Portable satellite communication station signal dynamic alignment method and device

Abstract

The invention discloses a portable satellite communication station signal dynamic alignment method and a portable satellite communication station signal dynamic alignment device, relates to the field of satellite-ground communication, and is used for optimizing the real-time performance and the stability of a communication link. The method comprises the following steps: constructing a multi-body collaborative orbit model for describing a three-dimensional dynamic coupling relationship of a portable satellite communication station, an unmanned aerial vehicle and a satellite; analyzing key influence factors and influence degrees of communication link performance through causal reasoning on the basis of the multi-subject collaborative orbit model and environmental observation data; performing dynamic channel state prediction on the key influence factors based on a multi-subject collaborative orbit model and a causal reasoning result; and performing antenna alignment adjustment and error compensation based on the dynamic channel state prediction result. According to the invention, the problem of insufficient adaptability in a complex dynamic environment in the prior art is solved, and the accuracy, real-time performance and adaptability of antenna alignment are improved.

Description

Technical Field

[0001] The invention relates to the field of satellite-to-earth communication, and in particular to a portable satellite communication station signal dynamic alignment method and device. Background Art

[0002] In complex dynamic environments, portable satellite communication stations have become key communication hubs in UAV scenarios and are widely used in disaster relief, military operations, remote monitoring and other tasks. However, with the rapid development of UAV technology, its high dynamic characteristics and multi-degree-of-freedom motion pose new challenges to satellite communications, especially in harsh environmental conditions, requiring an optimized communication strategy that can cope with dynamic alignment and environmental changes.

[0003] First of all, the high-speed flight and multi-degree-of-freedom attitude changes (such as pitch angle, roll angle and yaw angle) of drones place stringent requirements on the dynamic alignment of the antenna of portable satellite communication stations. Traditional static alignment or regular adjustment methods can no longer meet their high dynamic and real-time requirements. In the collaborative communication between low-orbit satellites and drones, signal offsets caused by alignment errors often lead to communication interruptions, significantly affecting link stability. In addition, due to the rapid changes in the drone's moving trajectory, the Doppler frequency shift of the signal further increases the complexity of channel adjustment.

[0004] Secondly, rainfall (or rain attenuation) is the main environmental factor affecting channel quality, and its impact is particularly significant in drone scenarios, especially in the higher frequency Ka band. Rainfall can cause significant attenuation of the signal on the propagation path, which is manifested as increased path loss, reduced signal-to-noise ratio, and even communication interruption. This phenomenon not only affects the communication quality, but also may lead to waste of link transmission power. Traditional technologies usually perform static compensation for channel status based on a single rain attenuation parameter, and fail to consider the complex coupling relationship between rain attenuation and drone trajectory changes. In a complex dynamic environment, rain attenuation and drone motion work together, making channel quality prediction, compensation, and optimization a multivariable coupling problem.

[0005] In addition, under the dual influence of rain attenuation and drone trajectory changes, the resource allocation of portable satellite communication stations faces serious conflicts. Traditional communication power and frequency allocation usually rely on preset rules or fixed thresholds, and lack the ability to adjust in real time for complex dynamic environments. The allocation of communication power and the dynamic adjustment of frequency need to be completed in real time under the trade-off of multiple factors. Traditional algorithms are insufficient in efficiency and adaptability when dealing with complex dynamic changes.

[0006] In summary, how to coordinate rain attenuation compensation, UAV trajectory prediction and channel resource allocation has become a key technical problem to improve the robustness and stability of communication links. Summary of the invention

[0007] The purpose of the present invention is to provide a method and device for dynamic alignment of portable satellite communication station signals to address all or part of the above-mentioned problems, so as to effectively combine the high dynamic motion characteristics of the UAV with the dynamic alignment of the portable satellite communication station in a complex environment, and optimize the real-time and stability of the communication link.

[0008] The technical solution adopted by the present invention is as follows:

[0009] A portable satellite communication station signal dynamic alignment method, comprising:

[0010] S1. Construct a multi-agent collaborative orbit model that describes the three-dimensional dynamic coupling relationship between portable satellite communication stations, drones, and satellites;

[0011] S2. Analyzing the key influencing factors and their influence degree of the communication link performance through causal reasoning based on the multi-agent collaborative trajectory model and environmental observation data;

[0012] S3. Based on the multi-agent collaborative trajectory model and causal reasoning results, dynamic channel state prediction is performed on the key influencing factors;

[0013] S4. Perform antenna alignment adjustment and error compensation based on the dynamic channel state prediction result.

[0014] In addition, the present application also provides a portable satellite communication station signal dynamic alignment device, including a processor and a storage medium, the storage medium stores a computer program, and when the processor runs the computer program, it executes the above-mentioned portable satellite communication station signal dynamic alignment method.

[0015] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0016] This application solves the problem of insufficient adaptability of traditional technologies in complex dynamic environments by constructing a multi-agent collaborative orbit model, analyzing link influencing factors through causal reasoning, predicting dynamic channel states, and implementing antenna adjustment and error compensation strategies based on prediction results. Compared with traditional technologies, this application improves the accuracy, real-time and adaptability of antenna alignment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will now be described by way of example with reference to the accompanying drawings, in which:

[0018] Figure 1 It is a flow chart of a method for dynamic alignment of signals of a portable satellite communication station provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0020] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other equivalent or alternative features having similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0021] In response to the problem that traditional portable satellite communication station signal alignment methods lack adaptability and real-time performance in high dynamic environments, the embodiments of the present application propose a portable satellite communication station signal dynamic alignment method and device, which effectively combines the high dynamic motion characteristics of the drone with the dynamic alignment of the portable satellite communication station in a rain attenuation environment, and optimizes the real-time and stability of the communication link.

[0022] The portable satellite communication station signal dynamic alignment method provided in the embodiment of the present application comprises the following steps:

[0023] S1. Construct a multi-agent collaborative orbit model that describes the three-dimensional dynamic coupling relationship among portable satellite communication stations, drones, and satellites.

[0024] As an optional implementation, the above step S1 includes the following sub-steps S11-S14, which are used to combine the high-dynamic motion characteristics of the UAV, the satellite's on-orbit operation rules and the ground station position to establish a trajectory model that can characterize multi-agent collaboration (communication).

[0025] S11. Obtain trajectory data of the portable satellite communication station, the UAV and the satellite respectively and unify the coordinate references.

[0026] In some feasible implementations, the sub-step S11 includes:

[0027] S11.1: Obtain the satellite's on-orbit operating parameters (i.e., trajectory data), which include the satellite's orbital radius R. s 、Satellite angular velocity ω s and the initial phase φ s , and record the geographical location information of the ground portable satellite communication station) g ,Y g ,Z g ), which is the trajectory data of the portable satellite communication station.

[0028] S11.2: Collect the trajectory data of the UAV flight, including the initial position of the UAV (X u (0),Y u (0),Z u (0)), speed vu (t) and heading angle α u (t), as well as the pitch angle, roll angle and yaw angle corresponding to the change of the drone’s attitude.

[0029] S11.3: The spatial positions of the above-mentioned multiple entities are described respectively by a unified earth-centered earth-fixed coordinate system, and the local coordinates of the UAV are aligned with the global coordinates of the ground portable satellite communication station and the satellite using a coordinate transformation matrix, so as to obtain the three-dimensional position vectors of the satellite, the UAV and the ground portable satellite communication station in the same coordinate reference, which are recorded as:

[0030] p s (t) = [X s (t),Y s (t),Z s (t)] T ,p u (t) = [X u (t),Y u (t),Z u (t)] T ,p g =[X g ,Y g ,Z g ] T

[0031] Among them, p s (t) represents the trajectory position of the satellite over time, p u (t) represents the flight position of the UAV over time, p g Indicates the fixed position of a ground-based portable satellite communications station.

[0032] S12. Establish a satellite orbit model describing the satellite's in-orbit position based on the satellite's trajectory data.

[0033] In some feasible implementations, the sub-step S12 includes:

[0034] S12.1: Based on the acquired satellite in-orbit parameters, the circular orbit model is used to parameterize the satellite position:

[0035]

[0036] Among them, R s is the orbit radius, ω s is the satellite angular velocity, φ s is the initial phase, h s is an approximately constant height.

[0037] S12.2: The satellite position vector p obtained in step S12.1 is s(t) Docking with the coordinate reference obtained in S11.3, the satellite in-orbit motion data consistent with the Earth-centered Earth-fixed coordinate system is obtained, providing basic input for the subsequent multi-agent collaborative orbit model.

[0038] S13. A motion model describing the flight trajectory of the UAV is established based on the trajectory data of the UAV, and the UAV attitude change information is coupled to obtain a UAV trajectory model.

[0039] In some feasible implementations, the sub-step S13 includes:

[0040] S13.1: Based on the speed and heading angle information of the UAV, the flight trajectory of the UAV is described using a discretized time-varying equation, which can be specifically expressed as:

[0041]

[0042] Where Δt is the discrete sampling time step, v u (t) is the speed of the UAV in the horizontal plane, α u (t) is the real-time heading angle of horizontal motion, Δh u (t) is the height increment of the UAV in the vertical direction.

[0043] S13.2: Based on step S13.1, in order to accurately describe the attitude change of the UAV, the pitch angle β u (t), roll angle γ u (t) and yaw angle δ u (t) Synchronously record and map these posture parameters to the three-dimensional coordinate transformation matrix R u (t), update the precise attitude vector A of the drone in the coordinate system u (t). This attitude vector plays a role in correcting the line of sight vector in the subsequent channel evaluation and antenna alignment process.

[0044] S14. Couple the satellite orbit model, the UAV orbit model and the portable satellite communication station trajectory data to obtain a multi-agent collaborative orbit model.

[0045] In some feasible implementations, the sub-step S14 includes:

[0046] S14.1: Based on the orbital model obtained in substeps S12 and S13, calculate the satellite's orbital position p s (t), UAV position p u (t) and its posture A u (t), location of the ground portable satellite communication station p g Overall coupling is performed to establish a multi-agent collaborative orbit model for channel assessment and antenna alignment.

[0047] S14.2: The multi-agent collaborative trajectory model obtained in step S14.1 is stored and called in subsequent steps to form a complete space-time coordinate sequence {p s (t),p u (t),A u (t),p g This time-space coordinate sequence can provide a high-precision input for subsequent rain attenuation compensation, Doppler frequency shift estimation, and antenna error correction processes.

[0048] S14.3: In the multi-agent collaborative trajectory model, in order to further evaluate the dynamic channel, the Doppler frequency shift f is defined in this embodiment. d (t) is used to characterize the deviation of the signal transmission frequency caused by the relative motion between the satellite and the drone:

[0049]

[0050] Among them, f 0 is the carrier center frequency, v s (t) and v u (t) represent the instantaneous velocity vectors of the satellite and the UAV, is the unit vector pointing from the transmitting end to the receiving end, and c is the propagation speed of the electromagnetic wave in free space. The Doppler frequency shift value obtained in this step can be used to correct the signal receiving frequency and alignment error of the portable satellite communication station in subsequent steps.

[0051] S2. Based on the multi-agent collaborative trajectory model and environmental observation data, the key influencing factors and their influence degree of communication link performance are analyzed through causal reasoning.

[0052] As an optional implementation, step S2 includes sub-steps S21-S24, which use the output of the multi-agent collaborative orbit model constructed in step S1 as input to perform causal modeling, identify the interaction mechanism between environmental and link influencing factors, and provide data support and logical basis for subsequent dynamic channel prediction and antenna error compensation.

[0053] S21. Obtain a multi-agent collaborative orbit model and load environmental observation data.

[0054] In some feasible implementations, the sub-step S21 includes:

[0055] S21.1: Call the multi-agent collaborative trajectory model {p s (t),p u (t),A u (t),p g}, where p s (t) represents the satellite's position in orbit, pu (t) represents the position of the UAV, A u (t) represents the drone attitude vector, p g Indicates the location of a ground-based portable satellite communication station.

[0056] S21.2: Based on S21.1, load the observation data set related to the external environment, including rainfall intensity r(t) and its temporal and spatial distribution, meteorological radar data, and geographic information system (GIS) data, and align these environmental observation data with the orbital model in the time dimension to obtain the environmental observation data sequence {E(t)}, where E(t) can be expressed as E(t) = [r(t), θ rain (t),…],θ rain (t) can be regarded as a rainfall direction or regional marker and used for spatial positioning in subsequent rain attenuation calculations.

[0057] S22. Based on the multi-agent collaborative trajectory model and environmental observation data, a causal relationship diagram is constructed to describe the causal path between environmental influencing factors and communication link performance, and key influencing factors and causal paths are identified from the causal relationship diagram.

[0058] In some feasible implementations, the sub-step S22 includes:

[0059] S22.1: Based on the multi-agent collaborative orbit model and environmental observation data obtained in substep S21, the causal path between different variables is clarified by constructing a causal relationship diagram. Potential factors that may cause communication interruption or link quality degradation are divided as dependent variables or independent variables. These potential factors include but are not limited to rainfall intensity r(t), Doppler frequency shift f d (t), antenna pointing deviation δ ant (t) and the UAV attitude change parameter A u (t).

[0060] S22.2: Define the meaning of the information flow of each node and directed edge in the causal relationship graph, as follows:

[0061] Node r(t): represents the rainfall intensity at the current moment;

[0062] Node f d (t): represents the Doppler frequency shift derived from the multi-agent collaborative orbit model constructed in step S1;

[0063] Node δ ant (t): represents the instantaneous pointing error of the antenna;

[0064] Node A u (t): indicates the change of the drone’s attitude;

[0065] Node Qlink (t): represents the link performance indicator (such as signal-to-noise ratio or signal receiving power).

[0066] In the embodiment of the present application, if there is r(t)→Q in the causal relationship graph link (t), it means that the rainfall intensity has a direct causal effect on the link performance index; if there is A u (t)→δ ant (t)→Q link (t), it means that the change of UAV attitude indirectly affects the link performance index through the antenna pointing deviation, and the other nodes in the causal relationship graph are similar. Based on this, the key influencing factors can be identified through the causal relationship graph.

[0067] S22.3: Screen the main causal paths that affect the quality of satellite communications in the application scenario of the embodiment of the present application, for example, according to empirical rules. Record the screened causal paths in the form of a directed acyclic graph (DAG) to lay a structural foundation for subsequent parameter learning and interference source location.

[0068] S23. Based on the cause-effect relationship diagram, the influence relationship of key influencing factors on the performance of the communication link is parameterized and characterized.

[0069] In some feasible implementations, the sub-step S23 includes:

[0070] S23.1: The effect of rainfall intensity r(t) on link channel loss is selected as the typical causal path, and the modified rain attenuation loss model is used to parameterize the effect. The additional path loss caused by rainfall is L rain (t), definition:

[0071] L rain (t) = α·r(t) β ·d,

[0072] Wherein, α and β are attenuation coefficients calibrated by experience or test data, and d is the path length of the signal propagating in the rainfall area. This formula is used in the embodiment of the present application to measure the degree of channel attenuation caused by rainfall intensity.

[0073] S23.2: Based on the relative velocity and position vectors obtained from the multi-agent collaborative orbit model, the representation of the Doppler shift is modified. Let the Doppler shift f d (t) Link performance index Q link The influence of (t) is described by a logarithmic attenuation relationship, which is expressed as:

[0074] Q link (t) = Q 0 -10log(1+k·|f d (t)|),

[0075] Among them, Q 0 is the benchmark link quality indicator, k is the ratio constant used to quantify the Doppler effect, |f d (t)| is the absolute value of the Doppler frequency shift.

[0076] The above relationship reflects the process in which the communication carrier offset increases due to the Doppler shift, thereby reducing the system demodulation performance.

[0077] S23.3: For antenna pointing deviation δ ant (t), in the embodiment of the present application, according to the cumulative effect of attitude change and alignment error, it is assumed that the influence of antenna deviation on link performance has a multiplicative amplification characteristic, which can be expressed as:

[0078] Q link (t)←Q link (t)exp(-λδ ant (t)),

[0079] Among them, λ is a parameter that describes the sensitivity of antenna alignment error.

[0080] The above formula is used together with the rain attenuation model of S23.1 and the Doppler model of S23.2 to construct the main causal equation system of the key influencing factors in the causal relationship diagram, so as to parameterize the influence of the key influencing factors on the performance of the communication link.

[0081] S24. Based on the causal relationship diagram and the parameterized characterization results of the key influencing factors, the influence of the key influencing factors in the causal path is evaluated to obtain the causal reasoning results.

[0082] As an optional implementation, the impact degree evaluated by the sub-step S24 includes the causal path contribution value of the key influencing factor on the communication link performance, and the importance of the key influencing factor on the communication link performance. In some feasible implementations, the sub-step S24 includes:

[0083] S24.1: Based on the causal relationship diagram constructed in sub-step S22 and the parameterized characterization results in sub-step S23, the rainfall intensity r(t), Doppler frequency shift f d (t), antenna pointing deviation δ ant (t) and the drone attitude A u (t) and other key nodes to quantitatively evaluate the causal effect, and obtain the description of these key influencing factors on the link performance index Q link (t) The contribution value of the causal path of direct or indirect impact.

[0084] S24.2: Sort the relative importance of each key influencing factor in the causal path to identify the most significant link interference sources under different flight periods or different rainfall conditions. If a factor (such as sudden rainfall) is found to cause significant degradation of the communication link in a short period of time, it will be marked as a high-priority warning factor. The subsequent steps will be to perform dynamic channel prediction and antenna compensation strategy design based on the priority marking results of each key influencing factor.

[0085] S24.3: Output the causal reasoning results of S24.1 and S24.2, and connect them with the dynamic channel prediction and resource scheduling module in the subsequent steps to achieve forward-looking identification and comprehensive consideration of environmental and link changes.

[0086] S3. Based on the multi-agent collaborative trajectory model and causal reasoning results, dynamic channel state prediction is performed on key influencing factors.

[0087] This step S3 uses the causal reasoning result output from step S2 and the multi-agent collaborative orbit model output from step S1 as the main input, and performs dynamic channel state prediction for complex rain attenuation environment, Doppler frequency shift and antenna pointing deviation, aiming to provide a priori basis with high timeliness and high precision for subsequent antenna alignment compensation, power control and resource scheduling. As an optional implementation, this step S3 includes the following sub-steps:

[0088] S31. Obtain key influencing factors from causal reasoning results and perform time synchronization.

[0089] In some feasible implementations, the sub-step S31 includes:

[0090] S31.1: Data integration and time synchronization

[0091] The key influencing factors (i.e., key variable set) are obtained from the causal reasoning results output from step S2, including rainfall intensity r(t), Doppler frequency shift f d (t), antenna pointing deviation δ ant (t) and the drone attitude A u (t). The coordinated orbit data of the portable satellite communication station, the UAV and the satellite are obtained from the multi-agent coordinated orbit model constructed in step S1, and are denoted as p s (t), p u (t), p g To ensure that multi-source data are aligned in the same time coordinate system, a unified multivariate time series dataset is constructed based on the minimum time difference criterion or interpolation method:

[0092]

[0093] Where Q link(t) is the link performance indicator at the current moment (such as signal-to-noise ratio or received power).

[0094] S31.2: Outlier detection and cleaning.

[0095] In a multivariate time series dataset In the process, missing values ​​caused by possible measurement anomalies or communication packet loss are detected. If missing values ​​occur, they are repaired by piecewise linear interpolation. If extreme outliers are detected, they are removed based on three times the standard deviation to form a relatively complete and consistent data set.

[0096] S31.3: Data initialization and segmentation.

[0097] The dataset The training set is divided into a training set and a validation set in the time dimension. The training set is used for model parameter learning, and the validation set is used for subsequent prediction effect evaluation and hyperparameter tuning. The training set is denoted as The validation set is During initialization, the periods of most severe rain attenuation and high-speed maneuvering of the UAV are marked based on prior knowledge to provide a reference for model fitting in subsequent steps.

[0098] S32. Construct a multi-dimensional state space model that describes the impact state of the communication link, and collaboratively constrain the states that have mutual impact.

[0099] In some feasible implementations, the sub-step S32 includes:

[0100] S32.1: Multidimensional state vector definition.

[0101] In the embodiment of the present application, dynamic channel prediction is regarded as a joint inference process of estimating the internal state variables of the system and the changes of external environmental factors. The state vector S(t) is defined to include the internal state of the communication link (such as channel attenuation coefficient, frequency deviation state) and the environmental impact state (such as rain attenuation level, Doppler component), which is specifically expressed as:

[0102] S(t)=[α ch (t),Δf(t),δ ant (t),r(t),…] T

[0103] where α ch (t) can be regarded as the instantaneous channel attenuation factor, and Δf(t) represents the dynamic quantization of the Doppler frequency shift.

[0104] S32.2: Definition of state transfer equation and observation equation.

[0105] In the multidimensional state space model, the state transition equation is defined as:

[0106] S(t+1)=F(S(t),U(t))+w(t)

[0107] Where F is the state transfer function, U(t) is the system input or interference (such as drone acceleration, rain attenuation distribution mutation), w(t) is the process noise, and the observation equation is defined as:

[0108] Y(t)=H(S(t))+v(t)

[0109] Where Y(t) represents the observable link performance (Q link (t)) is the function mapping of the orbit data, and v(t) is the measurement noise.

[0110] Based on the above equation, the evolution of S(t) is ensured to be physically interpretable, and the coupling relationship between key variables is embedded in F and the observation function H in combination with the causal graph information of step S2.

[0111] S32.3: Coordination constraints on causal priors.

[0112] Based on the causal relationship diagram generated in step S2, based on the Doppler frequency shift f d (t) for α ch (t) has a significant nonlinear effect, and this effect is explicitly added to the definition of F(S(t), U(t)), that is, ch The update of (t) introduces the sigmoid function:

[0113] α ch (t+1)=α ch (t)sigmoid(κf d (t)),

[0114] Where κ is the scaling factor, which is used to control the impact of Doppler frequency shift on the channel attenuation factor.

[0115] S33. Construct a timing prediction network that takes the state vector and observation vector of the multidimensional state space model at the current moment as input and takes the state vector and communication link performance at the next moment as output.

[0116] In some optional implementations, the sub-step S33 includes:

[0117] S33.1: High-order timing model structure design.

[0118] In the embodiment of the present application, in order to capture complex time series correlation, the multidimensional state vector S(t) and its observation vector Y(t) are used as input based on LSTM to learn short-term and long-term dependencies. The core mapping of the network is Φ(S(t)), which is used to predict the next moment S(t+1) and the link performance Q link(t+τ).

[0119] S33.2: Multi-step prediction and output mapping.

[0120] In order to obtain multi-step prediction results (t+1, t+2, ..., t+τ), the embodiment of the present application connects the decoupled output head to the last layer of LSTM and establishes mappings for the outputs at different prediction times:

[0121]

[0122] where h LSTM (t) is the hidden state vector of the LSTM network at time t, W o and b o are the learnable parameter matrices and biases.

[0123] Based on this strategy, multi-step prediction results can be obtained simultaneously in one forward propagation, improving real-time performance.

[0124] S33.3: Attention mechanism combining trajectory characteristics and posture coupling.

[0125] This application also takes into account the change of the drone's attitude A u (t) and the relative position of the satellite p s (t) is related to the link index, and an attention mechanism is added to the LSTM hidden layer. Let z t is the hidden layer output, and the context vector c is introduced at this moment t To measure the importance of drone attitude and satellite position to the current prediction:

[0126]

[0127] in,

[0128] e t,τ′ =ATTN(z τ′ ,A u (t),p s (t))

[0129] ATTN(·) is the attention scoring function, which comprehensively considers the attitude and satellite position on z τ′ weighted.

[0130] Based on this mechanism, important moments in highly dynamic scenes are represented, thereby improving the sensitivity and accuracy of the prediction.

[0131] S34. Construct the loss function of the time series prediction network.

[0132] In some feasible implementations, the sub-step S34 includes:

[0133] S34.1: Construct a multi-objective loss function.

[0134] In order to take into account different prediction objectives ( and ), the embodiment of the present application defines multi-objective loss during network training:

[0135]

[0136] Where S true (t) is the true state vector that may be measurable at some time or estimated a priori, α 1 With α 2 is the weight coefficient, which is used to balance the link indicator prediction error and the internal state estimation error.

[0137] S34.2: Regularization terms and causal constraints are integrated.

[0138] exist A regularization constraint on the model complexity is further added in Ω(Θ) to prevent the model from overfitting. In addition, based on the causal analysis results of step S2, it is shown that a certain key influencing factor has monotonicity or sparsity requirements on link performance. In the embodiment of the present application, a non-negative constraint is added, that is, the attenuation influence coefficient α of the rainfall intensity r(t) rain Apply α rain ≥0, thus ensuring the rationality of the model in a physical sense.

[0139] S35. Use the data obtained in sub-step S31 to train a time series prediction network.

[0140] In some optional implementations, the sub-step S35 includes:

[0141] S35.1: Initialization and batch training.

[0142] Initialize the network learnable parameters to small random values. Divide into small batches (Batch), and use forward propagation to calculate the loss function And use the Adam optimization algorithm to update the parameters.

[0143] S35.2: Gradient calculation and update formula

[0144] The update of Adam’s general parameter Θ can be expressed as:

[0145]

[0146] Where m Θ With v Θ are the first-order and second-order momentum estimates, β 1 ,β 2is a hyperparameter, η is the learning rate, and ∈ is a numerical stability term. Based on this update process, it is performed after each batch of training until the loss function converges or reaches the specified number of rounds.

[0147] S35.3: Dynamic learning rate and early stopping strategy.

[0148] During the training process, if it is found that the loss function no longer decreases significantly after several rounds, the learning rate η is dynamically adjusted, and based on the early stopping strategy, to prevent overfitting and excessive consumption of training time.

[0149] S36. Utilize the trained time series prediction network to predict the communication link performance and the multidimensional state space model using the current multidimensional state space model in the target scenario.

[0150] In some feasible implementations, the sub-step S36 performs short-term prediction and medium- and long-term trend prediction on the communication link performance, and perturbs the input data to improve the anti-interference ability of the prediction. Specifically, the sub-step S36 includes:

[0151] S36.1: Short-term prediction and adaptive output.

[0152] In actual satellite communications, in order to timely guide antenna alignment and power allocation, the embodiment of the present application proposes that the link performance index Q from t+1 to t+τ can be calculated in an online mode. link Make short-term predictions and output the predicted values And its credibility measurement. If the drone posture A is detected u (t) will change significantly, triggering the internal network (AttentionLayer is the previous attention layer) to further strengthen the weight of posture-related components.

[0153] S36.2: Medium- to long-term trend forecast.

[0154] In addition, since rainfall has a relatively slow spatial advancement characteristic, the embodiment of the present application also proposes an environmental prediction for a longer time scale (t+τ′, τ′>>τ). Under the existing model framework, an exponential regression form is established for the rainfall intensity r(t) and other slowly varying parameters, thereby providing a medium- and long-term channel quality trend estimation at the minute to hour level for coordinated scheduling and backup resource planning.

[0155] S36.3: Uncertainty analysis and diversity prediction.

[0156] In order to more accurately respond to sudden environmental changes, the present embodiment proposes to further use Monte Carlo simulation to perturb w(t) and v(t) to generate predicted distributions under different scenarios. For extreme conditions where significant rain attenuation or high-speed flight may occur, the probability of their impact at the tail of the distribution is evaluated so that the system can make early intervention decisions during the early warning stage.

[0157] In addition, in an optional implementation, step S3 further includes sub-step S37: outputting the dynamic channel prediction result of sub-step S36.

[0158] In some feasible implementations, the sub-step S37 includes:

[0159] S37.1: Generate a comprehensive forecast report.

[0160] Based on the short-term and medium-term prediction results obtained in sub-step S36, combined with the causal weights of key influencing factors such as attitude, rain attenuation, and Doppler frequency deviation, the predicted values ​​and uncertainty ranges are output in a structured format to form a visual or numerical comprehensive prediction report R pred The report contains real-time Estimated value and link risk assessment value (such as outage probability).

[0161] S37.2: Feedback and collaborative applications.

[0162] The comprehensive forecast report R pred Together with the state vector estimate Submit to the satellite communication station control module or the UAV navigation system. If it is detected that the rain attenuation will increase significantly in the short term, the antenna power increase command will be issued in advance or the backup frequency band will be switched; if it is detected that the UAV's attitude is about to change dramatically, the antenna pointing adjustment subsystem will be triggered to calibrate in advance, thereby reducing the risk of link interruption.

[0163] S4. Perform antenna alignment adjustment and error compensation based on dynamic channel state prediction results.

[0164] After completing the dynamic channel state prediction of rain attenuation, Doppler frequency shift and attitude disturbance factors, this step performs multi-stage optimization and execution on the antenna alignment strategy and error compensation scheme of the portable satellite communication station based on the prediction results and state estimation output by step S3, ensuring effective avoidance of link interruption and continuous guarantee of communication quality under high-speed maneuvers and complex environments.

[0165] In an optional implementation, step S4 includes the following sub-steps:

[0166] S41. Obtain dynamic channel state prediction results and obtain key information related to antenna alignment errors.

[0167] In some feasible implementations, the sub-step S41 includes:

[0168] S41.1: Analyze prediction results and uncertainty indicators.

[0169] Call the comprehensive forecast report R output in step S3 pred , extracting short-term and mid- to long-term The predicted value of the communication link quality is combined with the corresponding confidence interval or uncertainty quantification indicator.

[0170] S41.2: Extract antenna error sensitivity parameters.

[0171] Analysis and Antenna Alignment Error δ ant (t) Relevant sensitivity information, including causal inference or prediction model’s response to δ ant The importance evaluation value of sens , used for priority determination of subsequent antenna adjustment schemes.

[0172] S41.3: Multi-channel information fusion.

[0173] If the system has other channel monitoring (such as satellite backup link or surrounding weather radar information), its observation values ​​are uniformly integrated into the antenna adjustment process of the embodiment of the present application to further improve the completeness of the adjustment decision.

[0174] S42: construct a mapping model between antenna attitude and line of sight vector, and characterize the antenna pointing error based on the mapping model.

[0175] In some feasible implementations, the sub-step S42 includes:

[0176] S42.1: Define the antenna attitude vector.

[0177] In the embodiment of the present application, the elevation angle θ of the antenna body is EL (t), azimuth angle θ AZ (t) and the polarization angle θ POL (t) is combined into the antenna attitude vector Θ ant (t).

[0178] S42.2: Coupling of the sight line vector and position coordinates.

[0179] To facilitate subsequent alignment adjustments, define the sight vector The direction pointing to the target, and its relationship with the antenna attitude vector can be expressed as

[0180]

[0181] Where R(·) is the rotation matrix constructed by the attitude angle, u z It is the reference direction in the antenna coordinate system (for example, vertically upward by default).

[0182] Based on this, can be compared with the target position p in step S1 s (t) or p u (t) is compared to measure the pointing error of the antenna.

[0183] S42.3: Angular measure of antenna pointing error.

[0184] The target direction vector (starting from the relative position of the antenna) is Based on the inner product operation, the pointing error δ can be defined as ant (t) is:

[0185]

[0186] like and Perfect alignment, then the inner product is 1,δ ant (t)=0.

[0187] S43: Generate an antenna alignment strategy based on the dynamic channel state prediction result.

[0188] In some feasible implementations, the sub-step S43 includes:

[0189] S43.1: Generate a piecewise alignment strategy.

[0190] Combined with the short-term prediction given in step S3 The embodiment of the present application dynamically adjusts the antenna attitude in each sampling period based on factors such as the rainfall intensity r(t+1). If it is predicted that rain attenuation or Doppler shift will increase, the antenna angle alignment accuracy is prioritized, and the upper limit of the tracking rate of the pitch and azimuth axes is increased if necessary.

[0191] S43.2: Objective function definition.

[0192] During the segmented alignment process, the antenna attitude target at a moment is recorded as It is obtained through the following optimization objectives:

[0193]

[0194] where δ ant (t+1) is calculated based on the definition of (S42.3), Q link (t+1) is the expected value of link quality predicted in step S3, w 1 With w 2 is a weight factor used to trade off between alignment error and link performance.

[0195] S43.3: Priority inheritance and multi-objective balance.

[0196] If the prediction indicates that the attitude change is extremely drastic (for example, the drone turns at high speed), the severity can be determined according to the corresponding threshold or rule, and w is appropriately increased according to the pre-made rule. 1 To give priority to reducing antenna deviation; if it is predicted that rain attenuation is about to occur, increase w according to the pre-made rules 2 This highlights the need to ensure link performance.

[0197] S44. Perform transmit power compensation in accordance with the antenna alignment strategy.

[0198] In some feasible implementations, the sub-step S44 includes:

[0199] S44.1: Define the error compensation gain factor

[0200] In the embodiment of the present application, in order to further offset the effective receiving power attenuation caused by the pointing error, the transmit power compensation factor Γ(t) is configured while aligning, that is, Γ(t) and δ ant (t) has a negative logarithmic correlation:

[0201] Γ(t)=exp(-η 0 δ ant (t))

[0202] where η 0 It is an adjustable parameter.

[0203] If δ ant (t) is larger, then Γ(t) tends to a smaller value, increasing the transmission power P tx (t) or increase the RF gain G rx (t) to compensate for the loss caused by alignment deviation.

[0204] S44.2: Coordinated power allocation.

[0205] Taking into account the combined effects of rain attenuation, antenna deviation, and Doppler shift on the link, the embodiment of the present application proposes the following collaborative equation for power allocation:

[0206] P out (t) = P base +κ r r(t) β +κ δ δ ant (t)+κ f |f d (t)|

[0207] Where P out (t) is the actual transmission power, P base is the reference power, (κ r ,β),κδ ,κ f are the compensation coefficients for rain attenuation, pointing deviation, and Doppler frequency deviation, respectively. Based on the matching relationship between the output of this formula and Γ(t), the joint optimization of alignment and power is achieved.

[0208] In addition, as an optional implementation, step S4 may further include the following sub-steps:

[0209] S45. Establish constraints on antenna maneuvers in the antenna alignment strategy.

[0210] The sub-step S45 imposes corresponding constraints on the antenna alignment strategy based on the physical conditions of the antenna mobile platform. In some feasible implementations, the sub-step S45 includes:

[0211] S45.1: Physical limitations of mobile platforms

[0212] On an actual antenna maneuvering platform, there is an upper limit on the maneuvering speed of the pitch axis and the azimuth axis. And acceleration limit When it is predicted that the UAV will perform high-speed maneuvers at time t+1, it should be determined whether the required antenna rotation exceeds the limit. If it exceeds the physical capabilities of the platform, adjustments should be made in advance, such as adjusting to align at the maximum maneuvering speed or acceleration limit.

[0213] S45.2: Speed ​​limit planning.

[0214] To ensure that the maneuverable platform does not exceed the physical limit, the pitch angle is adjusted by Δθ EL (t+1) and the azimuth adjustment Δθ AZ (t+1) Speed ​​limit planning:

[0215]

[0216] The speed limit planning of the azimuth angle is similar. Based on this planning, it can be ensured that the rotation amount within each sampling period Δt does not exceed the limit, and the acceleration constraint can be further considered to avoid mechanical overload or speed oscillation.

[0217] S46: Execute the antenna alignment strategy and monitor the execution result.

[0218] In some feasible implementations, the sub-step S46 includes:

[0219] S46.1: Execute antenna alignment strategy online (including compensation for transmit power).

[0220] In completing the Power compensation P out(t) and multiple calculations of the antenna maneuver constraints, a comprehensive command is issued to the antenna execution module of the portable satellite communication station to achieve the synchronous execution of operations such as antenna attitude adjustment and transmission power gain.

[0221] S46.2: Real-time error and link quality detection.

[0222] During the execution of the system, the antenna pointing error δ at the current moment ant (t) and link index Q link (t) Monitor and compare with the prediction result in step S3. If the actual error or link attenuation is significantly higher than the model prediction, a safety strategy or secondary adjustment process is immediately triggered to prevent further degradation of communication quality.

[0223] S47. Perform dynamic iteration and feedback optimization based on monitoring results.

[0224] In some feasible implementations, the sub-step S47 includes:

[0225] S47.1: Error accumulation correction.

[0226] If the accumulated antenna alignment deviation is too large for a period of time (which can be determined by a preset threshold) or the link quality prediction and actual measurement deviation increases (which can be determined by the deviation of more than one adjacent time), the iterative correction process is entered. ant (t) is the difference between the predicted value in step S3 and the corrected value λ sens , κ δ And other key parameters make the subsequent alignment and compensation strategies more suitable for the real-time environment.

[0227] S47.2: Long-term feedback and causal graph updating.

[0228] On a longer time scale (for example, after performing several large-angle adjustments), the error compensation records and alignment logs are fed back to the causal reasoning module (step S2) and the dynamic prediction model (step S3) to enhance their adaptability to special situations (heavy rain attenuation, extreme Doppler shift, etc.) and continuously improve the closed-loop self-learning capability of the alignment method of the present application.

[0229] S47.3: Output the execution result.

[0230] Finally, the antenna alignment and error compensation results of the final iteration are packaged into a detailed execution log or data packet for output, including the final antenna attitude Θ ant (t), transmission power P out (t) and the observed δ ant (t) and Q link (t).

[0231] An embodiment of the present application also provides a portable satellite communication station signal dynamic alignment device, which includes a processor and a storage medium, the storage medium stores a computer program, and when the processor runs the computer program, it executes the portable satellite communication station signal dynamic alignment method of the above embodiment.

[0232] Through the above design, first of all, this application constructs a three-dimensional collaborative motion trajectory of low-orbit satellites, drones and ground portable satellite communication stations by constructing a multi-agent collaborative orbit model, and uses the attitude transformation matrix and trajectory coupling method to establish the basic model required for dynamic channel optimization. It is superior to the static pointing model or simple regular adjustment strategy in traditional technology. Traditional methods cannot effectively deal with the impact of high-speed flight and multi-degree-of-freedom attitude changes of drones on antenna pointing. This application explicitly introduces a dynamic parameterized description of Doppler frequency shift and antenna error in trajectory modeling, which not only improves the dynamic adaptability of the model, but also provides higher-precision input for subsequent dynamic alignment.

[0233] Secondly, in terms of environmental factor modeling and causal reasoning, this application quantifies the causal relationship between key link factors such as rain attenuation, Doppler frequency shift, and antenna pointing deviation through the construction of a causal relationship graph, and uses parameterized formulas to describe the coupling relationship between these key influencing factors. Traditional technologies usually perform static compensation based on a single environmental parameter and fail to effectively capture the dynamic coupling between rain attenuation and drone movement. The causal reasoning of this application improves the ability to identify factors affecting link performance and the logical interpretability. Through this method, this application provides a clear causal path and data support for the design of dynamic channel prediction and compensation strategies.

[0234] In addition, in terms of dynamic channel state prediction, this application uses a long short-term memory network (LSTM) combined with an attention mechanism to learn complex time series correlations, and uses causal priors to collaboratively constrain the prediction model. This method solves the problems of insufficient adaptability and precision caused by the use of heuristic algorithms or static optimization models in traditional technologies. In particular, this application clarifies the nonlinear effects of Doppler frequency shift, rain attenuation intensity and link performance, and further enhances the ability to capture time series correlations through multi-step prediction and attention mechanisms, making the prediction results more accurate and reliable.

[0235] Finally, in terms of antenna alignment and error compensation, this application proposes a segmented alignment strategy and a power allocation coordination mechanism by combining dynamic channel prediction results. Unlike traditional technologies that rely on fixed thresholds or simple rule adjustments for antenna alignment, this application optimizes the attitude angle based on dynamic comparison of the line of sight vector and the target position, and uses the collaborative modeling of power compensation factors and link influencing factors to achieve joint optimization of alignment and power control. That is, this closed-loop adjustment strategy effectively reduces the risk of link interruption.

[0236] To sum up, traditional technologies cannot cope with high dynamic changes in complex environments due to static and regular limitations. This application combines multi-dimensional modeling with dynamic adaptability and optimizes the entire process to improve the system's responsiveness and stability to complex dynamic environments.

[0237] The present invention is not limited to the above-mentioned specific embodiments, but extends to any new features or any new combination disclosed in this specification, as well as any new method or process steps or any new combination disclosed.

Claims

1. A portable satellite communication station signal dynamic alignment method, characterized in that: include: S1. Construct a multi-agent collaborative orbit model that describes the three-dimensional dynamic coupling relationship between portable satellite communication stations, drones, and satellites; S2. Analyzing the key influencing factors and their influence degree of the communication link performance through causal reasoning based on the multi-agent collaborative trajectory model and environmental observation data; S3. Based on the multi-agent collaborative trajectory model and causal reasoning results, dynamic channel state prediction is performed on the key influencing factors; S4. Perform antenna alignment adjustment and error compensation based on the dynamic channel state prediction result.

2. The portable satellite communication station signal dynamic alignment method as claimed in claim 1, characterized in that: The step S1 comprises: S11, respectively obtaining trajectory data of the portable satellite communication station, the UAV and the satellite and unifying the coordinate references; S12, establishing a satellite orbit model describing the satellite's on-orbit position based on the satellite's trajectory data; S13, establishing a motion model describing the flight trajectory of the UAV based on the trajectory data of the UAV, and coupling the attitude change information of the UAV to obtain the trajectory model of the UAV; S14, coupling the satellite orbit model, the UAV orbit model and the portable satellite communication station trajectory data to obtain the multi-agent collaborative orbit model.

3. The portable satellite communication station signal dynamic alignment method as claimed in claim 1, characterized in that: The step S2 comprises: S21, obtaining the multi-agent collaborative orbit model and loading environmental observation data; S22. Based on the multi-agent collaborative trajectory model and the environmental observation data, a causal relationship diagram is constructed to describe the causal path between environmental influencing factors and communication link performance, and key influencing factors and causal paths are identified from the causal relationship diagram; S23, parameterizing the influence of the key influencing factors on the communication link performance based on the cause-effect relationship diagram; S24. Based on the causal relationship diagram and the parameterized characterization results of the key influencing factors, the influence degree of the key influencing factors in the causal path is evaluated to obtain a causal reasoning result.

4. The portable satellite communication station signal dynamic alignment method as claimed in claim 3, characterized in that: The step of evaluating the influence of the key influencing factors in the causal path includes: Evaluate the causal path contribution value of the key influencing factors on the communication link performance; and, Evaluate the importance of the key influencing factors on the performance of the communication link.

5. The portable satellite communication station signal dynamic alignment method as claimed in claim 1, characterized in that: The step S3 comprises: S31, obtaining the key influencing factors from the causal reasoning results and performing time synchronization; S32, constructing a multidimensional state space model that describes the influence state of the communication link, and collaboratively constraining the states that have mutual influence; S33, constructing a timing prediction network that takes the state vector and observation vector of the multidimensional state space model at the current moment as input and the state vector and communication link performance at the next moment as output; S34, constructing a loss function of the time series prediction network; S35, using the data obtained in sub-step S31 to train the time series prediction network; S36. Utilize the trained time series prediction network to predict the communication link performance and the multidimensional state space model using the current multidimensional state space model in the target scenario.

6. The portable satellite communication station signal dynamic alignment method as claimed in claim 5, characterized in that: The timing prediction network is a multi-step prediction model; in step S3.6, short-term prediction and medium- and long-term trend prediction of the communication link performance are performed respectively, and the input data is disturbed.

7. The portable satellite communication station signal dynamic alignment method as claimed in claim 1, characterized in that: The step S4 comprises: S41, obtaining a dynamic channel state prediction result and obtaining key information related to the antenna alignment error; S42, constructing a mapping model between antenna attitude and sight vector, and characterizing antenna pointing error based on the mapping model; S43, generating an antenna alignment strategy based on the dynamic channel state prediction result; S44: Perform transmit power compensation in accordance with the antenna alignment strategy.

8. The portable satellite communication station signal dynamic alignment method as claimed in claim 7, characterized in that: The step S4 further comprises: S45. Establish constraints on antenna maneuvers in the antenna alignment strategy.

9. The portable satellite communication station signal dynamic alignment method according to claim 7 or 8, characterized in that: The step S4 further comprises: S46, executing the antenna alignment strategy and monitoring the execution result; S47. Perform dynamic iteration and feedback optimization based on monitoring results.

10. A portable satellite communication station signal dynamic alignment device, comprising a processor and a storage medium, wherein the storage medium stores a computer program, characterized in that: When the processor runs the computer program, it executes the portable satellite communication station signal dynamic alignment method as described in any one of claims 1-9.

Images