Reentry body trajectory prediction method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a reentry body trajectory prediction method and device, equipment and a storage medium. The method comprises the following steps: establishing a plurality of trajectory models based on a kinetic equation of a gliding section and different energy management strategy constraint conditions, simulating each trajectory model to obtain a corresponding simulation trajectory data sequence, preprocessing each simulation trajectory data sequence to obtain a plurality of training samples, inputting each training sample into a trajectory prediction model, and predicting the trajectory of the gliding section according to the trajectory prediction model. And performing trajectory classification training on the trajectory prediction model, performing trajectory prediction training for different trajectory types, and inputting current trajectory data of the reentry body into the trained trajectory prediction model to obtain predicted trajectory data of the reentry body. According to the method, the accuracy of reentry trajectory prediction can be improved.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of reentry body control technology, and in particular, to a reentry body trajectory prediction method, apparatus, device, and storage medium. Background Art

[0002] When the reentry body enters the atmosphere, its trajectory prediction and landing point calculation are key links to ensure safe recovery. However, due to the existence of a plasma sheath in the blackout area, which may cause communication interruption, sensor failure, or limited telemetry data bandwidth, only sparse latitude, longitude, and altitude information can be obtained.

[0003] Current technologies mainly fall into two categories: prediction methods based on physical models and prediction methods based on data-driven machine learning models. The prediction effect of the former depends on accurate initial states and real-time environmental parameters, while the latter requires continuous, high-frequency input data to train prediction models based on historical trajectory data. It has poor adaptability to sparse telemetry data, and the prediction error increases significantly with the increase in data missing rate. Therefore, the accuracy of these two prediction methods is relatively low. Summary of the Invention

[0004] The embodiments described herein provide a reentry body trajectory prediction method, apparatus, device, and storage medium, which can improve the accuracy of reentry body trajectory prediction.

[0005] In a first aspect, the present disclosure provides a method for predicting a reentry trajectory, comprising:

[0006] Based on the dynamic equations of the gliding segment and different energy management strategy constraints, multiple trajectory models are established; each trajectory model is simulated to obtain a corresponding simulation trajectory data sequence, where each simulation trajectory data in the simulation trajectory data sequence includes longitude, latitude and altitude; each simulation trajectory data sequence is preprocessed to obtain multiple training samples; each training sample is input into a trajectory prediction model, and the trajectory prediction model is trained for trajectory classification and trajectory prediction training is performed for different trajectory types; the current trajectory data of the reentry body is input into the trained trajectory prediction model to obtain the predicted trajectory data of the reentry body.

[0007] In some embodiments of the present disclosure, the trajectory prediction model includes a classifier and a prediction network, the prediction network is a Seq2Seq network, and includes an encoder, a decoder and a deformable cross attention mechanism layer, the output end of the classifier is connected to the input end of the encoder, the input end of the deformable cross attention mechanism layer is connected to the residual and the output end of the encoder, and the output end of the deformable cross attention mechanism layer is connected to the input end of the decoder.

[0008] In some embodiments of the present disclosure, inputting each training sample into a trajectory prediction model, performing trajectory classification training on the trajectory prediction model, and performing trajectory prediction training for different trajectory types includes:

[0009] Each training sample is input into the classifier to obtain the trajectory type corresponding to each training sample to train the classifier, and the trajectory type corresponds to the trajectory model one-to-one; each training sample and the trajectory type output by the corresponding classifier are input into the prediction network, the attention range is adjusted based on the trajectory type, and the prediction network is trained according to the residual and the training sample.

[0010] In some embodiments of the present disclosure, preprocessing each of the simulation trajectory data sequences to obtain a plurality of training samples includes:

[0011] The simulation trajectory data in each simulation trajectory data sequence are randomly discarded; based on piecewise cubic spline interpolation, the simulation trajectory data in each discarded simulation trajectory data sequence are fitted to obtain a trajectory data fitting function; an observation coordinate system is established with the fitting trajectory data at the end time of the black barrier area as the origin, and multiple trajectory data fitting functions are converted from the ground coordinate system to the observation coordinate system to obtain multiple observation trajectory data sequences; and multiple training samples are determined according to each observation trajectory data sequence.

[0012] In some embodiments of the present disclosure, determining the plurality of training samples according to each of the observation trajectory data sequences includes:

[0013] Based on the first time window, each of the observation trajectory data sequences is subjected to sliding cutting to obtain multiple first observation trajectory data cutting sequences to obtain multiple input samples; based on the second time window, each of the observation trajectory data sequences is subjected to sliding cutting to obtain multiple second observation trajectory data cutting sequences to obtain multiple output samples; the observation trajectory data in each of the input samples and each of the output samples is normalized to obtain the training sample.

[0014] In some embodiments of the present disclosure, the energy management strategy constraints include angle of attack constraints and / or roll angle constraints; the multiple trajectory models include a balanced gliding trajectory model, a jumping gliding trajectory model, a lateral C-shaped trajectory model and a lateral S-shaped trajectory model.

[0015] In some embodiments of the present disclosure, the reentry trajectory prediction method further includes:

[0016] The predicted trajectory data is disturbed to generate a plurality of random trajectories; and the landing point of the reentry body is predicted based on the landing point distribution of all the random trajectories.

[0017] In a second aspect, the present disclosure provides a reentry trajectory prediction device, comprising:

[0018] A simulation module is used to simulate each of the trajectory models to obtain a corresponding simulation trajectory data sequence, wherein each simulation trajectory data in the simulation trajectory data sequence includes longitude, latitude and altitude; a preprocessing module is used to preprocess each of the simulation trajectory data sequences to obtain multiple training samples; a training module is used to input each training sample into a trajectory prediction model, perform trajectory classification training on the trajectory prediction model, and perform trajectory prediction training for different trajectory types; a prediction module is used to input the current trajectory data of the reentry body into the trained trajectory prediction model to obtain the predicted trajectory data of the reentry body.

[0019] In a third aspect, the present disclosure provides an electronic device comprising a processor, wherein the processor is configured to execute a computer program stored in a memory, wherein the computer program, when executed by the processor, implements the steps of any one of the methods provided in the first aspect.

[0020] In a fourth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any method provided in the first aspect when executed by a processor.

[0021] In the technical solution provided by the present disclosure, multiple trajectory models are established based on the dynamic equations of the gliding segment and different energy management strategy constraints, each trajectory model is simulated to obtain a corresponding simulation trajectory data sequence, each simulation trajectory data sequence is preprocessed to obtain multiple training samples, each training sample is input into a trajectory prediction model, the trajectory prediction model is trained for trajectory classification, and trajectory prediction training is performed for different trajectory types, the current trajectory data of the re-entry body is input into the trained trajectory prediction model to obtain the predicted trajectory data of the re-entry body, the trajectory prediction model can be trained based on the simulation trajectory data of multiple trajectory models, energy management and trajectory prediction can be deeply integrated to realize multimodal trajectory prediction of the re-entry body, and energy management can also be embedded in the online prediction framework to dynamically correct the trajectory deviation caused by interference from the blackout area, thereby improving the accuracy of the re-entry body trajectory prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments will be briefly described below. It should be noted that the drawings described below only relate to some embodiments of the present disclosure and are not intended to limit the present disclosure.

[0023] Figure 1 A flowchart of a reentry trajectory prediction method provided by an embodiment of the present disclosure.

[0024] Figure 2 A schematic diagram of a coordinate system provided for an embodiment of the present disclosure.

[0025] Figure 3 and Figure 4 A schematic diagram of the process of obtaining training samples provided in an embodiment of the present disclosure.

[0026] Figure 5 This is an architectural diagram of the trajectory prediction model provided in an embodiment of the present disclosure.

[0027] Figure 6 A flowchart of another reentry trajectory prediction method provided by an embodiment of the present disclosure.

[0028] Figure 7 A schematic structural diagram of a reentry trajectory prediction device provided in an embodiment of the present disclosure.

[0029] Figure 8 A schematic structural diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the embodiments of the present disclosure more clear, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the described embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without creative work also fall within the scope of protection of the present disclosure.

[0031] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which the subject matter of the present disclosure belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the specification and the relevant art, and will not be interpreted in an idealized or overly formal manner unless otherwise explicitly defined herein. In addition, terms such as "first" and "second" are only used to distinguish one component (or a portion of a component) from another component (or another portion of a component).

[0032] When the reentry body returns to the atmosphere, there are problems such as the plasma sheath in the blackout area causing communication interruption, sensor failure or limited telemetry data bandwidth. In fact, only sparse latitude, longitude and altitude information can be obtained. The trajectory of the reentry body can be predicted through a physical model or a data-driven machine learning model.

[0033] Among them, the prediction method based on physical models establishes the motion equation of the re-entry body, combines parameters such as atmospheric drag, lift and gravitational acceleration, and performs numerical integration or simplified analytical solution calculations. However, its prediction effect is highly dependent on accurate initial conditions and real-time environmental parameters, making it difficult to guarantee the accuracy of the re-entry body trajectory prediction.

[0034] The prediction method of the data-driven machine learning model uses time series networks such as LSTM and Transformer. Training the prediction model based on historical trajectory data requires continuous and high-frequency input data. Therefore, it has poor adaptability to sparse telemetry data. The prediction error increases significantly with the increase of data missing rate, affecting the accuracy of re-entry trajectory prediction. In the sparse telemetry scenario in the blackout area, the re-entry trajectory prediction faces the problem of poor accuracy.

[0035] In view of this, the present disclosure provides a re-entry body trajectory prediction method, which establishes multiple trajectory models based on the dynamic equations of the gliding segment and different energy management strategy constraints, simulates each trajectory model to obtain a corresponding simulation trajectory data sequence, pre-processes each simulation trajectory data sequence to obtain multiple training samples, inputs each training sample into the trajectory prediction model, performs trajectory classification training on the trajectory prediction model, and performs trajectory prediction training for different trajectory types, inputs the current trajectory data of the re-entry body into the trained trajectory prediction model to obtain the predicted trajectory data of the re-entry body, and can train the trajectory prediction model based on the simulation trajectory data of multiple trajectory models, deeply integrate energy management and trajectory prediction to realize multimodal trajectory prediction of the re-entry body, and can also embed energy management into the online prediction framework to dynamically correct the trajectory deviation caused by interference from the blackout area, thereby improving the accuracy of the re-entry body trajectory prediction.

[0036] The technical solutions provided by the present disclosure are described in detail below with reference to several specific embodiments.

[0037] Figure 1 A flow chart of a reentry trajectory prediction method provided by an embodiment of the present disclosure is shown as follows: Figure 1 As shown, the specific steps of the re-entry trajectory prediction method include:

[0038] S101: Establish multiple trajectory models based on the dynamic equations of the gliding phase and different energy management strategy constraints.

[0039] For example, Figure 2 A schematic diagram of the coordinate system provided by the present disclosure, such as Figure 2 As shown in the figure, with the launch point of the reentry body as the origin, the ground coordinate system o-xyz is established. In the ground coordinate system o-xyz, the dynamic equation of the reentry body gliding phase is established:

[0040]

[0041] Where V is the velocity of the reentry body, γ is the trajectory inclination, that is, the angle between the velocity vector and the horizontal plane, χ is the trajectory deflection, that is, the angle between the projection of the velocity vector on the horizontal plane and the y-axis, ε is the roll angle, that is, the angle between the longitudinal axis of the reentry body and the horizontal plane, (x, y, z) is the position coordinate of the reentry body in the ground coordinate system o-xyz, m is the mass of the reentry body, in kg, L and D are the aerodynamic lift and aerodynamic drag on the reentry body, respectively, in N, and g is the acceleration due to gravity, in m / s 2 .

[0042] The energy management strategy constraints include angle of attack constraints and / or roll angle constraints, wherein the angle of attack constraints can be an angle of attack model of a jump glide trajectory, namely:

[0043]

[0044] Where α(t) is the function of the angle of attack changing with time t, α1 and α2 are the minimum and maximum angles of attack of the reentry body, respectively, and V1 and V2 are the minimum and maximum velocities of the reentry body, respectively.

[0045] The angle of attack constraint can also be a balanced angle of attack model of the gliding trajectory, that is:

[0046]

[0047] Where α is the angle of attack, S is the wing reference area, r is the trajectory radius, and ρ is the atmospheric density.

[0048] The roll angle constraint condition can be a roll angle model of a lateral C-shaped trajectory, namely:

[0049] ε(t)=C (4)

[0050] Where t is time and C is an angle constant.

[0051] The roll angle constraint condition can also be a roll angle model of a lateral S-shaped trajectory, that is:

[0052]

[0053] Wherein, ε1 is an angle constant and ε1∈(-30°, 60°), and t1 is a time constant.

[0054] Different trajectory models can be established based on the dynamic equations of the re-entry body gliding segment and different angle of attack constraints. For example, a jumping glide trajectory model can be established based on the dynamic equations of the re-entry body gliding segment and the angle of attack model of the jumping glide trajectory; and a balanced glide trajectory model can be established based on the dynamic equations of the re-entry body gliding segment and the angle of attack model of the balanced glide trajectory.

[0055] Alternatively, a trajectory model can be established based on the dynamic equations of the re-entry body gliding segment, the angle of attack constraints and different roll angle constraints. For example, a lateral C-shaped trajectory model can be established based on the dynamic equations of the re-entry body gliding segment, the angle of attack model of the balanced gliding trajectory and the roll angle model of the lateral C-shaped trajectory; and a lateral S-shaped trajectory model can be established based on the dynamic equations of the re-entry body gliding segment, the angle of attack model of the balanced gliding trajectory and the roll angle model of the lateral S-shaped trajectory.

[0056] In this way, the dynamic equations of the re-entry body glide phase and different energy management strategy constraints can be used to establish multiple trajectory models including the balanced glide trajectory model, the jumping glide trajectory model, the lateral C-shaped trajectory and the lateral S-shaped trajectory model.

[0057] S102: Simulate each trajectory model to obtain a corresponding simulation trajectory data sequence.

[0058] For example, each trajectory model can be simulated by controlling the angle of attack. For example, by controlling the angle of attack based on formula (2), a jumping gliding trajectory model can be simulated. By controlling the angle of attack based on formula (3), a balanced gliding trajectory model can be simulated.

[0059] In other embodiments, each trajectory model can be simulated by controlling the angle of attack and the roll angle. For example, by controlling the angle of attack based on formula (2) and the roll angle based on formula (3), a lateral C-shaped trajectory model can be simulated. By controlling the angle of attack based on formula (2) and the roll angle based on formula (4), a lateral S-shaped trajectory model can be simulated.

[0060] For the same trajectory model, by setting different initial velocities and / or altitudes for simulation, multiple continuous simulation trajectory data can be obtained, where the simulation trajectory data includes longitude, latitude and altitude, so that multiple simulation trajectory data sequences of each trajectory type can be obtained.

[0061] For example, by simulating a jump-glide trajectory model, a plurality of continuous jump-glide simulation trajectory data can be obtained, thereby obtaining a jump-glide simulation trajectory data sequence. By simulating a balanced glide trajectory model, a plurality of continuous balanced glide simulation trajectory data can be obtained, thereby obtaining a balanced glide simulation trajectory data sequence. By simulating a transverse C-shaped trajectory model, a plurality of continuous transverse C-shaped simulation trajectory data can be obtained, thereby obtaining a transverse C-shaped simulation trajectory data sequence. By simulating a transverse S-shaped trajectory model, a plurality of continuous transverse S-shaped simulation trajectory data can be obtained, thereby obtaining a transverse S-shaped simulation trajectory data sequence.

[0062] S103: Preprocess each simulation trajectory data sequence to obtain multiple training samples.

[0063] Exemplarily, each training sample includes an input sample and an output sample. First, based on the first time window, each simulation trajectory data sequence is subjected to sliding cutting to obtain multiple first simulation trajectory data cutting sequences, and a single first simulation trajectory data cutting sequence is used as an input sample to obtain multiple input samples, such as Figure 3 and Figure 4 As shown, Figure 3 and Figure 4 A schematic diagram of the process of obtaining training samples provided by an embodiment of the present disclosure. For example, the first time window is N seconds, and the cutting interval is T seconds, where N and T are positive integers, and T is less than N.

[0064] Secondly, based on the second time window, each simulation trajectory data sequence is slidingly cut to obtain multiple second simulation trajectory data cutting sequences, and each second simulation trajectory data cutting sequence is used as the output sample corresponding to the previous input sample of the second simulation trajectory data cutting sequence to obtain multiple output samples, such as Figure 3 and Figure 4 As shown, for example, the second time window is M seconds, and M can be greater than, less than, or equal to N.

[0065] Based on different task requirements, the output sample can be a continuous sequence of the corresponding input samples, such as Figure 3 As shown, and / or, the output samples can be an interval sequence of corresponding input samples, such as Figure 4 As shown, the present disclosure does not make any specific limitation to this.

[0066] Finally, the longitude, latitude, and altitude in each input sample are normalized, and the longitude, latitude, and altitude in each output sample are normalized, so as to normalize the simulation trajectory data in each input sample and each output sample. The normalized single input sample and the corresponding output sample constitute a training sample, thereby obtaining multiple training samples.

[0067] Based on the simulation trajectory data sequences of different trajectory types, training samples of different trajectory types can be obtained, that is, each training sample corresponds to a trajectory type, and the trajectory type is used as the label corresponding to the training sample.

[0068] S104 , based on the trajectory classification task and the trajectory prediction task, input each training sample into the trajectory prediction model, perform trajectory classification training on the trajectory prediction model, and perform trajectory prediction training for different trajectory types.

[0069] Exemplarily, each training sample and its corresponding label are input into the trajectory prediction model, and based on each input sample and its corresponding label, the trajectory prediction model is trained for trajectory classification, and for different trajectory types, the trajectory prediction model is trained for trajectory prediction based on each training sample, so that the trained trajectory prediction model can classify the trajectory of the re-entry body and predict the trajectory of the re-entry body for different trajectory types.

[0070] S105 , inputting the current trajectory data of the re-entry body into the trained trajectory prediction model to obtain the predicted trajectory data of the re-entry body.

[0071] Illustratively, based on the above embodiment, the trained trajectory prediction model can classify the trajectory of the reentry body based on the current trajectory data of the reentry body, and predict the trajectory of the reentry body based on the trajectory type of the reentry body to obtain the predicted trajectory data of the reentry body.

[0072] In summary, the embodiments of the present disclosure establish multiple trajectory models, simulate the multiple trajectory models, obtain simulated trajectory data sequences of multiple trajectory types, and thus obtain training samples corresponding to the multiple trajectory types. The trajectory prediction model is trained for trajectory classification through the training samples, and trajectory prediction training is performed for different trajectory types. The trajectory prediction model can be trained based on the simulated trajectory data of multiple trajectory models, and energy management and trajectory prediction can be deeply integrated to achieve multimodal trajectory prediction of the re-entry body. Energy management can also be embedded in the online prediction framework to dynamically correct trajectory deviations caused by interference from blackout areas, thereby improving the accuracy of re-entry trajectory prediction.

[0073] In some embodiments, Figure 5 This is an architecture diagram of the trajectory prediction model provided by the embodiment of the present disclosure, such as Figure 5 As shown in Figure 1, the trajectory prediction model includes a classifier and a prediction network. The prediction network is a Seq2Seq network consisting of an encoder, a decoder, and a deformable cross-attention mechanism layer. The output of the classifier is connected to the input of the encoder, the input of the deformable cross-attention mechanism layer is connected to the residual and the output of the encoder, and the output of the deformable cross-attention mechanism layer is connected to the input of the decoder.

[0074] For example, the classifier consists of an input layer, an LSTM network, and a fully connected network. The input to the classifier is the input sample, and the output is the probability distribution of the trajectory type. The prediction network uses the Seq2Seq framework as the basic framework, with both the encoder and decoder using LSTM networks, connected using a deformable cross-attention mechanism and residuals.

[0075] In this way, the embodiment of the present disclosure can shorten the time of re-entry trajectory prediction by setting a lightweight Seq2Seq prediction network, thereby improving the real-time performance of re-entry trajectory prediction.

[0076] In some embodiments, based on the trajectory prediction model architecture provided in the above embodiments, a specific description of a possible implementation method for executing S104 is as follows:

[0077] First, each training sample is input into the classifier to obtain the trajectory type corresponding to each training sample to train the classifier. The trajectory type corresponds to the trajectory model one by one.

[0078] Exemplarily, each input sample and its corresponding label are input into the classifier, and the classifier can classify the trajectory of the input sample. According to the trajectory type output by the classifier and the label corresponding to the input sample, the parameters of the classifier are adjusted until the loss function of the classifier converges and the training is completed, thereby obtaining a trained classifier.

[0079] Secondly, each training sample and its corresponding trajectory type output by the classifier are input into the prediction network, the attention range is adjusted based on the trajectory type, and the prediction network is trained based on the residual and training samples.

[0080] Exemplarily, based on the trajectory type received by the prediction network, the attention range is dynamically adjusted to train the prediction network for different trajectory types, and the residual connection is combined to suppress the accumulated error to improve the accuracy of the prediction network.

[0081] In this way, the embodiments of the present disclosure can establish a model that deeply integrates energy management and trajectory prediction, can adjust the attention range of the model according to the trajectory type, and combine residual connections to suppress cumulative errors, thereby improving the accuracy of the model.

[0082] In some embodiments, Figure 6 A flowchart of another reentry trajectory prediction method provided by an embodiment of the present disclosure is shown below. Figure 6 for Figure 1 Based on the illustrated embodiment, a possible implementation method for executing S103 is described in detail as follows:

[0083] S201 : randomly discarding simulation trajectory data in each simulation trajectory data sequence.

[0084] Exemplarily, the simulated trajectory data in the simulated trajectory data sequences of multiple trajectory types are randomly discarded to simulate blackout area interference.

[0085] S202 , fitting the simulation trajectory data in each discarded simulation trajectory data sequence based on piecewise cubic spline interpolation to obtain a trajectory data fitting function.

[0086] For example, based on the energy continuity assumption, piecewise cubic spline interpolation is used to fit the simulated trajectory data in the discarded simulated trajectory data sequences of multiple trajectory types to obtain trajectory data fitting functions of multiple trajectory types. The trajectory data fitting function can be expressed as:

[0087] S(t)=a i (tt i ) 3 +b i (tt i ) 2 +c i (tt i )+d i (6)

[0088] Where S(t) represents the longitude fitting function, latitude fitting function or altitude fitting function of the reentry body, t i Indicates the starting time of the i-th time interval, which is used to divide the interpolation interval of sparse data. i 、b i 、c i and d i represents the cubic polynomial coefficient of the i-th interval, which is solved by continuous optimization of energy derivatives, a i 、b i 、c i and d i The formula determines the smoothness and rationality of the trajectory in this interval.

[0089] For example, the jumping and gliding simulation trajectory data in the discarded jumping and gliding simulation trajectory data sequence are fitted to obtain the jumping and gliding trajectory data fitting function; the balanced gliding simulation trajectory data in the discarded balanced gliding simulation trajectory data sequence are fitted to obtain the balanced gliding trajectory data fitting function; the transverse C-shaped simulation trajectory data in the discarded transverse C-shaped simulation trajectory data sequence are fitted to obtain the transverse C-shaped trajectory data fitting function; the transverse S-shaped simulation trajectory data in the discarded transverse S-shaped simulation trajectory data sequence are fitted to obtain the transverse S-shaped trajectory data fitting function.

[0090] S203 , establishing an observation coordinate system with the fitting trajectory data at the end time of the blackout area as the origin, and converting the multiple trajectory data fitting functions from the ground coordinate system to the observation coordinate system to obtain multiple observation trajectory data sequences.

[0091] For example, the end time of the black barrier area is determined, and the end time of the black barrier area is substituted into the longitude fitting function, latitude fitting function and altitude fitting function respectively to obtain the fitting longitude, latitude and altitude of the black barrier area at the end time. The fitting longitude, latitude and altitude of the black barrier area at the end time are used as the origin to establish the observation coordinate system o'-x'y'z', as shown in Figure 2 shown.

[0092] The trajectory data fitting functions of the above-mentioned multiple trajectory types are functions under the ground coordinate system. By converting the trajectory data fitting functions of the multiple trajectory types from the ground coordinate system o-xyz to the observation coordinate system o'-x'y'z', the trajectory data fitting functions of the multiple trajectory types under the observation coordinate system o'-x'y'z' can be obtained. Based on the trajectory data fitting functions under the observation coordinate system o'-x'y'z', multiple observation trajectory data corresponding to each trajectory type can be obtained, thereby obtaining observation trajectory data sequences of the multiple trajectory types.

[0093] For example, based on the jumping glide trajectory data fitting function in the observation coordinate system o'-x'y'z', multiple jumping glide observation trajectory data can be obtained to obtain a jumping glide observation trajectory data sequence; based on the balance glide trajectory data fitting function in the observation coordinate system o'-x'y'z', multiple balance glide observation trajectory data can be obtained to obtain a balance glide observation trajectory data sequence; based on the lateral C-shaped trajectory data fitting function in the observation coordinate system o'-x'y'z', multiple lateral C-shaped observation trajectory data can be obtained to obtain a lateral C-shaped observation trajectory data sequence; based on the lateral S-shaped trajectory data fitting function in the observation coordinate system o'-x'y'z', multiple lateral S-shaped observation trajectory data can be obtained to obtain a lateral S-shaped observation trajectory data sequence.

[0094] S204: Determine a plurality of training samples according to each observation trajectory data sequence.

[0095] Exemplarily, first, based on a first time window, a sliding cut is performed on each observation trajectory data sequence to obtain multiple first observation trajectory data cut sequences. A single first observation trajectory data cut sequence is used as an input sample to obtain multiple input samples. For example, the first time window is N seconds, and the cut interval is T seconds, where N and T are positive integers, and T is less than N.

[0096] Next, based on the second time window, each observation trajectory data sequence is subjected to a sliding segmentation to obtain multiple second observation trajectory data segmentation sequences. Each second observation trajectory data segmentation sequence is used as the output sample corresponding to the previous input sample of the second observation trajectory data segmentation sequence to obtain multiple output samples. For example, the second time window is M seconds, where M can be greater than, less than, or equal to N.

[0097] Based on different task requirements, the output sample can be a continuous sequence of the corresponding input samples, such as Figure 3 As shown, and / or, the output samples can be an interval sequence of corresponding input samples, such as Figure 4 As shown, the present disclosure does not make any specific limitation to this.

[0098] Finally, the longitude, latitude, and altitude in each input sample are normalized respectively, and the longitude, latitude, and altitude in each output sample are normalized respectively to normalize the simulation trajectory data in each input sample and each output sample. The normalized single input sample and the corresponding output sample constitute a training sample, thereby obtaining a training sample.

[0099] In this way, the embodiment of the present disclosure simulates the interference of the black barrier area by randomly discarding the simulated trajectory data in each simulated trajectory data sequence, and fits the simulated trajectory data in each discarded simulated trajectory data sequence based on piecewise cubic spline interpolation to obtain a trajectory data fitting function, establishes an observation coordinate system with the fitted trajectory data at the end moment of the black barrier area as the origin, and converts multiple trajectory data fitting functions from the ground coordinate system to the observation coordinate system to obtain multiple observation trajectory data sequences. According to each observation trajectory data sequence, multiple training samples are determined, which can increase the number of training samples and avoid the problem of reduced trajectory prediction accuracy due to missing black barrier area data.

[0100] In some embodiments, after executing S105, the re-entry trajectory prediction method further includes:

[0101] S106 , perturbing the predicted trajectory data to generate multiple random trajectories.

[0102] Exemplarily, the atmospheric density disturbance obeys a uniform distribution, for example, ±10%, and the wind speed disturbance obeys a Gaussian distribution, for example, with a mean of ±20 m / s and a standard deviation of 5 m / s. Based on this, the predicted trajectory data is disturbed and multiple random trajectories are generated, for example, 1,000 random trajectories.

[0103] S107, predicting the landing point of the reentry body based on the landing point distribution corresponding to all random trajectories.

[0104] Exemplarily, the landing points of all random trajectories are determined to obtain the landing point distribution of all random trajectories, the kernel density component calculation method is used to calculate the landing point distribution density, and an 80% confidence region is marked to obtain the landing point of the re-entry body.

[0105] In this way, the predicted trajectory data is disturbed to generate multiple random trajectories. According to the landing point distribution corresponding to all random trajectories, the landing point of the reentry body is predicted. Monte Carlo simulation can be used to achieve rapid generation of the landing point area, meet the rapid response requirements, and improve the reentry body search efficiency and recovery success rate.

[0106] The present disclosure also provides a re-entry trajectory prediction device, Figure 7 A schematic diagram of the structure of a reentry trajectory prediction device provided by an embodiment of the present disclosure is shown in FIG. Figure 7 As shown, the reentry trajectory prediction device includes:

[0107] The model building module 110 is used to build multiple trajectory models based on the dynamic equations of the gliding segment and different energy management strategy constraints.

[0108] The simulation module 120 is used to simulate each trajectory model to obtain a corresponding simulation trajectory data sequence. Each simulation trajectory data in the simulation trajectory data sequence includes longitude, latitude and altitude.

[0109] The preprocessing module 130 is used to preprocess each simulation trajectory data sequence to obtain multiple training samples.

[0110] The training module 140 is used to input each training sample into the trajectory prediction model, perform trajectory classification training on the trajectory prediction model, and perform trajectory prediction training for different trajectory types.

[0111] The prediction module 150 is configured to input the current trajectory data of the reentry body into the trained trajectory prediction model to obtain the predicted trajectory data of the reentry body.

[0112] In some embodiments, the trajectory prediction model includes a classifier and a prediction network. The prediction network is a Seq2Seq network and includes an encoder, a decoder, and a deformable cross-attention mechanism layer. The output of the classifier is connected to the input of the encoder, the input of the deformable cross-attention mechanism layer is connected to the residual and the output of the encoder, and the output of the deformable cross-attention mechanism layer is connected to the input of the decoder.

[0113] In some embodiments, the training module 140 is further used to input each training sample into a classifier to obtain a trajectory type corresponding to each training sample to train the classifier, and the trajectory type corresponds to the trajectory model one-to-one; each training sample and the trajectory type output by its corresponding classifier are input into the prediction network, the attention range is adjusted based on the trajectory type, and the prediction network is trained based on the residual and the training sample.

[0114] In some embodiments, the preprocessing module 130 is further used to randomly discard the simulated trajectory data in each simulated trajectory data sequence; based on piecewise cubic spline interpolation, the simulated trajectory data in each discarded simulated trajectory data sequence is fitted to obtain a trajectory data fitting function; an observation coordinate system is established with the fitted trajectory data at the end time of the black barrier area as the origin, and multiple trajectory data fitting functions are converted from the ground coordinate system to the observation coordinate system to obtain multiple observation trajectory data sequences; and multiple training samples are determined according to each observation trajectory data sequence.

[0115] In some embodiments, the preprocessing module 130 is further used to perform sliding cutting on each observation trajectory data sequence based on the first time window to obtain multiple first observation trajectory data cutting sequences to obtain multiple input samples; perform sliding cutting on each observation trajectory data sequence based on the second time window to obtain multiple second observation trajectory data cutting sequences to obtain multiple output samples; and normalize the observation trajectory data in each input sample and each output sample to obtain a training sample.

[0116] In some embodiments, the energy management strategy constraints include attack angle constraints and roll angle constraints, and the multiple trajectory models include a balanced gliding trajectory model, a jumping gliding trajectory model, a lateral C-shaped trajectory model, and a lateral S-shaped trajectory model.

[0117] In some embodiments, the prediction module 150 is further configured to perturb the predicted trajectory data to generate multiple random trajectories; and predict the landing point of the reentry body based on the landing point distribution of all random trajectories.

[0118] The device provided by the embodiment of the present disclosure is used to execute the steps provided by any of the above method embodiments, has functional modules corresponding to the method embodiments, and has the beneficial effects of the method embodiments, which will not be repeated here.

[0119] The present disclosure further provides an electronic device, comprising: a processor, wherein the processor is configured to execute a computer program stored in a memory, wherein the computer program implements the steps of the method embodiment of the present disclosure when executed by the processor.

[0120] Figure 8 This is a schematic diagram of the structure of an electronic device provided by the present disclosure. Figure 8 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present disclosure is shown. Figure 8 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0121] like Figure 8As shown, the electronic device 12 is implemented as a general-purpose computing device. Components of the electronic device 12 may include, but are not limited to, one or more processors 16, a system memory 28, and a bus 18 connecting various system components (including the system memory 28 and the processor 16).

[0122] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0123] The electronic device 12 typically includes a variety of computer system readable media. These media can be any media that can be accessed by the electronic device 12, including volatile and non-volatile media, removable and non-removable media.

[0124] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). A magnetic disk drive for reading and writing to a removable non-volatile magnetic disk (such as a "floppy disk") may be provided, as well as an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical media). In these cases, each drive may be connected to the bus 18 via one or more data media interfaces. The system memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of various embodiments of the present disclosure.

[0125] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may include an implementation of a network environment. Program modules 42 generally perform the functions and / or methods described in the embodiments of the present disclosure.

[0126] The processor 16 executes at least one of the multiple programs stored in the system memory 28 to perform various functional applications and data processing, such as implementing the method embodiments provided in the embodiments of the present disclosure.

[0127] The present disclosure also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method embodiment when the computer program is executed by a processor.

[0128] Any combination of one or more computer-readable media can be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductors, or any combination thereof. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by an instruction execution system, device or device or used in combination with it.

[0129] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0130] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0131] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or wide area network (WAN) domain, or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0132] The present disclosure also provides a computer program product. When the computer program product is run on a computer, the computer is enabled to execute the steps of the above method embodiments.

[0133] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the apparatus and method according to multiple embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and the part of the module, program segment or instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0134] Unless the context clearly indicates otherwise, as used herein and in the appended claims, the singular includes the plural, and vice versa. Thus, when referring to the singular, the plural of the corresponding term is generally included. Similarly, the words "include" and "comprising" are to be interpreted as inclusive rather than exclusive. Likewise, the terms "include" and "or" should be interpreted as inclusive unless such interpretation is expressly prohibited herein. Where the term "example" is used herein, particularly when it follows a group of terms, the "example" is merely exemplary and illustrative and should not be considered exclusive or comprehensive.

[0135] Further aspects and scope of adaptability become apparent from the description provided herein. It should be understood that various aspects of the present disclosure can be implemented individually or in combination with one or more other aspects. It should also be understood that the description and specific embodiments herein are intended to be illustrative only and are not intended to limit the scope of the present disclosure.

[0136] Several embodiments of the present disclosure have been described in detail above, but it is obvious that those skilled in the art can make various modifications and variations to the embodiments of the present disclosure without departing from the spirit and scope of the present disclosure. The scope of protection of the present disclosure is defined by the appended claims.

Claims

1. A reentry trajectory prediction method, characterized in that: include: Based on the dynamic equations of the gliding phase and different energy management strategy constraints, multiple trajectory models are established; Simulating each of the trajectory models to obtain a corresponding simulation trajectory data sequence, wherein each simulation trajectory data in the simulation trajectory data sequence includes longitude, latitude and altitude; Preprocessing each of the simulation trajectory data sequences to obtain a plurality of training samples; Input each training sample into the trajectory prediction model, perform trajectory classification training on the trajectory prediction model, and perform trajectory prediction training for different trajectory types; The current trajectory data of the re-entering body is input into the trained trajectory prediction model to obtain the predicted trajectory data of the re-entering body.

2. The method according to claim 1, characterized in that The trajectory prediction model includes a classifier and a prediction network. The prediction network is a Seq2Seq network and includes an encoder, a decoder, and a deformable cross-attention mechanism layer. The output of the classifier is connected to the input of the encoder, the input of the deformable cross-attention mechanism layer is connected to the residual and the output of the encoder, and the output of the deformable cross-attention mechanism layer is connected to the input of the decoder.

3. The method according to claim 2, characterized in that Inputting each training sample into the trajectory prediction model, performing trajectory classification training on the trajectory prediction model, and performing trajectory prediction training for different trajectory types include: Inputting each of the training samples into the classifier to obtain the trajectory type corresponding to each of the training samples to train the classifier, wherein the trajectory type corresponds to the trajectory model in a one-to-one manner; Inputting each of the training samples and the trajectory type output by the corresponding classifier into the prediction network, adjusting the attention range based on the trajectory type, and training the prediction network according to the residual and the training samples.

4. The method according to claim 1, wherein The preprocessing of each simulation trajectory data sequence to obtain a plurality of training samples includes: Randomly discarding the simulation trajectory data in each simulation trajectory data sequence; Based on piecewise cubic spline interpolation, fitting the simulation trajectory data in each of the discarded simulation trajectory data sequences to obtain a trajectory data fitting function; An observation coordinate system is established with the fitting trajectory data at the end time of the blackout area as the origin, and a plurality of trajectory data fitting functions are converted from the ground coordinate system to the observation coordinate system to obtain a plurality of observation trajectory data sequences; A plurality of training samples are determined according to each of the observation trajectory data sequences.

5. The method according to claim 4, characterized in that The determining of the plurality of training samples according to each of the observation trajectory data sequences comprises: Based on the first time window, sliding cutting is performed on each of the observation trajectory data sequences to obtain multiple first observation trajectory data cutting sequences to obtain multiple input samples; Based on the second time window, sliding cutting is performed on each of the observation trajectory data sequences to obtain multiple second observation trajectory data cutting sequences to obtain multiple output samples; The observation trajectory data in each of the input samples and each of the output samples are normalized to obtain the training sample.

6. The method according to claim 1, characterized in that The energy management strategy constraints include attack angle constraints and / or roll angle constraints; the multiple trajectory models include a balanced gliding trajectory model, a jumping gliding trajectory model, a lateral C-shaped trajectory model and a lateral S-shaped trajectory model.

7. The method according to any one of claims 1 to 6, characterized in that The method further comprises: perturbing the predicted trajectory data to generate a plurality of random trajectories; The landing point of the reentry body is predicted based on the landing point distribution of all the random trajectories.

8. A reentry trajectory prediction device, characterized in that: include: A model building module is used to build multiple trajectory models based on the dynamic equations of the gliding segment and different energy management strategy constraints; A simulation module, configured to simulate each of the trajectory models to obtain a corresponding simulation trajectory data sequence, wherein each simulation trajectory data in the simulation trajectory data sequence includes longitude, latitude, and altitude; A preprocessing module, used for preprocessing each of the simulation trajectory data sequences to obtain a plurality of training samples; The training module is used to input each training sample into the trajectory prediction model, perform trajectory classification training on the trajectory prediction model, and perform trajectory prediction training for different trajectory types; The prediction module is used to input the current trajectory data of the re-entering body into the trained trajectory prediction model to obtain the predicted trajectory data of the re-entering body.

9. An electronic device, characterized in that: include: A processor, wherein the processor is configured to execute a computer program stored in a memory, wherein the computer program, when executed by the processor, implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images