Satellite collision avoidance method with full-process closed-loop quick response

Abstract

The invention discloses a full-process closed-loop quick response satellite collision avoidance method, which comprises the following steps of: 1, observing and acquiring an orbit state of a space object, and then carrying out multi-source data fusion so as to construct a space object database; 2, screening out high-risk collision sources which have the possibility of collision with the own satellite; 3, obtaining a determined high-risk collision source according to the track state historical record of the possible high-risk collision source and the own track state; 4, generating an initial evasion maneuvering scheme, and optimizing the initial evasion maneuvering scheme through a multi-objective optimization algorithm; and obtaining a final evading maneuvering scheme for subsequent execution. And 5, according to the engine startup and shutdown time and the thrust vector contained in the final evasion maneuver scheme, the satellite engine is controlled to work with the specified thrust vector within the given startup and shutdown time, and collision evasion maneuver is executed. Compared with a conventional method, the method has the advantage that the response speed is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of near-earth space, and specifically refers to a satellite collision avoidance method with a full-process closed-loop and rapid response. Background Art

[0002] With the increasing frequency of global space activities, the number of spacecraft has surged, and the number of space debris generated due to launches, failures, and collisions has increased. The number and density of space objects are increasing day by day, leading to an increasing number of space collision events. In order to protect the safety of spacecraft, protect China's space assets and national security, it is of great significance to study the collision avoidance method for satellites and other spacecraft.

[0003] Satellite collision avoidance generally includes three main parts: early warning, decision-making, and execution. Early warning refers to monitoring objects such as spacecraft and space debris in orbit through ground-based measurement and control stations, space-based satellites, etc., and forecasting their orbits to determine whether there is a high collision risk with one's own satellite. Decision-making means that when it is determined that there is a high collision risk between one's own satellite and a certain space object, multi-dimensional factors such as the health status of the satellite, maneuverability, mission constraints, and fuel consumption are comprehensively considered to formulate a collision avoidance strategy. Execution means that according to the formulated collision avoidance strategy, the orbit control maneuver command is uploaded to the satellite through the measurement and control station, and the on-board engine is started to perform orbit change maneuvers to avoid possible collisions.

[0004] In the conventional satellite collision avoidance method, the monitoring of space objects is mainly carried out through ground-based measurement and control stations and space-based navigation satellites. This method usually provides long-distance early warning information, and the collision warning result needs to be given 48 hours before the collision. Moreover, the space objects that can be monitored are limited, and the measurement accuracy is limited. For small space debris and non-cooperative spacecraft that approach suddenly and unpredictably, the monitoring effect of the conventional method is limited, and it is difficult to detect threats in advance to provide sufficient time for the decision-making, uploading, and execution of avoidance maneuvers.

[0005] To conduct collision early warning for a spacecraft, it is necessary to compare and calculate its orbit with the orbits of all other cataloged space objects. Since there are a large number of space objects and the orbit calculation amount is huge, an efficient screening algorithm is needed to filter out objects with low collision risks and only calculate the orbits of high-risk objects. Conventional collision source screening uses multi-level screening based on different orbit characteristics, and the process is relatively complex. Moreover, it requires manual operation and judgment, and the screening efficiency is low, so it cannot be deployed in the on-board computer with limited computing resources.

[0006] To formulate a collision avoidance maneuver plan, various factors such as the health status, maneuverability, mission constraints, and fuel consumption of the satellite need to be taken as optimization constraints and objectives for multi-objective optimization calculations, and repeated iterations are required to obtain an orbit control plan that meets the requirements. Conventional iterative optimization calculations are computationally intensive and have low computational efficiency, making it difficult to deploy on-board computers and unable to meet the requirements of rapid response.

[0007] In conventional methods, collision warning and collision avoidance decisions are often executed on the ground. The instructions for executing avoidance maneuvers need to be uplinked to the satellite through the ground measurement and control station, and then the maneuvers are executed through the on-board engine. However, the process of uplinking the satellite by the ground measurement station will cause a time delay, making it difficult to achieve immediate control. Moreover, it is difficult for the ground measurement and control station to conduct full-time measurement and control of the satellite, and instructions need to be uplinked when the satellite passes through the coverage area of its own measurement and control station. If the time from discovering a high-risk collision source to the expected collision is short and the satellite is not within the coverage range of its own measurement and control station at this time, the satellite will not be able to avoid the collision in time. Therefore, the conventional method of executing warnings and decisions on the ground and uplinking instructions greatly limits the ability to respond quickly to collision risks.

[0008] In addition, conventional collision avoidance methods usually consist of multiple independent links, and each link relies on manual operation, making it difficult to achieve an automated closed-loop process for avoiding multiple potential collision targets, which also limits the ability to respond quickly to collision risks. Summary of the Invention

[0009] The purpose of the present invention is to propose a satellite collision avoidance method with a full-process closed-loop rapid response in view of the deficiencies of the prior art. An intelligent algorithm based on a neural network is adopted, which can be deployed on an on-board computer with limited computing resources. Compared with the conventional scheme, it has higher computational efficiency and speed, and because it is directly deployed on the satellite and does not need to communicate with the ground measurement and control station, it can achieve a faster response speed than the conventional method.

[0010] To solve the above technical problems, the technical solution of the present invention is as follows:

[0011] A satellite collision avoidance method with a full-process closed-loop rapid response includes the following steps:

[0012] Step 1: Observe and obtain the orbital state of space objects, then perform multi-source data fusion, and further construct a space object database;

[0013] Step 2: Train an intelligent screening neural network model, and screen the orbital states of all space objects through the pre-trained intelligent screening neural network model to screen out high-risk collision sources that may collide with one's own satellite;

[0014] Step 3: Based on the orbital state history of possible high-risk collision sources and the own orbital state, determine the high-risk collision sources;

[0015] Step 4: Train an intelligent decision neural network model. Use the orbital states of the determined high-risk collision sources and the own satellite's orbital state as inputs, generate an initial avoidance maneuver plan through the pre-trained intelligent decision neural network model, and optimize the initial avoidance maneuver plan through a multi-objective optimization algorithm; obtain the final avoidance maneuver plan for subsequent execution;

[0016] Step 5: According to the engine on / off times and thrust vectors included in the final avoidance maneuver plan, control the on-board engine to work with the specified thrust vector within the given on / off times to perform a collision avoidance maneuver.

[0017] Step 6: After completing the maneuver, perform satellite state feedback to obtain the satellite's orbital state after the maneuver, evaluate the avoidance effect, and obtain the collision risk after avoidance. If the collision risk is still higher than the threshold, return to Step 4 and repeat the above process; if the collision risk is lower than the threshold, return to Step 1 and repeat the early warning and avoidance of potential high-risk collision sources.

[0018] Preferably, the orbital state refers to any physical quantity that can describe the position and velocity of a spacecraft in space.

[0019] Preferably, in Step 1, the methods for obtaining the orbital state of a space object include: ground station observation, GNSS satellite observation, on-board autonomous observation, and orbital prediction of known objects.

[0020] Preferably, the training method of the intelligent screening neural network model is as follows:

[0021] First, use the space object database as the input, apply geometric feature screening algorithms, time feature screening algorithms, and orbital rough prediction screening algorithms to screen the space objects, distinguish high-risk and low-risk collision sources, and label them, thereby establishing a machine learning sample set;

[0022] Construct a basic neural network model, and then use the supervised learning method to train the neural network model with the above machine learning sample set, optimize the network parameters. When the optimal network parameters are obtained, the training is completed to obtain the pre-trained intelligent screening neural network model.

[0023] Preferably, in Step 3, the method for obtaining the determined high-risk collision sources is as follows:

[0024] Step 3.1: Based on the orbital state history of the collision source and the own orbital state, obtain the orbital state and measurement error covariance matrix at the moment when the collision source is closest to the own satellite;

[0025] Step 3.2: Based on the orbital state and measurement error covariance matrix of the collision source at the closest moment, as well as the orbital state and measurement error covariance matrix of one's own satellite at the closest moment, evaluate the collision risk to determine the high-risk collision source.

[0026] Preferably, in Step 3.1, based on the historical record of the orbital state of the collision source, the initial measurement error covariance matrix is obtained; then, a high-precision dynamic model is used to perform high-precision prediction on the orbital state and initial measurement error covariance matrix of the collision source and one's own satellite at the initial moment, so as to obtain the orbital state and measurement error covariance matrix of the collision source and one's own satellite at the closest moment.

[0027] Preferably, in Step 3.2, the method for evaluating the collision risk is as follows:

[0028] First, calculate the collision probability, and the expression is as follows:

[0029] ;

[0030] where, and are the x and y components of the projection of the combined error covariance on the encounter plane, represents the distance between the two targets at the closest moment, and are the x and y components of the projection of the relative position vector at the closest moment on the encounter plane, represents the natural exponential function;

[0031] Construct a comprehensive collision risk assessment function, and the expression is as follows:

[0032] ;

[0033] where, a1 to a7 are weight coefficients, R, S, and W are the components of the relative position of the two targets at the closest moment in the radial, along-track, and normal directions of the orbit, L R 、L S 、L W are the reference approach distances in the radial, along-track, and normal directions of the orbit respectively, T1 and T2 are the time from the current moment to TCA (the closest moment) and the time from the most recent orbit measurement to TCA respectively, Q is the expert scoring value of the orbit determination quality considering the observation accuracy and orbit model accuracy, k and K are custom coefficients;

[0034] Calculate the collision risk assessment value of the high-risk collision source. When the collision risk of a certain space target is higher than the set threshold, it is considered that one's own satellite needs to avoid this target.

[0035] Preferably, the training method of the intelligent decision-making neural network model: in the satellite collision historical events, the orbital state of space objects, the satellite orbital state, constraints, and collision risks are used as input features, and the avoidance maneuver strategy is used as the output feature. A training sample set is constructed using a multi-objective optimization algorithm; a basic neural network model is established, and the neural network model is trained using the above training samples by a supervised learning method to obtain the optimal neural network parameters. After training, a pre-trained intelligent decision-making neural network model is obtained.

[0036] Preferably, in step 4, the method for optimizing the initial avoidance maneuver plan by a multi-objective optimization algorithm is as follows: The safety factor is represented by the collision risk F and needs to meet the condition:

[0037] ;

[0038] where is the collision risk threshold;

[0039] The time factor means that the on-off time of the engine for the avoidance maneuver is restricted by the warning time , the time when the two targets are closest , the start and end times of the ground measurable and controllable time interval , and the maximum on-time of the satellite orbit control engine . Then the on-off time of the satellite orbit control engine needs to meet the condition:

[0040] ;

[0041] The fuel factor refers to the fuel consumption during the engine startup in the avoidance maneuver. In the optimization, the fuel consumption should be reduced as much as possible. The expression of the fuel consumption is as follows:

[0042]

[0043] where represents the satellite mass, represents the engine thrust;

[0044] The mission factor refers to the function of the satellite itself and the working efficiency of the mission it performs. In the optimization, the working efficiency should be improved as much as possible, that is, the impact of the avoidance maneuver on the satellite's own mission is reduced.

[0045] The present invention has the following characteristics and beneficial effects:

[0046] Using a neural network to implement an intelligent screening algorithm and an intelligent decision-making algorithm, replacing or assisting the hierarchical screening and multi-objective optimization algorithms in conventional methods, can screen risk sources and make avoidance maneuver decisions with higher computational efficiency and faster computational speed, and improve the autonomous response ability of the system.

[0047] After performing a collision avoidance maneuver, the avoidance effect is evaluated in real time and fed back to the decision-making process to determine whether to continue the avoidance maneuver and dynamically correct the result. After completing the avoidance of the current target, it returns to the early warning stage to continue monitoring and avoiding the next potential collision target, achieving early warning and avoidance of multiple collision targets. Compared with the relatively scattered and independent collision avoidance links in the conventional method, this solution proposes a full-process closed-loop method to achieve an automated and rapid response to collision risks.

[0048] Each link of the present invention can be completed in whole or in part by the observation equipment and on-board computer carried by the satellite itself, reducing the communication limitations and time delays caused by data exchange with the ground. The conventional method usually uses ground equipment as the main observation and calculation carrier, and it is necessary to upload satellite control instructions to the satellite for maneuvering during the visible time interval of the satellite, making it difficult to handle emergently occurring collision risks and situations where ground measurement and control are not supported.

[0049] Use on-board autonomous observation equipment, ground-based measurement and control stations, and space-based navigation satellites to observe and monitor space objects, and adopt a multi-source data fusion method to improve the observation accuracy.

[0050] It solves the shortcomings of the conventional collision source screening method that the multi-level screening process based on different orbital characteristics is relatively complex, requires manual operation and judgment, has low screening efficiency, and cannot be deployed in the on-board computer with limited computing resources. By adopting an intelligent screening algorithm based on a neural network model and deploying it on the on-board computer after pre-training, it helps to achieve rapid collision source screening and solve this problem.

[0051] It solves the shortcomings of the conventional iterative optimization method that has a large amount of calculation, low calculation efficiency, is difficult to be deployed in the on-board computer, and cannot meet the requirements of rapid response. By adopting an intelligent decision-making algorithm based on a neural network model and deploying it on the on-board computer after pre-training, it helps to achieve rapid generation of avoidance maneuver plans and reduce the calculation amount of subsequent optimization and correction algorithms to solve this problem.

[0052] It solves the shortcomings of the conventional method that collision early warning and collision avoidance decisions are often executed on the ground, and it is necessary to upload the instruction to execute the avoidance maneuver to the satellite through the ground measurement and control station, making it impossible to handle emergently occurring collision risks and situations where ground measurement and control are not supported. By implementing all or part of the links through on-board equipment, the full process of satellite autonomous collision early warning, decision-making, and execution is realized, solving this problem.

[0053] In the conventional method, the observation accuracy of ground measurement and control stations is limited. This solution proposes a multi-source data fusion method that integrates ground-based measurement and control stations, space-based navigation satellites, and on-board autonomous observation equipment to construct a space object database, improving the observation accuracy of space objects by comprehensively considering observation data from multiple aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0055] Figure 1 It is a schematic flowchart of an embodiment of a satellite collision avoidance method with a full-process closed-loop and rapid response of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0057] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first", "second", etc. may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more.

[0058] In the description of the present invention, it should be noted that, unless otherwise clearly defined and limited, the terms "installed", "connected", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood through specific situations.

[0059] The present invention provides a satellite collision avoidance method with a full-process closed-loop and fast response, including a complete system for realizing satellite collision avoidance, and the specific technical details are not limited. As Figure 1 shown, it is mainly divided into three components, namely: the early warning link, the decision-making link, and the execution link.

[0060] In the early warning link, ground-based measurement and control stations, space-based navigation satellites, and on-board autonomous observation equipment are used to monitor and locate space objects. Combining with the known prediction results of space objects, a space object database is established through multi-source data fusion. The intelligent screening neural network model is used to screen out high-risk collision sources. When high-risk collision sources are detected in this link, an alarm is issued, and then the decision-making link of collision avoidance is entered. The decision-making link collects the orbital information of the collision risk source, the risk assessment information, and the information of its own satellite, and uses the intelligent decision-making neural network model to generate an avoidance maneuver plan, and corrects the maneuver plan by combining precise multi-objective optimization algorithms and mission constraints. The execution link executes the avoidance maneuver plan obtained in the decision-making link through the on-board attitude and orbit control system, and feeds back the orbital state and risk assessment information of the satellite after the execution maneuver to the decision-making link to evaluate the avoidance effect and whether to perform the next avoidance maneuver, completing the closed-loop of this method. Some steps of this solution adopt intelligent algorithms based on neural networks, which can be deployed in on-board computers with limited computing resources. Compared with conventional solutions, it has higher computing efficiency and computing speed, and because it is directly deployed on the satellite and does not need to communicate with ground measurement and control stations, it can achieve a faster response speed compared with conventional methods. The specific implementation steps are as follows.

[0061] Early warning link:

[0062] Step 1: Measure the orbital states of all observable space objects through ground-based measurement and control stations and space-based navigation satellites. The on-board ranging and angle measurement equipment is used for on-board autonomous observation to obtain the relative motion state of the space object relative to the satellite. Combining with the satellite's own orbital state, the orbital state of the space object is obtained. The orbital states of the space objects obtained from all the above observation methods are subjected to multi-source data fusion to construct a space object database.

[0063] It should be noted that in this embodiment, the orbital state refers to any physical quantity that can describe the position and velocity of a spacecraft in space. For example, Kepler orbital elements (hereinafter referred to as orbital elements), inertial space position and velocity coordinates, etc.

[0064] It should be noted that in this embodiment, a method combining space-based navigation satellites, ground-based measurement and control stations, and on-board autonomous observations is adopted to construct a space object database, and each part is executed on different equipment carriers. For the case where the computing power of the on-board computer is limited, all observation information needs to be aggregated to the ground measurement and control station, and the subsequent steps of the early warning link are executed on the ground computer. For the case where the on-board computer has sufficient computing power, all observation information can be aggregated to the on-board computer to execute the subsequent steps of the early warning link. In the implementation of this solution, a processor with sufficient computing power is preferably used as the on-board computer. In addition, the above several observation methods are optional, and in specific implementations or in emergencies, only some of these methods can be used for observation.

[0065] Step 2: After obtaining the space object database, use the pre-trained intelligent screening neural network model to screen the orbital states of all space objects therein, and screen out the high-risk collision sources that may collide with the own satellite.

[0066] It should be noted that there are various methods for training this intelligent screening neural network model, which belong to conventional technical means. There are various types of basic neural network models that can achieve the functions required by this solution, including but not limited to fully connected neural networks, recurrent neural networks, and convolutional neural networks.

[0067] In this embodiment, in a possible implementation manner, according to the orbital states of the space object and the own satellite, feature screening is performed, including geometric feature screening, time feature screening, and rough orbit prediction screening. The following briefly describes different screening methods.

[0068] Specifically, the geometric feature screening method is as follows: Let P o and A o be the perigee and apogee of a certain space target respectively, and P s and A s be the perigee and apogee of the own satellite respectively. When the condition is satisfied, it is considered that no collision will occur. According to this condition, space objects with a large difference in orbital altitude from the own satellite are screened out.

[0069] Furthermore, according to the orbital elements of the space object and the own satellite, through the Kepler orbit equation, the minimum distance between the two orbits is calculated. If it is greater than the threshold distance, it is considered that no collision will occur. The threshold distance is set according to specific application scenarios.

[0070] The time feature screening method is as follows: According to the orbital elements of the space object and the own satellite, based on the true anomaly, the time of closest approach (TCA) of the two orbits is obtained using the Kepler orbit equations, and then the collision risk time domain is obtained.

[0071] The method for rough orbit prediction screening is as follows: within the collision risk time domain obtained by the time feature screening method, the SGP4 / SDP4 prediction model is used for rough orbit prediction, and the prediction step length and threshold distance are set according to specific application requirements. If the distance between a certain space object and its own satellite is less than the threshold distance within a certain step length, it is considered that the target has a high collision risk and needs to enter the subsequent high-precision screening.

[0072] It can be understood that if the conditions for ground calculation and uploading instructions to the satellite are available, the above feature screening method or a method combined with intelligent screening can be directly adopted to screen high-risk collision sources.

[0073] Through the above method, as many historical data of space objects as possible that can be obtained are screened to distinguish high-risk and low-risk collision sources. The historical data of space objects are used as training samples, the orbital states therein are used as input features, and whether it is a high / low-risk collision source is used as a label. A basic neural network model is established, and the above training samples are used to train the neural network model using the supervised learning method to obtain the optimal neural network parameters, and a pre-trained intelligent screening neural network model is obtained. After being trained, the intelligent screening neural network model can directly output whether a space object state is a high-risk collision source when the space object state is input. The intelligent screening neural network model is deployed in the on-board computer to achieve fast intelligent screening and screen out possible high-risk collision sources.

[0074] Step 3: Determine the high-risk collision sources based on the historical orbital state records of the possible high-risk collision sources and the orbital state of one's own side.

[0075] Specifically, based on the historical orbital state records of the possible high-risk collision sources, the measurement error covariance matrix at the initial moment is obtained. The high-precision dynamic model is used to perform high-precision prediction on its orbital state and measurement error covariance matrix at the initial moment to obtain the orbital state and measurement error covariance matrix at the moment when the collision source is closest to its own satellite.

[0076] It should be noted that the high-precision dynamic model is a conventional model. The orbital state and measurement error covariance matrix at the initial moment are obtained from the space object database and used as the input for high-precision prediction. The orbital state and measurement error covariance matrix at the closest moment are obtained through high-precision prediction.

[0077] Based on the orbital state and measurement error covariance matrix of the high-risk collision source at the closest moment, and the orbital state and measurement error covariance matrix of its own satellite at the closest moment, the collision risk is evaluated. For the convenience of understanding, a quantitative evaluation method of collision risk is briefly introduced below.

[0078] First, calculate the collision probability. The relative position vector of the space target at the moment of closest approach to the satellite is perpendicular to the relative velocity vector. The two targets are located on a plane perpendicular to the relative velocity, and this plane is defined as the encounter plane. According to the orbital states of the two targets, the relative position vector, relative velocity vector and joint error covariance of the space target relative to the satellite are calculated and projected onto the encounter plane. The calculation formula for the collision probability is:

[0079]

[0080] in, and are the x and y components of the projection of the joint error covariance on the meeting plane, Represents the distance between two targets when they are closest. and are the x and y components of the projection of the relative position vector on the encounter plane at the moment of closest approach, Represents the natural exponential function.

[0081] Construct a comprehensive collision risk assessment function, the expression is as follows:

[0082]

[0083] Among them, a1 to a7 are weight coefficients, R, S, and W are the components of the relative position of the two targets at the closest moment in the orbital radial, track direction, and normal direction, and L R , L S , L W are the reference approach distances in the orbital radial, track direction, and normal direction, respectively; T1 and T2 are the time from the current moment to TCA and the time from the most recent orbit measurement to TCA, respectively; Q is the expert score of orbit determination quality considering observation accuracy and orbit model accuracy; k and K are custom coefficients. Specifically, the above weight coefficients, custom coefficients, and reference approach distances need to be set according to actual application requirements.

[0084] According to the collision risk comprehensive assessment function, the collision risk assessment value of the high-risk collision source is calculated. When the collision risk of a space target is higher than the set threshold, it is considered that the satellite needs to avoid the target and enter the collision avoidance decision-making stage.

[0085] Decision-making stage:

[0086] Step 4: The decision of the evasive maneuver plan is to generate the initial value through the intelligent decision-making neural network model. The orbital state of the collision target and the orbital state of the own satellite are input into the intelligent decision-making neural network model to obtain the initial value of the evasive maneuver plan. According to the orbital state, constraints, and collision risk of the collision target and the own satellite, the initial value of the evasive maneuver plan is optimized using a multi-objective optimization algorithm, the plan is constrained based on the target numerical range, and the optimized evasive maneuver plan is corrected to obtain the final evasive maneuver plan for subsequent execution. The role of the intelligent decision-making algorithm is to quickly generate a suitable evasive maneuver plan, reduce the amount of calculation of the subsequent multi-objective optimization algorithm, and improve the ability to respond quickly.

[0087] The basic idea of a satellite performing a collision avoidance maneuver is to use the onboard engine to generate thrust, change the satellite's orbit, and separate it from the orbit of the collision target to avoid a possible collision. There are two main types of satellite engine working modes: pulse maneuvers and continuous low-thrust maneuvers. In the former, the satellite has a sufficiently large thrust to significantly change the satellite's orbital speed instantly, and the engine's work can be regarded as a pulse. In the latter, the satellite has a smaller thrust and needs to be continuously pushed forward for a period of time to achieve effective avoidance. This scheme takes the continuous low-thrust collision avoidance method as an example. The avoidance maneuver scheme is expressed as ,in The engine start and shut down time. It is the components of the thrust vector in the radial, track and normal directions of the orbit when the engine is working.

[0088] Specifically, the training method of the intelligent decision-making neural network model is as follows:

[0089] First, a multi-objective algorithm is used to establish a machine learning sample set. Based on the historical satellite collision events, including the orbital state of the space object, the orbital state of the satellite, the constraints, and the collision risk in the historical events, a multi-objective optimization algorithm is used to solve the optimal avoidance maneuver strategy. The orbital state of the space object, the orbital state of the satellite, the constraints, and the collision risk are used as input features, and the avoidance maneuver strategy is used as the output feature to construct a training sample set.

[0090] Then, the neural network model is trained using the machine learning sample set. There are many types of basic neural network models that can achieve the functions required by this solution, including but not limited to fully connected neural networks, recursive neural networks, and convolutional neural networks. Use open source neural network toolboxes such as Pytorch to initialize the neural network model containing trainable network parameters, and use the supervised learning method to train the neural network model using the above machine learning sample set to optimize the network parameters. When the optimal network parameters are obtained, the training is completed.

[0091] After being trained, the neural network model can imitate the aforementioned multi-objective optimization algorithm. By inputting the orbital state of space objects, the satellite orbit state, constraints, and collision risks, it can directly output an avoidance maneuver plan that meets the requirements. Deploy this neural network model in the on-board computer to achieve rapid intelligent decision-making.

[0092] In this embodiment, after receiving a collision warning and determining the orbital state, constraints, and collision risks of the collision target and its own satellite, the intelligent decision-making algorithm carried by the satellite on-board computer gives the initial value of the avoidance maneuver plan, conducts high-precision orbit prediction on the satellite orbit state under the current plan, and calculates the constraints and optimization objectives. When the optimization objectives and constraints cannot meet the avoidance requirements, use the multi-objective optimization algorithm to optimize the avoidance maneuver plan.

[0093] Specifically, the multi-objective optimization algorithm is as follows:

[0094] The safety factor is represented by the collision risk F and needs to meet the condition:

[0095]

[0096] Where is the collision risk threshold value, which needs to be set according to the specific application scenario. The time factor means that the engine on / off time of the avoidance maneuver is restricted by the warning time , the time when the two targets are closest, the start and end times of the ground measurable and controllable time interval, and the maximum on-time of the satellite orbit control engine. The on / off time needs to meet the condition:

[0097]

[0098] The above optimization objectives include fuel factors and mission factors. The fuel factor refers to the fuel consumption during the engine startup in the avoidance maneuver, and fuel consumption should be reduced as much as possible during optimization. The expression of fuel consumption is as follows:

[0099]

[0100] Where represents the satellite mass, represents the engine thrust. In the case cited in this scheme, since the continuous small thrust mode is adopted, the thrust size is expressed as 。The mission factor refers to the functions of the satellite itself and the working efficiency of the tasks it performs. In the optimization, the working efficiency should be improved as much as possible, that is, the impact of the avoidance maneuver on the satellite's own mission should be reduced. The mission factor is set according to the specific satellite type and specific mission scenario deployed in this solution. For example, for Earth observation satellites, the mission factor can be represented by three indicators: coverage time resolution, total coverage duration, and coverage rate of mission target points.

[0101] The above multi-objective optimization algorithm means adjusting the avoidance maneuver plan X through the optimization algorithm so that the safety constraint and time factor in the constraint meet the aforementioned conditions, and making the fuel factor in the optimization objective as low as possible and the mission factor as high as possible. There are various specific implementation methods for the optimization algorithm, which are not limited in this solution.

[0102] Execution phase:

[0103] Step 5: After obtaining the avoidance maneuver strategy that meets all constraints and the optimization objective reaches the optimal, enter the execution phase. The following steps are divided into two execution paths according to whether the previous phase is executed on the on-board computer.

[0104] For the case where the computing power of the on-board computer is limited, the early warning and decision-making phase is mainly executed on the ground computer. Therefore, the avoidance maneuver plan needs to be uploaded to the satellite through the ground TT&C station and then executed. First, predict the orbital state of the satellite through a high-precision orbit prediction model to obtain the orbital state of the satellite from the current moment to a period of time before the closest approach time TCA to the collision target, compare it with the coverage ranges of all available ground TT&C stations, judge whether it is within the ground TT&C range, obtain the visible time interval, and the corresponding visible TT&C stations. In the visible time interval closest to the current time, upload the avoidance maneuver plan to the satellite from the corresponding visible TT&C station.

[0105] For the case where the on-board computer has sufficient computing power, the early warning and decision-making phase is executed on the on-board computer. The finally obtained avoidance maneuver plan has been stored in the on-board computer and does not need to be uploaded from the ground, and can directly enter the next step of the execution phase.

[0106] Given the optimized avoidance maneuver plan in the decision-making phase, perform the actuator response step, that is, according to the avoidance maneuver plan including the engine on / off time and thrust vector in it, control the on-board engine to work with the specified thrust vector within the given on / off time, and execute the collision avoidance maneuver.

[0107] Step 6: After the actuator response ends, perform satellite status feedback and avoidance effect evaluation. The satellite status feedback includes determining the current orbital status of the satellite through ground-based TT&C stations, space-based navigation satellite positioning, and on-board autonomous observation. The avoidance effect evaluation includes determining the current orbital status of the space object using the same method as the satellite status feedback, and calculating the collision risk F through the aforementioned method based on the orbital status and measurement error covariance matrix of the high-risk collision source at the current moment, and the orbital status and measurement error covariance matrix of one's own satellite at the current moment, and determining whether it is higher than the collision risk threshold . If F is higher than , then return to the decision-making link, repeat the steps of the above decision-making and execution links, and perform the avoidance maneuver again. If F is lower than , it is considered that the current high-risk collision source has been successfully avoided, the collision avoidance process ends, and return to the warning link to continue monitoring new collision risks. By feeding back the results after the avoidance maneuver to the previous link, the avoidance effect is dynamically corrected to achieve a full-process closed loop.

[0108] The above has described the embodiments of the present invention in detail with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, without departing from the principles and spirit of the present invention, various changes, modifications, substitutions, and variations to these embodiments including components still fall within the protection scope of the present invention.

Claims

1. A satellite collision avoidance method with full-process closed-loop and rapid response, characterized in that, It includes the following steps: Step 1: Observe and obtain the orbital state of a space object, then perform multi-source data fusion, and further construct a space object database; Step 2: Train an intelligent screening neural network model, and use the pre-trained intelligent screening neural network model to screen the orbital states of all space objects to screen out high-risk collision sources that may collide with one's own satellite; Step 3: Obtain the determined high-risk collision sources based on the historical orbital state records of the possible high-risk collision sources and one's own orbital state; Step 4: Train an intelligent decision-making neural network model, use the orbital states of the determined high-risk collision sources and the orbital states of one's own satellite as inputs, generate an initial avoidance maneuver plan through the pre-trained intelligent decision-making neural network model, and optimize the initial avoidance maneuver plan through a multi-objective optimization algorithm; obtain the final avoidance maneuver plan for subsequent execution; Step 5: According to the engine on / off times and thrust vectors included in the final avoidance maneuver plan, control the on-board engine to work with the specified thrust vector within the given on / off times to perform a collision avoidance maneuver; Step 6: After completing the maneuver, perform satellite state feedback to obtain the orbital state of the satellite after the maneuver, evaluate the avoidance effect to obtain the collision risk after avoidance; if the collision risk is still higher than the threshold, return to Step 4 and repeat the above process; If the collision risk is lower than the threshold, return to Step 1 and repeat the warning and avoidance of potential high-risk collision sources.

2. The satellite collision avoidance method with full-process closed-loop rapid response according to claim 1, characterized in that, The orbital state refers to any physical quantity that can describe the position and velocity of a spacecraft in space.

3. A satellite collision avoidance method with full-process closed-loop rapid response according to claim 1, characterized in that In Step 1, the methods for obtaining the orbital state of a space object include: ground station observation, GNSS satellite observation, on-board autonomous observation, and orbit prediction of known objects.

4. A method for satellite collision avoidance with full-process closed-loop fast response according to claim 1, characterized in that The training method of the intelligent screening neural network model is as follows: First, use the space object database as input, apply geometric feature screening algorithms, time feature screening algorithms, and orbit rough prediction screening algorithms to screen the space objects, distinguish high-risk and low-risk collision sources, and label them, and then establish a machine learning sample; Construct a basic neural network model, and then use the supervised learning method to train the neural network model with the above machine learning sample set to optimize the network parameters. When the optimal network parameters are obtained, the training is completed to obtain the pre-trained intelligent screening neural network model.

5. A satellite collision avoidance method with full-process closed-loop fast response according to claim 1, characterized in that, In Step 3, the method for obtaining the determined high-risk collision sources is as follows: Step 3.1: Based on the historical orbital state records of the collision source and one's own orbital state, obtain the orbital state and measurement error covariance matrix at the moment when the collision source is closest to one's own satellite; Step 3.2: Evaluate the collision risk based on the orbital state and measurement error covariance matrix of the collision source at the closest moment, the orbital state and measurement error covariance matrix of one's own satellite at the closest moment, and determine the high-risk collision source.

6. The satellite collision avoidance method with full-process closed-loop rapid response according to claim 5, characterized in that, In step 3.1, based on the orbital state history record of the collision source, the initial measurement error covariance matrix is obtained; then, a high-precision dynamics model is used to perform high-precision prediction on the orbital state and the initial measurement error covariance matrix of the collision source and the own satellite at the initial moment, so as to obtain the orbital state and the measurement error covariance matrix at the moment when the collision source and the own satellite are closest to each other.

7. A satellite collision avoidance method with full-process closed-loop fast response according to claim 5, characterized in that, In step 3.2, the method for evaluating the collision risk is as follows: First, calculate the collision probability, and the expression is as follows: ； Among them, and are the x and y components of the projection of the combined error covariance on the encounter plane, represents the distance when the two targets are closest, and are the x and y components of the projection of the relative position vector at the closest moment on the encounter plane, represents the natural exponential function; Construct a comprehensive evaluation function for collision risk, and the expression is as follows: ； where a1 to a7 are weight coefficients, R, S, and W are the components of the relative position at the moment when the two targets are closest in the radial, along-track, and normal directions of the orbit, and L R , L S , L W are the reference approach distances in the radial, along-track, and normal directions of the orbit respectively, T1 and T2 are the time from the current moment to TCA (i.e., the closest moment) and the time of the last orbit measurement distance to TCA respectively, Q is the scoring value of the orbit determination quality expert considering the observation accuracy and the orbit model accuracy, and k and K are custom coefficients; Calculate the collision risk evaluation value of the high-risk collision source. When the collision risk of a certain space target is higher than the set threshold, it is considered that the own satellite needs to avoid this target.

8. A method for satellite collision avoidance with full-process closed-loop fast response according to claim 1, characterized in that The training method of the intelligent decision neural network model: In the satellite collision historical events, the orbital state of the space object, the orbital state of the satellite, the constraints, and the collision risk are used as input features, and the avoidance maneuver strategy is used as the output feature. A training sample set is constructed by using a multi-objective optimization algorithm; Then, a basic neural network model is established, and the neural network model is trained by using the above training samples by means of supervised learning to obtain the optimal neural network parameters. After the training is completed, a pre-trained intelligent decision neural network model is obtained.

9. A satellite collision avoidance method with full-process closed-loop fast response according to claim 1, characterized in that In step 4, the method for optimizing the initial avoidance maneuver plan by using a multi-objective optimization algorithm is as follows: The safety factor is represented by the collision risk F, and the condition that needs to be satisfied is: ； Among them, is the threshold value of collision risk; The time factor means that the engine start-up and shut-down times for evasive maneuvers are subject to the warning time , the time at the moment when the two targets are closest , the start and end times of the ground measurable and controllable time interval , and the maximum start-up time of the satellite orbit control engine . Subject to these constraints, the start-up and shut-down times of the satellite orbit control engine need to meet the conditions: ； The fuel factor refers to the fuel consumption of the engine startup during the avoidance maneuver. In the optimization, the fuel consumption should be reduced as much as possible. The expression of the fuel consumption is as follows: ； wherein represents the satellite mass, represents the magnitude of the engine thrust; The mission factor refers to the function of the satellite itself and the working efficiency of the mission it performs. In the optimization, the working efficiency should be improved as much as possible, that is, the impact of the avoidance maneuver on the satellite's own mission is reduced.

10. A method for satellite collision avoidance with full-process closed-loop rapid response according to any one of claims 1-9, characterized in that, The basic network adopted by the intelligent screening neural network model and the intelligent decision neural network model is any one of the fully connected neural network, the recurrent neural network, and the convolutional neural network.

Images