A resource allocation system in a satellite-ground integrated network and a method thereof

Abstract

The application relates to the technical field of communication networks, in particular to a resource allocation system and method in a satellite-ground integrated network, which comprises a satellite-ground topology and channel evolution predictor, acquires satellite ephemeris, user GNSS positioning and movement vectors, ground base station topology and real-time CQI / RSRP measurement reports, generates future link state information, a service QoE impact and switching cost modeler, calculates transient interruption time of a switching action, generates switching cost information, a continuity risk quantization engine, generates continuity risk information containing the holding link risk probability, the instantaneous impact collapse probability and the switching link risk probability, a risk-avoiding resource decision maker, calculates holding link comprehensive cost and switching link comprehensive cost, calculates a dynamic switching damping threshold, and generates risk-avoiding decision information; the application effectively suppresses high-frequency ping-pong switching near edge state points and reduces signaling overhead caused by invalid switching.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication network, in particular to a resource allocation system and method in a satellite-terrestrial integrated network. BACKGROUND

[0002] With the rapid development of satellite-terrestrial integrated network, providing seamless coverage and high-quality service experience for high dynamic mobile users has become a key technical challenge, and ensuring service continuity is one of the core goals. In the integrated network, the user terminal needs to switch between the dynamically changing satellite link and the ground base station link.

[0003] However, the existing resource allocation and switching technology has significant limitations. Traditional decision-making mainly relies on instantaneous network indicators such as signal-to-noise ratio or throughput, and lacks forward-looking prediction of the future evolution state of the link, which is a short-sighted and passive response mechanism. Existing technologies generally ignore the transient transition overhead of switching actions, i.e. the physical interruption time caused by the switching process and its instantaneous impact cost on the QoE of specific services. This short-sighted decision-making mechanism based on instantaneous thresholds is prone to frequent ping-pong switching in the marginal area of link quality. This invalid switching not only increases the network signaling burden, but also seriously damages the fundamental value of user service continuity, causing service stalls or interruptions, and leading to a serious disconnection between network indicators and user experience. SUMMARY

[0004] To solve the above technical problems, the present application provides a resource allocation system and method in a satellite-terrestrial integrated network. Specifically, the technical solution of the present application is as follows:

[0005] A resource allocation system in a satellite-terrestrial integrated network, the system comprising:

[0006] A satellite-terrestrial topology and channel evolution predictor for obtaining satellite ephemeris, user GNSS positioning and movement vector, ground base station topology, and real-time CQI / RSRP measurement report; combining a pre-set orbit dynamics model and a pre-set wireless channel prediction model to predict the future state of all available links of the user, to generate future link state information;

[0007] A service QoE impact and switching cost modeler for obtaining user service flow type and QoE requirements; calculating the transient interruption time of switching actions; quantifying the QoE instantaneous drop value of specific services based on the transient interruption time; and generating switching cost information based on the QoE instantaneous drop value;

[0008] A continuity risk quantification engine configured to calculate a risk probability of staying with a current link based on the future link state information, the handover cost information, and a preset QoE acceptable threshold; calculate an instantaneous impact collapse probability; calculate a risk probability of switching to a target link; and generate continuity risk information including the risk probability of staying with the link, the instantaneous impact collapse probability, and the risk probability of switching to the link;

[0009] A risk-averse resource decision maker configured to calculate a staying link comprehensive cost and a switching link comprehensive cost based on the continuity risk information, a preset service priority weight, and a preset normalized resource cost; calculate a dynamic handover damping threshold based on the instantaneous impact collapse probability; and perform a decision based on the staying link comprehensive cost, the switching link comprehensive cost, and the dynamic handover damping threshold to generate risk-averse decision information.

[0010] Further, the future link state information includes a predicted signal-to-noise ratio, a predicted latency, and a predicted packet loss rate; the handover cost information includes the transient interruption time and the QoE instantaneous drop value; the continuity risk information includes the risk probability of staying with the link, the instantaneous impact collapse probability, and the risk probability of switching to the link; and the risk-averse decision information includes the staying link comprehensive cost, the switching link comprehensive cost, the dynamic handover damping threshold, and an optimal decision instruction.

[0011] Further, the satellite-to-ground topology and channel evolution predictor includes:

[0012] A data aggregation submodule configured to acquire satellite ephemeris data, terminal GNSS positioning and movement vectors, ground base station topology information, and real-time CQI / RSRP measurement reports;

[0013] A channel evolution prediction submodule configured to fuse a preset orbit dynamics model and a preset wireless channel prediction model based on the reports, the ephemeris data, the positioning and movement vectors, and the topology information to predict time series of signal-to-noise ratios, latencies, and packet loss rates of all available links in a future preset time;

[0014] A state information generation submodule configured to generate the future link state information based on the time series.

[0015] Further, the service QoE impact and handover cost modeler includes:

[0016] An interruption time calculation submodule configured to acquire a service flow type and calculate a transient interruption time based on a preset signaling interaction time, a preset physical layer synchronization time, and a preset data buffer reconstruction time;

[0017] a QoE impact modeling submodule configured to invoke a preset QoE mapping function specific to the service flow type, and calculate a QoE instantaneous drop value based on the transient interruption time and an estimated expected packet loss rate of a handover procedure;

[0018] a handover cost generation submodule configured to generate the handover cost information based on the transient interruption time and the QoE instantaneous drop value.

[0019] Further, the continuity risk quantification engine comprises:

[0020] a steady-state QoE prediction submodule configured to acquire the future link state information, and invoke a preset steady-state QoE prediction function to convert the steady-state QoE prediction value into a future steady-state QoE prediction value;

[0021] a risk probability calculation submodule configured to calculate a probability that the QoE is lower than a preset QoE acceptable threshold when the current link is maintained, to obtain a link-maintained risk probability, based on the future steady-state QoE prediction value; calculate a probability that the QoE collapses due to an instantaneous impact, to obtain an instantaneous impact collapse probability, based on the current QoE, the QoE instantaneous drop value and the QoE acceptable threshold; and calculate a probability that the future QoE is lower than the QoE acceptable threshold after the handover, to obtain a post-handover future collapse probability, based on the future steady-state QoE prediction value;

[0022] a risk information generation submodule configured to calculate a handover link risk probability by combining the instantaneous impact collapse probability and the post-handover future collapse probability, and generate the continuity risk information comprising the link-maintained risk probability, the instantaneous impact collapse probability and the handover link risk probability.

[0023] Further, the risk-avoiding resource decision maker comprises:

[0024] a comprehensive cost calculation submodule configured to acquire the link-maintained risk probability and the handover link risk probability, and calculate a link-maintained comprehensive cost and a handover link comprehensive cost respectively by combining a preset service priority weight and a preset normalized resource cost;

[0025] a damping threshold calculation submodule configured to acquire the instantaneous impact collapse probability, and calculate a dynamic handover damping threshold by combining a preset system sensitivity coefficient and the service priority weight;

[0026] a decision generation submodule configured to determine whether the handover link comprehensive cost is less than the link-maintained comprehensive cost minus the dynamic handover damping threshold, determine an optimal decision instruction as handover if the determination is positive, and determine the optimal decision instruction as link maintenance if the determination is negative, and generate the risk-avoiding decision information based on the optimal decision instruction.

[0027] Further, the wireless channel prediction model is a long short-term memory network model or a Markov model.

[0028] Further, the system is deployed at a ground network core network side, a MEC node or a satellite network ground gateway station.

[0029] A resource allocation method in a satellite-ground integrated network, comprising the following steps:

[0030] Obtaining satellite ephemeris, user GNSS positioning and movement vector, ground base station topology and real-time CQI / RSRP measurement report; fusing a preset orbit dynamics model and a preset wireless channel prediction model to predict the future state of all available links of the user to generate future link state information;

[0031] Obtaining user traffic flow type and QoE requirement; calculating transient interruption time of switching action; quantifying QoE instantaneous drop value of specific traffic based on the transient interruption time; and generating switching cost information based on the QoE instantaneous drop value;

[0032] Based on the future link state information, the switching cost information and a preset QoE acceptable threshold, calculating the risk probability of keeping the current link; calculating the instantaneous impact collapse probability; calculating the risk probability of switching to the target link; and generating continuity risk information containing the risk probability of keeping the link, the instantaneous impact collapse probability and the risk probability of switching the link;

[0033] Based on the continuity risk information, a preset traffic priority weight and a preset normalized resource cost, calculating the comprehensive cost of keeping the link and the comprehensive cost of switching the link; based on the instantaneous impact collapse probability, calculating a dynamic switching damping threshold; based on the comprehensive cost of keeping the link, the comprehensive cost of switching the link and the dynamic switching damping threshold, performing decision to generate risk avoidance decision information.

[0034] Compared with the prior art, the present application has the following beneficial effects:

[0035] 1. The present application realizes the transition from passive response to active avoidance, and through the satellite-ground topology and channel evolution predictor, the basis of resource decision is changed from the current instantaneous measurement value to the evolution curve of the future link state; this forward-looking prediction enables the system to actively perform switching to avoid the foreseeable QoE collapse risk in the future before the user enters the shadow area or the channel deteriorates, instead of waiting for signal loss and then responding passively;

[0036] 2. The application avoids instantaneous impact collapse, guarantees QoE continuity, quantifies the transient interruption time and its QoE instantaneous drop value for specific services caused by switching action through service QoE impact and switching cost modeler; the continuity risk quantization engine calculates the instantaneous impact collapse probability, decomposes the total switching risk into instantaneous impact and future channel risk, thereby avoiding switching to a link with good future channel but the switching action itself will cause service collapse, and guaranteeing user experience without freezing;

[0037] 3. The application effectively suppresses ping-pong switching and enhances decision robustness, and constructs a risk-averse resource decision-making closed loop to avoid QoE collapse risk; it calculates a dynamic switching damping threshold based on the instantaneous impact collapse probability; when the instantaneous impact risk is high, the dynamic threshold will increase significantly, requiring the net benefit brought by switching to be large enough to trigger, thereby effectively suppressing high-frequency ping-pong switching near the edge state point and reducing the signaling overhead caused by invalid switching;

[0038] 4. The application intelligently balances between QoE risk and resource cost, calculates the dimensionless comprehensive cost by uniformly weighting the risk probability and resource cost of the maintained link and the switching link through the comprehensive cost calculation submodule; the decision maker weighs between the risk cost and network resource consumption cost to guarantee user QoE continuity, so that the system can reject the link that is instantaneous optimal but unstable, and realize the resource optimal decision-making aiming at risk avoidance. BRIEF DESCRIPTION OF DRAWINGS

[0039] The application will be further explained in combination with the drawings and embodiments:

[0040] Figure 1 is a flowchart of the method of the application;

[0041] Figure 2 is a structural diagram of the system of the application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and advantages of the application clearer, the application will be further described in detail below in combination with specific embodiments.

[0043] Embodiment 1:

[0044] Please refer to Figure 1 A resource allocation system in a star-ground fusion network, the system comprising:

[0045] A satellite-terrestrial topology and channel evolution predictor is configured to obtain satellite ephemeris, user GNSS positioning and movement vectors, ground base station topology, and real-time CQI / RSRP measurement reports; combine a preset orbit dynamics model and a preset wireless channel prediction model to predict future states of all available links of a user to generate future link state information;

[0046] A service QoE impact and switching cost modeler is configured to obtain service flow types and QoE requirements of a user; calculate transient interruption time of a switching action; quantify a QoE instantaneous drop value of a specific service based on the transient interruption time; and generate switching cost information based on the QoE instantaneous drop value;

[0047] A continuity risk quantification engine is configured to calculate a risk probability of keeping a current link, calculate an instantaneous impact collapse probability, calculate a risk probability of switching to a target link, and generate continuity risk information including the risk probability of keeping the link, the instantaneous impact collapse probability, and the risk probability of switching the link based on the future link state information, the switching cost information, and a preset QoE acceptable threshold determined according to a service level agreement of service k or a related industry standard, such as a MOS score defined by ITU-T for VoIP;

[0048] A risk-averse resource decision maker is configured to calculate a keeping link comprehensive cost and a switching link comprehensive cost based on the continuity risk information, a preset service priority weight, and a preset normalized resource cost; calculate a dynamic switching damping threshold based on the instantaneous impact collapse probability; and perform a decision based on the keeping link comprehensive cost, the switching link comprehensive cost, and the dynamic switching damping threshold to generate risk-averse decision information.

[0049] The present application provides a resource allocation system in a satellite-terrestrial integrated network, which aims to solve the technical problem that ping-pong switching caused by pursuing instantaneous apparent indicators such as SINR or throughput seriously damages the fundamental value of user service continuity in the prior art; the system replaces the traditional short-sighted trigger based on instantaneous threshold by constructing a forward-looking decision-making closed loop aiming at avoiding QoE collapse risk;

[0050] In the embodiment, the system includes four core modules deployed on a MEC node of a ground network to realize low-delay decision-making.

[0051] A satellite-terrestrial topology and channel evolution predictor: the purpose of the predictor is to change the basis of resource decision-making from current measurement values to future evolution curves, realizing the transition from passive response to active avoidance;

[0052] In this embodiment, the satellite-to-ground topology and channel evolution predictor acquires satellite ephemeris, user GNSS positioning and movement vectors, ground base station topology, and real-time CQI / RSRP measurement reports in real time through a data aggregation interface.

[0053] After obtaining the above data, the predictor fuses two preset models:

[0054] The pre-defined orbital dynamics model refers to a deterministic model based on Keplerian mechanics and satellite orbital parameters; its function is to accurately predict a pre-defined prediction window in the future using satellite ephemeris and user GNSS positioning. Within, the geometry of all visible satellite links, including elevation angle, Doppler shift, and entry and exit times of the keyhole effect region;

[0055] The preset wireless channel prediction model refers to a probabilistic model trained based on historical data, such as an LSTM trained based on historical channel data and corresponding movement vectors. Its function is to use the time series of real-time CQI / RSRP measurement reports and user movement vectors to predict the evolution of the wireless environment of the terrestrial link, especially obstruction events such as entering urban canyons or tunnels.

[0056] Based on this fusion model, the predictor forecasts the future state of all available links for the user to generate structured future link state information; this information is packaged into a set of future-oriented state curves. ,in And output it to the continuous risk quantification engine;

[0057] Business QoE Impact and Switching Cost Modeler: The purpose of this modeler is to accurately quantify the transient transfer overhead that exists in the switching action itself and is ignored in the existing technology, that is, the hidden physical cost of the switching action to the user experience.

[0058] In this embodiment, the service QoE impact and switching cost modeler acquires the user service flow type during operation. and the QoE requirements of this business ;

[0059] This module performs calculations on the transient interruption time of the switching action. Transient interruption time Refers to the source link Release to target link During the availability period, the physical time of a complete interruption of the business flow is calculated using the following model:

[0060]

[0061] in, : total interruption time; unit: second; the result of this formula

[0062] : signaling interaction time required for handover; unit: second; according to the type of service, obtained from network configuration parameters;

[0063] : time required for terminal to complete physical layer synchronization on new link; unit: second; determined according to terminal hardware specification and 3GPP specification;

[0064] : time required for terminal side data buffer to be emptied, reconstructed and data packets reordered; unit: second; determined according to protocol stack standard;

[0065] After obtaining , this module quantifies the QoE instantaneous drop value of specific service based on transient interruption time ; the QoE instantaneous drop value refers to the instantaneous decrease value of user-perceived quality of experience caused by , which is quantified by a QoE impact model specific to service :

[0066]

[0067] wherein, : service type, obtained from PCF;

[0068] : QoE mapping function specific to service , such as E-Model of VoIP or proxy model trained for video service through offline data set; predefined or pre-trained by system;

[0069] : transient interruption time calculated in previous step;

[0070] : expected additional packet loss rate during handover ; estimated according to protocol state and initial synchronization quality of target link;

[0071] The innovation of this formula lies in that it maps the interruption time of physical layer to the experience loss of user-perceived layer; based on the QoE instantaneous drop value, this module generates handover cost information containing and , and outputs to subsequent module;

[0072] Continuous Risk Quantification Engine: The purpose of this engine is to provide a quantitative probabilistic basis for whether to switch or maintain the core issue of this invention; it no longer compares instantaneous SINR, but compares the total risk probability of future QoE collapse;

[0073] In this embodiment, the continuity risk quantification engine is based on future link status information and switching cost information. and the preset acceptable QoE threshold Perform calculations;

[0074] To achieve the above quantization, the engine calculates the risk probability of maintaining the current link. ; Risk probability of maintaining the current link This refers to the situation where the system decides to remain on the current link. In the future Within a certain period, the QoE of this link fell below The probability of is calculated using the following model:

[0075]

[0076] in, It is a preset steady-state QoE prediction function. predictor Converted to future QoE score; Based on The integral probability calculated from the predicted confidence interval; specifically, assuming that at any time t, the predicted QoE value is... If the predicted value follows a normal distribution with a mean of the predicted value and a standard deviation of the prediction confidence interval width (e.g., based on the variance estimate of the prediction model), then the probability can be calculated by taking the normal distribution at intervals less than 10 ... The cumulative distribution function value over the interval is obtained. The final... The maximum value of the probability calculated at all times within the time window can be taken to represent the worst-case risk within the time window;

[0077] The engine calculates the probability of instantaneous impact collapse. Probability of instantaneous impact collapse This refers to the effect caused by the switching action itself. Will a sudden drop immediately cause QoE to drop from its current value? Break Its calculation model is as follows:

[0078]

[0079] in, To predict QoE based on real-time measurements of the current link using a steady-state QoE prediction function. The calculated current QoE score, From the modeler;

[0080] The engine calculates the probability of risk when switching to the target link. Risk probability of switching to the target link This refers to the situation where the system decides to switch to the alternative link. The total probability that this leads to a QoE collapse; this definition is one of the key innovations of this invention, as it encompasses two failure paths: transient impact failure. Or the switch will fail in the future. Its calculation model is as follows:

[0081]

[0082] in, This formula is based on probability theory. Here, to simplify the model calculation, the instantaneous impact failure and the failure of the future link state after the handover are treated as approximately independent events. This approximation is reasonable in most scenarios because the instantaneous impact failure is mainly determined by short-term fixed overhead such as signaling interaction and hardware synchronization, while the future link failure is mainly determined by long-term factors such as user movement and environmental occlusion. The correlation between the two is weak, and the total failure risk is that neither fails. The engine generates content including , and The continuous risk information is output to the decision-maker;

[0083] Risk-averse resource decision-maker: The purpose of this decision-maker is to act as the central decision-making module of the system, making the final decision based on risk aversion rather than best efforts; it weighs the QoE risk and resource costs.

[0084] In this embodiment, the risk-averse resource decision-maker is based on continuous risk information. Preset business priority weights and the preset normalized resource cost The normalized resource cost It is a comprehensive measure of the various resources consumed by using link l, such as bandwidth, power, or spectrum costs, for example, through Normalization is performed in the following manner, where It is the original cost function. and The decision is based on the maximum and minimum possible original costs of all links in the system.

[0085] To execute the decision, the decision maker calculates the overall cost of maintaining the link. and the overall cost of switching links Comprehensive cost is a unified, dimensionless decision score for comparing the total cost of different actions:

[0086]

[0087]

[0088] wherein, is a preset hyper-parameter for balancing risk and cost;

[0089] To prevent ping-pong switching, the decision maker calculates a dynamic switching damping threshold based on the instantaneous impact collapse probability ; the dynamic switching damping threshold , also known as switching hysteresis, requires that the net benefit brought by switching must be large enough to trigger; the innovation of the present application lies in that the threshold is dynamic:

[0090]

[0091] wherein, is a preset system sensitivity coefficient;

[0092] The decision maker makes decisions based on the comprehensive cost of maintaining the link, the comprehensive cost of switching the link, and the dynamic switching damping threshold: only when does it perform switching; otherwise, it maintains the current link; the decision result is encapsulated as risk-avoiding decision information and delivered to the network execution layer;

[0093] The system of the embodiment of the present application changes the paradigm of resource allocation by constructing a complete technical closed loop including prediction, cost modeling, risk quantification, and avoidance decision; it is no longer a passive response to instantaneous measurement values, but actively avoids foreseeable future total risks of QoE collapse including transient transition overhead; it realizes the transition from passive response to active avoidance, such as actively switching 5 seconds before entering the tunnel instead of switching after signal loss after entering; it solves the problem of decoupling QoE and network indicators, and the system will deliberately reject instantaneous optimal but unstable links to ensure that the user experience is not stuck; through intelligent switching damping based on the , it effectively suppresses ping-pong switching and significantly reduces the signaling overhead caused by invalid switching;

[0094] It should be noted that the present embodiment mainly focuses on the service continuity optimization of a single user; in a multi-user scenario, the decision result of the present system can be used as the input of the upper-layer network resource manager, and the RRM combines the decision requests of multiple users and the global resource status to perform final resource allocation and admission control, so as to achieve global optimization.

[0095] Embodiment 2:

[0096] The future link state information includes: predicted signal-to-noise ratio, predicted latency and predicted packet loss rate; the switching cost information includes: the transient interruption time and the QoE instantaneous drop value; the continuity risk information includes: the keep link risk probability, the instantaneous impact collapse probability and the switching link risk probability; the risk avoidance decision information includes: the keep link comprehensive cost, the switching link comprehensive cost, the dynamic switching damping threshold and the optimal decision instruction.

[0097] On the basis of embodiment 1, the contents of the information artifacts transmitted between the four modules of the system are specifically defined in this embodiment to ensure the completeness and unambiguity of data delivery;

[0098] The future link state information: its purpose is to provide full-dimensional input required for calculating steady-state QoE; in this embodiment, it specifically includes the time series of predicted signal-to-noise ratio , predicted latency and predicted packet loss rate generated by the predictor; this set of vectors can more accurately map to the QoE score of a specific service than the individual SINR;

[0099] The switching cost information: its purpose is to encapsulate the physical cost and perceptual cost of the switching action; in this embodiment, it specifically includes the transient interruption time and the QoE instantaneous drop value calculated by the modeler;

[0100] The continuity risk information: its purpose is to provide all risk components required for decision-making to the decision maker; in this embodiment, it specifically includes the keep link risk probability , the instantaneous impact collapse probability and the switching link risk probability calculated by the risk quantification engine; the transmission of this triplet is crucial because the decision maker not only needs and to calculate cost, but also needs to independently calculate dynamic damping ;

[0101] The risk avoidance decision information: its purpose is to provide a traceable and complete decision result package; in this embodiment, it not only includes the final optimal decision instruction, but also includes the intermediate evidence for reaching the decision, i.e., the keep link comprehensive cost , the switching link comprehensive cost and the dynamic switching damping threshold ;

[0102] The embodiment ensures the standardization and completeness of data delivery between modules by structuring the definition of the four information artifacts inside the system. In particular, the instantaneous impact collapse probability is explicitly delivered as a separate data item in the continuity risk information, providing indispensable and decoupled input for risk-averse resource decision makers to implement risk-based dynamic damping, thereby enhancing the modularity and scalability of the system. The embodiment ensures the standardization and completeness of data delivery between modules by structuring the definition of the four information artifacts inside the system. In particular, the instantaneous impact collapse probability is explicitly delivered as a separate data item in the continuity risk information, providing indispensable and decoupled input for risk-averse resource decision makers to implement risk-based dynamic damping, thereby enhancing the modularity and scalability of the system.

[0103] Embodiment 3:

[0104] The satellite-ground topology and channel evolution predictor includes:

[0105] The data aggregation submodule is configured to obtain satellite ephemeris data, terminal GNSS positioning and movement vectors, ground base station topology information, and real-time CQI / RSRP measurement reports.

[0106] The channel evolution prediction submodule is configured to, based on the reports, the ephemeris data, the positioning and movement vectors, and the topology information, fuse a preset orbit dynamics model and a preset wireless channel prediction model to predict time series of signal-to-noise ratios, time delays, and packet loss rates of all available links within a preset future time.

[0107] The state information generation submodule is configured to generate the future link state information based on the time series.

[0108] Based on the embodiment 1, the embodiment specifically defines the internal implementation of the satellite-ground topology and channel evolution predictor, and decouples its function into three submodules.

[0109] The predictor includes:

[0110] The data aggregation submodule is configured to serve as a data entry of the system and collect information required for decision making from heterogeneous data sources. In the embodiment, the submodule is configured to obtain satellite ephemeris data from a satellite operation center, GNSS positioning and movement vectors reported by terminals, ground base station topology information in a network operation database, and real-time CQI / RSRP measurement reports reported by a physical layer through dedicated interfaces.

[0111] The channel evolution prediction submodule is configured to execute a core prediction algorithm. In the embodiment, the submodule is configured to, based on all the reports, data, vectors, and information obtained by the data aggregation submodule, fuse a preset orbit dynamics model and a preset wireless channel prediction model to predict time series of signal-to-noise ratios, time delays, and packet loss rates of all available links within a preset future time.

[0112] State information generation submodule: Its purpose is to encapsulate the original prediction results into a standardized output; in this embodiment, this submodule is used to package and format the time series generated by the channel evolution prediction submodule to generate the future link state information of Embodiment 2 for use by the continuity risk quantification engine.

[0113] This embodiment achieves clear architectural decoupling by decomposing the predictor function into three sub-modules: convergence, prediction, and encapsulation. By forcibly fusing two models, orbital dynamics and wireless channel, in the channel evolution prediction sub-module, it significantly improves the comprehensive prediction accuracy of heterogeneous links in the space-ground converged network, providing a high-confidence input for subsequent risk quantification.

[0114] Example 4:

[0115] The business QoE impact and switching cost modeler includes:

[0116] The interruption time calculation submodule is used to obtain the service flow type and calculate the transient interruption time based on the preset signaling interaction time, the preset physical layer synchronization time, and the preset data buffer reconstruction time.

[0117] The QoE impact modeling submodule is used to call a preset QoE mapping function specific to the service flow type; and calculate the instantaneous QoE drop value based on the transient interruption time and the estimated packet loss rate during the handover process.

[0118] The switching cost generation submodule is used to generate the switching cost information based on the transient interruption time and the instantaneous QoE drop value.

[0119] Based on Example 1, this example specifically defines the internal implementation of the business QoE impact and switching cost modeler, decoupling its functionality into three sub-modules:

[0120] The modeler includes:

[0121] Interruption time calculation submodule: Its purpose is to quantify the physical interruption duration of the handover; in this embodiment, this submodule is used to obtain the service flow type and, based on the preset signaling interaction time retrieved from the network configuration library. Preset physical layer synchronization time and preset data buffer rebuild time ,pass The transient interruption time was calculated. ;

[0122] The QoE impact modeling submodule aims to map physical interruption duration into perceived experience loss. In this embodiment, this submodule is used to call a function related to the service flow type. Pre-set QoE mapping function ; the sub-module calculates the transient interruption time inputted by the interruption time calculation sub-module and an estimated expected packet loss rate of the handover process , by calculating the QoE instantaneous drop value ; for example, for a video service, the function can be modeled as a non-linear model that can better reflect the perceptual saturation effect, for example , where A is the maximum QoE drop value, is the characteristic time constant of service perception, and this model can more accurately depict the dose-effect relationship that the user experience deteriorates sharply when the interruption time exceeds a certain threshold; in another embodiment, it can also be simplified as a linear model, such as , where and are impact coefficients obtained by regression analysis;

[0123] The handover cost generation sub-module is used to encapsulate the calculation results; in this embodiment, the sub-module is used to package the transient interruption time and the QoE instantaneous drop value calculated by the previous two sub-modules to generate the handover cost information of embodiment 2;

[0124] This embodiment decomposes the quantification of handover cost into two steps: physical interruption time calculation and perceptual QoE impact modeling; this decomposition enables the present application to evaluate the handover overhead in a fine and perceptual service manner; for example, an interruption of 230 ms can be disastrous for a VoIP service, but it can have little effect on a background file download service; this fine quantification is the key to avoiding the system from overprotecting non-critical services and underprotecting critical services.

[0125] Embodiment 5:

[0126] The continuity risk quantification engine includes:

[0127] The steady-state QoE prediction sub-module is used to obtain the future link state information; a pre-set steady-state QoE prediction function is called to convert it into a future steady-state QoE prediction value;

[0128] The risk probability calculation submodule is used to calculate, based on the predicted future steady-state QoE value, the probability that maintaining the current link will cause the QoE to fall below a preset acceptable QoE threshold, thus obtaining the link maintenance risk probability; based on the current QoE, the instantaneous QoE drop value, and the acceptable QoE threshold, the probability that an instantaneous impact will cause QoE collapse, thus obtaining the instantaneous impact collapse probability; and based on the predicted future steady-state QoE value, the probability that the future QoE of the link after the switch will fall below the acceptable QoE threshold, thus obtaining the future collapse probability after the switch.

[0129] The risk information generation submodule is used to calculate the switching link risk probability by combining the instantaneous impact collapse probability and the future collapse probability after the switch; and to generate the continuity risk information including the holding link risk probability, the instantaneous impact collapse probability and the switching link risk probability.

[0130] Based on Example 1, this example specifically defines the internal implementation of the continuous risk quantification engine, decoupling its functionality into three sub-modules:

[0131] The engine includes a steady-state QoE prediction submodule, whose purpose is to convert physical layer prediction into application layer experience prediction. In this embodiment, this submodule is used to obtain future link state information from the predictor and call a preset steady-state QoE prediction function. ,Will Convert into a time series of future steady-state QoE predictions The preset steady-state QoE prediction function For example, it could be a weighted model based on a specific business type k, such as... ,in It is a preset weight. It is a non-linear function that maps each physical indicator to a standard scoring range; or, the function can also be a deep neural network model trained with a large number of physical indicators and corresponding human scores.

[0132] Risk probability calculation submodule: Its purpose is to perform independent calculations of the three core risks; in this embodiment, this submodule:

[0133] based on Future steady-state QoE forecast Calculate its acceptable QoE threshold in the future. The probability of maintaining the link risk is obtained. ;

[0134] Based on the current QoE The instantaneous drop in QoE value transmitted from the modeler With QoE acceptable threshold ,pass Calculate the probability that a transient shock will cause QoE collapse, and obtain the transient shock collapse probability. ;

[0135] based on Future steady-state QoE forecast Calculate its future drop below The probability of future collapse after switching is obtained from the probability of switching. ;

[0136] Risk information generation submodule: Its purpose is to combine and encapsulate risks; in this embodiment, this submodule is used to combine the risk probability calculation submodule to generate the instantaneous shock collapse probability. and the probability of future crashes after switching ,pass The total probability of handover link risk is calculated. ; Generate includes , and Continuous risk information;

[0137] This embodiment clearly defines sub-modules. , and The calculation path for the three risks; it involves switching the total risk. Explicit decomposition into instantaneous impact risk and future channel risks By combining these factors, this invention can accurately identify and avoid a typical mistake that traditional decision-making makes: switching to a link with a good future channel, but the switching action itself will cause the service to crash; this multi-dimensional quantification of risk is the mathematical basis for achieving robust decision-making.

[0138] Example 6:

[0139] The risk-avoidance resource decision-maker includes: a comprehensive cost calculation submodule, used to obtain the probability of maintaining the link risk and the probability of switching the link risk; and to calculate the comprehensive cost of maintaining the link and the comprehensive cost of switching the link by combining the preset business priority weight and the preset normalized resource cost.

[0140] The damping threshold calculation submodule is used to obtain the instantaneous impact collapse probability; and calculate the dynamic switching damping threshold by combining the preset system sensitivity coefficient and the service priority weight.

[0141] A decision generation submodule is configured to determine whether the switching link comprehensive cost is less than the holding link comprehensive cost minus the dynamic switching damping threshold value; if yes, determine the optimal decision instruction as switching; if no, determine the optimal decision instruction as holding; and generate the risk-avoiding decision information based on the optimal decision instruction.

[0142] In the embodiment 1, the internal implementation of the risk-avoiding resource decision maker is specifically defined, and the function thereof is decoupled into three submodules:

[0143] The decision maker comprises a comprehensive cost calculation submodule, which is configured to unify the risk and the cost into the same dimension for comparison; in the embodiment, the submodule is configured to acquire the holding link risk probability and the switching link risk probability transmitted by the risk engine; combine the preset service priority weight and the preset normalized resource cost , and calculate the holding link comprehensive cost and the switching link comprehensive cost respectively;

[0144] The decision maker comprises a damping threshold value calculation submodule, which is configured to generate a dynamic switching brake to suppress the ping-pong effect; in the embodiment, the submodule is configured to acquire the instantaneous impact collapse probability transmitted by the risk engine; combine the preset system sensitivity coefficient and the service priority weight , and calculate the dynamic switching damping threshold value by ;

[0145] The decision maker comprises a decision generation submodule, which is configured to perform the final comparison and decision; in the embodiment, the submodule is configured to determine whether the switching link comprehensive cost is less than the holding link comprehensive cost minus the dynamic switching damping threshold value , that is ;

[0146] If yes, it indicates that the net benefit of switching is enough to overcome the instantaneous impact cost of switching , and the optimal decision instruction is determined as switching;

[0147] If no, the optimal decision instruction is determined as holding;

[0148] The submodule generates the risk-avoiding decision information of the embodiment 2 based on the optimal decision instruction;

[0149] The decision maker is constructed as a clear three-step closed loop in the embodiment; a damping threshold calculation sub-module is used to dynamically adjust the switching sensitivity ; when and are close and are high, the risk probability will significantly increase, resulting in the condition being difficult to meet, thereby inhibiting high-frequency ping-pong switching near the edge state point with high risk and low benefit, and greatly enhancing the robustness of system decision-making;

[0150] The decision-making mechanism has good robustness; in the case of high prediction uncertainty, the confidence interval of the predictor output will be widened, resulting in the risk probability and generally increasing, and the system will tend to make conservative decisions; when is extremely high, resulting in being huge, the decision-making condition becomes extremely difficult to meet, which effectively prevents the system from performing any switching that may cause the business to collapse instantaneously, even if the current link quality is poor, the basic usability is forced to be maintained, and the occurrence of a worse result is avoided, embodying the core idea of risk aversion. Embodiment 7:

[0151] The wireless channel prediction model is a long short-term memory network model or a Markov model.

[0152] On the basis of embodiment 3, the specific selection of the wireless channel prediction model is limited in the embodiment; the model is preferably a long short-term memory network LSTM model or a Markov model;

[0153] In the embodiment, the LSTM model is preferably used; the technical consideration is that LSTM is a special recurrent neural network and has a natural advantage in processing and predicting time series data; it can capture long-term dependencies and nonlinear characteristics in channel quality changes; for example, LSTM can learn a channel change pattern by learning a labeled data set containing user movement vectors and historical CQI sequences, thereby predicting a deterministic occlusion event caused by the user entering a city canyon or a tunnel, which is difficult to accurately predict by a traditional Markov model;

[0154] In another embodiment with limited computing resources, a Markov model can be used; the technical consideration is that the Markov model can simplify the channel state into a limited number of discrete states, and based on a labeled data set containing historical measurement data and corresponding channel states, the state transition probability is statistically obtained; the model has small computational overhead and is suitable for predicting channel fading with strong randomness but no obvious long-term pattern;

[0155] In another embodiment with limited computing resources, a Markov model can be used; the technical consideration is that the Markov model can simplify the channel state into a limited number of discrete states, and based on a labeled data set containing historical measurement data and corresponding channel states, the state transition probability is statistically obtained; the model has small computational overhead and is suitable for predicting channel fading with strong randomness but no obvious long-term pattern;

[0156] The embodiment can process time sequence dependence by using a model such as LSTM or Markov, so that the star-ground topology and channel evolution predictor has the ability to learn and predict the complex time sequence correlation in the wireless channel; this significantly improves the accuracy of future link state information, and is a key technical support for the invention to realize active regulation rather than passive response.

[0157] Embodiment 8:

[0158] The system is deployed on the ground network core network side, the MEC node or the satellite network ground gateway station.

[0159] In the system of any one of embodiments 1 to 7, the physical deployment position of the system is limited; the system can be flexibly deployed on the ground network core network side, the MEC node or the satellite network ground gateway station.

[0160] Deployed on the MEC node: this is the preferred embodiment of the invention; the technical consideration is that the MEC node is close to the network edge and the user terminal, and can obtain the GNSS and CQI / RSRP measurement report reported by the terminal with the lowest delay; at the same time, the decision result can also be executed as soon as possible; this deployment mode maximizes the response speed of the system to channel and position changes, and is especially suitable for vehicle networking automatic driving control signaling and other ultra-low latency, high reliability services;

[0161] Deployed on the ground network core network side: the technical consideration is that the core network has a global network topology and resource view, and can centrally manage and coordinate the resource allocation strategies of multiple MEC nodes and a large number of users, which is beneficial to realize global resource optimization and strategy coordination;

[0162] Deployed on the satellite network ground gateway station: the technical consideration is that the gateway station is the convergence point of the satellite link, and this deployment can make the system better master the real-time state of the satellite resources, and focus on the unified scheduling between the star-ground links;

[0163] The embodiment provides flexible deployment options to adapt to different network architectures and service requirements; deployment in the MEC node can minimize decision delay and guarantee ultra-low latency continuity of key services such as vehicle networking; deployment in the core network or gateway station is more conducive to global resource optimization and unified coordination of network strategies.

[0164] Embodiment 9:

[0165] Please refer to Figure 2 A resource allocation method in a star-ground integrated network, comprising the following steps:

[0166] Acquire satellite ephemeris, user GNSS positioning and movement vector, ground base station topology and real-time CQI / RSRP measurement report; fuse preset orbit dynamics model and preset wireless channel prediction model to predict future state of all available links of the user, so as to generate future link state information;

[0167] Acquire user service flow type and QoE requirement; calculate transient interruption time of switching action; based on the transient interruption time, quantize QoE instantaneous drop value of specific service; and based on the QoE instantaneous drop value, generate switching cost information;

[0168] Based on the future link state information, the switching cost information and preset QoE acceptable threshold, calculate risk probability of keeping current link; calculate instantaneous impact collapse probability; calculate risk probability of switching to target link; and generate continuity risk information containing the risk probability of keeping link, the instantaneous impact collapse probability and the risk probability of switching link;

[0169] Based on the continuity risk information, preset service priority weight and preset normalized resource cost, calculate keeping link comprehensive cost and switching link comprehensive cost; based on the instantaneous impact collapse probability, calculate dynamic switching damping threshold; based on the keeping link comprehensive cost, the switching link comprehensive cost and the dynamic switching damping threshold, execute decision, so as to generate risk avoidance decision information.

[0170] The application also provides a resource allocation method in a satellite-ground integrated network, which can be executed by the system of embodiment 1; the method comprises the steps which have been fully disclosed in the detailed description of the four modules of the system in the embodiment of embodiment 1;

[0171] The execution flow of the method is described below through a specific embodiment, and the example corresponds to the ping-pong switching suppression scenario mentioned in the background art:

[0172] Suppose that a user who is conducting VoIP service is currently connected to a ground link , and the signal quality is at the edge; at the same time, a LEO satellite passes, and the signal quality is slightly better; the traditional scheme will immediately switch; the method of the application executes the following steps:

[0173] The satellite-ground topology and channel evolution predictor acquires all inputs; the orbit dynamics model predicts that the channel will be good in the next 60 seconds, but will also remain stable; the system generates future link state information;

[0174] The service QoE impact and switching cost modeler starts;

[0175] Acquiring service type , ;

[0176] Computing transient interruption time : System determines this time of geo-satellite switching , , ; ;

[0177] Quantifying QoE transient drop value : System calls , inputs , computes that this time of switching has a great impact on VoIP service, ;

[0178] Generating switching cost information; continuity risk quantifying engine starts; acquiring inputs: , , ; computing risk probability of keeping current link : Based on the prediction of step one, its probability of being lower than 2.5 points ;

[0179] Computing transient impact collapse probability : Computing , i.e. ; since is true, collapse is a deterministic event, ;

[0180] Computing risk probability of switching to target link : Based on the prediction of step one, ; ;

[0181] Generating continuity risk information , , ;

[0182] Risk-avoiding resource decision maker starts; acquiring inputs: service priority , ; resource cost , ;

[0183] Computing comprehensive cost:

[0184]

[0185]

[0186] Computing dynamic switching damping threshold : Set , ;

[0187] Execute decision: judge i.e. whether the condition is false;

[0188] Generate risk avoidance decision information: instruction is to prohibit switching, keep the current ground link;

[0189] The method embodiment shows that the application quantifies the instantaneous impact cost of the switching action itself, which is unacceptable for VoIP services, resulting in the total risk cost of switching much higher than the cost of keeping the status quo , thus rejecting the switching that would occur in the traditional scheme and is harmful to user experience, ensuring the continuity of VoIP services; in another embodiment, the method can calculate much less than , thus performing active avoidance switching; this proves that the method can intelligently balance between avoiding future risks and avoiding instantaneous impacts according to the risks, ultimately realizing the fundamental value of ensuring service continuity.

[0190] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application and not to limit it, although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the application.

Claims

1.A resource allocation system in a satellite-terrestrial converged network, characterized by, The system comprises: a satellite-ground topology and channel evolution predictor configured to obtain satellite ephemeris, user GNSS positioning and movement vector, ground base station topology, and real-time CQI / RSRP measurement report; fuse a preset orbit dynamics model and a preset wireless channel prediction model to predict future states of all available links of the user to generate future link state information; a service QoE impact and switching cost modeler configured to obtain service flow type and QoE requirement of the user, calculate transient interruption time of switching action, quantify QoE instantaneous drop value of specific service based on the transient interruption time, and generate switching cost information based on the QoE instantaneous drop value; a continuity risk quantification engine configured to calculate risk probability of keeping current link based on the future link state information, the switching cost information, and a preset QoE acceptable threshold, calculate instantaneous impact collapse probability, calculate risk probability of switching to target link, and generate continuity risk information comprising the risk probability of keeping link, the instantaneous impact collapse probability, and the risk probability of switching link; a risk-averse resource decision maker configured to calculate keeping link comprehensive cost and switching link comprehensive cost based on the continuity risk information, a preset service priority weight, and a preset normalized resource cost, calculate dynamic switching damping threshold based on the instantaneous impact collapse probability, and perform decision making based on the keeping link comprehensive cost, the switching link comprehensive cost, and the dynamic switching damping threshold to generate risk-averse decision information. The future link state information comprises predicted signal-to-noise ratio, predicted delay, and predicted packet loss rate; the switching cost information comprises the transient interruption time and the QoE instantaneous drop value; the continuity risk information comprises the risk probability of keeping link, the instantaneous impact collapse probability, and the risk probability of switching link; and the risk-averse decision information comprises the keeping link comprehensive cost, the switching link comprehensive cost, the dynamic switching damping threshold, and optimal decision instruction. The satellite-ground topology and channel evolution predictor comprises: a data aggregation submodule configured to obtain satellite ephemeris data, terminal GNSS positioning and movement vector, ground base station topology information, and real-time CQI / RSRP measurement report; a channel evolution prediction submodule configured to fuse a preset orbit dynamics model and a preset wireless channel prediction model based on the report, the ephemeris data, the positioning and movement vector, and the topology information to predict time series of signal-to-noise ratio, delay, and packet loss rate of all available links within a future preset time; a state information generation submodule configured to generate the future link state information based on the time series. 2.The resource allocation system in the integrated satellite-terrestrial network according to claim 1, wherein, The service QoE impact and switching cost modeler comprises: 3.The resource allocation system in the integrated satellite-terrestrial network according to claim 1, wherein, an interruption time calculation submodule configured to obtain service flow type, and calculate transient interruption time based on a preset signaling interaction time, a preset physical layer synchronization time, and a preset data buffer reconstruction time. ​ ​ ​ 4.The resource allocation system in the integrated satellite-terrestrial network according to claim 1, wherein, ​ ​ The QoE impact modeling submodule is configured to call a preset QoE mapping function specific to the service flow type, and calculate a QoE instantaneous drop value based on the transient interruption time and an estimated expected packet loss rate of a switching process. The switching cost generation submodule is configured to generate the switching cost information based on the transient interruption time and the QoE instantaneous drop value. 5.The resource allocation system in the integrated satellite-terrestrial network according to claim 1, wherein, The continuity risk quantification engine includes: The steady-state QoE prediction submodule is configured to obtain the future link state information, and convert the future link state information into a future steady-state QoE prediction value by calling a preset steady-state QoE prediction function. The risk probability calculation submodule is configured to calculate a probability that the QoE is lower than a preset QoE acceptable threshold when the current link is maintained, to obtain a link-maintenance risk probability, based on the future steady-state QoE prediction value; calculate a probability that the QoE collapses due to an instantaneous impact, to obtain an instantaneous impact collapse probability, based on the current QoE, the QoE instantaneous drop value, and the QoE acceptable threshold; and calculate a probability that the future QoE is lower than the QoE acceptable threshold after switching, to obtain a post-switching future collapse probability, based on the future steady-state QoE prediction value. The risk information generation submodule is configured to calculate a switching link risk probability by combining the instantaneous impact collapse probability and the post-switching future collapse probability, and generate the continuity risk information including the link-maintenance risk probability, the instantaneous impact collapse probability, and the switching link risk probability. 6.The resource allocation system in the integrated satellite-terrestrial network according to claim 1, wherein, The risk-avoidance resource decision maker includes: The comprehensive cost calculation submodule is configured to obtain the link-maintenance risk probability and the switching link risk probability, and calculate a link-maintenance comprehensive cost and a switching link comprehensive cost, respectively, by combining a preset service priority weight and a preset normalized resource cost. The damping threshold calculation submodule is configured to obtain the instantaneous impact collapse probability, and calculate a dynamic switching damping threshold by combining a preset system sensitivity coefficient and the service priority weight. The decision generation submodule is configured to determine whether the switching link comprehensive cost is less than the link-maintenance comprehensive cost minus the dynamic switching damping threshold, determine an optimal decision instruction as switching if the determination is positive, and determine the optimal decision instruction as link maintenance if the determination is negative, and generate the risk-avoidance decision information based on the optimal decision instruction. 7.The resource allocation system in the integrated satellite-terrestrial network according to claim 1, wherein, The wireless channel prediction model is a long short-term memory network model or a Markov model. 8.The resource allocation system in the integrated satellite-terrestrial network according to claim 1, wherein, The system is deployed at a ground network core network side, a MEC node, or a satellite network ground gateway station. 9.A method for resource allocation in a satellite-terrestrial converged network, characterized in that, The method includes the following steps: Obtaining satellite ephemeris, user GNSS positioning and movement vector, ground base station topology, and real-time CQI / RSRP measurement report; Fusing a preset orbit dynamics model and a preset wireless channel prediction model to predict future states of all available links of the user, to generate future link state information; Obtaining user service flow type and QoE requirement; and calculating transient interruption time of a switching action; Based on the transient interruption time, quantifying a QoE instantaneous drop value of a specific service; and based on the QoE instantaneous drop value, generating switching cost information; calculating a risk probability of keeping the current link based on the future link state information, the handover cost information and a preset QoE acceptable threshold; calculating an instantaneous impact collapse probability; calculating a risk probability of switching to a target link; generating continuity risk information including the risk probability of keeping the link, the instantaneous impact collapse probability and the risk probability of switching the link; calculating a keeping link comprehensive cost and a switching link comprehensive cost based on the continuity risk information, a preset service priority weight and a preset normalized resource cost, calculating a dynamic switching damping threshold based on the instantaneous impact collapse probability, and performing a decision based on the keeping link comprehensive cost, the switching link comprehensive cost and the dynamic switching damping threshold to generate risk-avoiding decision information.

Images