An autonomous time synchronization method for low-orbit constellations based on Ka inter-satellite links

Abstract

The present invention provides a low-orbit constellation autonomous time synchronization method based on Ka inter-satellite links, which relates to the field of satellite navigation and time and frequency transfer. The method comprises: obtaining a ground station Ka-band standard timestamp signal and performing satellite clock training to obtain the constellation's initial anchor node; using time division multiplexing for bidirectional ranging and correcting motion delay to obtain corrected inter-satellite clock error data; fusing clock error data and suppressing long-term drift to generate a unified constellation time scale; performing dynamic topology optimization based on the unified constellation time scale to obtain an optimized network topology; maintaining a time base using Kalman filtering and polynomial prediction, and predicting frequency drift changes in combination with an LSTM neural network to obtain an autonomous timekeeping result; and verifying abnormal clock errors through majority voting by neighboring nodes. If the vote fails, the fault is isolated and the recovery value is reversely calculated. The present invention is intended to address the drawback of existing fiber time and frequency transfer technologies, which are limited by ground fixed facilities and cannot be applied to inter-satellite scenarios.

Description

Technical Field

[0001] The present invention relates to the field of satellite navigation and time and frequency transfer, and in particular to a low-orbit constellation autonomous time synchronization method based on Ka inter-satellite links. Background Art

[0002] With the rapid development of low-orbit giant constellations, high-precision time synchronization has become a core requirement for achieving integrated communication and navigation.

[0003] In existing technologies, two-way time transmission between satellites and the ground relies on ground reference stations, which has coverage blind spots and insufficient real-time performance. Inter-satellite time synchronization mostly uses UHF or laser links. The former has limited accuracy (1-10ns), while the latter is significantly affected by atmospheric interference and has low technical maturity.

[0004] Traditional inter-satellite time synchronization technology has the following bottlenecks: dynamic delay error: the high-speed movement of low-orbit satellites causes asymmetric signal propagation paths. The traditional two-way ranging model ignores the influence of acceleration and has insufficient correction accuracy; link calibration dependence: the delay of inter-satellite equipment requires frequent ground calibration, which makes it difficult to meet the needs of autonomous operation of the constellation; time scale drift: in the absence of an external benchmark, the internal clock error of the constellation is prone to accumulation and divergence, and long-term stability is poor.

[0005] Existing solutions, such as the GNSS carrier phase method, have high accuracy (sub-nanosecond level), but rely on external navigation satellite signals and cannot meet the independent operation requirements of low-orbit constellations; optical fiber time and frequency transmission is limited by ground fixed facilities and cannot be applied to inter-satellite scenarios.

[0006] Therefore, there is an urgent need for an autonomous time synchronization technology based on inter-satellite links that combines high precision and strong robustness. Summary of the Invention

[0007] The present invention provides a low-orbit constellation autonomous time synchronization method based on Ka inter-satellite links, which is used to solve the defect that fiber time and frequency transmission in the existing technology is limited by ground fixed facilities and cannot be applied to inter-satellite scenarios.

[0008] In one aspect, the present invention provides a method for autonomous time synchronization of a low earth orbit constellation based on a Ka inter-satellite link, comprising:

[0009] Obtain the Ka-band standard timestamp signal from the ground station, and use Kalman filtering and Doppler compensation to train the satellite clock to obtain the initial anchor node of the constellation.

[0010] Based on the constellation initial anchor node, time division multiplexing is used to bidirectionally measure the distance between the constellation initial anchor node and the adjacent satellites, and a single-point pseudo-range model is used to correct the motion delay to obtain the corrected inter-satellite clock error data.

[0011] Based on the corrected inter-satellite clock error data, clock error data fusion is performed, and a clock error prediction model is used to suppress long-term drift and generate a unified time scale for the constellation.

[0012] Based on the unified time scale of the constellation, dynamic topology optimization is performed to obtain the optimized network topology;

[0013] Based on the optimized network topology, Kalman filtering and polynomial prediction are used to maintain the time base, and the LSTM neural network is combined to predict frequency drift changes to obtain autonomous timekeeping results.

[0014] Based on the autonomous timekeeping results, the abnormal clock difference is verified by majority voting of neighboring nodes. If the vote fails, the fault is isolated and the recovery value is calculated reversely.

[0015] Furthermore, the ground station Ka-band standard timestamp signal is obtained, and the satellite clock is trained by combining Kalman filtering and Doppler compensation to obtain the constellation initial anchor node, including:

[0016] The ground station transmits a Ka-band standard timestamp signal, and the satellite receiver captures and demodulates the time information;

[0017] Based on the demodulated time information, the satellite atomic clock is used to generate a second pulse signal, and the satellite-to-ground clock difference is measured by a time interval counter.

[0018] Based on the satellite-to-ground clock error, a Kalman filter is used to smooth the clock error sequence, synchronously estimate the frequency error and frequency drift parameters, and generate the frequency adjustment command of the voltage-controlled crystal oscillator;

[0019] Based on the frequency adjustment instruction of the voltage-controlled crystal oscillator, the adjustment amount is loaded to the crystal oscillator control end through the digital-to-analog conversion module, and the satellite clock phase is gradually locked to the ground reference to obtain the initial anchor node of the constellation.

[0020] Furthermore, based on the constellation initial anchor node, time division multiplexing is used to bidirectionally measure the distance between the constellation initial anchor node and the adjacent satellites, and a single-point pseudo-range model is used to correct the motion delay to obtain the corrected inter-satellite clock error data, including:

[0021] Based on the initial anchor node of the constellation, a time-division multiplexing ranging link is established between the anchor node satellite and the adjacent satellite in the Ka band, and ranging signals are alternately transmitted and received according to the preset time slots.

[0022] Based on the ranging signal, the orbit and clock errors are decoupled by timestamp alignment to obtain two-way pseudorange observations.

[0023] Based on the two-way pseudorange observations, the single-point pseudorange reduction model is used to perform dynamic error correction to obtain the corrected pseudorange;

[0024] Based on the corrected pseudorange and combined with the relativistic effect correction term, the inter-satellite relative clock error is decoupled and obtained. The time delay accumulation error of the multi-hop path is modeled and distributed to compensate for it, and the corrected inter-satellite clock error data is obtained.

[0025] Furthermore, based on the corrected inter-satellite clock error data, clock error data fusion is performed, and a clock error prediction model is used to suppress long-term drift and generate a unified constellation time scale, including:

[0026] Uploading the corrected inter-satellite clock error data and inter-satellite measurement results to the distributed processing unit and pre-processing them to obtain pre-processed data;

[0027] Based on the pre-processed data, weights are dynamically assigned according to the Allan variance of the onboard atomic clock to obtain weighted fusion data;

[0028] Based on the weighted fused data, the first-level Kalman filter is used to eliminate single-clock noise, and the second-level Kalman filter is used to fuse the inter-satellite clock difference observations to generate a comprehensive time scale to obtain the optimized time scale.

[0029] Based on the optimized time scale, the clock error prediction model is used to suppress long-term drift and generate a unified time scale for the constellation.

[0030] Furthermore, based on the unified time scale of the constellation, dynamic topology optimization is performed to obtain the optimized network topology, including:

[0031] Based on the unified time scale of the constellation, link quality, network-wide clock error gradient, and orbit parameters are collected in real time. Link quality is evaluated using a scoring model, and abnormal nodes with deviations are marked and isolated.

[0032] After isolating the abnormal node, the clock error is reconstructed using the redundant observations of neighboring nodes. After verifying the residual, the node function is restored to obtain the optimized network topology.

[0033] Furthermore, based on the optimized network topology, Kalman filtering and polynomial prediction are used to maintain the time base, and combined with the LSTM neural network to predict frequency drift changes, autonomous timekeeping results are obtained, including:

[0034] Based on the optimized network topology, the inter-satellite clock error observations, atomic clock historical frequency offset data, and environmental parameters are aligned, outliers are removed, and time series features such as the clock error change rate are extracted to obtain pre-processed topological data.

[0035] Based on the pre-processed topological data, the constellation maintains temporal stability through two-level correction to obtain the corrected topological data;

[0036] Based on the corrected topological data, an LSTM neural network is deployed, and combined with the historical clock error series to predict the frequency drift changes in the next 12 hours, and the autonomous timekeeping results are obtained.

[0037] Furthermore, based on the autonomous timekeeping results, the abnormal clock error is verified through a majority vote of neighboring nodes. If the vote fails, the fault is isolated and the recovery value is calculated backwards, including:

[0038] Based on the autonomous timekeeping results, a majority voting mechanism is used to verify the clock error data of the target node to obtain the verification result;

[0039] Based on the verification results, an inter-satellite alarm signal is triggered, and the abnormal node is isolated from the synchronization network to obtain the clock error after isolation;

[0040] Based on the clock error after isolation, the theoretical clock error value of the faulty node is reversely calculated using historical clock error data and orbital dynamics models through the Kalman smoothing algorithm, compensating for the time deviation during the missing period and obtaining the reverse calculation recovery result.

[0041] Based on the reverse calculation recovery result, the estimated value is re-entered into the majority voting mechanism for secondary verification to obtain the recovery verification result;

[0042] Based on the recovery verification results, the inter-satellite link topology is updated and the time transfer path is optimized.

[0043] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for autonomous time synchronization of a low-orbit constellation based on a Ka inter-satellite link as described above is implemented.

[0044] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links.

[0045] On the other hand, the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned low-orbit constellation autonomous time synchronization methods based on Ka inter-satellite links.

[0046] The proposed method for autonomous time synchronization of low-orbit (LEO) constellations based on Ka inter-satellite links (ISLs) decouples orbit and clock information through time-division bidirectional, one-way ranging. This method, combined with a motion delay correction algorithm, suppresses dynamic path asymmetry errors, reducing inter-satellite clock error calculation errors. A closed-loop self-calibration mechanism estimates inter-satellite equipment delays in real time, reducing reliance on ground-based calibration. The proposed single-point pseudorange reduction correction algorithm requires only single-satellite velocity information to compensate for motion delays, reducing computational complexity compared to traditional two-way ranging models and adapting to a variety of LEO orbit configurations. A two-stage Kalman filter fusion and LSTM neural network prediction achieve dynamic timescale optimization, improving overall constellation clock stability and achieving long-term timekeeping capabilities superior to existing UHF or laser link solutions. A majority voting mechanism is used to detect and isolate faulty nodes, and a reverse calculation recovery algorithm is used to reconstruct abnormal clock errors. This method achieves low network-wide time base consistency errors and significantly enhances the constellation's anti-interference capability. The method is compatible with existing Ka-band ISLs, requiring no additional hardware and can be deployed through software upgrades. It is suitable for large-scale systems such as LEO communications, navigation augmentation, and remote sensing constellations, reducing operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 1 is a flow chart of a method for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links provided by an embodiment of the present invention;

[0049] Figure 2 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention;

[0050] Figure 3 This is a low-orbit constellation satellite-to-ground / inter-satellite time-frequency transfer model provided by an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of time-division inter-satellite link measurement provided by an embodiment of the present invention;

[0052] Figure 5 is a schematic diagram of motion delay provided by an embodiment of the present invention;

[0053] Figure 6 This is a schematic diagram of conventional bidirectional time-synchronized motion provided by an embodiment of the present invention;

[0054] Figure 7 This is a schematic diagram of single-point pseudorange two-way time synchronization provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0056] Figure 1 This is one of the flow charts of the low-orbit constellation autonomous time synchronization method based on the Ka inter-satellite link provided by an embodiment of the present invention.

[0057] like Figure 1 As shown, the embodiment of the present invention provides a method for autonomous time synchronization of a low-orbit constellation based on a Ka inter-satellite link, and the method mainly includes the following steps:

[0058] 11. Obtain the Ka-band standard timestamp signal from the ground station and perform satellite clock training using a combination of Kalman filtering and Doppler compensation to obtain the constellation's initial anchor node.

[0059] 12. Based on the constellation initial anchor node, use time division multiplexing to bidirectionally measure the distance between the constellation initial anchor node and the adjacent satellites, and use the single-point pseudo-range model to correct the motion delay to obtain the corrected inter-satellite clock error data;

[0060] 13. Based on the corrected inter-satellite clock error data, perform clock error data fusion and use the clock error prediction model to suppress long-term drift and generate a unified time scale for the constellation;

[0061] 14. Based on the unified time scale of the constellation, dynamic topology optimization is performed to obtain the optimized network topology;

[0062] 15. Based on the optimized network topology, Kalman filtering and polynomial prediction are used to maintain the time base, and the LSTM neural network is combined to predict frequency drift changes to obtain autonomous timekeeping results;

[0063] 16. Based on the autonomous timekeeping results, the abnormal clock error is verified through majority voting of neighboring nodes. If the vote fails, the fault is isolated and the recovery value is calculated in reverse.

[0064] In an embodiment of the present invention, by receiving the Ka-band standard timestamp signal from the ground station, Kalman filtering is used to suppress noise, and Doppler compensation is combined to correct the frequency deviation caused by high-speed motion, initial synchronization of the satellite clock and the ground reference is achieved, and the initial time reference anchor node of the constellation is established with an error of better than 0.5 ns. This provides a high-precision reference starting point for subsequent inter-satellite synchronization and solves the blind spot problem of traditional satellite-ground synchronization coverage. A time-division multiplexing two-way ranging mechanism is adopted to decouple the influence of orbital motion and clock error by staggered signal transmission and reception. In combination with a single-point pseudo-range model, only the velocity information of a single satellite is used to correct the path asymmetry error. The inter-satellite clock error measurement accuracy reaches 5 cm (corresponding to a delay of 0.17 ns), and the motion delay correction error is reduced, breaking through the traditional two-way ranging model's strong reliance on the velocity information of both parties. Based on the corrected inter-satellite clock error data, a two-stage Kalman filter architecture is adopted. The first stage eliminates single-clock noise, and the second stage integrates multi-satellite observations to generate a comprehensive time scale. Combined with the clock error prediction model to compensate for long-term drift, the overall constellation clock error stability is improved by 2 times, and the frequency drift rate is better than 2×10 -11 / day, achieving 30 days of baseline maintenance without ground intervention, with an absolute clock error drift of less than 25ns. Based on the unified time scale of the constellation, the system analyzes parameters such as inter-satellite link quality and satellite motion trajectory, dynamically adjusts the network topology, optimizes the time synchronization path, improves link utilization, reduces key path delay fluctuations, and enhances the robustness of time synchronization of the constellation in complex orbital configurations. Utilizing the optimized network topology and combining Kalman filtering to maintain the time benchmark, the system uses an LSTM neural network to predict the satellite clock frequency drift trend and achieve early compensation, with a short-term stability (Allan variance) of 1×10 -13 / τ, the long-term frequency prediction error is less than 5×10 -12 / day, significantly reducing dependence on external frequency references; time synchronization results are verified through a majority voting mechanism of neighboring nodes. If an abnormal clock error is detected, the faulty node is automatically isolated, and the time reference is restored based on reverse calculation of historical data. The fault detection response time is less than 1 second, and the recovery value error is better than 2ns. The system availability is improved to 99.99%, meeting the high-reliability operation requirements of giant constellations; through a layered architecture design, full-process closed-loop control is achieved from ground reference injection, inter-satellite high-precision measurement, dynamic error correction to autonomous reference maintenance. Simulation verification shows that in a typical low-orbit constellation scenario, the time synchronization accuracy is better than 10ns, the equipment delay self-calibration accuracy is 0.3ns, and it is compatible with a variety of orbit configurations (such as Walker, polar orbit, etc.). The link utilization rate exceeds 90%, which is significantly better than the traditional UHF / laser link solution, providing a key time and frequency support technology for the integration of low-orbit communication and navigation; reference Figure 3, showing the overall architecture of time and frequency transmission between low-orbit constellations and ground stations and satellites, including satellite-to-ground links (two-way ranging calibration module between ground stations and satellites), inter-satellite links (Ka-band two-way ranging unit between satellites) and time reference maintenance modules (multi-source data fusion and feedback control); the figure clearly marks the signal transmission path, Kalman filter training process and inter-satellite autonomous synchronization network topology, reflecting the dynamic transmission process of high-precision time and frequency reference from the ground to the inter-satellite and then to the entire constellation.

[0065] like Figure 1 As shown in FIG11 , the ground station Ka-band standard timestamp signal is obtained, and the satellite clock is trained by combining Kalman filtering and Doppler compensation to obtain the constellation initial anchor node, including:

[0066] 111. The ground station transmits a Ka-band standard timestamp signal, and the satellite receiver captures and demodulates the time information;

[0067] 112. Based on the demodulated time information, the satellite atomic clock is used to generate a second pulse signal, and the satellite-to-ground clock difference is measured by a time interval counter;

[0068] 113. Based on the satellite-to-ground clock error, a Kalman filter is used to smooth the clock error sequence, synchronously estimate the frequency error and frequency drift parameters, and generate the frequency adjustment instruction of the voltage-controlled crystal oscillator;

[0069] 114. Based on the frequency adjustment instruction of the voltage-controlled crystal oscillator, the adjustment amount is loaded to the crystal oscillator control end through the digital-to-analog conversion module, and the satellite clock phase is gradually locked to the ground reference to obtain the initial anchor node of the constellation.

[0070] In an embodiment of the present invention, a high-precision satellite-to-ground time reference transmission link is established. The ground station transmits a ranging signal carrying a UTC (Coordinated Universal Time) standard timestamp via the Ka band. The satellite receiver extracts time information using a high-sensitivity coherent demodulation technique, overcoming the Doppler frequency shift caused by the high-speed motion of low-orbit satellites. This achieves initial alignment of satellite-to-ground time information, with a demodulation error of less than 0.1 ns. This provides a reference for subsequent clock error measurements, solving the coverage blind spot problem of traditional satellite-to-ground time synchronization. Figure 4, describes the timing planning and signal interaction flow of two-way ranging in a time-division system between satellites. Satellite A and satellite B transmit ranging signals alternately in preset time slots, and decouple the orbit and clock errors by staggered transmission and reception. The figure marks the signal propagation path, the interpolation algorithm for returning pseudo-range observations to the same moment, and the device delay calibration module, which intuitively demonstrates the synchronization mechanism and error suppression principle of two-way ranging data; quantifies the time deviation between the satellite clock and the ground reference, and the satellite atomic clock (such as a rubidium clock or a hydrogen clock) generates a 1pps (pulse per second) signal, which is compared with the demodulated ground timestamp. The time interval counter (minutes) The system measures the time difference between the two edges with a resolution of 10ps, and the satellite-to-ground clock difference measurement accuracy reaches 0.01ns. It clarifies the initial deviation of the satellite clock relative to the ground reference, provides closed-loop feedback for taming control, and breaks through the ambiguity limitation of traditional one-way ranging. It suppresses measurement noise and predicts the dynamic characteristics of the clock difference. The Kalman filter fuses current and historical clock difference data, establishes a state equation to estimate the clock difference (Δt), frequency difference (Δf), and frequency drift (Δd / dt) parameters, and generates a continuous adjustment curve for the voltage-controlled crystal oscillator (VCXO). The variance of the clock difference estimation is reduced by 80%, and the frequency difference prediction error is less than 2×10 -13 / s, the output control command bandwidth covers 1mHz to 1kHz, which can adapt to the clock error changes in the dynamic environment of low-orbit satellites; realize the physical synchronization of satellite clock and ground reference, the digital-to-analog converter (DAC) converts the digital control quantity into analog voltage (resolution up to 0.1mV), drives the output frequency of the voltage-controlled crystal oscillator, and gradually reduces the satellite-ground clock error through the phase-locked loop (PLL). The phase locking time is less than 100 seconds and the steady-state clock error is less than 0.5ns. It generates the first high-precision time reference anchor node of the constellation, providing a reference starting point for subsequent inter-satellite synchronization, solving the convergence speed of traditional taming technology. The system solves the problem of slow speed; through the closed-loop control of "signal reception-clock error measurement-filter estimation-physical taming", sub-nanosecond synchronization of satellite clocks and ground references is achieved, and the initial reference frame for autonomous time synchronization of the constellation is constructed. Through the adaptive frequency offset compensation algorithm, the signal loss problem caused by the high-speed movement of low-orbit satellites relative to the ground station is overcome; the Kalman filter synchronously processes clock error, frequency error, and frequency drift, and its adaptability is increased by 3 times compared with traditional single-parameter control; the time-division two-way one-way ranging (TD-TWR) mechanism is designed to achieve orbit and clock error decoupling by staggered signal reception and transmission.

[0071] Establish geometric distance and relative clock difference solution equation:

[0072] ;

[0073] in, represents the relative clock error, that is, the time deviation between the two spacecraft, is the pseudorange observation value converted to the same time, is the device delay, The speed of light.

[0074] The closed-loop control bandwidth of the voltage-controlled crystal oscillator reaches 100Hz, and its ability to suppress short-term frequency fluctuations is better than traditional solutions. It provides a highly reliable initial time reference for the entire low-orbit constellation, extending the constellation's autonomous operation time to 30 days (compared to only 7 days with traditional solutions), and the absolute clock error drift is controlled within 25ns, significantly improving the giant constellation's time reference maintenance capability.

[0075] like Figure 1 As shown in 12, based on the constellation initial anchor node, time division multiplexing is used to bidirectionally measure the distance between the constellation initial anchor node and the adjacent satellites, and the single-point pseudo-range model is used to correct the motion delay to obtain the corrected inter-satellite clock error data, including:

[0076] 121. Based on the initial anchor node of the constellation, establish a Ka-band time-division multiplexing ranging link between the anchor node satellite and the adjacent satellite, and alternately transmit and receive ranging signals according to the preset time slots;

[0077] 122. Based on the ranging signal, orbit and clock error are decoupled by timestamp alignment to obtain two-way pseudorange observations;

[0078] 123. Based on the two-way pseudorange observations, a single-point pseudorange reduction model is used to perform dynamic error correction to obtain the corrected pseudorange;

[0079] 124. Based on the corrected pseudorange and combined with the relativistic effect correction term, the inter-satellite relative clock error is decoupled and the delay accumulation error of the multi-hop path is modeled and allocated for compensation to obtain the corrected inter-satellite clock error data.

[0080] In an embodiment of the present invention, a Ka-band time-division multiplexing ranging link is established between the anchor node satellite and the adjacent satellite, and ranging signals are alternately transmitted and received according to preset time slots to construct an efficient and anti-interference inter-satellite ranging channel. By using Ka-band (26.5-40GHz) time-division multiplexing (TDM) technology, non-overlapping time slots are allocated between the anchor node satellite and the adjacent satellite, and ranging signals are transmitted alternately to avoid signal conflicts caused by simultaneous transmissions of multiple satellites. The link utilization rate is improved, the probability of ranging conflicts is reduced to below 0.1%, the ranging signal bandwidth reaches 500MHz, and the distance resolution is excellent. The system is based on the 0.6m time-slot algorithm, which provides high-precision observation data for subsequent clock error calculation. The dynamic time slot allocation algorithm adapts to the changes in constellation topology, and the switching delay is less than 10ms, ensuring the link stability under the high-speed movement of low-orbit satellites. Based on the ranging signal, the orbit and clock error are decoupled through timestamp alignment to obtain two-way pseudo-range observation values, which are used to break through the coupling limitations of the traditional two-way ranging model on orbit dynamics. Through high-precision timestamp alignment technology, the satellite position vectors of the anchor node and the neighboring satellite at the time of transmission / reception are calculated to the same reference time, separating the influence of orbit motion and clock error on pseudo-range. The orbit-clock decoupling accuracy reaches 0.1ns, which is an improvement over traditional methods. The noise level of the two-way pseudorange observation value is less than 5cm (corresponding to a delay of 0.17ns), which meets the sub-nanosecond time synchronization requirements, supports dynamic scenarios with inter-satellite relative speeds of up to 7.8km / s, and adapts to the typical orbital characteristics of low-orbit constellations. Based on the two-way pseudorange observation value, a single-point pseudorange reduction model is used for dynamic error correction to obtain the corrected pseudorange, which is used to innovatively solve the path asymmetry error caused by the motion of low-orbit satellites. Traditional two-way ranging requires relying on the velocity information of both parties, while this model only requires The single-satellite velocity vector is projected onto the signal propagation direction, and the path delay difference is compensated through geometric relationships. The motion delay correction error is reduced from 2ns in the traditional method to 0.8ns, the correction efficiency is improved, and the standard deviation of the pseudorange correction residual is less than 1cm. The ranging deviation in dynamic environments is significantly suppressed, the computational complexity is reduced, and it is suitable for scenarios with limited onboard processor resources. A dynamic correction method based on single-point pseudorange reduction is proposed. Only single-satellite velocity information is needed to compensate for the path asymmetry error. The satellite velocity vector is projected onto the signal propagation direction through ground-fixed coordinate conversion. The correction formula is:

[0081] ;

[0082] in, is the corrected pseudorange, is the satellite speed, is the relative position vector, is the signal propagation time, is the speed of light;

[0083] Based on the corrected pseudoranges and combined with relativistic corrections, the inter-satellite relative clock error is decoupled and derived. The accumulated delay error of multi-hop paths is modeled and compensated for, resulting in corrected inter-satellite clock error data for high-precision relative clock error resolution and multi-hop network error suppression. Frequency shift corrections (such as gravitational and Doppler shifts) based on the general relativity framework are combined with graph-theoretic optimization algorithms to model and compensate for the accumulated delay error of multi-hop paths. After relativistic corrections, the clock error resolution accuracy reaches 0.05 ns, meeting the frequency stability requirements of atomic clocks. Multi-hop delay error allocation and compensation improves network-level clock error consistency by a factor of three, with path delay errors within five hops less than 0.3 ns. The inter-satellite clock error data update rate reaches 1 Hz, adapting to rapid topological changes in low-Earth orbit constellations. This step enables inter-satellite clock error measurement accuracy to the sub-nanosecond level (0.1 ns) and suppresses multi-hop network delay errors to within 0.3 ns, a tenfold improvement in accuracy compared to traditional UHF link solutions. This provides key technical support for maintaining an autonomous time base for low-Earth orbit constellations.

[0084] refer to Figure 5 , showing the asymmetric effect of the signal propagation path caused by the high-speed movement of the satellite, and comparing the geometric relationship differences between the traditional correction method and the single-point pseudorange reduction method, Figure 5 The left side shows the limitation of the traditional model which relies on the speed information of both parties. Figure 5 On the right is the dynamic correction method based on single-star velocity vector projection proposed by the present invention, marking the application scenarios of Earth-fixed coordinate transformation, velocity component decomposition and correction formula, revealing the suppression mechanism of motion delay error.

[0085] refer to Figure 6 This figure illustrates the geometric relationship of a traditional two-way ranging model, including the signal transmission and reception paths (two-way propagation delay difference) and trajectory offset between users S1 and S2. The figure also shows the position deviation at the transmission / reception time, the velocity-dependent correction module, and the uncompensated acceleration cumulative error. This reveals the accuracy bottleneck of traditional methods caused by their reliance on the dynamic parameters of both parties.

[0086] refer to Figure 7 This paper describes the single-point pseudorange calculation and correction mechanism, including the user S2 synchronous signal transmission and reception timing, S1 velocity vector projection, and the ground-fixed path correction module. The figure compares the delay errors before and after correction. Combined with simulation data of 5cm ranging noise, it highlights the engineering practicality of only requiring single-party dynamic parameters.

[0087] like Figure 1 As shown in 13, based on the corrected inter-satellite clock error data, clock error data fusion is performed, and a clock error prediction model is used to suppress long-term drift and generate a unified constellation time scale, including:

[0088] 131. Upload the corrected inter-satellite clock error data and inter-satellite measurement results to the distributed processing unit and perform pre-processing to obtain pre-processed data;

[0089] 132. Based on the pre-processed data, weights are dynamically assigned according to the Allan variance of the onboard atomic clock to obtain weighted fusion data;

[0090] 133. Based on the weighted fused data, the first-stage Kalman filter is used to eliminate the single clock noise, and the second-stage Kalman filter is used to fuse the inter-satellite clock difference observations to generate a comprehensive time scale to obtain the optimized time scale.

[0091] 134. Based on the optimized time scale, the clock error prediction model is used to suppress long-term drift and generate a unified time scale for the constellation.

[0092] In an embodiment of the present invention, a high-reliability data foundation is constructed, and the corrected inter-satellite clock error data and inter-satellite link measurement results (such as signal strength and Doppler shift) are uploaded to a distributed processing unit. Preprocessing operations such as outlier removal, data alignment, and noise filtering are performed to eliminate gross errors and outliers, ensuring the quality of input data. After preprocessing, the data integrity rate is improved to 99.9%, the outlier filtering rate reaches 95%, the data time-scale alignment accuracy is better than 0.1ns, the inter-satellite link transmission delay differences are eliminated, and the noise level is reduced to 3cm (corresponding to a delay of 0.1ns), providing a clean data source for subsequent fusion. The multi-source data fusion strategy is optimized, and the clock error stability is evaluated in real time based on the Allan variance of the onboard atomic clock, and the fusion weight is dynamically allocated. The Allan variance reflects the short-term stability (such as the 1-second sampling interval) and long-term drift characteristics of the clock. The dynamic weight allocation improves the stability of the fused clock error. Compared with the static weighting scheme, the high-stability rubidium clock (Allan variance 1×10 -11 ) weight ratio is automatically increased to suppress low-performance clock noise, and the fused data update rate reaches 1Hz, adapting to the rapid topology changes of low-orbit constellations; noise suppression and state estimation are achieved. The first-level filtering eliminates white noise and flicker noise for the clock error sequence of a single satellite, and the second-level filtering integrates multi-satellite observations to generate a globally consistent time scale. The first-level filtering reduces the power spectral density of single clock noise by 30dB. The short-term stability and constellation-level time scale consistency after the second-level filtering are better than 0.05ns, meeting the ITU-T G.811 clock level requirements. The filter convergence time is less than 100 seconds, adapting to the dynamic scenario of satellite network entry / exit; to ensure the long-term stability of the time scale, based on the optimized time scale, the autoregressive moving average (ARMA) model and periodic term compensation algorithm are used to predict the clock error changes in the next 24 hours and compensate for the frequency drift in advance. The clock error prediction error is less than 0.5ns (24-hour prediction period), which is 5 times higher than the traditional polynomial fitting, and the long-term frequency drift rate is suppressed to 5×10 -12 / day, meeting the constellation's 30-day autonomous operation requirement. The forecast model adaptively adjusts the periodic term to adapt to changes in the space environment such as the solar activity cycle; making the constellation's time scale stability reach 1×10 -13 / τ (short term) and 5×10 -12 / day (long term), meeting the stringent requirements of integrated communication and navigation on time and frequency benchmarks. Compared with traditional ground injection solutions, it has improved autonomy and significantly reduced dependence on ground stations.

[0093] like Figure 1 As shown in 14, based on the unified time scale of the constellation, dynamic topology optimization is performed to obtain the optimized network topology, including:

[0094] 141. Based on the unified time scale of the constellation, link quality, network-wide clock error gradient, and orbit parameters are collected in real time. Link quality is evaluated through a scoring model, and abnormal nodes with deviations are marked and isolated.

[0095] 142. After isolating the abnormal node, the clock error is reconstructed using the redundant observation values ​​of the neighboring nodes. After checking the residual, the node function is restored to obtain the optimized network topology.

[0096] In the embodiment of the present invention, an adaptive anti-interference intersatellite network topology is constructed. By real-time monitoring of link quality (such as signal-to-noise ratio, bit error rate), network-wide clock gradient (time deviation distribution between nodes) and orbit parameters (relative position, speed), a multi-parameter scoring model is used to quantify link health, and abnormal nodes that deviate from the normal range are dynamically marked. Link quality assessment: integrating signal-to-noise ratio (SNR>20dB is healthy), bit error rate (BER<1×10 -16 ) and data packet loss rate (<1%), with weights of 40% / 40% / 20%; clock gradient analysis: calculates the standard deviation of clock errors among nodes in the entire network. If a node causes a sudden increase in the local gradient (>0.3ns), an isolation warning is triggered; orbit parameter verification: combines ephemeris data to verify the rationality of node motion trajectories and eliminate false anomalies caused by orbit prediction errors; the abnormal node detection rate reaches 99% and the false alarm rate is less than 0.1%, significantly better than the traditional threshold comparison method; the link isolation response time is less than 500ms, preventing the spread of faults from affecting the time synchronization of the entire network. After isolation, the network connectivity retention rate is ≥95%, ensuring that the basic functions of the constellation are not impaired; to achieve rapid self-healing and functional recovery of faulty nodes, redundant clock observations of healthy neighboring nodes (at least three neighboring satellites) are used to reconstruct the clock of the isolated node using the weighted least squares method. The reconstructed residual (the difference from the predicted value) is checked to see if it meets the threshold (<0.2ns). If so, the node is restored to function; redundant observation selection: priority is given to nodes with high clock stability (Allan variance <1×10 -11), neighboring nodes with good link quality (SNR>25dB); clock error reconstruction algorithm: using weighted least squares method, the weight is inversely proportional to the stability of neighboring satellite clock error; residual verification mechanism: calculating the residual between the reconstructed value and the predicted value (based on historical data extrapolation). If the residual is less than 0.2ns for three consecutive measurements, the node is judged to have recovered, the utilization rate of redundant observations is improved, and the fault tolerance of the constellation to single-point failures is enhanced; the availability of the constellation network is improved, and the fault recovery time is shortened to minutes. Compared with the traditional ground manual intervention solution, the efficiency is improved, and the autonomous operation capability of the low-orbit constellation in a complex space environment is significantly enhanced.

[0097] like Figure 1 As shown in Figure 15, based on the optimized network topology, Kalman filtering and polynomial prediction are used to maintain the time base, and the LSTM neural network is combined to predict the frequency drift change to obtain the autonomous timekeeping results, including:

[0098] 151. Based on the optimized network topology, the inter-satellite clock error observations, atomic clock historical frequency deviation data, and environmental parameters are aligned, outliers are removed, and time series features such as the clock error change rate are extracted to obtain pre-processed topological data;

[0099] 152. Based on the pre-processed topological data, the constellation maintains temporal stability through two-level correction to obtain the corrected topological data;

[0100] 153. Based on the corrected topological data, the LSTM neural network is deployed and combined with the historical clock error series to predict the frequency drift changes in the next 12 hours to obtain the autonomous timekeeping results.

[0101] In an embodiment of the present invention, a data foundation for maintaining a high-precision time base is constructed. The inter-satellite clock error observation values, atomic clock historical frequency deviation data (such as frequency accuracy, aging rate) and environmental parameters (such as temperature, radiation dose) in the optimized network topology are aligned in time and space to eliminate gross errors (such as abnormal values ​​caused by signal jumps and equipment failures), and extract time series features such as clock error change rate and frequency drift trend; Time and space alignment: Use interpolation algorithms (such as cubic spline interpolation) to convert data from different satellites and at different times to a unified time base, and the spatial alignment error is less than 0.1ns; Outlier elimination: Based on the 3σ principle and the isolation forest algorithm, detect and eliminate clock error mutation points (such as jumps > 0.5ns), and the filtering rate reaches 95%; Feature extraction: Calculate the first-order difference (rate of change) and second-order difference (acceleration) of the clock error, and construct a multi-dimensional feature vector in combination with environmental parameters; Data integrity is improved, and the missing value filling error is less than 0.0 5ns, the feature dimension is compressed to within 10 dimensions, the computing efficiency is increased by 3 times, and it adapts to the resource limitations of the onboard processor. The correlation analysis of environmental parameters improves the accuracy of the frequency drift prediction model; through layered filtering to eliminate error sources and improve the stability of the time base, the first-level correction uses Kalman filtering to eliminate the short-term noise of the single satellite clock (such as white noise and flicker noise). The second-level correction fuses multi-satellite observations to generate a globally consistent time scale and suppress long-term drift; First-level correction: The Kalman filter model state vector contains clock error, frequency error, and frequency drift. The observation equation fuses the inter-satellite clock error observations with the atomic clock physics model (such as the rubidium clock aging rate model); Second-level correction: Using a federal filtering architecture, the local filtering results of each satellite are weighted and fused at the master control node, and the weights are dynamically adjusted according to the covariance of the clock error observations; the short-term stability (Allan variance) of a single clock reaches 1×10 -13 / τ, the long-term frequency drift rate is suppressed to 2×10 -11 / day; the time scale consistency after two-level correction is better than 0.03ns, meeting the clock grade requirements of ITU-T G.8272.1; the correction calculation delay is less than 50ms, adapting to the rapid dynamic scenarios of low-orbit constellations; using deep learning models to capture the nonlinear changes in frequency drift, extend the autonomous timekeeping period, train LSTM networks based on historical clock error sequences (such as the past 24 hours of data), predict frequency drift changes in the next 12 hours, and combine the two-level correction results to generate autonomous timekeeping instructions; network structure: two-layer LSTM (64 neurons per layer) + The fully connected layer takes clock error, frequency error, and environmental parameters as input, and outputs the predicted frequency drift for the next 12 hours. The training strategy uses a sliding window method to generate training samples (with a window length of 24 hours and a step length of 1 hour), and the loss function is the mean square error (MSE). Prediction fusion involves weighted fusion of the LSTM prediction value with the output of the physical model (such as the atomic clock aging model), with the weights dynamically adjusted based on the prediction confidence. The closed-loop process of "data preprocessing - two-stage correction - deep learning prediction" achieves high-precision autonomous maintenance of the low-orbit constellation time base. This allows the constellation's autonomous timekeeping accuracy to reach the sub-nanosecond level (0.1ns), and the timekeeping period to be extended to 72 hours, meeting the stringent requirements of integrated communication and navigation for time and frequency bases. Compared to traditional ground injection schemes, this scheme offers improved autonomy and significantly reduces dependence on ground stations.

[0102] like Figure 1 As shown in 16, based on the autonomous timekeeping results, the abnormal clock difference is verified by the majority vote of the neighboring nodes. If the vote fails, the fault is isolated and the recovery value is calculated in reverse, including:

[0103] 161. Based on the autonomous timekeeping results, a majority voting mechanism is used to verify the clock difference data of the target node to obtain the verification result;

[0104] 162. Based on the verification results, an inter-satellite alarm signal is triggered to isolate the abnormal node from the synchronization network to obtain the clock error after isolation;

[0105] 163. Based on the clock error after isolation, using historical clock error data and orbital dynamics models, the theoretical clock error value of the faulty node is reversely calculated using the Kalman smoothing algorithm to compensate for the time deviation during the missing period and obtain the reverse calculation recovery result;

[0106] 164. Based on the reverse calculation recovery result, the calculated value is re-entered into the majority voting mechanism for secondary verification to obtain the recovery verification result;

[0107] 165. Based on the recovery verification results, update the intersatellite link topology and optimize the time transfer path.

[0108] In an embodiment of the present invention, autonomous fault handling of the low-orbit constellation time synchronization network is achieved through a closed-loop process of "distributed voting-fast isolation-reverse calculation-secondary verification-topology optimization". The process first uses a weighted majority voting mechanism to verify the target node clock error data based on the autonomous timekeeping results, and achieves high-precision fault detection through neighboring satellite voting and dynamic threshold adjustment, with an anomaly detection rate of 98% and a false alarm rate of less than 0.2%. Subsequently, an inter-satellite alarm signal is triggered and the faulty node is isolated. Combined with the physical layer and logical layer isolation strategy and redundant link switching, the fault isolation success rate is ensured to reach 99% and the service interruption time is less than 1 second. Then, the Kalman smoothing algorithm is used to fuse historical data with the orbital dynamics model to reversely calculate the fault time. The theoretical clock difference value of the segment was calculated, with an error of <0.15ns and a computational complexity reduced by 60%. The calculated value was then verified twice, and the recovery verification pass rate was ensured to reach 90% by tightening the voting threshold, comparing historical trends, and performing redundant cross-validation. Finally, the inter-satellite link topology was dynamically updated based on the recovery results, and a multi-parameter weighted path planning algorithm was used to optimize the time transfer path, reducing the time transfer delay by 20% and achieving a network-wide time base stability of 0.02ns. This process innovatively combined weighted voting, Kalman smoothing, and dynamic topology optimization technologies, enabling the constellation's autonomous fault handling capability to reach over 90%, shortening service interruption time to seconds, and significantly improving the survivability and time synchronization robustness of low-orbit constellations in complex space environments.

[0109] Figure 2 It is a structural diagram of an electronic device provided by an embodiment of the present invention.

[0110] like Figure 2 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 may call logic instructions in the memory 630 to execute a low-orbit constellation autonomous time synchronization method based on the Ka inter-satellite link.

[0111] In addition, the logic instructions in the aforementioned memory 630 can be implemented in the form of a software functional unit and, when sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0112] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the low-orbit constellation autonomous time synchronization method based on the Ka inter-satellite link provided by the above methods.

[0113] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the low-orbit constellation autonomous time synchronization method based on the Ka inter-satellite link provided by the above methods.

[0114] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0115] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion 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, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for autonomous time synchronization of low earth orbit constellations based on Ka inter-satellite links, characterized in that: include: Obtain the Ka-band standard timestamp signal from the ground station, and use Kalman filtering and Doppler compensation to train the satellite clock to obtain the initial anchor node of the constellation. Based on the constellation initial anchor node, time division multiplexing is used to bidirectionally measure the distance between the constellation initial anchor node and the adjacent satellites, and a single-point pseudo-range model is used to correct the motion delay to obtain the corrected inter-satellite clock error data. Based on the corrected inter-satellite clock error data, clock error data fusion is performed, and a clock error prediction model is used to suppress long-term drift and generate a unified time scale for the constellation. Based on the unified time scale of the constellation, dynamic topology optimization is performed to obtain the optimized network topology; Based on the optimized network topology, Kalman filtering and polynomial prediction are used to maintain the time base, and the LSTM neural network is combined to predict frequency drift changes to obtain autonomous timekeeping results. Based on the autonomous timekeeping results, the abnormal clock difference is verified through majority voting of neighboring nodes. If the vote fails, the fault is isolated and the recovery value is calculated reversely.

2. The method for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links according to claim 1, characterized in that: Obtain the ground station Ka-band standard timestamp signal, and combine Kalman filtering and Doppler compensation to perform satellite clock training to obtain the constellation initial anchor node, including: The ground station transmits a Ka-band standard timestamp signal, and the satellite receiver captures and demodulates the time information; Based on the demodulated time information, the satellite atomic clock is used to generate a second pulse signal, and the satellite-to-ground clock difference is measured by a time interval counter. Based on the satellite-to-ground clock error, a Kalman filter is used to smooth the clock error sequence, synchronously estimate the frequency error and frequency drift parameters, and generate the frequency adjustment command of the voltage-controlled crystal oscillator; Based on the frequency adjustment instruction of the voltage-controlled crystal oscillator, the adjustment amount is loaded to the crystal oscillator control end through the digital-to-analog conversion module, and the satellite clock phase is gradually locked to the ground reference to obtain the initial anchor node of the constellation.

3. The method for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links according to claim 2, characterized in that: Based on the constellation initial anchor node, time division multiplexing is used to bidirectionally measure the distance between the constellation initial anchor node and the adjacent satellites. The single-point pseudo-range model is used to correct the motion delay to obtain the corrected inter-satellite clock error data, including: Based on the initial anchor node of the constellation, a time-division multiplexing ranging link is established between the anchor node satellite and the adjacent satellite in the Ka band, and ranging signals are alternately transmitted and received according to the preset time slots. Based on the ranging signal, the orbit and clock errors are decoupled by timestamp alignment to obtain two-way pseudorange observations. Based on the two-way pseudorange observations, the single-point pseudorange reduction model is used to perform dynamic error correction to obtain the corrected pseudorange; Based on the corrected pseudorange and combined with the relativistic effect correction term, the inter-satellite relative clock error is decoupled and obtained. The time delay accumulation error of the multi-hop path is modeled and distributed to compensate for it, and the corrected inter-satellite clock error data is obtained.

4. The method for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links according to claim 3, characterized in that: Based on the corrected inter-satellite clock error data, clock error data fusion is performed, and a clock error prediction model is used to suppress long-term drift and generate a unified constellation time scale, including: Uploading the corrected inter-satellite clock error data and inter-satellite measurement results to the distributed processing unit and pre-processing them to obtain pre-processed data; Based on the pre-processed data, weights are dynamically assigned according to the Allan variance of the onboard atomic clock to obtain weighted fusion data; Based on the weighted fused data, the first-level Kalman filter is used to eliminate single-clock noise, and the second-level Kalman filter is used to fuse the inter-satellite clock difference observations to generate a comprehensive time scale to obtain the optimized time scale. Based on the optimized time scale, the clock error prediction model is used to suppress long-term drift and generate a unified time scale for the constellation.

5. The method for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links according to claim 4, characterized in that: Based on the unified time scale of the constellation, dynamic topology optimization is performed to obtain the optimized network topology, including: Based on the unified time scale of the constellation, link quality, network-wide clock error gradient, and orbit parameters are collected in real time. Link quality is evaluated using a scoring model, and abnormal nodes with deviations are marked and isolated. After isolating the abnormal node, the clock error is reconstructed using the redundant observations of neighboring nodes. After verifying the residual, the node function is restored to obtain the optimized network topology.

6. The method for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links according to claim 5, characterized in that: Based on the optimized network topology, Kalman filtering and polynomial prediction are used to maintain the time base, and the LSTM neural network is combined to predict frequency drift changes to obtain autonomous timekeeping results, including: Based on the optimized network topology, the inter-satellite clock error observations, atomic clock historical frequency offset data, and environmental parameters are aligned, outliers are removed, and time series features such as the clock error change rate are extracted to obtain pre-processed topological data. Based on the pre-processed topological data, the constellation maintains temporal stability through two-level correction to obtain the corrected topological data; Based on the corrected topological data, an LSTM neural network is deployed, and combined with the historical clock error series to predict the frequency drift changes in the next 12 hours, and the autonomous timekeeping results are obtained.

7. The method for autonomous time synchronization of a low-orbit constellation based on Ka inter-satellite links according to claim 6, characterized in that: Based on the autonomous timekeeping results, the abnormal clock error is verified by a majority vote of neighboring nodes. If the vote fails, the fault is isolated and the recovery value is calculated in reverse, including: Based on the autonomous timekeeping results, a majority voting mechanism is used to verify the clock error data of the target node to obtain the verification result; Based on the verification results, an inter-satellite alarm signal is triggered, and the abnormal node is isolated from the synchronization network to obtain the clock error after isolation; Based on the clock error after isolation, the theoretical clock error value of the faulty node is reversely calculated using historical clock error data and orbital dynamics models through the Kalman smoothing algorithm, compensating for the time deviation during the missing period and obtaining the reverse calculation recovery result. Based on the reverse calculation recovery result, the estimated value is re-entered into the majority voting mechanism for secondary verification to obtain the recovery verification result; Based on the recovery verification results, the inter-satellite link topology is updated and the time transfer path is optimized.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for autonomous time synchronization of a low-orbit constellation based on a Ka inter-satellite link is implemented as described in any one of claims 1 to 7.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for autonomous time synchronization of a low-orbit constellation based on a Ka inter-satellite link is implemented as claimed in any one of claims 1 to 7.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for autonomous time synchronization of a low-orbit constellation based on a Ka inter-satellite link is implemented as claimed in any one of claims 1 to 7.

Images