A low-earth-orbit debris early warning and cleaning decision method and system based on multi-source fusion

Abstract

The application provides a low-orbit debris early warning and cleaning decision method and system based on multi-source fusion, and belongs to the cross field of spacecraft orbit mechanics and space debris management. The method comprises the following steps: acquiring orbit observation data of target low-orbit debris and a spacecraft, and performing pretreatment to generate standard orbit observation; performing three-level progressive fusion orbit determination on the standard orbit observation, outputting high-precision orbit state estimation and state error covariance matrix, and performing multi-level progressive collision risk screening and evaluation, and outputting a collision risk event; based on the high-precision orbit state estimation of the collision risk event, calculating a cleaning priority index and sorting to obtain a high-risk debris priority sorting list; selecting a candidate cleaning target set from the list, constructing a multi-objective optimization model, solving a Pareto optimal cleaning scheme set under the constraint condition, and outputting a cleaning task scheme. The application realizes low-orbit debris perception, early warning, cleaning and integrated closed-loop management.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of spacecraft orbital mechanics and space debris management, and in particular relates to a low-Earth orbit debris early warning and cleanup decision-making method and system based on multi-source fusion. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Low Earth Orbit (LEO) is a core region for the concentrated operation of spacecraft conducting remote sensing, communication, navigation enhancement, and scientific exploration. With the deployment of mega-constellations and the accumulation of historical launches, the amount of space debris in the LEO region continues to increase, and the frequency of encounters between debris and spacecraft, and between debris fragments themselves, has significantly increased. If high-energy collision events continue to occur, they may induce cascading collision risks, reducing the long-term availability of critical orbital resources.

[0004] Space debris management comprises at least two interrelated technical tasks: first, collision warning, which involves calculating the collision probability, geometry, and risk level of rendezvous events based on orbital observations and predictions, providing a basis for spacecraft maneuvering and avoidance decisions; and second, cleanup decision-making, which, under conditions of limited cleanup resources, determines priority targets, cleanup order, cleanup strategies, and mission timing from a large number of hazardous debris, providing feasible implementation plans for proactive debris removal missions. However, in existing engineering systems, these two tasks are typically performed independently by different systems, and a closed-loop technical process using a single set of highly reliable orbital data to connect warning and decision-making has not yet been established.

[0005] Existing collision warning schemes often rely on two-line orbital elements (TLE) or single ground-based radar or single optical station data sources. TLE data is generated based on the SGP4 simplified analytical model, which has limited orbital prediction accuracy and does not provide complete and reliable state covariance information. Single sensor data is also limited by observation geometry, weather factors, measurement type, and coverage arc, making it difficult to accurately estimate the target orbital state. Existing cleanup decision-making schemes mostly use simplified indicators such as static hazardous debris ranking, debris quality, or orbital altitude, or rely on expert experience rules for sorting, making it difficult to incorporate the latest collision warning results and the latest multi-source observation data in a timely manner. The dynamic risk events output by collision warnings cannot be directly converted into cleanup priorities and task planning inputs, forming information silos where "warning only warns and cleanup only cleans up," which seriously restricts the effective utilization of cleanup resources.

[0006] In addition, existing systems generally lack an uncertainty expression mechanism that runs through the entire chain of "sensor measurement - fusion orbit determination - collision probability calculation - cleanup priority ranking - cleanup scheme output", and also lack a closed-loop update mechanism that feeds back environmental changes and actual effects to the front-end model after the cleanup task is completed, which makes it impossible for the system to adaptively optimize based on the task execution results. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, this invention provides a low-orbit debris early warning and cleanup decision-making method and system based on multi-source fusion, aiming to solve the technical problems of independent collision early warning and cleanup decisions, insufficient accuracy of orbital state estimation, and lack of closed-loop adaptive update capability in the prior art.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for low-orbit debris early warning and cleanup decision-making based on multi-source fusion; A method for low-Earth orbit debris early warning and cleanup decision-making based on multi-source fusion includes: Acquire orbital observation data of target low-Earth orbit debris and spacecraft, and preprocess the data to generate standard orbital observations; A three-level progressive fusion orbit determination is performed on the standard orbit observations to output a high-precision orbit state estimate and state error covariance matrix; Based on the high-precision orbital state estimation and state error covariance matrix, a multi-level progressive collision risk screening and assessment is performed, and collision risk events are output. Based on the high-precision orbital state estimation of the collision risk event, the cleanup priority index of each debris is calculated and dynamically sorted to generate a priority sorting list of high-risk debris. Select a set of candidate cleanup targets from the high-risk fragment priority ranking list, construct a multi-objective optimization model, solve the Pareto optimal cleanup scheme set under constraints, and output the cleanup task scheme.

[0009] As a further technical solution, the three-level progressive fusion orbit determination includes the first level of fusion between homogeneous observation stations, the second level of fusion between heterogeneous observations across different types, and the third level of fusion assisted by space environment models.

[0010] As a further technical solution, the first-level homogeneous observation station fusion is as follows: for observation data of the same observation type but from different observation stations, weighted least squares and Gauss-Newton iteration are used for joint processing to output the local state vector and its covariance matrix after homogeneous fusion. The second-level heterogeneous observation cross-type fusion is as follows: local filters are configured for different observation types, and the target six-dimensional state vector and covariance matrix are independently estimated based on their own observation data; then the main fusion unit performs covariance cross-fusion on each local state estimate and outputs the fused state vector and conservative covariance matrix. The third-level space environment model assisted fusion is as follows: solar radiation flux and geomagnetic activity index are input into the high-precision atmospheric density model, the atmospheric drag density term in the orbit propagation model is corrected online, and the high-precision orbit state estimate and its state error covariance matrix are output.

[0011] As a further technical solution, based on the high-precision orbital state estimation and state error covariance matrix, a multi-level progressive collision risk screening and assessment is performed, outputting collision risk events, including: Based on the perigee and apogee altitudes of each target, determine whether there is any overlap in the orbital altitude ranges of two targets. If there is no overlap, the target pair is directly eliminated. Within the predetermined warning window, calculate the minimum orbital intersection distance between two targets within the forecast time window. If the distance exceeds the preset safety threshold, remove the current target pair. For the selected candidate rendezvous events, a high-fidelity numerical orbit propagator is used to propagate the state vectors and state error covariance matrices of the two targets to the estimated rendezvous time. At the rendezvous time, a relative position covariance matrix is ​​constructed, the collision probability is calculated using the true covariance matrix, and the confidence interval of the collision probability is calculated by scaling the covariance matrix. The collision probability and collision consequence indicators are combined, a comprehensive risk index is calculated according to a preset weighted model, and risk levels are divided according to preset risk level thresholds, ultimately outputting collision risk events.

[0012] As a further technical solution, a relative position covariance matrix is ​​constructed at the rendezvous point, the collision probability is calculated using the true covariance matrix, and the confidence interval of the collision probability is calculated by scaling the covariance matrix, including: Based on high-precision orbital state estimation and the state error covariance matrix, the covariance is propagated to the estimated rendezvous time; high-precision orbital state estimation and state error covariance matrix. Construct the target's state covariance matrix at the estimated rendezvous time, and determine the relative position covariance matrix based on the state covariance matrix; A collision plane is constructed with the direction of the relative velocity vector of the target as the normal. Two orthogonal unit vectors in the plane are selected to form a coordinate transformation matrix. The relative position and the relative position covariance matrix are projected onto the collision plane to obtain the two-dimensional mean of the relative position and the two-dimensional covariance matrix. The sum of the physical radii of the targets is used as the integral radius of the circular region of the combined collision section. The collision probability is then calculated by constructing a two-dimensional Gaussian integral formula. The two-dimensional covariance matrix is ​​scaled to calculate the upper and lower bounds of the collision probability, which are used as the confidence intervals for the collision probability.

[0013] As a further technical solution, the cleanup priority index is:

[0014] in, To clear the priority index; This represents the probability of fusion collision. For fragment quality; Orbit density factor; This refers to the orbital lifetime factor. The diffusion factor of collision consequences; , , This is a normalized reference value; , , , , Weighting coefficients to satisfy normalization constraints.

[0015] As a further technical solution, the multi-objective optimization model takes maximizing the total risk reduction, minimizing the total speed increment, and minimizing the total task duration as optimization objectives; The constraints include total speed increment budget constraints, single maneuver capability constraints, ground telemetry and control coverage constraints, illumination condition constraints, cleanup strategy matching constraints, and safe distance constraints.

[0016] The second aspect of this invention provides a low-orbit debris early warning and cleanup decision system based on multi-source fusion.

[0017] A low-Earth orbit debris early warning and cleanup decision-making system based on multi-source fusion, comprising: The multi-source data acquisition and preprocessing module is configured to: acquire orbital observation data of target low-Earth orbit debris and spacecraft, and preprocess the data to generate standard orbital observation data; The three-level progressive fusion orbit determination module is configured to perform three-level progressive fusion orbit determination on the standard orbit observations and output a high-precision orbit state estimate and state error covariance matrix. The multi-level progressive collision risk assessment module is configured to: perform multi-level progressive collision risk screening and assessment based on the high-precision orbital state estimation and state error covariance matrix, and output collision risk events; The data-driven dynamic sorting module is configured to: calculate the cleanup priority index of each debris based on the high-precision orbital state estimation of the collision risk event and perform dynamic sorting to generate a priority sorting list of high-risk debris; The multi-objective cleanup task planning module is configured to: select a set of candidate cleanup targets from the priority sorting list of high-risk fragments, construct a multi-objective optimization model, solve for the Pareto optimal cleanup scheme set under constraints, and output the cleanup task scheme.

[0018] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a low-orbit debris early warning and cleanup decision method based on multi-source fusion as described in the first aspect of the present invention.

[0019] A fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of a low-orbit debris early warning and cleanup decision method based on multi-source fusion as described in the first aspect of the present invention.

[0020] The above one or more technical solutions have the following beneficial effects: (1) This invention unifies multi-source orbital observation data, such as radar, optical, laser ranging, space-based sensors, GNSS, and public ephemeris, under the same spatiotemporal reference. Through a three-level progressive orbit determination architecture of homogeneous inter-station fusion, heterogeneous cross-type CI fusion, and space environment model-assisted correction, it effectively overcomes the influence of geometric limitations and system biases of single data source observations, and outputs a high-precision orbital state estimate with a complete 6×6 state error covariance matrix. Even when there are unknown correlations in the local estimates, the CI fusion algorithm can still ensure that the fusion covariance is a conservative upper bound of the true error, so that the subsequent collision probability calculation and orbit prediction have a quantifiable uncertainty basis, which significantly improves the credibility of the state estimate and the reliability of the prediction.

[0021] (2) This invention uses the true covariance matrix of HFOS after three-level fusion correction for high-fidelity numerical propagation. It employs a three-level progressive screening process: geometric coarse screening, MOID medium screening, and fine propagation. The collision probability is numerically integrated on the two-dimensional collision plane based on the true non-spherical covariance, and the confidence interval of the collision probability is output. Compared with existing schemes, this invention can more accurately describe the distribution of relative position uncertainty at the moment of intersection, avoids probability calculation errors caused by the assumption of spherical covariance, effectively suppresses the persistent engineering problems of excessively high false alarm rates and missed alarms, and improves the actual usability of early warning information.

[0022] (3) This invention uses multi-source fusion data as a unified information carrier throughout the entire process, and uses collision probability and comprehensive risk index as input factors for the cleanup priority index. This makes the cleanup ranking no longer dependent on static historical rankings or simplified empirical rules, but responds in real time to changes in early warning results driven by the latest observation data. Whenever new data triggers an update, the fusion collision probability, orbit density, and orbital lifetime in the cleanup priority index are updated synchronously, and the priority ranking list evolves dynamically. This breaks the traditional information silo pattern of "early warning only cares about early warning, and cleanup only cares about cleanup," and realizes end-to-end data connectivity from risk perception to cleanup decision-making.

[0023] (4) This invention also constructs a complete closed-loop feedback update mechanism. The success / failure status of the cleanup task, changes in the residual debris environment, and subsequent orbit monitoring results are fed back to the cataloging database, collision risk model, and cleanup priority index weight model. Based on this, the cleaned targets are cataloged, the orbital density statistics of the target's adjacent altitude layer and inclination zone are updated, and the covariance scaling factor, risk level threshold, and cleanup priority index weight coefficient are optimized online using Bayesian posterior update or stochastic gradient descent. This mechanism enables the system to continuously learn from historical tasks and correct model biases, adaptively track the dynamic evolution of the low-Earth orbit debris environment, and avoid system performance degradation due to changes in the space environment.

[0024] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0026] Figure 1 This is a flowchart of the method in the first embodiment.

[0027] Figure 2 This is a system structure diagram of the second embodiment. Detailed Implementation

[0028] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0030] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0031] Example 1 This embodiment discloses a low-Earth orbit debris early warning and cleanup decision-making method based on multi-source fusion. It obtains a high-precision orbit state with complete covariance by three-level progressive fusion of multiple types of orbit observation data. It solves the collision risk by multi-level screening combined with real covariance, dynamically calculates the debris cleanup priority index, generates a cleanup scheme through multi-objective optimization, and realizes integrated closed-loop control of low-Earth orbit debris perception, early warning and cleanup by relying on the closed-loop correction model based on the actual test results of the mission.

[0032] like Figure 1 As shown, a low-Earth orbit debris early warning and cleanup decision-making method based on multi-source fusion includes: Step S1: Acquire orbital observation data of the target low-Earth orbit debris and spacecraft, and perform preprocessing to generate standard orbital observation data.

[0033] In this embodiment, orbital observation data of the target low-Earth orbit debris or spacecraft are obtained from at least three different types of data sources. These include ground-based radar ranging / angle measurement / velocity measurement data, ground-based optical telescope right ascension and declination angle measurement data, ground-based or space-based laser ranging data, space-based optical sensor angular position information, cooperative spacecraft space-based GNSS position and velocity data, publicly available orbital element ephemeris data, and space environment monitoring data.

[0034] The acquired data undergoes preprocessing, specifically including: converting the time stamps of each data source to a unified time base, such as Coordinated Universal Time (UTC) or BeiDou Time (BJDT); and converting the spatial coordinates of each data source to a unified geocentric inertial coordinate system, such as GCRS / J2000.0. For ground-based radar data, a conversion can be performed from the station coordinate system to the Earth-fixed coordinate system and then to the geocentric inertial coordinate system; for space-based optical data, a conversion can be performed from the celestial coordinate system and orbital coordinate system to the geocentric inertial coordinate system; for TLE or other publicly available ephemeris data, a position and velocity state under the same reference frame can be converted using a resolving transponder.

[0035] Subsequently, outlier detection, online system error calibration, and data quality labeling are performed on the unified observation data. Preferably, the sensor system bias is estimated and corrected using a cooperative target with a known precise orbit, and an initial value for the measurement noise covariance and a quality label are added to each observation data. The final output is the standard orbit observations (OQ).

[0036] Step S2 involves performing a three-level progressive fusion orbit determination on the standard orbit observations, outputting a high-precision orbit state estimate and state error covariance matrix. The three-level progressive fusion orbit determination includes a first-level fusion between homogeneous observation stations, a second-level fusion across different types of heterogeneous observations, and a third-level fusion assisted by a space environment model.

[0037] In step S21, during the fusion process among the first-level homogeneous observation stations, weighted least squares and Gauss-Newton iteration are used to jointly process data of the same observation type but from different observation stations. For example, the ranging, angle and velocity data of multiple radar stations for the same target during the same passage are geometrically combined, or the synchronous / quasi-synchronous angle measurement data of multiple optical stations are used for multi-segment joint orbit determination to reduce the influence of geometric constraints and system biases in single-station observations.

[0038] Suppose that the same type of sensor (such as multiple radar stations) targets the same target in epochs. The observation vector is The measurement noise follows a zero-mean Gaussian distribution, and the covariance matrix is... Its observation equation is:

[0039] in, For the goal The six-dimensional state vector at time t. For the observation model function, for Measurement noise at time.

[0040] Using the state transition matrix Reference Era status Mapped to :

[0041] The optimal state estimate is solved using weighted least squares and Gauss-Newton iteration, with the cost function being the sum of squared residuals:

[0042] in, This is the weighted residual sum of squares cost function, used to measure the weighted deviation between predicted and actual observations; The total number of observation epochs participating in the fusion.

[0043] Levenberg The Marquardt method is used for iterative optimization, starting from the initial guess. Starting from this point, the iterative update formula (step k+1) is:

[0044] in This is the Levenberg-Marquardt damping factor, used to ensure stable convergence of the iteration; It is the identity matrix; It is a column vector consisting of all observations; The noise covariance matrix is ​​divided into blocks and diagonally. The Jacobian matrix for observation is defined as:

[0045] in, .

[0046] When the iteration proceeds to satisfy the convergence criterion (such as the state correction amount) (If the value is less than a preset threshold), then the optimal state estimate is obtained. This estimate is the local state vector output by the first-level homogeneous fusion:

[0047] Simultaneously, using the Jacobian matrix H and noise covariance R at convergence, and based on the least squares covariance propagation law, the uncertainty of the fusion state is given:

[0048] in, The formal error covariance matrix, used as the first-level fusion matrix, is employed for subsequent heterogeneous fusion and accuracy evaluation.

[0049] In step S22, during the second-level heterogeneous observation cross-type fusion process, local filters are configured for different observation types. Each local filter estimates the target six-dimensional state vector and its 6×6 covariance matrix based on its own observation data. Then, the main fusion unit performs covariance cross-fusion (CI) or other conservative fusion algorithms on each local state estimate to ensure that the fusion covariance is not underestimated even when there are unknown correlations in the local estimates.

[0050] Let N be the number of local filters actually participating in this fusion, where 2 The state estimate and its covariance matrix of the j-th local filter output are expressed as: j=1,2,...,N in, For the six-dimensional orbital state estimation corresponding to the j-th type of observation data, This is the corresponding state error covariance matrix.

[0051] Since there are unknown correlations among the estimation errors, the CI fusion algorithm is used in this embodiment. CI fusion state and covariance formula:

[0052]

[0053] in, The conservative covariance matrix after CI fusion; This is the fusion state vector after CI fusion; Estimate the fusion weights for the j-th local state, satisfying:

[0054]

[0055] The fusion weights are determined based on a preset uncertainty evaluation index. The evaluation index includes at least one of the determinant, logarithmic determinant, trace, or maximum eigenvalue of the fusion covariance matrix. For example, the optimal weights are determined when the objective is to minimize the determinant or trace of the fusion covariance matrix. Determination of (based on minimizing the determinant / trace of the covariance matrix):

[0056] or

[0057] in, The optimal fusion weight vector is the best value of each local filter weight obtained through optimization. To incorporate the determinant of the covariance matrix; To fuse the trace of the covariance matrix; This is the fusion weight vector. When N=2, we can let... , And perform a one-dimensional search in the interval [0,1]; when N>2, the fusion weight can be obtained by sequential quadratic programming, interior point method or other constrained numerical optimization methods.

[0058] This optimization problem belongs to either univariate search (when N=2) or convex optimization. CI fusion guarantees that even if the local estimation errors are perfectly correlated, the fused result will remain consistent. It is also always a conservative upper bound on the true error covariance matrix (i.e. (This is a positive semi-definite matrix), thus avoiding the risk of missed alarms. In summary, the second-level output is: a fused state vector. and conservative covariance matrix .

[0059] Step S23, third-level spatial environment model-assisted fusion.

[0060] After the first-level homogeneous inter-station fusion and the second-level heterogeneous cross-type CI fusion, a state estimate and its conservative covariance matrix incorporating geometric information from multiple sources have been obtained. However, due to atmospheric drag perturbations in the low Earth orbit region, geometric orbit determination relying solely on observation data cannot effectively constrain error divergence in the orbit prediction stage. Especially during geomagnetic storms or solar high years, atmospheric density model biases significantly degrade the accuracy of collision probability calculations for rendezvous events. Therefore, based on the second-level output, a third-level space environment model is further introduced to assist fusion. By estimating the atmospheric density proportionality correction factor online, external environment-driven data and existing orbit states are jointly filtered to output a final state estimate (HFOS) suitable for high-fidelity orbit prediction.

[0061] Specifically, the F10.7 solar radiation flux, Ap / Kp geomagnetic activity index, and high-precision atmospheric density model output are used as external drivers to perform online correction of the atmospheric drag density term in the orbital propagation model, reducing the amplification of orbital prediction errors during years of high solar activity or geomagnetic disturbances. To reduce prediction errors caused by atmospheric drag perturbations, the ballistic coefficients are extended to the state to be estimated. Let the extended state vector be... ,in This is the atmospheric density ratio correction factor.

[0062] Atmospheric drag acceleration continuity equation:

[0063] In the formula, This is the atmospheric drag acceleration vector; The reference density is based on the real-time F10.7 and Ap / Kp exponent outputs for the NRLMSISE-00 or HASDM model. Let be the velocity vector of the target relative to the rotating atmosphere. This is the drag coefficient; Effective windward cross-sectional area for the target; Quality for the target; Let be the position vector of the target in the geocentric inertial frame.

[0064] Using the orbit determination residual vector after the first and second stages are fused , To construct a system for preprocessed standard orbit observations, we can... The observation equation, using the observation residuals to... Online updates:

[0065] in, This is the residual term. The optimal scaling factor is calculated in real time using the above equation. This information is then fed back to the acceleration term in the numerical integrator, thereby enabling high-precision trajectory prediction.

[0066] After the above corrections, the second-level CI fusion state vector and conservative covariance matrix Using the initial values, numerical integration is performed under the drive of a high-precision dynamic model (including corrected atmospheric drag terms) to obtain the final high-precision orbital state estimate HFOS (High-Fidelity Orbit State) from the third-stage output:

[0067]

[0068] in, The six-dimensional state vector under the reference epoch; To retain the conservative covariance matrix output from the second-level CI fusion, the estimated values ​​of the surface-to-mass ratio A / m and the atmospheric density ratio correction factor are also output during the state estimation process.

[0069] The high-fidelity orbit state estimate (HFOS) output by the above steps includes target identifier, reference epoch, six-dimensional state vector, complete 6×6 state error covariance matrix, surface-to-mass ratio estimate, root mean square of orbit residuals, and list of data sources.

[0070] At this point, the three-stage progressive fusion orbit determination link is complete. The HFOS output from the third stage, while retaining the geometric constraints of multi-source observations, incorporates physical correction information from the atmospheric density model, and its state covariance... This more realistically reflects the evolution of orbit forecast uncertainty in the time domain. This state estimate will serve as the sole orbit input for subsequent multi-level progressive collision risk screening and assessment, ensuring that collision probability calculation and cleanup priority ranking are both based on the same highly reliable orbit benchmark with complete covariance, thereby achieving a data closed-loop connection from "observation fusion" to "risk decision-making".

[0071] Step S3: Based on the high-precision orbital state estimation and state error covariance matrix, perform multi-level progressive collision risk screening and assessment, and output collision risk events.

[0072] Using the output HFOS as input, the orbital geometry coarse screening is first performed: target pairs that are unlikely to intersect at close range are excluded based on whether the perigee and apogee altitude ranges of each target overlap.

[0073] Based on the target's perigee height and apogee altitude Determine if the orbital heights of the two targets overlap. For targets i and j, if the following conditions are met:

[0074] This indicates that the two target orbits do not overlap in spatial altitude, and the target pair is directly eliminated and will not proceed to the subsequent screening process.

[0075] Next, the minimum orbital intersection distance (MOID) is used for screening to exclude target pairs that have no possibility of approaching within the predetermined warning window or whose MOID is greater than the safety threshold.

[0076] For target pairs that pass through the geometric coarse sieve, within the predetermined early warning window Solve for the minimum orbital intersection distance (MOID) between two targets. Define the spatial relative distance between the two targets at time t. for:

[0077] Solving for MOID is equivalent to finding the time interval that makes MOID equal to 0. The moment of taking the minimum value That is, to find the roots of the following equation:

[0078] The root of the above equation is solved iteratively over the time interval using the Brent method or the bisection method to obtain the minimum distance. .like If so, then the target pair will be removed.

[0079] For the shortlisted candidate rendezvous events, a high-fidelity numerical orbit propagator is used to propagate the state vectors and covariance matrices of the two targets to the estimated rendezvous time. The transmitter may include the Earth's non-spherical gravitational field, the gravitational pull of the Sun and Moon (third bodies), solar radiation pressure and Earth's shadow model, an S2-corrected atmospheric drag model, and necessary relativistic corrections.

[0080] At the moment of intersection, a relative position covariance matrix is ​​constructed, and the collision probability is calculated using the true covariance matrix instead of assuming a spherical covariance. Preferably, Monte Carlo importance sampling, numerical integration, or the equivalent volume method can be used to calculate... .

[0081] (1) Formula for linear propagation of true covariance For target pairs that pass the intermediate screening, the high-precision orbital state estimate HFOS and its state error covariance matrix output by S2 are used to propagate the covariance to the estimated rendezvous time. :

[0082] in, This is the state transition matrix; This is the process noise matrix, used to absorb the effects of higher-order unmodeled perturbations.

[0083] (2) Construct the relative position covariance matrix Suppose targets A and B are in The covariance matrices are respectively and Then the relative position covariance matrix is:

[0084] in, , Take respectively , of Location sub-block. Here, the true non-spherical covariance is used directly, rather than the simplified isotropic spherical covariance, to ensure the accuracy of the collision probability calculation.

[0085] (3) Collision probability Two-dimensional numerical integration formula Suppose that target A and target B are in The closest in time, the relative velocity vector is Construct the collision encounter plane based on the relative velocity vector, and then construct a plane with... Let be the collision plane with the normal vector, and let the unit normal vector of the collision encounter plane be:

[0086] In vertical Choose mutually orthogonal unit basis vectors in the plane. , Construct the projection matrix from the collision encounter plane in three-dimensional space to the two-dimensional space. :

[0087] in , Let be an orthogonal unit vector in the plane.

[0088] The relative position between the centroids of the two targets and relative covariance Projecting onto this plane yields the two-dimensional average relative position. and the two-dimensional relative position covariance matrix :

[0089]

[0090] in, The projection matrix; It is a three-dimensional relative position vector.

[0091] Consider the physical radii of the two targets The combined collision cross section projects onto a two-dimensional plane as a circular region. ,radius ,radius The integral radius of the combined collision section defines the collision region on the two-dimensional collision plane. :

[0092] Collision probability By analyzing the two-dimensional Gaussian probability density function in the collision region Integrating, we get:

[0093] Because the integrand is highly nonlinear, this formula cannot be solved analytically directly. Instead, it can be calculated using numerical integration methods (such as two-dimensional Simpson quadrature) or Monte Carlo importance sampling methods.

[0094] (4) Confidence interval output In calculation At the same time, consider the true covariance matrix The uncertainty boundaries (upper and lower limits) for the formula Scaling is performed to calculate The 95% confidence interval, along with the collision probability. Together they are output as CRE (Collision Risk Event).

[0095] Then The collision consequence indicators, combined with other factors, form a comprehensive risk index R. These consequences indicators may include debris mass, relative velocity, orbital density factor, and orbital lifetime factor. Therefore, the multidimensional comprehensive risk index R is:

[0096] In the formula, This is a high-fidelity nonlinear collision probability, used to characterize the probability of an encounter occurring; The mass of the target fragment is used to characterize the kinetic energy destructive potential of the collision. The preset normalized reference quality; The orbital density factor of the target orbit is used to characterize the potential risk of cascading collisions in its region; The orbital lifetime of the target debris is used to characterize the time risk of its long-term stay in space. This is the preset normalized reference lifetime; The weighted composite weights of the risk factors in each dimension satisfy the normalization constraint. .

[0097] Step S4: Based on the high-precision orbital state estimation of the collision risk event, calculate the cleanup priority index of each debris and dynamically sort them to generate a priority sorting list of high-risk debris.

[0098] Based on the output collision risk events and debris state attributes, a cleanup priority index (CPI) is calculated. Preferably, the CPI is a weighted composite of five factors: fused collision probability... Fragment quality Orbit density factor Orbital life and collision consequence diffusion factor .

[0099] One feasible form of CPI calculation is as follows:

[0100] in , , As a normalized reference value, to The weighting coefficients are set to satisfy the normalization constraints. The weights can be given by expert experience, or determined by using historical cleanup data through random forest regression, Bayesian optimization, or online learning methods.

[0101] Whenever new multi-source observation data enters the system and triggers an update, the fusion collision probability... Fragment quality Orbit density factor Orbital life and collision consequence diffusion factor All of these can be updated synchronously, so that the CPI ranking list reflects the latest orbital environment and the latest collision risks, rather than a static historical ranking.

[0102] Step S5: Select a set of candidate cleanup targets from the high-risk fragment priority ranking list, construct a multi-objective optimization model, solve the Pareto optimal cleanup scheme set under constraints, and output the cleanup task scheme.

[0103] Receive the CPI ranking list output from step S4, construct a multi-objective optimization model, and output the Pareto optimal cleanup scheme set (CMP) to resolve the conflict between risk reduction, fuel consumption, and mission duration. Select the top K debris from the output CPI descending list as the candidate cleanup target set. The value of K is determined by the total velocity increment budget of the cleanup spacecraft, the upper limit of mission duration, payload capacity, mission window, and safety constraints.

[0104] The core contradiction in cleanup mission planning lies in the conflicting relationships between risk reduction benefits, fuel consumption, and mission duration. Maximizing risk reduction may require visiting multiple targets with significantly different orbital planes, leading to a sharp increase in velocity increments; conversely, optimizing fuel consumption may favor targets with similar orbits, but at the expense of overall risk reduction. Therefore, this step models the cleanup sequence optimization as a three-objective optimization problem. Thus, in this embodiment, a multi-objective optimization model is constructed: Objective 1 is to maximize the total risk reduction, i.e., maximizing the sum of the CPI values ​​or risk reduction benefits of the targets included in the cleanup sequence; Objective 2 is to minimize the total velocity increment. This includes the speed increments required for target transfers, rendezvous and approach, and cleanup operations; objective three is to minimize the total mission duration. Since the three objectives conflict, the preferred solution is the Pareto optimal solution set.

[0105] (1) Multi-objective optimization mathematical model Let the cleanup sequence be S, and the objective function vector be:

[0106] in, Represents the i-th target The cleanup priority index; Indicates the total duration of the task; The total velocity increment consists of three parts:

[0107] in, This is the initial orbital insertion / transfer velocity increment; This represents the speed increment for the transfer between targets; To clear the operation speed increment.

[0108] During the transfer between targets, to reduce propellant consumption, the natural drift effect of the right ascension of the ascending node (RAAN) caused by the Earth's J2 term perturbation can be fully utilized. By selecting the starting time, the residual can be optimized. This reduces out-of-plane maneuvering energy consumption.

[0109] (2) Constraints and Pareto Optimality Constraints are set, including total budget constraints, single maneuver capability constraints, ground tracking and control coverage constraints, illumination condition constraints, cleanup strategy matching constraints, and safety distance constraints. Cleanup strategies can be matched using a "debris characteristics - cleanup strategy" knowledge base. For example, small debris can be matched with laser ablation or non-contact momentum transfer strategies, medium-sized or high-spin debris can be matched with net / trawling capture strategies, and large, structurally intact targets can be matched with robotic arm capture or docking deorbiting propulsion strategies.

[0110] In orbit transfer optimization, the natural RAAN drift of the ascending node caused by Earth's J2 term perturbation can be utilized. By selecting the mission start time and transfer orbit parameters, the RAAN drift difference between the cleanup spacecraft and the next target can be compensated for the orbital surface difference during the transfer time, reducing out-of-plane maneuver propellant consumption. Optimization algorithms can employ NSGA-II, MOEA / D, particle swarm optimization, or mixed integer programming, among others.

[0111] The final output is the Cleanup Mission Plan (CMP), which includes the target cleanup sequence, cleanup strategies for each target, a summary of the transfer trajectory, Δv decomposition, task sequence, risk reduction amount, and robustness assessment.

[0112] Furthermore, it also includes step S6, the cleanup effect feedback and model closed-loop update step. After the cleanup mission is completed, the system front-end will be fed back with information such as whether the target was successfully captured, whether it was deorbited, changes in the residual debris environment, observation data from the spacecraft's attached sensors, and subsequent orbit monitoring results.

[0113] Based on the feedback results, the debris cataloging database is updated, cleared targets are marked as cleared or re-entered, and the spatial distribution statistics of debris in the target's adjacent height layer and dip zone are refreshed, and the orbital density factor is updated. .

[0114] Based on the changes in predicted risk before cleanup and measured risk after cleanup, the collision probability calculation parameters, covariance scaling factor, risk level threshold, and CPI weight coefficient are updated online using Bayesian posterior, least squares recalibration, or stochastic gradient descent. After the above is completed, return to step S1 to enter the next round of closed-loop iteration.

[0115] Example 2 This embodiment discloses a low-orbit debris early warning and cleanup decision system based on multi-source fusion; like Figure 2 As shown, a low-Earth orbit debris early warning and cleanup decision-making system based on multi-source fusion includes: The multi-source data acquisition and preprocessing module is configured to: acquire orbital observation data of target low-Earth orbit debris and spacecraft, and preprocess the data to generate standard orbital observation data; The three-level progressive fusion orbit determination module is configured to perform three-level progressive fusion orbit determination on the standard orbit observations and output a high-precision orbit state estimate and state error covariance matrix. The multi-level progressive collision risk assessment module is configured to: perform multi-level progressive collision risk screening and assessment based on the high-precision orbital state estimation and state error covariance matrix, and output collision risk events; The data-driven dynamic sorting module is configured to: calculate the cleanup priority index of each debris based on the high-precision orbital state estimation of the collision risk event and perform dynamic sorting to generate a priority sorting list of high-risk debris; The multi-objective cleanup task planning module is configured to: select a set of candidate cleanup targets from the priority sorting list of high-risk fragments, construct a multi-objective optimization model, solve for the Pareto optimal cleanup scheme set under constraints, and output the cleanup task scheme.

[0116] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0117] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in Example 1.

[0118] Example 4 The purpose of this embodiment is to provide an electronic device.

[0119] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in a low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in Embodiment 1.

[0120] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0121] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0122] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for low-Earth orbit debris early warning and cleanup decision-making based on multi-source fusion, characterized in that, include: Acquire orbital observation data of target low-Earth orbit debris and spacecraft, and preprocess the data to generate standard orbital observations; A three-level progressive fusion orbit determination is performed on the standard orbit observations to output a high-precision orbit state estimate and state error covariance matrix; Based on the high-precision orbital state estimation and state error covariance matrix, a multi-level progressive collision risk screening and assessment is performed, and collision risk events are output. Based on the high-precision orbital state estimation of the collision risk event, the cleanup priority index of each debris is calculated and dynamically sorted to generate a priority sorting list of high-risk debris. Select a set of candidate cleanup targets from the high-risk fragment priority ranking list, construct a multi-objective optimization model, solve the Pareto optimal cleanup scheme set under constraints, and output the cleanup task scheme.

2. The low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in claim 1, characterized in that, The three-level progressive fusion orbit determination includes the first level of fusion between homogeneous observation stations, the second level of fusion between heterogeneous observations across different types, and the third level of fusion assisted by space environment models.

3. The low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in claim 2, characterized in that, The first-level homogeneous observation station fusion is as follows: for observation data of the same observation type but from different observation stations, weighted least squares and Gauss-Newton iteration are used for joint processing to output the local state vector and its covariance matrix after homogeneous fusion; The second-level heterogeneous observation cross-type fusion is as follows: local filters are configured for different observation types, and the target six-dimensional state vector and covariance matrix are independently estimated based on their own observation data; then the main fusion unit performs covariance cross-fusion on each local state estimate and outputs the fused state vector and conservative covariance matrix. The third-level space environment model assisted fusion is as follows: solar radiation flux and geomagnetic activity index are input into the high-precision atmospheric density model, the atmospheric drag density term in the orbit propagation model is corrected online, and the high-precision orbit state estimate and its state error covariance matrix are output.

4. The low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in claim 1, characterized in that, Based on the high-precision orbital state estimation and state error covariance matrix, a multi-level progressive collision risk screening and assessment is performed, outputting collision risk events, including: Based on the perigee and apogee altitudes of each target, determine whether there is any overlap in the orbital altitude ranges of two targets. If there is no overlap, the target pair is directly eliminated. Within the predetermined warning window, calculate the minimum orbital intersection distance between two targets within the forecast time window. If the distance exceeds the preset safety threshold, remove the current target pair. For the selected candidate rendezvous events, a high-fidelity numerical orbit propagator is used to propagate the state vectors and state error covariance matrices of the two targets to the estimated rendezvous time. At the rendezvous time, a relative position covariance matrix is ​​constructed, the collision probability is calculated using the true covariance matrix, and the confidence interval of the collision probability is calculated by scaling the covariance matrix. The collision probability and collision consequence indicators are combined, a comprehensive risk index is calculated according to a preset weighted model, and risk levels are divided according to preset risk level thresholds, ultimately outputting collision risk events.

5. The low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in claim 1, characterized in that, At the moment of rendezvous, a relative position covariance matrix is ​​constructed. The collision probability is calculated using the true covariance matrix, and the confidence interval for the collision probability is calculated by scaling the covariance matrix, including: Based on high-precision orbital state estimation and the state error covariance matrix, the covariance is propagated to the estimated rendezvous time; high-precision orbital state estimation and state error covariance matrix. Construct the target's state covariance matrix at the estimated rendezvous time, and determine the relative position covariance matrix based on the state covariance matrix; A collision plane is constructed with the direction of the relative velocity vector of the target as the normal. Two orthogonal unit vectors in the plane are selected to form a coordinate transformation matrix. The relative position and the relative position covariance matrix are projected onto the collision plane to obtain the two-dimensional mean of the relative position and the two-dimensional covariance matrix. The sum of the physical radii of the targets is used as the integral radius of the circular region of the combined collision section. The collision probability is then calculated by constructing a two-dimensional Gaussian integral formula. The two-dimensional covariance matrix is ​​scaled to calculate the upper and lower bounds of the collision probability, which are used as the confidence intervals for the collision probability.

6. The low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in claim 1, characterized in that, The cleanup priority index is: in, To clear the priority index; This represents the probability of fusion collision. For fragment quality; Orbit density factor; This refers to the orbital lifetime factor. The diffusion factor of collision consequences; , , This serves as a normalized reference value. , , , , Weighting coefficients to satisfy normalization constraints.

7. The low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in claim 1, characterized in that, The multi-objective optimization model aims to maximize the total risk reduction, minimize the total speed increment, and minimize the total task duration. The constraints include total speed increment budget constraints, single maneuver capability constraints, ground telemetry and control coverage constraints, illumination condition constraints, cleanup strategy matching constraints, and safe distance constraints.

8. A low-orbit debris early warning and cleanup decision-making system based on multi-source fusion, characterized in that, include: The multi-source data acquisition and preprocessing module is configured to: acquire orbital observation data of target low-Earth orbit debris and spacecraft, and preprocess the data to generate standard orbital observation data; The three-level progressive fusion orbit determination module is configured to perform three-level progressive fusion orbit determination on the standard orbit observations and output a high-precision orbit state estimate and state error covariance matrix. The multi-level progressive collision risk assessment module is configured to: perform multi-level progressive collision risk screening and assessment based on the high-precision orbital state estimation and state error covariance matrix, and output collision risk events; The data-driven dynamic sorting module is configured to: calculate the cleanup priority index of each debris based on the high-precision orbital state estimation of the collision risk event and perform dynamic sorting to generate a priority sorting list of high-risk debris; The multi-objective cleanup task planning module is configured to: select a set of candidate cleanup targets from the priority sorting list of high-risk fragments, construct a multi-objective optimization model, solve for the Pareto optimal cleanup scheme set under constraints, and output the cleanup task scheme.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the low-orbit debris early warning and cleanup decision-making method based on multi-source fusion as described in any one of claims 1-7.

Images