Space debris near-earth orbit cleaning system based on cooperation of beidou satellite and unmanned aerial vehicle

Abstract

This invention discloses a space debris near-Earth orbit cleanup system based on the collaboration of BeiDou satellites and unmanned aerial vehicles (UAVs), relating to the field of debris cleanup. The system includes: configuring a parent satellite's Ku-band phased array radar and linking it with an imaging payload; setting radar scanning cycles and identification threshold parameters; setting strategies for echo acquisition, observation timescales, and raw data retention; performing feature extraction based on echo acquisition and the imaging payload; setting rules for debris entry generation and merging updates; performing preliminary material determination and mass estimation based on feature extraction results; and setting a method for calculating threat scores and determining candidate cleanup targets. By constructing a dynamic debris database covering the target orbit, rapidly solidifying integer ambiguities based on inter-satellite dual differences and LAMBDA, and employing a material-adaptive laser / mechanical anchoring process driven by material confidence and mass ranges, the system achieves a closed-loop capability with a response time of no more than 30 minutes from identification to release, 10Hz relative position / velocity updates, and end-point alignment.

Description

Technical Field

[0001] This invention relates to the field of debris removal, and in particular to a space debris near-Earth orbit removal system based on the collaboration of BeiDou satellites and unmanned aerial vehicles. Background Technology

[0002] The increasing number of near-Earth orbit debris and the growing risk of collisions necessitate continuous cataloging, prioritization, and on-orbit removal of debris within the orbital zone. Existing ground / single-platform detection or passive relocation solutions have limitations in terms of observation frequency, material identification, and safety of close-range operations, making it difficult to meet the operational and maintenance requirements for rapid response and multi-aircraft coordinated towing.

[0003] However, there are still two problems in the existing technology. First, the precision of close rendezvous is limited by the insufficient accuracy of inter-satellite differential and time synchronization, which makes it difficult to obtain centimeter-level relative positioning and 10Hz updates stably, increasing the risk of collision and mismatch. Second, the cleanup strategy is mostly a one-size-fits-all approach, lacking adaptive deceleration based on material / rotation rate and multi-machine force-torque balance distribution mechanism, making it difficult to achieve steady-state dragging and attitude control under force constraints. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a space debris near-Earth orbit cleanup system based on the collaboration of BeiDou satellites and unmanned aerial vehicles (UAVs). This system aims to achieve high-frequency observation and integer ambiguity fixation through a combination of a parent satellite's Ku-band phased array radar and BeiDou satellite dual-frequency carrier differential detection, combined with 10Hz real-time filtering and parent satellite clock broadcasting to achieve centimeter-level relative positioning and 1×10⁻⁶ [unclear - possibly referring to a specific measurement or measurement range]. -7 The system achieves time synchronization on the order of seconds; it also determines the confidence level of materials based on observed feature vectors and selects laser or mechanical anchoring strategies according to confidence level / rotation rate. Combined with distributed MPC+CBF and chain force distribution, it significantly improves the safety and success rate of close-range approach, anchoring and multi-machine towing.

[0005] Therefore, this application provides a space debris near-Earth orbit cleanup system based on the collaboration of BeiDou satellites and UAVs, including the following modules: The debris database module is used to configure the parent satellite's Ku-band phased array radar and link it with the imaging payload. It sets the radar scanning period and identification threshold parameters, as well as the echo acquisition, observation timescale, and raw data retention strategies. Based on the echo acquisition and imaging payload, it performs feature extraction, sets the rules for debris entry generation, merging, and updating, performs preliminary material determination and mass estimation based on the feature extraction results, and sets the calculation method for threat scores and the determination of candidate cleanup targets.

[0006] The high-precision positioning module is used to configure the inter-satellite observation link and dual-frequency receiving capability, receive observations for preprocessing, set inter-satellite differential and double-difference to eliminate gross errors, perform integer ambiguity resolution and fast solidification, perform filtering fusion and status output, and set the parent satellite clock broadcast and joint time synchronization strategy.

[0007] The feature determination module is used to configure confidence level and quality interval confirmation, set proximity triggering and confirmation process, set objective function for multi-target task allocation, set constraints and feasible domain limits, set distributed allocation implementation examples and distinguishing descriptions, and set formation pattern generation and release batch command output.

[0008] The proximity control module is used to configure the release window and initial safety parameters, form an initial formation through the release action, perform distributed MPC to gradually approach and coordinate the formation, realize real-time collision and safety constraints through the control barrier function, set the end alignment and anchoring trigger criteria and issue anchoring preparation instructions, execute degradation strategy and fault reconstruction, and write back the task log.

[0009] The material adaptation module is used to configure the laser unit, mechanical anchor, and force sensor. It selects the laser working mode based on material confidence and rotation rate, calculates the laser energy and impulse requirements and converts them into emission parameters, executes laser ablation timing and online fragmentation risk monitoring decisions, performs mechanical anchoring preparation and clamping force setting, and sets and distributes multi-machine anchoring and force distribution rules.

[0010] The chain drag module is used to configure anchor points and work inputs, configure force and torque balance constraints, calculate real-time force distribution, execute force balance closed loop and residual monitoring, set planned track migration and along-line monitoring, execute triggered replacement, multi-machine redistribution and robust reconfiguration, and execute recycling docking, charging and reuse maintenance.

[0011] In some specific embodiments, the fragment library building module specifically includes: Configure the Ku-band phased array radar of the host satellite and link it with the imaging payload. Set the radar scanning period and identification threshold parameters, and set the echo acquisition, observation time scale and raw data retention strategies. Perform feature extraction based on echo acquisition and imaging payload.

[0012] Rules for generating and merging fragment entries are set, preliminary material determination and quality estimation are performed based on feature extraction results, and a method for calculating threat scores and determining candidate cleanup objects is set.

[0013] In some specific embodiments, the high-precision positioning module specifically includes: Configure inter-satellite observation links and dual-frequency receiving capabilities, and preprocess the received observations, setting inter-satellite differential and double-difference to eliminate gross errors.

[0014] Perform integer ambiguity resolution and rapid solidification, perform filtering fusion and status output, and set the host satellite clock broadcast and joint time synchronization strategy.

[0015] In some specific embodiments, the feature determination module specifically includes: Configure confidence level and quality interval confirmation, set up proximity triggering and confirmation process, and set the objective function for multi-objective task allocation.

[0016] Set constraints and feasible domain limits, set examples and distinguishing features for distributed allocation implementation, and set batch command output for formation generation and release.

[0017] In some specific embodiments, the proximity control module specifically includes: Configure the release window and initial security parameters, form an initial formation through the release action, and execute distributed MPC to gradually approach and coordinate with the formation.

[0018] Real-time collision and safety constraints are achieved by controlling the barrier function, end alignment and anchoring trigger criteria are set and anchoring preparation instructions are issued, degradation strategies and fault reconstruction are executed, and the task log is written back.

[0019] In some specific embodiments, the material adaptation module specifically includes: The system is equipped with a laser unit, a mechanical anchor, and a force sensor. The laser operating mode is selected based on material confidence and rotation rate. The laser energy and impulse requirements are calculated and converted into emission parameters.

[0020] Execute laser ablation timing and online fracture risk monitoring decisions, execute mechanical anchoring preparation and clamping force setting, and set and distribute multi-machine anchoring and force distribution rules.

[0021] In some specific embodiments, the chain drag-and-drop module specifically includes: Configure anchor points and operation inputs, configure force and torque balance constraints, calculate real-time force distribution, and execute force balance closed-loop and residual monitoring.

[0022] Before performing chain drag, configure and read the following input data: spatial position vector of each anchor point, real-time clamping force measurement of each drone, and maximum sustainable pulling force of each drone. The target debris mass estimation range and target migration requirements, the home planet recovery window and recharging capacity. Establish system-level force and moment balance equations as hard constraints for drag control: Translational balance:

[0023] In the formula: For the first The tensile force vector exerted by the Taiwanese drone on the debris. For the actual mass of the fragments, The desired drag acceleration vector.

[0024] Torque balance:

[0025] In the formula: For the center of mass of the fragment to the first The position vector of the anchor point of the Taiwanese drone. Let the moment of inertia matrix of the fragment be... This is the angular acceleration vector of the fragment.

[0026] To satisfy the equality constraints and minimize the risk of single-machine overload, a real-time secondary optimization solution is performed.

[0027] After the force distribution is solved, each participating UAV performs closed-loop force tracking control: reading the force sensor output and... PI / PID Alternatively, the propulsion command can be adjusted based on the model-based force controller, and the system continuously calculates the force residual and sends the residual to the monitoring module.

[0028] Set up planned track migration and along-line monitoring, execute triggered replacement, multi-machine redistribution and robust reconfiguration, and perform recovery docking, charging and reuse maintenance.

[0029] Based on the force distribution and desired acceleration Calculate the orbital migration trajectory and the expected cumulative .

[0030] Fault detection employs two mechanisms: residual detection and majority voting residual detection. To determine if a single drone has failed, a majority voting mechanism is used where participating drones vote on key observations. If the majority vote exceeds 50%, the system is considered to be in a state of anomaly.

[0031] Replacement and Reconstruction Rules: If a single machine is determined to be faulty, the required number of participants will be recalculated according to the multi-machine allocation rules. The system uses a distributed task reconfiguration algorithm to select a replacement drone from the in-service drones to take over the anchor point or compensate for the force. If the replacement drone is insufficient to maintain safe towing, it will perform evacuation or abort and return to the home planet according to priority.

[0032] After the target migration and cleanup tasks are completed, the return and reuse process is executed: each UAV returns and docks according to the provided high-precision positioning and time reference, with the optimal docking tolerance being position. And attitude error The mother planet allocates charging resources to each docked drone and optimizes the charging power. It implements modular self-inspection and maintenance processes and records maintenance checklists, with the design goal of ensuring that the drone can be reused at least once. Second-rate.

[0033] During the drag-and-drop and migration process, the system periodically outputs the following fields to the task planning module and the fragment item set: real-time gripping force of each drone. Distributed tension residual force Current force distribution scheme, cumulative The system estimates the health level of anchor points and records events. When a policy switch or fault reconstruction occurs, it immediately writes back the data and notifies the task planning module for dynamic rescheduling.

[0034] In summary, the space debris near-Earth orbit cleanup system based on BeiDou satellite and UAV collaboration provided in this application achieves a closed-loop capability with a response time of no more than 30 minutes from identification to release, 10Hz relative position / velocity updates, and end-effector alignment by constructing a dynamic debris database covering the target orbit, rapidly solidifying integer ambiguities based on inter-satellite dual differences and LAMBDA, and a material-adaptive laser / mechanical anchoring process driven by material confidence and mass range. The accompanying chain-drag force-torque balance secondary planning and multi-UAV reconfiguration mechanism enable stable orbit migration or recovery under thrust-limited and chain-collaboration conditions, and support UAV reuse and fault-robust reconfiguration, thereby reducing secondary debris generation and improving cleanup efficiency and on-orbit reuse rate. Attached Figure Description

[0035] Figure 1 This is an overall framework diagram of a space debris near-Earth orbit cleanup system based on the collaboration of BeiDou satellites and drones, provided in the embodiments of this application. Detailed Implementation

[0036] Please refer to Figure 1 This illustrates the process of one embodiment of a space debris near-Earth orbit cleanup system based on the collaboration of BeiDou satellites and drones, according to this disclosure.

[0037] like Figure 1 As shown, the space debris near-Earth orbit cleanup system based on the collaboration of BeiDou satellites and drones includes the following steps: The debris database module is used to configure the parent satellite's Ku-band phased array radar and link it with the imaging payload. It sets the radar scanning period and identification threshold parameters, as well as the echo acquisition, observation timescale, and raw data retention strategies. Based on the echo acquisition and imaging payload, it performs feature extraction, sets the rules for debris entry generation, merging, and updating, performs preliminary material determination and mass estimation based on the feature extraction results, and sets the calculation method for threat scores and the determination of candidate cleanup targets.

[0038] The high-precision positioning module is used to configure the inter-satellite observation link and dual-frequency receiving capability, receive observations for preprocessing, set inter-satellite differential and double-difference to eliminate gross errors, perform integer ambiguity resolution and fast solidification, perform filtering fusion and status output, and set the parent satellite clock broadcast and joint time synchronization strategy.

[0039] The feature determination module is used to configure confidence level and quality interval confirmation, set proximity triggering and confirmation process, set objective function for multi-target task allocation, set constraints and feasible domain limits, set distributed allocation implementation examples and distinguishing descriptions, and set formation pattern generation and release batch command output.

[0040] The proximity control module is used to configure the release window and initial safety parameters, form an initial formation through the release action, perform distributed MPC to gradually approach and coordinate the formation, realize real-time collision and safety constraints through the control barrier function, set the end alignment and anchoring trigger criteria and issue anchoring preparation instructions, execute degradation strategy and fault reconstruction, and write back the task log.

[0041] The material adaptation module is used to configure the laser unit, mechanical anchor, and force sensor. It selects the laser working mode based on material confidence and rotation rate, calculates the laser energy and impulse requirements and converts them into emission parameters, executes laser ablation timing and online fragmentation risk monitoring decisions, performs mechanical anchoring preparation and clamping force setting, and sets and distributes multi-machine anchoring and force distribution rules.

[0042] The chain drag module is used to configure anchor points and work inputs, configure force and torque balance constraints, calculate real-time force distribution, execute force balance closed loop and residual monitoring, set planned track migration and along-line monitoring, execute triggered replacement, multi-machine redistribution and robust reconfiguration, and execute recycling docking, charging and reuse maintenance.

[0043] In some specific embodiments, the fragment library building module specifically includes: Configure the Ku-band phased array radar of the host satellite and link it with the imaging payload. Set the radar scanning period and identification threshold parameters, and set the echo acquisition, observation time scale and raw data retention strategies. Perform feature extraction based on echo acquisition and imaging payload.

[0044] A Ku-band phased array radar is configured on the parent satellite platform. When the system has an optical imaging payload or an infrared imaging payload available, the parent satellite platform triggers the optical imaging payload or the infrared imaging payload to perform linked observations within the radar observation time window.

[0045] The mothership Ku-band phased array radar refers to a radar unit deployed on the mothership platform, operating in the Ku-band frequency band, and possessing phased array scanning capability and pulse compression echo acquisition function. The linkage observation time window refers to the tolerable time window during radar sector scanning and before and after radar observation, used to ensure time matching between imaging and radar observation.

[0046] The single sector scan period of the Ku-band phased array radar of the parent satellite is set to 10 minutes, the spatial resolution of the radar is limited to no more than 0.1 meters, and the minimum target identification size threshold of the radar is limited to 0.10 meters.

[0047] The parent satellite's Ku-band phased array radar acquires echo data using pulse compression. The echo data includes echo amplitude, carrier phase, polarization channel data, and Doppler spectrum. The parent satellite platform generates an observation time tag for each observation, with a time accuracy of no more than 0.1 seconds. The parent satellite platform retains the raw echo data in the database for no less than 72 hours and connects it to subsequently generated fragment entries with a searchable index.

[0048] Based on the echo data, pulse compression and Doppler spectrum estimation are performed to obtain the equivalent radar cross-section estimate, polarization ratio, spectral width, and spectral peak information, respectively. Based on the optical imaging data or the infrared imaging data, radiometric calibration and geometric correction are performed to obtain reflectivity spectral band characteristics or brightness temperature curves. The radar characteristics and the imaging characteristics constitute an observation feature vector, which includes the following fields: RCS, polarization ratio, spectral width, spectral peak information, reflectivity spectral band characteristics, imaging boundary size estimation, and observation time stamp.

[0049] Using the measured value of the observed feature vector as the observation input, the orbital state vector of the debris is filtered and estimated. The orbital state vector consists of the position vector in the geocentric inertial coordinate system, the velocity vector in the geocentric inertial coordinate system, and the rotation rate.

[0050] The filtered estimation uses an extended Kalman filter or an unscented Kalman filter to perform prediction and observation updates. The filtered estimation results must output the position covariance matrix and velocity covariance matrix as uncertainty measures, and the orbital state estimation and uncertainty are updated at a frequency of no less than once every 10 minutes.

[0051] Rules for generating and merging fragment entries are set, preliminary material determination and quality estimation are performed based on feature extraction results, and a method for calculating threat scores and determining candidate cleanup objects is set.

[0052] For each observation that meets the identification threshold, a fragment entry is generated and written into the fragment library. The fragment entry includes the following fields: fragment unique identifier, observation time tag, geocentric inertial coordinate system position vector, geocentric inertial coordinate system velocity vector, size estimate, shape estimate, rotation rate estimate, preliminary material estimate, position covariance matrix, velocity covariance matrix, observation source identifier, and original echo index.

[0053] When a new observation is less than 0.5 meters away from an existing debris entry and the observation time interval is less than 30 minutes, the new observation is incorporated into the existing debris entry, and the position and velocity of the entry are updated with the latest estimate from the filter, and the uncertainty and observation count of the entry are also updated.

[0054] The fragments are identified by a material determination method based on the observed feature vector. The material categories include at least: aluminum alloy, steel, composite materials and ceramics. The material determination outputs a material classification label and a confidence value.

[0055] When the confidence value is greater than or equal to 0.8, the material determination result serves as the direct basis for subsequent approach and cleanup strategies. When the confidence value is between 0.5 and 0.8, close-range confirmation is triggered to obtain further spectral or imaging data to ultimately determine the material category. A mass estimation interval is calculated based on the size estimate and the typical density range of the material category. The mass estimation is specifically obtained as follows:

[0056] In the formula: This is an estimated mass value for the fragment. The density value is a typical value for the material category. The volume estimate is the value of the fragment, which is calculated from the size estimate based on a preset geometric assumption. The mass estimate range includes the uncertainty range formed by the upper and lower limits of the typical density of the material and the uncertainty of the size estimate.

[0057] The debris threat score is defined as the weighted sum of the debris size normalized score and the debris collision probability normalized score, and the threat score is expressed as:

[0058] In the formula: Assess the threat level of the fragments. As the weight of the size normalization score, The weights of the collision probability normalization score, To score for size normalization, To normalize the collision probability score, in one embodiment, the weight example is taken as a value. , The size normalization score With the collision probability normalized score The specific normalization method can be defined by the task planning module and used as the preferred implementation method.

[0059] The rule for identifying the fragments as candidate cleanup objects is as follows: If... If the fragment is marked as a candidate for cleanup, the independent rule is: if Miqie Then it will be forcibly included in the candidate cleanup list, among which: For the size estimation of the fragments, Let the collision probability between the fragment and the potential target be given. The collision probability is calculated by taking the position covariance matrix, velocity covariance matrix, fragment size estimate, and potential target state uncertainty as inputs, and obtaining them through the covariance intersection or close-range encounter radius method. The symbols used in this claim are used to represent specific solutions, which can be described as preferred embodiments in the specification.

[0060] The set of debris entries outputs a list of candidate cleanup targets to the task planning module. The output fields include: unique debris identifier, latest orbital state estimate, position covariance matrix, velocity covariance matrix, mass estimation range, material label and confidence level, threat score and original echo index. The output adopts a structured machine-readable data format and controls the access permissions of the original echo data according to the task security level.

[0061] In some specific embodiments, the high-precision positioning module specifically includes: Configure inter-satellite observation links and dual-frequency receiving capabilities, and preprocess the received observations, setting inter-satellite differential and double-difference to eliminate gross errors.

[0062] The mother satellite platform and the UAV are equipped with their own BeiDou receiving units to establish inter-satellite link observation capabilities. The BeiDou receiving units have dual-frequency carrier phase observation capabilities and can output carrier phase raw observation streams and code pseudorange observation streams.

[0063] BeiDou inter-satellite link dual-frequency observation refers to the dual-frequency carrier phase observation and code pseudorange observation obtained through BeiDou satellite inter-satellite links or satellite-platform links, which are used to achieve high-precision differential positioning and time synchronization. The carrier phase raw observation stream refers to the sequence of continuous carrier phase measurement values ​​output by the receiving unit according to the observation period.

[0064] The observations of a single satellite receiver pair are described using a carrier phase observation model, which is expressed as follows:

[0065] In the formula: For carrier phase observations, The geometric distance between the satellite and the receiver. The wavelength of the carrier wave. The integer ambiguity of the carrier phase. To detect noise and other minor error terms, cycle slip detection and observation integrity verification are performed on the carrier phase observation, while code pseudorange observation is retained for subsequent ambiguity initial value estimation and scale constraint.

[0066] To eliminate common errors such as satellite clock bias and receiver clock bias, inter-satellite differential and double-difference processing are employed: First, single-difference processing is performed on the carrier phase of the two satellites by the same receiver to remove receiver clock bias. Then, the single-difference values ​​of the two receivers for the same satellite are subtracted to form a double-difference observation. An example of double-difference observation is as follows:

[0067] In the formula: These are double-difference carrier phase observations. The difference is the geometric distance between the two differences. For double-difference integer ambiguity, This is double-difference observation noise.

[0068] Based on the aforementioned double-difference observations, the inter-satellite double-difference method is used to eliminate common terms caused by large-scale clock errors and propagation, thereby making the integer ambiguity resolution a feasible integer estimation problem.

[0069] The inter-satellite differential and double-difference steps described herein are one of the innovative implementation details of this application, and are particularly significant in achieving real-time double-difference processing and subsequent integer ambiguity fixation in an on-orbit inter-satellite link environment.

[0070] Perform integer ambiguity resolution and rapid solidification, perform filtering fusion and status output, and set the host satellite clock broadcast and joint time synchronization strategy.

[0071] For the double-difference integer ambiguity Integer least squares analysis is performed to obtain integer solutions. This integer analysis can be implemented using the LAMBDA algorithm, and the integer analysis steps can be represented as minimizing the cost function:

[0072] In the formula: The integer ambiguity vector solution obtained from the analysis. This is the floating-point ambiguity estimation vector. The floating-point ambiguity covariance matrix. express n 3D integer space.

[0073] The integer resolution is fixed when an acceptable confidence level is reached, and the fixed integer ambiguity is used to transform the double-difference observation into a precise geometric baseline solution, thereby achieving centimeter-level relative positioning. Preferably, the success rate of fixing the integer ambiguity is completed within 1 second or less to support subsequent high-frequency positioning updates at 10Hz.

[0074] Based on the double-difference observations with fixed or floating-point ambiguity, the carrier phase and code pseudorange observations are input into a real-time filter to estimate the relative position and relative velocity. The real-time filter can be an extended Kalman filter or an unscented Kalman filter. The filter outputs at least the following state quantities and updates them at a frequency of 10 Hz: relative position vector, relative velocity vector, and positioning covariance matrix. When the ambiguity cannot be fixed in real time, the filter maintains the continuity of the solution in a floating-point ambiguity manner and outputs a suboptimal accuracy solution.

[0075] The filter design should include a state prediction model and a measurement update model. The state prediction model should incorporate a relative dynamics model and possible non-conservative disturbances to ensure the continuity of the solution under short-term occlusion or short-term link loss.

[0076] Using the reference clock of the home satellite platform as the system time base, the home satellite platform periodically broadcasts time synchronization packets to the UAV swarm and other on-orbit nodes. The time synchronization packets contain reference time identifiers and clock offset correction information. The receiving end corrects its local clock based on the propagation delay estimation of the time synchronization packets and carrier observations. The time offset correction can be represented by a simplified model:

[0077] In the formula: This is the time corrected to the receiver's local time. Reference time for the home planet For receiver clock deviation, To estimate propagation delay, the receiver clock offset is controlled within a certain range using the joint time synchronization strategy. Within the range of seconds, it supports 10Hz relative positioning updates and high-precision timing control.

[0078] When the ambiguity cannot be fixed within the preset time window, or the covariance of the floating-point ambiguity solution exceeds the quality threshold, the system should output the current best floating-point solution in a degraded form and note the accuracy level of the solution in the output. The output interface provides the following fields to the proximity control module and the task planning module in a structured format: relative position vector, relative velocity vector, positioning covariance matrix, ambiguity fixation flag, time synchronization deviation value, and observation link status indication.

[0079] In some specific embodiments, the feature determination module specifically includes: Configure confidence level and quality interval confirmation, set up proximity triggering and confirmation process, and set the objective function for multi-objective task allocation.

[0080] The generated set of debris entries and the output high-precision relative positioning and time reference are used as inputs, including but not limited to: debris entry fields, the remaining propellant / remaining energy / health status of the UAV swarm, the home planet recovery window, and ground mission priorities.

[0081] Online classification is performed based on the observed feature vectors of the fragment entries to output material confidence scores. With material labels, when At that time, the material label serves as a direct basis for the approach and cleaning strategy. At that time, a close-up verification procedure is triggered to obtain further spectral or imaging data to ultimately confirm the material label.

[0082] in: The confidence level for material determination, with a range of values. Simultaneously, based on the size estimate and the typical density range of the material, a mass estimation range is output. This interval is then written into the fragment entry for subsequent decision-making.

[0083] When close-range confirmation is triggered, the mission planning module generates a close-range observation task and assigns at least one UAV to perform close-range confirmation observations. Close-range confirmation observations include short-range spectral sampling or high-resolution imaging. The close-range confirmation results will be written back to the fragment entries with observation timestamps and updated. And the quality estimation range.

[0084] Using the candidate fragment set and the drone swarm status as input, a multi-objective optimization objective function for task allocation is formulated, specifically as follows:

[0085] In the formula: For the overall goal, To be assigned to the The estimated total speed change required for the Taiwanese drone to approach and complete its mission. For the first The estimated time for the Taiwanese drone to complete its assigned task. For the first Quantitative indicators of the risks of drone missions carried out by Taiwan These are non-negative weighting coefficients, set by the task planning module according to task priority. The selected values ​​are: The weights are adjustable parameters, and the objective function is solved according to the constraints to obtain the optimal task allocation.

[0086] Set constraints and feasible domain limits, set examples and distinguishing features for distributed allocation implementation, and set batch command output for formation generation and release.

[0087] The multi-objective optimization is solved under the constraint that the estimated propellant consumption of each UAV must not exceed its remaining propellant budget, denoted by: The number of cooperating drones required for a single fragment must not exceed the number of available drones and must not be less than the required lower limit, denoted by the following symbol: The task execution window is subject to time constraints; the task execution window must fall within the time interval specified by the mother planet recycling window and the task priority.

[0088] in: For the first Taiwan's remaining budget limit for drone transmission. The number of drones assigned to a certain fragment. The minimum number of collaborative drones required for this fragment.

[0089] The task allocation is preferably implemented as a distributed auction allocation to improve robustness and online rescheduling capability. The distributed auction allocation can be CBBA or a functionally equivalent distributed matching algorithm. The distributed implementation allows each UAV to base its allocation on local estimations. , The remaining energy participates in the bidding, and the task allocation result is finally generated through consensus convergence. The distributed implementation has distinctive benefits in on-orbit high latency / link instability scenarios.

[0090] Based on the optimal allocation results, a task allocation table is generated for each UAV. The task allocation table includes fields such as: target fragment ID, predetermined approach time window, and estimated time window. Estimate execution time The required anchor point location list, initial formation suggestions, release batch number and priority, and the task allocation table must be sent to the mother satellite release module or UAV ground dispatch unit no later than 30 minutes after task identification.

[0091] The task allocation table is output to the proximity control module and the home satellite release module, and the close-range confirmation results, actual observations, and execution feedback should be written back to the debris entry set for updating. Quality estimation range and threat score are used to support the replanning of subsequent tasks.

[0092] In some specific embodiments, the proximity control module specifically includes: Configure the release window and initial security parameters, form an initial formation through the release action, and execute distributed MPC to gradually approach and coordinate with the formation.

[0093] The home satellite platform is configured to select a release window based on the mission allocation table and the capabilities of the home satellite release module, and initial safety parameters after release are set, including the minimum safe separation distance between UAVs after release and the maximum permissible relative speed at the moment of release. These parameters are set as follows: ， ,in, The minimum safe separation distance between any two drones after release. To release the relative velocity between any UAV and its nearest neighbor or reference point at the instant of release, the configuration ensures that the initial state of the formation after release has sufficient geometric spacing and low relative velocity, so as to reduce the collision and control complexity in the subsequent close encounter phase.

[0094] The UAVs are released in batches by the mothership platform to form a predetermined initial formation, and local approach commands are issued to each released UAV. The local approach commands include the following fields: target fragment ID, predetermined approach time window, expected terminal relative position, expected terminal attitude, expected terminal relative speed and priority identifier.

[0095] Distributed MPC is executed to achieve precise approach under constraints of limited propulsion and multi-machine cooperation. The discretized dynamics model and prediction cost function are as follows:

[0096] In the formula: For drones at any time The relative state vector includes position error, velocity error, and attitude error. The control input vector includes propulsion pulses and attitude control commands. For the system discretization matrix, This represents the model perturbation term. The prediction cost function is:

[0097] In the formula: For reference end state, This is a weighted matrix for state and control. To predict the number of time-domain steps, the system calculates the future steps in each control cycle using the aforementioned MPC. The optimal control sequence for the first step, and only the first step Step-by-step control is implemented, followed by rolling recalculation. In distributed implementation, each UAV performs local calculations while exchanging necessary boundary information with neighboring UAVs to achieve formation coordination. The parameters are: , .

[0098] Real-time collision and safety constraints are achieved by controlling the barrier function, end alignment and anchoring trigger criteria are set and anchoring preparation instructions are issued, degradation strategies and fault reconstruction are executed, and the task log is written back.

[0099] By defining a barrier function based on the distance between the two bodies and embedding it as a real-time inequality constraint into MPC, the minimum safe margin is guaranteed.

[0100] And apply the CBF condition:

[0101] In the formula: drones and The position vector, As a positive definite phase function, the textual explanation states that using the CBF condition as an inequality constraint in MPC ensures that, even under propulsion limitations or sudden disturbances, the distance between UAVs remains no less than [value missing] during the control solution process. The innovative implementation of this CBF combined with distributed MPC is a key innovation of this system, which helps to achieve real-time security assurance during the multi-machine collaborative approach phase.

[0102] An end-point alignment criteria are set. An anchoring preparation signal is issued and the anchor deployment procedure begins when the following conditions are simultaneously met: position error threshold, attitude error threshold, end-point relative velocity threshold, and positioning uncertainty threshold. The criteria are as follows:

[0103]

[0104]

[0105]

[0106] In the formula: Let be the end position error vector. For end-effector attitude error, The terminal relative rate, Let be the location covariance matrix. The threshold for the location covariance trace value is preferably... If the above criteria are met, it indicates that the UAV has reached a state where it can be safely contacted and anchored in terms of spatial position, attitude and speed, and the system will then switch to the anchoring execution sequence.

[0107] Degradation conditions and handling procedures are set: When the positioning uncertainty or ambiguity quality index exceeds the safety threshold, or when a single drone's propulsion / attitude control malfunctions, the following actions are taken in priority order: suspend approach and return to the predetermined standby formation, replace the drone performing the operation, or terminate the entire mission and withdraw. An example of the degradation judgment is provided below.

[0108] In the formula: The preferred threshold for the positional uncertainty of task termination is... When performing a downgrade, the system automatically triggers a distributed task refactoring process to reallocate tasks or call up backup drones. At the same time, the status, diagnostic reasons, and real-time logs are written back to the fragment entry set and task logs for auditing and subsequent learning.

[0109] The system outputs propulsion / attitude control commands, CBF safety constraint status flags, anchoring preparation flags, and current positioning uncertainty indicators for each control cycle to the proximity execution unit, and feeds back key events of the proximity process to the mission planning and debris item set.

[0110] In some specific embodiments, the material adaptation module specifically includes: The system is equipped with a laser unit, a mechanical anchor, and a force sensor. The laser operating mode is selected based on material confidence and rotation rate. The laser energy and impulse requirements are calculated and converted into emission parameters.

[0111] Before performing the cleaning task, the laser unit, mechanical anchor hook and the force sensor on the anchor hook are configured and verified. The laser unit has adjustable peak value and preferably operates at a wavelength of 1064nm with an output power range of 100W to 1kW. The mechanical anchor hook has an extension and retraction mechanism and is equipped with a real-time force sensor. The adjustable range of the anchor hook clamping force is preferably 50N to 200N, and the clamping force measurement accuracy is preferably ±5N.

[0112] By material confidence With fragment rotation rate Determine the laser operating mode: When and When laser deceleration is the primary mode, it should be selected first. or When using this method, prioritize a hybrid or mechanically-dominated mode that combines close-range confirmation with mechanical anchoring.

[0113] in: To determine the confidence level of the material, Let be the angular velocity of the fragment's rotation around its center of mass. The adaptive selection strategy, which uses the rotation rate switching threshold as a distinguishing factor, selects different laser timing sequences or directly switches to mechanical anchoring based on material and dynamic characteristics to reduce the risk of secondary fragmentation and failure.

[0114] To achieve the required deceleration Using the energy-impulse empirical coefficient Estimate the required laser energy Specifically:

[0115] From the above formula, we can derive:

[0116] In the formula: The rate at which the target fragment needs to be altered during the ablation phase. The energy and impulse coefficients are calibrated by ground experiments and recorded according to the material type. The total laser energy required. The estimated mass of the fragments is taken from the center or a conservative upper limit of the mass estimation interval. Mapped to transmit power-time parameters:

[0117] In the formula: The average emission power is limited by the laser unit's capabilities and mode. The above describes the engineering energy-impulse estimation method for this system, representing the total irradiation duration. The empirical coefficients obtained from ground calibration are innovative refinements and can be written as subordinate claims or preferred embodiments.

[0118] Execute laser ablation timing and online fracture risk monitoring decisions, execute mechanical anchoring preparation and clamping force setting, and set and distribute multi-machine anchoring and force distribution rules.

[0119] During laser ablation, the power and pulse parameters are controlled simultaneously according to the selected operating mode to activate the fragmentation / debris monitoring circuit and estimate the fragmentation probability index. Set decision rules: If Then the laser will stop and switch to mechanical control mode. in: To estimate the probability of fragmentation in real time, The preferred example for determining the fragmentation threshold is as follows: The probability of fragmentation can be estimated by online sensor indicators and ground / airborne material brittleness parameters after correction. If the risk is high, lasering should be avoided to prevent the generation of many small fragments. This online decision-making logic is a key point for differentiation and safety.

[0120] Laser deceleration reaches or approaches the target Afterwards, or when directly selecting mechanical mode, perform anchoring preparation: deploy the anchor hook mechanism and approach the anchoring point; at the moment of contact, calculate the mass range and the desired drag acceleration. Calculate the required clamping force Specifically:

[0121] In the formula: For the required clamping force, For the expected drag acceleration, an example is... , For safety reasons, the preferred option is... To cover the uncertainty of the force.

[0122] like If the clamping force of a single machine is within its upper limit, then single-machine clamping is sufficient; otherwise, multi-machine anchoring is performed. During anchoring, the force sensor is read in real time, and closed-loop control is used to bring the actual clamping force closer to the limit. If the clamping error is less than ±5N, the clamping is recorded as successful.

[0123] When the clamping force of a single machine is insufficient At that time, the minimum number of drones required to participate will be determined. Specifically:

[0124] In the formula: The maximum sustainable pulling force (N) of a single UAV is determined by the UAV's structure and propulsion capability and recorded in the status field. The total dragging force is then distributed according to the capabilities of multiple UAVs, and torque balance is considered in the distribution to avoid generating adverse overturning torque. The torque balance criterion is: selecting the anchor point position and the force direction of each UAV to minimize the net torque with respect to the centroid of the fragment. The force distribution rule and torque balance criterion of multi-UAV anchoring are the implementation innovations of this application, which are used to ensure the stability of the fragment attitude during chain dragging.

[0125] Laser power timing and emitted energy , estimate the generation of The anchor point sequence, the real-time clamping force of each drone, and the clamping success flag are written back to the fragment entry set and the task log. The task planning module is notified in real time when a strategy switch occurs so that dynamic rescheduling can be carried out.

[0126] In some specific embodiments, the chain drag-and-drop module specifically includes: Configure anchor points and operation inputs, configure force and torque balance constraints, calculate real-time force distribution, and execute force balance closed-loop and residual monitoring.

[0127] Before performing chain drag, configure and read the following input data: spatial position vector of each anchor point, real-time clamping force measurement of each drone, and maximum sustainable pulling force of each drone. The target debris mass estimation range and target migration requirements, the home planet recovery window and recharging capacity. Establish system-level force and moment balance equations as hard constraints for drag control: Translational balance:

[0128] In the formula: For the first The tensile force vector exerted by the Taiwanese drone on the debris. For the actual mass of the fragments, The desired drag acceleration vector.

[0129] Torque balance:

[0130] In the formula: For the center of mass of the fragment to the first The position vector of the anchor point of the Taiwanese drone. Let the moment of inertia matrix of the fragment be... This is the angular acceleration vector of the fragment.

[0131] To satisfy the equality constraints and minimize the risk of single-machine overload, a real-time quadratic optimization solution is performed, including the objective function and the constraints, specifically: The objective function is:

[0132] The constraints are:

[0133] Furthermore, the following inequality constraints are applied to each drone:

[0134] In the formula: To determine the real-time availability coefficient, the above problem is a quadratic programming problem with linear equations and a quadratic objective. It is solved in real time within a distributed or centralized controller, with an update frequency of 1Hz–10Hz.

[0135] After the force distribution solution is obtained, each participating UAV performs closed-loop force tracking control: it reads the force sensor output and adjusts the propulsion command using a PI / PID or model-based force controller. The system continuously calculates the force residuals and sends them to the monitoring module. Specifically:

[0136] In the formula: For the first The actual clamping / tension force measured by the Taiwanese drone The target tension issued by the controller, if Exceeding the threshold And the duration exceeds If this occurs, a replacement or compensation procedure will be triggered.

[0137] Set up planned track migration and along-line monitoring, execute triggered replacement, multi-machine redistribution and robust reconfiguration, and perform recovery docking, charging and reuse maintenance.

[0138] Based on the force distribution and desired acceleration Calculate the orbital migration trajectory and the expected cumulative Specifically:

[0139] In the formula: To drag from startup to time Cumulative change in velocity, example of a migration decision rule: if quality estimation Then, the priority is to move towards the cemetery track. And if the reentry conditions are met, the reentry plan is evaluated, taking into account ground recovery capabilities. Monitoring along the route includes: anchor point wear estimation, debris attitude changes, and anchor point displacement deviations. If the monitored indicators exceed the safety threshold, adjustments are made immediately. Or it may trigger the termination of the downgrade.

[0140] Fault detection employs two mechanisms: residual detection and majority voting residual detection. To determine if a single drone has failed, a majority voting mechanism is used where participating drones vote on key observations. If the majority vote exceeds 50%, the system is considered to be in a state of anomaly.

[0141] Replacement and Reconstruction Rules: If a single machine is determined to be faulty, the required number of participants will be recalculated according to the multi-machine allocation rules. The system uses a distributed task reconfiguration algorithm to select a replacement drone from the in-service drones to take over the anchor point or compensate for the force. If the replacement drone is insufficient to maintain safe towing, it will perform evacuation or abort and return to the home planet according to priority.

[0142] After the target migration and cleanup tasks are completed, the return and reuse process is executed: each UAV returns and docks according to the provided high-precision positioning and time reference, with the optimal docking tolerance being position. And attitude error The mother planet allocates charging resources to each docked drone and optimizes the charging power. It implements modular self-inspection and maintenance processes and records maintenance checklists, with the design goal of ensuring that the drone can be reused at least once. Second-rate.

[0143] During the drag-and-drop and migration process, the system periodically outputs the following fields to the task planning module and the fragment item set: real-time gripping force of each drone. Distributed tension residual force Current force distribution scheme, cumulative The system estimates the health level of anchor points and records events. When a policy switch or fault reconstruction occurs, it immediately writes back the data and notifies the task planning module for dynamic rescheduling.

[0144] In practical applications, the above describes a system where a Ku-band phased array radar is configured on the home satellite platform, linked with optical and infrared payloads. The single sector scan cycle is set to ten minutes, the spatial resolution is no greater than 0.1 meters, and the minimum target identification size is limited to 0.1 meters. The home satellite generates a time stamp for each observation and stores the raw echo data in the database for no less than seventy-two hours. The observation results are processed through pulse compression and Doppler spectral analysis to generate observation feature vectors, which are then written into debris entries. These debris entries include fields such as a unique identifier, observation time, geocentric inertial coordinate system position and velocity estimation, size and shape estimation, rotation rate estimation, radar cross-section estimation, preliminary material determination, and uncertainty measurement.

[0145] Dual-frequency BeiDou receiver units are deployed on the mother satellite and each UAV to obtain carrier phase and code pseudorange observations. Inter-satellite differential and double-difference processing are used, and the ambiguity is quickly resolved and solidified using the integer least squares method to obtain centimeter-level relative positioning. The relative position and velocity estimates are output at a frequency of ten times per second. The mother satellite broadcasts time synchronization packets and corrects the propagation delay of the receiver clock to make the system time synchronization reach the order of 10 to the power of negative 7 seconds. If the integer solidification step does not meet the standard in a short time, the system outputs floating-point solutions and marks the current accuracy level in the data for subsequent control downgrade use.

[0146] An online classifier is run based on the observed feature vectors to output material confidence values ​​and material labels. When the material confidence is greater than or equal to 0.8, the material label is used as the direct basis for the approach and cleanup strategy. If the material confidence is between 0.5 and 0.8, close-range confirmation observation is triggered to supplement spectral or high-resolution imaging. Task allocation aims to minimize the overall cost. The cost comprehensively considers the required speed change, task completion time, and execution risk. The example weights are 50% for speed change, 30% for time, and 20% for risk. The allocation is carried out within 30 minutes under the conditions of meeting the propellant budget, formation size, and time window constraints.

[0147] The mothership is released in batches according to the mission allocation to form an initial formation. After release, the minimum safe separation distance between any two UAVs must not be less than five meters. At the moment of release, any relative speed must not exceed 0.2 meters per second. Each UAV runs a distributed model predictive controller to gradually approach. The prediction time domain is about two seconds and the control execution time domain is about 0.5 seconds. The control barrier function is embedded in the constraint to ensure real-time collision limiting. The position error of the approach terminal must not be greater than 0.1 meters, the attitude error must not exceed two degrees, and the terminal relative speed must not exceed 0.05 meters per second. After the above conditions are met, the UAV enters the anchoring preparation sequence.

[0148] Based on the material confidence level and fragment rotation rate, a laser ablation or mechanically driven cleaning scheme is selected: when the material confidence level is not lower than 0.8 and the rotation rate does not exceed one arcsecond per second, laser deceleration is preferred; otherwise, mechanical anchoring or a hybrid strategy is implemented after close-range confirmation. The total laser energy and irradiation time required are estimated based on the energy-impulse empirical coefficient calibrated by ground tests. The fragmentation risk index is monitored online. If the fragmentation risk index reaches or exceeds 0.2, the laser is immediately stopped and mechanical anchoring is switched to. After the anchor extends, the clamping force is set according to the mass estimation and expected drag acceleration. The clamping force range is set to 50 Newtons to 200 Newtons. The clamping control closed-loop accuracy requirement is ±5 Newtons. The dragging stage can only begin after the force sensor confirms successful clamping.

[0149] During the chain-dragging phase, force and torque balance are used as hard constraints. That is, the resultant force of the pulling force applied by each UAV is equal to the mass of the fragment multiplied by the set dragging acceleration, and the total torque generated by each anchor point is kept approximately zero about the center of mass of the fragment to maintain attitude stability. Based on the above constraints, real-time secondary planning is performed to minimize the load ratio of each UAV and meet the upper limit of the sustainable pulling force of each UAV and the real-time availability coefficient. The UAV performs force tracking closed-loop execution and monitors the force residual. The residual judgment threshold is the larger of five Newtons and one-tenth of the sustainable pulling force of a single UAV. If the residual exceeds the threshold continuously for five seconds, a replacement or compensation procedure is triggered and the participating UAVs are reassigned using a distributed reconstruction algorithm. After the mission is completed, each UAV returns and docks with the home satellite with high-precision positioning. The position docking tolerance is limited to 0.02 meters and the attitude error is limited to ±1 degree. The home satellite provides the docking UAV with a charging capacity of no less than 2 kilowatts and performs modular self-checks and maintenance. The goal is to achieve at least ten reuses of UAVs and mission robustness with controllable faults.

Claims

1. A space debris near-Earth orbit cleanup system based on the coordinated use of BeiDou satellites and unmanned aerial vehicles, characterized in that: Includes the following modules: The debris database module is used to configure the Ku-band phased array radar of the parent satellite and link it with the imaging payload. It sets the radar scanning cycle and identification threshold parameters, and sets the echo acquisition, observation time scale and raw data retention strategies. It performs feature extraction based on echo acquisition and imaging payload, sets the rules for debris entry generation and merging update, performs preliminary material judgment and mass estimation based on feature extraction results, and sets the calculation method for threat score and the judgment of candidate cleanup objects. The high-precision positioning module is used to configure the inter-satellite observation link and dual-frequency receiving capability, receive observations for preprocessing, set inter-satellite differential and dual-difference to eliminate gross errors, perform integer ambiguity resolution and fast solidification, perform filtering fusion and status output, and set the parent satellite clock broadcast and joint time synchronization strategy. The feature determination module is used to configure confidence level and quality interval confirmation, set the proximity triggering and confirmation process, set the objective function for multi-objective task allocation, set constraints and feasible domain limits, set distributed allocation implementation examples and distinguishing descriptions, and set formation pattern generation and release batch command output. The proximity control module is used to configure the release window and initial safety parameters, form an initial formation through the release action, execute distributed MPC to gradually approach and coordinate formation, realize real-time collision and safety constraints through the control barrier function, set the end alignment and anchoring trigger criteria and issue anchoring preparation instructions, execute degradation strategy and fault reconstruction, and write back the task log. The material adaptation module is used to configure the laser unit, mechanical anchor, and force sensor. It selects the laser working mode based on material confidence and rotation rate, calculates the laser energy and impulse requirements and converts them into emission parameters, executes laser ablation timing and online fragmentation risk monitoring decisions, executes mechanical anchoring preparation and clamping force setting, and sets and distributes multi-machine anchoring and force distribution rules. The chain drag module is used to configure anchor points and work inputs, configure force and torque balance constraints, calculate real-time force distribution, execute force balance closed loop and residual monitoring, set planned track migration and along-line monitoring, execute triggered replacement, multi-machine redistribution and robust reconfiguration, and execute recycling docking, charging and reuse maintenance.

2. The space debris near-Earth orbit cleanup system based on the coordinated use of BeiDou satellites and unmanned aerial vehicles as described in claim 1, characterized in that, The fragment library building module specifically includes: S100.1 Configure the Ku-band phased array radar of the parent satellite and link it with the imaging payload. Set the radar scanning period and identification threshold parameters, and set the echo acquisition, observation time scale and raw data retention strategy. Perform feature extraction based on echo acquisition and imaging payload. A Ku-band phased array radar is configured on the mother satellite platform. When the system has an optical imaging payload or an infrared imaging payload available, the mother satellite platform triggers the optical imaging payload or the infrared imaging payload to perform linked observation within the radar observation time window. The single sector scan period of the Ku-band phased array radar of the parent satellite is set to 10 minutes, the spatial resolution of the radar is limited to no more than 0.1 meters, and the minimum target identification size threshold of the radar is limited to 0.10 meters. The parent satellite Ku-band phased array radar acquires echo data using pulse compression. The echo data includes echo amplitude, carrier phase, polarization channel data, and Doppler spectrum. The parent satellite platform generates an observation time tag for each observation, with a time accuracy of no more than 0.1 seconds. The parent satellite platform retains the raw echo data in the database for no less than 72 hours. Based on the echo data, pulse compression and Doppler spectrum estimation are performed to obtain the equivalent radar cross-section estimate, polarization ratio, spectral width, and spectral peak information, respectively. Based on the optical imaging data or the infrared imaging data, radiometric calibration and geometric correction are performed to obtain reflectivity spectral band characteristics or brightness temperature curves. The radar characteristics and the imaging characteristics constitute an observation feature vector, which includes the following fields: RCS, polarization ratio, spectral width, spectral peak information, reflectivity spectral band characteristics, imaging boundary size estimation, and observation time stamp. Using the measured value of the observed feature vector as the observation input, the orbital state vector of the debris is filtered and estimated. The orbital state vector consists of the position vector in the geocentric inertial coordinate system, the velocity vector in the geocentric inertial coordinate system, and the rotation rate. The filtered estimation uses an extended Kalman filter or an unscented Kalman filter to perform prediction and observation updates. The filtered estimation results must output the position covariance matrix and velocity covariance matrix as uncertainty measures, and the orbital state estimation and uncertainty are updated at a frequency of no less than once every 10 minutes. S100.

2. Set the rules for generating and merging fragment entries, perform preliminary material judgment and quality estimation based on feature extraction results, and set the calculation method for threat score and the judgment of candidate cleanup objects. For each observation that meets the identification threshold, a fragment entry is generated and written into the fragment library. The fragment entry includes the following fields: fragment unique identifier, observation time tag, geocentric inertial coordinate system position vector, geocentric inertial coordinate system velocity vector, size estimate, shape estimate, rotation rate estimate, preliminary material estimate, position covariance matrix, velocity covariance matrix, observation source identifier, and original echo index. The fragments are identified by a material determination method based on the observed feature vector. The material categories include at least: aluminum alloy, steel, composite materials and ceramics. The material determination outputs a material classification label and a confidence value. When the confidence value is greater than or equal to 0.8, the material determination result serves as the direct basis for subsequent approach and cleanup strategies. When the confidence value is between 0.5 and 0.8, close-range confirmation is triggered to obtain further spectral or imaging data to finally determine the material category. The mass estimation range is calculated based on the size estimate and the typical density range of the material category. The debris threat score is defined as the weighted sum of the debris size normalized score and the debris collision probability normalized score; The set of debris entries outputs a list of candidate cleanup targets to the task planning module. The output fields include: unique debris identifier, latest orbital state estimate, position covariance matrix, velocity covariance matrix, mass estimation range, material label and confidence level, threat score and original echo index. The output adopts a structured machine-readable data format and controls the access permissions of the original echo data according to the task security level.

3. The space debris near-Earth orbit cleanup system based on the coordinated use of BeiDou satellites and unmanned aerial vehicles as described in claim 1, characterized in that, The high-precision positioning module specifically includes: Configure inter-satellite observation links and dual-frequency receiving capabilities, and preprocess the received observations, setting inter-satellite differential and double-difference to eliminate gross errors; The mother satellite platform and the UAV are equipped with their respective BeiDou receiving units to establish inter-satellite link observation capabilities. The BeiDou receiving unit has dual-frequency carrier phase observation capabilities and can output carrier phase raw observation stream and code pseudorange observation stream. The observations of a single satellite receiver pair are described using a carrier phase observation model; To eliminate common errors such as satellite clock bias and receiver clock bias, inter-satellite differential and double-difference processing is adopted: first, single-difference processing is performed on the carrier phase of the two satellites by the same receiver to remove the receiver clock bias, and then the single-difference values ​​of the two receivers on the same satellite are subtracted to form double-difference observations; Based on the aforementioned double-difference observations, the inter-satellite double-difference method is used to eliminate common terms caused by large-scale clock errors and propagation, thereby making the integer ambiguity resolution a feasible integer estimation problem. Perform integer ambiguity resolution and fast solidification, perform filtering fusion and status output, and set the host star clock broadcast and joint time synchronization strategy; For the double-difference integer ambiguity Integer least squares analysis is performed to obtain integer solutions, and the integer analysis can be implemented using the LAMBDA algorithm. Based on the double-difference observation with fixed or floating-point ambiguity, the carrier phase and code pseudorange observations are input into a real-time filter to estimate the relative position and relative velocity. The real-time filter can be an extended Kalman filter or an unscented Kalman filter. The filter outputs at least the following state quantities and updates them at a frequency of 10 Hz: relative position vector, relative velocity vector, and positioning covariance matrix. When the ambiguity cannot be fixed in real time, the filter maintains the continuity of the solution in floating-point ambiguity mode and outputs a suboptimal accuracy solution. Using the reference clock of the mother satellite platform as the system time reference, the mother satellite platform periodically broadcasts time synchronization packets to the UAV swarm and other on-orbit nodes. The time synchronization packets contain reference time identifiers and clock deviation correction information. The receiving end corrects its local clock based on the propagation delay estimation of the time synchronization packets and carrier observations. When the ambiguity cannot be fixed within the preset time window, or the covariance of the floating-point ambiguity solution exceeds the quality threshold, the system should output the current best floating-point solution in a degraded form and note the accuracy level of the solution in the output. The output interface provides the following fields to the proximity control module and the task planning module in a structured format: relative position vector, relative velocity vector, positioning covariance matrix, ambiguity fixation flag, time synchronization deviation value, and observation link status indication.

4. The space debris near-Earth orbit cleanup system based on the coordinated use of BeiDou satellites and unmanned aerial vehicles as described in claim 1, characterized in that, The feature determination module specifically includes: Configure confidence level and quality range confirmation, set up proximity triggering and confirmation process, and set the objective function for multi-objective task allocation; The generated set of debris entries and the output high-precision relative positioning and time reference are used as inputs, including but not limited to: debris entry fields, the remaining propellant / remaining energy / health status of the UAV swarm, the home planet recovery window and ground mission priority; Online classification is performed based on the observed feature vectors of the fragment entries to output material confidence scores. With material labels, when At that time, the material label serves as a direct basis for the approach and cleaning strategy. At that time, a close-up verification procedure is triggered to obtain further spectral or imaging data to ultimately confirm the material label; When close-range confirmation is triggered, the mission planning module generates a close-range observation task and assigns at least one UAV to perform close-range confirmation observations. Close-range confirmation observations include short-range spectral sampling or high-resolution imaging. The close-range confirmation results will be written back to the fragment entries with observation timestamps and updated. And the quality estimation range; Using the fragment candidate set and the drone swarm status as input, a multi-objective optimization objective function for task allocation is formulated; Set constraints and feasible domain limits, set examples and distinguishing features for distributed allocation implementation, and set batch command output for formation generation and release; The multi-objective optimization is solved under the constraint that the estimated propellant consumption of each UAV must not exceed its remaining propellant budget, denoted by: The number of cooperating drones required for a single fragment must not exceed the number of available drones and must not be less than the required lower limit, denoted by the following symbol: Task time window constraints: The task execution window must fall within the time interval specified by the mother planet recovery window and the task priority. The task allocation is preferably implemented as a distributed auction allocation to improve robustness and online rescheduling capability. The distributed auction allocation can be CBBA or a functionally equivalent distributed matching algorithm. The distributed implementation allows each UAV to base its allocation on local estimations. , The task allocation result is generated through consistent convergence by bidding with the remaining energy. The distributed implementation has distinctive benefits in on-orbit high latency / link instability scenarios. Based on the optimal allocation results, a task allocation table is generated for each UAV. The task allocation table includes fields such as: target fragment ID, predetermined approach time window, and estimated time window. Estimate execution time The required anchor point location list, initial formation suggestions, release batch number and priority, the task allocation table must be sent to the mother satellite release module or UAV ground dispatch unit no later than 30 minutes after task identification; The task allocation table is output to the proximity control module and the home satellite release module, and the close-range confirmation results, actual observations, and execution feedback should be written back to the debris entry set for updating. Quality estimation range and threat score are used to support the replanning of subsequent tasks.

5. The space debris near-Earth orbit cleanup system based on the coordinated use of BeiDou satellites and unmanned aerial vehicles as described in claim 1, characterized in that, The proximity control module specifically includes: S400.1 Configure the release window and initial security parameters, form an initial formation through the release action, and execute distributed MPC to gradually approach and coordinate with the formation; Configure the mothership platform to select the release window based on the mission allocation table and the capabilities of the mothership release module, and set the initial safety parameters after release, including the minimum safe separation distance between UAVs after release and the maximum permissible relative speed at the moment of release; The UAVs are released in batches by the mothership platform to form a predetermined initial formation, and local approach commands are issued to each released UAV. The local approach commands include the following fields: target fragment ID, predetermined approach time window, expected terminal relative position, expected terminal attitude, expected terminal relative speed and priority identifier. Execute distributed MPC to achieve precise approach under constraints of limited propulsion and multi-machine collaboration; Real-time collision and safety constraints are achieved by controlling the barrier function, setting end alignment and anchoring trigger criteria and issuing anchoring preparation instructions, executing degradation strategies and fault reconstruction, and writing back the task log; By defining a barrier function based on the distance between two bodies and embedding it as a real-time inequality constraint into MPC; Set the end alignment criteria. When the following conditions are met simultaneously, an anchoring preparation signal is issued and the anchor deployment procedure begins: position error threshold, attitude error threshold, end relative velocity threshold, and positioning uncertainty threshold. Set downgrade conditions and handling procedures: When the positioning uncertainty or ambiguity quality index exceeds the safety threshold, or when a single drone propulsion / attitude control fails, execute the following in order of priority: pause approach and return to the predetermined standby formation, replace the drone performing the task, or terminate the entire mission and withdraw. The system outputs propulsion / attitude control commands, CBF safety constraint status flags, anchoring preparation flags, and current positioning uncertainty indicators for each control cycle to the proximity execution unit, and feeds back key events of the proximity process to the mission planning and debris item set.

6. The space debris near-Earth orbit cleanup system based on BeiDou satellite and UAV collaboration as described in claim 1, characterized in that, The material adaptation module specifically includes: The system is equipped with a laser unit, mechanical anchor, and force sensor. The laser operating mode is selected based on material confidence and rotation rate. The laser energy and impulse requirements are calculated and converted into emission parameters. Before performing the cleaning task, the laser unit, mechanical anchor hook and the force sensor on the anchor hook are configured and verified. The laser unit has adjustable peak value and preferably operates at a wavelength of 1064nm and has an output power range of 100W to 1kW. The mechanical anchor hook has an extension and retraction mechanism and is equipped with a real-time force sensor. The adjustable range of the anchor hook clamping force is preferably 50N to 200N, and the clamping force measurement accuracy is preferably ±5N. By material confidence With fragment rotation rate Determine the laser operating mode: When and When laser deceleration is the primary mode, it should be selected first. or When using this method, prioritize a hybrid or mechanically-dominated mode that combines close-range confirmation with mechanical anchoring. To achieve the required deceleration Using the energy-impulse empirical coefficient Estimate the required laser energy ; Execute laser ablation timing and online fragmentation risk monitoring decisions, execute mechanical anchoring preparation and clamping force settings, and set and distribute multi-machine anchoring and force distribution rules; During laser ablation, the power and pulse parameters are controlled simultaneously according to the selected operating mode to activate the fragmentation / debris monitoring circuit and estimate the fragmentation probability index. Set decision rules: If Then the laser will stop and switch to mechanical control mode. Laser deceleration reaches or approaches the target Afterwards, or when directly selecting mechanical mode, perform anchoring preparation: deploy the anchor hook mechanism and approach the anchoring point; at the moment of contact, calculate the mass range and the desired drag acceleration. Calculate the required clamping force ; like If the clamping force of a single machine is within its upper limit, then single-machine clamping is sufficient; otherwise, multi-machine anchoring is performed. During anchoring, the force sensor is read in real time, and closed-loop control is used to bring the actual clamping force closer to the limit. Record successful clamping when the clamping error is less than ±5N; When the clamping force of a single machine is insufficient At that time, the minimum number of drones required to participate will be determined. ; Laser power timing and emitted energy , estimate the generation of The anchor point sequence, the real-time clamping force of each drone, and the clamping success flag are written back to the fragment entry set and the task log. The task planning module is notified in real time when a strategy switch occurs so that dynamic rescheduling can be carried out.

7. The space debris near-Earth orbit cleanup system based on the coordinated use of BeiDou satellites and unmanned aerial vehicles as described in claim 1, characterized in that, The chain-style drag-and-drop module specifically includes: Configure anchor points and operation inputs, configure force and torque balance constraints, calculate real-time force distribution, and execute force balance closed-loop and residual monitoring; Before performing chain drag, configure and read the following input data: spatial position vector of each anchor point, real-time clamping force measurement of each drone, and maximum sustainable pulling force of each drone. The target debris mass estimation range and target migration requirements, the home planet recovery window and recharging capacity. Establish system-level force balance and torque balance equations as hard constraints for drag control; To satisfy the equality constraints and minimize the risk of single-machine overload, a real-time secondary optimization solution is performed. After the force distribution is solved, each participating UAV performs closed-loop force tracking control: reading the force sensor output and... PI / PID Alternatively, the propulsion command can be adjusted based on the model-based force controller, and the system continuously calculates the force residuals and sends the residuals to the monitoring module; Set up planned track migration and along-line monitoring, execute triggered replacement, multi-machine redistribution and robust reconfiguration, and perform recovery docking, charging and reuse maintenance; Based on the force distribution and desired acceleration Calculate the orbital migration trajectory and the expected cumulative ; Fault detection employs two mechanisms: residual detection and majority voting residual detection. To determine if a single drone has failed, a majority voting mechanism is used where participating drones vote on key observations. If the majority vote exceeds 50%, the system is considered to be in a state of anomaly. Replacement and Reconstruction Rules: If a single machine is determined to be faulty, the required number of participants will be recalculated according to the multi-machine allocation rules. The distributed task reconfiguration algorithm selects a replacement drone from the in-service drones to take over the anchor point or compensate for the force. If it is insufficient to maintain safe towing, the evacuation or termination is carried out according to priority and the drone returns to the home planet. After the target migration and cleanup tasks are completed, the return and reuse process is executed: each UAV returns and docks according to the provided high-precision positioning and time reference, with the optimal docking tolerance being position. And attitude error The mother planet allocates charging resources to each docked drone and optimizes the charging power. It implements modular self-inspection and maintenance processes and records maintenance checklists, with the design goal of ensuring that the drone can be reused at least once. Second-rate; During the drag-and-drop and migration process, the system periodically outputs the following fields to the task planning module and the fragment item set: real-time gripping force of each drone. Distributed tension residual force Current force distribution scheme, cumulative The system estimates the health level of anchor points and records events. When a policy switch or fault reconstruction occurs, it immediately writes back the data and notifies the task planning module for dynamic rescheduling.

Images