Satellite-borne interference source Doppler observation strategy optimization method

Abstract

The invention discloses a spaceborne interference source Doppler observation strategy optimization method, which comprises the following steps: acquiring an original radio frequency signal and satellite orbit data, predicting mutation characteristics in Doppler data, and generating a self-adaptive observation scheme; according to the self-adaptive observation scheme, energy-aware resource scheduling is carried out, observation is executed, and enhanced Doppler observation data is obtained; and carrying out positioning calculation and performance evaluation based on the enhanced Doppler observation data, and generating parameters for iteratively optimizing an observation strategy. According to the invention, intelligent matching of the observation resource and the signal information content is realized, and the resource utilization efficiency, the key event capturing capability and the final positioning precision are improved.

Description

Technical Field

[0001] The present invention relates to satellite communication technology, in particular to a Doppler observation strategy optimization method for satellite-borne interference sources. Background Art

[0002] In the field of spaceborne wireless communications and non-cooperative signal positioning, low-Earth orbit (LEO) satellite platforms, due to their flexible orbital coverage, offer significant advantages in detecting and observing ground-based interference sources. Passive positioning of non-cooperative interference sources based on Doppler shift information is a key technical approach in wireless communication interference monitoring and positioning. In practical applications, the reception quality and frequency observation accuracy of interference source signals are influenced by a combination of factors. To improve spectrum efficiency and interference source observation accuracy, optimizing the observation strategy of the spaceborne platform has become the core and prerequisite for improving the performance of the entire positioning system.

[0003] In the broader field of wireless communication resource optimization, particularly in emerging communication networks assisted by intelligent reflecting surfaces (IRSs), researchers have proposed a variety of advanced joint optimization methods to cope with complex signal environments. For example, some work employs traditional optimization theories such as block coordinate descent (BCD) and Lagrangian dual transformation to iteratively solve base station beamforming and IRS phase shifts. Deep reinforcement learning (DRL) methods have been introduced to address higher-dimensional state and action spaces. For example, in a multi-user MISO system integrated with a drone IRS, one study utilized the deep deterministic policy gradient (DDPG) algorithm to achieve joint optimization of IRS trajectories and reflection coefficients. Other researchers designed a segmented DRL framework with a segmented action space to improve joint beamforming performance, avoiding DRL algorithms from falling into local optima. These methods demonstrate great potential for handling highly dynamic, multivariable resource allocation problems and provide a viable technical path for solving complex wireless communication system optimization problems.

[0004] However, despite the existence of advanced optimization frameworks in related fields, direct application of these methods to the specific scenario of Doppler observation of spaceborne interference sources still faces profound and unresolved technical bottlenecks. These bottlenecks primarily stem from the contradiction between the unique event-driven information structure of Doppler observation missions and the absolute scarcity of onboard resources. Specifically, existing technologies suffer from serious deficiencies in the forward-looking nature of observation strategies and the semantic depth of state perception, resulting in blind resource allocation and inefficient capture of critical information. Summary of the Invention

[0005] The purpose of the invention is to propose a Doppler observation strategy optimization method for satellite-borne interference sources to solve the problems of blind observation resource allocation and inaccurate prediction of mutation characteristics in the existing technology.

[0006] Technical solutions and optimization methods for Doppler observation strategies of satellite-borne interference sources include:

[0007] Obtain the original RF signal of the onboard interference source and the precise satellite orbit data, predict the mutation characteristics in the Doppler data, and generate an adaptive observation plan;

[0008] According to the adaptive observation scheme, energy-aware resource scheduling is performed and observations are performed to obtain enhanced Doppler observation data;

[0009] Positioning solution and performance evaluation are performed based on the enhanced Doppler observation data, and parameters for iterative optimization of the observation strategy are generated.

[0010] According to one aspect of the present application, obtaining enhanced Doppler observation data includes:

[0011] Identify energy accumulation windows defined by low elevation angle periods in adaptive observing schemes, and high-value observing periods defined by Doppler mutation characteristics;

[0012] In the energy accumulation window, by reducing the working mode of the receiver, the energy saved is accumulated and counted into the energy reserve unit;

[0013] When entering a high-value observation period, evaluate the matching degree between the balance of the energy storage unit and the power consumption demand of that period;

[0014] Under the condition that the balance of the energy reserve unit meets the power consumption requirement, high-density sampling is authorized to obtain enhanced Doppler observation data.

[0015] According to one aspect of the present application, evaluating the matching degree between the balance of the energy storage unit and the power consumption demand during the period includes:

[0016] Construct an energy-accuracy benefit function to characterize the relationship between observation accuracy and energy consumption;

[0017] Based on the energy-accuracy benefit function, a segmented benefit evaluation matrix is ​​generated for different power consumption modes during high-value observation periods.

[0018] Establish a multi-constraint optimization model with the goal of maximizing total benefits and integrating energy storage unit balance and mode switching cost as constraints;

[0019] The multi-constraint optimization model is solved to obtain the optimal power allocation scheme, and high-density sampling is performed accordingly.

[0020] According to one aspect of the present application, it further includes:

[0021] When the assessment determines that the balance of the energy storage unit cannot meet the power consumption demand, the degradation strategy is initiated;

[0022] The degradation strategy calculates the ratio of the current remaining energy to the expected demand of the subsequent task and compares it with at least one preset degradation threshold;

[0023] If the ratio is lower than the degradation threshold, a downgraded sampling plan is generated to reduce the total energy consumption. The downgraded sampling plan is achieved by increasing the sampling interval of non-critical observation areas or shrinking the spatial range of high-density sampling.

[0024] Observations are performed according to a downsampling plan.

[0025] According to one aspect of the present application, when the original RF signal is lower than a preset threshold, the process of obtaining enhanced Doppler observation data is implemented by using a step-by-step Doppler frequency cascade estimation method, specifically:

[0026] The frequency grid is adaptively adjusted based on the spectrum energy distribution to perform preliminary spectrum peak location and obtain a rough Doppler estimate.

[0027] Based on the coarse Doppler estimate, the intermediate Doppler estimate is obtained by injecting the reverse Doppler sequence and searching for the solution that minimizes the spectral flatness of the compensated signal.

[0028] Based on the Doppler mean estimation, a virtual mirror signal is constructed and iteratively adjusted to minimize the time rate of change of the beat frequency component generated by its cross-correlation with the original signal, and the Doppler precise estimation is obtained. Based on this, the enhanced Doppler observation data is sampled.

[0029] According to one aspect of the present application, a Doppler precision estimate is obtained by constructing a virtual mirror signal and iteratively adjusting the signal, including:

[0030] Using the conjugate of the original RF signal and taking the Doppler mean value as the frequency basis, a virtual mirror signal is constructed;

[0031] Calculate the cross-correlation function between the original RF signal and the virtual mirror signal, and separate the beat frequency component therefrom;

[0032] The frequency used to construct the virtual mirror signal is iteratively adjusted until the time rate of change of the beat frequency component converges to a preset minimum value, and the converged frequency is determined as the Doppler precise estimate.

[0033] According to one aspect of the present application, when acquiring enhanced Doppler observation data, when sampling is performed in a non-uniform manner, a data reconstruction process is also included, specifically:

[0034] Obtaining a precise Doppler estimate and using it to pre-compensate the non-uniformly sampled observation data to obtain a compensated stationary signal;

[0035] For the compensated stationary signal, solve the pre-built compressed sensing convex optimization model to obtain the reconstructed stationary signal component;

[0036] The frequency trend represented by the Doppler precise estimate is re-assigned to the reconstructed stationary signal component to generate enhanced Doppler observation data.

[0037] According to one aspect of the present application, predicting a mutation feature in Doppler data comprises:

[0038] Perform spectrum analysis on the original RF signal and, combined with the satellite's precise orbit data, extract the time-series Doppler shift data through a frequency tracking loop;

[0039] Perform time-frequency transformation on the time series Doppler frequency shift data to generate a time-frequency spectrum and process it into a grayscale image;

[0040] Extract local binary patterns and gray-level co-occurrence matrices from grayscale images as texture features and construct multi-dimensional texture feature vectors;

[0041] The multidimensional texture feature vector is fed into a pre-trained convolutional neural network to parse and output the mutation type and mutation time in the Doppler data.

[0042] According to one aspect of the present application, the prediction of the mutation characteristics in Doppler data may also be:

[0043] Extract the time-series Doppler frequency shift data from the original RF signal and calculate the corresponding Doppler second-order derivative sequence;

[0044] The historical transit mutation point knowledge base is called up and combined with the current satellite precise orbit data to generate the initial predicted mutation position;

[0045] Real-time monitoring of the Doppler second-order derivative sequence. When the measured characteristics of the sequence deviate from the expected characteristics of the initially predicted mutation position and exceed a preset threshold, a correction mechanism is triggered.

[0046] The correction mechanism fuses the initial prediction with the measured features and outputs a corrected mutation window.

[0047] According to one aspect of the present application, the correction mechanism includes:

[0048] A Kalman filter is used to fuse the initial predicted mutation position with the measured features extracted from the Doppler second-order derivative sequence;

[0049] The Kalman filter outputs an optimized corrected mutation position and a confidence weight coefficient representing the credibility of the corrected position;

[0050] The boundary of the modified mutation window is adaptively set according to the modified mutation position and the confidence weight coefficient.

[0051] Beneficial effects: This application addresses the problem of blind resource allocation caused by the inability to foresee key events by introducing a forward-looking mutation prediction mechanism. It accurately predicts future high-value observation windows, driving energy pools and non-uniform sampling strategies to optimally plan and target energy and sampling resources across time and in response to future events. Furthermore, it addresses the problem of unclear optimization objectives caused by the superficial state perception of existing methods, ensuring the accuracy and efficiency of the optimization process. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of a Doppler observation strategy optimization method for a satellite-borne interference source provided by an embodiment of the present invention.

[0053] Figure 2 This is a flow chart of obtaining enhanced Doppler observation data provided by an embodiment of the present invention.

[0054] Figure 3 This is a flow chart for evaluating the matching degree between the balance of an energy storage unit and the power consumption demand in the time period, provided by an embodiment of the present invention.

[0055] Figure 4 This is a flow chart of obtaining a precise Doppler estimation by constructing a virtual mirror signal and iteratively adjusting it, provided by an embodiment of the present invention.

[0056] Figure 5 This is a flow chart for predicting mutation characteristics in Doppler data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0058] It should be noted that the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or are inherent to these processes, methods, products or apparatus.

[0059] Research has found that existing optimization methods generally lack the ability to proactively predict and plan for future critical events. Doppler observations from LEO satellites are a typical event-driven process, with the vast majority of their valuable positioning information concentrated within brief, sudden window periods when the Doppler frequency rate of change peaks. Existing DRL methods, such as DDPG, are inherently state-driven, making reactive decisions based on the current channel or signal state. This mechanism cannot foresee impending sudden window periods, and therefore cannot proactively enter energy-saving mode to conserve energy during stable periods with low observation efficiency (such as low elevation angles). It also fails to plan ahead and prepare for high-density sampling before sudden window periods arrive. This reactive, rather than proactive, approach results in suboptimal allocation of energy and sampling, two of the most valuable onboard resources, for future high-value events.

[0060] Furthermore, the problem of static and blind allocation of observation resources manifests itself in two aspects. First, there is the waste and mismatch of sampling resources. The information content of the Doppler signal is extremely uneven throughout the entire transit arc. The vast majority of its positioning information is contained in the brief, sudden window where the frequency changes dramatically. During the remaining, vast, stable periods, the signal changes slowly, and the information density is extremely low. Adopting a fixed, uniform sampling strategy inevitably results in a large number of redundant observations during the stable periods, wasting precious onboard storage and computing resources. Furthermore, during the most critical sudden window, the fixed sampling rate is likely to fail to meet the Nyquist requirement for capturing transient signal details, resulting in information loss. This is a typical mismatch between resources and information needs.

[0061] Second, energy management is crude and inefficient. The static power budget cannot be dynamically adjusted based on the real-time progress of the mission and the actual signal conditions. It lacks an intelligent energy pooling mechanism, proactively degrading operation during low-elevation periods with poor signal quality and low observation value to conserve energy. This precious energy can then be allocated to high-value periods with high signal-to-noise ratio and good observation value, thereby achieving optimal temporal allocation of energy resources throughout the mission cycle. Existing methods lack the semantic depth of state perception, specifically a lack of in-depth understanding and extraction of the intrinsic physical characteristics of Doppler signals. In IRS-assisted communications, the state input for DRL optimization is typically macroscopic metrics such as channel state information (CSI) or signal-to-noise ratio (SINR). However, in Doppler positioning tasks, the value of observations is determined not by the average signal strength but by its inherent high-order dynamic features that characterize kinematic characteristics, such as the peak of the Doppler second-order derivative, the texture fingerprint of the signal's time-spectral spectrum, or the strength of the signal's singularity. Existing optimization frameworks lack a dedicated semantic layer, making it impossible to extract these deep, physically meaningful mutational features from raw observational data as a basis for decision-making. This superficial understanding of state perception prevents the optimization algorithm from identifying truly valuable observation periods, and the optimization process loses its precise focus, making it difficult to fundamentally improve the efficiency of capturing key information and the ultimate success rate of the mission. In real onboard environments, especially when dealing with weak interfering signals, simple derivative or frequency domain analysis methods are extremely susceptible to noise interference. Random spikes in the noise may be mistaken for mutations (false alarms), while actual weak mutation features may be submerged in the noise and go undetected (missed alarms).

[0062] like Figure 1 As shown in the figure, a Doppler observation strategy optimization method for satellite-borne interference sources is proposed, including:

[0063] Obtain the original RF signal of the onboard interference source and the precise satellite orbit data, predict the mutation characteristics in the Doppler data, and generate an adaptive observation plan.

[0064] For example, the onboard receiver acquires intermediate frequency (IF) sampling data and performs a fast Fourier transform (FFT) to identify the spectral peaks of the interfering signal. GPS timing information and satellite attitude measurement data are combined for time synchronization and coordinate system conversion. A frequency tracking loop is then used to extract the instantaneous Doppler shift value sequence. Simultaneously, the signal's power spectral density is calculated and used as a metric for evaluating subsequent data quality. Based on this, the Doppler rate of change characteristics are calculated and analyzed. Specifically, based on the instantaneous Doppler shift value sequence extracted in the previous step, the first-order derivative is calculated using a five-point difference method. The second-order Doppler derivative sequence is then calculated using a Savitzky-Golay filter. By setting a threshold, periods where the absolute value of the second-order derivative exceeds the threshold are identified and marked as potential mutation intervals. The corresponding satellite elevation angle range is also recorded. A prediction-correction dual-loop mechanism is used to locate mutation points. This mechanism draws on a knowledge base of historical transit mutation points, which stores statistical feature vectors (including mean and variance) of mutation points under similar conditions over the past 30 days, for example. Based on the current satellite orbital elements and geographic location, the initial mutation point location for this transit is matched and predicted from a knowledge base. During the observation process, the Doppler second-order derivative sequence is monitored in real time. When the deviation between the measured and predicted values ​​exceeds a preset threshold (e.g., 15%), a correction mechanism is triggered, fusing the predicted and measured information to output a corrected mutation window. Based on the corrected mutation window, a non-uniform sampling strategy is generated. This strategy aims to perform dense sampling within the mutation interval and sparse sampling within the stationary interval, thereby conserving resources while ensuring data quality. Furthermore, based on this sampling strategy and the power consumption of the receiver in different operating modes, the expected energy cost distribution of the entire observation mission can be estimated.

[0065] According to the adaptive observation scheme, energy-aware resource scheduling is performed and observations are executed to obtain enhanced Doppler observation data.

[0066] Specifically, this step can be implemented through energy-aware dynamic resource scheduling, multipath effect identification, and collaborative information enhancement. Regarding resource scheduling, the system reads the non-uniform sampling strategy and power consumption requirement estimates to activate the energy pooling mechanism. During periods of low satellite elevation angle (e.g., less than 20°), known as energy accumulation windows, the system accumulates energy savings by reducing the receiver's operating mode (e.g., switching from a 15W high-precision mode to a 3W energy-saving mode). When the satellite enters a high-value observation period, defined by Doppler mutation characteristics, the system assesses whether energy reserves meet the power consumption requirements for that period. If so, high-density sampling observations are authorized. Regarding acquisition of observation data, the system executes the finalized sampling plan. During the observation process, multipath effects can be exploited for information enhancement. Specifically, the presence of multipath effects is identified by analyzing the time-varying characteristics of the signal's power spectral density. If multipath is detected, a subspace estimation algorithm, such as the MUSIC algorithm, can be used to separate the direct path signal component and one or more reflected path signal components from the received signal. By estimating the Doppler frequency shift differences and relative time delays of each path and utilizing the geometric constraints attached to the reflection path to construct an overdetermined set of equations, solving this set of equations can obtain Doppler measurement values ​​with enhanced parameter estimation accuracy as part of the enhanced Doppler observation data.

[0067] Positioning solution and performance evaluation are performed based on the enhanced Doppler observation data, and parameters for iterative optimization of the observation strategy are generated.

[0068] For example, a state vector containing the interference source's position, velocity, and acceleration is established. Based on enhanced high-precision Doppler measurements and precise satellite orbit data, a nonlinear observation equation is constructed. An extended Kalman filter (EKF) is used to recursively update this state vector to estimate the interference source's motion state. The clean, motion-compensated Doppler data is input into an iterative positioning algorithm, which solves for the interference source's geographic coordinates by combining multi-time observation constraints. Metrics such as the positioning error ellipse and geometric dilution of precision (GDOP) are then calculated to evaluate the positioning performance of the transit and generate a quantitative positioning quality score Q. In this embodiment, to achieve adaptive iterative optimization of the observation strategy, the positioning quality score Q is combined with the final energy reserve unit balance and the accuracy of the predicted catastrophe point to form a multi-dimensional task performance metric. Based on this multi-dimensional metric, a reward function value R is calculated to evaluate the overall performance of the observation task. This reward function value R serves as feedback to update the internal parameters of the deep reinforcement learning network through a gradient backpropagation algorithm (e.g., the actor and critic parameter updates used in the TD3 network). The updated internal parameters of the network constitute the optimization strategy parameter set for the next task iteration optimization observation strategy.

[0069] like Figure 2As shown, according to one aspect of the present application, enhanced Doppler observation data is obtained, including: identifying an energy accumulation window defined by a low elevation angle period in an adaptive observation scheme, and a high-value observation period defined by a Doppler mutation characteristic; within the energy accumulation window, accumulating saved energy values ​​by lowering the working mode of the receiver, and counting them into an energy reserve unit; when entering a high-value observation period, evaluating the matching degree between the balance of the energy reserve unit and the power consumption requirement of the period; under the condition that the balance of the energy reserve unit meets the power consumption requirement, authorizing the execution of high-density sampling to obtain enhanced Doppler observation data.

[0070] The energy accumulation window refers to the period when the satellite's elevation angle relative to the ground station is below a preset threshold (e.g., 20°). During this period, signal propagation loss is typically high and multipath effects are severely impacted, resulting in relatively low observation data quality. Therefore, it serves as a suitable window for implementing energy-saving strategies to accumulate energy. High-value observation periods are typically defined by the modified mutation window, when Doppler frequency is predicted to undergo significant changes.

[0071] Further, if Figure 3 As shown in the figure, the matching degree between the balance of the energy reserve unit and the power consumption demand in the period is evaluated, specifically including: constructing an energy-precision benefit function that characterizes the relationship between observation accuracy and energy consumption; based on the energy-precision benefit function, generating a segmented benefit evaluation matrix for different power consumption modes in the high-value observation period; establishing a multi-constraint optimization model with the goal of maximizing the total benefit and integrating the balance of the energy reserve unit (energy reserve account) and the mode switching cost as constraints; solving the multi-constraint optimization model to obtain the optimal power consumption allocation plan, and performing high-density sampling based on the optimal power consumption allocation plan.

[0072] For example, the specific implementation of this process is as follows: In order to quantify the relationship between energy input and the return of observation data accuracy, the benefit function U(E) = α·log a (1+β(E / E ref ) ); where U(E) represents the accuracy benefit value obtained by investing energy E; E represents the energy allocated to the receiver during a specific observation period; α is the accuracy weighting factor, which can be set to α=0.8, for example; β is the energy sensitivity coefficient, which can be set to β=2.5, for example; E ref is the standard transit energy consumption benchmark value, which serves as a reference quantity. a can be 2, e, or 10. This logarithmic function reflects the principle of diminishing marginal benefits, that is, the initial energy input can bring significant accuracy improvement, but as the energy input increases, the benefit growth will tend to be flat. Based on this benefit function, a segmented benefit evaluation matrix M can be generated for different power consumption modes during high-value observation periods. benefit , each element M in the matrix ijRepresents the benefit value that can be brought by adopting the jth power consumption mode in the i-th observation segment. Establish and solve the multi-constraint optimization model. The objective function of the model is: maxΣ(M ij ·x ij ), where x ij is the variable for selecting the jth power consumption mode in the i-th observation segment; it aims to maximize the total benefit during the entire observation mission. The constraints include: energy constraint Σ(P seg [i]·t seg [i]·x ij )≤E avail , ensuring that the total energy consumption does not exceed the reserve, where P seg [i] is the power consumption of the selected power consumption mode in the i-th observation segment, t seg [i] is the duration of the i-th observation segment, E avail The upper limit of the system's energy reserve; continuity constraint | mode i -mode i+1 |≤1, avoid drastic jumps in the receiver working mode, where mode i is the number of the working mode adopted by the receiver in the i-th segment; and the mode switching cost, for example, each switch consumes an additional 0.1J. In this embodiment, the dynamic programming algorithm is used to solve the above model to obtain the optimal power consumption allocation solution X opt And the corresponding working mode sequence mode[i].

[0073] Optionally, in some implementations, the algorithm used to solve the multi-constraint optimization model can also be other optimization algorithms, such as genetic algorithms, simulated annealing algorithms, or greedy algorithms with lower computational complexity when rapid real-time adjustments are required. Furthermore, the form of the energy-accuracy benefit function U(E) can be adjusted based on the specific satellite mission requirements and hardware characteristics. For example, a piecewise linear function or an exponential function can be used for modeling. Finally, observations are performed based on the optimal power consumption allocation scheme. This embodiment resolves the contradiction between the limited on-orbit resources (energy) and the high performance requirements of observation missions, enabling refined and intelligent resource management.

[0074] Furthermore, in order to cope with the situation that the actual energy consumption may not be consistent with the expected energy consumption during the on-orbit execution, the energy intelligent allocation process also includes a real-time adjustment mechanism. This mechanism performs an evaluation at a fixed time period (for example, every 10 seconds) to calculate the actual consumed energy E actual and the energy E that should be consumed according to plan planned When the absolute value of the deviation |ΔE| exceeds the preset adjustment threshold (for example, 0.5J), it indicates that there is an unplanned energy consumption change that cannot be ignored. At this time, the system will immediately recalculate the benefit evaluation matrix M for the remaining transit period.benefit_new However, to ensure real-time performance, the system no longer uses complex dynamic programming algorithms. Instead, it uses a fast greedy algorithm with lower computational complexity (e.g., O(n)) to update the operating mode for subsequent periods and generate a real-time mode switching instruction sequence cmd[i]. This rapid feedback and adjustment loop ensures that the satellite's energy efficiency remains near optimal throughout the entire transit period.

[0075] According to one aspect of the present application, it also includes: when the assessment determines that the balance of the energy reserve unit cannot meet the power consumption demand, starting the degradation strategy; the degradation strategy calculates the ratio of the current remaining energy to the expected demand of the subsequent task, and compares the ratio with at least one preset degradation threshold; if the ratio is lower than the degradation threshold, generating a degradation sampling plan aimed at reducing the total energy consumption, the degradation sampling plan is achieved by increasing the sampling interval of non-critical observation areas or shrinking the spatial range of high-density sampling; performing observations according to the degradation sampling plan.

[0076] Among them, the degradation strategy is a robust mechanism to ensure the completion of the core task. Its motivation is to avoid the complete failure of the entire observation task (especially the observation of high-value mutation periods) due to energy exhaustion, and to achieve graceful degradation of the system. Specifically, the degradation strategy is implemented as follows: the system continuously monitors the remaining energy E remain and the expected demand for subsequent tasks E needed , and calculate the energy ratio r energy =E remain / E needed . E needed It can be dynamically calculated based on the subsequent observation plan and power consumption estimation model that have not yet been executed. energy The system generates a downgrade sampling plan based on the different intervals in which the system is located. The system presets at least one, preferably multiple, downgrade thresholds to achieve hierarchical downgrade management. For example, two downgrade thresholds can be set: a first-level downgrade threshold (e.g., 0.7) and a second-level downgrade threshold (e.g., 0.4). Triggering and executing the first-level downgrade: When 0.4≤r energy <0.7, triggering the first level downgrade. At this time, the system will update the original sampling plan and generate a first level downgrade sampling plan. The core idea is to sacrifice the secondary and preserve the primary. The specific measure can be to increase the sampling interval of all non-mutation windows (i.e., stable intervals) by a fixed ratio based on the original plan, such as increasing it by 50%. This can reduce the overall energy consumption without affecting the data density of the core mutation area. Triggering and executing the second level downgrade: When monitoring r energyWhen it is <0.4, it indicates that the energy situation is extremely critical, and the second-level downgrade is triggered. The second-level downgrade strategy is more stringent and aims to protect the core. Specific measures may be that the system will generate a sampling plan in emergency mode, retaining only the high-density and intensive sampling plan within a very small range (for example, ±3°) before and after the determined corrected mutation window. All other areas outside this range, regardless of whether the original plan is dense or sparse, are forced to switch to the energy-saving mode and basic sampling rate with the lowest power consumption. This ensures that even under extremely unfavorable energy conditions, the observation mission of the highest-value scientific event, the Doppler mutation, can be completed with a high probability (for example, ensuring that at least 80% of high-value observation missions are completed), thereby improving the mission success rate and reliability of the entire observation system.

[0077] Optionally, in some embodiments, the adjustment of the degradation strategy can be smoother. For example, the increase in the sampling interval can be energy In addition, the boundary of the shrinking high-density sampling space range can also be dynamic, rather than fixed ±3°, and its specific size can be determined by comprehensive consideration of r energy and the predicted value of mutation intensity.

[0078] This embodiment solves the problem of on-board energy resource waste caused by extensive and inefficient energy management. Specifically, the introduction of the energy pool mechanism enables the observation strategy to have energy planning capabilities across time scales. By actively lowering the receiver operating mode in energy accumulation windows with low observation benefits such as low elevation angles, the saved energy is deposited into a virtual account, and then optimally withdrawn according to the energy-precision benefit function during the predicted high-value observation period. Combined with a multi-constraint optimization model that takes total benefit maximization as the goal and integrates mode switching costs, it ensures that every bit of energy is used to the best effect. The traditional static power budget with a huge margin is transformed into a sophisticated optimal energy flow management that is dynamically adjusted in real time with the mission, thereby improving the utilization efficiency of on-board energy, the most scarce resource.

[0079] According to one aspect of the present application, predicting mutation features in Doppler data can be as follows: extracting time-series Doppler frequency shift data from the original radio frequency signal and calculating the corresponding Doppler second-order derivative sequence; calling a historical transit mutation point knowledge base and combining it with the current satellite precise orbit data to generate an initial predicted mutation position; monitoring the Doppler second-order derivative sequence in real time, and triggering a correction mechanism when the measured characteristics of the sequence deviate from the expected characteristics of the initial predicted mutation position and exceed a preset threshold; the correction mechanism fuses the initial prediction and measured characteristics to output a corrected mutation window for generating an adaptive observation plan.

[0080] Furthermore, the correction mechanism includes: using a Kalman filter to fuse the initially predicted mutation position with the measured features extracted from the Doppler second-order derivative sequence; the Kalman filter outputs an optimized corrected mutation position and a confidence weight coefficient that characterizes the credibility of the corrected position; based on the corrected mutation position and the confidence weight coefficient, the boundary of the corrected mutation window is adaptively set, where a higher confidence weight corresponds to a narrower window, and a lower confidence weight corresponds to a wider window.

[0081] In this embodiment, the system first reads all historical transit observation records over a period of time (e.g., 30 days). These records contain various relevant factors, such as the satellite elevation angle, atmospheric conditions, and season, at the time of each successful observation of the mutation point. Using algorithms such as cluster analysis, mutation point data with similar environmental conditions (such as similar orbit type, season, and geographical region) are grouped together. For each group of data, its statistical feature vector F is calculated. hist , which contains the mean, variance, skewness and other statistics of the sudden elevation angle under this type of conditions to quantify its distribution characteristics. A classification index table is established for these feature vectors so that subsequent tasks can quickly retrieve and match according to the current conditions. Before a new transit mission begins, the system obtains the current orbital elements and satellite geographic location, and uses this information to match the classification index table established in the first step to find the historical condition category that is most similar to the current conditions. The system then extracts the statistical feature vector F corresponding to the category hist To improve the accuracy of the prediction, the model also combines the atmospheric refractive index and ionospheric total electron content (TEC) values ​​obtained in real time by other sensors, and through the framework of Bayesian reasoning, integrates historical statistical priors and real-time environmental information to finally calculate the initial predicted elevation angle θ for this transit. pred and its confidence interval [θ low ,θ high During the satellite transit, the system continuously monitors the Doppler second-order derivative sequence calculated in real time. When the absolute value of the second-order derivative of three consecutive sampling points exceeds the preset trigger threshold (for example, 0.5Hz / s), the system will trigger the signal. 2 ), the system determines that it has captured a real mutation signal and immediately triggers the correction procedure. After the trigger, the system extracts the measured mutation characteristics (such as the elevation angle when the mutation occurs) and calculates its difference with the initial predicted elevation angle θ pred The deviation Δθ between them. At this point, the Kalman filter is called, which will θ pred As the prior prediction value, the measured mutation elevation angle is used as the observation value, and the optimal estimate that combines the two types of information is generated through the state update process of the filter, that is, the corrected elevation angle θ corr , and output the confidence weight coefficient w that represents the credibility of the correction result corrThe system is based on the modified elevation angle θ corr and confidence weight coefficient w corr , an adaptive window algorithm is used to dynamically determine the final mutation window boundary [θ start ,θ end ]. The rule of this algorithm is that when w corr When the value is high (such as w corr >0.8), indicating that the system is very confident in the judgment of the mutation location, and a narrower tight window will be set at this time, such as [θ corr -2°, θ corr +2°] to concentrate resources. On the contrary, when w corr When the value is low (such as w corr <0.5), indicating that there is a large uncertainty, the system will set a wider conservative window, such as [θ corr -5°, θ corr +5°] to ensure that the target is not missed. This dynamically adjusted window will be directly used to generate subsequent non-uniform sampling strategies, thereby achieving precise targeted delivery of observation resources.

[0082] This embodiment addresses the failure to capture critical information due to prediction lag and low robustness by introducing two forward-looking mutation feature prediction methods: a prediction-correction dual-loop mechanism and texture fingerprint recognition. Specifically, the prediction-correction mechanism optimally integrates statistical priors from historical data (via a knowledge base) with current evidence from real-time observations (via second-order derivative monitoring) within a Kalman filter framework. This eliminates the need for the system to confirm mutations after the fact, instead enabling the formation of a high-probability prediction window before a mutation occurs, thus achieving forward-looking decision-making. The texture fingerprint recognition method, on the other hand, transforms one-dimensional time series analysis into a two-dimensional image recognition problem, leveraging a convolutional neural network (CNN) to extract physically meaningful patterns from the deep textures of the time-spectrogram. Its robustness far exceeds that of traditional single-threshold decision methods, particularly in low signal-to-noise ratio environments, improving its ability to capture subtle mutation features. This ensures that observation strategies can accurately target the highest-value observation periods in the future, providing a reliable foundation for the targeted allocation of all subsequent resources.

[0083] Alternatively, as Figure 5As shown in FIG, the mutation characteristics in Doppler data can also be predicted by: performing spectrum analysis on the original radio frequency signal, and combining it with the satellite precise orbit data, extracting the time-series Doppler frequency shift data through the frequency tracking loop; performing time-frequency transformation on the time-series Doppler frequency shift data to generate a time-frequency spectrum, and processing the time-frequency spectrum into a grayscale image; extracting the local binary pattern and the grayscale co-occurrence matrix from the grayscale image as texture features, and constructing a multidimensional texture feature vector; feeding the multidimensional texture feature vector into a pre-trained convolutional neural network, parsing and outputting the mutation type and mutation time in the Doppler data, and generating an adaptive observation scheme.

[0084] In this embodiment, the Doppler frequency shift time series f obtained after preprocessing is D (t) is subjected to a time-frequency transform. Specifically, a short-time Fourier transform (STFT) can be used. For example, a sliding time window (the window length can be set to 5 seconds, and the sliding step can be set to 0.5 seconds) is constructed, and a Fourier transform is performed on the data segments within each window to generate a time-spectrogram S(t, f). A time-spectrogram is a two-dimensional representation, with the x-axis representing time t and the y-axis representing frequency f. The brightness or color of each point (t, f) in the time-spectrogram represents the energy intensity of the signal at that frequency at that moment. The generated time-spectrogram S(t, f) is normalized, its energy intensity values ​​are mapped to an integer range of [0, 255], and converted into a grayscale image I(x, y). In this way, the temporal variation pattern of the Doppler signal is intuitively converted into the texture pattern in the image. Features that characterize its local and global texture properties are extracted from the grayscale image I(x, y). In this embodiment, at least the following are included: Local Binary Patterns (LBP): LBP compares the grayscale values ​​of a pixel with other pixels in its neighborhood and encodes the comparison results into binary numbers, thereby effectively capturing the microscopic structural information of the image, such as edges, corners, and spots. Gray-Level Co-occurrence Matrix (GLCM): GLCM reflects the comprehensive information of the image grayscale about direction, adjacent intervals, and change amplitude by calculating the probability of occurrence of grayscale values ​​of two pixel pairs separated by a certain distance and direction in the image. Multiple statistical features such as contrast, correlation, energy, and homogeneity can be further calculated from GLCM. Gabor filter response: Gabor filter performs well in extracting local spatial frequency features of images and can capture texture information from different scales and directions. The multiple features extracted from the above are combined to construct a high-dimensional mutation texture feature vector V texture For example, a 128-dimensional feature vector can be constructed, which fully describes the fingerprint of the Doppler signal in the time-frequency domain. textureThe network is fed into a pre-trained convolutional neural network (CNN). A convolutional neural network is a deep learning model whose network structure (for example, containing 3 convolutional layers and 2 fully connected layers) is particularly suitable for processing image and high-dimensional feature vector classification tasks. The network is trained on a large amount of labeled historical data (including various mutation types) to learn the complex mapping relationship between specific texture feature vectors and specific mutation patterns. The network receives V texture After that, the classification results of the current data window will be output, including the mutation type (for example, step type, pulse type, oscillation type or gradual change type) and the mutation time t mut accurate estimate of .

[0085] Optionally, in some scenes with low signal-to-noise ratio and weak signal, in order to enhance the significance of texture features, before generating the grayscale image I(x, y), a cumulative average of multiple consecutive time-spectrograms can be performed, that is, I enh =Σ(I i ) / N, where N is the number of accumulated frames. This method effectively suppresses random noise and highlights the persistent, real texture structures associated with mutation events, thereby improving the success rate and accuracy of mutation detection in weak signal environments.

[0086] Regarding the steps of constructing a multidimensional texture feature vector, it is further as follows: a multidimensional texture feature vector, such as a 128-dimensional feature vector, can be specifically constructed by splicing multiple different types of texture features to achieve a comprehensive and non-redundant description of the time-frequency spectrum information. An exemplary construction method is: extracting the statistical histogram features of the local binary pattern (LBP) from the grayscale image I(x, y); calculating the gray-level co-occurrence matrix (GLCM) of the image, and extracting multiple statistics such as contrast, correlation, energy, and homogeneity from it; constructing a set of Gabor filter groups covering different scales and directions, and using them to filter the image, and taking its response energy as the Gabor texture feature. These three sets of features (LBP features, GLCM features, Gabor features) are spliced ​​in a predetermined order to form the final 128-dimensional feature vector V that can fully characterize the Doppler mutation texture fingerprint. texture .

[0087] According to one aspect of the present application, the steps of generating an adaptive observation scheme include a process of generating a non-uniform sampling strategy, which includes: based on the identified Doppler mutation window, calculating the ratio of the maximum value to the average value of the second-order derivative of Doppler within this window; adaptively determining the proportional coefficient in the inverse proportional sampling function according to this ratio, where a larger ratio corresponds to a smaller proportional coefficient; applying the inverse proportional sampling function containing the adaptive proportional coefficient to generate a dense sampling interval for the mutation interval and a sparse sampling interval for the stationary interval; constructing a transition zone at the boundary between the dense sampling interval and the sparse sampling interval, and using a sigmoid function to smoothly connect the dense and sparse sampling intervals to ensure a continuous and shock-free change in the sampling rate.

[0088] Specifically, after determining the mutation window, the system first calculates the maximum value d 2 f max of the absolute value of the second-order derivative of Doppler within this window and the average value d 2 f[[ID=I0]] avg . Calculate the ratio r = d 2 f max / d 2 f avg , and this ratio r is used to evaluate the severity of this mutation. The system uses a piecewise function to adaptively determine the inverse proportional coefficient k according to the r value: when r > 10, set k = 0.1; when 5 < r ≤ 10, set k = 0.2; when r ≤ 5, set k = 0.5. This makes the sampling strategy not only respond to instantaneous changes but also to the macroscopic characteristics of the overall event. Generate a sampling function Δt = k / |d 2 f / dt 2 | containing the adaptive coefficient k. To avoid sudden changes in the sampling rate at the boundaries of different intervals, the system sets a transition zone at the boundaries of the mutation window (such as [θ start -1°, θ start and [θ end , θ end +1°]). Within the transition zone, a sigmoid function is used to smoothly connect the dense sampling interval Δt dense and the sparse sampling interval Δt sparse . This smoothing function can be specifically expressed as: Δt trans =Δt sparse +(Δt dense -Δt<00,00083>) / (1 + exp(-10(θ - θ boundary))). It can ensure that the rate of change of the time interval between adjacent sampling points does not exceed the preset smooth threshold, such as 50%, thereby ensuring the smoothness of hardware execution. After generating the sampling interval function Δt(θ), an energy consumption estimation model is constructed. This model maps different sampling intervals Δt to instantaneous power consumption P(θ) based on the power consumption characteristics of the receiver (for example, 15W in high-precision mode and 3W in energy-saving mode). For example, when Δt<0.5s, corresponding to the high-precision mode, P(θ)=15W; when Δt>2s, corresponding to the energy-saving mode, P(θ)=3W; in the middle area, the power consumption value can be linearly interpolated. By integrating the instantaneous power consumption P(θ) of the entire transit arc, the expected total energy consumption E of this mission can be calculated. total , and generate a segmented power consumption requirement sequence P seg [i] Starting from the initial time t0 of the mission, according to the continuous sampling interval function Δt(θ), the orbit prediction is used to obtain θ at each moment. i , and recursively calculate each sampling time t i+1 =t i +Δt(θ i ), generate a complete non-uniform sampling time sequence {t i Before outputting the final executable sampling plan, the system will also verify it to ensure that it meets basic quality requirements, such as verifying that the number of sampling points in the mutation window is not less than 20 and that the sampling point interval in the stable area does not exceed 5 seconds.

[0089] This embodiment addresses the waste and mismatch of sampling resources caused by fixed uniform sampling by designing an adaptive inverse-proportional sampling function and combining it with a sigmoid function for smooth transitions. Specifically, the introduction of the inverse-proportional sampling function achieves a precise negative correlation between the sampling rate and the signal's information content (characterized by the second-order Doppler derivative). This ensures the most intensive sampling at information-rich abrupt points and sparse sampling in information-sparse stationary regions, thus enabling on-demand allocation of sampling resources. Furthermore, the proportionality coefficient k is adaptively adjusted based on the overall severity of the abrupt event (via the ratio r), achieving both macro- and micro-level adaptation. The use of a sigmoid function for smooth transitions at interval boundaries ensures continuous, impact-free sampling of the sampling rate, improving the stability of subsequent signal processing and protecting the hardware life of the onboard actuators. This optimizes the utilization of limited sampling resources.

[0090] According to one aspect of the present application, when the original RF signal is below a threshold, the process of obtaining enhanced Doppler observation data is implemented using a step-by-step Doppler frequency cascade precision estimation method. Specifically, the method comprises: performing preliminary spectral peak positioning based on a frequency grid adaptively adjusted according to the spectral energy distribution to obtain a coarse Doppler estimate; based on the coarse Doppler estimate, obtaining an intermediate Doppler estimate by injecting a reverse Doppler sequence and searching for a solution that minimizes the spectral flatness of the compensated signal; based on the intermediate Doppler estimate, minimizing the time rate of change of the beat frequency component generated by cross-correlation with the original signal by constructing a virtual mirror signal and iteratively adjusting it to obtain a final fine Doppler estimate. Based on the fine Doppler estimate, enhanced Doppler observation data is sampled.

[0091] This embodiment achieves a balance between computational efficiency and estimation accuracy through a coarse-to-fine strategy. It first uses a method with low computational cost to quickly narrow the frequency uncertainty range, and then uses a more computationally expensive but more accurate method to perform a fine search, thereby efficiently obtaining a high-precision estimation result. The specific implementation method includes: Coarse estimation based on adaptive frequency grid evolution. The goal of this stage is to quickly locate the approximate position of the Doppler frequency from a larger frequency uncertainty range (for example, ±500Hz). Based on the preliminary Doppler shift estimate f D_coarse , initialize the frequency search grid G0. For example, G0={f i |f i =f D_coarse +i×Δf, i∈[-100, 100]}, where the initial grid resolution Δf can be set to 10Hz. The calculated signal is at each grid point f i The spectral energy E on i The grid density is adaptively adjusted based on the energy distribution. Specifically, in areas with higher energy (such as E i >0.7×max(E)), the grid is refined (for example, the resolution is changed to Δf / 4); in the area with lower energy (for example, E i <0.3×max(E)), the grid is sparser (for example, the resolution is changed to 4Δf). After several (for example, 5) iterations, the grid points will naturally gather around the true spectrum peak. At this time, in the densest grid area, the peak position is fitted by parabolic interpolation and other methods to output a high-precision Doppler rough estimate f D_grid , its accuracy can reach about ±50Hz.

[0092] In this stage, the frequency estimation problem is transformed into an optimization problem by using signal processing technology. Specifically, based on the obtained f D_grid , in a smaller range (e.g. [f D_grid -50Hz, fD_grid +50Hz]), a series of tentative inverse Doppler sequences exp(-j2πf probe ×t). Multiply these sequences one by one with the original signal s(t) to obtain the compensated signal s comp (t)=s(t)×exp(-j2πf probe ×t). When the test frequency f probe When the frequency is exactly equal to the true Doppler frequency, the Doppler effect is compensated, s comp The spectrum of (t) will become the flattest. Calculate each s comp The spectrum flatness index F=σ(|S comp (f)|) / μ(|S comp (f)|), where σ and μ are the standard deviation and mean of the spectrum amplitude respectively. Search for the f that minimizes the spectrum flatness index F. probe , which is the Doppler mean value f D_inject To accelerate convergence, optimization algorithms such as the golden section search can be used. This can improve the estimation accuracy to approximately ±0.1 Hz.

[0093] like Figure 4 As shown, a Doppler precise estimate is obtained by constructing a virtual mirror signal and iteratively adjusting it, including: constructing a virtual mirror signal using the conjugate of the original RF signal and taking the Doppler mean estimate as the frequency basis; calculating the cross-correlation function between the original RF signal and the virtual mirror signal, and separating the beat frequency component from the cross-correlation function; iteratively adjusting the frequency used to construct the virtual mirror signal until the time rate of change of the beat frequency component converges to a preset minimum value, and determining the converged frequency as the Doppler precise estimate.

[0094] Specifically, based on the median estimate f D_inject , construct a virtual mirror signal s mirror (t)=s*(t)×exp(j4πf D_inject ×t), where s*(t) is the conjugate of the original signal. Calculate the original signal s(t) and the mirror signal s mirror (t) is the cross-correlation function R(τ). When there is a small frequency difference between the two, the cross-correlation result will contain a beat frequency component with a very low frequency. The beat frequency f beat It can be extracted from the phase difference of adjacent sampling points of the cross-correlation function. Iteratively fine-tune the frequency f used to construct the image signal D_inject , and observe the time rate of change of the beat frequency df beat / dt. When f D_inject When it is consistent with the true Doppler frequency, df beat / dt will approach zero theoretically. beatWhen / dt converges to the preset minimum value, the iteration ends, and the frequency used to generate the image signal is the final Doppler precision estimate f D_mirror , its accuracy can reach ±0.01Hz or even higher. Optionally, a three-level estimation error model can be constructed, and the errors at each level can be back-propagated and compensated through the Kalman filter, further improving the final accuracy to, for example, below ±0.001Hz.

[0095] According to one aspect of the present application, the step of obtaining enhanced Doppler observation data, when sampling is performed in a non-uniform manner, also includes a data reconstruction process, specifically: obtaining a Doppler fine estimate, and using the Doppler fine estimate to pre-Doppler compensate the non-uniformly sampled observation data to obtain a compensated stationary signal; solving a pre-constructed compressed sensing convex optimization model for the compensated stationary signal to obtain a reconstructed stationary signal component; and re-assigning the frequency trend represented by the Doppler fine estimate to the reconstructed stationary signal component to generate enhanced Doppler observation data.

[0096] Standard compressed sensing algorithms are not very effective in processing non-stationary signals with strong time-varying characteristics (such as Doppler signals). This embodiment first uses high-precision Doppler estimation to remove the non-stationary trend, thereby transforming the difficult non-stationary signal reconstruction problem into a relatively easy stationary signal reconstruction problem. The specific implementation method is: the non-uniformly sampled observation data y is combined with the reverse Doppler sequence exp(-j2πf D_final ×t) to obtain an approximately stationary signal s comp Where f D_final is the final Doppler precision estimate obtained by the above cascade precision estimation. comp , solve the pre-built compressed sensing convex optimization model: min||α||1subject to||s comp -ΦD stationary ×α|| 2 ≤ε. Where α is the signal in a certain transformation domain (defined by the dictionary D stationary Definition, for example, the sparse coefficients under the discrete cosine transform (DCT) basis, Φ is the measurement matrix that characterizes the non-uniform sampling process, and ε is the noise tolerance. This optimization problem can be efficiently solved using algorithms such as the alternating direction multiplier method (ADMM). The sparse coefficient α is obtained by solving final After that, through the inverse transformation D stationary ×α final The stationary signal component can be reconstructed. By re-adding the previously stripped Doppler frequency trend, a complete, high-resolution reconstructed signal can be obtained.

[0097] The description of the cascaded fine estimation method and data reconstruction process can be further refined as follows: In the coarse estimation based on adaptive frequency grid evolution, adaptive adjustment specifically refers to: in the high energy region (such as spectral energy E i Greater than 70% of the maximum energy), the grid resolution is refined to one-fourth of the original (Δf / 4); in low energy areas (such as E i Less than 30% of the maximum energy), the grid resolution will be sparse to four times the original (4Δf). After the entire cascaded fine estimation process, a cascaded error backpropagation compensation step can be added to further improve the estimation accuracy. Specifically, a three-level estimation error model is constructed to calculate the coarse estimation error ε1, the medium estimation error ε2, and the fine estimation error ε3 respectively. After estimating the statistical characteristics of the errors at each level through the Kalman filter, a backpropagation compensator is designed. For example, the following formula is used to compensate the estimation results of the first two levels: f D_grid_new =f D_grid -α1·ε3 and f D_inject_new =f D_inject -α2·ε3; where the compensation coefficients can be set to α1=0.3 and α2=0.7. After compensation, the intermediate estimation and fine estimation steps can be re-executed, and this process can be iterated until the final fine estimation error |ε3| is less than a very small threshold, such as 0.001Hz, thereby obtaining the final Doppler fine estimate f with extremely high confidence. D_final .

[0098] During the data reconstruction process, the Doppler-assisted reconstruction method cleverly transforms the difficult non-stationary signal reconstruction problem into a relatively easy stationary signal reconstruction problem. When the alternating direction multiplier method (ADMM) is used to solve it, the convergence speed and peak signal-to-noise ratio (PSNR) are improved.

[0099] According to one aspect of the present application, the step of obtaining enhanced Doppler observation data also includes a four-fold feature cross-validation process for confirming Doppler mutation events, which process includes: extracting time-series Doppler frequency shift data from the observation data; performing at least four heterogeneous feature extraction analyses on the time-series Doppler frequency shift data in parallel; cross-validating the consistency of the features output by the four analyses in time and amplitude; when the consistency meets the preset criteria, it is determined to be a reliable mutation event, and by weighted fusion of the four sets of features, a fused mutation feature parameter set including the mutation moment, duration and intensity is generated.

[0100] Specifically, four types of heterogeneous feature extraction analysis include:

[0101] Feature recognition based on wavelet multi-scale decomposition. The analysis uses Daubechies-4 wavelet basis to analyze the Doppler frequency shift time series f D (t) is decomposed into 5 layers. By calculating the detail coefficient d of each layerj The energy distribution of the data is calculated and the energy mutation detection threshold is set (for example, the energy ratio within the sliding window exceeds 3 times) to identify the scale and time of the mutation. The specific starting and ending points of the mutation are t start_w and t end_w Positioning is performed on detail coefficients d3 and d4, and the mutation duration Δt is calculated w and mutation amplitude A w , and finally generate the wavelet domain mutation feature vector F wavelet .

[0102] Instantaneous feature extraction based on time-frequency joint analysis. This analysis is for f D (t) performs a short-time Fourier transform (STFT), where the window length can be set to 0.5 seconds and the overlap rate can be set to 75%, to generate a time-frequency spectrum S(t, f). The instantaneous frequency trajectory f is extracted by the ridge tracking algorithm. inst (t), and calculate its rate of change df inst / dt. When |df inst When / dt| exceeds the preset threshold (e.g. 5Hz / s), it is identified as a frequency jump point and the jump amplitude Δf is extracted jump and the jump slope k jump etc., generate the time-frequency domain mutation feature vector F timefreq .

[0103] Singularity quantification based on multi-resolution singularity detection. This analysis calculates the Doppler second-order derivative d 2 f D / dt 2 The continuous wavelet transform modulus maximum line is obtained and its evolution trajectory at different scales is tracked. The singularity strength of the signal is quantified by calculating the Lipschitz exponent α=log|W(a, t)| / log(a), where W(a, t) is the wavelet coefficient at scale a and time t. Points with α<0.5 are identified as strong singular points and used as the core position of the mutation t. singular , and extract the singularity peak α min and the width of the singular region w singular Equal features, generate singular eigenvector F singular .

[0104] Analysis of the time-varying characteristics of spectrum flatness based on reverse Doppler injection. This is a newly added verification dimension. In this analysis, a reverse Doppler sequence is generated at intervals of 0.1 seconds during the candidate mutation period and the signal is injected. By observing the time-varying characteristics of the spectrum flatness F(t), it is found that at the real mutation point, F(t) will show a step due to the drastic change in the statistical characteristics of the signal. By calculating the time derivative of the flatness |dF / dt|, when its value exceeds three times the standard deviation of the background fluctuation (i.e. |dF / dt|>3σ) F ), the presence of the mutation can be confirmed.

[0105] After obtaining four sets of features, the system performs cross-validation and fusion. Specifically, the temporal consistency C between the features is calculated. t and amplitude consistency C A When C t >0.8 and C A When the value is >0.7, it is determined to be a reliable mutation. For confirmed mutations, the estimation results of the four sets of features are fused by weighted averaging based on the signal-to-noise ratio of each method, and finally a high-confidence fused mutation feature parameter set is output, which reduces the false positive rate.

[0106] According to one aspect of the present application, the steps of generating parameters for iteratively optimizing the observation strategy include: after positioning solution and performance evaluation, obtaining the positioning quality score, and combining the final balance of the energy reserve unit of this task and the accuracy of the mutation point prediction to generate multi-dimensional task performance indicators; integrating multi-dimensional task performance indicators such as positioning quality score, energy utilization efficiency and prediction accuracy to calculate the reward function value for evaluating the comprehensive performance of this observation task; applying the reward function value as a feedback signal to update the internal parameters of the deep reinforcement learning network through the gradient backpropagation algorithm; the updated internal parameters of the deep reinforcement learning network constitute the optimization strategy parameter set for iteratively optimizing the observation strategy.

[0107] Specifically, after a transit observation mission is completed, the system first collects performance indicators from various sub-modules that can comprehensively reflect the execution effect of this mission, including at least: Positioning quality score Q: This score is output by the positioning solution and accuracy assessment module. It is a direct quantification of the quality of completion of the core goal of this mission: interference source positioning. It can be obtained by comprehensively calculating the size of the solved positioning error ellipse and GDOP value and other parameters. The higher the Q value, the more accurate the positioning; Energy utilization efficiency index: This indicator is directly represented by the final balance of the energy reserve unit in the energy pool mechanism. A higher final balance means that this mission consumes less energy while completing the established goals, reflecting higher energy utilization efficiency; Mutation point prediction accuracy index: This indicator is obtained by comparing the initial predicted elevation angle θ output by the mutation point prediction module pred The final mutation position θ obtained after real-time correction corr The smaller the deviation, the more accurate the prediction. The calculation of the comprehensive reward function value is to fuse the above multi-dimensional performance indicators into a single scalar value - the reward function value R. The reward function is a core concept in deep reinforcement learning, which provides a quantitative evaluation of the quality of each complete behavior (a complete transit observation) of the intelligent agent (in this case, the observation strategy decision network). A reasonable reward function design is the key to the success of deep reinforcement learning. In this embodiment, R can be designed as a weighted sum: R=wq Q+w e ·E final -w p ·A error ; Among them E final is the final balance of the energy reserve unit; A error is the absolute value of the error in the prediction of the mutation point; w q , w e , w p are weight coefficients for positioning quality, energy efficiency, and prediction accuracy, respectively. These weight coefficients can be set according to the overall priority of the task. For example, in a task with the highest accuracy as the primary goal, w q The weight of can be set relatively high.

[0108] The calculated reward value R serves as the core feedback signal, updating the deep reinforcement learning network pre-installed in the system. In this embodiment, the network uses the TD3 (Twin Delayed Deep Deterministic Policy Gradient) architecture, a deep reinforcement learning algorithm suitable for continuous action spaces. The TD3 network consists of an actor network and two critic networks. The actor network is responsible for outputting specific policies (for example, key parameters that determine the sampling strategy), while the critic network is responsible for evaluating the quality of the actor network's output policies. The reward value R is primarily used to update the critic network, enabling it to more accurately assess the value of state-action pairs. Through the gradient backpropagation algorithm, the actor network adjusts its parameters in a direction that leads the critic network to a higher evaluation value. This process essentially allows the actor to modify its performance based on the critic's feedback, hoping to receive a higher reward (reward value R) next time. After this update process, parameters such as weights and biases within the TD3 network are changed. These updated network parameters themselves constitute the optimized policy parameter set used to iteratively optimize the observation policy. At the beginning of the next new observation mission, the system will use this evolved TD3 network to make decisions, for example, adjusting the weights of the mutation point prediction model or optimizing the generation parameters of the sampling strategy.

[0109] Alternatively, in some implementations, other advanced deep reinforcement learning algorithms, such as SAC (Soft Actor-Critic) or PPO (Proximal Policy Optimization), can be used in place of TD3. The reward function R can also be more complex, for example by introducing nonlinear terms or taking into account other dimensions such as task completion time. This embodiment enables continuous self-evolution of the policy, thereby maintaining long-term efficiency and intelligence.

[0110] This embodiment, by constructing a multi-dimensional reward function encompassing positioning quality, energy efficiency, and prediction accuracy, and applying a deep reinforcement learning algorithm for closed-loop iteration, empowers the entire observation system with self-evolution capabilities. This addresses the fundamental issue of existing optimization frameworks, which are unable to learn and evolve strategies based on the final comprehensive mission results. Specifically, the final outcome of a complete transit observation mission—including the degree of completion of the core objective (positioning quality Q), the economic efficiency of resource consumption (final energy balance), and the accuracy of the process prediction (prediction error)—is integrated into a single, quantified reward value. This reward value serves as the final judgment on the entire strategic behavior and drives the update of the internal parameters of the deep reinforcement learning network through gradient backpropagation. This enables the system to learn from each success and failure, autonomously and continuously optimizing its internal decision-making model. This allows it to gradually adapt to the changing on-orbit environment without human intervention, maintaining long-term efficiency and intelligence.

[0111] According to one aspect of the present application, the step of obtaining enhanced Doppler observation data also includes an information enhancement process utilizing the multipath effect, which includes: using a subspace estimation algorithm to separate a direct path signal component and at least one reflected path signal component from a received signal; estimating the Doppler frequency shift and relative time delay of the direct path and at least one reflected path respectively to obtain a Doppler difference sequence and a relative time delay sequence; jointly processing the Doppler difference sequence and the relative time delay sequence, and utilizing the geometric constraints attached to the reflection path to construct an overdetermined set of equations; and by solving the overdetermined set of equations, obtaining a Doppler measurement value with enhanced parameter estimation accuracy as a component of the enhanced Doppler observation data.

[0112] Specifically, the presence of multipath effect is detected by analyzing the power spectrum density time series of the received signal. When the power spectrum shows multiple peaks, and the interval and intensity relationship of these peaks conform to the typical multipath propagation geometry model, it can be determined that there is a multipath effect. Once the presence of multipath is confirmed, the system will use a high-resolution subspace estimation algorithm, preferably an improved MUSIC (Multiple Signal Classification) algorithm, to separate the signal components. The algorithm can construct a signal subspace projection matrix to project the mixed received signal onto different subspaces, thereby separating the signal components from the direct path and the signal components from one or more reflected paths (for example, reflected by the earth's surface). After successfully separating the signal components, the system independently estimates the parameters of each component to obtain: the Doppler frequency shift f0 of the direct path; the Doppler frequency shift f of the i-th reflected path i ; Relative delay τ of the i-th reflected path relative to the direct path iBased on these estimates, the Doppler difference sequence Δf can be calculated i =f i -f0 and relative delay sequence τ i The reflection path is not random; it carries geometric information about the propagation environment. For example, for a path that reflects once from the Earth’s surface, its path length and arrival angle have strict geometric constraints on the position of the satellite, the location of the interference source, and the location of the reflection point on the Earth’s surface. This embodiment mathematically transforms this known geometric constraint relationship into a series of constraint equations. These geometric constraint equations are combined with the observation equation (i.e., Δf) obtained through parameter estimation. i and τ i Since the number of reflection paths (and therefore the number of observation equations) is usually greater than the number of unknown parameters to be solved (such as the correction value of the interference source position), this naturally constructs an overdetermined system of equations.

[0113] The overdetermined system of equations refers to a system of equations in which the number of equations is greater than the number of unknowns, and usually has no exact solution. For the constructed overdetermined system of equations, the weighted least squares method is used to solve it. The goal of the solution is to find a set of estimated values ​​for the unknown parameters so that the sum of the squared residuals of all equations is minimized. Among them, weighting means that different weights can be assigned to the corresponding equations according to the signal-to-noise ratio or strength of the signals of different reflection paths. Paths with high signal-to-noise ratios have a greater influence in the solution. By solving this overdetermined system of equations, Doppler measurements with enhanced parameter estimation accuracy can be obtained. This enhanced data integrates information from the direct path and all utilized reflection paths. Its accuracy and reliability, especially in low elevation angle areas where multipath effects are significant, will be better than the results obtained using only the direct path signal.

[0114] Alternatively, in some implementations, other subspace algorithms, such as the ESPRIT algorithm, may be employed. The geometric constraint model can also be expanded as needed, for example, to account for the curvature of the Earth or to introduce a more sophisticated ground reflection model. This embodiment expands the system's ability to acquire information under adverse observation conditions.

[0115] This embodiment improves the performance lower limit and data quality of the system under harsh observation conditions. On the one hand, for weak signal scenarios, the three-level cascade estimation strategy consisting of adaptive grid coarse estimation, reverse injection intermediate estimation and virtual beat frequency fine estimation, by gradually reducing the uncertainty range from coarse to fine, takes into account the computational efficiency while ensuring extremely high final accuracy, and solves the problem that traditional single FFT or phase-locked loop is prone to failure under low signal-to-noise ratio. On the other hand, for the multipath effect that is prevalent in low elevation angle periods, it is no longer regarded as pure interference, but is separated by a subspace algorithm, and its additional geometric constraint information is used to construct an overdetermined set of equations for solution. Converting additional propagation paths into additional sources of information improves the data accuracy in the period with the worst observation quality in traditional methods, and enhances the robustness of the entire system and the ability to obtain information in the entire arc segment.

[0116] According to one aspect of the present application, the step of obtaining enhanced Doppler observation data also includes an intelligent data splicing and reconstruction process, which includes: dividing the observation data into at least one stable data segment and one sudden transition data segment based on the identified mutation characteristics; evaluating the statistical characteristics of each data segment, including smoothness and rate of change, and adaptively assigning different interpolation strategies to data segments with different characteristics; constructing a fifth-order polynomial transition function at the junction of the stable data segment and the sudden transition data segment that can ensure the continuity of the Doppler second-order derivative; applying the transition function to smoothly splice the data segments using different interpolation strategies to reconstruct a continuous and smooth complete data sequence.

[0117] In this embodiment, the entire transit observation data sequence is segmented according to the fusion mutation feature parameter set. Specifically, the confirmed mutation center time t mut As the boundary, the data is divided into three main parts: the stationary section before the mutation [t start , t mut -Δt mut ]、Sudden transition section[t mut -Δt mut , t mut +Δt mut ], and the plateau after the mutation [t mut +Δt mut , t end ], where t start is the starting time point of the observation data sequence, Δt mut is the half width of the mutation window, t end is the end time point of the observed data sequence. For each data segment, the system will independently extract its statistical features, such as the mean μ i , variance σ i 2 , and the autocorrelation function ACF i(τ), thereby constructing the segmented feature description matrix F seg . Based on the segmented feature description matrix F seg , evaluate the characteristics of each data segment and tailor the most appropriate interpolation strategy for it. The interpolation strategy refers to the method of constructing new data points between discrete data points. Evaluation indicators can include: Smoothness index: S i =σ i 2 / μ i 2 ; Rate of change index: R i =max|df / dt|. The adaptive assignment rule can be: for the stationary data segment (S i and R i are all small, such as S i <0.1 and R i <1): Use cubic spline interpolation. This method can generate a globally smooth curve, which is very suitable for describing smooth and slowly changing signals. i Medium, for example, 0.1≤S i <0.5: Use piecewise cubic Hermite interpolation. The advantage of this method is that it can ensure the monotonicity of the interpolation curve and avoid unnecessary oscillations in monotonic data (i.e., Runge phenomenon). i Larger or R i Large): Select radial basis function (RBF) interpolation. RBF interpolation has strong flexibility and fitting capabilities for processing complex and irregularly distributed data points.

[0118] In order to avoid creases or sharp points at the intersection of data segments using different interpolation strategies, this embodiment constructs a transition function that ensures high continuity based on the boundary smoothing of fifth-order polynomials. Specifically, at each intersection point t boundary Set up very short transition zones before and after (for example, t boundary ±0.5s), and in this area a fifth-order polynomial transition function p(t)=a0+a1t+a2t is used 2 +a3t 3 +a4t 4 +a5t 5 The motivation for choosing a fifth-order polynomial is that it has six undetermined coefficients (a0 to a5), which allows it to simultaneously satisfy six boundary conditions: that is, at the start and end of the transition zone, its function value (p), first-order derivative value (p', representing velocity), and second-order derivative value (p'', representing acceleration) are all equal to the interpolation functions of the left and right segments. By solving these six simultaneous boundary conditions, the coefficients of the polynomial can be determined, thus constructing not only C 1 Continuous (tangent continuous) and C2 Continuous (curvature continuous) smooth splicing. This means that not only is there no inflection point at the splicing, but even the smoothness of the inflection is continuous.

[0119] After completing the interpolation and concatenation of all data segments, the reconstructed data sequence is complete. The system also performs several final checks: Continuity check: Calculates the differences between adjacent data points to ensure there are no jumps. Smoothness check: Calculates local curvature to ensure there are no sharp inflection points. Physical plausibility check: Ensures that the reconstructed Doppler shift value is within its maximum physically possible range. A comprehensive reconstruction quality indicator, Q, is calculated, and a data quality report is generated to provide a reliability reference for subsequent applications using the data.

[0120] Through the above steps, this embodiment can intelligently reconstruct incomplete and discontinuous raw data that may be generated due to various reasons (such as non-uniform sampling and signal interruption) into a high-quality, complete and smooth final data product.

[0121] The intelligent data splicing and reconstruction process can be further refined as follows: After the multi-scale data segmentation step based on mutation features, the segment feature description matrix F constructed by the system seg and the inter-segment relationship matrix R seg , which will be used as the direct input for the subsequent adaptive interpolation strategy selection. In the integrity verification and quality assessment steps, the comprehensive reconstruction quality index Q can be designed as a weighted sum, for example: Q = w1·C cont +w2·C smooth +w3·C physics . Among them C cont 、C smooth and C physics The Q values ​​represent the quantitative scores for continuity, smoothness, and physical plausibility, respectively. Weights w1, w2, and w3 can be configured based on task requirements, for example, 0.4, 0.3, and 0.3, respectively. Ultimately, the system outputs a high-quality reconstructed data sequence along with a data quality report. This report identifies the Q value and any potentially problematic time periods identified during the test for subsequent analysis.

[0122] In a specific embodiment, it is assumed that a high-confidence Doppler mutation window has been identified and the following input parameters are available: Mutation window: It is determined that the mutation window of this transit occurs at the satellite elevation angle [θ start ,θ end ]=[30°,45°]. Second-order derivative characteristics: According to calculation, in this window, the maximum value of the Doppler second-order derivative is d 2 f max =1.2Hz / s 2 , the average value is d2 f avg =0.1Hz / s 2 Boundary sampling interval: According to the task requirements, the target sampling interval Δt in the stable sparse area sparse =2s; at the boundary of the mutation window, the target sampling interval Δt in the dense area dense =0.2s. Calculate the ratio of mutation severity r: r=d 2 f max / d 2 f avg =1.2 / 0.1=12; Determine the coefficient k based on the r value: According to the piecewise function rule, since r=12>10, it indicates that the mutation is very drastic, so the smallest proportional coefficient k=0.1 is selected. At the point where the mutation is most drastic, |d 2 f / dt 2 |=d 2 f max =1.2Hz / s 2 The calculated sampling interval is Δt=k / |d 2 f / dt 2 |=0.1 / 1.2≈0.083s. In actual implementation, this value may be limited by the system's minimum sampling interval capability (e.g., 0.1s), so the actual sampling interval can be set to 0.1s. At a point in the mutation window where the change is relatively slow, assuming |d 2 f / dt 2 |=0.2Hz / s 2 The calculated sampling interval is Δt=0.1 / 0.2=0.5s. The boundary from the stable region to the mutation region, i.e. θ boundary =30°. Apply the S-type function for smooth connection, the formula is: Δt trans =Δt sparse +(Δt dense -Δt sparse ) / (1+exp(-a(θ-θ boundary ))); Substitute the parameters (assuming the steepness coefficient a=10): Δt trans =2+(0.2-2) / (1+exp(-10(θ-30))); calculate the sampling interval at a point just entering the transition zone, for example, when θ=30.1°: exp(-10(30.1-30))=exp(-1)≈0.368Δt trans=2+(-1.8) / (1+0.368)≈2-1.316=0.684s; It can be seen that when the elevation angle just crosses the 30° boundary by 0.1°, the sampling interval has rapidly but smoothly decreased from 2s in the stable area to about 0.684s, and will continue to decrease smoothly to the target interval of 0.2s in the dense area as θ increases further. The step-like mutation from 2s to 0.2s at 30° is avoided. By iterating the elevation angle or time of the entire transit arc and repeating the above calculation, a complete, steplessly variable non-uniform sampling time sequence {t i The system also verifies the sequence, for example, checking to ensure that the total number of sampling points within the [30°, 45°] window is no less than 20, and that the sampling interval in the stable region does not exceed 5 seconds, to ensure the effectiveness and completeness of the final plan. This example clearly demonstrates the transformation of abstract mathematical models and rules into a concrete, executable, and intelligent observation sampling plan.

[0123] According to one aspect of the present application, real-time energy monitoring and adaptive adjustment are also included. This step is implemented by the system continuously and frequently monitoring the voltage of the spacecraft's main power bus and the output current of the battery pack using dedicated sensors. The system has internally set safety thresholds for battery voltage. During normal operation, if the bus voltage drops rapidly within a short period of time, and the magnitude of the drop exceeds this preset safety threshold, the system determines that abnormally high power consumption or battery failure may have occurred, and immediately triggers emergency energy management mode. In this emergency mode, the system's decision-making logic shifts from maximizing efficiency to ensuring survival and core missions. The system immediately executes pre-set power reduction plans, such as temporarily shutting down or degrading non-critical scientific payloads, and prioritizes power supply to the currently ongoing, highest-value observation mission (e.g., intensive sampling within the sudden change window). Simultaneously, the system dynamically updates an executable sampling schedule based on current real-time power consumption and remaining battery capacity. This schedule may reduce some subsequent observation points, but its core goal is to ensure that at least 80% of the core portion of the current high-value observation mission is completed. This mechanism provides ultimate safety assurance for the entire observation mission.

[0124] According to one aspect of the present application, the original GNSS observations contain a variety of error sources (such as ionospheric delay, multipath effect, cycle slip, gross error, etc.), which must be eliminated or suppressed through a systematic processing process. Therefore, before extracting the Doppler shift data, a multi-source heterogeneous filtering and fusion step based on the dual-frequency GNSS observations is included. Specifically, the dual-frequency pseudorange observations (ρ L1 , ρ L2 ), dual-frequency carrier phase observation (Φ L1, Φ L2 ) and the signal-to-noise ratio sequence SNR[i] of each channel. Based on these raw data, the quality index of each observation is calculated. For example, by combining pseudorange and carrier phase, the pseudorange multipath index MP=|ρ L1 -ρ L2 -2.546(Φ L1 -Φ L2 )|, the size of this value can reflect the degree of influence of multipath error. Based on the signal-to-noise ratio and multipath index, the initial weight is calculated for each observation, such as the pseudorange weight w ρ =SNR 2 / (SNR 2 +MP 2 ), carrier phase weight w Φ =exp(-σ Φ 2 / σ ref 2 ), where σ Φ is the standard deviation of the carrier phase noise corresponding to the current observation segment or a certain observation point, σ ref is the ideal carrier phase noise level used as a standard reference. All these weights are finally combined into the observation weight matrix W obs , used for the subsequent filtering process.

[0125] Construct a state vector X containing satellite position (x, y, z), velocity (vx, vy, vz), receiver clock error and clock drift. Build a state transfer matrix based on the satellite's orbital dynamics model. In the Kalman filter update step, read W obs The system dynamically adjusts the observation noise covariance matrix R, thereby assigning different degrees of confidence to observations of different qualities. Furthermore, the system uses a fading memory factor λ to dynamically adjust the process noise Q, enabling the filter to adaptively compensate for satellite maneuvers or model errors, thus achieving adaptive Kalman filter state estimation.

[0126] In order to eliminate the largest error source of ionospheric delay, the system uses dual-frequency pseudorange and carrier phase to construct the ionospheric-free combined observation quantity P IF and L IF At the same time, in order to detect the possible cycle slips (i.e., the sudden jump of the whole cycle ambiguity) in the carrier phase observation, a geometrically independent combined observation GF=λ1Φ is constructed. L1 -λ2Φ L2 , where λ1 is the carrier wavelength of the L1 band, and λ2 is the carrier wavelength of the L2 band. By performing sliding window detection on the GF sequence, we can effectively identify the time when a cycle slip occurs and generate a cycle slip identification sequence for subsequent processing.

[0127] After Kalman filtering, in order to further eliminate the possible gross errors in the data (i.e., outliers far away from the normal data group), the system adopts the robust estimation method. Specifically, the standardized residual v of each observation is calculated. std . Using the IGG-III robustness weight function, according to v std The size of |v recalculate the weight for each observation. For example, when |v std When | is very small (such as less than 1.5), the weight is 1; when |v std When | is very large (such as greater than 2.5), the weight is directly set to 0; in the middle area, the weight is based on |v std The size of | is smoothly transitioned. The residual calculation and weight update are performed iteratively until the weight converges. Finally, those observations with a weight of 0 are regarded as gross errors and eliminated from the solution, thus outputting a robust positioning solution X with high accuracy and reliability, eliminating various errors. robust And the corresponding positioning precision dilution DOP.

[0128] According to one aspect of the present application, the step of obtaining enhanced Doppler observation data may further include a joint reconstruction process based on multi-resolution analysis. The specific implementation of this process is as follows: obtaining downsampled or non-uniformly sampled observation data, including global data y with a low sampling rate L and possible high-sampling rate local data segments y H . Although the signal itself is very complex, it is sparse in a certain transform domain, that is, it can be represented by very few non-zero coefficients. In order to find the most suitable transform domain, an over-complete dictionary D is constructed. The over-complete dictionary is a matrix containing a variety of different basis functions. For example, it also contains the discrete cosine transform (DCT) basis for representing the stationary component, the Gabor atoms for representing the oscillating component, and a variety of wavelet bases for representing transients and singular points. The greedy algorithm of orthogonal matching pursuit (OMP) is used to solve the sparse coefficient vector α. Its goal is to find the sparsest α (that is, the one with the least non-zero items) so that the signal y can be approximately represented as Dα.

[0129] Perform compressed sensing measurement matrix optimization. The measurement matrix Φ mathematically describes the process from the original high-resolution signal to the actual low-resolution or non-uniformly sampled data. In order to ensure that the original signal can be successfully reconstructed from the undersampled data, the measurement matrix Φ must meet the key condition of the Restricted Isometry Property (RIP). To achieve this, the system can first initialize Φ with a random Gaussian matrix, and then iteratively optimize it using a gradient descent method through a cross-correlation minimization algorithm. The goal of the optimization is to make Φ'Φ (where Φ' is the transpose of Φ) as close to the unit matrix as possible, thereby ensuring that the measurement matrix Φ has good RIP characteristics, for example, its RIP constant is less than 0.3.

[0130] After constructing the overcomplete dictionary D and the optimized measurement matrix Φ, the signal reconstruction problem can be formulated as a convex optimization problem, whose mathematical form is: min||α||1subject to||y-ΦDα|| 2 ≤ε. The meaning of this formula is: among all sparse solutions α that satisfy the requirement that the error between the reconstructed signal and the actual observation value y after measurement does not exceed the noise level ε, find the solution with the smallest L1 norm ||α||1. This problem can be solved using an efficient algorithm called the alternating direction method of multipliers (ADMM). ADMM decomposes the original problem into several more easily solvable sub-problems by introducing auxiliary variables and augmented Lagrangian functions, and iterates and updates alternately until the convergence criterion (for example, the ratio of the norm of the residual to the solution is less than 10) is achieved. -4 ) is satisfied. Finally, the optimal sparse coefficient α is obtained final , the reconstructed signal is x rec =Dα final .

[0131] In order to further improve the quality of the reconstructed signal, especially when processing signals containing both slowly varying trends and high-frequency details, the system performs rec Perform multi-resolution fusion. Specifically, x rec Decomposed into low-frequency components x low (e.g. frequency below 10 Hz) and high frequency components x high (Frequency is higher than 10Hz). For low frequency component x low , bicubic interpolation and other smoothing methods can be used to further improve its resolution and visual effect. high , then its original high-resolution features are maintained. In the overlapping area of ​​data processing at different resolutions, a weighted fusion method is used for smooth transition, for example, using a cosine window function for weighting to ensure the phase continuity of the final fusion signal. The final output high-resolution reconstructed signal x final, its quality can be comprehensively evaluated by multiple indicators such as mean square error (MSE), peak signal-to-noise ratio (PSNR) and structural similarity (SSIM).

[0132] According to one aspect of the present application, the original observed Doppler shift is the result of the superposition of two motion components: the motion of the satellite relative to the Earth, and the possible motion of the interference source itself. Conventional positioning algorithms typically assume that the interference source is stationary, so its motion can become a major source of error. Therefore, before performing positioning, a predictive Doppler compensation step can be included. Specifically, this involves utilizing the estimated motion parameters of the interference source output by a state estimation module (e.g., one based on an extended Kalman filter). These parameters include the best estimate of the interference source's current position, velocity, and even acceleration. Based on these motion state parameters, a forward prediction is performed using kinematic equations to predict the interference source's trajectory over a short period of time (e.g., 5 to 10 seconds) in the future. Based on this predicted trajectory, the expected Doppler shift change caused solely by the interference source's motion during this period is accurately calculated. The calculated expected Doppler shift change caused by the interference source's motion is subtracted from the original, mixed observation data. This process is known as forward compensation. After compensation, the resulting data sequence ideally contains only the Doppler shift component caused by satellite motion—that is, compensated, pure Doppler data. This predictively compensated pure Doppler data is then input into the subsequent iterative positioning algorithm for solution. Because the primary error sources in the input data have been preemptively eliminated, the convergence speed of the positioning algorithm and the ultimate positioning accuracy are significantly improved. Optionally, in some embodiments, the length of the prediction time window can be adaptively adjusted based on the intensity of the interference source's motion. The more intense the motion, the shorter the prediction window to ensure prediction accuracy.

[0133] This invention systematically addresses two core technical bottlenecks by introducing a forward-looking mutation event prediction mechanism and semantically deep state perception capabilities. First, addressing the problem of observation strategies lacking foresight and leading to blind resource allocation, two prediction methods empower the system with the ability to foresee future critical events. The first is a dual-loop prediction-correction mechanism. By integrating prior statistics from a historical knowledge base with current evidence from real-time observations, it generates high-probability prediction windows before the mutation window arrives. The second is a texture fingerprint recognition method that utilizes a convolutional neural network to learn and identify deep patterns from the signal's time-spectrogram that indicate impending mutations. This is a critical prerequisite for all subsequent resource optimization. Based on this prediction, an energy pool mechanism proactively degrades operation during stable periods of low observation efficiency to accumulate energy. This energy is then optimized according to an energy-accuracy benefit function and concentrated on upcoming high-value mutation windows. Furthermore, a non-uniform sampling strategy based on an adaptive inverse proportional function precisely allocates sampling resources to the periods of highest information density based on these predictions. This transforms the traditional static, passive resource allocation model into a dynamic, forward-looking, intelligent planning model guided by future critical events.

[0134] Secondly, to address the lack of semantic depth in state perception, which leads to unclear optimization objectives, a deep state perception system was constructed that understands the intrinsic physical characteristics of Doppler signals. Rather than relying solely on macroscopic indicators such as signal strength, this system delves into the dynamic nature of the signal, integrating high-order physical quantities that directly characterize kinematic properties, such as Doppler second-order derivatives, texture features of the time-spectrogram, and the strength of the signal's singularity, as the core of state perception. For example, through cross-validation of four heterogeneous features, the system quantifies and confirms the physical characteristics of sudden events from different dimensions, providing a clear, robust, and physically meaningful target for optimization decisions. Importantly, the deep reinforcement learning closed-loop optimization framework employed features a reward function that incorporates not only the final localization result but also an evaluation of the accuracy of the prediction process. This incentivizes the intelligent agent (observation policy network) to continuously learn to gain a deeper understanding of these physical characteristics, thereby achieving continuous self-evolution of the semantic depth of state perception.

[0135] The preferred embodiments of the present invention are described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the scope of protection of the present invention.

Claims

1. A Doppler observation strategy optimization method for satellite-borne interference sources, characterized in that: include: Obtain the original RF signal of the onboard interference source and the precise satellite orbit data, predict the mutation characteristics in the Doppler data, and generate an adaptive observation plan; According to the adaptive observation scheme, energy-aware resource scheduling is performed and observations are performed to obtain enhanced Doppler observation data; Positioning solution and performance evaluation are performed based on the enhanced Doppler observation data, and parameters for iterative optimization of the observation strategy are generated.

2. The method according to claim 1, characterized in that Obtain enhanced Doppler observation data, including: Identify energy accumulation windows defined by low elevation angle periods in adaptive observing schemes, and high-value observing periods defined by Doppler mutation characteristics; In the energy accumulation window, by reducing the working mode of the receiver, the energy saved is accumulated and counted into the energy reserve unit; When entering a high-value observation period, evaluate the matching degree between the balance of the energy storage unit and the power consumption demand of that period; Under the condition that the balance of the energy reserve unit meets the power consumption requirement, high-density sampling is authorized to obtain enhanced Doppler observation data.

3. The method according to claim 2, characterized in that Evaluate the matching degree between the balance of the energy reserve unit and the power consumption demand during the period, including: Construct an energy-accuracy benefit function to characterize the relationship between observation accuracy and energy consumption; Based on the energy-accuracy benefit function, a segmented benefit evaluation matrix is ​​generated for different power consumption modes during high-value observation periods. Establish a multi-constraint optimization model with the goal of maximizing total benefits and integrating energy storage unit balance and mode switching cost as constraints; The multi-constraint optimization model is solved to obtain the optimal power allocation scheme, and high-density sampling is performed accordingly.

4. The method according to claim 2, characterized in that Also includes: When the assessment determines that the balance of the energy storage unit cannot meet the power consumption demand, the degradation strategy is initiated; The degradation strategy calculates the ratio of the current remaining energy to the expected demand of the subsequent task and compares it with at least one preset degradation threshold; If the ratio is lower than the degradation threshold, a downgraded sampling plan is generated to reduce the total energy consumption. The downgraded sampling plan is achieved by increasing the sampling interval of non-critical observation areas or shrinking the spatial range of high-density sampling. Observations are performed according to a downsampling plan.

5. The method according to claim 1, wherein When the original RF signal is lower than the preset threshold, the process of obtaining enhanced Doppler observation data is achieved by using a step-by-step Doppler frequency cascade estimation method, specifically: The frequency grid is adaptively adjusted based on the spectrum energy distribution to perform preliminary spectrum peak location and obtain a rough Doppler estimate. Based on the coarse Doppler estimate, the intermediate Doppler estimate is obtained by injecting the reverse Doppler sequence and searching for the solution that minimizes the spectral flatness of the compensated signal. Based on the Doppler mean estimation, a virtual mirror signal is constructed and iteratively adjusted to minimize the time rate of change of the beat frequency component generated by its cross-correlation with the original signal, and the Doppler precise estimation is obtained. Based on this, the enhanced Doppler observation data is sampled.

6. The method according to claim 5, characterized in that By constructing a virtual mirror signal and iteratively adjusting it, a precise Doppler estimation is obtained, including: Using the conjugate of the original RF signal and taking the Doppler mean value as the frequency basis, a virtual mirror signal is constructed; Calculate the cross-correlation function between the original RF signal and the virtual mirror signal, and separate the beat frequency component therefrom; The frequency used to construct the virtual mirror signal is iteratively adjusted until the time rate of change of the beat frequency component converges to a preset minimum value, and the converged frequency is determined as the Doppler precise estimate.

7. The method according to claim 5, characterized in that When acquiring enhanced Doppler observation data, when sampling is performed in a non-uniform manner, a data reconstruction process is also included, specifically: Obtaining a precise Doppler estimate and using it to pre-compensate the non-uniformly sampled observation data to obtain a compensated stationary signal; For the compensated stationary signal, solve the pre-built compressed sensing convex optimization model to obtain the reconstructed stationary signal component; The frequency trend represented by the Doppler precise estimate is re-assigned to the reconstructed stationary signal component to generate enhanced Doppler observation data.

8. The method according to claim 1, characterized in that Predicting mutation signatures in Doppler data, including: Perform spectrum analysis on the original RF signal and, combined with the satellite's precise orbit data, extract the time-series Doppler shift data through a frequency tracking loop; Perform time-frequency transformation on the time series Doppler frequency shift data to generate a time-frequency spectrum and process it into a grayscale image; Extract local binary patterns and gray-level co-occurrence matrices from grayscale images as texture features and construct multi-dimensional texture feature vectors; The multidimensional texture feature vector is fed into a pre-trained convolutional neural network to parse and output the mutation type and mutation time in the Doppler data.

9. The method according to claim 1, characterized in that Predicting mutation features in Doppler data can also be: Extract the time-series Doppler frequency shift data from the original RF signal and calculate the corresponding Doppler second-order derivative sequence; The historical transit mutation point knowledge base is called up and combined with the current satellite precise orbit data to generate the initial predicted mutation position; Real-time monitoring of the Doppler second-order derivative sequence. When the measured characteristics of the sequence deviate from the expected characteristics of the initially predicted mutation position and exceed a preset threshold, a correction mechanism is triggered. The correction mechanism fuses the initial prediction with the measured features and outputs a corrected mutation window.

10. The method according to claim 9, characterized in that Correction mechanisms include: A Kalman filter is used to fuse the initial predicted mutation position with the measured features extracted from the Doppler second-order derivative sequence; The Kalman filter outputs an optimized corrected mutation position and a confidence weight coefficient representing the credibility of the corrected position; The boundary of the modified mutation window is adaptively set according to the modified mutation position and the confidence weight coefficient.

Images