An emergency communication satellite mobile broadband triple play system

Abstract

The application discloses an emergency communication satellite mobile broadband three-network fusion system, and relates to the technical field of wireless communication networks, comprising: a multi-mode fusion access gateway, which is used for providing communication interfaces with a telecommunication network, a Unicom network, a mobile network, a satellite broadband network and an ad hoc network node; and a six-dimensional state sensing unit, which is used for collecting six-dimensional network state data in time domain, frequency domain, space domain, energy domain, service domain and environment domain. The emergency communication satellite mobile broadband three-network fusion system realizes five-dimensional heterogeneous network resource collaborative scheduling of the telecommunication network, the Unicom network, the mobile network, the satellite broadband network and the ad hoc network node; the system can sense the network state evolution trend in advance, quantitatively evaluate the reliability of the prediction result, and spontaneously evolve to the state with the most balanced resource utilization and the highest overall order in the multi-network collaborative process, so that the fusion efficiency of multi-system networks in a dynamic environment is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication network technology, specifically to an emergency communication satellite mobile broadband triple-play system. Background Technology

[0002] With the increasing frequency of global climate change and various public emergencies, emergency communication capabilities have become a key indicator for measuring the modernization level of a country's disaster prevention, mitigation, and relief system. In the event of major natural disasters such as earthquakes, floods, and forest fires, terrestrial public communication networks often become paralyzed due to power outages, damaged fiber optic cables, or flooded equipment, turning disaster areas into information islands and severely hindering command and dispatch and information transmission for rescue operations. Therefore, establishing a stable, reliable, and high-bandwidth emergency communication system capable of rapid deployment under extremely harsh conditions is crucial.

[0003] Currently, the main emergency communication technologies include satellite communication vehicles, drone relays, and portable satellite terminals. Among these, satellite communication, with its wide coverage and lack of terrain limitations, has become a key means of restoring communication in the early stages of a disaster. For example, patent document CN120343659A describes a "satellite broadband self-organizing network converged communication method and system," which obtains the state parameters of satellite links and self-organizing network nodes, generates network state feature vectors, and makes dynamic switching decisions based on spectrum interference thresholds. This achieves efficient coordination between satellite and self-organizing network resources, and can, to a certain extent, cope with complex electromagnetic environments and dynamic topology changes, thereby improving communication reliability.

[0004] While the aforementioned patents achieve intelligent integration and dynamic link switching between satellite and ad hoc networks, demonstrating significant advantages in improving communication reliability, their application scenarios focus more on network collaboration under wide-area coverage. They do not address how to deeply integrate satellite broadband networks with existing or rapidly restored terrestrial mobile communication networks in disaster areas to achieve true "three-network convergence." Furthermore, in extreme emergency scenarios involving widespread power and network outages, the system's energy security, unified access and scheduling of signals from multiple operators, rapid on-site network setup, and service priority assurance still require further optimization. Issues include insufficient system integration, low efficiency in multi-network collaboration, and inflexible on-site deployment. Therefore, existing emergency communication systems struggle to meet the integrated command and dispatch needs of multiple teams, multiple standards, high concurrency, and large bandwidth at large disaster sites. Summary of the Invention

[0005] The purpose of this invention is to provide an emergency communication satellite mobile broadband triple-network converged system to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an emergency communication satellite mobile broadband triple-play system, comprising: Multimode converged access gateway is used to provide communication interfaces with telecommunications networks, China Unicom networks, mobile networks, satellite broadband networks, and self-organizing network nodes; The six-dimensional state perception unit is used to collect six-dimensional network state data in the time domain, frequency domain, spatial domain, energy domain, service domain, and environmental domain. The fusion decision control center is connected to the six-dimensional state perception unit and the multi-mode fusion access gateway, respectively. The fusion decision control center has a built-in multi-source heterogeneous network state cyclic screening and fusion decision model. The model is used to perform nonlinear dynamic prediction on the six-dimensional network state data to generate advanced state prediction values, and to perform uncertainty quantification on the advanced state prediction values ​​to generate credibility evaluation parameters. Then, based on the credibility evaluation parameters and the six-dimensional network state data, multi-round game entropy reduction optimization is performed to generate the fusion decision instruction with the minimum system entropy value. The multi-mode converged access gateway responds to the converged decision command by establishing a dynamic converged transmission path among the telecommunications network, the China Unicom network, the mobile network, the satellite broadband network, and the self-organizing network node.

[0007] Preferably, the six-dimensional state sensing unit includes: The time-domain sensing module is used to acquire the time-domain waveform and signal-to-noise ratio time series of network signals; The frequency domain sensing module is used to collect spectrum occupancy and interference frequency distribution; The airspace awareness module is used to collect the real-time location of self-organizing network nodes and the boundary of satellite coverage area; The energy domain sensing module is used to collect the backup power supply duration of the base stations of the telecommunications network, the Unicom network and the mobile network, the power amplifier power margin of the satellite broadband network, the remaining energy of the self-organizing network node and the remaining power supply duration of the microgrid unit. The service domain awareness module is used to collect the type, priority, and quality of service requirements of the current transmission service; The environmental domain sensing module is used to collect environmental temperature, humidity, and air pressure parameters.

[0008] Preferably, the multi-source heterogeneous network state cyclic screening and fusion decision model includes a chaos prediction unit. The chaos prediction unit is used to reconstruct the phase space of the key parameter time series in the six-dimensional network state data, calculate the maximum Lyapunov exponent of the reconstructed phase space to determine the chaotic characteristics of the key parameter time series, and use a weighted first-order local method to predict the future state of the key parameter time series in advance, generating the predicted value of the advanced state.

[0009] Preferably, the multi-source heterogeneous network state cyclic screening and fusion decision model further includes a cloud model credibility assessment unit, which is connected to the chaos prediction unit. The cloud model credibility assessment unit is used to collect historical prediction error data and generate digital features of the error cloud, including expected value, entropy value and hyperentropy value. The cloud model credibility assessment unit is also used to take the deviation between the advanced state prediction value and the current true value as cloud droplet input, and combine the digital features of the error cloud to generate credibility cloud parameters of the advanced state prediction value. The credibility cloud parameters include prediction expectation value, prediction entropy value and prediction over-entropy value. The cloud model credibility assessment unit is also used to classify the credibility of the advanced state prediction value based on the prediction entropy value and the prediction hyper-entropy value.

[0010] Preferably, the multi-source heterogeneous network state cyclic screening and fusion decision model further includes a multi-agent game entropy reduction optimization unit, which is connected to the cloud model credibility evaluation unit. The multi-agent game entropy reduction optimization unit is used to define the telecommunications network, the Unicom network, the mobile network, the satellite broadband network, and the self-organizing network node as five game agents, each of which has its own set of available resource strategies. The multi-agent game entropy reduction optimization unit is also used to construct a system entropy function, which is the sum of the resource utilization Shannon entropies of the five game agents; the multi-agent game entropy reduction optimization unit is also used to execute the first round of the game to solve the Nash equilibrium strategy combination in the current state and calculate the first system entropy value. The multi-agent game entropy reduction optimization unit is also used to take the advanced state prediction value as the game environment input, execute the second round of game to solve the Nash equilibrium strategy combination in the future state, and calculate the second system entropy value. The multi-agent game entropy reduction optimization unit is also used to use the credibility cloud parameter as the penalty term of the payoff function, execute the third round of game to solve the modified Nash equilibrium strategy combination, and calculate the entropy value of the third system. The multi-agent game entropy reduction optimization unit is also used to compare the entropy values ​​of the first system, the second system, and the third system. When the reduction in the entropy value of the system in two adjacent rounds is less than the preset entropy reduction threshold, the Nash equilibrium strategy combination of the third round of the game is output as the fusion decision instruction.

[0011] Preferably, the multi-source heterogeneous network state cyclic screening and fusion decision model further includes a two-way verification and self-learning evolution unit, which is connected to the chaos prediction unit, the cloud model credibility assessment unit and the multi-agent game entropy reduction optimization unit, respectively. The bidirectional verification and self-learning evolution unit is used to collect real network state data after the execution of the fusion decision instruction, and to calculate the prediction error between the real network state data and the advanced state prediction value. The bidirectional verification and self-learning evolution unit is also used to positively feed the prediction error back to the chaotic prediction unit to correct the phase space reconstruction parameters, feed it back to the cloud model credibility evaluation unit to update the digital features of the error cloud, and feed it back to the multi-agent game entropy reduction optimization unit to adjust the payoff function weight coefficients. The bidirectional verification and self-learning evolution unit is also used to periodically re-input historical real network state data into the chaos prediction unit, the cloud model credibility assessment unit, and the multi-agent game entropy reduction optimization unit to reenact the decision process, verify the decision effect after parameter correction, and solidify or roll back the parameter correction based on the verification results.

[0012] Preferably, the bidirectional verification and self-learning evolution unit also has a built-in experience knowledge base, which is used to store historical successful decision-making cases and their corresponding six-dimensional network state data feature vectors. The bidirectional verification and self-learning evolution unit is also used to perform similarity matching between the current six-dimensional network state data and the feature vectors in the experience knowledge base. When the matching similarity exceeds a preset threshold, the decision strategy of the matched historical successful decision case is directly called as the fusion decision instruction.

[0013] Preferably, it also includes a smart microgrid unit, which is connected to the fusion decision control center. The smart microgrid unit includes a photovoltaic power generation module, a diesel power generation module, an energy storage battery module, and a backup capacitor module. The fusion decision control center is also used to use the remaining power supply duration collected by the energy domain sensing module as a hard constraint condition for the game agent during the multi-round game entropy reduction optimization process.

[0014] Preferably, the fusion decision instruction includes a primary fusion path, a backup fusion path, and an emergency fallback path; The primary fusion path is used to carry high-priority real-time services, the backup fusion path is used to carry medium-priority non-real-time services, and the emergency fallback path is the BeiDou short message communication queue, which is used to transmit critical coordinate data when all network links are interrupted.

[0015] Preferably, the multi-mode converged access gateway includes a telecommunications network access module, a China Unicom network access module, a China Mobile network access module, a satellite broadband modem, and a self-organizing network radio frequency transceiver module. The telecommunications network access module, the China Unicom network access module, and the China Mobile network access module respectively support the 2G, 4G, and 5G network standards of their respective operators. The satellite broadband modem supports dynamic switching of multiple satellite frequency bands, and the self-organizing network radio frequency transceiver module supports multi-hop relay routing protocols.

[0016] This invention provides an emergency communication satellite mobile broadband triple-play system. It has the following beneficial effects: Through the multi-source heterogeneous network state cyclic screening and fusion decision model built into the fusion decision control center, the five-dimensional heterogeneous network resource collaborative scheduling of telecommunications networks, Unicom networks, mobile networks, satellite broadband networks and self-organizing network nodes is realized. The system can proactively perceive the network state evolution trend, quantitatively assess the credibility of the prediction results, and spontaneously evolve towards the state with the most balanced resource utilization and the highest overall orderliness in the process of multi-network collaboration, effectively improving the fusion efficiency of multi-standard networks in dynamic environments.

[0017] The system can dynamically establish a three-level transmission guarantee system of primary converged path, backup converged path and emergency backup path according to the real-time priority of rescue operations. The entire system improves the stability and resource utilization efficiency of multi-network converged communication in extreme disaster environments through a closed-loop architecture of multi-dimensional perception, advanced prediction, reliable assessment, game optimization and self-learning evolution. Attached Figure Description

[0018] Figure 1 This is an internal processing flowchart of the converged decision control center of an emergency communication satellite mobile broadband triple-network converged system according to the present invention; Figure 2 This is a flowchart illustrating the experience knowledge base matching and self-learning process of an emergency communication satellite mobile broadband triple-network converged system according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 and Figure 2 This invention provides a technical solution: an emergency communication satellite mobile broadband triple-play system, comprising: Multimode converged access gateway is used to provide communication interfaces with telecommunications networks, China Unicom networks, mobile networks, satellite broadband networks, and self-organizing network nodes; The six-dimensional state perception unit is used to collect six-dimensional network state data in the time domain, frequency domain, spatial domain, energy domain, service domain, and environmental domain. The fusion decision control center is connected to the six-dimensional state perception unit and the multi-mode fusion access gateway. The fusion decision control center has a built-in multi-source heterogeneous network state cyclic screening and fusion decision model. The model is used to perform nonlinear dynamic prediction on the six-dimensional network state data to generate advanced state prediction values, and to quantify the uncertainty of the advanced state prediction values ​​to generate credibility assessment parameters. Then, based on the credibility assessment parameters and the six-dimensional network state data, multi-round game entropy reduction optimization is performed to generate the fusion decision instruction with the minimum system entropy value. The multi-mode converged access gateway responds to converged decision commands and establishes dynamic converged transmission paths between telecommunications networks, China Unicom networks, mobile networks, satellite broadband networks, and self-organizing network nodes.

[0021] It should be further explained that, in the specific implementation, the multi-source heterogeneous network state cyclic screening and fusion decision model built into the fusion decision control center processes the time domain, frequency domain, spatial domain, energy domain, service domain and environmental domain data collected by the six-dimensional state perception unit.

[0022] The multi-source heterogeneous network state cyclic screening and fusion decision-making model is mounted on the hardware platform of the fusion decision control center. This hardware platform adopts a heterogeneous computing architecture combining an embedded multi-core processor and a field-programmable gate array (FPGA), and uses high-speed static random access memory (SRAM) and non-volatile flash memory modules to realize data storage and operation. The embedded multi-core processor is responsible for model logic scheduling and decision output, the FPGA undertakes six-dimensional data parallel processing and real-time prediction operation, and the non-volatile flash memory is used to store historical state data and model parameters. The key preset thresholds in the model are all obtained through training and on-site calibration with a large amount of measured data in emergency communication scenarios. The entropy reduction threshold is set according to the accuracy requirements of system resource balancing scheduling and is used to judge whether the system orderliness has reached a stable state during the game iteration process. The similarity threshold is determined based on the feature matching accuracy requirements of historical successful decision cases to ensure that the optimal historical strategy can be quickly matched in similar scenarios. The credibility threshold is combined with the error tolerance range of network state prediction to distinguish the credibility level of the advanced state prediction value. All thresholds can be dynamically adjusted according to the complexity of the actual emergency scenario.

[0023] The model first performs phase space reconstruction on the time series of key parameters in the six-dimensional network state data, and determines the chaotic characteristics of the sequence by calculating the maximum Lyapunov exponent of the reconstructed phase space. For time series exhibiting chaotic characteristics, a weighted first-order local method is used for advance prediction, generating advance state prediction values ​​containing multiple future time steps.

[0024] Subsequently, the model inputs historical prediction error data into the inverse cloud generator to generate three digital features of the error cloud: expected value, entropy value, and over-entropy value. The model then uses the deviation between the current advanced state prediction and the actual value as cloud droplet input, and combines these with the three digital features of the error cloud to generate a confidence cloud parameter for the current advanced state prediction through the forward cloud generator. This confidence cloud parameter includes the predicted expected value, predicted entropy value, and predicted over-entropy value. Based on the magnitude of the predicted entropy value and predicted over-entropy value, the model classifies the advanced state prediction into three levels: high confidence, medium confidence, and low confidence.

[0025] The model defines five game agents: a telecommunications network, a China Unicom network, a mobile network, a satellite broadband network, and ad hoc network nodes. Each game agent has its own set of available resource strategies. The model constructs a system entropy function that is the sum of the Shannon entropies of the resource utilization rates of these five game agents.

[0026] The model performs a first round of game analysis to solve for the Nash equilibrium strategy combination in the current state and calculates the first system entropy. The model then uses high-confidence-level forward state predictions as input to the game environment and performs a second round of game analysis to solve for the Nash equilibrium strategy combination in the future state and calculates the second system entropy. Finally, the model uses the confidence cloud parameter as a penalty term in the payoff function and performs a third round of game analysis to solve for the modified Nash equilibrium strategy combination and calculate the third system entropy.

[0027] The model compares the entropy values ​​of the first, second, and third systems. When the decrease in system entropy value between two adjacent rounds is less than a preset entropy reduction threshold, the model determines that the system has converged to an ordered state and outputs the Nash equilibrium strategy combination of the third round of the game as a fusion decision instruction to the multi-mode fusion access gateway.

[0028] In response to the convergence decision command, the multi-mode converged access gateway performs resource scheduling and link switching among its internal telecommunications network access module, Unicom network access module, mobile network access module, satellite broadband modem, and self-organizing network radio frequency transceiver module, thereby establishing a dynamic converged transmission path among telecommunications network, Unicom network, mobile network, satellite broadband network, and self-organizing network nodes.

[0029] The six-dimensional state perception unit includes: The time-domain sensing module is used to acquire the time-domain waveform and signal-to-noise ratio time series of network signals; The frequency domain sensing module is used to collect spectrum occupancy and interference frequency distribution; The airspace awareness module is used to collect the real-time location of self-organizing network nodes and the boundary of satellite coverage area; The energy domain sensing module is used to collect data on the backup power supply duration of base stations in telecommunications networks, China Unicom networks, and mobile networks; the power amplifier margin of satellite broadband networks; the remaining energy of self-organizing network nodes; and the remaining power supply duration of microgrid units. The service domain awareness module is used to collect the type, priority, and quality of service requirements of the current transmission service; The environmental domain sensing module is used to collect environmental temperature, humidity, and air pressure parameters.

[0030] It should be further explained that, in specific implementations, six-dimensional state perception refers to the comprehensive collection of operational status data of emergency communication networks from six dimensions: time domain, frequency domain, spatial domain, energy domain, service domain, and environmental domain. This covers all elements of network signal, spectrum, location, energy, services, and environment. Its technical essence is to achieve comprehensive perception of network status, providing complete data support for subsequent prediction and decision-making. The rationale for this is that network status in emergency communication scenarios is influenced by multiple coupled factors, and single-dimensional perception cannot meet the needs of dynamic scheduling. Game-theoretic entropy reduction optimization refers to transforming multi-network resource scheduling into a multi-agent game process, reducing system entropy through iterative game theory, and improving network resource allocation. The orderliness and balance, in physical terms, are to allow the multi-network converged system to evolve from a disordered state to an ordered and stable state. The basis for this is the mature application of the entropy increase principle and multi-agent game theory in the field of resource scheduling. The entropy reduction threshold is a critical value used to determine whether the game iteration has converged. It represents the minimum effective magnitude of the reduction in the system's entropy value. Reaching this threshold indicates that the system's resource allocation has become stable and there is no need to continue iterating. The technical evaluation standard for energy domain awareness takes the remaining energy support time of each network node and power supply unit as the core, combined with the energy consumption rate and the priority of business needs, to evaluate whether the energy supply can meet the communication needs of the current and future periods, and ensure that energy resources and network scheduling are coordinated and matched. The time-domain sensing module in the six-dimensional state-aware unit collects in-phase orthogonal data of the satellite link and the self-organizing network link in real time through the baseband chip of the satellite modem and the self-organizing network node. After performing a sliding window Fourier transform on the in-phase orthogonal data, the signal-to-noise ratio time series and the bit error rate time series are extracted.

[0031] The frequency domain sensing module scans the licensed frequency bands used by telecommunications networks, China Unicom networks, and mobile networks, as well as the downlink frequency bands of satellite broadband networks, using a software-defined radio receiver to collect power spectral density data for each frequency band. Frequency points with power spectral density exceeding the background noise baseline are marked as interference frequencies, and their frequency coordinates and interference intensity are recorded, generating a spectrum occupancy distribution map and a list of interference frequency distributions.

[0032] The airspace perception module collects the real-time longitude, latitude, and altitude coordinates of the self-organizing nodes through the GPS receiver and the BeiDou positioning module, and calculates the projection boundary of the satellite beam coverage area on the ground through satellite orbit ephemeris data and beam pointing angle, generating a set of self-organizing node position coordinates and a set of satellite coverage area boundary coordinates.

[0033] The energy domain sensing module reads the remaining backup battery capacity and remaining fuel level of the generator at each base station in the telecommunications, China Unicom, and China Mobile networks through the base station network management interface to calculate the backup power supply duration of the base stations. It reads the current transmission power level and maximum allowable transmission power of the satellite broadband network through the satellite power amplifier status monitoring circuit to calculate the power amplifier's power margin. It reads the remaining battery capacity and total battery capacity of the self-organizing network nodes through the power monitoring chip to calculate the remaining energy percentage of the nodes. Finally, it reads the state of charge of the energy storage batteries and the remaining fuel level of the diesel generator through the battery management system and fuel level sensor of the smart microgrid unit to calculate the remaining power supply duration that the microgrid unit can provide to external systems.

[0034] The service domain awareness module uses deep packet inspection technology to analyze the data stream passing through the multi-mode fusion access gateway, identifying the five-tuple information and application layer protocol type of the data stream. It categorizes the data stream into voice command, video transmission, environmental data, and location information, and assigns transmission priority and quality of service levels to each type of service according to the preset rules of the rescue site command system.

[0035] The environmental domain sensing module collects ambient temperature, relative humidity, and atmospheric pressure parameters through digital temperature sensors, humidity sensors, and barometric pressure sensors distributed on the surface of radio frequency components and in the base station equipment room.

[0036] The multi-source heterogeneous network state cyclic screening and fusion decision model includes a chaotic prediction unit. The chaotic prediction unit is used to reconstruct the phase space of the key parameter time series in the six-dimensional network state data, calculate the maximum Lyapunov exponent of the reconstructed phase space to determine the chaotic characteristics of the key parameter time series, and use the weighted first-order local method to predict the future state of the key parameter time series in advance, generating the advanced state prediction value.

[0037] It should be further explained that, in a specific implementation, the chaos prediction unit obtains the signal-to-noise ratio time series of the satellite link or the moving speed time series of the self-organizing network nodes from the time domain perception module of the six-dimensional state perception unit as the key parameter time series, and uses the CC method to calculate the delay time and embedding dimension of the key parameter time series for phase space reconstruction.

[0038] During reconstruction, the one-dimensional time series is first mapped to an m-dimensional phase space determined by the delay time and embedding dimension to form a set of phase points. Then, the maximum Lyapunov exponent is calculated using a small-data-volume method on the reconstructed phase space. During the calculation, the nearest neighbor of each phase point is found in the reconstructed phase space, and the separation angle is constrained to avoid neighboring points being on the same trajectory. Then, the logarithmic distance between each phase point and its nearest neighbor is calculated and averaged. The time average obtained by fitting using the least squares method is the maximum Lyapunov exponent. If the maximum Lyapunov exponent is greater than zero, the time series of the key parameter is determined to have chaotic characteristics, and the subsequent prediction process is initiated. If the maximum Lyapunov exponent is not greater than zero, the traditional time series prediction method is switched to.

[0039] When performing phase space reconstruction, the chaotic prediction unit uses the CC method to determine the delay time and embedding dimension adapted to the changes in the state of the emergency communication network by performing segmented statistical analysis and correlation analysis on the time series of key parameters. Parameter selection is completed solely based on the distribution characteristics and changing trends of the sequence data. When calculating the maximum Lyapunov exponent using the small data volume method, the phase points in the reconstructed phase space are traversed, and the neighboring feature points of each phase point are found and their distance change patterns are statistically analyzed. The chaotic characteristics of the time series are determined based on the trend of distance change. The entire process is based on data traversal and feature comparison. When using the weighted first-order local method for prediction, weights are assigned based on the spatial distance between the current state point and its neighboring points. The closer the neighboring points are, the higher their weights are. The future state change pattern is obtained by linear fitting through the evolution trend of neighboring points, thereby generating the advanced state prediction value. The parameter fitting process is completed through iterative calibration using historical network state data, and the fitting rules are continuously corrected to adapt to the network state changes under different emergency scenarios.

[0040] For a key parameter time series exhibiting chaotic characteristics, the chaotic prediction unit searches for multiple neighboring points of the current state point in the reconstructed phase space. A weighted first-order local method is used to assign weights to each neighboring point, with a larger weight for each neighboring point that is closer to the current state point. Then, a weighted linear fit is performed on the set of neighboring points to obtain the evolutionary pattern. Based on this evolutionary pattern, the phase coordinates for multiple future time steps are iteratively calculated. The first component of the future phase coordinates is then mapped back to the one-dimensional time series to obtain the advanced state prediction value.

[0041] The chaos prediction unit also incorporates a long short-term memory (LSTM) neural network model, which simultaneously learns and predicts the time series of the key parameter to generate auxiliary prediction values. The chaos prediction unit dynamically adjusts the weighted fusion coefficients based on the errors of the chaos predictions and the LTM predictions from recent historical periods. These weighted fusion coefficients are then multiplied by both the leading-state prediction from the chaos predictions and the auxiliary prediction from the LTM model, and summed to generate the final fused leading-state prediction, which is output to the cloud model credibility evaluation unit.

[0042] The multi-source heterogeneous network state cyclic screening and fusion decision model also includes a cloud model credibility assessment unit, which is connected to the chaos prediction unit. The cloud model credibility assessment unit is used to collect historical prediction error data and generate digital features of the error cloud. The digital features include expected value, entropy value and hyperentropy value. The cloud model credibility assessment unit is also used to take the deviation between the advanced state prediction value and the current true value as cloud droplet input, and combine the digital features of the error cloud to generate the credibility cloud parameters of the advanced state prediction value. The credibility cloud parameters include the prediction expectation value, the prediction entropy value, and the prediction over-entropy value. The cloud model credibility assessment unit is also used to classify the credibility of the advanced state prediction value based on the prediction entropy value and the prediction hyper-entropy value.

[0043] It should be further explained that, in the specific implementation, the cloud model credibility assessment unit obtains the advanced state prediction value from the chaos prediction unit, and at the same time obtains the true value of the key parameter corresponding to the advanced state prediction value from the six-dimensional state perception unit.

[0044] The cloud model credibility assessment unit has a built-in historical error database to store the deviation data between the predicted and actual values ​​at various historical moments within a preset time window. The cloud model credibility assessment unit periodically reads a set of historical prediction error data from this historical error database as input samples and inputs this set of input samples into the inverse cloud generator. The inverse cloud generator first calculates the sample mean of this set of input samples as the expected value, then calculates the first-order sample absolute central moment of this set of input samples and uses the expected value to calculate the entropy value, and finally calculates the sample variance of this set of input samples and uses the entropy value to calculate the hyperentropy value. The inverse cloud generator outputs the error cloud digital features composed of the expected value, entropy value, and hyperentropy value and stores them in the internal register of the cloud model credibility assessment unit.

[0045] After obtaining the advanced state prediction value at the current moment, the cloud model credibility assessment unit reads the true value of the most recent time point before the current moment from the six-dimensional state perception unit as a reference value, and calculates the difference between the advanced state prediction value and the reference value as the value of the current cloud droplet. This value of the current cloud droplet is input into the forward cloud generator. The forward cloud generator uses the most recently updated error cloud digital features before the current moment as the digital feature parameters of the cloud model, and uses the value of the current cloud droplet as the quantitative value of the cloud droplet to generate the certainty of the current cloud droplet relative to the error cloud distribution. This certainty is the credibility of the advanced state prediction value at the current moment. The inverse cloud generator in the cloud model credibility assessment unit directly generates the expected value, entropy value, and hyperentropy value of the error cloud by statistically analyzing the central tendency, dispersion, and fluctuation characteristics of historical prediction error data. It extracts core feature parameters solely based on the overall distribution pattern of the error data. The forward cloud generator uses the deviation between the predicted and actual values ​​as input data, combines the digital features of the error cloud for cloud droplet mapping and certainty calculation, and obtains the credibility cloud parameters of the predicted value through data mapping and feature matching. The entire process is implemented based on the standard data processing flow of the cloud model. Those skilled in the art can deploy the algorithm using conventional cloud model implementation methods. The digital features are continuously updated and iteratively optimized using newly added prediction error data. The forward cloud generator outputs the current prediction value credibility cloud parameters, which include the predicted value itself and a two-dimensional credibility parameter composed of entropy and hyperentropy values.

[0046] The cloud model credibility assessment unit internally sets a high-credibility entropy threshold and a medium-credibility entropy threshold, with the high-credibility entropy threshold being lower than the medium-credibility entropy threshold. The unit compares the entropy value in the current prediction's credibility cloud parameters with both the high-credibility and medium-credibility entropy thresholds. If the entropy value is lower than the high-credibility entropy threshold and the over-entropy value is lower than the preset over-entropy threshold, the predicted state is marked as high-credibility. If the entropy value is greater than or equal to the high-credibility entropy threshold but less than the medium-credibility entropy threshold, the predicted state is marked as medium-credibility. If the entropy value is greater than or equal to the medium-credibility entropy threshold, the predicted state is marked as low-credibility.

[0047] The cloud model credibility assessment unit outputs the credibility cloud parameters of the current predicted value, labeled with credibility levels, to the multi-agent game entropy reduction optimization unit. For predicted values ​​with high credibility levels, the entropy and hyper-entropy values ​​in their credibility cloud parameters are used as factors in the penalty term of the payoff function during subsequent game play. For predicted values ​​with low credibility levels, the multi-agent game entropy reduction optimization unit abandons the use of these predicted values ​​and instead uses the current true value for the game.

[0048] The multi-source heterogeneous network state cyclic screening and fusion decision model also includes a multi-agent game entropy reduction optimization unit. The multi-agent game entropy reduction optimization unit is connected to the cloud model credibility evaluation unit. The multi-agent game entropy reduction optimization unit is used to define the telecommunications network, the Unicom network, the mobile network, the satellite broadband network, and the self-organizing network nodes as five game agents. Each game agent has its own set of available resource strategies. The multi-agent game entropy reduction optimization unit is also used to construct the system entropy function, which is the sum of the Shannon entropies of the resource utilization of the five game agents; the multi-agent game entropy reduction optimization unit is also used to execute the first round of the game to solve the Nash equilibrium strategy combination in the current state and calculate the first system entropy value; The multi-agent game entropy reduction optimization unit is also used to take the advanced state prediction value as the game environment input, execute the second round of game to solve the Nash equilibrium strategy combination in the future state, and calculate the entropy value of the second system. The multi-agent game entropy reduction optimization unit is also used to use the credibility cloud parameter as the penalty term of the payoff function to execute the third round of the game to solve the modified Nash equilibrium strategy combination and calculate the entropy value of the third system. The multi-agent game entropy reduction optimization unit is also used to compare the entropy values ​​of the first system, the second system, and the third system. When the reduction in the entropy value of the system in two adjacent rounds is less than the preset entropy reduction threshold, the Nash equilibrium strategy combination of the third round of the game is output as the fusion decision instruction.

[0049] It should be further explained that, in the specific implementation, the multi-agent game entropy reduction optimization unit obtains the real-time status data of the telecommunications network, the Unicom network, the mobile network, the satellite broadband network, and the self-organizing network nodes at the current moment from the six-dimensional state perception unit. This includes the available bandwidth, transmission latency, packet loss rate, remaining energy or battery life, current load rate, and priority queue of the services carried by each network.

[0050] The multi-agent game entropy reduction optimization unit defines the five network entities mentioned above as five independent game agents and constructs a set of available resource strategies for each game agent. The available resource strategy set for the telecom network game agent includes the bandwidth allocation ratio and access priority level of its 2G, 4G, and 5G network slices; the available resource strategy sets for the China Unicom and China Mobile game agents are similar. The available resource strategy set for the satellite broadband network game agent includes the power level and modulation / coding scheme of its different frequency band links. The available resource strategy set for the ad hoc network node game agent includes the hop count selection and transmit power level of its relay path.

[0051] The multi-agent game entropy reduction optimization unit constructs a system entropy function. This function calculates the ratio of each player's current resource utilization rate to its maximum available resource as a probability value. This probability value is then substituted into the Shannon entropy formula to obtain the individual entropy value for each player. Finally, the individual entropy values ​​of the five players are summed to obtain the system entropy value. A lower system entropy value indicates a higher overall orderliness in the resource allocation across the five networks.

[0052] The replication dynamic equation in the multi-agent game entropy reduction optimization unit adopts an iterative update logic based on strategy payoffs. The strategy selection probability is adjusted according to the payoff value of each agent's current strategy. The higher the payoff, the greater the probability of the strategy being selected. By iteratively updating the strategy combination of each agent, the solution of Nash equilibrium adopts an iterative optimization method, continuously traversing the available resource strategy set of each agent until a stable strategy combination is found in which no agent can improve payoff by changing the strategy alone. The system entropy function calculates the degree of balance of resource utilization of each agent by statistically analyzing the resource utilization distribution characteristics of the five agents. The more balanced the resource allocation, the lower the system entropy and the higher the system orderliness. The entire calculation process is implemented based on the statistical characteristics of resource utilization. Those skilled in the art can complete the implementation based on the conventional methods of communication network resource scheduling.

[0053] The multi-agent game entropy reduction optimization unit executes the first round of the game, using the real-time state data at the current moment as the game environment, and adopts the replication dynamic equation to evolve the strategy with the goal of maximizing the self-payment of each game agent. It solves for the Nash equilibrium strategy combination in the current state and calculates the first system entropy value corresponding to the Nash equilibrium strategy combination according to the system entropy value function.

[0054] The multi-agent game entropy reduction optimization unit obtains advanced state prediction values ​​marked with high credibility levels from the cloud model credibility assessment unit, substitutes these advanced state prediction values ​​as game environment parameters for future moments into the payoff functions of each game agent, re-employs the replication dynamic equation to perform strategy evolution, solves for the Nash equilibrium strategy combination in the future state, and calculates the second system entropy value corresponding to the Nash equilibrium strategy combination in the future state according to the system entropy value function.

[0055] The multi-agent game entropy reduction optimization unit obtains the entropy and hyper-entropy values ​​from the current prediction confidence cloud parameters from the cloud model confidence assessment unit, and introduces these entropy and hyper-entropy values ​​as penalty factors into the payoff function of each game agent. Specifically, the original payoff function value is multiplied by the difference between the entropy value and the hyper-entropy value to obtain the corrected payoff function value. After introducing this confidence penalty term, the strategy evolution is re-executed using the real-time state data at the current moment as the game environment to obtain the corrected Nash equilibrium strategy combination, and the third system entropy value corresponding to the corrected Nash equilibrium strategy combination is calculated according to the system entropy value function.

[0056] The multi-agent game entropy reduction optimization unit has a preset entropy reduction threshold, which is used to determine whether the system has converged to an ordered state. Three rounds of game play form the basic iterative process for the system to perform multi-agent game entropy reduction optimization. These three rounds correspond to the current state game, future state game, and credibility correction game, respectively. Iteration continues until the entropy reduction is less than the preset threshold twice consecutively. This is based on the convergence judgment rule of the three-round basic game process. That is, after completing the three rounds of basic game play, if the entropy reduction in two adjacent rounds is less than the preset threshold twice consecutively, the iteration is terminated and a fusion decision instruction is output. If the convergence condition is not met, the result of the third round of game play is used as the new initial strategy combination, and the iterative process of the three rounds of basic game play is repeated until the convergence condition is met. The correspondence between the basic iterative process and the convergence judgment rule is completely consistent in its overall expression logic.

[0057] The multi-agent game entropy reduction optimization unit calculates the difference between the entropy value of the second system and the entropy value of the first system as the first entropy reduction magnitude, and calculates the difference between the entropy value of the third system and the entropy value of the second system as the second entropy reduction magnitude. The first and second entropy reduction magnitudes are then compared with the entropy reduction thresholds, respectively.

[0058] If both the first and second entropy reduction magnitudes are less than the entropy reduction threshold, the system is considered converged. The modified Nash equilibrium strategy combination obtained from the third round of game solving is output as the final fusion decision instruction to the multi-mode fusion access gateway. If either the first or second entropy reduction magnitude is greater than or equal to the entropy reduction threshold, the system is considered not yet converged. The multi-agent game entropy reduction optimization unit uses the modified Nash equilibrium strategy combination obtained from the third round of game solving as the new initial strategy combination. It repeats the iterative process of introducing advanced state prediction values ​​for the second round of game solving and introducing a credibility penalty term for the third round of game solving until the reduction magnitude of the system entropy value output by the two adjacent rounds of game solving is less than the entropy reduction threshold twice consecutively. The iteration stops when the Nash equilibrium strategy combination obtained from the last iteration is output as the final fusion decision instruction.

[0059] The multi-source heterogeneous network state cyclic screening and fusion decision-making model also includes a two-way verification and self-learning evolution unit, which is connected to the chaos prediction unit, the cloud model credibility evaluation unit, and the multi-agent game entropy reduction optimization unit, respectively. The bidirectional verification and self-learning evolutionary unit is used to collect real network state data after the execution of fusion decision instructions and calculate the prediction error between the real network state data and the advanced state prediction value. The two-way verification and self-learning evolution unit is also used to positively feed the prediction error to the chaotic prediction unit to correct the phase space reconstruction parameters, to feed it to the cloud model credibility evaluation unit to update the digital features of the error cloud, and to feed it to the multi-agent game entropy reduction optimization unit to adjust the weight coefficients of the payoff function. The two-way verification and self-learning evolution unit is also used to periodically re-input historical real network state data into the chaos prediction unit, cloud model credibility assessment unit, and multi-agent game entropy reduction optimization unit to reenact the decision process, verify the decision effect after parameter correction, and solidify or roll back the parameter correction based on the verification results.

[0060] It should be further explained that, in the specific implementation, the bidirectional verification and self-learning evolution unit obtains the actual network state data generated after the execution of the fusion decision instruction from the multi-mode fusion access gateway. This actual network state data includes the signal-to-noise ratio, transmission delay, packet loss rate, and remaining bandwidth on the actually established fusion transmission path. Simultaneously, the bidirectional verification and self-learning evolution unit obtains the advanced state prediction value generated before the decision moment from the chaos prediction unit, and calculates the difference between the actual network state data and the corresponding advanced state prediction value as the prediction error. This prediction error includes signal-to-noise ratio prediction error, delay prediction error, and bandwidth prediction error.

[0061] The bidirectional verification and self-learning evolution unit encodes the prediction error and generates positive feedback instructions, which are then sent to the chaos prediction unit, the cloud model credibility assessment unit, and the multi-agent game entropy reduction optimization unit. Upon receiving the positive feedback instructions, the chaos prediction unit adjusts the delay time and embedding dimension used in its internal phase space reconstruction based on the magnitude of the prediction error. If the prediction error is consistently positive for several consecutive periods and its value increases, the delay time is reduced to enhance sensitivity to rapid changes; if the prediction error is consistently negative for several consecutive periods, the embedding dimension is increased to capture longer-term evolutionary patterns. Upon receiving the positive feedback instructions, the cloud model credibility assessment unit inputs the prediction error as a new error sample into its internal inverse cloud generator to update the expected value, entropy value, and hyperentropy value of the error cloud. Upon receiving the positive feedback instructions, the multi-agent game entropy reduction optimization unit adjusts the weight coefficients in the payoff functions of each game agent based on the prediction error. For network entities with large prediction errors, the weight of the bandwidth term in their payoff function is reduced, while the weight of the delay term is increased.

[0062] The bidirectional verification and self-learning evolution unit is equipped with a timer that triggers a reverse verification process at preset time intervals. After the reverse verification process is initiated, the bidirectional verification and self-learning evolution unit reads multiple sets of six-dimensional network state data and their corresponding real decision results stored in the previous time period from the historical database. These multiple sets of six-dimensional network state data are then re-input into the chaos prediction unit, the cloud model credibility assessment unit, and the multi-agent game entropy reduction optimization unit. These three units then re-execute a complete advance prediction, credibility assessment, and multi-round game entropy reduction optimization process and generate a new set of historical reenactment decision instructions.

[0063] The two-way verification and self-learning evolutionary unit compares the new set of historical reenactment decision instructions with the actual historical decision instructions executed in the previous time period, calculating the degree of difference between the two in terms of path selection, resource allocation ratio, and switching timing. If the degree of difference is less than a preset difference threshold, the parameters of each unit are determined to be set to a valid state and remain unchanged.

[0064] If the difference is greater than or equal to a preset difference threshold, the average system entropy value corresponding to the new historical reenactment decision instruction and the average system entropy value corresponding to the historical real decision instruction are further calculated. If the average system entropy value of the new historical reenactment decision instruction is lower than the average system entropy value of the historical real decision instruction, the parameter correction is deemed valid. The bidirectional verification and self-learning evolution unit stores the currently corrected parameters of the chaos prediction unit, cloud model credibility assessment unit, and multi-agent game entropy reduction optimization unit as benchmark parameters for subsequent operation. If the average system entropy value of the new historical reenactment decision instruction is not lower than the average system entropy value of the historical real decision instruction, the parameter correction is deemed invalid. The bidirectional verification and self-learning evolution unit sends a rollback instruction to the chaos prediction unit, cloud model credibility assessment unit, and multi-agent game entropy reduction optimization unit, restoring the parameters of each unit to their values ​​before the start of this reverse verification process.

[0065] The two-way verification and self-learning evolutionary unit also has a built-in experience knowledge base, which is used to store historical successful decision-making cases and their corresponding six-dimensional network state data feature vectors. The bidirectional verification and self-learning evolution unit is also used to perform similarity matching between the current six-dimensional network state data and the feature vectors in the experience knowledge base. When the matching similarity exceeds the preset threshold, the decision-making strategy of the matched historical successful decision-making case is directly called as the fusion decision instruction.

[0066] It should be further explained that, in the specific implementation, the bidirectional verification and self-learning evolution unit internally constructs an experience knowledge base. This experience knowledge base is stored in non-volatile memory in the form of a distributed hash table, used to store historical successful decision cases and their corresponding six-dimensional network state data feature vectors. Each historical successful decision case includes the fusion decision instruction executed at that time and the stable range of system entropy value reached after execution. Each six-dimensional network state data feature vector consists of normalized values ​​from six dimensions: the mean square error of the signal-to-noise ratio time series collected by the time-domain perception module, the interference frequency density collected by the frequency-domain perception module, the variance of node movement speed collected by the spatial-domain perception module, the average remaining power supply duration collected by the energy-domain perception module, the proportion of high-priority services collected by the service-domain perception module, and the temperature value collected by the environmental-domain perception module.

[0067] Before generating fusion decision instructions through multi-round game entropy reduction optimization, the bidirectional verification and self-learning evolutionary unit first obtains the current six-dimensional network state data from the six-dimensional state perception unit, and generates a feature vector of the current six-dimensional network state data according to the same feature extraction method as historical cases. The current feature vector is then compared with each historical feature vector stored in the experience knowledge base using Euclidean distance calculation; the smaller the Euclidean distance, the higher the similarity between the two feature vectors.

[0068] The bidirectional verification and self-learning evolutionary unit has a preset similarity threshold, which is expressed as the reciprocal of the Euclidean distance. When the reciprocal of the Euclidean distance between a historical feature vector and the current feature vector is greater than this similarity threshold, the two are considered to be successfully matched. The bidirectional verification and self-learning evolutionary unit reads the historical successful decision cases corresponding to the successfully matched historical feature vector from the experience knowledge base, and directly outputs the fusion decision instructions from the historical successful decision cases to the multi-mode fusion access gateway for execution. At the same time, the operation processes of the current chaos prediction unit, cloud model credibility assessment unit, and multi-agent game entropy reduction optimization unit are paused to reduce system power consumption.

[0069] If the reciprocal of the Euclidean distance between the current feature vector and all historical feature vectors in the experience knowledge base is not greater than the similarity threshold, then no successful matching case is determined. The bidirectional verification and self-learning evolution unit notifies the chaos prediction unit, cloud model credibility assessment unit, and multi-agent game entropy reduction optimization unit to initiate a complete process of advance prediction, credibility assessment, and multi-round game entropy reduction optimization to generate a fusion decision instruction. After the instruction is executed, the current six-dimensional network state data feature vector, the generated fusion decision instruction, and the stable range of system entropy value reached after the instruction execution are associated and stored in the experience knowledge base as new historical successful decision cases for subsequent matching.

[0070] The experience knowledge base also has a periodic cleanup mechanism. When the number of stored cases exceeds the preset capacity limit, historical cases are deleted in order of increasing system entropy value, while excellent cases with lower system entropy values ​​are retained first.

[0071] It also includes smart microgrid units, which are connected to the integrated decision control center. The smart microgrid units include photovoltaic power generation modules, diesel power generation modules, energy storage battery modules, and backup capacitor modules. The fusion decision control center is also used to treat the remaining power supply duration as a hard constraint condition for the game agent during the multi-round entropy reduction optimization process, based on the remaining power supply duration collected by the energy domain sensing module.

[0072] It should be further explained that, in the specific implementation, the smart microgrid unit includes a photovoltaic power generation module, a diesel power generation module, an energy storage battery module, and a backup capacitor module. The photovoltaic power generation module converts solar energy into DC power through solar photovoltaic panels, which is then connected to the DC bus via a maximum power point tracking controller. The diesel power generation module generates AC power from a synchronous generator driven by a diesel engine, which is then connected to the DC bus after being rectified. The energy storage battery module consists of a lithium-ion battery pack and a battery management system, and is connected to the DC bus via a bidirectional DC-DC converter. The backup capacitor module consists of a supercapacitor bank and an equalization control circuit, and is connected to the DC bus via another bidirectional DC-DC converter. Multiple DC-DC step-down converters are also connected to the DC bus to provide stable operating voltages to the multi-mode fusion access gateway, the six-dimensional state perception unit, the fusion decision control center, and the radio frequency components of each network entity.

[0073] The energy domain awareness module reads the state of charge and health status of the energy storage battery module through the battery management system, reads the remaining fuel level of the diesel generator module through the fuel level sensor, and monitors the output power of the photovoltaic generator module and the terminal voltage of the backup capacitor module in real time through the voltage and current sampling circuit. The energy domain awareness module calculates the remaining power supply time of the diesel generator module based on the remaining fuel level and the rated fuel consumption rate of the diesel generator module, the remaining power supply time of the energy storage battery module based on the state of charge of the energy storage battery module and the current total load power, and the remaining power supply time of the backup capacitor module based on the terminal voltage and capacitance of the backup capacitor module. The energy domain awareness module sends the remaining power supply time of each of the diesel generator module, energy storage battery module, and backup capacitor module, as well as the real-time output power of the photovoltaic generator module, as energy domain status data to the fusion decision control center.

[0074] The power supply switching logic of the smart microgrid unit adopts a priority-based switching rule. It prioritizes the use of photovoltaic power generation modules to provide clean energy power. When the photovoltaic power generation cannot meet the system load demand, it automatically switches to the energy storage battery module for power supply. If the energy storage battery power is lower than a preset threshold, the diesel power generation module is activated. The backup capacitor module provides instantaneous power compensation at the moment of power supply switching to avoid power interruption. The power allocation algorithm adopts an on-demand allocation strategy. Based on the real-time power consumption requirements of the multi-mode fusion access gateway, the six-dimensional state perception unit, and each communication module, the power supply is allocated according to the service priority. Communication components corresponding to high-priority services receive sufficient power supply first, while low-priority components are allocated power on demand. At the same time, the remaining capacity of each power supply module is monitored in real time, and the power output ratio is dynamically adjusted to ensure the stable operation of the power supply system.

[0075] Before the first round of the multi-agent game entropy reduction optimization unit, the fusion decision control center uses the remaining power supply duration sent by the energy domain sensing module as a hard constraint to filter the backup power supply duration of base stations in the telecommunications, China Unicom, and China Mobile networks. Specifically, for network agents corresponding to base stations whose backup power supply duration is less than the remaining power supply duration of the diesel generator module, the high-power 5G massive MIMO transmission mode is temporarily disabled in their available resource policy set. For self-organizing network node game agents, the remaining energy percentage of their nodes is coupled with the remaining power supply duration of the energy storage battery module to obtain an energy efficiency weight coefficient, which is directly multiplied into the payoff function of the self-organizing network node game agent.

[0076] While outputting fusion decision commands, the fusion decision control center also generates microgrid dispatch commands based on the power consumption requirements of network entities involved in the primary and backup fusion paths, and sends these commands to the smart microgrid unit. The smart microgrid unit, according to these dispatch commands, prioritizes allocating the real-time output power of the photovoltaic power generation modules to the radio frequency components on the primary fusion path. When the output power of the photovoltaic power generation modules is insufficient, the energy storage battery module and the diesel generator module are activated sequentially to supplement power supply. At the moment the diesel generator module starts up, the backup capacitor module provides surge power compensation to prevent DC bus voltage drops.

[0077] The convergence decision-making instructions include the primary convergence path, the backup convergence path, and the emergency fallback path; The primary fusion path is used to carry high-priority real-time services, the backup fusion path is used to carry medium-priority non-real-time services, and the emergency fallback path is the BeiDou short message communication queue, which is used to transmit critical coordinate data when all network links are interrupted.

[0078] It should be further explained that, in the specific implementation, the fusion decision control center uses the Nash equilibrium strategy combination output by the multi-agent game entropy reduction optimization unit to parse out three fusion transmission paths with different priorities and service levels.

[0079] The primary converged path consists of network links corresponding to the strategy combination that minimizes system entropy and maximizes payoff after game-theoretic entropy reduction optimization. Specifically, this primary converged path includes 5G network slices with the highest current signal-to-noise ratio and sufficient remaining bandwidth in telecommunications, Unicom, or mobile networks; Ku-band links with the most stable available frequency bands and sufficient power amplifier power margin in satellite broadband networks; and relay node combinations with the highest percentage of remaining energy and the fewest hops among self-organizing network nodes. The primary converged path is allocated to carry high-priority real-time services such as voice command and dispatch instructions, real-time video keyframes, and location information. The multi-mode converged access gateway reserves fixed bandwidth resources and queue buffers for these high-priority services on this path and enables forward error correction coding and automatic retransmission request mechanisms to ensure transmission reliability.

[0080] The backup converged path consists of network links corresponding to the strategy combination with the second-lowest system entropy value and the second-best payoff function value after game-theoretic entropy reduction optimization. Specifically, this backup converged path includes 4G network slices with moderate load and sufficient battery life in telecommunications, Unicom, or mobile networks; C-band links in satellite broadband networks that are currently available but pose some interference risk; and combinations of relay nodes with moderate remaining energy but wide coverage in self-organizing network nodes. The backup converged path is allocated to carry medium-priority non-real-time services such as environmental monitoring data, non-real-time image files, and log files. The multi-mode converged access gateway adopts a best-effort transmission mode for these medium-priority services on this path. When the primary converged path experiences congestion or interruption, the backup converged path can be temporarily upgraded to carry some critical services.

[0081] The emergency backup path is independent of the telecommunications network, China Unicom network, China Mobile network, satellite broadband network, and self-organizing network nodes. It consists of a BeiDou short message communication module and its independent antenna, independent power supply circuit, and independent buffer queue. The integrated decision control center triggers the emergency backup path activation command when it detects that all available network links are simultaneously interrupted or when the backup power supply duration of the base station, the satellite signal-to-noise ratio, and the remaining energy of the self-organizing network nodes reported by the six-dimensional status perception unit are all lower than the preset emergency threshold.

[0082] The multi-mode fusion access gateway encapsulates the coordinate data of the rescue team, the status data of key equipment, and the simple text instructions collected by the six-dimensional status perception unit according to the BeiDou short message communication protocol, generates a data packet that does not exceed the single transmission length limit of BeiDou short message, stores it in an independent cache queue, and sends the data packet to the BeiDou satellite through the BeiDou short message communication module at a preset transmission interval. At the same time, it listens to the downlink broadcast channel of the BeiDou satellite to receive acknowledgments or short message instructions from the ground command center.

[0083] The emergency backup power supply circuit is powered by a standby capacitor module of the smart microgrid unit. The standby capacitor module can maintain the Beidou short message communication module for at least 24 hours of operation when the main power supply of the system fails.

[0084] The emergency backup path is a complete emergency communication link with the BeiDou short message communication queue as the core data transmission carrier. The entire path includes a BeiDou short message communication module, an independent antenna, an independent power supply circuit, and a BeiDou short message communication queue. The BeiDou short message communication queue is used to buffer key coordinate data and simple instructions to be transmitted, while the BeiDou short message communication module, independent antenna, and independent power supply circuit provide hardware support for queue data transmission.

[0085] The multi-mode converged access gateway includes a telecom network access module, a Unicom network access module, a mobile network access module, a satellite broadband modem, and a self-organizing network radio frequency transceiver module. The telecom network access module, Unicom network access module, and mobile network access module support the 2G, 4G, and 5G network standards of their respective operators. The satellite broadband modem supports dynamic switching of multiple satellite frequency bands, and the self-organizing network radio frequency transceiver module supports multi-hop relay routing protocols.

[0086] It should be further explained that, in the specific implementation, the multi-mode converged access gateway, as the physical interface device for data exchange between the entire system and external network entities, has a motherboard, a backplane, and multiple pluggable communication module slots inside its chassis. Telecom network access modules, Unicom network access modules, mobile network access modules, satellite broadband modems, and self-organizing network RF transceiver modules are inserted into these multiple communication module slots as independent boards and connected to the motherboard via the backplane bus.

[0087] The telecom network access module integrates a multi-band RF transceiver chip, a baseband processing chip, and a user identification card slot corresponding to the telecom operator. The multi-band RF transceiver chip supports the 2G, 4G, and 5G frequency bands allocated by the telecom network. The baseband processing chip has a built-in telecom network protocol stack and supports network slicing selection. The user identification card slot contains a user identification card provided by the telecom operator for network authentication and access. The internal structure of the China Unicom and China Mobile network access modules is similar to that of the telecom network access module, integrating multi-band RF transceiver chips, baseband processing chips, and user identification card slots corresponding to China Unicom and China Mobile, respectively, and each supporting the corresponding operator's 2G, 4G, and 5G network standards.

[0088] The satellite broadband modem integrates an intermediate frequency (IF) processing unit, an analog-to-digital converter (ADC), and a programmable digital signal processor (PDS). The PDS is loaded with software code supporting modulation and demodulation algorithms for multiple satellite frequency bands, including C-band, Ku-band, and Ka-band. Based on the target satellite frequency band and modulation scheme specified in the fusion decision-making instructions sent by the fusion decision-making control center, the satellite broadband modem dynamically loads the corresponding software code and adjusts the local oscillator frequency of the IF processing unit to achieve frequency band switching and parameter reconfiguration for the satellite link.

[0089] The ad hoc network RF transceiver module integrates a software-defined radio (SDK) architecture-based RF front-end and a field-programmable gate array (FPGA). The FPGA contains hardware logic supporting a multi-hop relay routing protocol for ad hoc networks, including an on-demand distance vector routing protocol and an optimized link-state routing protocol. The ad hoc network RF transceiver module employs a combination of these two protocols. The on-demand distance vector routing protocol dynamically finds transmission paths between ad hoc network nodes, initiating path requests only when data transmission is needed, thus reducing node power consumption. The optimized link-state routing protocol maintains the ad hoc network topology in real time, quickly sensing changes in node location and link status. The routing calculation method comprehensively evaluates link quality, remaining energy, and transmission distance between nodes, prioritizing nodes with superior link quality, sufficient remaining energy, and fewer hops as relay nodes. The routing table is updated by periodically broadcasting link-state information. Those skilled in the art can complete the protocol deployment and path calculation using conventional methods for ad hoc network multi-hop relay implementations. The self-organizing network RF transceiver module dynamically updates the routing table and adjusts the bias voltage of the power amplifier at the RF front end to change the transmit power based on the relay path and transmit power level specified in the fusion decision command.

[0090] The hardware interface of the multi-mode converged access gateway adopts standardized industrial communication interfaces, including Ethernet interface, universal serial bus interface, radio frequency interface and serial port. The Ethernet interface is used for data interaction with the converged decision control center. The radio frequency interface is connected to the base stations of the three major operators, satellite antenna and self-organizing network radio frequency unit respectively. The serial port is used for status parameter configuration and debugging. The interface protocol adopts the general TCP / IP network protocol, radio frequency communication standard protocol and satellite communication dedicated protocol. The data interaction format adopts a standardized data packet structure, including frame header, data field, check bit and frame trailer. The frame header is used to identify the data type and source, the data field carries business data and status information, and the check bit is used to ensure the accuracy of data transmission. Those skilled in the art can implement it by using the interface design and protocol adaptation methods of conventional communication gateways.

[0091] The motherboard connects to the telecom network access module, the Unicom network access module, the mobile network access module, the satellite broadband modem, and the self-organizing network RF transceiver module via a backplane bus. The central processing unit on the motherboard runs a gateway control program. This program receives convergence decision instructions from the convergence decision control center, parses the network module identifiers and resource allocation parameters corresponding to the primary and backup convergence paths contained in the instructions, and generates module control commands based on these identifiers and parameters, sending them to the corresponding network modules. Simultaneously, the gateway control program is also responsible for aggregating and queuing uplink data streams received from each network module according to service priority, and distributing downlink data streams from on-site rescue terminal equipment to the corresponding network modules for transmission.

[0092] This system achieves collaborative scheduling of five-dimensional heterogeneous network resources, including telecommunications networks, China Unicom networks, mobile networks, satellite broadband networks, and self-organizing network nodes, through the multi-source heterogeneous network state cyclic screening and fusion decision model built into the fusion decision control center.

[0093] The model first reconstructs the phase space and calculates the maximum Lyapunov exponent from the six-dimensional network state data to identify the chaotic characteristics of the key parameter time series and make advanced state predictions. Then, it quantifies the prediction results into confidence cloud parameters containing expected value, entropy value, and hyperentropy value through inverse cloud generator and forward cloud generator, and classifies the uncertainty of the advanced prediction value. Next, it defines the five network entities as game agents, with the goal of minimizing the system entropy function. It conducts multiple rounds of game iteration by introducing advanced prediction value and confidence penalty term until the reduction of system entropy value in two adjacent rounds converges, and then outputs the fusion decision instruction.

[0094] This mechanism enables the system to anticipate network state evolution trends, quantitatively assess the reliability of prediction results, and spontaneously evolve towards the state with the most balanced resource utilization and the highest overall orderliness during multi-network collaboration, effectively improving the integration efficiency of multi-system networks in dynamic environments.

[0095] In field applications, this system can dynamically establish a three-tiered transmission guarantee system—primary fusion path, backup fusion path, and emergency fallback path—based on the real-time priority of rescue operations. The primary fusion path carries high-priority services such as voice commands and video keyframes, the backup fusion path transmits non-real-time services such as environmental monitoring data, and the emergency fallback path maintains minimum backhaul capability of coordinate data through an independent power supply module powered by BeiDou short messages when all network links are interrupted.

[0096] The deep coupling between the smart microgrid unit and the energy domain sensing module enables the system to adjust the resource strategy set of each network in real time based on the backup power endurance of the base station and the remaining energy of the nodes, prioritizing the operating time of key radio frequency components in power-constrained scenarios. The bidirectional verification and self-learning evolution unit corrects the parameters of each model through positive error feedback, and verifies and solidifies the effective adjustments through reverse history replay. At the same time, the case matching mechanism of the experience knowledge base can directly call historical successful strategies in similar scenarios.

[0097] The entire system improves the stability and resource utilization efficiency of multi-network converged communication in extreme disaster environments through a closed-loop architecture of multi-dimensional perception, advanced prediction, reliable assessment, game optimization and self-learning evolution.

[0098] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0099] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An emergency communication satellite mobile broadband triple-play system, characterized in that, include: Multimode converged access gateway is used to provide communication interfaces with telecommunications networks, China Unicom networks, mobile networks, satellite broadband networks, and self-organizing network nodes; The six-dimensional state perception unit is used to collect six-dimensional network state data in the time domain, frequency domain, spatial domain, energy domain, service domain, and environmental domain. The fusion decision control center is connected to the six-dimensional state perception unit and the multi-mode fusion access gateway, respectively. The fusion decision control center has a built-in multi-source heterogeneous network state cyclic screening and fusion decision model. The model is used to perform nonlinear dynamic prediction on the six-dimensional network state data to generate advanced state prediction values, and to perform uncertainty quantification on the advanced state prediction values ​​to generate credibility evaluation parameters. Then, based on the credibility evaluation parameters and the six-dimensional network state data, multi-round game entropy reduction optimization is performed to generate the fusion decision instruction with the minimum system entropy value. The multi-mode converged access gateway responds to the converged decision command by establishing a dynamic converged transmission path among the telecommunications network, the China Unicom network, the mobile network, the satellite broadband network, and the self-organizing network node.

2. The emergency communication satellite mobile broadband triple-play system according to claim 1, characterized in that: The six-dimensional state sensing unit includes: The time-domain sensing module is used to acquire the time-domain waveform and signal-to-noise ratio time series of network signals; The frequency domain sensing module is used to collect spectrum occupancy and interference frequency distribution; The airspace awareness module is used to collect the real-time location of self-organizing network nodes and the boundary of satellite coverage area; The energy domain sensing module is used to collect the backup power supply duration of the base stations of the telecommunications network, the Unicom network and the mobile network, the power amplifier power margin of the satellite broadband network, the remaining energy of the self-organizing network node and the remaining power supply duration of the microgrid unit. The service domain awareness module is used to collect the type, priority, and quality of service requirements of the current transmission service; The environmental domain sensing module is used to collect environmental temperature, humidity, and air pressure parameters.

3. The emergency communication satellite mobile broadband triple-play system according to claim 1, characterized in that: The multi-source heterogeneous network state cyclic screening and fusion decision model includes a chaos prediction unit. The chaos prediction unit is used to reconstruct the phase space of the key parameter time series in the six-dimensional network state data, calculate the maximum Lyapunov exponent of the reconstructed phase space to determine the chaotic characteristics of the key parameter time series, and use a weighted first-order local method to predict the future state of the key parameter time series in advance, generating the advanced state prediction value.

4. The emergency communication satellite mobile broadband triple-play system according to claim 3, characterized in that: The multi-source heterogeneous network state cyclic screening and fusion decision model also includes a cloud model credibility assessment unit, which is connected to the chaos prediction unit. The cloud model credibility assessment unit is used to collect historical prediction error data and generate digital features of the error cloud. The digital features include expected value, entropy value and hyperentropy value. The cloud model credibility assessment unit is also used to take the deviation between the advanced state prediction value and the current true value as cloud droplet input, and combine the digital features of the error cloud to generate credibility cloud parameters of the advanced state prediction value. The credibility cloud parameters include prediction expectation value, prediction entropy value and prediction over-entropy value. The cloud model credibility assessment unit is also used to classify the credibility of the advanced state prediction value based on the prediction entropy value and the prediction hyper-entropy value.

5. The emergency communication satellite mobile broadband triple-play system according to claim 4, characterized in that: The multi-source heterogeneous network state cyclic screening and fusion decision model further includes a multi-agent game entropy reduction optimization unit, which is connected to the cloud model credibility evaluation unit. The multi-agent game entropy reduction optimization unit is used to define the telecommunications network, the Unicom network, the mobile network, the satellite broadband network, and the self-organizing network node as five game agents, each of which has its own set of available resource strategies. The multi-agent game entropy reduction optimization unit is also used to construct a system entropy function, which is the sum of the resource utilization Shannon entropies of the five game agents; the multi-agent game entropy reduction optimization unit is also used to execute the first round of the game to solve the Nash equilibrium strategy combination in the current state and calculate the first system entropy value. The multi-agent game entropy reduction optimization unit is also used to take the advanced state prediction value as the game environment input, execute the second round of game to solve the Nash equilibrium strategy combination in the future state, and calculate the second system entropy value. The multi-agent game entropy reduction optimization unit is also used to use the credibility cloud parameter as the penalty term of the payoff function, execute the third round of game to solve the modified Nash equilibrium strategy combination, and calculate the entropy value of the third system. The multi-agent game entropy reduction optimization unit is also used to compare the entropy values ​​of the first system, the second system, and the third system. When the reduction in the entropy value of the system in two adjacent rounds is less than the preset entropy reduction threshold, the Nash equilibrium strategy combination of the third round of the game is output as the fusion decision instruction.

6. The emergency communication satellite mobile broadband triple-play system according to claim 5, characterized in that: The multi-source heterogeneous network state cyclic screening and fusion decision model further includes a two-way verification and self-learning evolution unit, which is connected to the chaos prediction unit, the cloud model credibility evaluation unit and the multi-agent game entropy reduction optimization unit, respectively. The bidirectional verification and self-learning evolution unit is used to collect real network state data after the execution of the fusion decision instruction, and to calculate the prediction error between the real network state data and the advanced state prediction value. The bidirectional verification and self-learning evolution unit is also used to positively feed the prediction error back to the chaotic prediction unit to correct the phase space reconstruction parameters, feed it back to the cloud model credibility evaluation unit to update the digital features of the error cloud, and feed it back to the multi-agent game entropy reduction optimization unit to adjust the payoff function weight coefficients. The bidirectional verification and self-learning evolution unit is also used to periodically re-input historical real network state data into the chaos prediction unit, the cloud model credibility assessment unit, and the multi-agent game entropy reduction optimization unit to reenact the decision process, verify the decision effect after parameter correction, and solidify or roll back the parameter correction based on the verification results.

7. The emergency communication satellite mobile broadband triple-play system according to claim 6, characterized in that: The bidirectional verification and self-learning evolution unit also has a built-in experience knowledge base, which is used to store historical successful decision-making cases and their corresponding six-dimensional network state data feature vectors. The bidirectional verification and self-learning evolution unit is also used to perform similarity matching between the current six-dimensional network state data and the feature vectors in the experience knowledge base. When the matching similarity exceeds a preset threshold, the decision strategy of the matched historical successful decision case is directly called as the fusion decision instruction.

8. The emergency communication satellite mobile broadband triple-play system according to claim 1, characterized in that: It also includes a smart microgrid unit, which is connected to the fusion decision control center. The smart microgrid unit includes a photovoltaic power generation module, a diesel power generation module, an energy storage battery module, and a backup capacitor module. The fusion decision control center is also used to use the remaining power supply duration collected by the energy domain sensing module as a hard constraint condition for the game agent during the multi-round game entropy reduction optimization process.

9. The emergency communication satellite mobile broadband triple-play system according to claim 1, characterized in that: The fusion decision-making instructions include a primary fusion path, a backup fusion path, and an emergency fallback path; The primary fusion path is used to carry high-priority real-time services, the backup fusion path is used to carry medium-priority non-real-time services, and the emergency fallback path is the BeiDou short message communication queue, which is used to transmit critical coordinate data when all network links are interrupted.

10. The emergency communication satellite mobile broadband triple-play system according to claim 1, characterized in that: The multi-mode converged access gateway includes a telecom network access module, a Unicom network access module, a mobile network access module, a satellite broadband modem, and a self-organizing network radio frequency transceiver module. The telecom network access module, the Unicom network access module, and the mobile network access module respectively support the 2G, 4G, and 5G network standards of their respective operators. The satellite broadband modem supports dynamic switching of multiple satellite frequency bands, and the self-organizing network radio frequency transceiver module supports multi-hop relay routing protocols.

Images