Collision warning method based on maneuver state segmented trajectory prediction
By adopting a segmented orbit prediction method based on maneuver status, orbit predictions are identified and processed in segments. Combined with adaptive risk criterion switching, the problems of orbit prediction model mismatch and insufficient single criterion are solved, thereby improving the accuracy and stability of orbit prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIHANG UNIV
- Filing Date
- 2026-05-26
- Publication Date
- 2026-07-14
AI Technical Summary
Existing technologies fail to effectively distinguish different maneuvering states of targets within the forecast window in orbit prediction, leading to easy mismatch in orbit prediction models, excessive reliance on covariance quality in collision probability calculation, and difficulty in taking into account both missed and false alarms by a single risk criterion, especially when low-Earth orbit satellites frequently maneuver, resulting in insufficient prediction accuracy.
A segmented trajectory prediction method based on maneuver state is adopted. By identifying the maneuver state of the target, the trajectory prediction process is divided into multiple segments, an appropriate trajectory prediction strategy is selected, and risk criteria are adaptively switched according to the prediction reliability, including collision probability, maximum collision probability, BOX region, and risk duration.
It improves the targeting and accuracy of orbit prediction, reduces missed and false alarms, and is suitable for engineering application scenarios involving high-precision orbits of main satellites and publicly available TLE data of other space objects. It also has good scalability.
Smart Images
Figure CN122392368A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of space situational awareness, spacecraft orbit prediction and on-orbit collision warning technology, and specifically provides a collision warning method based on segmented orbit prediction under maneuvering conditions. Background Technology
[0002] In recent years, the rapid development of large-scale low-Earth orbit (LEO) constellations, remote sensing satellite ensembles, and on-orbit servicing missions has led to a continuous increase in the number of LEO space targets and a significant rise in close-proximity events between them. For space stations, remote sensing satellites, communication satellites, and various experimental payloads, routine collision screening using publicly available orbital data has become fundamental to orbital safety management. In engineering practice, the main satellite generates a position and velocity sequence within a warning window based on current precise orbit determination results or position and velocity status, combined with dynamic extrapolation and known maneuvering information. For other space objects, publicly available TLE data combined with the SGP4 model remains the most commonly used orbit prediction combination, offering advantages such as low acquisition cost, wide coverage, and rapid updates.
[0003] However, existing Time-Like-Ended Leak (TLE) warning procedures typically do not differentiate between different maneuvering states of a target within the forecast window. Instead, they employ a uniform orbital extrapolation model and a fixed, single risk criterion for collision screening. While this type of procedure is adequate for routine applications with long-term stable orbits, it is significantly inadequate for low-Earth orbit satellites exhibiting continuous orbital increases or decreases, frequent orbital maintenance, or unclear maneuvering boundaries. On one hand, the uniform extrapolation model struggles to reflect the trend changes in orbital elements during maneuvers, leading to discrepancies in the Time of Close Approach (TCA) and miss distance (denoted as TLE). and meeting speed ( The estimation bias increases; on the other hand, the collision probability... Highly sensitive to covariance quality, when stable covariance is difficult to obtain under publicly available data conditions, simply relying on... It can easily lead to missed alarms or false alarms.
[0004] Existing technologies have addressed issues such as maneuver detection, orbit fitting, collision probability calculation, and rapid BOX region screening. However, these methods typically focus on a single stage, lacking a unified, engineered closed-loop process, and have not organically integrated key stages such as "maneuver status identification—segmented orbit prediction—uncertainty estimation—risk criterion switching—rolling updates." Especially when the main satellite can obtain high-precision orbits, while other space objects primarily rely on publicly available TLE data, improving the accuracy of orbit predictions for other space objects exhibiting maneuvering signs, and selecting more appropriate risk criteria based on prediction reliability, remains a critical issue restricting the accuracy and stability of collision warning systems. Summary of the Invention
[0005] To overcome the above-mentioned shortcomings, this invention is proposed to provide solutions or at least partially solve the problems in the prior art, such as the easy mismatch of orbit prediction models when other space objects show signs of maneuvering, the excessive reliance on covariance quality in collision probability calculation, and the difficulty of taking into account both missed and false alarms with a single risk criterion.
[0006] This invention provides a collision warning method based on segmented trajectory prediction under maneuvering conditions, comprising: Based on whether the main satellite has high-precision orbit data within the warning window, the orbit prediction method for the main satellite and other space objects is adaptively determined; If the main satellite has the high-precision orbit data, the main satellite uses the high-precision orbit data as the orbit input, and the other space objects use a segmented orbit prediction method based on maneuver state recognition. If the main satellite does not have the high-precision orbit data, then the main satellite and the other space objects will all use the segmented orbit prediction method based on maneuver state recognition. Based on the forecast reliability index obtained from the segmented trajectory forecast based on the maneuver state recognition, the system adaptively switches between multiple preset risk criteria, including collision probability criteria, maximum collision probability criteria, BOX region criteria, closest distance criteria, and risk duration criteria, and outputs the risk assessment result with the switched criteria.
[0007] Preferably, the segmented trajectory prediction method based on maneuver state recognition includes: For the other space objects or the main satellite, identify their maneuvering states during the analysis period. The maneuvering states include at least a continuous orbit raising phase, a continuous orbit lowering phase, an orbit holding phase, a steady state phase, and a maneuvering boundary uncertainty phase. Using the detected maneuvering moment or state transition moment as the dividing point, the historical TLE sequence of the other space objects or the main satellite is divided into multiple adjacent orbital segments, each orbital segment corresponding to a maneuvering state; Based on the maneuvering status type of each orbital segment, an appropriate orbital forecasting strategy is selected for forecasting, and different effective forecast durations are set for different types of orbital segments.
[0008] Preferably, for the other space objects or the main satellite, identifying their maneuvering status during the analysis period includes: For the historical TLE sequences of the other space objects, and for the historical TLE sequences of the main satellite when the main satellite does not have the high-precision orbit data, the following steps are performed: Extract the semi-major axis and orbital height from historical TLE sequences and perform smoothing. Calculate the rate of change of orbital altitude, the rate of change of semi-major axis, the second difference of semi-major axis, and the number of sign flips of the rate of change of semi-major axis; When the orbital height change rate is greater than the first threshold and the semi-major axis change rate is greater than the second threshold for N consecutive epochs, it is determined to be a continuous track lifting segment. When the orbital height change rate is less than the first negative threshold and the semi-major axis change rate is less than the second negative threshold for N consecutive epochs, it is determined to be a continuous orbital descent segment. When the number of sign flips within the statistical window exceeds the third threshold or the absolute value of the second difference of the semi-major axis exceeds the fourth threshold, it is determined to be an uncertain segment of the maneuver boundary. When the number of symbol flips is no greater than the third threshold and the orbital height fluctuates periodically within a preset small range, it is determined to be an orbital holding segment; The remaining time periods are considered to be in a steady state.
[0009] Preferably, an appropriate orbit prediction strategy is selected based on the maneuvering state type of each orbit segment, and different effective prediction durations are set for different types of orbit segments, including: For the steady-state and track-maintaining phases, a baseline orbit extrapolation model is used for prediction. For the continuous track raising and continuous track lowering sections, a time-related drift correction term is superimposed on the baseline extrapolation model. The drift correction term is represented as a time polynomial function in the radial, tangential and normal directions, respectively, and its coefficients are obtained by least squares fitting of recent historical epochs. For the uncertain segment of the maneuver boundary, a conservative envelope that expands linearly with time is constructed in the radial, tangential and normal directions, and the shortest effective forecast duration is set.
[0010] Preferably, the effective forecast duration satisfies the following conditions: the effective forecast duration of the steady-state segment is longer than the effective forecast duration of the track-holding segment, the effective forecast duration of the track-holding segment is longer than the effective forecast duration of the continuous track-raising segment or the continuous track-lowering segment, and the effective forecast duration of the maneuver boundary uncertainty segment is the shortest.
[0011] Preferably, the forecast reliability index is calculated based on at least several of the following factors: The freshness of orbital data update time relative to the current moment; The root mean square error of the historical orbit prediction of the other space objects or the main satellite; The forecast duration of the current early warning window; Stability score based on the number of maneuver state switching times within the most recent time window.
[0012] Preferably, the forecast confidence index is used to classify forecasts into high confidence level, medium confidence level, or low confidence level by comparing them with preset high confidence threshold and low confidence threshold; wherein the high confidence threshold is greater than the low confidence threshold.
[0013] Preferably, the specific rule for adaptively switching among multiple preset risk criteria is as follows: When the forecast confidence level is high and the two-dimensional intersection plane covariance is valid, the joint criterion of collision probability criterion and nearest neighbor distance criterion is adopted. When the forecast confidence level is medium confidence, a joint criterion of maximum collision probability criterion, nearest distance criterion and risk duration criterion is adopted; When the forecast confidence level is low or the covariance quality indicator is invalid, a conservative early warning mode is entered, and risk identification is carried out using the BOX region criterion, the nearest neighbor distance threshold, and the upper bound of the maximum collision probability.
[0014] Preferably, the duration of the risk is: the length of time during which the relative distance remains below the medium-risk threshold, or the length of time during which the relative trajectory remains within the medium-risk screening area.
[0015] Preferably, it also includes a rolling update step: When the update triggering conditions are met, the collision warning method based on segmented trajectory prediction of maneuver status is re-implemented; The update triggering conditions include: receiving new high-precision orbit data from the main satellite, receiving new TLE data from other space objects, the current time reaching the preset next update time, or a change in the risk assessment results.
[0016] The beneficial effects of this invention are as follows: (1) It can significantly improve the accuracy of trajectory prediction for continuously maneuvering targets. By first identifying the maneuvering state and then performing segmented modeling, it avoids the obvious mismatch that occurs in the traditional unified extrapolation model near the continuous track lifting, continuous track lowering and maneuvering boundaries, thereby improving the stability of the nearest proximity time and nearest proximity distance estimation.
[0017] (2) Achieve linkage between forecast results and risk criteria. Adaptively switch between Pc, MaxPc, BOX region method, distance threshold and risk duration based on forecast confidence level. Improve risk resolution in high confidence scenarios and enhance conservatism in low confidence scenarios, taking into account both missed alarm control and false alarm suppression.
[0018] (3) Applicable to engineering application scenarios where high-precision orbit of the main satellite coexists with publicly available TLE data of other space objects. This invention uses the precise orbit or position and velocity status of the main satellite within the future early warning window as the input of the main satellite, and the continuous TLE sequence of other space objects as the main screening object input; when the main satellite cannot obtain a high-precision orbit, the system can also degenerate into a unified processing mode based on TLE, which has strong engineering adaptability.
[0019] (4) It has good scalability. The maneuver identification threshold, covariance default model, credibility scoring method, risk threshold and update frequency can all be configured for different satellite types, different orbital altitudes and different mission scenarios, making it suitable for expansion to multi-satellite parallel early warning, constellation-level operational safety analysis and special monitoring missions. Attached Figure Description
[0020] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein: Figure 1 This is a schematic flowchart of a collision warning method based on segmented track prediction in maneuver state according to an embodiment of the present invention. Detailed Implementation
[0021] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.
[0022] like Figure 1 As shown, the present invention provides a collision warning method based on segmented trajectory prediction under maneuvering conditions, comprising: S0, Orbit Data Acquisition and Preprocessing The process involves acquiring the current high-precision orbital data, position and velocity status, or known maneuver information of the main satellite, as well as the TLE sequences of other space objects to be screened within the historical analysis period. The main satellite's orbital data can be derived from precise orbit determination, dynamic extrapolation, onboard navigation, or telemetry and control calculations. If the main satellite cannot obtain the aforementioned high-precision orbital data, its TLE sequence can be used and processed in the same way as other space objects.
[0023] The TLE sequences of the other space objects, as well as the corresponding sequences of the main satellite when using TLE, are sorted by epoch time and deduplicated, anomaly detected, and marked as missing.
[0024] For adjacent epochs with an interval greater than a preset threshold (For example, 24 hours, this threshold can be adjusted according to the target orbital altitude and maneuver frequency) When the TLE data before and after that moment is divided into different subsequences, erroneous interpolation across maneuver segments can be avoided.
[0025] Extracting average motion from TLE eccentricity ,inclination Right ascension of ascending node Perigeal argument and the near point angle Isoorbital elements, and further calculate the semi-major axis. Perimeter altitude and apogee altitude Isoorbital characteristics, where the semi-major axis, perigee altitude, and apogee altitude can be calculated using the following formula: in, The gravitational constant of Earth, The average radius of the Earth This is an epochal index.
[0026] To reduce the impact of random fluctuations in orbital data on maneuver identification, moving median filtering, moving average, or robust smoothing are applied to the semi-major axis, perigee / apogee altitude, and mean motion sequence to obtain a more stable sequence of orbital change characteristics. For missing positions with small epoch intervals that do not cross maneuver breakpoints, nearest neighbor completion or linear interpolation can be used for completion; for time gaps that may cross maneuver breakpoints, no cross-segment interpolation is performed.
[0027] Step S1: Based on whether the main satellite has high-precision orbit data within the warning window, adaptively determine the orbit prediction method between the main satellite and other space objects; If the main satellite has the high-precision orbit data, the main satellite uses the high-precision orbit data as the orbit input, and the other space objects use a segmented orbit prediction method based on maneuver state recognition. If the main satellite does not have the high-precision orbit data, then the main satellite and the other space objects will all use the segmented orbit prediction method based on maneuver state recognition.
[0028] In this embodiment, the high-precision orbit data of the main satellite can be derived from various methods such as precise orbit determination, dynamic extrapolation, on-board navigation, or telemetry and control calculations. The high-precision orbit data can be a position and velocity sequence within a future warning window, or it can be a precise orbit prediction result. Other space objects typically refer to low-Earth orbit mega-satellite constellations, space debris, or other non-cooperative targets, and their publicly available orbit data is usually a two-line element (TLE) sequence.
[0029] Adaptive features include automatically selecting the orbit prediction path of the main satellite and other space objects based on whether the main satellite has high-precision orbit data within the warning window; and automatically switching between multiple risk criteria to output risk assessment results based on the prediction credibility index obtained during the segmented orbit prediction process.
[0030] In an alternative implementation, when the high-precision orbital data of the main satellite only covers part of the warning window, for the uncovered time period, the main satellite can also temporarily degenerate into the same TLE segmented forecasting method as the other space objects to ensure the integrity of the warning window.
[0031] In one embodiment, for other space objects, or for the main satellite when a high-precision orbit cannot be obtained, the orbital altitude change rate is constructed based on the smoothed orbital altitude and semi-major axis sequence. , semi-major axis change rate Second difference of semi-major axis and the number of symbol flips : in, For the first The orbital height corresponding to each epoch For the first The semi-major axis corresponding to each epoch, For the first The time corresponding to each epoch. The rate of change of orbital altitude, The rate of change of the semi-major axis It is the second difference of the semi-major axis. For statistics window Number of sign flips in the rate of change of the inner semi-major axis For symbolic functions, For indicator functions, For the current epoch index, This is the epoch index within the window.
[0032] In one embodiment, the segmented trajectory prediction method based on maneuver state identification in step S1 includes: Step S11: For the other space objects or the main satellite, identify their maneuvering status within the analysis period, including: for the historical TLE sequence of the other space objects, and for the historical TLE sequence of the main satellite when the main satellite does not have the high-precision orbit data, extract the semi-major axis and orbital altitude from the historical TLE sequence and perform smoothing; calculate the orbital altitude change rate, semi-major axis change rate, second difference of semi-major axis, and sign flip number of the semi-major axis change rate.
[0033] Preferably, moving median filtering or moving average is used to smooth the semi-major axis and track height sequences to reduce the impact of random fluctuations on maneuver recognition.
[0034] The judgment rules are as follows: When continuous Each epoch satisfies and At that time, it was determined to be a section of continuous track lifting; When continuous Each epoch satisfies and At that time, it was determined to be a continuous descending trajectory segment; When in the statistics window Internal satisfaction or At that time, it was determined to be an uncertain segment of the maneuver boundary; when When the orbital altitude fluctuates periodically within a small range, it is determined to be the orbital maintenance period; the rest of the time period is determined to be the steady state period.
[0035] For length less than Isolated short segments are merged with adjacent state segments to reduce the impact of misjudgments on subsequent forecasts.
[0036] in, The threshold for determining the rate of change of orbital altitude. The threshold value representing the rate of change of the semi-major axis. The threshold for determining the number of sign flips. The threshold value for determining the second difference of the semi-major axis is represented by . The threshold for determining the orbital hold-up status. Indicates the length of the statistical window with the sign flipped. This represents the number of epochs in which the conditions for determining a continuous track-raising or continuous track-lowering section must be met consecutively. This represents the minimum length threshold for isolated short segments.
[0037] In an alternative approach, maneuver status can also be identified by using auxiliary judgment based on orbital element changes such as the rate of change of orbital inclination and the rate of change of right ascension of the ascending node, in order to improve the accuracy of maneuver identification.
[0038] Step S12: Using the detected maneuver time or state transition time as the dividing point, divide the historical TLE sequence of the other space objects or the main satellite into multiple adjacent orbital segments, each orbital segment corresponding to a maneuver state.
[0039] Step S13: Select an appropriate orbit forecasting strategy for forecasting based on the maneuvering status type of each orbit segment, and set different effective forecast durations for different types of orbit segments.
[0040] Specifically, for the steady-state and orbit-maintaining phases, a baseline orbit extrapolation model is used for forecasting. This baseline extrapolation model can be the SGP4 model, a semi-analytical extrapolation model based on orbital elements, or other orbit forecasting models suitable for publicly available orbital data. However, considering that such targets are typically accompanied by periodic small orbital corrections, the effective forecast duration should be shorter than that of the steady-state phase, and the warning update frequency should be increased to reduce forecast bias caused by the accumulation of small maneuvers.
[0041] For sections with continuous track elevation and continuous track descent, a time-dependent drift correction term is superimposed on the baseline extrapolation model. This drift correction term is expressed as a polynomial function of time in the radial (R), tangential (T), and normal (N) directions, respectively, and its coefficients are obtained through least-squares fitting of recent historical epochs. For example, the drift correction term can be expressed as: The corrected target location can be represented as: in, Indicates the corrected orbital position; This indicates the orbital position obtained from the baseline extrapolation model; This represents a time-dependent drift correction term; Relative to reference time Time offset; Indicates the current prediction time; Indicates the start time or fitting reference time of the current orbital segment; , , These represent the radial, tangential, and normal directions, respectively. These are the coefficients of the first and second terms of the drift correction term in the three directions of the RTN.
[0042] The above coefficients were obtained through recent The least-squares fitting of each epoch yields the objective function, which can be written as: in, This represents the parameter estimation results obtained through least squares fitting. Represents the vector of parameters to be estimated. Indicates the first The reference orbit position corresponding to each fitted epoch. Indicates the first At the time of the fitted epoch, This indicates the length of the historical epoch window used for parameter fitting.
[0043] For the uncertain segment of the maneuver boundary, only short-time forecasts are retained, and a conservative envelope that expands over time is constructed in the three directions of the RTN coordinate system to form a safe boundary in low-confidence scenarios. Let the prediction reference time be... The predicted time is t, and the time offset is... The conservative envelopes of the uncertain segment of the maneuver boundary in the three directions of the RTN are respectively expressed as: in, These represent the initial envelope widths in the radial, tangential, and normal directions, respectively. These represent the envelope expansion rates in the corresponding directions, and both are non-negative parameters. Furthermore, different effective forecast durations are assigned to different orbital segments. , and and satisfy one of them , , and These represent the effective forecast durations for the steady-state segment, the maneuvering segment, and the uncertain segment at the maneuvering boundary, respectively.
[0044] Step S2: Segmented orbit prediction and candidate target selection.
[0045] Collision candidate targets are screened based on the orbital input of the main satellite and the predicted orbits of other space objects. To reduce computational burden, firstly, targets with obviously no collision possibility are removed by apogee altitude screening, orbital plane geometry screening, and analysis time window screening; then, the relative motion of the remaining candidate targets is calculated to obtain the rendezvous time (…). ), meeting distance ( ) and meeting speed ( Only when Only when the relative position enters the coarse screening BOX area will it enter the detailed risk assessment stage.
[0046] Step S3: Estimate the forecast uncertainty of the selected candidate targets and calculate the forecast credibility index, which characterizes the forecast quality.
[0047] Specifically, for the main satellite and other space objects, a back-substitution process of "predicting backward from historical epochs—comparing with reference states" is performed within the most recent M historical back-substitution windows to obtain orbit prediction residual samples for each window. For the primary satellite, if precise orbit, position-velocity sequence, or their extrapolated results are used as input, its error covariance can be obtained from the orbit determination accuracy, history fitting residuals, or given covariance information. When the primary satellite only uses TLE, the same history back substitution process as other space objects can be performed. For other space objects, after transforming the residuals to the RTN coordinate system, the position error covariance matrix P is constructed: in, The position error covariance matrix is constructed in the RTN coordinate system. The number of historical windows For the first The orbital prediction residual vector corresponding to each iteration window The residual sample mean. For transpose operation, To add a regularization term to the covariance matrix, where It is a non-negative small constant. It is the identity matrix. This serves as an index for the historical retrospective window.
[0048] When historical samples of candidate targets are insufficient, a default diagonal covariance model partitioned by target category, orbital altitude level, or update frequency can be used. The relative covariance between the primary target and candidate targets is denoted as... Then projected onto the intersection plane, a two-dimensional covariance matrix is obtained for calculating the collision probability or upper bound probability. At the same time, the freshness of the orbital data is updated. orbital fitting residual score Prediction duration penalty item and maneuverability stability score Construct a forecast reliability index : in, This represents the time difference between the corresponding orbital data and the current time. For the main satellite, it corresponds to the time difference between the update of precise orbit or position and velocity data. For other space objects, it corresponds to the TLE update time difference. RMSE represents the root mean square error of the fit or history back substitution. Indicates the duration of this prediction; This represents the number of times the state has switched within the most recent window. This is the stability decay parameter.
[0049] when When it is determined to be a high-confidence forecast; when When it is determined to be a moderately reliable forecast; when The forecast was determined to be of low confidence; among which For high confidence threshold, The threshold is low confidence, and .
[0050] Step S4: Adaptive switching of risk criteria Based on the forecast reliability index, the system adaptively switches between multiple preset risk criteria. These preset risk criteria include collision probability criteria, maximum collision probability criteria, BOX region criteria, closest proximity criteria, and risk duration criteria, and outputs the risk assessment result using the switched criteria.
[0051] The specific rules for switching are as follows: when Furthermore, when the covariance of the two-dimensional intersection plane is valid, the collision probability is used. and intersection distance ( The joint criteria of ( ) are used to issue early warnings; among them It can be based on the two-dimensional Gaussian distribution of the intersection plane and the joint hardware radius. Earn points through gameplay.
[0052] when When using the maximum collision probability and the intersection distance ( and duration of risk The joint criteria, among which This indicates that the relative distance remains below the medium-risk threshold. Or the length of time that the relative trajectory remains within the moderate screening area. The longer the risk duration, the higher the intersection risk, and even a slightly larger intersection distance at a single point should be taken seriously.
[0053] when If the covariance quality indicator is invalid, enter the conservative early warning mode and prioritize the use of the BOX area method, intersection distance threshold, maximum collision probability upper bound, and high-frequency update mechanism for risk identification.
[0054] For other space objects exhibiting characteristics of continuous orbit raising, continuous orbit lowering, or uncertain maneuver boundaries, the system automatically shortens their forecast window and increases the recalculation frequency; when the main satellite can only use TLE data, the same strategy is also applied to the main satellite to avoid missed warnings due to model mismatch.
[0055] S5. Warning Level Generation and Rolling Updates The system compares the switched risk criterion output with multiple preset warning thresholds to generate warning levels including low risk, medium risk, and high risk. Simultaneously, it outputs the main target identifier, candidate target identifier, maneuver status label, credibility index C, selected risk criterion, rendezvous time TCA, and rendezvous distance. Meeting speed Collision probability or maximum collision probability The next update time and recommendation analysis window.
[0056] In addition, it also includes a rolling update step: when new high-precision orbit data of the main satellite is received, new TLE data of other space objects is received, the current time reaches the preset next update time, or the risk assessment results change, the above steps S0 to S5 are re-executed (or at least steps S1 to S5 are re-executed), forming a closed-loop early warning process of "data update - preprocessing - state re-identification - segmented forecast - risk reassessment - level update".
[0057] The technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles and objectives of the present invention, those skilled in the art can make equivalent changes or substitutions to the original technical features, specific structures, materials, connection methods, arrangement methods, etc., and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.
Claims
1. A collision warning method based on segmented trajectory prediction under maneuvering conditions, characterized in that, include: Based on whether the main satellite has high-precision orbit data within the warning window, the orbit prediction method for the main satellite and other space objects is adaptively determined; If the main satellite has the high-precision orbit data, the main satellite uses the high-precision orbit data as the orbit input, and the other space objects use a segmented orbit prediction method based on maneuver state recognition. If the main satellite does not have the high-precision orbit data, then the main satellite and the other space objects will all use the segmented orbit prediction method based on maneuver state recognition. Based on the forecast reliability index obtained from the segmented trajectory forecast based on the maneuver state recognition, the system adaptively switches between multiple preset risk criteria, including collision probability criteria, maximum collision probability criteria, BOX region criteria, closest distance criteria, and risk duration criteria, and outputs the risk assessment result with the switched criteria.
2. The method according to claim 1, characterized in that, Segmented trajectory prediction methods based on maneuver state recognition include: For the other space objects or the main satellite, identify their maneuvering states during the analysis period. The maneuvering states include at least a continuous orbit raising phase, a continuous orbit lowering phase, an orbit holding phase, a steady state phase, and a maneuvering boundary uncertainty phase. Using the detected maneuvering moment or state transition moment as the dividing point, the historical TLE sequence of the other space objects or the main satellite is divided into multiple adjacent orbital segments, each orbital segment corresponding to a maneuvering state; Based on the maneuvering status type of each orbital segment, an appropriate orbital forecasting strategy is selected for forecasting, and different effective forecast durations are set for different types of orbital segments.
3. The method according to claim 2, characterized in that, For the other space objects or the main satellite, identify their maneuvering status during the analysis period, including: For the historical TLE sequences of the other space objects, and for the historical TLE sequences of the main satellite when the main satellite does not have the high-precision orbit data, the following steps are performed: Extract the semi-major axis and orbital height from historical TLE sequences and perform smoothing. Calculate the rate of change of orbital altitude, the rate of change of semi-major axis, the second difference of semi-major axis, and the number of sign flips of the rate of change of semi-major axis; When the orbital height change rate is greater than the first threshold and the semi-major axis change rate is greater than the second threshold for N consecutive epochs, it is determined to be a continuous track lifting segment. When the orbital height change rate is less than the first negative threshold and the semi-major axis change rate is less than the second negative threshold for N consecutive epochs, it is determined to be a continuous orbital descent segment. When the number of sign flips within the statistical window exceeds the third threshold or the absolute value of the second difference of the semi-major axis exceeds the fourth threshold, it is determined to be an uncertain segment of the maneuver boundary. When the number of symbol flips is no greater than the third threshold and the orbital height fluctuates periodically within a preset small range, it is determined to be an orbital holding segment; The remaining time periods are considered to be in a steady state.
4. The method according to claim 2, characterized in that, Based on the maneuvering status type of each orbital segment, an appropriate orbital prediction strategy is selected for prediction, and different effective prediction durations are set for different types of orbital segments, including: For the steady-state and track-maintaining phases, a baseline orbit extrapolation model is used for prediction. For the continuous track raising and continuous track lowering sections, a time-related drift correction term is superimposed on the baseline extrapolation model. The drift correction term is represented as a time polynomial function in the radial, tangential and normal directions, respectively, and its coefficients are obtained by least squares fitting of recent historical epochs. For the uncertain segment of the maneuver boundary, a conservative envelope that expands linearly with time is constructed in the radial, tangential and normal directions, and the shortest effective forecast duration is set.
5. The method according to claim 4, characterized in that, The effective forecast duration satisfies the following conditions: the effective forecast duration of the steady-state segment is longer than the effective forecast duration of the track-holding segment; the effective forecast duration of the track-holding segment is longer than the effective forecast duration of the continuous track-raising segment or the continuous track-lowering segment; and the effective forecast duration of the maneuver boundary uncertainty segment is the shortest.
6. The method according to claim 1, characterized in that, The forecast reliability index is calculated based on at least several of the following factors: The freshness of orbital data update time relative to the current moment; The root mean square error of the historical orbit prediction of the other space objects or the main satellite; The forecast duration of the current early warning window; Stability score based on the number of maneuver state switching times within the most recent time window.
7. The method according to claim 6, characterized in that, The forecast reliability index is used to classify forecasts into high reliability, medium reliability, or low reliability levels by comparing them with preset high reliability and low reliability thresholds. The high confidence threshold is greater than the low confidence threshold.
8. The method according to claim 7, characterized in that, The specific rules for adaptively switching among multiple preset risk criteria are as follows: When the forecast confidence level is high and the two-dimensional intersection plane covariance is valid, the joint criterion of collision probability criterion and nearest neighbor distance criterion is adopted. When the forecast confidence level is medium confidence, a joint criterion of maximum collision probability criterion, nearest distance criterion and risk duration criterion is adopted; When the forecast confidence level is low or the covariance quality indicator is invalid, a conservative early warning mode is entered, and risk identification is carried out using the BOX region criterion, the nearest neighbor distance threshold, and the upper bound of the maximum collision probability.
9. The method according to claim 8, characterized in that, The duration of the risk is defined as the length of time during which the relative distance remains below the medium-risk threshold, or the length of time during which the relative trajectory remains within the medium-risk screening area.
10. The method according to claim 1, characterized in that, It also includes a rolling update step: When the update triggering condition is met, the method described in claim 1 is executed again; The update triggering conditions include: receiving new high-precision orbit data from the main satellite, receiving new TLE data from other space objects, the current time reaching the preset next update time, or a change in the risk assessment results.