Method and device for identifying intention of cluster spacecraft

Abstract

The invention discloses an intention recognition method and device for cluster spacecrafts, and relates to the technical field of monitoring and management of spacecrafts, and the method comprises the steps: obtaining the initial values of the position and speed of each spacecraft, and employing a label multi-Bernoulli algorithm based on interactive multiple models, estimating a position sequence, a speed sequence and maneuvering time of each spacecraft in a future preset time period, and further applying the trained first neural network to obtain a single-target motion intention of each spacecraft in the future preset time period; and inputting the position sequence, the speed sequence and the corresponding single-target motion intention of each spacecraft in the future preset time period into a second neural network, and outputting the cluster motion intention of each spacecraft in the future preset time period. According to the invention, the intention of the cluster spacecraft in the maneuvering scene can be accurately identified.

Description

Technical Field

[0001] The present application relates to the technical field of cluster spacecraft monitoring and management, and particularly to a method and device for identifying the intentions of cluster spacecraft. Background Art

[0002] In recent years, the rapid development of space exploration and utilization has led to a significant increase in the number of artificial objects in orbit. With the continuous deployment of low-Earth orbit satellite constellations (such as Starlink), the number of satellites is expected to grow exponentially. Therefore, challenges such as space traffic congestion and collision risks have become increasingly apparent. At the same time, with the development of spacecraft orbit and attitude control capabilities, it has become possible for spacecraft to work together in a cluster to carry out high-quality on-orbit services and disaster monitoring operations. However, reconfiguring cluster spacecraft through orbital maneuvers may pose potential collision risks to other operating spacecraft. In addition, the maneuvering behaviors of cluster spacecraft with unknown intentions also complicate target identification and catalog maintenance. Therefore, it is necessary to monitor and manage cluster spacecraft for intention recognition. Summary of the Invention

[0003] The purpose of the present application is to provide a method and device for identifying the intentions of cluster spacecraft, which can accurately identify the intentions of cluster spacecraft in a maneuvering scenario.

[0004] To achieve the above purpose, the present application provides the following solutions:

[0005] In a first aspect, the present application provides a method for identifying the intentions of cluster spacecraft, including:

[0006] Obtaining the observed values of the position and velocity of each spacecraft;

[0007] According to the previous cluster motion intention recognition results, the current position and velocity observed values of each spacecraft, applying the labeled multi-Bernoulli algorithm based on the interacting multiple models, to obtain the position sequence, velocity sequence and maneuvering time of each spacecraft within a future preset time period;

[0008] According to the position sequence, velocity sequence and maneuvering time of each spacecraft within the future preset time period, applying the trained first neural network, to obtain the single-target motion intention of each spacecraft within the future preset time period;

[0009] Inputting the position sequence, velocity sequence and corresponding single-target motion intention of each spacecraft within the future preset time period into the trained second neural network, and outputting the cluster motion intention of each spacecraft within the future preset time period; spacecraft with the same cluster motion intention belong to the same cluster.

[0010] In a second aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the above-mentioned method for identifying the intention of cluster spacecraft.

[0011] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned method for identifying the intention of cluster spacecraft is implemented.

[0012] In a fourth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the above-mentioned method for identifying the intention of cluster spacecraft is implemented.

[0013] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0014] The present application provides a method and device for identifying the intention of cluster spacecraft. By obtaining the observed values of the position and velocity of each spacecraft and applying the labeled multi-Bernoulli algorithm based on the interacting multiple models, the position sequence, velocity sequence, and maneuver time of each spacecraft within a future preset time period are obtained. This step realizes single-object maneuver detection (i.e., the maneuver behavior detection of a single spacecraft). According to the position sequence, velocity sequence, and maneuver time of each spacecraft within the future preset time period, a trained first neural network is applied to obtain the single-object motion intention of each spacecraft within the future preset time period. This step realizes single-object intention recognition (i.e., the maneuver behavior intention recognition of a single spacecraft). The position sequence, velocity sequence, and corresponding single-object motion intention of each spacecraft within the future preset time period are input into a second neural network to output the cluster motion intention of each spacecraft within the future preset time period. This step realizes cluster clustering analysis (i.e., the division of the cluster to which each spacecraft belongs). The present application finally completes the accurate identification of the intention of cluster spacecraft through the processes of single-object maneuver detection, single-object intention recognition, and cluster clustering analysis, thereby accurately predicting the future trajectory of cluster spacecraft, reducing spacecraft collisions, and strengthening space traffic management. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is an application environment diagram of a method for identifying the intention of cluster spacecraft in an embodiment of the present application;

[0017] Figure 2 The flowchart of a method for intention recognition of a cluster spacecraft provided by an embodiment of the present application;

[0018] Figure 3 The schematic diagram of the maneuver intention cluster network structure provided by an embodiment of the present application;

[0019] Figure 4 The schematic diagram of the own orbit coordinate system provided by an embodiment of the present application;

[0020] Figure 5 The schematic diagram of the single-target motion intention and the corresponding initial values provided by an embodiment of the present application;

[0021] Figure 6 The schematic diagram of the cluster motion intention provided by an embodiment of the present application;

[0022] Figure 7 The schematic diagram of the structure of a BiGRU-Multi-HeadAttention neural network provided by an embodiment of the present application;

[0023] Figure 8 The schematic diagram of the structure of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0024] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0025] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0026] The method for intention recognition of a cluster spacecraft provided by an embodiment of the present application can be applied to, for example Figure 1In the application environment shown. Among them, the terminal communicates with the server through the network. The data storage system can store the data that the server needs to process. The data storage system can be set separately, integrated on the server, or placed on the cloud or other servers. The terminal can send the observed values of the position and speed of each spacecraft to the server. After the server receives the observed values of the position and speed of each spacecraft, the server applies the labeled multi-Bernoulli algorithm based on the interacting multiple models according to the previous cluster motion intention recognition results, the current position and speed observed values of each spacecraft, and obtains the position sequence, speed sequence, and maneuver time of each spacecraft within a preset future time period; according to the position sequence, speed sequence, and maneuver time of each spacecraft within the preset future time period, the trained first neural network is applied to obtain the single-object motion intention of each spacecraft within the preset future time period; the position sequence, speed sequence, and the corresponding single-object motion intention of each spacecraft within the preset future time period are input into the trained second neural network, and the cluster motion intention of each spacecraft within the preset future time period is output. The server can feedback the obtained cluster motion intention of each spacecraft within the preset future time period to the terminal. In addition, in some embodiments, the intention recognition method of the cluster spacecraft can also be implemented separately by the server or the terminal. For example, the terminal can directly perform intention recognition on the observed values of the position and speed of each spacecraft, or the server can obtain the observed values of the position and speed of each spacecraft from the data storage system and perform intention recognition.

[0027] Among them, the terminal can be, but is not limited to, various desktop computers, laptop computers, smartphones, tablet computers, Internet of Things devices, and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0028] In an exemplary embodiment, as Figure 2 and Figure 3 shown, a method for recognizing the intention of cluster spacecraft is provided. This method is executed by a computer device, and specifically can be executed separately by a computer device such as a terminal or a server, or jointly executed by a terminal and a server. In the embodiments of the present application, taking this method applied to the Figure 1 server in as an example for illustration, it includes the following steps 101 to step 104.

[0029] Step 101, obtain the observed values of the position and speed of each spacecraft.

[0030] Step 102: Based on the previous cluster motion intention recognition results of each spacecraft, as well as the observed values of the current position and velocity, apply the Interacting Multiple Model with Labeled Multi-Bernoulli (IMM-LMB) algorithm to obtain the position sequence, velocity sequence, and maneuvering time of each spacecraft within a preset future time period.

[0031] Step 103: Based on the position sequence, velocity sequence, and maneuvering time of each spacecraft within a preset future time period, apply the trained first neural network to obtain the single-target motion intention of each spacecraft within a preset future time period. The single-target motion intention includes station-keeping, hovering, oscillation, coplanar orbiting, non-coplanar orbiting, flyby, hopping, and spiraling.

[0032] Step 104: Input the position sequence, velocity sequence, and corresponding single-target motion intention of each spacecraft within a preset future time period into the trained second neural network, and output the cluster motion intention of each spacecraft within a preset future time period. The cluster motion intention includes surrounding, escaping, on-orbit maintenance, assembling, and dispersion camouflage; spacecraft with the same cluster motion intention belong to the same cluster.

[0033] Implementing the above steps 101 to 104, the present application designs a Maneuvering Intention and Clustering Network (MIC-Net), that is, the intention recognition method for cluster spacecraft of the present application. This method decouples the intention recognition problem of cluster spacecraft hierarchically, reduces the computational complexity without simplifying the problem, and completes the single-target maneuver detection, single-target intention recognition, and cluster clustering analysis processes through the maneuver layer, intention layer, and cluster layer respectively, and finally completes the intention recognition of cluster spacecraft.

[0034] In another exemplary embodiment of the present application, in step 102, the existing IMM-LMB algorithm is used for single-target maneuver detection. To track a maneuvering target (spacecraft), the single-target state can be augmented as where, in the existing IMM-LMB algorithm, x represents the state vector of the single target, and the state includes position and velocity information. represents the motion mode of the target, where, are all possible motion modes. In the Random Finite Set (RFS), represents the label of the target, which is used to distinguish different spacecraft. Therefore, the single-target state can be augmented as For a multi-target filter, the multi-target state of n targets It can be defined by the following formula.

[0035]

[0036] Assume that the prior density of the multi-target is the LMB RFS in the augmented space, then the density of the target is expressed by the following formula.

[0037]

[0038] Among them, π represents the density of the target; r l represents the probability that target l exists; p (l) (m) is the probability that target l is in mode m; p (l) (·|m) is the spatial distribution of the target given mode m. The spatial distribution of each target is characterized by its joint probability distribution:

[0039] p (l) (x,m) = p (l) (x|m)p (l) (m)

[0040] Among them, p (l) (x,m) represents the spatial distribution of target l in state x and mode m; p (l) (x|m) represents the spatial distribution of target l in state x given that it is in mode m.

[0041] For the prediction step, assume that the survival probability of the trajectory is independent of the current motion mode, then the predicted spatial distribution of target l at the next step is expressed by the following formula.

[0042]

[0043] Among them, represents the survival probability that the predicted target l is in state x′ and mode m′ at the next step; represents the survival probability that the predicted target l is in state x′ at the next step; x′ represents the state of the target at the next step; m′ represents the motion mode of the target at the next step.

[0044] The prediction of the joint distribution can be decomposed into the probability that the target is in mode m after prediction and the corresponding spatial distribution The specific calculation is shown in the following formula.

[0045]

[0046] Among them, f(m|m′) represents the transition probability of the target from mode m to mode m′; p (l) (m′) represents the probability that target l is in mode m′; f m(x|x′) represents the transition probability of the target from state x to state x′ in mode m; p (l) (x′|m′) represents the spatial distribution of the target l being in state x′ given that it is in mode m′; η S (l) represents the normalization constant.

[0047] Among them, the normalization constant η S (l) is calculated by the formula: η S (l) = ∑p (l) (m′) ∫p s (l) (x′) p (l) (x′|m′) dx′.

[0048] Furthermore, the predicted existence probability is calculated by the formula:

[0049] Therefore, the predicted LMB RFS of the IMM-LMB filter is represented by the following π + formula.

[0050]

[0051] The multi-target posterior is still an LMB RFS, represented by the following formula.

[0052] π(·|Z) = {(r (l) , p (l) (m) p (l) (·|m))}

[0053] Among them, π(·|Z) represents the density of multi-targets given that the observation value is Z; π represents the density, and Z is the observation value of all targets.

[0054]

[0055] Among them, Θ represents the tracking label of the measurement assignment θ; represents a set; J + → {0, 1,... |Z|}, J + represents the number of groups of measurement assignments. The measurement assignment is unique. represents the updated component weight, w + (J + ) represents the weight of different measurement assignment groups in the multi-target posterior process; represents the normalization parameter after the measurement value is assigned.

[0056] The measurement update of the spatial distribution conditional on mode m is obtained by the following formula.

[0057]

[0058]

[0059] where p (l,θ) (x|m) represents the spatial distribution of the target l in state x given that it is in mode m after the measurement values have been assigned; represents the normalization constant for all targets after the measurement values have been assigned; ψ Z (x, l; θ) is the likelihood function; p D (x, l) is the detection probability of a single target; g(z θ(l) |x, l) is the likelihood function of z for a single target; κ(z θ(l) ) is the clutter intensity; z represents the measurement value of a single target. θ(l) represents the measurement value assigned to target l.

[0060] The updated probability p (l,θ) (m) of the target being in mode m is calculated by marginalizing the state x from the spatial distribution p (l,θ) (x, m) of target l after the measurement values have been assigned, as shown in the following equation.

[0061]

[0062] In another exemplary embodiment of the present application, before the neural networks in steps 103 and 104 are actually applied, they need to be trained. Before training, first design the intentions of the cluster spacecraft. The research on the intention recognition of spacecraft is based on relative motion dynamics. Therefore, first establish the orbital coordinate system PXYZ of one's own spacecraft, with the centroid of one's own spacecraft P as the origin of the coordinate system, the PZ axis along the radial direction of the spacecraft orbit, the PX axis perpendicular to the PZ axis and along the track direction, and the PY axis perpendicular to the orbital plane and determined by the right-hand rule, as Figure 4 shown. Assuming that the spacecraft all move along near-circular orbits and are not affected by perturbation forces, the relative dynamics equation, that is, the Clohessy-Wiltshire equation (CW equation), can be obtained.

[0063]

[0064] Further solving the CW equation gives:

[0065]

[0066] where x, y, z are the three-axis positions of the target spacecraft relative to one's own spacecraft, is the three-axis velocity, is the three-axis acceleration, and x0, y0, z0 are the initial values of the three-axis positions; is the initial value of the three-axis velocity, and n is the orbital angular velocity. By setting different initial position and velocity values, 8 single-target motion intentions can be obtained. Define I = {i1, i2, i3, i4, i5, i6, i7, i8} as the set of target spacecraft motion intentions. The motion schematic diagrams and initial values are as follows Figure 5 shown. The single-target motion intentions are: fixed-point, hover, oscillation, coplanar fly-around, non-coplanar fly-around, fly-by, jump, and spiral.

[0067] Since there will be perturbations in the actual scenarios of cluster spacecraft operation, and the intention recognition method studied in this application ignores these perturbations and only studies intention recognition theoretically. Additionally, the larger and more diverse the training sample size, the more accurate the neural network prediction results in steps 103 and 104.

[0068] In current research on cluster spacecraft, their motion states are mostly set as accompanying flight or fly-around. However, there may be multiple intentions in actual scenarios. Therefore, the present invention designs and defines 5 kinds of cluster spacecraft motion intentions as follows Figure 6 shown, namely: surround, escape, on-orbit maintenance, assembly, and dispersion camouflage. Based on the above 5 kinds of cluster spacecraft motion intentions, 4 kinds of simulation maneuver scenarios of cluster spacecraft can be designed, including the surround-to-escape scenario, the assembly-to-surround scenario, the on-orbit maintenance scenario, and the dispersion camouflage scenario. The spacecraft realizes the conversion of intentions through pulse maneuvers. After the intention design is completed, the first neural network is further trained with the single-target motion intention as the label, and the second neural network is trained with the cluster motion intention as the label.

[0069] The process of training the first neural network is as follows. In step 103, according to the position sequence, velocity sequence, and maneuver time of each spacecraft within a future preset time period, the trained first neural network is applied to obtain the single-target motion intention of each spacecraft within the future preset time period, specifically including:

[0070] (1) Set the initial simulation values of the position and velocity of each spacecraft under different simulation maneuver scenarios.

[0071] The simulation maneuver scenarios include the aforementioned surround-to-escape scenario, assembly-to-surround scenario, on-orbit maintenance scenario, and dispersion camouflage scenario.

[0072] (2) According to the initial simulation values of the position and velocity of each spacecraft, apply the labeled multi-Bernoulli algorithm based on the interacting multiple models to obtain the position simulation sequence, velocity simulation sequence, and simulation maneuver time of each spacecraft within the future simulation preset time period.

[0073] (3) According to the simulated maneuver time, divide the position simulation sequence and velocity simulation sequence of each spacecraft within the future simulated preset time period into the position simulation sequence and velocity simulation sequence before maneuver and the position simulation sequence and velocity simulation sequence after maneuver.

[0074] (4) Determine the single-target motion intention before simulation maneuver of each spacecraft (including the aforementioned fixed-point, hovering, oscillation, coplanar orbiting, non-coplanar orbiting, flyby, jump, and spiral) according to the position simulation sequence and velocity simulation sequence of each spacecraft before maneuver.

[0075] (5) Determine the single-target motion intention after simulation maneuver of each spacecraft (including the aforementioned fixed-point, hovering, oscillation, coplanar orbiting, non-coplanar orbiting, flyby, jump, and spiral) according to the position simulation sequence and velocity simulation sequence of each spacecraft after maneuver.

[0076] (6) Use the position simulation sequence and velocity simulation sequence of each spacecraft before maneuver and the position simulation sequence and velocity simulation sequence after maneuver as inputs, and the corresponding single-target motion intention before simulation maneuver and single-target motion intention after simulation maneuver as labels to train the first neural network. When the network loss converges during the training process, obtain the trained first neural network. Steps (1) to (6) train the first neural network through the simulation data of spacecraft under different maneuver scenarios.

[0077] (7) According to the position sequence, velocity sequence, and maneuver time of each spacecraft within the future preset time period, apply them to the trained first neural network to obtain the single-target motion intention of each spacecraft within the future preset time period.

[0078] In another exemplary embodiment of the present application, the observed values of the positions and velocities of the cluster spacecraft are input into the maneuver layer, and then the IMM-LMB algorithm is used to estimate the position, velocity, and maneuver time of the target. Before inputting the data into the intention layer and the cluster layer, it is divided into the pre-maneuver stage and the post-maneuver stage. For the target l (i.e., spacecraft l) that maneuvers at time k, by extracting the maximum a posteriori (MAP) state of l from LMBπ 1:k-1 the pre-maneuver trajectory S pre of the target l can be obtained, while the post-maneuver stage S post is the MAP state extracted from LMBπ k:n The data corresponding to the pre-maneuver state S pre is used to identify the initial intention of the target, while the post-maneuver stage S postThe data is used to identify the intention after maneuver. Therefore, in step (7), according to the position sequence, velocity sequence and maneuver time of each spacecraft within a preset future time period, the trained first neural network is applied to obtain the single-object motion intention of each spacecraft within the preset future time period, specifically including:

[0079] (1) According to the maneuver time, the position sequence and velocity sequence of each spacecraft within the preset future time period are divided into the position sequence and velocity sequence before maneuver and the position sequence and velocity sequence after maneuver.

[0080] (2) The position sequence and velocity sequence before maneuver and the position sequence and velocity sequence after maneuver are input into the trained first neural network, and the single-object initial motion intention before maneuver and the single-object motion intention after maneuver of each spacecraft are output; the single-object motion intention includes the single-object initial motion intention and the single-object motion intention after maneuver.

[0081] In another exemplary embodiment of the present application, the design of the BiGRU-Multi-HeadAttention neural network of the intention layer corresponding to step 103 can be divided into an input layer, a BiGRU layer, a multi-head attention mechanism layer and an output layer, as Figure 7 shown. The input layer obtains the estimated position and velocity time series before and after the target maneuver from the maneuver layer, and through unifying the data set, data format and batch size, and adding the intention label of the target (for the training of the neural network), a data structure that can be processed by the BiGRU layer is obtained.

[0082] GRU aims to solve the long-term dependence problem in the Recurrent Neural Network (RNN). It combines a reset gate and an update gate. The update gate determines the degree to which the unit updates its content, while the reset gate determines the degree to which the information of the previous hidden state is discarded. In addition, GRU combines the cell state and the hidden state, making the model simpler than the Long Short-Term Memory (LSTM) while maintaining performance and improving the convergence efficiency. The calculation process is as follows:

[0083]

[0084] In GRU, the update gate z t is calculated using a formula that combines the input at the current time step x t with the hidden state at the previous time step h t-1 , the weight matrices W z and U z of the update gate, and the bias b zCombined. The composition of the reset gate is structurally similar to the update gate, both maintaining and updating their respective weight matrices and biases. Specifically, the reset gate is parameterized by its weight matrix W r and U r as well as bias b r These parameters independently control the behavior of the gate and determine the degree to which previous hidden state information is discarded during the calculation process.

[0085] BiGRU is a bidirectional GRU that can capture the association between forward and backward features in time series information better than GRU, thus improving the network performance.

[0086] The multi-head attention mechanism extends the traditional attention mechanism, enabling the model to simultaneously focus on different aspects of the input sequence.

[0087]

[0088] Output = Concat(head1, head2,... head h )W O

[0089] where Q i = QW i Q K i = KW i K V i = VW i V . Q represents the query vector, K is the key vector, and K T is the transpose of the key vector. The similarity between the query and the key is evaluated through the dot product, which is used to assign weights to the corresponding values V. d k is the dimension of the key, used for scaling to avoid excessive dot products.

[0090] The output layer uses the softmax function to calculate the probabilities of different intents, and finally selects the intent with the highest probability as the output single-target motion intent.

[0091] As mentioned above, BiGRU can effectively process long-term sequence data, while the multi-head attention mechanism can enhance the generalization ability of the model. Therefore, the BiGRU-Multi-HeadAttention neural network can identify the intent of a single target.

[0092] In another exemplary embodiment of the present application, in the cluster layer corresponding to step 104, the maneuver layer and the intention layer process the motion and intention data of the spacecraft, and the cluster layer performs cluster clustering operations based on the combined input of the maneuver layer and the intention layer. The cluster layer also uses a BiGRU-Multi-HeadAttention neural network, but modifies the dimension of the input data. The input data includes the output of the maneuver layer and the output of the intention layer (single-target motion intention label), and the input data also includes a target label.

[0093] Target label l: This label comes from the IMM-LMB algorithm and is used to distinguish individual targets in the cluster.

[0094] Single-target motion intention label i: This label is the result of the single-target intention recognition process. The single-target motion intention label reflects the current state or purpose of the spacecraft, and it can change as the spacecraft executes maneuvers.

[0095] Specifically, the cluster motion intention label is determined by the network output. For a given cluster, although an individual spacecraft may change its intention before and after maneuvers, the cluster motion intention label remains unchanged. This ensures that spacecraft within the same cluster are consistently grouped, although their maneuver states or intentions may change. This method uses a BiGRU-Multi-HeadAttention neural network to process data and generate a dynamic and accurate spacecraft cluster that reflects the current and evolving behavior. The cluster layer inputs the recognition result of the cluster spacecraft as the cluster motion intention label into the maneuver layer. When executing the first intention recognition method, the cluster motion intention label does not need to be input into the maneuver layer. Here, C represents the clustering label, and C and l will be used as the basis for grouping and gating in the LMB algorithm. Grouping and gating are operations performed before updating the LMB filter, which allows updates to be executed in parallel on each group to greatly reduce the calculation time.

[0096] The specific technical effects of the present application are as follows:

[0097] First, there is currently no detailed definition of the intentions of cluster spacecraft, and the present application defines 5 kinds of intentions of cluster spacecraft and 4 kinds of simulation maneuver scenarios of cluster spacecraft.

[0098] Second, there is currently no research on the intention recognition of cluster spacecraft. The present application designs a MIC-Net to hierarchically solve the problem of intention recognition of cluster spacecraft.

[0099] Finally, in the intention layer, compared with the BiGRU, BiGRUAttention, and BiGRU-SelfAttention neural networks, the BiGRU-Multi-HeadAttention neural network of the present application has better intention recognition results, specifically manifested as higher accuracy and lower loss function.

[0100] The present application also provides an application scenario, which applies the above-mentioned method for identifying the intention of cluster spacecraft. Specifically: The method for identifying the intention of cluster spacecraft provided in this embodiment can be applied to the scenario of space traffic management during the operation of spacecraft. This scenario includes a data collection link, an intention recognition link, and a traffic management link; the data collection link is used to collect the initial motion values of each spacecraft; the intention recognition link is used to identify the intention of the cluster motion of spacecraft based on the collected initial motion values of each spacecraft; the traffic management link is used to perform space traffic management according to the recognition result of the intention recognition link. The method for identifying the intention of cluster spacecraft provided in this embodiment belongs to the intention recognition link.

[0101] In an exemplary embodiment, a computer device is provided. This computer device can be a server or a terminal, and its internal structure diagram can be as Figure 8 shown. This computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of this computer device is used to store the data of the intention recognition results of cluster spacecraft. The input / output interface of this computer device is used to exchange information between the processor and external devices. The communication interface of this computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it realizes a method for identifying the intention of cluster spacecraft.

[0102] Those skilled in the art can understand that Figure 8 the structure shown in

[0103] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, it implements the steps in the above-mentioned method embodiments.

[0104] In an exemplary embodiment, a computer program product is provided, including a computer program which, when executed by a processor, implements the steps in the above method embodiments.

[0105] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0106] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0107] The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0108] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0109] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for intention recognition of a cluster spacecraft, characterized in that The method for intention recognition of the cluster spacecraft includes: Obtaining the observed values of the position and velocity of each spacecraft; According to the previous cluster motion intention recognition results of each spacecraft, the observed values of the current position and velocity, and applying the labeled multi-Bernoulli algorithm based on the interacting multiple models, obtaining the position sequence, velocity sequence, and maneuver time of each spacecraft within a future preset time period; According to the position sequence, velocity sequence, and maneuver time of each spacecraft within a future preset time period, applying the trained first neural network to obtain the single-object motion intention of each spacecraft within a future preset time period; Inputting the position sequence, velocity sequence, and the corresponding single-object motion intention of each spacecraft within a future preset time period into the trained second neural network, and outputting the cluster motion intention of each spacecraft within a future preset time period; The spacecraft with the same cluster motion intention belong to the same cluster.

2. The method for identifying the intention of a cluster spacecraft according to claim 1, characterized in that According to the position sequence, velocity sequence, and maneuver time of each spacecraft within a future preset time period, applying the trained first neural network to obtain the single-object motion intention of each spacecraft within a future preset time period, specifically including: Setting the initial simulation values of the position and velocity of each spacecraft under different simulation maneuver scenarios; According to the initial simulation values of the position and velocity of each spacecraft, applying the labeled multi-Bernoulli algorithm based on the interacting multiple models to obtain the position simulation sequence, velocity simulation sequence, and simulation maneuver time of each spacecraft within a future simulation preset time period; According to the simulation maneuver time, dividing the position simulation sequence and velocity simulation sequence of each spacecraft within a future simulation preset time period into the position simulation sequence and velocity simulation sequence before maneuver and the position simulation sequence and velocity simulation sequence after maneuver; Determining the single-object motion intention before simulation maneuver of each spacecraft according to the position simulation sequence and velocity simulation sequence before maneuver of each spacecraft; Determining the single-object motion intention after simulation maneuver of each spacecraft according to the position simulation sequence and velocity simulation sequence after maneuver of each spacecraft; Using the position simulation sequence and velocity simulation sequence before maneuver and the position simulation sequence and velocity simulation sequence after maneuver of each spacecraft as inputs, and using the corresponding single-object motion intention before simulation maneuver and single-object motion intention after simulation maneuver as labels to train the first neural network. When the network loss converges during the training process, the trained first neural network is obtained; According to the position sequence, velocity sequence, and maneuver time of each spacecraft within a future preset time period, applying the trained first neural network to obtain the single-object motion intention of each spacecraft within a future preset time period.

3. The method for identifying the intention of a cluster spacecraft according to claim 1 or 2, characterized in that According to the position sequence, velocity sequence, and maneuver time of each spacecraft within a future preset time period, applying the trained first neural network to obtain the single-object motion intention of each spacecraft within a future preset time period, specifically including: According to the maneuver time, dividing the position sequence and velocity sequence of each spacecraft within a future preset time period into the position sequence and velocity sequence before maneuver and the position sequence and velocity sequence after maneuver; Input the position sequence and velocity sequence before maneuver and the position sequence and velocity sequence after maneuver into the trained first neural network, and output the single-object initial motion intention before maneuver and the single-object motion intention after maneuver for each spacecraft; the single-object motion intention includes the single-object initial motion intention and the single-object motion intention after maneuver.

4. The method for identifying the intention of a cluster spacecraft according to claim 1, characterized in that, The first neural network and the second neural network are neural networks based on the multi-head attention mechanism and the bidirectional gated recurrent unit.

5. The method for identifying the intention of a cluster spacecraft according to claim 4, characterized in that, The first neural network includes an input layer, a BiGRU layer, a multi-head attention mechanism layer, and an output layer connected in sequence.

6. The method for identifying the intention of a cluster spacecraft according to claim 1, wherein The single-object motion intention includes fixed-point, hovering, oscillation, coplanar orbiting, non-coplanar orbiting, flyby, jump, and spiral. The swarm motion intention includes: surrounding, escaping, on-orbit maintenance, assembling, and dispersion camouflage.

7. The method for identifying the intention of a cluster spacecraft according to claim 2, characterized in that, The simulation maneuver scenarios include: surrounding-to-escape scenario, assembling-to-surrounding scenario, on-orbit maintenance scenario, and dispersion camouflage scenario.

8. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the intention recognition method for cluster spacecraft according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intention recognition method for cluster spacecraft according to any one of claims 1-7.

10. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the intention recognition method for cluster spacecraft according to any one of claims 1-7.

Images