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings. The following embodiments are implemented based on the technical solution of the present invention, providing detailed implementation methods and specific operating procedures; however, the scope of protection of the present invention is not limited to the following embodiments.

[0039] like Figure 1 As shown, this invention proposes an adaptive bidirectional forwarding detection method. First, the round-trip delay sequence of the satellite link is parameterized and modeled. A lightweight prediction model and configuration parameters are generated through model accuracy verification and window segmentation. Then, the onboard node uses the configuration parameters to predict the propagation delay at the current moment, and dynamically adjusts the BFD detection cycle based on the prediction results to achieve adaptive optimization of link detection overhead and state awareness accuracy.

[0040] S1. Obtain the round-trip time delay sequence (RTT) corresponding to the satellite link, perform parameterized modeling on the RTT sequence, and generate a detection parameter configuration file.

[0041] S1.1 Obtain the RTT sequence corresponding to the satellite link:

[0042] S1.1.1 Establish an orbital propagation model based on the satellite's two-line elements (TLE), and convert the orbital elements into satellite state vectors in Earth-Centered Inertial (ECI) coordinates. : in, Represents the three-dimensional position vector of the satellite. This represents the corresponding velocity vector.

[0043] According to time step Discrete sampling of the orbit yields the position sequences of the source satellites. and the position sequence of the target satellite The relative motion state of the satellites is constructed based on the position sequences of the source and target satellites, providing input for subsequent link visibility analysis.

[0044] S1.1.2 Extracting the visible window of the satellite link based on spatial geometric constraints, wherein the spatial geometric constraints include distance constraints, occlusion constraints, and field of view constraints, specifically:

[0045] like Figure 2 As shown, this invention determines whether a satellite link meets communication requirements based on the spatial geometric relationship between satellites. First, the satellite distance between two satellites at any given time is calculated. : in, and These represent the three-dimensional position vectors of the two satellites, respectively.

[0046] Distance constraints are based on maximum communication distance constraints. To determine whether the communication distance requirement is met at the current moment, where This represents the maximum communication distance between two satellites at any given time.

[0047] Occlusion constraint refers to the constraint that is satisfied when At that time, it was assumed that the link was not blocked by Earth, in which This indicates the closest distance from the Earth's center along the line connecting the two stars. Represents the Earth's radius. Indicates the thickness of the atmosphere.

[0048] Field of view constraint refers to when At that time, the field of view limitation is satisfied, where This represents the field of view angle of satellite i and j relative to the line connecting the satellites.

[0049] When the distance constraint, occlusion constraint, and field of view constraint are all satisfied, the current moment is marked as visible, and the time periods that are continuously visible are merged to obtain the link visibility window. ,in, Indicates the start timestamp of the currently visible window. Indicates the end timestamp.

[0050] S1.2 Perform parameterized modeling and compression of the RTT sequence within the link's visible window.

[0051] like Figure 3 and Figure 5 As shown, this invention addresses the characteristic that satellite link propagation delay changes continuously with orbital motion by performing lightweight modeling of the propagation delay sequence within the visible window.

[0052] S1.2.1 Calculate the round-trip propagation delay based on satellite distance. : Where d(t) is the distance to the satellite and c is the speed of light.

[0053] Obtain the discrete time delay sequence within the window : in, Let k be the k-th discrete sampling time within the visible window, where k = 1, 2, ..., n, and n is the total number of sampling points within the visible window.

[0054] S1.2.2 Normalization of the window timeline: in, For normalized time variables, and t represents the link visibility window. The current sampling time within. Trend term is used. Describing long-term changes: in, , and These are the coefficients of the quadratic term, the linear term, and the constant term, respectively.

[0055] Calculate residuals : .

[0056] S1.2.3, Based on the residual Construct a propagation delay prediction model that includes a polynomial trend term and an Nth-order harmonic term: in, For the propagation delay and prediction model, N is the harmonic order, which increases adaptively from 1 to 5. , , The first The amplitude, frequency, and phase parameters of the first harmonic.

[0057] S1.2.4, Judging the Fitting Quality of the Propagation Delay Prediction Model and Handling the Visible Window:

[0058] Calculate the maximum absolute error of the current propagation delay prediction model. : in, Indicates the actual RTT value. This indicates the predicted RTT value from the fitted model.

[0059] And based on the error value e, the following determination is performed:

[0060] like If the model fitting error meets the requirements, stop harmonic order increase and output the current fitted model;

[0061] like If the harmonic order N has reached the set upper limit, then the current window will no longer be divided, and the current fitted model will be output.

[0062] like If the harmonic order N has reached the set upper limit, then the current window fitting error is determined to be unsatisfactory. The visible window is then bisected into two sub-windows at the midpoint, and the sub-windows are added back to the queue. The process returns to S1.2.2 to re-execute the progressive fitting logic.

[0063] S1.3, Inverse Parameter Solution and Lightweight Model Derivation:

[0064] After the progressive fitting converges, the normalized coefficients obtained based on the normalized time axis are uniformly inversely solved into physical second-level coefficients.

[0065] By inversely solving the parameters, a lightweight model fitting coefficient is finally obtained, and the coefficient sequence and model type identifier are encapsulated in JSON (JavaScript Object Notation). In an optional implementation, the detection parameter configuration file may also include link state auxiliary parameters for configuring the BFD control message sending period and dynamic timeout detection threshold, which are sent to the target satellite node as the final configuration parameters.

[0066] S2. The onboard node obtains the model parameters corresponding to the propagation delay prediction model;

[0067] S2.1 The onboard node parses the detection parameter configuration file, reads the model parameters, and restores the propagation delay prediction model.

[0068] S2.2 Calculate and predict round-trip delay based on current running time. :

[0069] Since the model parameters have been pre-calculated, the onboard node only needs to perform model parameter parsing and algebraic calculations to recover the propagation delay prediction model and obtain the predicted round-trip delay corresponding to the current time based on the current time information.

[0070] In one optional implementation, the onboard node parses the link status auxiliary parameters in the detection parameter configuration file and stores them as the configuration basis for subsequent steps S3 and S4. In this embodiment, the link propagation status information includes the Doppler frequency shift parameters of the laser link; the time delay change information includes the fitting parameters of the propagation time delay prediction model and their variation law; the link risk characteristic information includes inter-satellite geometric relationship parameters, model fitting error, and other information.

[0071] S3, Adaptive adjustment of BFD transmission period

[0072] like Figure 4 and Figure 6 As shown, this invention calculates the BFD control message transmission period based on the pre-round-trip delay and uses a preset minimum transmission period as a lower limit constraint to obtain the actual BFD transmission period. : in, Minimum transmission period; This is the adaptive scaling factor.

[0073] When the prediction latency is small, i.e. At that time, with the minimum transmission period This serves as the actual transmission cycle, preventing excessively high transmission frequencies from wasting bandwidth.

[0074] In this embodiment, the onboard node configures an adaptive scaling factor by combining the link propagation status information and delay change information parsed in S2.1. For example, when the Doppler frequency shift rate exceeds a preset threshold, indicating severe inter-satellite relative motion, the adaptive scaling factor can be appropriately reduced to increase the BFD control message transmission frequency.

[0075] S4. Generate dynamic timeout detection threshold

[0076] This invention uses the product of the actual BFD transmission period and a preset detection multiplier as the base time for timeout determination, and adds a delay jitter margin to the base time for timeout determination to obtain a dynamic timeout detection threshold. : Wherein, DM is the detection multiplier; This is a margin for time delay jitter.

[0077] In this embodiment, the onboard node generates a dynamic timeout detection threshold by combining the link risk feature information parsed in S2.1. For example, when the model fitting error increases, the delay jitter margin can be appropriately increased to adjust the dynamic timeout detection threshold towards a conservative state, thereby reducing the probability of misjudgment caused by abnormal fluctuations in propagation delay in complex space environments.

[0078] S5, Link Fault Detection

[0079] The onboard node sends BFD control messages according to the actual BFD transmission period determined in step S3, and judges the link status based on the dynamic timeout detection threshold generated in step S4.

[0080] Get the current detection time And the time when the last valid BFD control message was received from the peer. Calculate the message reception time interval : in, This indicates the time interval between the current detection time and the time of the most recent valid message reception.

[0081] The message receiving time interval With dynamic timeout detection threshold Comparison:

[0082] When satisfied During this period, maintain the normal link status; upon receiving a new valid BFD control message, Update to the reception time of the new valid BFD control message.

[0083] When satisfied At that time, it is determined that the link has failed.

[0084] Method gain effect:

[0085] To verify the effectiveness and robustness of the adaptive bidirectional forwarding detection method described in this invention, this embodiment constructs a highly dynamic hybrid constellation space network application scenario and conducts a comprehensive performance evaluation.

[0086] 1. Scene topology and monitoring object settings

[0087] 1.1 This embodiment takes the BDS-3 heterogeneous hybrid constellation system as the application background, selects the inclined geosynchronous orbit IGSO-1 satellite as the source monitoring node, and selects the medium Earth orbit MEO satellite as the target tracking node to form an IGSO-MEO cross-orbit high dynamic satellite monitoring link;

[0088] 1.2 The system uses the SDP4 orbital perturbation model to deduce the three-dimensional position sequence of the satellite in the inertial coordinate system. The time step for the deduction is set to 5 minutes, and the total duration is set to 24 hours.

[0089] 1.3. Taking into account the triple geometric constraints of maximum communication distance, earth occlusion and field of view, the visible window of the physically continuous communication link is automatically extracted as the effective spatiotemporal domain for protocol execution.

[0090] 2. Modeling of random disturbances in spatial links

[0091] Based on deterministic propagation delay, random disturbance factors are introduced according to the actual operating characteristics of space links to simulate non-ideal states such as link queuing jitter and sudden packet loss. The robustness of the adaptive bidirectional forwarding detection method in complex space environments is verified by setting different link quality states. The simulation superimposed queuing jitter and Markov packet loss models, and the parameter settings are shown in Table 1.

[0092] Table 1. Core Simulation Parameter Configuration Table:

[0093] 3. Protocol parameters and control group (baseline) settings

[0094] 3.1 The core operating parameter configuration of the adaptive BFD protocol of this invention is shown in Table 1.

[0095] 3.2 Setting a high-frequency baseline (Fixed 100ms): A fixed detection period of 100ms represents an aggressive detection strategy that seeks sensitivity in extreme conditions;

[0096] 3.3 Setting a standard baseline (Fixed 150ms): A fixed detection period of 150ms represents a standard engineering compromise strategy that balances overhead and response speed;

[0097] 3.4 Setting an equivalent overhead baseline (Fixed Mean-Tx): The average packet transmission frequency configured in this baseline is consistent with the adaptive strategy of this invention, eliminating performance interference caused by differences in resource consumption. Under completely equal bandwidth constraints, the architectural superiority of the adaptive mechanism in the time-domain coordination and state awareness layers is verified, directly proving that the performance improvement of this invention comes from the mechanism itself, rather than simply reducing the number of BFD control messages.

[0098] 4. Definition of verification indicators

[0099] 4.1 Control Overhead / Bandwidth Usage: The total number of BFD control messages sent by the protocol per hour represents the degree of consumption of satellite bandwidth resources;

[0100] 4.2 Number of false status judgments: Count the number of false interruption judgments caused by non-real link failures, that is, false alarm events triggered by physical disturbances such as random channel jitter and sudden deep fading, which are used to characterize the anti-interference capability and operational reliability of the protocol.

[0101] 4.3 Fault Detection Delay: The time interval from the moment a real link fault occurs to the time it takes for the protocol to complete the fault determination, characterizing the protocol's response speed to real link anomalies.

[0102] 5. Conclusion on Overall Performance and Beneficial Effects

[0103] 5.1. An independent random seed was introduced to perform 100 rounds of discrete event Monte Carlo parallel simulation. The comprehensive performance evaluation results of the adaptive strategy of this invention and each static baseline are shown in Table 2.

[0104] 5.2 Compared to the traditional high-frequency baseline (Fixed 100ms), the fault detection latency of the adaptive strategy of this invention increases from 629ms to 851ms, an increase of 222ms; at the same time, the control message overhead per hour decreases from 36,000 packets to 20,350 packets, and the bandwidth overhead decreases by 43.5%.

[0105] 5.3 The average number of false alarms decreased from 2.28 times / window to 0.12 times / window, a reduction of 94.7%, indicating that the present invention can reduce the number of false alarms while reducing control message overhead;

[0106] Table 2. Comparison of overall system performance under different detection strategies:

[0107] 5.4 Under the condition that the control message overhead is close to that of the equivalent overhead baseline (Fixed Mean-Tx), the average number of misjudgments of the adaptive strategy of the present invention is 0.12 times / window, which is 27.9% of the baseline of 0.43 times / window, indicating that the present invention can reduce state misjudgments under similar control message overhead.

[0108] 5.5 The adaptive method described in this invention can adjust the BFD control message sending period and dynamic timeout detection threshold according to the propagation delay prediction results, reduce state misjudgment in high dynamic satellite link environments, and provide link state detection results for upper-layer routing protocols.

Claims

1. An adaptive bidirectional forwarding detection method, characterized in that, The implementation process of this method is as follows: S1. Obtain the round-trip delay sequence corresponding to the satellite link, perform parameterized modeling on the round-trip delay sequence to obtain a propagation delay prediction model that describes the link delay variation law, and generate a detection parameter configuration file. The propagation delay prediction model includes: a trend term describing the trend of round-trip delay changes, and a periodic term describing the periodic change pattern. S1.1 Extract time-domain features from the round-trip delay sequence within the historical visible link window to obtain a trend term to characterize the delay change trend and a periodic term to characterize the periodic change pattern. S1.2 Construct a propagation delay prediction model based on the trend term and periodic term, and optimize the model parameters by fitting. S1.

3. Encapsulate the optimized model parameters into a parsable detection parameter configuration file for the onboard node; S2. The onboard node obtains the model parameters corresponding to the propagation delay prediction model; S2.

1. The onboard node parses the detection parameter configuration file and restores the propagation delay prediction model described in S1 based on the model parameters; S2.

2. Input the propagation delay prediction model based on the current time information to obtain the predicted round-trip delay of the link at the current time. S3. The onboard node adjusts the transmission period of BFD control messages based on the predicted round-trip time delay. S4. The onboard node generates a dynamic timeout detection threshold based on the BFD transmission cycle. S5. The onboard node sends BFD control messages according to the BFD transmission cycle and determines the link status based on the dynamic timeout detection threshold. If a response message is received from the peer within the dynamic timeout detection threshold, the detection status is updated; otherwise, the link is determined to be faulty. Step S3, which refers to the onboard node adjusting the transmission period of BFD control messages based on the predicted round-trip time, means: The candidate transmission period is calculated based on the predicted round-trip time and adaptive scaling factor at the current moment; the adaptive scaling factor is configured according to the link propagation characteristics and service latency requirements so that the BFD control message transmission period is adapted to the link propagation characteristics and service requirements. When the candidate transmission period is less than the preset minimum transmission period, the preset minimum transmission period is used as the actual BFD transmission period; otherwise, the candidate transmission period is used as the actual BFD transmission period. Step S4, which involves generating a dynamic timeout detection threshold based on the BFD transmission period, includes: The timeout determination base time is determined based on the timing relationship during continuous BFD control message transmission. The timeout determination base time is adjusted by combining the preset detection multiplier and the delay jitter margin to obtain the dynamic timeout detection threshold; the preset detection multiplier is the detection multiplier parameter defined by the BFD protocol, and its specific value is pre-configured according to the service type and fault detection requirements, which is used to determine the number of times the peer response message is not received continuously during the link fault determination process.

2. The adaptive bidirectional forwarding detection method according to claim 1, characterized in that, The trend term and periodic term in the propagation delay prediction model described in step S1 are obtained by feature extraction and parameter fitting of the round-trip delay sequence within the historical visible link window, and the corresponding model parameters are encapsulated in the detection parameter configuration file for storage and / or transmission. The trend term is used to characterize the long-term evolution law of satellite link propagation delay over time, and the periodic term is used to characterize the periodic change law of link propagation delay caused by satellite motion.

Images