GNSS / space-to-space link distributed cooperative real-time orbit determination method and device for mega constellation

Abstract

The application provides a GNSS / inter-satellite link distributed cooperative real-time orbit determination method and device for a mega constellation, first, each satellite in a cluster performs preliminary orbit determination based on local GNSS observation data and IMU data to obtain preliminary determination results of each satellite in the cluster; then, a cluster head satellite collects the preliminary determination results to determine relative orbit parameters in the cluster; then, a backbone node collects the relative orbit parameters to determine a regional absolute orbit reference; a global anchor satellite determines unified constellation orbit parameters; finally, according to the regional absolute orbit reference and the unified constellation orbit parameters, orbit determination results of each satellite are determined. Through a three-level hierarchical cooperative architecture, the application disperses data processing and determination tasks to different level nodes, avoids data processing bottlenecks of a central node and communication link congestion problems in centralized technology, reduces complex interaction between nodes and inter-satellite link resource occupation in a thousand-star level scale in traditional distributed technology, and shortens iteration convergence time.

Description

Technical Field

[0001] This invention relates to the field of satellite orbit determination technology, and in particular to a distributed collaborative real-time orbit determination method and apparatus for GNSS / inter-satellite links for mega-constellations. Background Technology

[0002] With the rise of large low-Earth orbit constellations, their orbit determination missions face enormous challenges. Traditional orbit determination methods relying on ground stations are insufficient to meet the needs of real-time management of tens of thousands of satellites. Using onboard GNSS receivers for orbit determination is the mainstream approach, but when GNSS signals are interfered with or interrupted, orbit determination accuracy drops sharply or even fails.

[0003] With the exponential growth in the number of satellites, the amount of data and computational load that central nodes need to process has increased dramatically, leading to computational bottlenecks. Simultaneously, all data needs to be transmitted back to the center, resulting in severe communication link congestion and enormous communication overhead, failing to meet real-time requirements. While traditional distributed orbit determination methods are decentralized, at a scale of thousands of satellites, the complex communication interactions and iterative convergence processes between nodes still consume a significant amount of valuable inter-satellite link resources, and the convergence time increases significantly with the scale. Summary of the Invention

[0004] This invention provides a distributed collaborative real-time orbit determination method and apparatus for GNSS / inter-satellite links for mega-constellations, which solves the problem that the convergence time of distributed orbit determination in the prior art increases significantly with the scale.

[0005] The first aspect of this invention provides a distributed, cooperative, real-time orbit determination method for GNSS / inter-satellite links in a mega-constellation. The mega-constellation is divided into multiple orbital plane clusters, each cluster comprising multiple satellites. Several global anchor satellites are evenly distributed within the mega-constellation. Each cluster has a cluster head satellite, and a first inter-satellite link exists between satellites within the cluster. A regional backbone node is established within multiple adjacent orbital plane clusters, and a second inter-satellite link exists between the backbone nodes. The method includes: Preliminary orbit calculations are performed on the satellites within each cluster based on local GNSS observation data and IMU data to obtain preliminary calculation results for each satellite within the cluster; The cluster head satellite summarizes the preliminary calculation results of each satellite in the cluster to determine the relative orbital parameters within the cluster.

[0006] The backbone nodes collect the relative orbital parameters of each cluster and solve for the absolute orbital reference of the region; Global anchor satellites aggregate reference values ​​from various regions to determine unified orbital parameters for the constellation. Based on the regional absolute orbit reference and the unified orbit parameters of the constellation, the orbit calculation results for each satellite are determined.

[0007] In one possible implementation, satellites within each cluster perform preliminary orbit calculations based on local GNSS observation data and IMU data to obtain preliminary calculation results for each satellite within the cluster, including: Satellites with GNSS receivers deployed within the cluster collect multi-mode GNSS observation data and IMU data, and perform preprocessing to obtain preprocessed observation data; Satellites within a cluster that do not have GNSS receivers deployed request raw GNSS observation data and data confidence scores from adjacent GNSS satellites within the cluster via the first inter-satellite link, forming a joint observation dataset. The sliding window is a local orbit determination sliding time window maintained by each satellite, and the window length is dynamically adjusted according to the satellite's orbital dynamics and the quality of the observation data. Each satellite performs preliminary orbit calculations based on preprocessed observation data and joint observation datasets, using a simplified orbit dynamics model.

[0008] In one possible implementation, the cluster head satellite aggregates the preliminary solution results from all satellites within the cluster to determine the relative orbital parameters within the cluster, including: Based on the preliminary calculation results of each satellite in the cluster and the weighted consensus algorithm, the states of the satellites in the cluster are fused to obtain state fusion data; the weights in the fusion process are determined according to the confidence level of the preliminary calculation results of each satellite. Using the state of the cluster head satellite as a reference, the state deviation of each satellite in the cluster relative to the cluster head satellite is calculated based on the state fusion data to construct the relative orbital parameters within the cluster.

[0009] In one possible implementation, the backbone nodes collect the relative orbital parameters of each cluster and calculate the regional absolute orbital reference, including: Based on the backbone node's own GNSS and IMU data, calculate its own absolute orbital parameters; Based on the relative orbit parameters and absolute orbit parameters of each cluster, calculate the transformation parameters between the coordinate system of each cluster and the coordinate system of the backbone node, and perform coordinate transformation on the relative orbit parameters of each cluster. The inter-satellite ranging data with adjacent backbone nodes is obtained as a constraint and jointly solved with the relative orbital parameters of each cluster after coordinate transformation to determine the absolute orbital parameters of the region. The residuals of the optimized regional absolute orbit parameters are evaluated. If they meet the accuracy requirements, they are determined as the regional absolute orbit reference and distributed to each cluster. Otherwise, the inter-satellite ranging data of other backbone nodes are obtained as constraints, and the regional absolute orbit parameters are recalculated.

[0010] In one possible implementation, global anchor satellites aggregate reference values ​​from various regions to determine unified orbital parameters for the constellation, including: Global anchor satellites calculate the absolute orbital parameters of the anchor points based on their own multi-mode GNSS and laser ranging data. Using its own anchoring benchmark as a reference, the systematic deviation of the benchmark values ​​in each region is calculated based on the absolute track parameters of the anchor point and the deviation correction model. The regional correction coefficient is then determined, and the regional benchmark data is corrected based on the regional correction coefficient. Under the constraint of inter-satellite ranging between anchor points, the unified orbital parameters of the constellation are obtained based on the regional reference data after deviation correction.

[0011] In one possible implementation, the orbital calculation results for each satellite are determined based on the regional absolute orbital reference and the unified orbital parameters of the constellation, including: The cluster head satellite receives constellation unified orbit parameters and regional correction coefficients forwarded by backbone nodes in the region, and aligns them with its own stored regional absolute orbit reference in terms of time and coordinate system. The cluster head satellite uses the unified orbital parameters of the constellation as a reference, combines them with the regional absolute orbital reference, and corrects the regional reference deviation through a weighted fusion algorithm to output the corrected regional absolute orbital reference. Based on the corrected regional absolute orbit reference and the relative orbit parameters within the cluster, the cluster head satellite calculates the preliminary absolute orbit solution for each satellite within the cluster; The cluster head satellite collects inter-satellite ranging data between satellites within the cluster as a local constraint, and performs joint calculation with the preliminary absolute orbit results to obtain the orbit calculation results.

[0012] In one possible implementation, the cluster head satellite collects inter-satellite ranging data between satellites within the cluster as a local constraint, and performs joint calculation with the preliminary absolute orbit results to obtain the orbit calculation results, including: Retrieve the preliminary absolute orbit results of each satellite in the cluster and perform timestamp calibration; Based on the preliminary absolute orbit results and constrained by effective inter-satellite ranging data, a weighted least squares joint solution model is constructed, and the weights are dynamically allocated according to the data quality. The orbital solution results are determined based on the joint solution model.

[0013] In one possible implementation, the method further includes: real-time monitoring of satellite status within the cluster, triggering cluster reconstruction when preset conditions are met; updating the inter-satellite link topology based on the reconstructed cluster structure and LSTM orbit prediction model; and dynamically adjusting the communication frequency of each level of inter-satellite link based on the updated inter-satellite link topology.

[0014] In one possible implementation, the method further includes: when a satellite failure is detected, calculating alternative orbital parameters for the failed satellite based on its historical orbital data and information on neighboring satellites.

[0015] A second aspect of the present invention provides a distributed, cooperative, real-time orbit determination device for GNSS / inter-satellite links for mega-constellations. The mega-constellation is divided into multiple orbital plane clusters, each cluster comprising multiple satellites. Several global anchor satellites are evenly distributed within the mega-constellation. Each cluster has a cluster head satellite, and a first inter-satellite link exists between satellites within the cluster. A regional backbone node is located within multiple adjacent orbital plane clusters, and a second inter-satellite link exists between the backbone nodes. The device comprises: The satellite calculation module is used to perform preliminary orbit calculations for satellites within each cluster based on local GNSS observation data and IMU data, and obtain preliminary calculation results for each satellite within the cluster; The cluster-level solution module is used by the cluster head satellite to summarize the preliminary solution results of each satellite in the cluster and determine the relative orbital parameters within the cluster.

[0016] The regional solution module is used by backbone nodes to collect the relative orbital parameters of each cluster and solve for the regional absolute orbital reference. The anchor point calculation module is used to summarize the reference values ​​of various regions for global anchor point satellites and determine the unified orbital parameters of the constellation. The final calculation module is used to determine the orbit calculation results of each satellite based on the regional absolute orbit reference and the unified orbit parameters of the constellation.

[0017] Compared to traditional technologies, this invention provides a distributed collaborative real-time orbit determination method and apparatus for GNSS / inter-satellite links for mega-constellations. First, satellites within each cluster perform preliminary orbit calculations based on local GNSS observation data and IMU data to obtain preliminary calculation results for each satellite within the cluster. Then, the cluster head satellite summarizes the preliminary calculation results of each satellite within the cluster to determine the relative orbit parameters within the cluster. Next, backbone nodes collect the relative orbit parameters of each cluster and calculate the regional absolute orbit reference. Global anchor point satellites summarize the reference values ​​of each region to determine the unified orbit parameters for the constellation. Finally, based on the regional absolute orbit reference and the unified orbit parameters for the constellation, the orbit calculation results for each satellite are determined. This invention employs a three-tiered collaborative architecture: in-cluster preliminary calculation – cluster head summarizing relative parameters – backbone node calculation of regional benchmarks – global anchor point determination of unified parameters. This architecture distributes data processing and calculation tasks across different levels of nodes, avoiding the data processing bottlenecks and communication link congestion issues of the central node in centralized technologies. It also reduces the complex interactions between nodes and the inter-satellite link resource consumption in traditional distributed technologies at the scale of thousands of satellites, while shortening the iteration convergence time. Ultimately, this achieves low computational load, low communication overhead, and high real-time performance for orbit determination of giant constellations. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the implementation of the distributed collaborative real-time orbit determination method for GNSS / inter-satellite links for mega-constellations provided in this embodiment of the invention. Detailed Implementation

[0019] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating the implementation of the distributed cooperative real-time orbit determination method for GNSS / inter-satellite links for mega-constellations provided in this invention. Figure 1 As shown, the mega-constellation is divided into multiple orbital plane clusters, each cluster containing multiple satellites; several global anchor satellites are evenly distributed within the mega-constellation; a cluster head satellite is set within each cluster, and there are first inter-satellite links between satellites within the cluster; a regional backbone node is set within multiple adjacent orbital plane clusters, and there are second inter-satellite links between the backbone nodes. The method includes: S110, each cluster of satellites performs preliminary orbit calculations based on local GNSS observation data and IMU data to obtain preliminary calculation results for each satellite in the cluster; S120, the cluster head satellite summarizes the preliminary calculation results of each satellite in the cluster and determines the relative orbital parameters within the cluster.

[0021] S130, the backbone nodes collect the relative orbital parameters of each cluster and solve for the absolute orbital reference of the region; S140: Global anchor point satellites summarize reference values ​​from various regions to determine unified orbital parameters for the constellation; S150 determines the orbit calculation results for each satellite based on the regional absolute orbit reference and the unified orbit parameters of the constellation.

[0022] In this embodiment of the invention, within each orbital plane cluster, some satellites are equipped with multi-mode GNSS receivers to acquire multi-system observation data. All satellites synchronously acquire angular and linear acceleration data output by inertial measurement units (IMUs). Statistical criteria combined with the isolated forest algorithm are used to remove outlier observations from the GNSS data, ensuring data reliability. For the IMU data, zero-bias compensation and drift correction are performed periodically on the attitude reference values ​​output by star sensors to suppress the cumulative error of the inertial data and provide a high-quality observational foundation for subsequent calculations.

[0023] Each satellite within the cluster maintains a dynamic sliding time window. The window length is adaptively adjusted based on the satellite's orbital dynamics (such as orbital curvature and rate of change of velocity) and the quality of the observation data. Smaller windows are used in highly dynamic scenarios to quickly track state changes, while larger windows are used in stable scenarios to accumulate data and improve accuracy. Satellites without deployed GNSS receivers acquire in-window observation data and confidence indices from neighboring GNSS satellites via inter-satellite links within the cluster, forming a joint observation dataset of "its own IMU data + neighboring GNSS data." Based on this dataset, all satellites use a simplified dynamic model that considers only the core orbital perturbation term to perform preliminary orbit calculations using a weighted least squares algorithm, outputting preliminary results for position, velocity, and clock bias.

[0024] The cluster head satellite receives the initial calculation results from all satellites via the intra-cluster inter-satellite link. It first performs data integrity verification and validity screening, eliminating obviously abnormal results and sending back retransmission requests. Based on valid results, the cluster head uses a locally weighted consensus algorithm for iterative fusion. During the fusion process, weights are dynamically allocated according to the confidence level of each satellite's observation data. GNSS satellite weights are associated with their observation data quality indicators, while non-GNSS satellite weights are associated with the confidence level of neighboring GNSS data and the correction effect of their own IMU data. After iterative convergence, using the cluster head's own fusion results as a reference, the relative positions, velocities, and clock biases between all satellites within the cluster and the cluster head are calculated to form intra-cluster relative orbital parameters. Finally, the reliability of the parameters is verified through an overall consensus assessment; if qualified, the parameters are distributed to all satellites within the cluster.

[0025] The regional backbone nodes receive the relative orbit parameters from multiple orbital plane clusters under their jurisdiction, first filtering out cluster data with poor consistency. Based on data from their deployed multi-mode GNSS receivers and IMUs, the backbone nodes perform their own absolute orbit calculations using a precise dynamic model that considers multi-order orbit perturbations, obtaining high-precision absolute orbit parameters as the anchoring reference for regional orbit determination. Using this reference, the Helmholtz transform algorithm is used to calculate the transformation parameters between the local coordinate system of each cluster and the absolute coordinate system of the backbone nodes. Through coordinate transformation, the relative orbit parameters of all clusters are unified to the absolute coordinate system, obtaining the preliminary absolute orbit values ​​of each cluster's satellites, while eliminating systematic deviations between clusters caused by local calculations.

[0026] Backbone nodes acquire high-precision inter-satellite ranging data with adjacent backbone nodes through regional backbone links. After outlier removal, this data is used as a strong constraint in the calculation process. Based on the preliminary absolute orbit values ​​of each cluster after coordinate transformation, and combined with the inter-satellite ranging constraints of adjacent backbone nodes, a weighted least squares algorithm is used for joint calculation, and orbit parameters are optimized through data fusion. After the calculation is completed, the overall accuracy of the regional absolute orbit parameters is evaluated. If it meets the preset requirements, it is determined as the regional absolute orbit benchmark and distributed to all clusters under its jurisdiction; if the accuracy does not meet the requirements, more ranging data from adjacent backbone nodes are added and the calculation is repeated to ensure the reliability of the regional benchmark.

[0027] Global anchor satellites collect absolute orbital references for each region and filter valid data through periodic communication with regional backbone nodes. Based on their deployed high-precision observation equipment (multi-mode GNSS receivers, laser ranging equipment, etc.), the anchor satellites use a precise model considering complex perturbation and relativistic effects to calculate their own absolute orbital parameters, forming a high-precision global orbit determination anchoring reference. Using this reference, the systematic deviation between each regional reference and the global reference is calculated, and the deviation is corrected by incorporating Earth orientation parameters to obtain regional reference data in a unified coordinate system. All corrected regional reference data are fused, and inter-satellite ranging constraints between anchors are introduced to perform a global joint solution. After iterative convergence, the unified orbital parameters of the constellation are obtained and periodically distributed to each regional backbone node.

[0028] After receiving the unified orbit parameters of the constellation, the regional backbone nodes synchronize their time and coordinate system with their stored regional absolute orbit reference. A weighted fusion algorithm is then used to correct the deviation between the regional reference and the global parameters, forming a corrected regional reference, which is then distributed to each cluster head satellite. Using the corrected regional reference as a reference, and combining it with the previously constructed intra-cluster relative orbit parameters, the cluster head satellites obtain the preliminary absolute orbit results for each satellite within the cluster through a calculation method of "reference value + relative deviation". Subsequently, high-precision inter-satellite ranging data is collected among the satellites within the cluster as local constraints, and this data is jointly solved with the preliminary results using a weighted least squares algorithm to optimize the orbit parameters. After passing accuracy verification, the final orbit calculation results for each satellite are determined and distributed to the corresponding satellite. Simultaneously, consistency verification between the overall regional results and global parameters is completed, and the results are updated periodically according to the global calibration cycle.

[0029] In some embodiments, satellites within each cluster perform preliminary orbit calculations based on local GNSS observation data and IMU data to obtain preliminary calculation results for each satellite within the cluster. This includes: satellites with GNSS receivers deployed within the cluster collecting multi-mode GNSS observation data and IMU data, and performing preprocessing to obtain preprocessed observation data; satellites without GNSS receivers deployed within the cluster requesting raw GNSS observation data and data confidence levels within their sliding windows from adjacent GNSS satellites within the cluster via a first inter-satellite link to form a joint observation dataset; wherein, the sliding window is a local orbit determination sliding time window maintained by each satellite, and the window length is dynamically adjusted according to the satellite orbit dynamic characteristics and observation data quality; each satellite performs preliminary orbit calculations based on the preprocessed observation data and the joint observation dataset, using a simplified orbit dynamics model.

[0030] In this embodiment of the invention, to control hardware costs while ensuring computational accuracy, a differentiated observation equipment deployment strategy is adopted within the cluster: some satellites are equipped with multi-mode GNSS receivers, capable of simultaneously receiving observation signals from multiple satellite navigation systems, while all satellites are equipped with IMU devices to collect inertial data such as angular acceleration and linear acceleration. This deployment method ensures the existence of an absolute observation benchmark within the cluster and achieves dynamic state awareness of all satellites in the cluster through general-purpose IMU devices, providing a basic data source for subsequent joint computation. Satellites without deployed GNSS receivers acquire observation data through intra-cluster communication links, forming full cluster data coverage.

[0031] The acquired raw data needs to be preprocessed to remove errors and interference, ensuring the reliability of the solution. For GNSS observation data, statistical tests combined with the isolated forest algorithm are used to identify and remove outliers, which may be caused by factors such as signal obstruction and electromagnetic interference. At the same time, the data is time-synchronized to ensure that the observation data from different satellites are on a unified time reference. For IMU data, zero-bias compensation and drift correction are performed periodically using the attitude reference values ​​provided by the star sensor to suppress the long-term accumulated errors of inertial data. This ensures that the preprocessed IMU data can accurately reflect the dynamic motion state of the satellite, providing high-quality input for subsequent solutions.

[0032] Each satellite maintains a local orbit-determining sliding time window. Its core function is to select observation data within a certain period as the solution sample, balancing solution accuracy and real-time performance. The window length is not fixed but dynamically adjusted based on the satellite's orbital dynamics and the quality of the observation data: when the satellite is in a highly dynamic operating state (such as orbital maneuvering or traversing complex perturbation regions) or when the quality of the observation data fluctuates significantly, a shorter window is used to quickly track state changes; when the satellite is operating stably and the data quality is excellent, a longer window is used to accumulate more effective data and improve solution accuracy. The latest observation data is updated in real time within the window, and expired data is removed to always maintain the timeliness of the sample. The orbit-determining sliding time window is the core mechanism for achieving a balance between "accuracy and real-time performance" locally on the satellite. Its adjustment strategy adopts a "quantization threshold trigger + multi-dimensional index linkage" mode. The specific logic and operational details are as follows: First, the basic window configuration is defined, the solution cycle is fixed at 1 second, the initial window length is set to 10 cycles (i.e., 10 seconds), and three preset levels are set: "short window 5-8 seconds, standard window 9-12 seconds, and long window 13-15 seconds". The adjustment triggering requires simultaneous evaluation of two major quantitative indicators: orbit dynamics and observation data quality. For orbit dynamics, high dynamic states are determined by real-time calculated orbit curvature (threshold ≥ 0.001 rad / km) and velocity change rate (threshold ≥ 0.5 m / s²), triggering a short window. When both are ≤ 50% of the thresholds (i.e., orbit curvature ≤ 0.0005 rad / km, velocity change rate ≤ 0.25 m / s²), a stable state is determined, triggering a long window; otherwise, the standard window is maintained. For data quality, the GNSS geometrical precision factor (GDOP) and IMU corrected residuals are introduced as auxiliary indicators: if GDOP ≥ 5 or IMU corrected residual ≥ 0.15 m, the window is forcibly shortened to 8 seconds even if the orbit is dynamically stable; if GDOP ≤ 3 and IMU corrected residual ≤ 0.08 m, the window can be extended to 13 seconds even if the orbit is slightly dynamic. The window update adopts a "first-in, first-out" mechanism. One new data record is added every solution cycle, and the oldest expired data record is removed simultaneously. At the same time, a quality check is performed every 3 cycles. If two consecutive checks find that the proportion of abnormal data in the window is ≥10%, the window is temporarily shortened by 2 seconds and the data update frequency is increased to 0.8 seconds / time until the proportion of abnormal data is ≤5% and the normal configuration is restored.

[0033] For satellites without deployed GNSS receivers, a data request is initiated to neighboring GNSS satellites via a dedicated first inter-satellite link within the cluster. The request includes raw GNSS observation data from the neighboring GNSS satellites within a sliding window, as well as data confidence metrics characterizing data reliability (such as observation noise variance and signal-to-noise ratio). After receiving the data, the satellites without deployed GNSS receivers fuse their pre-processed IMU data with the acquired GNSS data to form a joint observation dataset of "inertial data + external GNSS data." This approach not only addresses the issue of some satellites lacking GNSS observation capabilities but also improves the completeness and reliability of the dataset through multi-source data fusion.

[0034] To accommodate the limited onboard computing resources while meeting the accuracy requirements of the initial solutions, a simplified orbital dynamics model is adopted as the core of the solution. This model retains key influencing factors such as Earth's central gravity and major orbital perturbations, while ignoring some minor perturbations to reduce computational complexity, achieving a balance between accuracy and computational efficiency. The model parameters are dynamically adapted based on satellite orbital altitude, operating region, and other characteristics. For example, atmospheric drag is appropriately enhanced in low-Earth orbit regions, while the influence of solar radiation pressure is emphasized in high-Earth orbit regions, ensuring the model's adaptability in different scenarios.

[0035] The simplified orbital dynamics model adopts a "layered calculation, progressive priority" logic. By identifying key forces and clarifying the calculation process, it ensures accuracy while adapting to onboard computing power. The calculation details and adaptation rules for each layer are as follows: The model as a whole follows the calculation order of "baseline first, then correction, then adaptation." The calculation priority of the three modules decreases sequentially, and the calculation of each layer needs to be based on the superposition and correction of the results of the previous layer. The first layer is the basic gravitational term (highest calculation priority), which is the core driving force of orbital motion. The calculation is based on the law of universal gravitation, and the core formula is: a1 = -GM·r / |r|³, where a1 is the gravitational acceleration vector, and GM is the Earth's gravitational constant (standard value 3.986004418×10¹). 4 m³ / s², pre-stored on-board and periodically updated via global calibration), r is the satellite's position vector relative to the Earth's center of mass (initial value from the previous cycle's calculation results, updated recursively in real time via IMU data); this layer's calculations are not simplified in any way to ensure the accuracy of the basic orbital shape. The second layer consists of the dominant perturbation terms (lower priority in calculation), retaining only the J2 term (oblateness perturbation) from the Earth's non-spherical perturbation. Because the Earth's equatorial radius is approximately 21 km longer than the poles, the satellite experiences an additional gravitational torque, which accounts for over 95% of all perturbation terms, far exceeding higher-order terms such as J3-J6 (each single-order influence percentage <1%); the formula for calculating the acceleration of the J2 term is: a² = (3GM·J²·R) / ( ... e ²) / (2|r| 5)·[ (5z² / |r|² - 1)r - 2z·(0,0,1) ], where J2 is the Earth's oblateness coefficient (standard value 1.08263 × 10 ], - ³, on-board curing), R e The Earth's equatorial radius is 6378137m (pre-stored parameter), z is the component of the satellite's position vector along the Earth's polar axis, and the orbital parameters (semi-major axis, eccentricity, etc.) are taken from the orbital results calculated using the fundamental gravitational terms. During calculation, the initial orbit is first obtained through the fundamental gravitational terms, and then a2 is superimposed to complete the perturbation correction. The third layer is the scene correction term (lowest calculation priority, dynamically activated), triggered by orbital altitude: low Earth orbit (≤2000km) activates atmospheric drag correction, using formula a. 31 = -0.5·ρ·v·C_d·S / m·v, where ρ is the atmospheric density (calculated using the pre-stored NRLMSISE-00 model on the satellite, combined with the satellite's real-time altitude and solar activity index), v is the satellite's velocity vector relative to the atmosphere, C_d is the drag coefficient (initial value is 1.0, adjusted ±0.1 every 10 solution cycles via GNSS observation residual feedback), S is the satellite's windward area (pre-stored design value), and m is the satellite's mass (fixed parameter); for medium-high orbit (>2000km), solar radiation pressure correction is activated using the cannonball model, formula a 32 = P·C_r·S / m·u, where P is the solar constant (1367W / m², pre-stored), C_r is the solar radiation pressure coefficient (0.95-1.05 based on satellite surface material, dynamically calibrated), and u is the unit vector of sunlight (calculated using star sensor attitude data); thrust correction is temporarily activated during orbital maneuvers, directly accessing the thrust acceleration data a collected by the IMU. 33 The entire model calculation process is as follows: first, obtain the basic orbital acceleration through a1, then superimpose a2 to complete the core correction, and finally superimpose a according to the scenario. 31 / a 32 / a 33 The final output total acceleration is used for orbital integration calculation; this process reduces the amount of calculation by more than 60% compared to the full perturbation model, while ensuring that the initial solution deviation is ≤0.3m through accurate calculation of key terms.

[0036] All satellites performed preliminary orbit calculations based on their respective datasets: satellites with GNSS deployment used preprocessed GNSS and IMU fused data, while satellites without GNSS deployment used a joint observation dataset. Orbit parameters were solved using a weighted least squares algorithm. During the calculation process, weights were dynamically assigned based on data confidence levels; data with high confidence (such as high-quality GNSS data and corrected IMU data) received higher weights to improve the reliability of the results. The final output included preliminary orbit calculation results for each satellite, including core parameters such as position, velocity, and clock bias. These results will serve as the foundational data for subsequent cluster head aggregation and fusion, supporting the construction of relative orbit parameters within the cluster.

[0037] In some embodiments, the cluster head satellite aggregates the preliminary calculation results of each satellite in the cluster to determine the relative orbital parameters within the cluster, including: fusing the states of the satellites in the cluster based on the preliminary calculation results of each satellite in the cluster and a weighted consensus algorithm to obtain state fusion data; wherein the weights in the fusion process are determined based on the confidence level of the preliminary calculation results of each satellite; and using the state of the cluster head satellite as a reference, calculating the state deviation of each satellite in the cluster relative to the cluster head satellite based on the state fusion data to construct the relative orbital parameters within the cluster.

[0038] In this embodiment of the invention, the cluster head satellite receives preliminary solution results uploaded by all satellites in the cluster in batches via the first inter-satellite link within the cluster. Each result contains core parameters: position vector (x, y, z) and velocity vector (v). x , v The data includes the clock error value t (v_z), the clock error value t, and the corresponding residuals res (i.e., the fitting deviation between the solution result and the observed data). After reception, a two-layer verification is performed: the first layer is integrity verification, checking whether the parameter fields of each satellite are complete, and results missing key parameters (such as position or residuals) are directly marked as invalid; the second layer is validity screening, using the 3σ criterion (σ is the standard deviation of the solution residuals of all satellites), eliminating abnormal results with residuals res > 3σ, usually with the abnormal elimination ratio controlled within 3%. For results marked as invalid or abnormal, the cluster head satellite sends a retransmission request to the corresponding satellite through the inter-satellite link, requesting the upload of complete and reliable preliminary solution results to ensure the quality of input data for subsequent fusion calculations.

[0039] The cluster head satellite takes the filtered, valid preliminary solution results as input and initiates a weighted consensus algorithm to perform state fusion. The core objective is to converge the state estimates of all satellites within the cluster to a unified intermediate benchmark value. The algorithm's iterative process is clear: First, the cluster head initializes its own state as a temporary benchmark (taken from its own preliminary solution results; if its own results are abnormal, the average of the valid results within the cluster is used). Second, in each iteration, the cluster head collects the current state estimates of each satellite within the cluster, combines them with preset satellite weights, and updates its own state using a weighted average formula: X_head_new = Σ(w_i ×X_i) / Σw_i, where X_head_new is the updated state of the cluster head (including position, velocity, and clock bias), w_i is the weight of the i-th satellite, and X_i is the weight of the i-th satellite. The first step is to update the current state of each satellite. The second step is to update the state of each satellite. The third step is to update the state of each satellite synchronously using the same weighted average formula. The fourth step is to repeat the second and third steps until the state changes of all satellites in the cluster meet the convergence threshold (position change ≤ 0.05m, velocity change ≤ 0.01m / s, clock difference change ≤ 1ns) in two consecutive iterations. At this point, the final state of the cluster head is the unified state fusion data of the cluster.

[0040] The satellite weights in the fusion process are determined by the confidence level of the preliminary solution results for each satellite. Both the confidence level assessment and weight calculation employ explicit mathematical logic and differentiate between satellite types and appropriate indicators: For satellites with deployed GNSS receivers, the confidence level is determined by two core indicators: the geometrical precision factor (GDOP) of the GNSS observation data and the preliminary solution residuals res. The smaller the GDOP value and the smaller the residuals res, the higher the confidence level. The weight calculation formula is w_i = (1 / GDOP_i)× (1 / res_i) / Σ[(1 / GDOP_j) × (1 / res_j)] (j traverses all valid GNSS satellites within the cluster). Normalization ensures that the sum of all satellite weights is 1. For satellites without deployed GNSS receivers, the confidence level depends on two types of indirect indicators. The average confidence level of the acquired neighbor GNSS data (indirectly represented by the neighbor GDOP value, denoted as neighbor_GDOP_avg) and the corrected residual imu_res of the IMU data are weighted by the formula w_k = (1 / neighbor_GDOP_avg) × (1 / imu_res_k) / Σ[(1 / neighbor_GDOP_avg_j) × (1 / imu_res_j)] (j traverses all valid non-GNSS satellites in the cluster). This ensures that the weight allocation is strongly bound to the data reliability and avoids low-quality data from affecting the fusion results.

[0041] After state fusion is completed, using the state fusion data of the cluster head satellite as the absolute reference, the state deviation of each satellite in the cluster is calculated one by one, ultimately constructing complete relative orbital parameters within the cluster. The deviation calculation covers three core dimensions, all using vector subtraction logic: relative position deviation (Δx, Δy, Δz) = state fusion position vector of the i-th satellite - state fusion position vector of the cluster head; relative velocity deviation (Δv) = state fusion position vector of the i-th satellite - state fusion position vector of the cluster head satellite; relative velocity deviation (Δv) = state fusion position vector of the i-th satellite - state fusion position vector of the cluster head satellite. x , Δv Δv_z) = State fusion velocity vector of the i-th satellite - State fusion velocity vector of the cluster head; Relative clock error Δt = State fusion clock error of the i-th satellite - State fusion clock error of the cluster head. After calculation, the "cluster head state fusion data (as the absolute reference within the cluster) + the three-dimensional position deviation, three-dimensional velocity deviation, and relative clock error of all satellites" are packaged and bound to form complete intra-cluster relative orbit parameters. Finally, a consistency check is performed: the root mean square (RMS) of all satellite deviation values ​​is calculated. If RMS ≤ 0.2m, the parameters are deemed valid and sent to all satellites within the cluster via the first inter-satellite link; if RMS > 0.2m, 2-3 rounds of weighted consistency iterations are added and the deviations are recalculated until the accuracy requirements are met, ensuring that the relative orbit parameters can support subsequent regional collaborative solutions.

[0042] In some embodiments, the backbone nodes collect the relative orbit parameters of each cluster and calculate the regional absolute orbit reference, including: calculating their own absolute orbit parameters based on their own GNSS and IMU data; calculating the transformation parameters between the coordinate system of each cluster and the coordinate system of the backbone nodes according to the relative orbit parameters of each cluster and their own absolute orbit parameters, and performing coordinate transformation on the relative orbit parameters of each cluster; obtaining inter-satellite ranging data with adjacent backbone nodes as constraints, and jointly calculating the relative orbit parameters of each cluster after coordinate transformation to determine the regional absolute orbit parameters; evaluating the residuals of the optimized regional absolute orbit parameters, and if they meet the accuracy requirements, determining them as the regional absolute orbit reference and distributing them to each cluster; otherwise, obtaining inter-satellite ranging data from other backbone nodes as constraints and recalculating the regional absolute orbit parameters.

[0043] In this embodiment of the invention, the backbone node, as the anchoring core for regional orbit determination, first independently calculates high-precision absolute orbit parameters based on its onboard multi-mode GNSS receiver and IMU device. The GNSS receiver synchronously acquires observation data (including pseudorange and carrier phase data) from multiple navigation systems, while the IMU device continuously outputs angular acceleration and linear acceleration data. Both are synchronized through timestamp calibration. The calculation employs a precise orbital dynamics model, which includes the Earth's central gravitational term and the Earth's non-spherical perturbation J2 term. Simultaneously, atmospheric drag or solar radiation pressure correction terms are dynamically activated based on orbital altitude. The algorithm combines weighted least squares and Kalman filtering. Initial orbit parameters are first solved using weighted least squares, and then Kalman filtering is used to fuse IMU data in real time to optimize the calculation results. Finally, the absolute position vector (x, y, z) and absolute velocity vector (v) of the backbone node itself are output. x , v The positional accuracy is controlled at the centimeter level, using the values ​​of v_z and clock difference to provide reliable anchor points for subsequent regional benchmark construction.

[0044] Backbone nodes receive cluster relative orbit parameters uploaded by cluster head satellites from 3-5 orbital plane clusters under their jurisdiction via a dedicated second inter-satellite link within the region. Each set of parameters includes the cluster head satellite's intra-cluster reference state (relative position, velocity), the state deviations of all satellites within the cluster relative to the cluster head, and parameter consistency residuals. After reception, the backbone nodes perform two layers of verification: the first layer is integrity verification, checking whether the parameters contain key fields such as cluster head reference and satellite deviations; parameters missing fields are directly marked as invalid. The second layer is validity screening, removing parameters with consistency residuals exceeding a preset threshold (usually set to 0.2m) to prevent low-quality data from interfering with regional calculations. For invalid or unqualified parameters, the backbone nodes send retransmission requests to the corresponding cluster head until valid relative orbit parameters for all clusters are collected.

[0045] The backbone nodes use their own calculated absolute orbit parameters as a reference to calculate the transformation parameters between the local coordinate system of each cluster and the absolute coordinate system of the backbone nodes, achieving a unified transformation from intra-cluster relative parameters to the absolute coordinate system. The transformation employs the Helmert transformation model, with three core parameters: rotation angles (Δα, Δβ, Δγ), translations (ΔX, ΔY, ΔZ), and scaling factor k. The calculation logic uses the cluster head satellite as a bridge: extracting the intra-cluster reference state of the cluster head from the relative orbit parameters of each cluster, and combining this with the absolute orbit state of the backbone nodes at that moment, the transformation parameters are solved through least-squares fitting. The rotation angle is used to correct the attitude deviation between the intra-cluster coordinate system and the absolute coordinate system, the translation compensates for the origin offset between the two coordinate systems, and the scaling factor corrects the scale error of intra-cluster relative measurements. After the parameters are solved, the relative state deviations of all satellites within each cluster are substituted into the transformation formula to complete the transformation from the intra-cluster local coordinate system to the backbone node absolute coordinate system, obtaining the preliminary absolute orbit values ​​for each satellite.

[0046] To improve the reliability of regional absolute orbit parameters, backbone nodes acquire two-way inter-satellite ranging data with adjacent backbone nodes via a second inter-satellite link. This data, after preprocessing, is used as a strong constraint in the joint solution. Ranging data preprocessing includes carrier phase smoothing (improving accuracy to the centimeter level), 3σ criterion outlier removal (removal rate ≤3%), and calculation of the root mean square error (RMS) of the ranging data to characterize data confidence. The joint solution is based on the preliminary absolute orbit values ​​after coordinate transformation of each cluster, using the inter-satellite ranging data from adjacent backbone nodes as constraints to construct a global optimization model. In the model, the weights of each cluster's data are allocated according to its consistency residual (the smaller the residual, the higher the weight), while the weights of the inter-satellite ranging data are dynamically adjusted according to the RMS (higher weight is used when RMS ≤ 0.03m). Through multiple rounds of iterative solving, the satellite orbit information and ranging constraints of all clusters are integrated to eliminate inter-cluster solution bias, ultimately outputting unified absolute orbit parameters covering the entire region.

[0047] After the joint calculation is completed, the backbone nodes assess the accuracy of the regional absolute orbit parameters by calculating the residuals. The core assessment indicators are the orbit fitting residual and the ranging constraint residual. The orbit fitting residual is the root mean square deviation between the preliminary absolute orbit values ​​of each cluster of satellites after conversion and the joint calculation result. The ranging constraint residual is the root mean square deviation between the measured inter-satellite ranging values ​​and the theoretical ranging values ​​calculated based on the regional absolute orbit parameters. If both residuals are ≤ preset thresholds (typically 0.15m for the fitting residual and 0.05m for the ranging residual), the parameters are deemed to meet the accuracy requirements and are determined as the regional absolute orbit reference. Subsequently, the data is transmitted to the cluster head satellites of each cluster via the second inter-satellite link, providing an absolute reference for the final orbit calculation of satellites within the cluster. If the residuals do not meet the requirements, the backbone nodes will expand the inter-satellite ranging constraint range, acquire more ranging data from adjacent backbone nodes, and re-execute the joint calculation process until the residuals meet the accuracy requirements, ensuring the reliability and consistency of the regional absolute orbit reference.

[0048] In some embodiments, the global anchor satellite aggregates reference values ​​from various regions to determine the unified orbit parameters of the constellation, including: the global anchor satellite calculates the absolute orbit parameters of the anchor point based on its own multi-mode GNSS and laser ranging data; using its own anchoring reference as a reference, it calculates the systematic deviation of the reference values ​​of each region according to the absolute orbit parameters of the anchor point and the deviation correction model, determines the regional correction coefficient, and corrects the regional reference data according to the regional correction coefficient; under the inter-satellite ranging constraints between anchor points, the unified orbit parameters of the constellation are obtained based on the deviation-corrected regional reference data.

[0049] In this embodiment of the invention, the global anchor satellite serves as the highest-level reference for constellation orbit determination. Relying on its onboard multi-mode GNSS receiver and spaceborne laser ranging equipment, it calculates high-precision absolute orbital parameters for the anchor satellite. The multi-mode GNSS receiver simultaneously acquires pseudorange and carrier phase observation data from multiple navigation systems, providing a continuous orbital observation foundation. The spaceborne laser ranging equipment obtains centimeter-level or even millimeter-level distance observations through Earth- or Moon-level laser ranging, used to correct systematic errors in GNSS observations. The calculation employs a precise orbital dynamics model considering complex perturbation effects, with core components including Earth's central gravitational term, Earth's non-spherical perturbation J2-J6 term, solar radiation pressure correction term, atmospheric drag correction term (for low-Earth orbit anchors), and general relativistic effect correction term. The algorithm uses extended Kalman filtering to weight and fuse GNSS observation data and laser ranging data according to confidence level, optimizing orbital parameters in real time. The final output is the absolute position vector, absolute velocity vector, and clock error value of the anchor satellite, with position accuracy controlled within centimeter levels, providing a stable and reliable anchoring reference for global orbit unification.

[0050] The global anchor satellite receives absolute orbital reference values ​​for all regions of the entire constellation in batches via dedicated periodic communication links with regional backbone nodes. Each regional reference includes the absolute orbital parameters of all satellites within the region, regional solution residuals, and reference confidence indices. After reception, the anchor satellite performs a rigorous two-layer verification process: the first layer is integrity verification, checking each regional reference to ensure it contains key fields such as satellite orbital parameters and solution residuals; references lacking key information are directly marked as invalid. The second layer is validity screening, using the 3σ criterion to eliminate low-quality references with regional solution residuals exceeding a threshold (typically set at 0.2m), while simultaneously verifying the reference confidence index, retaining only valid references with a confidence level ≥ 95%. For invalid or low-quality regional references, the anchor satellite sends retransmission requests to the corresponding regional backbone nodes until valid reference data for all regions of the entire constellation is collected, ensuring the input quality of the global solution.

[0051] The anchor satellite uses its own precisely calculated absolute orbit parameters as the global reference benchmark. A deviation correction model is used to calculate the systematic deviation of the benchmark values ​​for each region, generating regional correction coefficients and completing data correction. The systematic deviation mainly includes three types: translational deviation (coordinate system origin offset), rotational deviation (coordinate system attitude inconsistency), and scale deviation (measurement scale error) between the regional and global benchmarks. The deviation calculation employs a least-squares fitting logic: the absolute orbit parameters of the backbone nodes in each regional benchmark are extracted and paired with the absolute orbit parameters of the anchor satellite at the same time. These are then substituted into the deviation correction model, and the three types of deviation components are solved through iterative fitting, transforming them into corresponding regional correction coefficients (translation correction, rotation correction angle, and scale correction factor). Subsequently, the absolute orbit parameters of all satellites within each region are substituted into the correction formula, and the regional systematic deviation is eliminated through the correction coefficients. This unifies the benchmark data of all regions to the global coordinate system of the anchor satellite, laying a unified data foundation for global joint calculation.

[0052] To achieve global consistency of orbital parameters across the entire constellation, anchor satellites introduce inter-anchor ranging constraints, which are then jointly used with corrected regional reference data to calculate unified orbital parameters for the constellation. First, global anchor satellites acquire two-way inter-anchor ranging data via high-frequency inter-satellite links (communication frequency above 5Hz). This data is then smoothed by carrier phase to improve ranging accuracy to the centimeter level. Outliers are removed using the 3σ criterion and introduced as a strong constraint in the calculation. Subsequently, a global joint optimization model is constructed: using the corrected absolute orbital parameters of each regional satellite as the base observations, and the inter-anchor ranging data as the strong constraint, weights are dynamically allocated based on data confidence levels. Regional reference data weights are assigned according to their solution residuals (smaller residuals result in higher weights), while inter-anchor ranging data are given high weights (typically ≥0.8). Through multiple rounds of iterative weighted least squares calculation, the orbital information of all satellites in the entire constellation is integrated with the inter-anchor constraints, eliminating residual biases between regions and bringing the global orbital parameters to the optimal solution. After the final solution is verified for global consistency (the root mean square deviation of all satellite orbit parameters from the global solution is ≤0.1m), it is determined as the unified orbit parameters for the constellation, providing a unified global benchmark for the final orbit calculation of all satellites in the constellation.

[0053] In some embodiments, the orbit calculation results of each satellite are determined based on the regional absolute orbit reference and the constellation unified orbit parameters. This includes: the cluster head satellite receiving the constellation unified orbit parameters and regional correction coefficients forwarded by the regional backbone nodes, and aligning them with its own stored regional absolute orbit reference in terms of time and coordinate system; the cluster head satellite using the constellation unified orbit parameters as a reference, combined with the regional absolute orbit reference, correcting the regional reference deviation through a weighted fusion algorithm, and outputting the corrected regional absolute orbit reference; the cluster head satellite calculating the preliminary absolute orbit calculation results for each satellite in the cluster based on the corrected regional absolute orbit reference and the relative orbit parameters within the cluster; and the cluster head satellite collecting inter-satellite ranging data between satellites within the cluster as local constraints, performing joint calculations with the preliminary absolute orbit results to obtain the orbit calculation results.

[0054] In this embodiment of the invention, the cluster head satellite receives constellation-wide unified orbit parameters (including global coordinate system definition, unified orbit reference value, and solution residuals) and regional correction coefficients forwarded by the regional backbone nodes via the first inter-satellite link, while simultaneously retrieving its own stored regional absolute orbit reference. Regarding time alignment, using the timestamp of the constellation-wide unified orbit parameters as a reference, the time deviation of the regional absolute orbit reference is corrected using an onboard clock calibration algorithm, and linear interpolation is used to compensate for the time difference between different data acquisition times, ensuring that the time synchronization error between the two is ≤10ms. Regarding coordinate system alignment, based on the clearly defined global coordinate system (such as the geocentric coordinate system) of the constellation-wide unified orbit parameters, the local coordinate system of the regional absolute orbit reference is mapped to the global coordinate system through a coordinate transformation matrix, eliminating system deviations caused by differences in coordinate system definitions and laying a foundation for spatiotemporal consistency for subsequent fusion and correction.

[0055] The cluster head satellite uses the unified orbital parameters of the constellation as the global optimal benchmark, combined with the regional absolute orbital benchmark, and corrects the deviation between the regional benchmark and the global benchmark through a weighted fusion algorithm. The weight allocation adopts a quantization logic: the weight of the unified orbital parameters of the constellation is determined based on its global solution residuals; the weight is 0.8-0.9 when the residual is ≤0.05m, 0.6-0.7 when the residual is >0.05m and ≤0.1m, and 0.4-0.5 when the residual is >0.1m. The weight of the regional absolute orbital benchmark is allocated based on its own solution residuals; the weight is 0.5-0.6 when the residual is ≤0.1m, 0.3-0.4 when the residual is >0.1m and ≤0.2m, and 0.1-0.2 when the residual is >0.2m. The fusion calculation follows the logic of "weighted summation of reference values ​​+ deviation correction". The core formula is: corrected regional reference = (constellation unified parameter × global weight + regional reference × regional weight) / (global weight + regional weight). At the same time, the regional correction coefficient is superimposed to compensate for systematic deviations. Finally, the corrected regional absolute orbit reference is output, ensuring that its deviation from the global parameters is ≤3cm.

[0056] Based on a modified regional absolute orbit reference, the cluster-head satellite first extracts its own absolute orbital core parameters under that reference, including the absolute position vector (x0, y0, z0) and the absolute velocity vector (v). x0 , v The parameters are: 0, v_z0) and absolute clock difference t0. Then, the previously constructed intra-cluster relative orbital parameters are retrieved, namely, the three-dimensional position deviation (Δx, Δy, Δz) and three-dimensional velocity deviation (Δv) of each satellite relative to the cluster head. x , Δv The parameters are Δv_z and relative clock bias Δt. Through a vector superposition calculation logic of "reference parameters + relative bias", the preliminary absolute orbit calculation results for each satellite are obtained: Absolute position = cluster head absolute position + relative position bias, Absolute velocity = cluster head absolute velocity + relative velocity bias, Absolute clock bias = cluster head absolute clock bias + relative clock bias. During the calculation process, the validity of the bias data is simultaneously verified. If the relative bias of a satellite exceeds a preset threshold (position bias > 0.5m, velocity bias > 0.08m / s), the bias value is discarded and replaced with the average bias of adjacent satellites within the cluster to ensure the reasonableness of the preliminary results.

[0057] To strengthen the local constraints of the solution results, the cluster head satellite collects two-way inter-satellite ranging data between all satellites within the cluster through the first inter-satellite link (communication frequency 10Hz). After the collection is completed, multiple steps of preprocessing are performed: in the first step, the carrier phase smoothing technique is adopted to smooth the ranging value through the phase observation data of multiple consecutive cycles, and the ranging accuracy is improved to the centimeter level (≤3cm); in the second step, the 3σ criterion is used to identify and eliminate ranging outliers, calculate the root mean square error (RMS) of all ranging data, and eliminate the abnormal data with RMS>0.05m to ensure that the proportion of valid data is ≥97%; in the third step, timestamp matching is performed to correspond the preprocessed ranging data with the timestamps of the preliminary absolute orbit solution results one by one, eliminate the deviation caused by time asynchrony, and form a high-quality local constraint data set.

[0058] The cluster head satellite constructs a weighted least squares joint solution model, takes the preprocessed inter-satellite ranging data as local strong constraints, and fuses and optimizes it with the preliminary absolute orbit solution results. The weight assignment in the model is strongly bound to the data quality: the weight of the inter-satellite ranging data is dynamically adjusted according to its RMS. When RMS≤0.03m, the weight is taken as 0.7-0.8; when 0.03m<RMS≤0.05m, the weight is taken as 0.5-0.6; the weight of the preliminary absolute orbit solution results is assigned according to the solution residuals of the corrected regional benchmark. When the residual≤0.1m, the weight is taken as 0.6-0.7; when the residual>0.1m and ≤0.15m, the weight is taken as 0.4-0.5. The solution process adopts an iterative solution method. After each round of iteration, the model fitting residual is calculated. If the fitting residuals of two consecutive rounds of iteration are both ≤0.1m, it is determined that the solution converges, the iteration is stopped and the optimized orbit parameters are output; if the fitting residual is still >0.1m after 5 iterations, the inter-satellite ranging data is re-collected (doubling the sampling volume) and the weight of the ranging data is increased to 0.9, and the solution process is executed again.

[0059] After the joint calculation is completed, the cluster head satellite determines the final orbit calculation result through dual verification: the first is residual verification, which calculates the fitting residual between the optimized orbit parameters and the inter-satellite ranging data, as well as the deviation from the corresponding position of the unified orbit parameters of the constellation. If the fitting residual is ≤0.1m and the deviation from the global parameters is ≤0.15m, the result is considered qualified. The second is consistency verification, which calculates the root mean square error (RMS) of the final orbit parameters of all satellites in the cluster. If the RMS is ≤0.12m, it indicates that the consistency of the results within the cluster meets the standard. After the dual verification is qualified, the final orbit calculation result of each satellite (including absolute position, absolute velocity, absolute clock error and calculation confidence) is sent to the corresponding satellite through the first inter-satellite link. The satellite stores the results and uses them for subsequent orbit prediction and attitude control. If the verification fails, GNSS / IMU observation data for 1-2 calculation cycles are supplemented, and the preliminary calculation and joint calculation process is re-executed until the result meets the accuracy requirements, ensuring the reliability and global consistency of the orbit calculation results of all satellites in the cluster.

[0060] In some embodiments, the cluster head satellite collects inter-satellite ranging data among satellites within the cluster as a local constraint, and performs joint calculation with the preliminary absolute orbit results to obtain the orbit calculation result. This includes: retrieving the preliminary absolute orbit results of each satellite within the cluster and performing timestamp calibration; constructing a weighted least squares joint calculation model based on the effective inter-satellite ranging data as a constraint and the preliminary absolute orbit results, and dynamically allocating weights according to data quality; and determining the orbit calculation result based on the joint calculation model.

[0061] In this embodiment of the invention, the cluster head satellite first retrieves the preliminary absolute orbit calculation results of all satellites within the cluster from the local storage module. These results include the core parameters of each satellite: absolute position vector (x, y, z), absolute velocity vector (v...). x , v The data includes the initial orbit data (v_z), absolute clock difference t, and corresponding residuals res (characterizing the reliability of the preliminary results). To ensure spatiotemporal synchronization with subsequent inter-satellite ranging data, a timestamp calibration process is initiated: using the unified sampling timestamp of the inter-satellite ranging data within the cluster as a reference (usually the satellite system time, with microsecond-level accuracy), the time deviation of the preliminary orbit results is corrected using an onboard clock synchronization algorithm. For satellite data with large timestamp deviations (>10ms), linear interpolation is used to compensate for changes in orbital parameters at different times. For example, the position correction amount at the time of deviation is calculated based on the satellite velocity vector. Ultimately, the time synchronization error between the preliminary orbit results of all satellites and the ranging data is ≤5ms, eliminating spatiotemporal deviation interference for joint calculation.

[0062] The cluster head satellite collects bidirectional inter-satellite ranging data in batches from all satellites within the cluster via a dedicated first inter-satellite link (communication frequency ≥10Hz to ensure data timeliness). This means each pair of adjacent satellites sends ranging signals to each other, forming a "two-way observation" to offset some system errors. After acquisition, a three-stage preprocessing is performed: the first stage is carrier phase smoothing, which smooths the raw ranging values ​​using carrier phase data from 5-8 consecutive sampling periods, improving ranging accuracy from decimeters to centimeters (≤3cm); the second stage is outlier removal, using the 3σ criterion (σ is the standard deviation of the ranging data) to identify and delete outliers exceeding the range of [mean - 3σ, mean + 3σ], while simultaneously calculating the root mean square error (RMS) of the remaining data; the third stage is validity screening, retaining only ranging data with RMS ≤ 0.05m as valid constraints, ensuring that the percentage of valid data is ≥97%, and avoiding low-quality data from affecting the calculation accuracy.

[0063] After completing the initial result calibration and ranging data preprocessing, the cluster head satellite performs a data spatiotemporal matching process: from the calibrated initial orbit results, satellite position parameters that completely correspond to the timestamp of each valid ranging data are extracted, forming a one-to-one correspondence dataset of "satellite pair position - measured ranging value". Subsequently, an inter-satellite ranging observation model is constructed. The core logic is: for any satellite pair (satellite A, satellite B), its theoretical ranging value r_AB can be calculated using the distance formula between two points, i.e., r_AB = √[(x_A - x_B)² + (y_A - y_B)² + (z_A - z_B)²], where (x_A, y_A, z_A) and (x_B, y_B, z_B) are the initial absolute orbital positions of satellites A and B, respectively; the residual of the observation model is defined as the difference between the measured ranging value and the theoretical ranging value, i.e., v = r_measured - r_theoretical. The goal of the joint solution is to minimize the residual by correcting the satellite position parameters.

[0064] Based on the spatiotemporally matched dataset, a weighted least squares joint solution model is constructed for the cluster head satellites. The core of the model is "minimizing the weighted sum of squared residuals". The model variables are the position corrections (Δx_i, Δy_i, Δz_i) of each satellite within the cluster. That is, the solution results are optimized by correcting the initial orbital position parameters (x_i' = x_i + Δx_i, y_i' = y_i + Δy_i, z_i' = z_i + Δz_i). To reflect the reliability differences of different data, a dynamic weight matrix is ​​introduced into the model: the weights are negatively correlated with data quality, that is, the higher the data quality, the larger the weight, ensuring that high-quality data plays a dominant role in the solution results. The mathematical expression of the model is: min Σ(w_ij × v_ij²), where w_ij is the weight of the satellite for the ranging data at (i,j), and v_ij is the corresponding residual. By differentiating the objective function and setting the derivative to zero, a system of linear equations for the position corrections can be obtained, providing a mathematical basis for subsequent solutions.

[0065] The weight allocation employs a "type-based, threshold-based" quantization logic to ensure a strong correlation between weights and data reliability: For preliminary absolute orbit results, the weight w_orbit is allocated based on the calculated residual res. When res≤0.1m, w_orbit=0.6-0.7; when 0.1m<res≤0.2m, w_orbit=0.4-0.5; when res>0.2m, w_orbit=0.2-0.3. For inter-satellite ranging data, the weight w_measurement is allocated based on RMS. When RMS≤0.03m, w_measurement=0.7-0.8; when 0.03m<RMS≤0.05m, w_measurement=0.5-0.6; when RMS>0.05m, the data is directly discarded. After the weights are determined, they are input into a system of linear equations, and the position correction is solved using the Cholesky decomposition method to obtain the optimized satellite position parameters. Then, iterative calculation is performed: the optimized position parameters are substituted into the observation model to recalculate the residuals. If the mean residual is ≤0.1m and the change in residuals between two consecutive iterations is ≤0.01m, the calculation is considered to have converged. If it has not converged, the weight allocation is adjusted (e.g., the weight of the ranging data is increased by 10%-20%) and the solution is repeated. The number of iterations is controlled at 3-5 times to balance accuracy and computing power.

[0066] After iterative convergence, the cluster head satellite determines the final orbit calculation result through dual verification: the first is residual verification, which calculates the fitting residual between the theoretical ranging value and the measured value corresponding to the optimized orbit parameters. If the fitting residual is ≤0.1m and the position deviation between the optimized orbit and the initial orbit is ≤0.3m, the calculation accuracy is considered to meet the standard. The second is consistency verification, which calculates the root mean square error (RMS) of the optimized orbit parameters of all satellites in the cluster. If the RMS is ≤0.12m, it indicates that the orbit results within the cluster are consistent. After the dual verification is qualified, the final orbit calculation result of each satellite (including the optimized absolute position, velocity, clock error and calculation confidence) is sent to the corresponding satellite through the first inter-satellite link. The satellite stores the results for subsequent orbit prediction and attitude control. If the verification fails, inter-satellite ranging data for 1-2 cycles is collected again, the weight allocation is optimized, and the calculation process is executed again until the result meets the accuracy requirements, ensuring the reliability and consistency of the orbit calculation of satellites within the cluster.

[0067] In some embodiments, the method further includes: real-time monitoring of satellite status within the cluster, triggering cluster reconstruction when preset conditions are met; updating the inter-satellite link topology based on the reconstructed cluster structure and LSTM orbit prediction model; and dynamically adjusting the communication frequency of each level of inter-satellite link based on the updated inter-satellite link topology.

[0068] In this embodiment of the invention, the cluster head satellite collects and monitors the core status of all satellites in the cluster in real time through the inter-satellite link within the cluster, including key indicators such as satellite operating status (normal / failed), orbital position and spatial distribution, and calculation accuracy (calculation residual). Three types of reconstruction trigger conditions are preset, and cluster reconstruction is initiated when any one of the conditions is met: First, the proportion of failed satellites in the cluster is ≥15% (or the number of failed satellites is ≥2), resulting in insufficient coverage of observation data within the cluster; second, satellites move across orbital planes or perform orbital maneuvers, causing the maximum spatial distance between satellites in the cluster to exceed 500km, which is beyond the effective communication range of the inter-satellite link; third, the calculation residual of the cluster head satellite is >0.3m for 3 consecutive cycles, or the relative orbital parameter consistency residual within the cluster continues to exceed the standard, making it impossible to guarantee the reliability of the calculation. After the reconstruction is triggered, a "pre-election + fast switching" strategy is adopted: the satellites with the best data quality (such as the highest GNSS observation confidence) and computing power redundancy ≥40% are selected as candidate cluster heads from within the cluster. The new cluster head is determined by voting among the satellites in the cluster (each satellite casts 1 vote based on the neighbor status, and the satellite with a vote rate ≥60% is elected). The reconstruction time is controlled within 10 seconds. During this period, data transmission is maintained through backup inter-satellite links to avoid orbit determination interruption.

[0069] After cluster reconstruction, the inter-satellite link topology is dynamically updated using an LSTM orbit prediction model, based on the new cluster structure (satellite composition and spatial distribution). The LSTM model is trained using historical satellite orbital parameters (position and velocity), outputting predicted future satellite positions 30 minutes in advance, with a prediction error ≤1km within 10 minutes, providing accurate data for link planning. Topology updates follow the principles of "distance priority, reliability priority": the first inter-satellite link within a cluster prioritizes connecting adjacent satellites with an orbital distance ≤300km, ensuring data transmission latency ≤50ms; the second inter-satellite link within a region prioritizes connecting backbone nodes with a residual error ≤0.1m, forming a mesh topology to reduce communication hops (≤2 hops) and improve regional data transmission efficiency; communication links between global anchors and backbone nodes prioritize nodes with an inter-satellite distance ≤5000km to avoid signal attenuation and latency caused by long-distance links. The topology update cycle is 5 minutes, synchronously adapting to dynamic changes in satellite orbits and cluster structure adjustments to ensure link connection stability.

[0070] Based on the updated inter-satellite link topology (link connection relationships, transmission distance, and load status), and combined with the real-time status of orbit determination calculations, the communication frequency of each level of inter-satellite link is dynamically adjusted. The adjustment logic is strongly tied to data transmission requirements and link load: when orbit determination calculations are in a stable state (cluster relative orbit parameter consistency residual ≤ 0.1m, regional absolute orbit reference residual ≤ 0.15m), and link load ≤ 70%, the communication frequency is reduced. The first inter-satellite link is reduced to 5Hz, the second inter-satellite link to 10Hz, and the communication frequency between anchor points and backbone nodes to 0.5Hz to reduce onboard power consumption; when in a dynamic state (new satellite joining the network within 1 hour, sudden changes in satellite status, calculation residuals > 0.2m for two consecutive cycles), or when link load ≥ 85%, the communication frequency is increased. The first inter-satellite link frequency was increased to 15Hz, the second inter-satellite link frequency was increased to 25Hz, and the communication frequency between the anchor point and the backbone node was increased to 2Hz to ensure real-time data transmission. During the transition period, a linear adjustment strategy was adopted to avoid link congestion caused by frequency changes, and the inter-satellite link resource utilization rate was maintained at 80%-90%, reducing power consumption by 30%-40% compared with the fixed frequency mode.

[0071] Cluster status monitoring, cluster reconstruction, topology updates, and communication frequency adjustments form a collaborative optimization closed loop, ensuring the system adapts to dynamic satellite changes and orbit determination requirements. Cluster status monitoring provides the trigger for reconstruction, responding promptly to anomalies such as satellite failures and orbital deviations; cluster reconstruction ensures the integrity of the cluster structure and the reliability of the solution, laying the foundation for topology updates; LSTM model-driven topology updates adapt link connections to the new cluster structure and satellite orbit changes, reducing invalid link occupation; dynamic communication frequency adjustments achieve efficient resource allocation and power consumption optimization based on topology load and solution status. The entire closed-loop process iterates every 5-10 minutes, ensuring inter-satellite link connectivity and low latency, reducing system energy consumption through dynamic adjustments, and guaranteeing the real-time performance and accuracy stability of distributed orbit determination, giving the system strong adaptability and robustness.

[0072] In some embodiments, the method further includes: when a satellite failure is detected, calculating alternative orbital parameters for the failed satellite based on its historical orbital data and information on neighboring satellites.

[0073] In this embodiment of the invention, the cluster head satellite monitors the working status and data transmission of each satellite in the cluster in real time through the first inter-satellite link within the cluster. Monitoring indicators include the frequency of satellite orbit calculation result uploads, data integrity, and communication link connectivity. If no valid data feedback is received from a satellite for three consecutive calculation cycles (or a preset time threshold), and there is still no response after two retransmission requests, it is initially determined that the satellite may be faulty. Further analysis of the inter-satellite ranging status with the faulty satellite (if multiple neighboring satellites cannot obtain valid ranging data from the faulty satellite) confirms the satellite's failure. The failure determination process strictly follows the principle of "multi-indicator verification and multi-satellite cross-confirmation" to avoid misjudgments due to non-failure factors such as temporary communication interruptions and data transmission delays, ensuring the accuracy of the failure status determination.

[0074] Upon confirmation of satellite failure, the cluster head satellite immediately retrieves historical orbital data from its local storage module. It prioritizes high-quality orbital parameters (including absolute position, velocity, clock bias, and corresponding residuals) from the 12 hours prior to failure, ensuring that the residuals of the filtered data are all ≤0.2m and that the data sampling intervals are uniform (≤1 minute) to fully reflect the orbital trend of the failed satellite. Simultaneously, it collects real-time orbital parameters (absolute position, velocity) and historical inter-satellite ranging data (effective ranging values ​​over the past 10 cycles, accuracy ≤3cm) from 3-5 neighboring satellites with an orbital distance ≤300km from the failed satellite via inter-satellite links. After collection, the historical ranging data is preprocessed to remove outliers and calculate the mean and deviation of the ranging measurements, forming a complete input dataset consisting of "historical orbital trend + neighboring satellite correlation data".

[0075] Based on collected historical orbit data and information from neighboring satellites, a fusion calculation method combining "orbit trend prediction + neighbor constraint correction" is employed to solve for the alternative orbit parameters of the failed satellite. First, using the historical orbit data of the failed satellite, a simplified orbital dynamics model (retaining the Earth's central gravitational term and J2 perturbation term) combined with a linear prediction algorithm is used to calculate the preliminary orbit prediction value for the next 24 hours. This prediction value primarily reflects the satellite's natural orbital trend. Then, using the real-time orbit parameters of neighboring satellites as a benchmark, and combining historical inter-satellite ranging data before failure, the relative orbital relationship (relative position deviation, relative velocity deviation) between the failed satellite and neighboring satellites is fitted, and the preliminary orbit prediction value is constrained and corrected based on this relative relationship. The relative deviation between the predicted orbit and the real-time orbits of neighboring satellites is minimized using a least-squares algorithm, eliminating the accumulated error of the prediction model. Finally, the alternative orbit parameters for the failed satellite for the next 24 hours are output, including the absolute position, velocity, and corresponding confidence index every minute, ensuring that the deviation between the alternative orbit and the actual orbit before failure is ≤5m within 12 hours.

[0076] This invention provides a distributed, collaborative, real-time orbit determination device for GNSS / inter-satellite links for mega-constellations. The mega-constellation is divided into multiple orbital plane clusters, each cluster comprising multiple satellites. Several global anchor point satellites are evenly distributed within the mega-constellation. Each cluster has a cluster head satellite, and a first inter-satellite link exists between satellites within the cluster. A regional backbone node is located within multiple adjacent orbital plane clusters, and a second inter-satellite link exists between the backbone nodes. The device includes: The satellite calculation module is used to perform preliminary orbit calculations for satellites within each cluster based on local GNSS observation data and IMU data, and obtain preliminary calculation results for each satellite within the cluster; The cluster-level solution module is used by the cluster head satellite to summarize the preliminary solution results of each satellite in the cluster and determine the relative orbital parameters within the cluster.

[0077] The regional solution module is used by backbone nodes to collect the relative orbital parameters of each cluster and solve for the regional absolute orbital reference. The anchor point calculation module is used to summarize the reference values ​​of various regions for global anchor point satellites and determine the unified orbital parameters of the constellation. The final calculation module is used to determine the orbit calculation results of each satellite based on the regional absolute orbit reference and the unified orbit parameters of the constellation.

[0078] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0079] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A distributed cooperative real-time orbit determination method for GNSS / inter-satellite links for giant constellations, characterized in that, The mega-constellation is divided into multiple orbital plane clusters, each cluster comprising multiple satellites; several global anchor satellites are evenly distributed within the mega-constellation; a cluster head satellite is positioned within each cluster, and a first inter-satellite link exists between satellites within the cluster; a regional backbone node is positioned within multiple adjacent orbital plane clusters, and a second inter-satellite link exists between the backbone nodes; the method includes: Preliminary orbit calculations are performed on the satellites within each cluster based on local GNSS observation data and IMU data to obtain preliminary calculation results for each satellite within the cluster; The cluster head satellite summarizes the preliminary calculation results of each satellite in the cluster to determine the relative orbital parameters within the cluster; The backbone nodes collect the relative orbital parameters of each cluster and solve for the absolute orbital reference of the region; Global anchor satellites aggregate reference values ​​from various regions to determine unified orbital parameters for the constellation. Based on the absolute orbital reference of the region and the unified orbital parameters of the constellation, the orbital calculation results of each satellite are determined.

2. The GNSS / inter-satellite link distributed cooperative real-time orbit determination method for mega-constellations according to claim 1, characterized in that, Preliminary orbit calculations are performed on satellites within each cluster based on local GNSS observation data and IMU data, yielding preliminary calculation results for each satellite within the cluster, including: Satellites with GNSS receivers deployed within the cluster collect multi-mode GNSS observation data and IMU data, and perform preprocessing to obtain preprocessed observation data; Satellites within a cluster that do not have GNSS receivers deployed request raw GNSS observation data and data confidence scores from adjacent GNSS satellites within the cluster via the first inter-satellite link, forming a joint observation dataset. The sliding window is a local orbit determination sliding time window maintained by each satellite, and the window length is dynamically adjusted according to the satellite's orbital dynamics and the quality of the observation data. Each satellite performs preliminary orbit calculations based on preprocessed observation data and joint observation datasets, using a simplified orbit dynamics model.

3. The GNSS / inter-satellite link distributed cooperative real-time orbit determination method for mega-constellations according to claim 2, characterized in that, The cluster head satellite summarizes the preliminary calculation results of each satellite in the cluster to determine the relative orbital parameters within the cluster, including: Based on the preliminary calculation results of each satellite in the cluster and the weighted consensus algorithm, the states of the satellites in the cluster are fused to obtain state fusion data; the weights in the fusion process are determined according to the confidence level of the preliminary calculation results of each satellite. Using the state of the cluster head satellite as a reference, the state deviation of each satellite in the cluster relative to the cluster head satellite is calculated based on the state fusion data to construct the relative orbital parameters within the cluster.

4. The distributed cooperative real-time orbit determination method for GNSS / inter-satellite links for mega-constellations according to claim 3, characterized in that, The backbone nodes collect the relative orbital parameters of each cluster and calculate the regional absolute orbital reference, including: Based on the backbone node's own GNSS and IMU data, calculate its own absolute orbital parameters; Based on the relative orbit parameters and absolute orbit parameters of each cluster, calculate the transformation parameters between the coordinate system of each cluster and the coordinate system of the backbone node, and perform coordinate transformation on the relative orbit parameters of each cluster. The inter-satellite ranging data with adjacent backbone nodes is obtained as a constraint and jointly solved with the relative orbital parameters of each cluster after coordinate transformation to determine the absolute orbital parameters of the region. The residuals of the optimized regional absolute orbit parameters are evaluated. If they meet the accuracy requirements, they are determined as the regional absolute orbit reference and distributed to each cluster. Otherwise, the inter-satellite ranging data of other backbone nodes are obtained as constraints, and the regional absolute orbit parameters are recalculated.

5. The GNSS / inter-satellite link distributed cooperative real-time orbit determination method for mega-constellations according to claim 4, characterized in that, Global anchor satellites aggregate reference values ​​from various regions to determine unified orbital parameters for the constellation, including: Global anchor satellites calculate the absolute orbital parameters of the anchor points based on their own multi-mode GNSS and laser ranging data. Using its own anchoring benchmark as a reference, the systematic deviation of the benchmark values ​​in each region is calculated based on the absolute track parameters of the anchor point and the deviation correction model. The regional correction coefficient is then determined, and the regional benchmark data is corrected based on the regional correction coefficient. Under the constraint of inter-satellite ranging between anchor points, the unified orbital parameters of the constellation are obtained based on the regional reference data after deviation correction.

6. The GNSS / inter-satellite link distributed cooperative real-time orbit determination method for mega-constellations according to claim 5, characterized in that, Based on the regional absolute orbit reference and the unified orbit parameters of the constellation, the orbit calculation results for each satellite are determined, including: The cluster head satellite receives constellation unified orbit parameters and regional correction coefficients forwarded by backbone nodes in the region, and aligns them with its own stored regional absolute orbit reference in terms of time and coordinate system. The cluster head satellite uses the unified orbital parameters of the constellation as a reference, combines them with the regional absolute orbital reference, and corrects the regional reference deviation through a weighted fusion algorithm to output the corrected regional absolute orbital reference. Based on the corrected regional absolute orbit reference and the relative orbit parameters within the cluster, the cluster head satellite calculates the preliminary absolute orbit solution for each satellite within the cluster; The cluster head satellite collects inter-satellite ranging data between satellites within the cluster as a local constraint, and performs joint calculation with the preliminary absolute orbit results to obtain the orbit calculation results.

7. The GNSS / inter-satellite link distributed cooperative real-time orbit determination method for mega-constellations according to claim 6, characterized in that, The cluster head satellite collects inter-satellite ranging data among satellites within the cluster as local constraints, and performs joint calculations with the preliminary absolute orbit results to obtain the orbit calculation results, including: Retrieve the preliminary absolute orbit results of each satellite in the cluster and perform timestamp calibration; Based on the preliminary absolute orbit results and constrained by effective inter-satellite ranging data, a weighted least squares joint solution model is constructed, and the weights are dynamically allocated according to the data quality. Based on the joint solution model, the orbit solution results are determined.

8. The GNSS / inter-satellite link distributed cooperative real-time orbit determination method for mega-constellations according to claim 1, characterized in that, The method further includes: Real-time monitoring of satellite status within the cluster; triggering cluster reconfiguration when preset conditions are met. Based on the reconstructed cluster structure and LSTM orbit prediction model, update the inter-satellite link topology; Based on the updated inter-satellite link topology, the communication frequency of each level of inter-satellite link is dynamically adjusted.

9. The distributed cooperative real-time orbit determination method for GNSS / inter-satellite links for mega-constellations according to claim 1, characterized in that, The method further includes: When a satellite failure is detected, alternative orbital parameters for the failed satellite are calculated based on its historical orbital data and information on neighboring satellites.

10. A distributed collaborative real-time orbit determination device for GNSS / inter-satellite links for giant constellations, characterized in that, The mega-constellation is divided into multiple orbital plane clusters, each cluster comprising multiple satellites; several global anchor satellites are evenly distributed within the mega-constellation; a cluster head satellite is positioned within each cluster, and a first inter-satellite link exists between satellites within the cluster; a regional backbone node is positioned within multiple adjacent orbital plane clusters, and a second inter-satellite link exists between backbone nodes; the device includes: The satellite calculation module is used to perform preliminary orbit calculations for satellites within each cluster based on local GNSS observation data and IMU data, and obtain preliminary calculation results for each satellite within the cluster; The cluster-level solution module is used by the cluster head satellite to summarize the preliminary solution results of each satellite in the cluster and determine the relative orbital parameters within the cluster. The regional solution module is used by backbone nodes to collect the relative orbital parameters of each cluster and solve for the regional absolute orbital reference. The anchor point calculation module is used to summarize the reference values ​​of various regions for global anchor point satellites and determine the unified orbital parameters of the constellation. The final calculation module is used to determine the orbit calculation results of each satellite based on the regional absolute orbit reference and the unified orbit parameters of the constellation.

Images