Satellite link dynamic regulation and control device and method

Abstract

The invention relates to a satellite link dynamic regulation and control device and method. The device comprises an upper-layer strategy scheduler, a middle-layer predictor and a lower-layer real-time controller. And the upper-layer strategy scheduler obtains a link state prediction result output by the middle-layer predictor and a current link state reported by the lower-layer real-time controller, selects an initial strategy, a dynamic regulation and control strategy or a conventional communication recovery strategy according to task requirements, and issues a link parameter configuration instruction. And a middle-layer predictor performs link state prediction by using a pre-trained random forest model based on an input feature set extracted in a historical sliding window. And the lower-layer real-time controller executes transmission power and modulation coding scheme adjustment according to the instruction, executes emergency regulation and control when the link margin is abnormal, and gradually adjusts to a conventional communication mode after the link is recovered. According to the invention, dynamic prediction and rapid adaptive regulation and control of the satellite link can be realized.

Description

Technical Field

[0001] This application relates to the field of satellite communication, and particularly to a satellite link dynamic regulation device and method. Background Art

[0002] With the rapid development of satellite communication technology, low-orbit satellite systems have been widely used in fields such as long-distance communication, data backhaul, and emergency communication. The stability and reliability of satellite communication links are crucial for the overall performance of the communication system. However, satellite links are affected by various factors, including satellite trajectory changes, terminal position variations, communication environment interference, and atmospheric condition fluctuations, and the link signal quality exhibits obvious dynamic change characteristics.

[0003] Existing satellite communication link management methods usually rely on fixed configurations or preset link parameters, such as fixed transmit power or modulation and coding schemes, and trigger adjustments through simple thresholds when the link state changes. This method has a lag in response, cannot predictively regulate according to the link quality trend in a timely manner, and is prone to taking remedial measures only after the link deteriorates, resulting in an increased risk of communication interruption. Summary of the Invention

[0004] In order to be able to combine the real-time link state with trend prediction, adjust the link configuration in advance, and quickly respond to emergencies, this application provides a satellite link dynamic regulation device and method.

[0005] In a first aspect, a satellite link dynamic regulation device provided by this application adopts the following technical solution: A satellite link dynamic regulation device includes an upper-layer policy scheduler, a middle-layer predictor, and a lower-layer real-time controller; The upper-layer policy scheduler is used to obtain the link state prediction result output by the middle-layer predictor and the current link state information reported by the lower-layer real-time controller, select and execute a link regulation policy based on task requirements, where the link regulation policy includes an initial policy, a dynamic regulation policy, and a restore normal communication policy, and issue a link parameter configuration instruction to the lower-layer real-time controller; The middle-layer predictor is used to extract an input feature set within a set historical sliding time window based on the link real-time data collected by the lower-layer real-time controller, and input the input feature set into a pre-trained random forest model for link state prediction, where the pre-trained random forest model is constructed based on historical link operation data and trained with the classification label of whether to trigger a link adjustment instruction as the training target; The lower-layer real-time controller is used to perform transmit power adjustment and / or modulation and coding scheme adjustment according to the link parameter configuration instructions issued by the upper-layer policy scheduler, and perform emergency power boost and modulation degradation operations to stabilize link communication when it detects that the link signal-to-noise ratio is lower than the burst threshold within the preset burst response duration. At the same time, it feeds back the emergency status information to the upper-layer policy scheduler to perform progressive power recovery and modulation mode improvement after receiving the recovery configuration instructions issued by the upper-layer policy scheduler, and complete the fallback of the link to the normal communication mode.

[0006] By adopting the above technical solutions, the present invention realizes dynamic prediction and hierarchical regulation of satellite communication links. Specifically, the upper-layer policy scheduler flexibly selects the initial policy, dynamic regulation policy or recovery to normal communication policy at different stages based on the link state prediction results provided by the middle-layer predictor and the link states reported in real time by the lower-layer real-time controller, in combination with task requirements, to ensure that the link configuration can dynamically adapt to changes in the link environment. The middle-layer predictor predicts the future margin change trend of the link based on the link statistical features extracted from the historical sliding window and using the pre-trained random forest model, so as to achieve pre-decision before the link degradation shows obvious anomalies, and avoid link interruption caused by passive response. The lower-layer real-time controller not only performs regular adjustment according to the upper-layer instructions, but also immediately performs emergency power boost and modulation degradation operations when it detects a sharp drop in the link margin, ensuring that the communication link can still maintain the basic transmission capacity under harsh conditions. After the link is restored, through the progressive power and modulation recovery strategy, it effectively prevents the link oscillation problem caused by too fast recovery.

[0007] Through the above mechanism, the present invention can effectively reduce the link interruption rate and increase the stable communication duration of the link under different channel conditions in the environment where the satellite is moving at high speed and the channel changes frequently. At the same time, by introducing a random forest model trained based on real historical data in the link prediction stage, it can improve the recognition accuracy of link trend changes, reduce unnecessary adjustment actions, and avoid waste of link resources. After adopting the progressive recovery process, it can smoothly return to the high-efficiency communication configuration after the link signal improves, and avoid link instability and repeated degradation caused by too fast power or modulation recovery. The introduction of the emergency adjustment mechanism enables the link to be protected from interruption within milliseconds even in the case of sudden strong interference or rapid occlusion, ensuring the communication continuity and service availability of remote base stations or terminals.

[0008] Optionally, the upper-layer policy scheduler includes: A status receiving unit, configured to receive the link state prediction results output by the middle-layer predictor and receive the current link operation status information reported by the lower-layer real-time controller; A strategy selection unit, configured to select a current link regulation strategy to be executed from an initial strategy, a dynamic regulation strategy, and a recovery normal communication strategy according to task requirements and comprehensive link status information; A decision generation unit, configured to generate a transmit power configuration instruction and / or a modulation and coding scheme configuration instruction based on the selected link regulation strategy; A configuration distribution unit, configured to send the transmit power configuration instruction and / or the modulation and coding scheme configuration instruction to a lower-layer real-time controller to implement the execution of link regulation instructions; A strategy update unit, configured to enter a rapid decision-making loop after receiving an emergency status notification from the lower-layer real-time controller, and re-evaluate and update the executed link regulation strategy based on real-time status.

[0009] By adopting the above technical solution, the upper-layer policy scheduler can comprehensively judge the link risk situation based on the trend prediction provided by the middle-layer predictor and the current link status reported by the lower-layer real-time controller. By dynamically switching the initial strategy, the dynamic regulation strategy, or the recovery strategy through the strategy selection unit, the link configuration can be adjusted in advance, and the probability of link interruption can be reduced. The decision generation unit and the configuration distribution unit cooperate to ensure the rapid and accurate transmission of link parameter instructions to the execution layer. The strategy update unit re-evaluates the status in a timely manner in case of emergencies, shortens the time from anomaly detection to regulation effectiveness, and ensures that the link still has communication capabilities in a rapidly deteriorating environment.

[0010] Optionally, the middle-layer predictor includes: A feature extraction unit, configured to extract an input feature set based on received link real-time data within a set historical sliding time window, where the input feature set covers statistical features of non-high-frequency and fast-fluctuating data, including received signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, and satellite elevation angle; A model inference unit, configured to input the input feature set into a pre-trained random forest model and output a link status prediction result, including a predicted value of link margin or a link failure risk level within a future time period; A confidence evaluation unit, configured to evaluate the confidence of the output result of the model inference unit and send the prediction result and the confidence index to the upper-layer policy scheduler together to assist in strategy selection; A prediction update interface unit, configured to provide a training data collection and offline model update interface to support the periodic training and optimization of the random forest model based on newly collected link operation data.

[0011] By adopting the above technical solutions, the middle-layer predictor can extract statistical features reflecting the link trend within a set time window, avoid high-frequency noise interference, and improve the stability of feature input. By feeding the input feature set into a pre-trained random forest model, it is possible to achieve a forward-looking prediction of the future margin state of the link. The confidence evaluation unit judges the quality of the prediction results, reducing the risk of incorrect regulation caused by low-confidence predictions. The prediction update interface unit optimizes the model parameters by introducing new link data, improving the prediction accuracy under different satellite transit conditions, and supporting the adaptability to the dynamic changes of the link in long-term applications.

[0012] Optionally, the lower-layer real-time controller includes: A link monitoring unit, configured to collect link signal status data at a preset time interval, including signal-to-noise ratio, bit error rate, and frame loss rate, and report the data to the middle-layer predictor and the upper-layer policy scheduler; An instruction execution unit, configured to control the transmitter to perform corresponding parameter adjustments according to the transmit power configuration instruction and / or modulation and coding scheme configuration instruction issued by the upper-layer policy scheduler; An emergency event detection unit, configured to determine whether the link signal-to-noise ratio drops below the emergency threshold within the emergency response duration, and trigger an emergency adjustment when the condition is met; An emergency adjustment unit, configured to immediately increase the transmit power to the upper limit when detecting a link anomaly, switch to a low-order modulation and coding scheme to stabilize the link communication, and feedback the emergency status information to the upper-layer policy scheduler; A progressive recovery control unit, configured to gradually reduce the transmit power and restore the high-order modulation scheme at a set rhythm after receiving the recovery configuration instruction, and fall back to the normal communication mode on the premise of ensuring link stability.

[0013] By adopting the above technical solutions, the lower-layer real-time controller can periodically collect signal status data through the link monitoring unit to form a real-time feedback link of the link status. The instruction execution unit can accurately adjust the transmit power or modulation method according to the upper-layer instruction to ensure that the link parameter configuration takes effect quickly. The emergency event detection unit can judge whether the signal-to-noise ratio drops below the threshold within milliseconds and trigger an emergency adjustment in time to prevent the link from disconnecting. The emergency adjustment unit ensures communication connectivity under extreme conditions by quickly increasing the transmit power and reducing the modulation level. The progressive recovery control unit reduces the transmit power and increases the modulation rate in stages after the link is stable, avoiding new link instability problems caused by too fast adjustment during the recovery process.

[0014] In a second aspect, a satellite link dynamic regulation method provided by the present application adopts the following technical solutions: A satellite link dynamic regulation method includes the following steps: S1. When a satellite transit starts or a link is established, the upper-layer policy scheduler sets an initial policy according to the task requirements, and sends the initial transmit power configuration, the initial modulation and coding scheme configuration, and the initial margin threshold parameter to the lower-layer real-time controller. The lower-layer real-time controller sets the initial transmit power and the initial modulation mode according to the initial configuration and starts link monitoring. Among them, the execution policies of the upper-layer policy scheduler include the initial policy, the dynamic regulation policy, and the recovery of normal communication policy; S2. The lower-layer real-time controller continuously collects real-time link data and organizes it into an input feature set. The middle-layer predictor continuously performs link state prediction based on the input feature set. The upper-layer policy scheduler obtains the link state prediction result output by the middle-layer predictor and the current link state reported by the lower-layer real-time controller at a set decision cycle. When the predicted link margin value is lower than the preset safety threshold, the upper-layer policy scheduler pre-adjusts the link configuration policy; S3. The upper-layer policy scheduler sends an adjustment instruction to the lower-layer real-time controller according to the pre-adjusted link configuration policy to adjust the transmit power configuration and / or the modulation and coding scheme configuration; S4. After the lower-layer real-time controller adjusts the link parameters according to the adjustment instruction sent by the upper-layer policy scheduler, it performs fine power adjustment based on the current link measurement results to callback the actual signal-to-noise ratio to near the target margin, and sets the preset threshold for emergency event detection; S5. The lower-layer real-time controller detects whether the signal-to-noise ratio is lower than the preset threshold within the burst duration in the link monitoring loop. If it is lower, it raises the transmit power to the upper limit value and switches to a lower-order modulation and coding scheme to stabilize the link communication; S6. After the lower-layer real-time controller completes the emergency response, it sends an emergency status notification to the upper-layer policy scheduler. After receiving the emergency notification, the upper-layer policy scheduler immediately enters the decision-making loop to update the link regulation policy; S7. During the period when the upper-layer policy scheduler updates the policy, the lower-layer real-time controller continuously maintains the link communication with a preset high-power and low-rate configuration. After the link signal quality improves, it gradually reduces the transmit power. At the same time, if the upper-layer policy scheduler instructs to maintain the high-reliability configuration, the lower-layer real-time controller stops the power recovery action according to the upper-layer instruction to give priority to ensuring the link stability; S8. When the link signal quality significantly rebounds and stabilizes, the middle-layer predictor re-predicts the link state based on the new input features. The upper-layer policy scheduler makes a decision to restore the normal policy according to the middle-layer prediction result and the link state, and sends an instruction to the lower-layer real-time controller to reduce the transmit power and restore the original modulation and coding scheme; S9. After receiving the recovery instruction sent by the upper-layer policy scheduler, the lower-layer real-time controller performs a progressive recovery action to gradually reduce the transmit power and increase the modulation and coding rate until the link is restored to the normal communication mode.

[0015] By adopting the above technical solutions, the present invention can, during satellite transit or at the initial stage of link establishment, set initial link parameters through the upper-layer policy scheduler to ensure that the link has sufficient startup margin. Based on the link characteristics extracted within the sliding window, the middle-layer predictor uses a random forest model to predict the future margin change trend of the link, enabling the upper-layer scheduler to identify potential risks in advance and actively adjust the transmit power or modulation mode before the link degrades, reducing the probability of link interruption. After receiving the adjustment instruction, the lower-layer real-time controller performs fine power adjustment and, when there is a sudden change in the link signal, through the emergency event detection and emergency adjustment mechanism, increases the transmit power by milliseconds and reduces the modulation level to effectively prevent link failure. After receiving the emergency notice, the upper-layer scheduler quickly enters the policy update process to avoid communication loss caused by the inability of static configuration to adapt to environmental changes. During the link quality improvement stage, the lower-layer real-time controller can autonomously perform power reduction to restore communication efficiency and stop the restoration action when receiving the instruction from the upper layer to maintain a high-reliability configuration, avoiding new link oscillations caused by blind switching. Finally, through progressive restoration, the transmit power and modulation rate smoothly transition to the normal communication mode, while ensuring link stability, restoring system energy efficiency, and extending the effective communication duration during satellite transit. This solution overall realizes the predictive control of link status, real-time emergency processing, and recovery rhythm control, dynamically matches the optimal configuration according to the link status at different stages, and improves the communication availability and efficiency of satellite links in complex environments.

[0016] Optionally, S2 includes the following sub-steps: S201. The lower-layer real-time controller continuously collects real-time link data at a preset sampling interval, organizes and statistics the real-time link data according to the set sliding time window, and obtains an input feature set including indicators such as link signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, and satellite elevation angle. Among them, the length of the sliding time window is 1 second to 5 seconds, the step size is 0.5 second, and the input feature set does not include high-frequency fast-fluctuating data that requires millisecond-level response. S202. The middle-layer predictor inputs the input feature set into a pre-trained random forest model. The random forest model is trained based on historical link operation data and uses whether a link adjustment instruction is triggered as a classification label to perform link status prediction and output a prediction result of the link margin status within a preset future time period. S203. The middle-layer predictor calculates the link failure risk level and / or link safety margin according to the prediction result, and sends the prediction result and confidence index to the upper-layer policy scheduler together for assisting in the selection and update of link control strategies.

[0017] By adopting the above technical solutions, the lower-layer real-time controller organizes the link real-time data in a sliding window manner, avoiding instantaneous fluctuation interference and ensuring that the input features reflect the true trend of the link. Based on the stable input features, the middle-layer predictor uses the random forest model trained with historical data to classify and predict the future link margin state, realizing the early identification of link risks. By calculating the link failure risk level or safety margin and combining the confidence level to output to the upper-layer policy scheduler, the link regulation decision can actively intervene before the link deteriorates, reducing the probability of link interruption and unnecessary frequent adjustments, and improving the accuracy and timeliness of link configuration adjustment.

[0018] Optionally, the S2 includes the following sub-steps: S211. The upper-layer policy scheduler receives the link state prediction result output by the middle-layer predictor and the current link state information reported by the lower-layer real-time controller. The link state prediction result includes the link margin change trend within a preset future time period, and the current link state information includes the real-time signal-to-noise ratio, bit error rate, frame loss rate, and transmit power level. S212. The upper-layer policy scheduler performs trend analysis on the link state prediction result to judge whether there is a risk of insufficient future link margin and the degree of link quality degradation. S213. The upper-layer policy scheduler performs real-time evaluation on the current link state information to judge whether the current link is close to the failure boundary, including whether there is a situation where the signal-to-noise ratio is close to the lowest decoding threshold or the bit error rate rises sharply. S214. The upper-layer policy scheduler makes a joint determination of the trend analysis result and the real-time evaluation result. If the prediction result and / or the current state indicates that the link margin is lower than the preset safety threshold, a link pre-adjustment configuration policy is generated to determine whether to adjust the transmit power and / or switch the modulation and coding scheme. S215. The upper-layer policy scheduler formulates the adjustment amplitude and adjustment priority based on the pre-adjustment configuration policy and prepares to send the corresponding link parameter configuration instruction to the lower-layer real-time controller.

[0019] By adopting the above technical solutions, the upper-layer policy scheduler can synchronously obtain the future link margin change trend and the current real-time state of the link, realizing the forward-looking and immediate identification of link risks. By trend analysis to judge future risks and real-time evaluation to confirm whether it is close to failure currently, a pre-adjustment configuration policy can be generated before the link degradation is significant. Through joint determination and adjustment priority setting, the adjustment amplitude can be reasonably allocated according to the link change rate, avoiding premature or late adjustment, reducing the communication interruption rate, and at the same time reducing the link fluctuation caused by frequent switching, and improving the link continuous communication duration and regulation efficiency.

[0020] Optionally, the pre-training step of the random forest model includes: Construct a training data set based on historical link operation data collected over multiple satellite transit cycles. The historical link operation data includes link signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, and satellite elevation angle, and extract features from the data according to a set time window to form an input feature vector; According to the link regulation records corresponding to the historical link data, mark whether a link parameter adjustment instruction has been triggered as a classification label. Among them, the link parameter adjustment instruction includes a transmit power adjustment instruction and a modulation and coding scheme switching instruction; Use the input feature vector and the classification label to construct a supervised learning sample, and train a random forest classification model so that the model can output a prediction result on whether there is a regulation requirement for the link in a future preset time period; Perform cross-validation and generalization ability evaluation on the trained random forest model to determine the optimal model parameter configuration, including the number of decision trees, the maximum tree depth, and the feature selection strategy.

[0021] By adopting the above technical solutions, the random forest model can establish a training set based on the historical link operation data of multiple satellite transit cycles, effectively covering different communication environment changes. By extracting key features such as link signal strength and signal-to-noise ratio and annotating classification labels with actual regulation records, the model training process directly targets the link regulation requirements, improving the operational relevance of the prediction results. The classification model formed through supervised learning can judge in advance whether the future link needs to adjust the configuration, reducing the risk of link interruption. The model parameter configuration optimized through cross-validation and generalization evaluation improves the prediction accuracy and stability of the link state under different orbital positions and environmental conditions.

[0022] Optionally, S7 includes the following sub-steps: S71. During the period when the upper-layer policy scheduler enters the link regulation policy update, the lower-layer real-time controller maintains the current high-power configuration and low-rate modulation and coding scheme to ensure the continuity and reliability of link communication; S72. The lower-layer real-time controller detects the recovery of the link signal-to-noise ratio during the link monitoring process. If the signal quality improves and is higher than the set recovery margin threshold, it will independently perform the action of gradually reducing the transmit power and monitor the change of the link quality; S73. When the lower-layer real-time controller receives the instruction from the upper-layer policy scheduler to maintain the high-reliability setting, it stops the independent recovery action and maintains the current high-power and low-rate configuration until it receives a new recovery instruction; S74. During the period of maintaining the high-reliability configuration, the lower-layer real-time controller continuously monitors the link state and periodically reports the signal state indicators to the upper-layer policy scheduler to assist the upper-layer in subsequent decision-making. Among them, the signal state indicators include signal-to-noise ratio, bit error rate, and frame loss rate.

[0023] By adopting the above technical solutions, the lower-layer real-time controller maintains a high-power and low-rate configuration during the upper-layer policy update, avoiding link interruption due to signal fluctuations. By detecting the rise of the signal-to-noise ratio, it autonomously and gradually reduces the power, realizing the dynamic optimization of link load and energy consumption. When receiving the instruction from the upper layer to maintain a high-reliability configuration, it can immediately stop power recovery to prevent the link from becoming unstable again due to too fast self-adjustment. By periodically reporting the link signal status, the upper layer can timely grasp the change of link quality and reasonably decide the subsequent policy switching time, thereby extending the stable operation time of the link and reducing unnecessary adjustment actions.

[0024] Optionally, S9 includes the following sub-steps: S91. After receiving the recovery configuration instruction sent by the upper-layer policy scheduler, the lower-layer real-time controller gradually reduces the transmit power according to the set progressive recovery strategy and the preset step size. The amplitude of each power reduction is less than or equal to 1 dB, and the link signal quality is continuously monitored after each power adjustment; S92. During the power reduction process, if the link signal-to-noise ratio still remains higher than the set recovery margin threshold and the bit error rate and frame loss rate are within the normal range, continue to execute the next power reduction operation until the target transmit power configuration is reached; S93. When the transmit power is restored to the target power configuration, the lower-layer real-time controller gradually increases the modulation and coding scheme rate according to the set recovery rule, increasing one modulation order or coding rate each time, and monitoring the change of the link bit error rate after each switch; S94. If during the recovery process, the link signal-to-noise ratio drops below the recovery margin threshold or the bit error rate rises sharply, immediately abort the current recovery operation and roll back to the previous stable configuration state to ensure the stability of link communication; S95. After completing the power recovery and modulation rate recovery, the lower-layer real-time controller notifies the upper-layer policy scheduler of the status of link recovery completion and enters the normal communication mode.

[0025] By adopting the above technical solutions, the lower-layer real-time controller can gradually reduce the transmit power in small steps according to the set after receiving the recovery configuration instruction. After each power adjustment, combined with signal quality monitoring, it ensures that the recovery process is controlled and the link is stable. When the link signal-to-noise ratio and bit error rate meet the conditions, continue to recover to the target power configuration to avoid energy consumption waste. After the power recovery is completed, gradually increase the modulation rate to improve communication efficiency, and at the same time monitor the change of the bit error rate in real time to prevent the link from becoming unstable due to too fast rate increase. If link degradation occurs during the recovery process, it can be aborted and rolled back in time to ensure that link communication is not interrupted. Report the status in time after recovery to facilitate the upper-layer policy to be updated synchronously. Description of the Drawings

[0026] Figure 1 The module connection diagram of a satellite link dynamic regulation device in an embodiment of the present invention is shown.

[0027] Figure 2 The flowchart of a satellite link dynamic regulation method in an embodiment of the present invention is shown. Detailed implementation manners

[0028] The following further describes the present application in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] In the following description, for the purpose of explanation, many specific details are set forth in order to provide a thorough understanding of the inventive concept. As part of this specification, some of the drawings in the present disclosure represent structures and devices in block diagram form to avoid complicating the disclosed principles. For clarity, not all features of the actual specific implementation need to be described. In addition, the language used in the present disclosure has been mainly selected for readability and guidance purposes and may not have been selected to delimit or define the subject matter of the invention, and thus recourse to the required claims is necessary to determine such inventive subject matter. References to "a specific implementation" or "specific implementations" in the present disclosure mean that the specific features, structures, or characteristics described in connection with that specific implementation are included in at least one specific implementation, and multiple references to "a specific implementation" or "specific implementations" should not be construed as necessarily all referring to the same specific implementation.

[0030] Unless explicitly defined, the terms "a", "an", and "the" are not intended to refer to a singular entity but include general categories for which specific examples can be used for illustration. Thus, the use of the term "a" or "an" can mean any number of at least one, including "one", "one or more", "at least one", and "one or more than one". The term "or" means any one of the options and any combination of the options, including all the options, unless the options are explicitly indicated to be mutually exclusive. The phrase "at least one of" when combined with a list of items refers to a single item in the list or any combination of items in the list. The phrase does not require all of the listed items, unless explicitly so defined.

[0031] Referring to Figure 1 , the present invention discloses a satellite link dynamic regulation device, including an upper-layer policy scheduler, a middle-layer predictor, and a lower-layer real-time controller.

[0032] The upper-layer policy scheduler is used to obtain the link state prediction results output by the middle-layer predictor and the current link state information reported by the lower-layer real-time controller, and select and execute link regulation policies based on task requirements. Among them, the link regulation policies include an initial policy, a dynamic regulation policy, and a recovery to normal communication policy, and send link parameter configuration instructions to the lower-layer real-time controller; The middle-layer predictor is used to extract an input feature set within a set historical sliding time window based on the link real-time data collected by the lower-layer real-time controller, and input the input feature set into a pre-trained random forest model for link state prediction. Among them, the pre-trained random forest model is constructed based on historical link operation data and trained with the training objective of whether the classification label triggers a link adjustment instruction; The lower-layer real-time controller is used to perform transmit power adjustment and / or modulation and coding scheme adjustment according to the link parameter configuration instructions issued by the upper-layer policy scheduler, and perform emergency power boost and modulation degradation operations to stabilize link communication when it detects that the link signal-to-noise ratio is lower than the burst threshold within the preset burst response duration. At the same time, it feeds back the emergency status information to the upper-layer policy scheduler to perform progressive power recovery and modulation mode improvement after receiving the recovery configuration instructions issued by the upper-layer policy scheduler, and complete the fallback of the link to the normal communication mode.

[0033] The upper-layer policy scheduler is a global decision-making and control module, responsible for integrating link prediction information and real-time state information, formulating a regulation plan according to task priorities and communication policy requirements, and generating link parameter configuration instructions. The middle-layer predictor undertakes the prediction calculation task, extracts features from the link historical data within the sliding time window, and uses the pre-trained random forest model to classify and predict the future link margin state to identify potential risks in advance. The lower-layer real-time controller, as an execution and monitoring module, is responsible for real-time collecting link operation data, executing configuration instructions, and performing fast adjustment actions when the link mutates to ensure immediate link stability and gradually adjust back to the normal communication state during the recovery process.

[0034] In terms of the signal control path, link signal operation data (such as SNR, BER, power level, etc.) is first collected by the lower-layer real-time controller and uploaded to the middle-layer predictor and the upper-layer policy scheduler in real time; the middle-layer predictor performs prediction based on the collected data and outputs the prediction results to the upper layer; the upper-layer policy scheduler combines the prediction information and the current state to evaluate whether policy adjustment is needed, and generates transmit power and modulation parameter instructions according to the results. The instructions are transmitted to the lower-layer real-time controller through the control link to complete the link configuration closed-loop.

[0035] In this application, the main reason for dividing the system into a three-layer structure is as follows: The upper-layer policy scheduler has the ability to comprehensively weigh global policies and manage priorities, which is conducive to dynamically adjusting policies based on different business scenarios; the middle-layer predictor focuses on data processing and model inference, avoiding direct coupling between the upper-layer decision-making logic and the underlying perception data, and improving scalability; the lower-layer real-time controller independently undertakes link state collection and instruction execution, ensuring the timeliness of response and the granularity of control.

[0036] Specifically, the upper-layer policy scheduler includes a status receiving unit, a policy selection unit, a decision generation unit, a configuration distribution unit, and a policy update unit. These units work together to achieve global perception of the link, policy formulation, and distribution of control instructions. The status receiving unit is responsible for obtaining the link state prediction result from the middle-layer predictor and receiving the real-time link state data reported by the lower-layer real-time controller. These data include the future margin change trend, the current signal-to-noise ratio, the bit error rate, the transmit power level, etc., which are the core basis for the upper-layer policy scheduler to make policy decisions.

[0037] The status receiving unit is used to receive the link state prediction result output by the middle-layer predictor and the current link operating state information reported by the lower-layer real-time controller; the policy selection unit is used to select the current link control policy to be executed from the initial policy, the dynamic regulation policy, and the recovery of normal communication policy according to the task requirements and the comprehensive link state information; the decision generation unit is used to generate a transmit power configuration instruction and / or a modulation and coding scheme configuration instruction based on the selected link control policy; the configuration distribution unit is used to send the transmit power configuration instruction and / or the modulation and coding scheme configuration instruction to the lower-layer real-time controller to implement the execution of the link control instruction; the policy update unit is used to enter a rapid decision-making loop after receiving an emergency status notification from the lower-layer real-time controller and re-evaluate and update the executed link control policy based on the real-time status.

[0038] After receiving the prediction result and the current link state, the policy selection unit will select an appropriate link control policy according to the set task type (such as delay-sensitive, energy efficiency priority, or high-reliability tasks). Taking the example of a satellite transmitting a low-Earth orbit data segment, if it is a short-time high-priority data window, the policy selection unit will preferentially select a high-margin policy with the goal of link reliability; if the prediction result shows that the link quality will deteriorate in the next ten seconds but is currently stable, the policy selection unit will anticipate entering the dynamic regulation policy state and actively resist the upcoming link degradation by gradually increasing the transmit power and reducing the modulation method.

[0039] The decision-making generation unit generates specific parameter configuration instructions based on the policy type output by the policy selection unit, in combination with link status indicators, historical regulation success rates, and adjustment granularity requirements. For example, in the case where it is predicted that the link margin will drop below the safety threshold after 3 seconds in the future, the instruction that the decision-making generation unit may issue is "increase the transmit power by 1.5 dB within the next 2 seconds and switch the current 64QAM modulation mode to 16QAM". If the system is in a low-priority task state, the generated instruction may be an energy-saving configuration such as "slowly decrease by 0.5 dB after a 5-second delay".

[0040] The configuration distribution unit is responsible for transmitting the above-generated parameter instructions to the lower-layer real-time controller through the link management interface, ensuring that the policy can be quickly converted into actual actions and is accompanied by a timestamp or synchronization mark to avoid deviation in the regulation effect caused by the incorrect timing of parameter activation.

[0041] The policy update unit is responsible for re-entering the fast decision-making loop in the event of a sudden link anomaly. When the lower-layer real-time controller reports an emergency event, the policy update unit will first pause the current original policy execution path, conduct a review and evaluation of whether the previous regulation was effective, and at the same time call the most recent prediction information to regenerate the regulation plan. For example, if a power slow-rise policy was adopted in the previous regulation but the link still suddenly became intermittent, the policy update unit may mark the current state as "insufficient response" and give priority to selecting a combination of "instantaneously increasing power + downregulating modulation" in the next event.

[0042] In a specific application scenario, for example, when a low-earth-orbit remote sensing satellite passes through the mid-latitude cloud region during the downlink of image data, the link quality may be affected by the combined effects of water vapor attenuation and elevation angle changes. At this time, the status receiving unit continuously receives the link status information uploaded by the lower-layer real-time controller, indicating that the current signal-to-noise ratio has started to decline to near the critical value, and at the same time the mid-layer predictor outputs that the probability of the margin dropping below the safety threshold within the next 5 seconds exceeds 80%. Based on this, the policy selection unit determines that the current state is "predictable degradation", and labels this data segment as the "non-interruptible" category in combination with the task identifier, and selects to execute the dynamic regulation policy instead of the delay recovery or energy-saving mode.

[0043] The decision-making generation unit generates a combined configuration based on the current state: increase the transmit power by 2 dB, lower the modulation mode from 64QAM to 16QAM, and set the regulation duration to 7 seconds. The configuration distribution unit then immediately generates a regulation instruction frame with a time tag and issues it to the lower-layer real-time controller through the high-priority control channel to ensure that the underlying control instruction is put into execution before the link state continues to deteriorate.

[0044] If the above regulation still fails to prevent further deterioration of the link metrics during the actual process, the lower-layer real-time controller will trigger an emergency event report. After receiving this notification, the policy update unit quickly calls the historical features in the previous prediction cycle and adjusts the policy path in combination with the current anomaly trigger conditions. At this time, the policy update unit will abandon the original "prediction compensation" path and enter the "quick fallback" path, and re-call the policy selection unit to execute the high-margin guarantee mode, generating a forced configuration with the transmit power increased to the maximum and the safety threshold setting relaxed to ensure the continuous availability of the link.

[0045] Once the link recovers to a relatively high signal-to-noise ratio and the bit error rate stabilizes below the target level, and the new round of prediction by the middle-layer predictor shows that the link trend is stable in the short term, the status receiving unit combines the latest link status to determine that the link has met the recovery conditions. At this time, the policy selection unit switches to the recovery normal communication policy, and the decision generation unit generates phased instructions of "gradually reducing the power and increasing the modulation rate", and the configuration distribution unit pushes the instructions to the lower layer in batches until the link recovers to the normal configuration with the lowest power and the highest modulation efficiency.

[0046] Specifically, the middle-layer predictor, as the core module for link prediction calculation, mainly includes a feature extraction unit, a model inference unit, a confidence evaluation unit, and a prediction update interface unit. Its main responsibility is to make a forward-looking prediction of the future state of the link based on historical link data and current sampling data, and assist the upper-layer policy scheduler to actively adjust the configuration before the link degrades.

[0047] The feature extraction unit is used to extract an input feature set based on the received real-time link data within a set historical sliding time window. The input feature set covers the statistical features of non-high-frequency fast-fluctuating data, including received signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, and satellite elevation angle; the model inference unit is used to input the input feature set into a pre-trained random forest model and output the link state prediction result, including the predicted value of the link margin or the link failure risk level in the future period; the confidence evaluation unit is used to evaluate the confidence of the output result of the model inference unit and send the prediction result and the confidence index to the upper-layer policy scheduler together to assist in policy selection; the prediction update interface unit is used to provide a training data collection and offline model update interface to support the periodic training and optimization of the random forest model based on the newly collected link operation data.

[0048] The feature extraction unit is used to construct a sliding window from the link operation data reported by the lower-layer real-time controller and extract stable statistical features within this time window. The selected features include, but are not limited to, the mean and slope of the received signal strength RSSI, the short-term variance of the signal-to-noise ratio SNR, the mean bit error rate BER per unit time, the change frequency of the modulation and coding scheme within the time window, the historical stability of the current transmission power level, and the change rate of the current satellite elevation angle, etc. These features can effectively reflect the link trend and avoid misjudgment caused by instantaneous mutations. For example, during a certain link operation, when the signal-to-noise ratio slowly decreases within the past 5 seconds, the bit error rate slightly increases, the elevation angle decreases faster, and the transmission power is at a medium to high level and remains unchanged, the feature extraction unit will output a set of feature vectors indicating the link degradation trend.

[0049] The model inference unit inputs the feature set output by the feature extraction unit into a pre-trained random forest classification model. This model is trained based on the real link operation data within multiple satellite transit cycles. During the training process, whether a link adjustment instruction has been triggered is used as the classification label for supervised learning, enabling the model to not only judge whether the link has deteriorated but also pay more attention to whether the link needs to be adjusted in the current state. In the above example, the model inference unit will output a prediction result of "the link margin will drop below the preset threshold within the next 2 seconds" and return a "regulation requirement level" of medium to high level, indicating that a control response should be made in advance.

[0050] The confidence evaluation unit determines the consistency and reliability of the prediction results output by the model inference. For the prediction consistency of each decision tree within the forest, this unit will calculate the prediction confidence interval and generate the corresponding confidence score. For example, in a prediction cycle, 87% of the subtrees in the model judge that adjustment is required, and the output time points of each tree are concentrated between 1.5 and 2.5 seconds in the future. The confidence evaluation unit will assign a "high confidence level" label to this prediction and recommend that the upper-layer policy scheduler give priority to referring to this prediction result for early adjustment.

[0051] The prediction update interface unit is responsible for realizing the sustainable optimization of the model. On the one hand, it supports the automatic archiving of the data after the actual link operation back to the model training pool to form continuously accumulated data samples; on the other hand, it provides an interface to support the periodic update of the model or the switching of model versions with different training parameters under different orbital heights and regional environmental conditions. For example, when the satellite enters the vicinity of the equator and the link perturbation mode is different from that in the mid-latitudes, the prediction update interface can switch to the model version trained for the equatorial link environment to improve the adaptability of the prediction in different scenarios.

[0052] In a typical example, a certain low-Earth orbit satellite is performing a task of downlinking remote sensing data. The signal-to-noise ratio (SNR) of the link within the prediction window slowly drops from 13 dB to 10 dB, and at the same time, the bit error rate (BER) slightly increases but remains below the trigger threshold, and the modulation mode is maintained at 16QAM. The middle-layer predictor outputs a conclusion of "there is a need to adjust the link within the next 1.5 seconds, and the confidence level reaches 91%" through feature extraction, model inference, and confidence evaluation. Based on this, the upper-layer policy scheduler generates a fine-tuning configuration of "increasing the transmit power by 1 dB and keeping the modulation scheme unchanged". In a system without a prediction mechanism, such a state may be considered not to trigger adjustment yet. However, the middle-layer predictor realizes the early perception of link degradation through the mapping relationship between the learned trend features and historical adjustment data, thus significantly reducing the passive response window before communication interruption.

[0053] Specifically, the lower-layer real-time controller, as the execution layer of link regulation, undertakes key functions such as link state data acquisition, configuration instruction execution, handling of sudden anomalies, and recovery control. It mainly includes a link monitoring unit, an instruction execution unit, a sudden event detection unit, an emergency adjustment unit, and a progressive recovery control unit inside. The lower-layer controller perceives the link state in real time at a high sampling frequency and ensures that the upper-layer policy is timely and effectively transformed into specific control actions in the actual link.

[0054] The link monitoring unit is used to collect link signal state data at a preset time interval, including SNR, BER, and frame loss rate, and report the data to the middle-layer predictor and the upper-layer policy scheduler; the instruction execution unit is used to control the transmitter to make corresponding parameter adjustments according to the transmit power configuration instruction and / or modulation and coding scheme configuration instruction issued by the upper-layer policy scheduler; the sudden event detection unit is used to judge whether the link SNR drops below the sudden threshold within the sudden response duration and trigger emergency adjustment when the condition is met; the emergency adjustment unit is used to immediately increase the transmit power to the upper limit and switch to a low-order modulation and coding scheme to stabilize link communication when a link anomaly is detected, and feedback the emergency state information to the upper-layer policy scheduler; the progressive recovery control unit is used to gradually reduce the transmit power and restore the high-order modulation scheme at a set rhythm after receiving the recovery configuration instruction, and return to the normal communication mode on the premise of ensuring link stability.

[0055] The link monitoring unit is used to collect the signal status metrics of the current communication link at preset time intervals (e.g., every 200 milliseconds), including key parameters such as signal-to-noise ratio (SNR), bit error rate (BER), frame loss rate (FER), current transmit power level, current modulation and coding scheme, etc., and perform statistical processing in a sliding window manner. These data are not only used for its own judgment and control, but also synchronously uploaded to the middle-layer predictor and the upper-layer policy scheduler to provide a unified link status basis for the entire system. For example, in a certain satellite transit segment, the link monitoring unit detects that the signal-to-noise ratio shows a fluctuating downward trend, dropping from 14.5 dB to 11.2 dB, and the frame loss rate rises to 0.03, and the system thus judges that the link is entering an unstable area.

[0056] The instruction execution unit is used to receive and decode the link configuration instructions issued by the upper-layer policy scheduler, and convert them into control signals recognizable by the underlying transmitter, and control the adjustment of parameters such as transmit power, modulation and coding scheme, coding rate, etc. For example, when the upper layer issues an instruction to increase the transmit power by 2 dB and switch the modulation mode from 64QAM to 16QAM, the instruction execution unit will first confirm the current link parameter status, and then trigger the specific power amplifier adjustment logic and modulation and demodulation module configuration switch according to the command content to ensure that the link parameters complete the transition without interrupting the communication.

[0057] The emergency event detection unit is used to identify rapid fluctuations in the link state and determine whether a margin sudden drop anomaly is triggered within a short period of time (e.g., 100 milliseconds). The specific criterion is whether the signal-to-noise ratio is continuously lower than the preset burst threshold within the burst duration, or the frame loss rate exceeds a certain set threshold within a unit time, etc. When the judgment condition is met, this unit will immediately notify the emergency adjustment unit to activate the fallback configuration. For example, during a communication process when a satellite passes through an occlusion area, the link undergoes a rapid change in the signal-to-noise ratio from 11 dB to 7 dB within a short period of time. At this time, the emergency event detection unit continuously detects the anomaly within three sampling periods and immediately triggers an emergency response.

[0058] The emergency adjustment unit is used to quickly stabilize the link state. After the burst detection is triggered, this unit will immediately increase the transmit power to the maximum set value (e.g., 25 dBm), and at the same time reduce the modulation mode to a more robust scheme (e.g., switch from 16QAM to QPSK) to improve the link reliability with the shortest response time and avoid the loss of current data or communication interruption. This process is usually completed within one sampling period without waiting for the upper-layer policy to reissue instructions, ensuring that the system has the local fast emergency response ability. After completing the emergency response, this unit will also package and send the current adjustment result and status mark to the upper-layer policy scheduler for it to enter the subsequent policy correction process.

[0059] The progressive recovery control unit is used to execute a phased recovery strategy for power and modulation scheme according to the upper-layer recovery configuration instruction after the link is stable. Its control logic usually includes setting the step of reducing the transmit power (such as not exceeding 1 dB each time) and gradually increasing the modulation and coding rate, and re-detecting the link status after each parameter change. Only when the signal-to-noise ratio still meets the target margin is it allowed to enter the next adjustment. For example, when the link status recovers well and the signal-to-noise ratio is stable above 14 dB, this unit will reduce the power by 0.5 dB every 30 seconds, and gradually restore the modulation mode to 16QAM, then to 64QAM, and monitor whether the bit error rate is still within the acceptable range (such as <0.01) after each change. If the bit error rate soars after a certain step, it will immediately roll back to the previous stable configuration.

[0060] Taking the downlink of image data of a certain remote sensing satellite as an example, when the link passes through the earth's shadow, the signal attenuates violently. The emergency detection unit triggers an emergency adjustment within milliseconds, and the system quickly switches to the high-power + low-rate mode to ensure the uninterrupted transmission of key data. Subsequently, when the satellite elevation angle rises and the link quality recovers, the system smoothly restores the link to the normal configuration within 90 seconds through the progressive recovery control unit, and there is no data transmission interruption or repeated adjustment jitter during the whole process. This architecture of high-frequency monitoring + local execution + upper-layer cooperation ensures the timeliness, stability and continuity of the regulation.

[0061] Furthermore, the present application also discloses a satellite link dynamic regulation method, referring to Figure 2 , including the following steps S1-S9.

[0062] S1. When detecting the start of a satellite pass or the establishment of a link, the upper-layer policy scheduler sets an initial policy according to the task requirements, and sends the initial transmit power configuration, the initial modulation and coding scheme configuration, and the initial margin threshold parameter to the lower-layer real-time controller. The lower-layer real-time controller sets the initial transmit power and the initial modulation mode according to the initial configuration, and starts link monitoring; among them, the execution policies of the upper-layer policy scheduler include the initial policy, the dynamic regulation policy and the policy of restoring normal communication.

[0063] Furthermore, in this embodiment, the S1 includes the following sub-steps S11-S14.

[0064] S11. Detect the start of a satellite pass or the establishment of a link event; S12. The upper-layer policy scheduler sets the initial link configuration policy according to the task requirements, including the initial transmit power, the initial modulation and coding scheme, and the initial link margin threshold; S13. The upper-layer policy scheduler sends the initial link configuration parameters to the lower-layer real-time controller; S14. The lower-layer real-time controller sets the transmit power and modulation and coding scheme according to the initial link configuration, and activates the link monitoring function.

[0065] In S11, the system determines that the satellite is about to enter the target service area or the link has been successfully established by monitoring the satellite orbit data and the ground terminal link establishment status. At this node, the system triggers the initialization of the link dynamic management process. For example, when a low-orbit remote sensing satellite is about to enter the visible window of the ground station, the satellite orbit control system sends an "enter communication window" instruction to the communication module, and the link dynamic regulation device determines that the link establishment condition is met accordingly.

[0066] In S12, the upper-layer policy scheduler determines the initial link configuration policy according to the current task demand type (such as data backhaul, real-time command, or low-priority monitoring). The initial configuration policy includes at least setting the transmit power level (for example, setting it to 80% of the full power to balance energy consumption and link stability), selecting the modulation and coding scheme (for example, initially using 16QAM to balance rate and link margin), and determining the initial link margin threshold (such as setting the target signal-to-noise ratio to 12 dB).

[0067] In S13, the upper-layer policy scheduler generates a link parameter instruction containing the initial transmit power, initial modulation and coding scheme, and initial margin threshold, and sends it to the lower-layer real-time controller through the control channel. The parameter instruction needs to be attached with a synchronization flag and an effective time to ensure the synchronous switching of each parameter and avoid short-term anomalies at the initial stage of the link transit. For example, if the instruction sets the effective time to 15:30:00 UTC, the lower-layer real-time controller will uniformly adjust the link parameters at the corresponding time to ensure the smooth completion of the link initialization phase.

[0068] In S14, according to the received initial link configuration, the lower-layer real-time controller immediately sets the output power level of the transmit power module, configures the modulation and coding module to switch to the specified modulation scheme, and enables the link monitoring function, starting to collect real-time link data at a preset frequency (such as once every 200 ms), including but not limited to indicators such as signal-to-noise ratio, bit error rate, and frame loss rate. Taking a specific example, after the initial transmit power is set to 23 dBm and the modulation and coding is set to 16QAM 1 / 2 rate, the lower-layer real-time controller starts monitoring and collects link status data five times per second to provide a basis for subsequent dynamic prediction and regulation.

[0069] S2. The lower-layer real-time controller continuously collects real-time link data and organizes it into an input feature set. The middle-layer predictor continuously performs link state prediction based on the input feature set. The upper-layer policy scheduler obtains the link state prediction result output by the middle-layer predictor and the current link state reported by the lower-layer real-time controller at a set decision-making cycle. When the predicted value of the link margin is lower than the preset safety threshold, the upper-layer policy scheduler pre-adjusts the link configuration policy.

[0070] Specifically, the S2 includes the following sub-steps S201 - S203.

[0071] S201. The lower-layer real-time controller continuously collects real-time link data at a preset sampling interval, and organizes and statistically processes the real-time link data according to a set sliding time window to obtain an input feature set including indicators such as link signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, and satellite elevation angle. Among them, the length of the sliding time window is from 1 second to 5 seconds, and the step size is 0.5 seconds. The input feature set does not include high-frequency and fast-fluctuating data that requires millisecond-level response. S202. The middle-layer predictor inputs the input feature set into a pre-trained random forest model, which is trained based on historical link operation data and uses whether a link adjustment instruction is triggered as a classification label to perform link state prediction and output the prediction result of the link margin state within a preset future time period. S203. The middle-layer predictor calculates the link failure risk level and / or link safety margin according to the prediction result, and sends the prediction result and confidence index to the upper-layer policy scheduler together to assist in the selection and update of the link control strategy.

[0072] In sub-step S201, the lower-layer real-time controller continuously collects real-time link data at a preset sampling interval (for example, every 200 milliseconds), and organizes and statistically processes it according to a set sliding time window (with a length from 1 second to 5 seconds and a step size of 0.5 seconds). The collected data includes indicators such as link signal strength (RSSI), signal-to-noise ratio (SNR), bit error rate (BER), modulation and coding scheme (MCS level), transmit power level, and satellite elevation angle. Through the organization of the sliding window, outliers caused by high-frequency instantaneous fluctuations of the link can be effectively filtered out, improving the stability and prediction reliability of the input data. Taking a specific example, when the satellite passes over the ground station at a speed of 7.6 km / s, there is a rapid fluctuation phenomenon in the link signal-to-noise ratio. By calculating the mean and variance of SNR within a 5-second window, the link attenuation trend can be accurately captured, rather than misjudging the link quality due to single-point noise.

[0073] In sub-step S202, the middle-layer predictor inputs the input feature set obtained in S201 into a pre-trained random forest model. The random forest model is constructed based on historical link operation data, and during the training process, whether it is necessary to trigger link parameter adjustment is used as the classification label. By using the forest structure composed of multiple decision trees, the overfitting risk of a single tree model to specific abnormal data can be effectively reduced, and the classification accuracy of link trend changes under different environmental conditions can be improved. Taking an actual application scenario as an example, during a certain mid-latitude region transit, the link signal-to-noise ratio continuously decreased by 3 dB, and the random forest model accurately predicted that the link would fall below the margin lower limit within 2 seconds, thus prompting the system to adjust the configuration in advance to avoid link interruption.

[0074] In sub-step S203, the middle-layer predictor calculates the link failure risk level and / or the link safety margin according to the model inference result, and sends the prediction result together with the confidence index to the upper-layer policy scheduler. The confidence index is calculated based on the consistency of the prediction results of each decision tree inside the forest. For example, in a prediction, 95% of the decision trees all judge that adjustment is needed within the next 3 seconds, then the output confidence is 95%. The upper-layer policy scheduler decides whether to immediately execute the pre-adjustment policy according to the confidence value, so as to avoid false triggering caused by low-confidence predictions.

[0075] Specifically, in S2, it also includes sub-steps S211 - S215 where the upper-layer policy scheduler makes decisions based on the prediction result and the link status information.

[0076] S211. The upper-layer policy scheduler receives the link state prediction result output by the middle-layer predictor and the current link state information reported by the lower-layer real-time controller. The link state prediction result includes the link margin change trend within a preset future time period, and the current link state information includes the real-time signal-to-noise ratio, bit error rate, frame loss rate, and transmission power level. S212. The upper-layer policy scheduler conducts a trend analysis on the link state prediction result to judge whether there is a risk of insufficient link margin in the future and the degree of link quality deterioration. S213. The upper-layer policy scheduler conducts a real-time evaluation on the current link state information to judge whether the current link has approached the failure boundary, including whether there is a situation where the signal-to-noise ratio is close to the lowest decoding threshold or the bit error rate rises sharply. S214. The upper-layer policy scheduler makes a joint determination of the trend analysis result and the real-time evaluation result. If the prediction result and / or the current state indicate that the link margin is lower than the preset safety threshold, a link pre-adjustment configuration policy is generated to determine whether to adjust the transmission power and / or switch the modulation and coding scheme. S215. The upper-layer policy scheduler formulates the adjustment range and adjustment priority based on the pre-adjustment configuration policy, and prepares to send the corresponding link parameter configuration instruction to the lower-layer real-time controller.

[0077] In sub-step S211, the upper-layer policy scheduler receives the link state prediction result output by the middle-layer predictor and the current link state information reported by the lower-layer real-time controller. The prediction result includes the change trend of the link margin within the next several seconds, and the link state information includes the real-time SNR, BER, FER, and transmission power level. Taking an actual application as an example, in a certain maritime base station connection, the prediction result indicates that the link margin will decrease by more than 4 dB in the next 5 seconds, and the current SNR is only 2 dB higher than the minimum decoding threshold. Based on this, the upper-layer policy scheduler enters the emergency pre-adjustment mode.

[0078] In sub-step S212, the upper-layer policy scheduler performs trend analysis based on the prediction result, judges whether there is a risk of insufficient link margin, and evaluates the link quality degradation speed. For example, through the joint analysis of the SNR decrease rate and the BER increase trend, it can be identified whether the link configuration needs to be adjusted quickly.

[0079] In sub-step S213, the upper-layer policy scheduler performs real-time evaluation based on the link state reported by the lower-layer real-time controller, and judges whether it is close to the link failure boundary. The judgment basis includes but is not limited to: whether the current SNR is less than 1 dB within the set margin safety boundary, or whether the BER shows an exponential increase.

[0080] In sub-step S214, the upper-layer policy scheduler makes a joint determination of the trend analysis result and the real-time evaluation result. If one or both of them indicate that the link safety margin is lower than the preset threshold, a link pre-adjustment configuration policy is generated. For example, in a certain transit section, the trend analysis shows that the link will become unstable in the next 2 seconds, and the real-time evaluation finds that the current BER has approached the warning line. The system generates a combined adjustment strategy of increasing the power by 1.5 dB and downgrading the modulation scheme by one level.

[0081] In sub-step S215, the upper-layer policy scheduler formulates the specific parameter adjustment range and adjustment priority according to the pre-adjustment configuration policy, and prepares to send the corresponding instruction to the lower-layer real-time controller. For example, the power increase is executed first, and then gradually switched to a more robust modulation and coding scheme to avoid further signal deterioration during the link switching process. Specifically, the pre-training steps of the random forest model include: Construct a training data set based on the historical link operation data collected during multiple satellite transit cycles. The historical link operation data includes link signal strength, SNR, BER, modulation and coding scheme, transmission power level, and satellite elevation angle, and perform feature extraction on the data according to the set time window to form an input feature vector; According to the link regulation records corresponding to the historical link data, mark whether a link parameter adjustment instruction has been triggered as a classification label, where the link parameter adjustment instruction includes a transmit power adjustment instruction and a modulation and coding scheme switching instruction; Construct a supervised learning sample using the input feature vector and the classification label, and train a random forest classification model so that the model can output a prediction result on whether there is a regulation requirement for the link within a preset time period in the future; Perform cross-validation and generalization ability evaluation on the trained random forest model to determine the optimal model parameter configuration, including the number of decision trees, the maximum tree depth, and the feature selection strategy.

[0082] For the pre-training step of the random forest model, in this embodiment, a training data set is constructed using historical link operation data collected over multiple satellite transit cycles, and feature vectors including link signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, and satellite elevation angle are extracted, and whether a link parameter adjustment instruction has been triggered is marked as a supervised learning label. By using a large amount of actual operation data as training samples, the adaptability and accuracy of the model under different orbits, different geographical regions, and different meteorological conditions can be improved. After training, the optimal model hyperparameters, including the number of decision trees, the maximum tree depth, and the feature subset selection method, are determined through cross-validation and generalization ability evaluation to ensure that the model has both high prediction accuracy and can effectively avoid overfitting in actual deployment. For example, in a certain training set, by setting the forest scale to 100 trees and the maximum tree depth to 10 layers, the prediction accuracy of the link regulation requirement on the test data finally reaches 93%, and the false trigger rate is kept low (less than 5%).

[0083] S3. The upper-layer policy scheduler issues an adjustment instruction to the lower-layer real-time controller according to the pre-adjusted link configuration policy to adjust the transmit power configuration and / or the modulation and coding scheme configuration.

[0084] In S3, after the upper-layer policy scheduler completes the link state trend analysis, real-time state evaluation, and joint determination in step S2, if it is determined that the link margin has fallen below the preset safety threshold or is about to fall below the safety margin threshold, an adjustment instruction is issued to the lower-layer real-time controller according to the pre-generated pre-adjusted configuration policy. The adjustment instruction at least includes a transmit power configuration instruction and / or a modulation and coding scheme configuration instruction, and the instruction content and the issuing rhythm can be dynamically adjusted according to the degree of link deterioration, the prediction confidence level, and the service demand priority.

[0085] In a specific implementation, the upper-layer policy scheduler determines the target value of the adjustment parameter and the adjustment method based on the link trend prediction and real-time status. For example, during a satellite transit, if the prediction shows that the link signal-to-noise ratio will drop below 8 dB in the next 3 seconds and the current bit error rate is already close to the maximum allowable threshold, the upper-layer policy scheduler can decide to increase the transmit power by 1.5 dB and reduce the modulation mode from 64QAM to 16QAM. At the same time, to avoid link oscillations caused by parameter mutations, the upper-layer policy scheduler can also set the parameter adjustment step in the instruction. For example, the power is gradually increased by 0.5 dB each time, and the modulation switch takes effect with a delay of 1 second to ensure a smooth transition of the control action.

[0086] In the generation of the transmit power configuration instruction, the scheduler needs to calculate the target transmit power value and the adjustment path based on the current transmit power level and the maximum power limit allowed by the system. For example, if the current power is 20 dBm, the maximum allowed is 24 dBm, and it is predicted that an increase of about 2 dB is needed to maintain the margin, the target power of 22 dBm is sent down, and the instruction is completed in two steps (an increase of 1 dB each step). This setting can effectively reduce the risk of link instability caused by a large one-time adjustment.

[0087] In the generation of the modulation and coding scheme configuration instruction, the scheduler determines the target modulation rate based on the current MCS level and the preset MCS downgrading rules. For example, under the condition that the current MCS is 64QAM 3 / 4, according to the degree of link degradation, it is preferentially switched to 16QAM 1 / 2 to improve the link's tolerance to signal degradation. If the predicted link recovery time is short, a temporary low-rate switch can be adopted; if it is predicted that the link degradation will last for a long time, a continuous low-order modulation configuration is adopted.

[0088] S4. After the lower-layer real-time controller adjusts the link parameters according to the adjustment instruction issued by the upper-layer policy scheduler, it independently performs fine power adjustment based on the current link measurement results to callback the actual signal-to-noise ratio to near the target margin and sets the preset threshold for emergency event detection.

[0089] In S4, the lower-layer real-time controller first completes the preliminary adjustment of link parameters according to the transmission power configuration instruction and / or modulation and coding scheme configuration instruction issued by the upper-layer policy scheduler. Subsequently, based on the current link measurement results, including indicators such as real-time signal-to-noise ratio (SNR), bit error rate (BER), and frame loss rate (FER), the lower-layer real-time controller performs fine power adjustment operations to accurately callback the actual link state near the target margin. For example, after the preliminary adjustment, if the transmission power is increased by 1.5 dB and the link SNR reaches 13.8 dB, while the target margin is 14 dB, the lower-layer real-time controller can increase the power by 0.2 dB through fine-tuning, so that the link SNR is further close to 14 dB, avoiding the deviation of link performance from the target range due to insufficient or excessive initial adjustment.

[0090] During the fine power adjustment process, the lower-layer real-time controller dynamically evaluates the effect of each adjustment according to the link feedback, usually approaching the target margin range step by step with a step less than 0.5 dB, and setting a short stable observation period (such as 500 ms) after the adjustment to ensure that the adjustment action does not cause new link jitters. Taking an actual application as an example, when a low-orbit satellite performs a data backhaul task, there is a 0.5 dB deviation in the link SNR after the initial configuration. The lower-layer real-time controller completes the callback within 1 second through two small-amplitude power adjustments, so that the link still maintains stable transmission during the high-load stage.

[0091] After completing the fine power adjustment, the lower-layer real-time controller sets a preset threshold for emergency event detection for rapid abnormal capture during subsequent link monitoring. The preset threshold is set according to the current link target margin and actual operating status, usually including the signal-to-noise ratio drop threshold and the bit error rate sudden increase threshold. For example, when the target signal-to-noise ratio is 14 dB, the emergency event detection threshold can be set to be lower than 12 dB, and the bit error rate sudden increase threshold is set to exceed 0.02. Such a setting ensures that when the link deteriorates sharply, an emergency adjustment can be triggered within milliseconds.

[0092] S5. The lower-layer real-time controller detects whether the signal-to-noise ratio is lower than the preset threshold within the burst duration in the link monitoring loop. If so, it increases the transmission power to the upper limit value and switches to a low-order modulation and coding scheme to stabilize the link communication.

[0093] In S5, the lower-layer real-time controller continuously detects the change of the link signal-to-noise ratio in the link monitoring loop based on the preset threshold for emergency event detection set in the previous S4 step. When it detects that the signal-to-noise ratio is lower than the preset burst threshold within the burst response duration (for example, set to 100 milliseconds to 300 milliseconds), it triggers an emergency response mechanism. The setting of the burst response duration is used to filter out occasional small fluctuations in the link and only respond to continuous and large-amplitude burst deteriorations to avoid ineffective regulation caused by over-sensitivity.

[0094] For example, during a satellite data transmission process, the lower-layer real-time controller monitors that the signal-to-noise ratio drops sharply from 13 dB to 10 dB within two consecutive sampling periods (each period is 200 milliseconds), and is lower than the preset burst threshold of 12 dB. Then, it is determined that a burst degradation event has occurred, and subsequent emergency control operations are immediately triggered.

[0095] After detecting a burst anomaly, the lower-layer real-time controller first performs an operation to increase the transmission power, quickly increasing the transmission power to the maximum value allowed by the system or the preset upper limit value. For example, the transmission power is directly increased from the originally set 22 dBm to 24 dBm (the upper limit of the system's maximum transmission power). This increase operation has the highest priority, aiming to restore the link signal-to-noise ratio to the normal range in the shortest time and prevent data transmission interruption.

[0096] At the same time, the lower-layer real-time controller synchronously performs an operation to reduce the modulation and coding scheme, switching from the originally used high-order modulation method (such as 64QAM or 16QAM) to a more robust low-order modulation method (such as QPSK or BPSK). For example, after detecting a burst anomaly, the system directly switches the modulation method from 16QAM 3 / 4 to QPSK 1 / 2, effectively improving the link's anti-interference ability in a low signal-to-noise ratio environment, thereby maintaining the successful decoding rate of data frames.

[0097] The above-mentioned power increase and modulation reduction actions are usually completed within one control cycle, and the entire response process can be closed-loop within 500 milliseconds, ensuring that the link can still maintain basic communication functions under conditions such as occlusion, interference, or other burst degradations. For example, when the satellite passes through a strong rainfall cloud area, resulting in severe attenuation of the link, this system can complete emergency control within 200 milliseconds when the link performance drops suddenly, avoiding a large amount of data loss or disconnection of the communication link.

[0098] S6. After the lower-layer real-time controller completes the burst response, it sends an emergency status notification to the upper-layer policy scheduler. After receiving the emergency notification, the upper-layer policy scheduler immediately enters the decision-making loop to update the link control policy.

[0099] In S6, after the lower-layer real-time controller completes the emergency control operations of increasing the transmission power and reducing the modulation, it immediately generates an emergency status notification and reports it to the upper-layer policy scheduler through the control channel. The emergency status notification contains at least the current link signal-to-noise ratio, bit error rate, frame loss rate, current transmission power level, current modulation and coding scheme, and the abnormal trigger timestamp. By carrying complete link status information, the upper layer can accurately understand the severity of the emergency event and the current link operating status, providing a basis for subsequent decision-making.

[0100] For example, during a satellite's passage through a high-density cloud layer, when the signal-to-noise ratio (SNR) detected by the lower-layer real-time controller drops below the burst threshold, the lower-layer real-time controller performs emergency regulation. It increases the transmission power to 24 dBm and reduces the modulation mode to QPSK. The emergency status notification records the burst trigger time (e.g., UTC 15:32:10), the current SNR of 10.5 dB, the bit error rate increased to 0.015, and marks the current transmission power and modulation mode configuration. This information is synchronously uploaded to the upper-layer policy scheduler within 200 milliseconds.

[0101] Upon receiving the emergency status notification, the upper-layer policy scheduler does not wait for the next regular decision cycle but immediately enters a fast decision loop. The so-called fast decision loop means that the upper-layer scheduler skips the normal scheduling rhythm and re-judges the link status in a high-priority and fast-evaluation manner to determine whether further power increase, extension of the high-power maintenance time, or modification of the original recovery plan is required.

[0102] In a specific implementation, the upper-layer policy scheduler first compares the actual link status in the emergency notification with the previous predicted data to confirm whether there is a prediction deviation. If the deviation exceeds the set tolerance range (e.g., the actual SNR drops more than 3 dB from the prediction), it immediately re-evaluates the applicability of the current policy. Subsequently, according to the latest link status, it dynamically modifies the subsequent link regulation plan. For example, if the expected link recovery is delayed, the upper layer can issue an instruction to the lower layer to maintain the high-power and low-rate configuration until a new set time node, or directly instruct to take stronger protection measures.

[0103] S7. During the period when the upper-layer policy scheduler updates the policy, the lower-layer real-time controller continuously maintains the link communication with a preset high-power and low-rate configuration. After the link signal quality improves, it gradually reduces the transmission power. At the same time, if the upper-layer policy scheduler instructs to maintain a high-reliability configuration, the lower-layer real-time controller stops the power recovery action according to the upper-layer instruction to prioritize ensuring the link stability.

[0104] Specifically, the S7 includes the following sub-steps S71 - S74.

[0105] S71. During the period when the upper-layer policy scheduler enters the link regulation policy update, the lower-layer real-time controller maintains the current high-power configuration and low-rate modulation and coding scheme to ensure the continuity and reliability of the link communication; S72. During the link monitoring process, the lower-layer real-time controller detects the recovery of the link SNR. If the signal quality improves and is higher than the set recovery margin threshold, it independently performs the action of gradually reducing the transmission power and monitors the change of the link quality; S73. When the lower-layer real-time controller receives the instruction from the upper-layer policy scheduler to maintain the high-reliability setting, it stops the independent recovery action and maintains the current high-power and low-rate configuration until it receives a new recovery instruction; S74. During the period of maintaining a high-reliability configuration, the lower-layer real-time controller continuously monitors the link status and periodically reports the signal status metrics to the upper-layer policy scheduler to assist the upper layer in subsequent decision-making. Among them, the signal status metrics include signal-to-noise ratio, bit error rate, and frame loss rate.

[0106] In S7, after the lower-layer real-time controller completes the emergency response and sends an emergency status notification to the upper layer, if the upper-layer policy scheduler is still in the process of policy update, the lower-layer real-time controller needs to enter the stable holding mode. In this mode, the lower-layer real-time controller maintains a high transmit power output and a low-order modulation and coding scheme according to the current link configuration to ensure continuous communication of the link during the stage of drastic environmental changes. For example, when the transmit power of the link is increased to 24 dBm and the modulation mode is reduced to QPSK 1 / 2 after a sudden anomaly, the lower-layer real-time controller continues to maintain this high-reliability configuration without active adjustment before receiving a new instruction, ensuring that the link has sufficient anti-attenuation ability.

[0107] Meanwhile, the lower-layer real-time controller continuously monitors key metrics such as the signal-to-noise ratio and bit error rate of the link. When it is detected that the signal-to-noise ratio of the link recovers and stabilizes above the preset recovery margin threshold (such as recovering to more than the target margin +2 dB and lasting for a set threshold, for example, more than 3 seconds), the autonomous progressive recovery logic is triggered. Specifically, the lower-layer real-time controller gradually reduces the transmit power in a set small step (such as reducing the power by 0.5 dB each time), and at the same time, after each power adjustment, it re-detects the signal-to-noise ratio and bit error rate of the link to confirm whether the link status continues to be maintained within the safe range.

[0108] S8. When the link signal quality significantly recovers and stabilizes, the middle-layer predictor re-predicts the link status based on the new input features, and the upper-layer policy scheduler makes a decision to resume the normal policy according to the middle-layer prediction result and the link status, and issues an instruction to the lower-layer real-time controller to reduce the transmit power and restore the original modulation and coding scheme.

[0109] In S8, first, the middle-layer predictor re-performs the link status prediction based on the new link input feature set. The input feature set is based on the data collected by the link monitoring unit during the signal recovery period, and after being organized and statistically processed through a set sliding time window, it includes indicators such as the latest signal-to-noise ratio mean, bit error rate change trend, frame loss rate change rate, current transmit power level, and modulation and coding scheme. By re-inputting it into the pre-trained random forest model, the middle-layer predictor can output prediction information such as whether the link margin will continue to be maintained well in a certain period in the future and whether there is a risk of deterioration again.

[0110] For example, in a certain transit mission, after the satellite communication link passes through the mountain shadow area, the signal-to-noise ratio rises from 12 dB to 16 dB and remains stable for more than 5 seconds. The middle-layer predictor re-predicts based on the new link data and concludes that the link failure risk in the next 10 seconds is extremely low (less than 5%), and the link margin estimate continues to be above the safety threshold, providing a prediction basis for the upper-layer policy scheduler to resume the regular communication policy.

[0111] Subsequently, the upper-layer policy scheduler obtains the latest link state prediction results output by the middle-layer predictor within the set decision cycle and conducts a comprehensive evaluation in combination with the real-time link state reported by the lower-layer real-time controller (such as the current SNR, BER, FER). If the link signal quality meets the recovery conditions, for example, the signal-to-noise ratio is stable above the recovery margin, the bit error rate is lower than the set threshold, and the prediction shows no major risks in the future, the upper-layer policy scheduler determines that it can switch to resume the regular communication policy.

[0112] Based on this decision, the upper-layer policy scheduler generates a recovery instruction and sends it to the lower-layer real-time controller. The recovery instruction at least includes an instruction to reduce the transmit power to the target normal level and an instruction to increase the modulation and coding scheme rate to the original high-efficiency mode. For example, if the transmit power is increased to 24 dBm and the modulation method is reduced to QPSK 1 / 2 during the burst response, the upper-layer policy scheduler issues an instruction to gradually reduce the power to 20 dBm and gradually restore the modulation method to the original configuration of 16QAM or 64QAM.

[0113] During the instruction issuance process, to avoid new severe fluctuations during the link recovery process, progressive adjustment parameters can be set in the recovery instruction, such as the power reduction not exceeding 1 dB each time, and the modulation and coding increase only by one level each time, and at the same time, attach the link index monitoring requirements during the recovery process to ensure that the link quality is verified after each adjustment, and if an abnormality occurs, it can be immediately rolled back.

[0114] S9. After receiving the recovery instruction issued by the upper-layer policy scheduler, the lower-layer real-time controller performs progressive recovery actions to gradually reduce the transmit power and increase the modulation and coding rate until the link is restored to the regular communication mode.

[0115] Specifically, S9 includes the following sub-steps: S91. After receiving the recovery configuration instruction issued by the upper-layer policy scheduler, the lower-layer real-time controller gradually reduces the transmit power according to the set progressive recovery strategy and the preset step size, with the power reduction amplitude each time being less than or equal to 1 dB, and continuously monitors the link signal quality after each power adjustment; S92. During the power reduction process, if the link signal-to-noise ratio (SNR) still remains higher than the set recovery margin threshold and the bit error rate (BER) and frame loss rate are within the normal range, then continue to perform the next power reduction operation until the target transmit power configuration is reached; S93. When the transmit power has been restored to the target power configuration, the lower-layer real-time controller gradually increases the modulation and coding scheme rate according to the set recovery rules, increasing the modulation order or coding rate by one level each time, and monitors the change in the link BER after each switch; S94. If during the recovery process, the link SNR drops below the recovery margin threshold or the BER rises sharply, then immediately abort the current recovery operation and roll back to the previous stable configuration state to ensure stable link communication; S95. After the power recovery and modulation rate recovery are completed, the lower-layer real-time controller notifies the upper-layer policy scheduler of the link recovery completion status and enters the normal communication mode.

[0116] In sub-step S91, after receiving the recovery configuration instruction sent by the upper-layer policy scheduler, the lower-layer real-time controller gradually reduces the transmit power according to the set progressive recovery strategy and at a preset step size. The amplitude of each power adjustment does not exceed 1 dB, and the link signal quality is continuously monitored after each power reduction. For example, during a link burst, the transmit power is increased to 24 dBm. During the recovery process, it is reduced by 0.5 dB each time. After 200 milliseconds of monitoring and confirmation that neither the SNR nor the BER has deteriorated, the next power reduction is then performed. By making small-step adjustments and real-time monitoring, the problem of the link re-deteriorating due to too rapid power recovery can be effectively avoided.

[0117] In sub-step S92, if during the power reduction process, the link SNR still remains higher than the set recovery margin threshold (e.g., higher than 12 dB), and the BER and frame loss rate are within the normal range (such as BER less than 0.01 and FER less than 0.02), then the lower-layer real-time controller continues to perform the next power reduction operation. If the link metrics continue to meet the requirements, the system can complete multiple step adjustments continuously and finally restore the transmit power to the target configuration. For example, during a certain actual recovery process, the link SNR remained above 14 dB after each 0.5 dB drop in power. The system continuously performed 8 step adjustments and finally restored the power from 24 dBm to the set target of 20 dBm.

[0118] In sub-step S93, when the transmission power has successfully recovered to the target configuration, the lower-layer real-time controller gradually increases the modulation and coding scheme rate according to the recovery rule set in the recovery instruction. Each increase only raises one modulation order or coding rate level, and the change in the link bit error rate is monitored again after each switch. For example, it is increased from QPSK 1 / 2 to 16QAM 1 / 2 and then gradually transitions to 16QAM 3 / 4. After each increase, the BER needs to be maintained within a stable range for at least 5 consecutive sampling periods (e.g., every 200 ms of sampling) before the next rate increase can continue, ensuring a smooth transition of the link during the modulation switch.

[0119] In sub-step S94, if the link signal-to-noise ratio drops below the recovery margin threshold during the recovery process, or the bit error rate rises sharply beyond the set tolerance range (e.g., the BER suddenly increases to more than 0.03), the current recovery operation is immediately aborted and the system reverts to the previous stable configuration state. For example, during a recovery process, after the modulation mode is increased from 16QAM 1 / 2 to 16QAM 3 / 4, the link bit error rate rises to 0.04, and the lower-layer real-time controller immediately reverts to 16QAM 1 / 2, and the transmission power is maintained at the level of the previous stage before recovery, thus preventing further deterioration of the link quality.

[0120] In sub-step S95, when the transmission power recovery and modulation rate recovery are completed, the lower-layer real-time controller notifies the upper-layer policy scheduler of the link recovery completion status. The recovery completion notification includes the current link signal-to-noise ratio, bit error rate, frame loss rate, final transmission power level, and modulation and coding scheme configuration. Through the status report, the upper-layer policy scheduler can confirm that the link has returned to the normal communication mode and is re-included in the normal regulation and management cycle. For example, in a certain satellite mission, after the link recovery is completed, the lower-layer real-time controller reports that the transmission power has recovered to 20 dBm, the modulation mode has recovered to 64QAM 3 / 4, the SNR is stable above 16 dB, and the BER is below 0.005, indicating that the link has been fully restored to the task expectation standard.

[0121] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0122] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In practical applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0123] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A satellite link dynamic regulation device, characterized in that, It includes an upper-layer policy scheduler, a middle-layer predictor, and a lower-layer real-time controller; The upper-layer policy scheduler is used to obtain the link state prediction result output by the middle-layer predictor and the current link state information reported by the lower-layer real-time controller, and select and execute a link regulation policy based on task requirements. Among them, the link regulation policy includes an initial policy, a dynamic regulation policy, and a recovery of normal communication policy, and sends a link parameter configuration instruction to the lower-layer real-time controller; The middle-layer predictor is used to extract an input feature set within a set historical sliding time window based on the link real-time data collected by the lower-layer real-time controller, and input the input feature set into a pre-trained random forest model for link state prediction. Among them, the pre-trained random forest model is constructed based on historical link operation data and trained with whether the classification label triggers a link adjustment instruction as the training target; The lower-layer real-time controller is used to execute transmit power adjustment and / or modulation and coding scheme adjustment according to the link parameter configuration instruction issued by the upper-layer policy scheduler, and perform emergency power boost and modulation degradation operations to stabilize link communication when it detects that the link signal-to-noise ratio is lower than the burst threshold within a preset burst response duration. At the same time, it feeds back the emergency state information to the upper-layer policy scheduler to perform progressive power recovery and modulation mode improvement after receiving the recovery configuration instruction issued by the upper-layer policy scheduler, and complete the fallback of the link to the normal communication mode.

2. The satellite link dynamic regulation device according to claim 1, wherein The above-mentioned upper-layer policy scheduler includes: A status receiving unit, which is used to receive the link state prediction result output by the middle-layer predictor and the current link operation state information reported by the lower-layer real-time controller; A policy selection unit, which is used to select the current link regulation policy to be executed from the initial policy, the dynamic regulation policy, and the recovery of normal communication policy according to the task requirements and the comprehensive link state information; A decision generation unit, which is used to generate a transmit power configuration instruction and / or a modulation and coding scheme configuration instruction based on the selected link regulation policy; A configuration sending unit, which is used to send the transmit power configuration instruction and / or the modulation and coding scheme configuration instruction to the lower-layer real-time controller to implement the execution of the link regulation instruction; A policy update unit, which is used to enter a fast decision-making loop after receiving an emergency state notification from the lower-layer real-time controller, and re-evaluate and update the executed link regulation policy based on the real-time state.

3. The satellite link dynamic regulation device according to claim 2, wherein The above-mentioned middle-layer predictor includes: A feature extraction unit, which is used to extract an input feature set within a set historical sliding time window based on the received link real-time data. The input feature set covers the statistical features of non-high-frequency and fast-fluctuating data, including received signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, satellite elevation angle; A model inference unit, which is used to input the input feature set into the pre-trained random forest model and output a link state prediction result, including a predicted value of the link margin or a link failure risk level within a future time period; A confidence evaluation unit, which is used to evaluate the confidence of the output result of the model inference unit and send the prediction result and the confidence index to the upper-layer policy scheduler together to assist in policy selection; A prediction update interface unit is used to provide an interface for training data collection and offline model update to support the periodic training optimization of the random forest model based on newly collected link operation data.

4. The satellite link dynamic regulation device according to claim 3, characterized in that, The lower-layer real-time controller includes: A link monitoring unit, which is used to collect link signal status data at a preset time interval, including signal-to-noise ratio, bit error rate, and frame loss rate, and report the data to the middle-layer predictor and the upper-layer policy scheduler; An instruction execution unit, which is used to control the transmitter to perform corresponding parameter adjustments according to the transmit power configuration instruction and / or modulation and coding scheme configuration instruction issued by the upper-layer policy scheduler; An emergency event detection unit, which is used to judge whether the link signal-to-noise ratio drops below the emergency threshold within the emergency response duration, and trigger an emergency adjustment when the condition is met; An emergency adjustment unit, which is used to immediately increase the transmit power to the upper limit when a link anomaly is detected, switch to a low-order modulation and coding scheme to stabilize link communication, and feedback the emergency status information to the upper-layer policy scheduler; A progressive recovery control unit, which is used to gradually reduce the transmit power and restore the high-order modulation scheme at a set rhythm after receiving the recovery configuration instruction, and fallback to the normal communication mode on the premise of ensuring link stability.

5. A method for dynamically regulating a satellite link, characterized in that For the satellite link dynamic regulation device as described in any one of claims 1-4, it includes the following steps: S1. When a satellite transit starts or a link is established, the upper-layer policy scheduler sets an initial policy according to the task requirements, and issues an initial transmit power configuration, an initial modulation and coding scheme configuration, and an initial margin threshold parameter to the lower-layer real-time controller. The lower-layer real-time controller sets the initial transmit power and the initial modulation method according to the initial configuration and starts link monitoring; among them, the execution policies of the upper-layer policy scheduler include an initial policy, a dynamic regulation policy, and a recovery to normal communication policy; S2. The lower-layer real-time controller continuously collects link real-time data and organizes it into an input feature set. The middle-layer predictor continuously performs link state prediction based on the input feature set. The upper-layer policy scheduler obtains the link state prediction result output by the middle-layer predictor and the current link state reported by the lower-layer real-time controller at a set decision period. When the predicted link margin value is lower than the preset safety threshold, the upper-layer policy scheduler pre-adjusts the link configuration policy; S3. The upper-layer policy scheduler issues an adjustment instruction to the lower-layer real-time controller according to the pre-adjusted link configuration policy to adjust the transmit power configuration and / or the modulation and coding scheme configuration; S4. After the lower-layer real-time controller adjusts the link parameters according to the adjustment instruction issued by the upper-layer policy scheduler, it performs a fine power adjustment based on the current link measurement result to callback the actual signal-to-noise ratio to near the target margin, and sets the preset threshold for emergency event detection; S5. The lower-layer real-time controller detects whether the signal-to-noise ratio is lower than the preset threshold within the emergency duration in the link monitoring loop. If it is lower, it increases the transmit power to the upper limit value and switches to a low-order modulation and coding scheme to stabilize link communication; S6. After the lower-layer real-time controller completes the burst response, it sends an emergency status notification to the upper-layer policy scheduler. After receiving the emergency notification, the upper-layer policy scheduler immediately enters the decision-making loop to update the link regulation policy; S7. During the period when the upper-layer policy scheduler updates the policy, the lower-layer real-time controller continuously maintains the link communication with a preset high-power and low-rate configuration. After the link signal quality improves, it gradually reduces the transmission power. At the same time, if the upper-layer policy scheduler instructs to maintain the high-reliability configuration, the lower-layer real-time controller stops the power recovery action according to the upper-layer instruction to prioritize ensuring the link stability; S8. When the link signal quality significantly rebounds and stabilizes, the middle-layer predictor re-predicts the link state based on the new input features. The upper-layer policy scheduler makes a decision to restore the normal policy according to the middle-layer prediction result and the link state, and issues an instruction to the lower-layer real-time controller to reduce the transmission power and restore the original modulation and coding scheme; S9. After receiving the restoration instruction issued by the upper-layer policy scheduler, the lower-layer real-time controller performs a progressive restoration action to gradually reduce the transmission power and increase the modulation and coding rate until the link is restored to the normal communication mode.

6. The satellite link dynamic regulation method according to claim 5, characterized in that The above S2 includes the following sub-steps: S201. The lower-layer real-time controller continuously collects the link real-time data at a preset sampling interval, and organizes and statistics the link real-time data according to the set sliding time window to obtain an input feature set including indicators such as link signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmission power level, satellite elevation angle, etc. Among them, the length of the sliding time window is 1 second to 5 seconds, the step size is 0.5 seconds, and the input feature set does not include high-frequency and fast-fluctuating data that requires millisecond-level response; S202. The middle-layer predictor inputs the input feature set into a pre-trained random forest model. The random forest model is trained based on historical link operation data and uses whether to trigger a link regulation instruction as the classification label to perform link state prediction and output the prediction result of the link margin state within a preset future time period; S203. The middle-layer predictor calculates the link failure risk level and / or link safety margin according to the prediction result, and sends the prediction result and the confidence index to the upper-layer policy scheduler together for assisting in the selection and update of the link regulation policy.

7. The satellite link dynamic regulation method according to claim 5, characterized in that The above S2 includes the following sub-steps: S211. The upper-layer policy scheduler receives the link state prediction result output by the middle-layer predictor and the current link state information reported by the lower-layer real-time controller. The link state prediction result includes the link margin change trend within a preset future time period, and the current link state information includes real-time signal-to-noise ratio, bit error rate, frame loss rate, and transmission power level; S212. The upper-layer policy scheduler performs a trend analysis on the link state prediction result to judge whether there is a risk of insufficient future link margin and the degree of link quality deterioration; S213. The upper-layer policy scheduler performs a real-time evaluation on the current link state information to judge whether the current link has approached the failure boundary, including whether there is a situation where the signal-to-noise ratio is close to the lowest decoding threshold or the bit error rate rises sharply; S214. The upper-layer policy scheduler jointly determines the trend analysis result and the real-time evaluation result. If the predicted result and / or the current state indicates that the link margin is lower than the preset safety threshold, it generates a link pre-adjustment configuration policy to determine whether to adjust the transmit power and / or switch the modulation and coding scheme. S215. The upper-layer policy scheduler formulates the adjustment amplitude and adjustment priority based on the pre-adjustment configuration policy and prepares to send the corresponding link parameter configuration instruction to the lower-layer real-time controller.

8. The satellite link dynamic regulation method according to claim 7, wherein The pre-training steps of the random forest model include: Construct a training data set based on the historical link operation data collected during multiple satellite transit cycles. The historical link operation data includes link signal strength, signal-to-noise ratio, bit error rate, modulation and coding scheme, transmit power level, and satellite elevation angle, and perform feature extraction on the data according to a set time window to form an input feature vector. According to the link regulation records corresponding to the historical link data, mark whether a link parameter adjustment instruction has been triggered as a classification label, where the link parameter adjustment instruction includes a transmit power adjustment instruction and a modulation and coding scheme switching instruction. Construct a supervised learning sample using the input feature vector and the classification label, and train a random forest classification model so that the model can output a prediction result on whether there is a regulation requirement for the link within a preset time period in the future. Perform cross-validation and generalization ability evaluation on the trained random forest model to determine the optimal model parameter configuration, including the number of decision trees, the maximum tree depth, and the feature selection strategy.

9. The satellite link dynamic regulation method according to claim 5, characterized in that S7 includes the following sub-steps: S71. During the period when the upper-layer policy scheduler enters the link regulation policy update, the lower-layer real-time controller maintains the current high-power configuration and low-rate modulation and coding scheme to ensure the continuity and reliability of link communication. S72. The lower-layer real-time controller detects the recovery of the link signal-to-noise ratio during the link monitoring process. If the signal quality improves and is higher than the set recovery margin threshold, it autonomously performs the action of gradually reducing the transmit power and monitors the change of the link quality. S73. When the lower-layer real-time controller receives the instruction from the upper-layer policy scheduler to maintain the high-reliability setting, it stops the autonomous recovery action and maintains the current high-power and low-rate configuration until it receives a new recovery instruction. S74. During the period of maintaining the high-reliability configuration, the lower-layer real-time controller continuously monitors the link state and periodically reports the signal state indicators to the upper-layer policy scheduler to assist the upper layer in subsequent decision-making. Among them, the signal state indicators include signal-to-noise ratio, bit error rate, and frame loss rate.

10. The satellite link dynamic regulation method according to claim 3, characterized in that S9 includes the following sub-steps: S91. After receiving the recovery configuration instruction sent by the upper-layer policy scheduler, the lower-layer real-time controller gradually reduces the transmit power according to the set progressive recovery strategy and the preset step amplitude. The amplitude of each power reduction is less than or equal to 1 dB, and it continuously monitors the link signal quality after each power adjustment. S92. During the power reduction process, if the link signal-to-noise ratio still remains higher than the set recovery margin threshold, and the bit error rate and frame loss rate are within the normal range, continue to perform the next power reduction operation until the target transmit power configuration is reached. S93. After the transmission power is restored to the target power configuration, the lower-layer real-time controller gradually increases the modulation and coding scheme rate according to the set restoration rules, increasing the modulation order or coding rate by one level each time, and monitoring the change in the link bit error rate after each switch; S94. If during the restoration process, the link signal-to-noise ratio drops below the restoration margin threshold or the bit error rate rises sharply, the current restoration operation is immediately aborted and the system reverts to the previous stable configuration state to ensure stable link communication; S95. After completing the power restoration and modulation rate restoration, the lower-layer real-time controller notifies the upper-layer policy scheduler of the completion of the link restoration and enters the normal communication mode.

Images