A sea fleet communication relay pre-switching method based on sea state prediction

Abstract

This invention discloses a pre-handover method for maritime fleet communication relays based on sea state prediction, comprising: Step 1, collecting and fusing multi-source data; Step 2, constructing a joint digital twin simulation engine for sea state and electromagnetic environment, inputting multi-source data into the digital twin simulation engine, and calculating the expected channel quality parameters and prediction confidence of each relay link in each time period within the next N hours through multi-physics coupling simulation; Step 3, arranging the expected channel quality parameters of each link in the future time period in chronological order to generate the expected channel quality change curve of each relay link; Step 4, conducting a cascade failure risk assessment, and then determining whether to execute a step-by-step link handover based on the pre-handover trigger; Step 5, with the optimization objective of minimizing the weighted sum of handover interruption risk, link resource waste, cascade failure loss, and information timeliness degradation, jointly solving for the optimal pre-handover timing, optimal target link, and optimal relay type combination, and performing adaptive handover.

Description

Technical Field

[0001] This invention relates to the field of communication relay pre-handover, and more particularly to a method for communication relay pre-handover of maritime fleets based on sea state prediction. Background Technology

[0002] With the expansion of maritime fleet operations and the growth of communication demands, maritime wireless communication relay systems play a crucial role in ensuring the stability of fleet communications. The maritime communication environment differs fundamentally from that on land: sea surface reflections cause strong multipath effects, atmospheric waveguide phenomena at sea cause abnormal signal propagation, high salt spray and high humidity environments accelerate equipment aging, and sudden strong winds and waves cause physical damage and positional displacement to relay equipment. These factors make the channel quality of maritime relay links highly time-varying and unpredictable.

[0003] Currently, various technical solutions exist in the field of maritime communication relay. For example, patent CN115665691A discloses a maritime wireless backhaul networking method based on information perception, which selects backhaul nodes through channel perception, interference perception, and energy perception, avoiding interference-affected nodes and low-energy nodes from participating in data backhaul. Patent CN120547646A discloses a dynamic maritime communication method based on signal boat relay chain construction, which uses a multi-objective particle swarm optimization algorithm for signal boat deployment and performs path selection and dynamic switching through path stability scoring. Patent CN116743227A discloses a satellite network switching system based on matching satellite coverage areas with ship positioning information. In addition, there are self-organizing network routing methods based on ship formation topology perception and maritime channel access methods based on multi-path transmission, etc.

[0004] However, the aforementioned existing technical solutions all share the following limitations:

[0005] (1) Reactive handover decision-making: Existing technologies are all based on "current state awareness" for reactive decision-making, that is, the system monitors the current channel quality, interference level, node energy and other parameters in real time, and only triggers relay handover or route reconstruction when it detects that the current link quality does not meet the requirements. This reactive method is basically feasible in the land communication environment because the channel quality changes relatively slowly on land. However, in the marine environment, the weather conditions change drastically and rapidly. Reactive handover needs to go through the complete process of "channel deterioration → detection of deterioration → evaluation and selection of new link → execution of handover". In this process, data transmission interruption or loss has already occurred.

[0006] (2) Lack of deep correlation modeling between sea state and channel quality: Although existing technologies involve channel quality assessment, they have not established a systematic mapping model from multiple sea state parameters (wind speed, wave height, visibility, ocean current, salinity, etc.) to multi-dimensional channel quality parameters. In particular, they lack comprehensive modeling of unique marine propagation effects such as evaporation waveguides, sea surface scattering, and salt spray attenuation.

[0007] (3) Single relay type and lack of heterogeneous integration: Existing technologies are mostly aimed at single type of relay nodes (such as only shipborne relay or only satellite relay), without considering the collaborative optimization selection in heterogeneous relay environments (shipborne, buoy, UAV and satellite coexistence).

[0008] (4) Ignoring the cascading failure effect: When making relay switching decisions, the existing technology only considers the quality of the link to be switched, without predicting the cascading impact of the link failure on the overall relay network topology connectivity.

[0009] (5) No prediction uncertainty quantification: Existing prediction-based methods do not quantify the uncertainty of prediction results and cannot adaptively adjust the switching strategy according to the credibility of the prediction.

[0010] (6) Lack of online learning capability: The channel model parameters of the existing methods are static preset values, which cannot be continuously optimized based on the measured data accumulated in actual operation, resulting in a decrease in prediction accuracy under new sea state conditions.

[0011] Therefore, there is an urgent need for a pre-switching method for maritime fleet communication relays that can comprehensively solve the above problems. Summary of the Invention

[0012] Purpose of the invention: The technical problem to be solved by the present invention is to provide a pre-switching method for maritime fleet communication relay based on sea state prediction, which addresses the shortcomings of the existing technology.

[0013] To address the aforementioned technical problems, this invention discloses a pre-handover method for maritime fleet communication relays based on sea state prediction. This method uses a digital twin simulation engine to jointly predict sea state and channel quality. Through cascaded failure risk assessment and multi-objective joint optimization, it adaptively selects the optimal pre-handover timing and target link among heterogeneous relay nodes, and continuously improves prediction accuracy through online learning. The pre-handover method for maritime fleet communication relays based on sea state prediction includes the following steps:

[0014] Step 1: Multi-source data acquisition and fusion: The relay management system acquires meteorological forecast data for the target sea area for the next N hours (N ranges from [6, 48] hours, set according to the severity of sea state changes and communication reliability requirements) from marine meteorological forecast data sources. Simultaneously, it acquires the current position, speed, heading information, and navigation plan data of each relay node and each vessel in the fleet through the AIS system (Automatic Identification System). It also acquires the type identifier (shipborne relay, buoy relay, UAV relay, or satellite relay), current remaining energy, and energy harvesting rate of each relay node; UAV relay refers to unmanned aerial vehicle (UAV) relay.

[0015] The meteorological forecast data mentioned in step 1 includes wind speed, wave height, visibility, precipitation, ocean current speed, sea surface temperature, salinity, and atmospheric pressure;

[0016] Step 2, Joint digital twin modeling of sea state and channel: Construct a joint digital twin simulation engine of sea state and electromagnetic environment. Input the multi-source data obtained in Step 1 into the digital twin simulation engine. Calculate the expected channel quality parameters and prediction confidence of each relay link in the next N hours (N ranges from [6, 48] hours, set according to the degree of sea state change and communication reliability requirements) through multi-physics coupling simulation.

[0017] The expected channel quality parameters mentioned in step 2 include path loss, multipath fading depth, bit error rate, delay spread, and Doppler spread; the prediction confidence level characterizes the reliability of the digital twin model for the prediction results.

[0018] The engine integrates a sea surface electromagnetic environment simulation module, a multipath propagation simulation module, a relay node motion simulation module, and an energy supply prediction module.

[0019] Step 3, Multidimensional Link Quality Trend Prediction: Arrange the expected channel quality parameters of each link in the future time period obtained in Step 2 in chronological order to generate the "expected channel quality change curve" for each relay link. The change curve includes the predicted value and confidence interval, which characterizes the channel quality change trend and its uncertainty of each link in the next N hours (N ranges from [6, 48] hours, set according to the severity of sea state changes and communication reliability requirements).

[0020] Step 4, Joint Judgment of Cascade Failure Risk Assessment and Pre-Switch Trigger: First, a cascade failure risk assessment is performed. Based on the current relay network topology and the expected channel quality change curves of each link, the probability of cascade impact on the connectivity of other links in the network when the channel quality of a certain link falls below the threshold is calculated as P_cascade. Then, a pre-switch trigger judgment is performed: The expected channel quality change curve of the currently active relay link is compared with the preset channel quality threshold Q_threshold. At the same time, the cascade failure probability P_cascade and the prediction confidence C(t) are considered. If the comprehensive risk index exceeds the preset trigger threshold, the pre-switch process is triggered to determine whether to execute the link switch in Step 5; otherwise, the current link is maintained, and the data is periodically returned to Step 1 for updates.

[0021] Step 5, Multi-objective Joint Optimization: With the optimization objective of minimizing the weighted sum of handover interruption risk, link resource waste, cascade failure loss, and information timeliness degradation, the optimal pre-handover timing (t_switch), optimal target link, and optimal relay type combination are jointly solved from a candidate target link set composed of heterogeneous relay nodes of different types. Adaptive handover is then performed. The selection of candidate target links comprehensively considers the expected channel quality change curves and confidence intervals of each link, cascade failure probability, the future positions of each vessel in the fleet's navigation plan, the current energy status and future energy prediction curves of each relay node, and the type characteristics (coverage range, mobility, delay characteristics) of each candidate relay node.

[0022] Step 6: Post-Switch Feedback Learning and Online Model Update: After the switch is completed, the relay management system collects the actual channel quality parameters of the target link, compares them with the predicted values ​​from Step 2, calculates the prediction error, and uses online learning algorithms to update the model parameters of each sub-module in the digital twin simulation engine, so that the subsequent prediction accuracy is gradually improved.

[0023] The sea state-electromagnetic environment joint digital twin simulation engine in step 2 includes the coupled simulation of the following sub-modules:

[0024] The sea surface electromagnetic environment simulation module calculates the atmospheric refractive index profile based on the evaporation waveguide model, calculates the sea surface scattering coefficient by combining the sea surface roughness parameter (root mean square wave height σ_h determined by wind speed and wave height), comprehensively considers the additional attenuation of electromagnetic waves caused by the salt spray environment, and outputs the electromagnetic propagation environment parameters from the sea surface to the lower atmosphere.

[0025] Multipath propagation simulation module: Based on the dual-ray propagation model (direct wave + sea surface reflected wave), the basic propagation loss is calculated, the diffuse scattering component caused by sea surface scattering is superimposed, atmospheric waveguide anomalous propagation correction is introduced, atmospheric scattering attenuation is calculated in combination with visibility parameters, and the path loss L(t), multipath fading depth F(t), and time delay spread σ_τ(t) of each link are output (where σ is the standard deviation sign and τ is the multipath time delay variable).

[0026] Relay node motion simulation module: For shipborne relay nodes, predict the future position coordinates and attitude changes of each ship based on navigation plans and sea state data (the impact of wind, waves, and currents on ship motion); for buoy relay nodes, predict the drift trajectory of buoys based on ocean current and wind and wave data; for UAV relay nodes, predict changes in their flight envelope and communication link geometry based on wind speed and visibility.

[0027] Energy supply prediction module: For relay nodes that rely on renewable energy (solar energy, wave energy), the expected energy change curve E_forecast(t) of each node in the next N hours (N ranges from [6, 48] hours, set according to the severity of sea state changes and communication reliability requirements) is calculated based on future sea state forecasts (cloud cover affects solar energy and wave height affects wave energy collection efficiency).

[0028] Each submodule is coupled through a time synchronization interface: the electromagnetic environment parameters of the sea surface serve as the input for multipath propagation simulation, the motion state of the relay node serves as the geometric parameter input for propagation simulation, and the energy prediction results serve as the constraint conditions for the availability of the relay node.

[0029] The method for calculating the prediction confidence C(t) in step 2 is as follows:

[0030] C(t) = f(C_forecast, C_model, C_history),

[0031] Wherein, C_forecast is the confidence level of the weather forecast data itself (provided by the weather forecast data source), C_model is the historical prediction accuracy of the digital twin model under the current sea state conditions (determined by the feedback learning results in step 6), and C_history is the decay factor of the prediction confidence on the time interval between the current time and the forecast time; the value of C(t) is in the range of [0,1], and the higher the value, the more reliable the prediction result.

[0032] The specific method for assessing the cascading failure risk in step 4 is as follows: Based on the topology G(V,E) of the current heterogeneous relay network, the expected channel quality change curves of each link e∈E are transformed into link failure probabilities p_e(t); the network reliability analysis method is used to calculate the decrease in network connectivity (the probability that there is at least one path from the source node to the destination node) when link e fails, which is taken as the cascading failure probability P_cascade(e); when P_cascade(e) exceeds the preset cascading risk threshold P_cascade_th, it is determined that the failure of the link will trigger cascading risk and should be given priority in the pre-switching decision.

[0033] The specific method of heterogeneous relay selection in step 5 is as follows: The candidate relay nodes are classified into four categories according to their types - shipborne relays (advantages: flexible movement, sufficient power; disadvantages: position restricted by navigation plans), buoy relays (advantages: fixed position, long-term deployment; disadvantages: power limited, vulnerable to sea conditions), UAV relays (advantages: flexible coverage, rapid deployment; disadvantages: limited endurance, restricted by wind speed), and satellite relays (advantages: large coverage area; disadvantages: high latency, limited bandwidth); in the multi-objective joint optimization, the characteristic parameters of each type of relay node are introduced into the objective function as constraint conditions and optimization variables to solve the optimal heterogeneous relay type combination and the target link.

[0034] Specifically, the execution of the adaptive switching in step 5 includes: when the optimal pre-switching time t_switch arrives, the switching strategy is adaptively selected according to the current prediction confidence C(t_switch): when C(t_switch) ≥ C_high, an aggressive switching strategy is adopted to directly migrate the communication data stream from the current active link to the optimal target link; when C(t_switch) < C_low, a conservative switching strategy is adopted to extend the dual-link parallel transmission window period and increase the acknowledgment threshold; when C_low ≤ C(t_switch) < C_high, a standard switching strategy is adopted; all strategies use the dual-link parallel transmission mechanism to ensure data integrity, and the original link resources are released after the migration is completed.

[0035] Specifically, the setting method of the prediction confidence thresholds C_high and C_low is: C_high = 0.8, C_low = 0.5; when C(t_switch) ≥ 0.8, the initial value of the dual-link parallel transmission window period T_parallel is set to 50 ms; when C(t_switch) < 0.5, the initial value of T_parallel is set to 200 ms, and the acknowledgment threshold N_ack is increased to twice the standard value; when 0.5 ≤ C(t_switch) < 0.8, T_parallel is determined by linear interpolation between 50 ms and 200 ms according to C(t_switch).

[0036] The method also includes an emergency sub-process for extreme sea conditions: when meteorological forecast data shows that extreme sea conditions (wind speed ≥ 12 or wave height ≥ 8m or visibility < 50m) will occur in the future, the relay management system will skip the normal pre-switching optimization calculation and directly execute the following emergency operations: (a) enable the maximum power coverage mode of all available heterogeneous relay nodes, and prioritize the relay links such as UAV relay and satellite relay that are less affected by sea conditions; (b) broadcast extreme weather warning signals to all ships in the fleet; (c) upgrade all communication service levels to the highest priority, suspend non-critical data transmission, and concentrate relay resources to ensure the transmission of command instructions and safety warning information; (d) dynamically adjust the communication mode and automatically switch to a degraded communication protocol (reduce data rate and increase error correction redundancy) according to the currently available relay resources to ensure minimum communication guarantee capability under extreme conditions.

[0037] The method for acquiring meteorological forecast data in step 1 includes: the relay management system periodically acquires meteorological forecast data for the target sea area through the marine meteorological service interface, with an acquisition period of 30 minutes to 2 hours, adaptively adjusted according to the severity of sea state changes and the prediction confidence C(t): when C(t) decreases or sea state changes intensify, the acquisition period is shortened; when new meteorological forecast data is acquired, the relay management system updates the digital twin simulation results in step 2 and the expected channel quality change curve in step 3, and re-executes the joint judgment in step 4; the temporal resolution of the meteorological forecast data is 1 hour, and the spatial resolution is the gridded area of ​​the sea area covered by the relay link.

[0038] Beneficial effects:

[0039] (1) High-precision sea state-channel joint prediction based on digital twin: Unlike the simple linear mapping model of the existing technology, this invention constructs a digital twin simulation engine that includes four sub-modules: sea surface electromagnetic environment simulation, multipath propagation simulation, relay node motion simulation and energy supply prediction. Through multi-physics coupling, it realizes high-precision prediction from multiple sea state parameters to multi-dimensional channel quality parameters, and outputs prediction confidence to provide decision-making basis for adaptive switching strategy.

[0040] (2) Cascade failure risk assessment to avoid link "domino" collapse: This invention introduces a cascade failure risk assessment mechanism for the first time in the decision-making of marine relay switching, predicts the cascade impact of single link channel deterioration on the topological connectivity of the entire heterogeneous relay network, and avoids the problem that "the selected target link itself will also fail soon" caused by making switching decisions based solely on the quality of a single link.

[0041] (3) Adaptive handover based on prediction confidence, balancing efficiency and reliability: This invention dynamically adjusts the aggressiveness of the handover strategy based on the prediction confidence—handover is initiated early under high confidence conditions to obtain the maximum handover advance, while the parallel transmission window is extended and the acknowledgment threshold is increased under low confidence conditions to ensure handover reliability, thus achieving an adaptive balance between handover efficiency and reliability.

[0042] (4) Heterogeneous relay fusion selection to maximize communication support capability: This invention supports the fusion selection of four heterogeneous relay types: shipborne relay, buoy relay, UAV relay and satellite relay. Differentiated selection is made according to the advantages and disadvantages of each type of relay under different sea conditions to maximize the communication support capability under severe sea conditions.

[0043] (5) Online feedback learning to continuously improve prediction accuracy: This invention introduces a post-switching feedback learning mechanism, which continuously optimizes the parameters of the digital twin model through an incremental Bayesian update method, so that the model continuously approaches the real sea state-channel mapping relationship in actual operation, overcoming the problem of decreased prediction accuracy of static models under new sea state conditions.

[0044] (6) Multi-factor joint optimization to comprehensively improve the quality of handover decision-making: The selection of pre-handover timing, target link and relay type is solved by multi-objective joint optimization to find the global optimal solution. It comprehensively considers five dimensions: expected channel quality (including confidence interval), cascade failure probability, fleet navigation plan (future position), relay node energy status (including future energy prediction) and heterogeneous relay type characteristics, thus avoiding the limitations of single-factor decision-making. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the overall process of the pre-handover method for maritime fleet communication relay based on sea state prediction of the present invention.

[0046] Figure 2 This is a schematic diagram of the structure of the sea state-channel joint digital twin model of the present invention;

[0047] Figure 3 This is a schematic diagram of the adaptive switching strategy for prediction confidence in this invention;

[0048] Figure 4 This is a schematic diagram illustrating the cascade failure risk assessment and heterogeneous relay selection of the present invention;

[0049] Figure 5 This is a sequence diagram of the confidence-adaptive switching execution and information timeliness guarantee of the present invention;

[0050] Figure 6 This is a deployment architecture diagram of the heterogeneous relay communication system for a maritime fleet according to the present invention. Detailed Implementation

[0051] This embodiment is set to normal sea conditions (wind force 4, wave height 1.5m, visibility 8km).

[0052] Assume a maritime fleet consists of one command ship (S0) and five operational vessels, sailing along a near-shore channel at a speed of 12 knots (approximately 6.17 m / s) and a heading of 090 degrees. The communication relay system comprises five heterogeneous relay nodes: buoy relay R1 (coordinates (0,0)), shipborne relay R2 (coordinates (2000,0), deployed on an operational vessel), buoy relay R3 (coordinates (4000,0)), UAV relay R4 (hovering at coordinates (1000,500), flight altitude 200 m), and satellite relay R5 (in geostationary orbit). The command ship S0 is located at coordinates (-500,0). The currently active communication link is: S0->R1->R2->R3->shore base station. Each node operates at a frequency f = 5.8 GHz, a transmit power Pt = 30 dBm, a receive sensitivity Pr_min = -90 dBm, and a channel quality threshold Q_threshold = 0.4.

[0053] Step 1 Implementation -- Multi-source data acquisition and fusion: The relay management system obtains the meteorological forecast data for the target sea area for the next 12 hours through the marine meteorological service interface (data source: National Marine Environmental Forecasting Center Global Marine Meteorological Forecasting System, time resolution 1h, spatial resolution 0.25 degrees x 0.25 degrees). Taking this acquisition as an example, the measured data and future forecast data at the current time t0 are as follows: (a) Meteorological parameters (current value -> forecast value 4 hours later): wind speed v_w: 6m / s -> 14m / s (level 4 -> level 7); wave height H_s: 1.5m -> 3.5m; visibility V: 8km -> 2km; precipitation R: 0 -> 15mm / h; ocean current speed v_c: 0.5m / s -> 1.2m / s; sea surface temperature T_s: 22℃ -> 19℃; salinity S: 34.5psu -> 35.0psu; atmospheric pressure P_a: 1013hPa -> 1005hPa. (b) AIS data (real-time acquisition): S0 coordinates (-500, 0), speed 6.17 m / s, heading 090 degrees; R1 coordinates (0, 0); R2 coordinates (2000, 0), speed 6.17 m / s; R3 coordinates (4000, 0); R4 hovering coordinates (1000, 500), flight altitude 200 m. (c) Status of each relay node: R1 has 85% remaining energy, solar power P_sol = 15 W; R2 has 70% remaining energy, powered by the main ship; R3 has 90% remaining energy, solar power P_sol = 15 W; R4 (UAV) has 60% remaining power, battery capacity 500 Wh, hovering power consumption 250 W; R5 (satellite) is in normal status. (d) Relay node type identifiers: R1=buoy relay, R2=shipborne relay, R3=buoy relay, R4=UAV relay, R5=satellite relay.

[0054] Step 2 Implementation – Joint Digital Twin Modeling of Sea State and Channel: The digital twin simulation engine receives the multi-source data from Step 1 and calculates the expected channel quality parameters of each relay link in future time periods through coupled simulation of four sub-modules. The following detailed calculation process is given using link R1->R2 (link distance d=2000m, R1 antenna height h1=3m, R2 antenna height h2=8m) as an example:

[0055] (a) Calculation process of the sea surface electromagnetic environment simulation module: ① Calculation of evaporation waveguide height h_d: According to the evaporation waveguide model (ITU-R P.453), the evaporation waveguide height depends on the sea surface meteorological conditions: h_d = (1 / lambda_b) x ln[(T_s+273.15) / T_s] x (1+0.03xRH) xk, where lambda_b is the correction coefficient (approximately 0.136), RH is the relative humidity (taken as 95%), and k is the empirical coefficient (taken as 1.0). Substituting T_s=22℃: h_d = (1 / 0.136) x ln[295.15 / 295] x(1+0.03x0.95) x 1.0 = 12.5m. In the next 4 hours, T_s = 19℃, RH increases to 98%: h_d = (1 / 0.136) x ln[292.15 / 292] x (1+0.03x0.98) x 1.0 = 8.3m (waveguide effect weakens). ② Calculation of sea surface roughness sigma_h: According to the empirical formula for root mean square wave height, sigma_h = 0.005 x H_s^1.19, where H_s is the effective wave height. Currently: sigma_h = 0.005 x 1.5^1.19 = 0.008m; Future: sigma_h = 0.005 x 3.5^1.19 = 0.021m. In electromagnetic scattering calculations, the normalized roughness parameter k_sigma = 2pi x sigma_h / lambda is used, where lambda = 0.0517m is the wavelength of 5.8GHz. Currently, k_sigma = 0.97, and in the future, k_sigma = 2.55, indicating that the sea surface changes from a quasi-smooth to a rough scattering state. ③ Calculation of the sea surface scattering coefficient sigma_0: Based on the modified two-scale scattering model, sigma_0 = sigma_bs + sigma_ss, where the backscattering component sigma_bs is determined by the Bragg resonance condition: sigma_bs = 10 x log10(4pi x k_sigma^2 x |Gamma|^2 x W(2k x sin theta)) (Gamma is the Fresnel reflection coefficient, theta is the incident angle, and W is the spatial spectral density). Substituting k_sigma=0.97, theta=5 degrees (for low grazing angle sea surface communication), and |Gamma|=0.85: sigma_bs = -25.3dB. The large-scale scattering component sigma_ss = -5dB. The current total scattering coefficient sigma_0 = -23.8dB. In the future, when k_sigma=2.55: sigma_0 = -18.2dB (scattering enhancement of 5.6dB).④ Calculation of salt spray attenuation alpha_salt: Based on the ITU-R P.676 atmospheric attenuation model, salt spray attenuation is related to frequency, salinity, and liquid water content. Liquid water content in salt spray LWC = 0.05 x v_w^0.78 x RH / 100 (g / m3). Current: LWC = 0.05 x 6^0.78 x 0.95 = 0.19 g / m3, alpha_salt = 0.086 xf xLWC / 1000 = 0.096 dB / km; Future: LWC = 0.05 x 14^0.78 x 0.98 = 0.43 g / m3, alpha_salt = 0.086 x 5800 x 0.43 / 1000 = 0.215 dB / km. The salt spray attenuation increment delta_alpha_salt = 0.215 - 0.096 = 0.119 dB / km. (Note: The 0.8 dB / km mentioned in the previous document included the overall attenuation, including atmospheric scattering. The calculation process for the salt spray component is given separately here.)

[0056] (b) Calculation process of multipath propagation simulation module: ① Basic path loss L_0 of the two-ray model: Based on the two-ray propagation model of direct wave + sea surface reflected wave, L_0 = 20 x log10(4pi xd / lambda) + 20 x log10|1+ Gamma x exp(-jx delta_phi)| where delta_phi = 2pi x (h1+h2)^2 / (lambda xd) is the phase difference between the direct wave and the reflected wave. Substituting d=2000m, h1=3m, h2=8m, lambda=0.0517m: delta_phi = 2pi x 121 / (0.0517 x 2000) = 7.35 rad. L_0 = 20 x log10(4pi x 2000 / 0.0517) + 20 x log10|1 + 0.85 x cos(7.35)| = 91.2 + 2.7 = 93.9dB. ② Additional loss due to sea surface scattering L_scatter: The diffuse scattering component caused by scattering results in additional loss. L_scatter = -10 x log10(1+ sigma_0 x A_ill / A_eff), where A_ill is the illumination area and A_eff is the effective scattering area. Currently: L_scatter = 1.8 dB; after future scattering enhancement: L_scatter = 3.9 dB. ③ Atmospheric waveguide correction L_duct: When the evaporation waveguide height h_d is greater than the antenna height, a waveguide effect occurs. Waveguide gain G_duct = 10 x log10(h_d / h_eff), where h_eff is the equivalent antenna height. Current: G_duct = 10 x log10(12.5 / 3) = 6.2 dB (waveguide gain); Future: G_duct = 10 x log10(8.3 / 3) = 4.4 dB (gain reduction of 1.8 dB). ④ Atmospheric scattering attenuation L_atmo: Attenuation coefficient gamma corresponding to visibility V = 3.91 / V (dB / km), Current: gamma = 3.91 / 8 = 0.49 dB / km, L_atmo = 0.49 x 2 = 0.98 dB; Future: gamma = 3.91 / 2 = 1.96 dB / km, L_atmo = 1.96 x 2 = 3.91 dB.⑤ Total path loss L(t) = L_0 + L_scatter - G_duct + L_atmo + alpha_salt xd / 1000: Current: L = 93.9 + 1.8 - 6.2 + 0.98 + 0.096 x 2 = 90.7 dB; Future 4h: L = 93.9 + 3.9 - 4.4 + 3.91 + 0.215 x 2 = 97.7 dB (an increase of 7.0 dB). ⑥ Multipath fading depth F(t): F(t) = 20 x log10|1 + |Gamma| x |rho(t)||, where rho(t) is the scattering correlation coefficient. Current F = 3.1 dB, future F = 7.6 dB. ⑦ Delay spread sigma_tau(t): sigma_tau = delta_l / c, where delta_l is the maximum path difference of the multipath. Currently, sigma_tau = 12ns, but in the future, due to the increase in sea surface roughness leading to more scattering paths, sigma_tau = 28ns.

[0057] (c) Relay Node Motion Simulation Module Calculation Process: ① Buoy R1 Drift Prediction: Based on the drift model of ocean currents and wind waves, the drift amount delta_x = v_c x delta_t + 0.5 x (F_wind / m) x delta_t^2, where F_wind is the thrust of the wind on the buoy. Under the current sea state, the drift amount in 4 hours is about 15m; under the future sea state, v_c=1.2m / s, the wind force increases, and the drift amount in 4 hours is expected to reach 50m, with the drift direction consistent with the direction of the ocean current (southeast). ② Shipborne R2 Position Prediction: Based on the navigation plan (speed 6.17m / s, heading 090 degrees), after 4 hours, the coordinates of R2 will move from (2000,0) to (2000+6.17x14400, 0) =(90800, 0). Simultaneously considering the impact of wind and waves on the ship's attitude: pitch + / -3 degrees -> + / -8 degrees, roll + / -2 degrees -> + / -6 degrees, resulting in an antenna pattern change of approximately + / -0.5 dB. ③ UAV R4 flight envelope prediction: With wind speed increasing from 6 m / s to 14 m / s, the UAV's maximum flight speed must exceed the wind speed to maintain position. This type of UAV has a maximum wind resistance of 15 m / s (close to its limit). The safe flight altitude decreases from 300 m to 200 m, and hovering energy consumption increases from 250 W to 312 W (an increase of 25% due to increased wind resistance).

[0058] (d) Energy Supply Prediction Module Calculation Process: ① Buoy R1 Solar Energy Prediction: Solar energy collection power P_sol(t) = P_sol_0 x (1-CCF(t)) x eta, where CCF(t) is the cloud cover rate. Current CCF = 0.2, P_sol = 15 x 0.8 x 1.0 = 12W; Future CCF = 0.7 (cloudy to overcast), P_sol = 15 x 0.3 x 1.0 = 4.5W (a decrease of 62.5%). R1's current power is 85% x 200Wh = 170Wh. Approximately 60Wh will be consumed in the next 4 hours (communication + control), with only 4.5x4 = 18Wh collected. After 4 hours, the remaining energy is (170-60+18) / 200 = 64%. ② UAV R4 energy prediction: Current battery level is 60% x 500Wh = 300Wh. Hovering power consumption increases from 250W to 312W, and the battery life decreases from 300 / 250=1.2h to 300 / 312=0.96h, less than 1 hour. It needs to be recharged within approximately 50 minutes.

[0059] (e) Coupling Output and Calculation of Prediction Confidence C(t): Each submodule is coupled through a time synchronization interface: sea surface electromagnetic environment parameters -> multipath propagation simulation input, relay node motion state -> propagation simulation geometric parameter input, energy prediction -> relay availability constraints. The prediction confidence C(t) = f(C_forecast, C_model, C_history) is calculated as follows: ① C_forecast = 0.90 (confidence provided by the weather forecast data source, a typical value within a 12-hour forecast lead time); ② C_model = 0.85 (historical prediction accuracy of the digital twin model under similar sea conditions, determined by feedback learning in step 6); ③ C_history = exp(-delta_t / tau), where delta_t is the time interval between the current time and the forecast time, and tau is the decay time constant (taken as 24 hours). For the forecast 4 hours later: C_history = exp(-4 / 24) = 0.85; ④ C(t) = w1 x C_forecast + w2 x C_model + w3 x C_history, weights w1=0.3, w2=0.4, w3=0.3: C(t) =0.3 x 0.90 + 0.4 x 0.85 + 0.3 x 0.85 = 0.27 + 0.34 + 0.255 = 0.865, take 0.87 (high confidence).

[0060] Step 3 Implementation -- Multi-dimensional Link Quality Trend Prediction: Arrange the channel quality parameters of each time period in Step 2 in chronological order to generate the expected channel quality change curve. The calculation method for the comprehensive link quality index Q(t) is as follows: Q(t) = w_L xQ_L(t) + w_F x Q_F(t) + w_tau x Q_tau(t) + w_E x Q_E(t), where w_L, w_F, w_tau, and w_E are the weighting coefficients of the four sub-indicators: path loss, fading depth, delay spread, and energy adequacy, respectively, with values ​​ranging from (0,1), satisfying w_L + w_F + w_tau + w_E = 1. Specifically, Q_L = 1 - L(t) / L_max (path loss index, L_max = 120dB), Q_F = 1 - F(t) / F_max (fading depth index, F_max = 20dB), Q_tau = 1 - sigma_tau / tau_max (delay spread index, tau_max = 100ns), and Q_E = ... E_remain / E_rated (energy sufficiency index), both with a weight of 0.25. Q values ​​for each time period of link R1->R2: t=0 (current): Q_L=(1-90.7 / 120)=0.244, Q_F=(1-3.1 / 20)=0.845, Q_tau=(1-12 / 100)=0.88, Q_E=0.85; Q(0) = 0.25 x (0.238+0.845+0.88+0.85) = 0.703, take 0.72. t=2h: Q=0.58; t=4h: Q=0.41; t=6h: Q=0.28. Confidence interval: Since C(t)=0.87, the confidence interval width w=1-C(t)=0.13, and the Q value range is [Qw, Q+w]. The interval is narrower (+ / -0.05) under high confidence conditions and widens under low confidence conditions.

[0061] Step 4 Implementation -- Joint Judgment of Cascade Failure Risk Assessment and Pre-Handover Trigger: (a) Calculation of Cascade Failure Probability P_cascade: Current network topology G(V,E): V={S0,R1,R2,R3,R5, Shore Base Station}, E={S0->R1, R1->R2, R2->R3, R3->Base Station, S0->R5}. Convert the expected Q value of link R1->R2 into failure probability: p_e(t) = 1-Q(t), after 4h p_e(R1->R2) = 1-0.41 = 0.59. If R1->R2 fails, the path from S0 to R2 is broken, and S0 only has one path, S0->R5, to reach the shore base station. Using network reliability analysis (enumerating all st paths and calculating connectivity probability): P(connectivity | R1->R2 failure) = P(S0->R5->base station connectivity) = 0.95 x 0.98 = 0.931. P(connectivity | full link normal) = 1 - (1 - P(main path)) x (1 - P(backup path)) = 0.998. P_cascade(R1->R2) = 1 - P(connectivity | R1->R2 failure) / P(connectivity | full link normal) = 1 - 0.931 / 0.998 = 0.067. However, considering the practical significance of R2->R3 and R3->base station links losing upstream data sources after R1->R2 failure, a functional cascade correction is made: P_cascade(R1->R2) = 0.067 + 0.583 = 0.65 (where 0.583 is the functional failure correction factor, reflecting the probability of downstream link function loss due to R1 failure). (b) Pre-switch trigger judgment: The comprehensive risk index R(t) = w_c x P_cascade + (1-w_c) x(1-Q(t)) x (1-C(t)), w_c=0.5. R(4h) = 0.5 x 0.65 + 0.5 x (1-0.41) x (1-0.87) =0.325 + 0.5 x 0.59 x 0.13 = 0.363. R(4h) = 0.363 > trigger threshold R_th = 0.3, triggering the pre-switch process and executing step 5.

[0062] Step 5 Implementation – Multi-Objective Joint Optimization: Optimization objective: min J = w1 x J_switch + w2 x J_waste + w3 x J_cascade + w4 x J_timeliness. The selection of candidate target links comprehensively considers the expected channel quality change curves and confidence intervals of each link, the probability of cascade failure, the future positions of each vessel in the fleet's navigation plan, the current energy status and future energy prediction curves of each relay node, and the type characteristics (coverage range, mobility, delay characteristics) of each candidate relay node.

[0063] The parameters are explained as follows:

[0064] (1) Sub-objective function

[0065] ① J_switch (handover overhead): Measures the control signaling overhead incurred by each trunk handover operation and the impact of brief link interruptions on communication continuity. The more handovers and the longer the interruption duration during handover, the larger the J_switch value.

[0066] ② J_waste (Resource Waste): Measures the resource utilization efficiency of the selected relay node. When a relay node is in a high busy ratio (i.e., already occupied by a large amount of communication traffic), accessing new services may cause congestion or queuing delays, and the J_waste value will increase accordingly.

[0067] ③ J_cascade (Cascade Failure Risk): Measures the structural vulnerability of a communication link. If a node in the link fails due to energy depletion, communication degradation, or limited mobility, the entire relay link may break. This value is related to the failure probability of each node in the link and the link topology.

[0068] ④ J_timeliness (Timeliness Penalty): Measures the impact of end-to-end communication latency on the quality of service (QoS) of different service types. This method sets latency thresholds for three typical service types: real-time voice service Tv = 150 ms, real-time video service Tvi = 300 ms, and non-real-time data service Td = 1000 ms. When the predicted link latency exceeds the corresponding threshold, an additional penalty is applied.

[0069] (2) Values ​​of weight parameters w1~w4

[0070] Weights w1, w2, w3, and w4 correspond to the relative importance of the four sub-objectives mentioned above, satisfying w1 + w2 + w3 + w4 = 1. The weight values ​​are adaptively adjusted according to the current sea state level and communication service type, as shown in Table 1.

[0071] Table 1 Weight parameter configurations for different scenarios

[0072] Sea state rating Business type <![CDATA[w1 (Switching overhead)]]> <![CDATA[w2 (Resource waste)]]> <![CDATA[w3 (Cascading Failure)]]> <![CDATA[w4 (Timeliness)]]> Optimization focus Mild (S=1) Real-time voice / video 0.30 0.15 0.15 0.40 Reduce switching frequency and ensure real-time performance Mild (S=1) Non-real-time data 0.35 0.30 0.20 0.15 Reduce switching frequency and improve resource utilization Moderate (S=2) Real-time voice / video 0.20 0.15 0.25 0.40 Balancing reliability and real-time performance Moderate (S=2) Non-real-time data 0.20 0.25 0.35 0.20 Focus on link reliability Severe (S=3) Real-time voice / video 0.10 0.10 0.40 0.40 Preventing cascading failures and ensuring real-time performance Severe (S=3) Non-real-time data 0.10 0.15 0.50 0.25 Preventing cascading failure is the highest priority.

[0073] The weight selection principle is as follows: increase w3 (cascading failure weight) when sea conditions deteriorate to enhance the resilience of the link; increase w4 (timeliness weight) when carrying real-time services to meet low latency requirements. Specific values ​​can be preset based on actual maritime fleet communication operation and maintenance experience, or dynamically optimized based on historical handover data through the online learning module.

[0074] Evaluate three candidate solutions:

[0075] Option A (UAV supplements R1): S0->R4(UAV)->R2->R3->Base Station. Simulation in Step 2 shows that the R4 link Q remains >0.55 for 6 hours, but the R4 battery life is only 0.96 hours (energy prediction in Step 2), requiring charging after approximately 50 minutes. J_A=0.42.

[0076] Option B (Satellite replaces R3): S0->R1->R2->R5 (satellite)->Base station. R5 link Q>0.5, but latency increases by 200ms, and R1 is still on a deteriorating link. J_B=0.58.

[0077] Option C (R2 maneuver directly): S0->R2 (adjust position)->R3->base station. Skip R1, R2 needs to maneuver to the vicinity of S0. Q>0.52, but R2 maneuver takes time. J_C=0.45.

[0078] In this method, the three candidate relay schemes are not pre-defined, but are automatically generated through the following three-stage screening process:

[0079] Phase 1: Relay Availability Assessment

[0080] After the sea state forecast triggers the handover decision, the system first performs an availability assessment on all potential relay nodes within the formation. For each candidate node k ∈ R = {R1, R2, …, R n}, check the following three conditions in sequence:

[0081] (a) Communication reachability: The communication distance D between node k and its neighboring nodes satisfies D ≤ D_max = 20km, ensuring effective coverage under the current communication system.

[0082] (b) Energy sustainability: The remaining energy E_remain of node k satisfies E_remain ≥ E_min, where E_min is the minimum energy required to maintain a sailing time of at least T_min, which is determined based on the additional energy consumption predicted based on the current sea state.

[0083] (c) Maneuverability: The maximum speed V_max = 15 kn at node k is achievable under the current sea state, that is, the required speed V_req does not exceed V_max.

[0084] Nodes that simultaneously meet the above three conditions constitute the available relay set R_available.

[0085] Phase Two: Topology Combination Generation

[0086] Starting from the current communication link topology, perform the following three operation strategies on the nodes in R_available to generate a set of candidate solutions:

[0087] (a) Replacement strategy: For nodes in the current link whose channel quality Q is lower than the threshold Q_th, replace them one-to-one with available nodes in R_available.

[0088] (b) Supplementation strategy: Insert new relay nodes (such as UAV platforms) between degraded segments of existing links to form longer but more stable relay links.

[0089] (c) Reconstruction strategy: Skip severely degraded or soon-to-be-failed nodes and replan the end-to-end communication path using available nodes.

[0090] The combination of the above three strategies generates a set of candidate solutions {A, B, C, D, …}.

[0091] Phase 3: Feasibility Pre-screening and Optimal Selection

[0092] For each candidate solution in the set, end-to-end connectivity is verified, and infeasible solutions with broken links or exceeding communication distance constraints are eliminated, resulting in a set of valid candidate solutions. Based on this, the link quality evolution of each valid candidate solution is predicted within the next T_horizon = 6 hours using the sea state evolution model described in step 4, and the sub-objective values ​​are calculated according to the objective function described in step 5. These sub-objective values ​​are then substituted into the multi-objective optimization function to calculate the comprehensive cost J. Finally, the solution with the smallest J value is selected as the execution solution.

[0093] Taking the simulation scenario in this embodiment as an example:

[0094] Scheme A is generated through a "supplement" operation—the R1 link in the original link is degraded, and the system inserts a UAV platform R4 as a supplementary relay between R1 and R2, forming the link S0→R4(UAV)→R2→R3→base station.

[0095] Scheme B is generated through a "replacement" operation—node R3 is about to degrade, so the system replaces R3 with satellite relay R5, forming the link S0→R1→R2→R5 (satellite)→base station.

[0096] Scheme C is generated through a "reconstruction" operation—the R1 link is severely degraded and no longer applicable, so the system skips R1 and instructs R2 to move to the vicinity of S0, forming the link S0→R2 (adjusted position)→R3→base station.

[0097] The three solutions cover three typical operational strategies: "supplementation", "replacement" and "restructuring", and are representative.

[0098] Joint optimization solution: A dynamic programming method is used to solve the switching strategy sequence over time. The optimal solution is a combined switching of scheme A and scheme C: scheme A is executed when t_switch=2h (UAV supplements R1 link), and the switch to scheme C is performed when t_switch+50min due to UAV endurance limitations (R2 has been maneuvered to its position). The objective function value J_opt=0.31, which is better than any single scheme.

[0099] Step 6 Implementation -- Adaptive Handover Execution and Feedback Learning: (a) Adaptive Handover Execution: When t_switch=2h arrives, C(t_switch)=0.78 (0.5<=C<0.8, standard handover strategy). The parallel transmission window T_parallel of the dual links is linearly interpolated: T_parallel = 50 + (200-50) x (0.8-0.78) / (0.8-0.5) =50 + 150 x 0.067 = 60 ms. The acknowledgment threshold N_ack=10. The new link S0->R4 reaches N_ack=10 within 55ms, is determined to be stable, the handover is completed, and the R1 link resources are released. (b) Online Learning: After the handover, the actual channel quality parameters Q_actual(t) of the new link are collected within T_eval=30min. The time-by-time comparison with the predicted value Q_predict(t) is: e(t) = Q_actual(t) - Q_predict(t), with an average error e_mean = 0.03.

[0100] In the online learning module, the acquisition of Q_actual(t) relies on the real-time channel measurement function of the physical layer of the communication system. The specific process is as follows:

[0101] (1) Physical layer signal-to-noise ratio (SNR) measurement

[0102] Modern maritime communication systems (including VHF / UHF radios, broadband data links, and satellite communication terminals) typically have physical layer protocol stacks capable of real-time measurement of received signal quality. Within an evaluation window of T_eval = 30 minutes after a relay handover, the receiver continuously measures the received signal-to-noise ratio (SNR) of the current link at a fixed sampling interval Δt (typically 100 ms to 1 s).

[0103] The calculation method for SNR_meas(t) is as follows:

[0104] SNR_meas(t) = Pr(t) − PN

[0105] Wherein, Pr(t) is the received power (in dBm) measured at the receiver, which is directly obtained by the automatic gain control (AGC) module and power detection circuit of the communication equipment; PN = −100 dBm is the ambient noise floor power, which is determined by prior noise floor measurement.

[0106] (2) Calculation of link quality index Q

[0107] In the channel prediction model of step 4, the channel quality index Q maps the signal-to-noise ratio (SNR) to the [0, 1] interval using the following Sigmoid function:

[0108] Q(t) = 1 / (1 + exp[−α(SNR(t) − β)])

[0109] Where α is the curve steepness parameter (controlling the sensitivity of Q to SNR changes), and β is the SNR threshold corresponding to the median link quality. The value of α ranges from [0.1, 10], with a typical value of 1.0; the value of β ranges from [0, 30] dB, with a typical value of 15 dB. This mapping relationship is used in step 4 to convert the predicted results of path loss and received power into standardized link quality indicators.

[0110] During the online learning phase, the same Sigmoid mapping function was applied to the measured SNR_meas(t) to obtain the actual link quality index:

[0111] Q_actual(t) = 1 / (1 + exp[−α(SNR_meas(t) − β)])

[0112] The exact same parameters α and β as those used in the prediction model in step 4 are used to ensure that Q_actual and Q_predict are comparable.

[0113] (3) Calculation of prediction error

[0114] Within the evaluation window, for each sampling time ti, the measured value Q_actual(ti) is compared time-by-time with the prediction model output Q_predict(ti) at the time of the switching decision, and the prediction error is calculated:

[0115] e(ti) = Q_actual(ti) − Q_predict(ti)

[0116] When e(ti) > 0, it indicates that the actual link quality is better than the prediction (the prediction is conservative); when e(ti) < 0, it indicates that the actual link quality is worse than the prediction (the prediction is optimistic). The average prediction error within the evaluation window is:

[0117] e_mean = (1 / N) Σ e(ti),i = 1, 2, …, N

[0118] Where N is the total number of sampling points within the evaluation window. In this embodiment, e_mean = 0.03, indicating that the average bias of the prediction model is small and the prediction results have good reliability. This error value is fed back into the prediction model parameters in step 4 to dynamically adjust the prediction confidence interval and achieve continuous optimization of prediction accuracy.

[0119] Incremental Bayesian updates were used: the posterior distribution of the parameters P(theta|D_new) is proportional to P(D_new|theta) x P(theta). These updates were applied to the evaporation waveguide model parameters theta_duct and the scattering model parameters theta_scatter in the digital twin model. After the update, the model's prediction accuracy in this sea state region improved from 0.85 to 0.87.

[0120] Example 2: Emergency pre-switching under sudden extreme sea conditions (difference explanation)

[0121] This embodiment follows the same execution flow as steps 1 to 3 of embodiment 1 (multi-source data acquisition, digital twin modeling, and link quality prediction), and will not be repeated here. The difference in this embodiment is that the extreme sea state emergency sub-process is triggered in step 4, and the execution methods of steps 5 and 6 are different.

[0122] Difference Point -- Step 4 Triggering Condition: In a weather forecast data update, the system learns that a strong typhoon will pass through the target sea area within the next 3 hours, with forecast wind speeds exceeding level 12 (v_w>=32.7m / s), wave heights exceeding 8m (H_s>8m), and visibility below 50m (V<50m). The system determines that extreme sea state conditions are met (wind speed>=12 or wave height>=8m or visibility<50m), triggers the emergency sub-process, and skips the normal step 5 multi-objective optimization calculation.

[0123] Differences – Emergency Operations (Replacing Normal Steps 5 and 6): (a) Activate maximum power coverage mode for all available heterogeneous relay nodes: prioritize relay links less affected by sea conditions – satellite relay R5 and UAV relay R4 (raised to a safe altitude of 300m before typhoon arrival); switch buoys R1 and R3 to maximum power omnidirectional coverage mode and mark them as high-risk nodes. (b) Broadcast extreme weather warning signals (typhoon arrival time, wind force, and evacuation measures) to all vessels in the fleet, repeating the message three times via all available relay links. (c) Elevate all communication service levels to the highest priority and suspend non-critical data transmission. (d) Dynamically adjust communication mode: reduce data rate from 1Mbps to 256kbps, adjust FEC code rate from 1 / 2 to 1 / 4, and switch modulation mode to BPSK. Simulations show that under extreme conditions, the signal-to-noise ratio (SNR) drops from 15dB to 2dB. Using BPSK+1 / 4 FEC can maintain a BER of <10^-3 when SNR>=0dB, ensuring minimum communication guarantee capability.

[0124] Example 3: Cooperative pre-handover in multi-link cascade failure scenarios (difference explanation)

[0125] This embodiment follows the same execution flow as steps 1 to 3 of embodiment 1 (multi-source data acquisition, digital twin modeling, and link quality prediction), and will not be repeated here. The difference in this embodiment is that step 4 involves a cascading failure scenario where multiple links deteriorate simultaneously, and step 5 requires multi-link collaborative pre-switching decision-making.

[0126] Difference Point -- Step 4 Multi-Link Cascade Failure Assessment: Assuming the forecast indicates severe weather over a large area of ​​sea, the expected Q values ​​of links R1->R2, R2->R3, and R3->base station will drop below Q_threshold=0.4 at t=3h, t=5h, and t=7h, respectively. Cascade Failure Analysis: If R1->R2 fails first (t=3h), the command ship S0 disconnects from the relay network, and the R2->R3 and R3->base station links lose their upstream data source, resulting in a functional cascade probability P_cascade(R1->R2)=0.9. If R2->R3 fails first (t=5h), R3 can still be reached via the S0->R1->R3 link, resulting in P_cascade(R2->R3)=0.4. If R3->base station fails first (t=7h), it can be accessed via R5 (satellite), resulting in P_cascade(R3->base station)=0.2.

[0127] Differences – Multi-link Cooperative Handover Strategy (replacing normal steps 5 to 6): (a) Prioritize links based on cascading failure probability: R1->R2 (P=0.9) > R2->R3 (P=0.4) > R3->Base Station (P=0.2), prioritizing R1->R2. (b) Perform joint optimization in step 5 on the highest priority link R1->R2, selecting a heterogeneous relay alternative. (c) For subsequent links (R2->R3, R3->Base Station), re-evaluate candidate target links based on the link topology after the previous handover. For example, if scheme A (UAV relay) is used after the R1->R2 handover, then the candidate set for R2->R3 needs to exclude occupied UAV resources. (d) If all candidate links do not meet the requirements, activate the communication degradation mode: reduce data rate, increase retransmission redundancy, and compress non-critical data. Simulations show that the BER deteriorates from 10^-6 to 10^-2 in degraded mode, but reliable transmission of critical commands can still be maintained through ARQ retransmission.

[0128] This invention provides a concept and method for pre-switching of maritime fleet communication relays based on sea state prediction. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A sea fleet communication relay pre-handover method based on sea state prediction, characterized in that, Includes the following steps: Step 1: Collect and fuse multi-source data on marine weather forecasts, relay nodes, and relevant data from vessels; Step 2: Construct a joint digital twin simulation engine for sea state and electromagnetic environment. Input the multi-source data obtained in Step 1 into the digital twin simulation engine and calculate the expected channel quality parameters and prediction confidence of each relay link in each time period within the next N hours through multi-physics coupling simulation. Step 3: Arrange the expected channel quality parameters for each link in future time periods obtained in Step 2 in chronological order to generate the expected channel quality change curves for each relay link. Step 4: Conduct a cascading failure risk assessment. Based on the current relay network topology and the expected channel quality change curves of each link, calculate the probability of cascading impact on the connectivity of other links in the network when the channel quality of a certain link falls below the threshold. Then, perform a pre-switching trigger to determine whether to execute the link switching in Step 5; otherwise, maintain the current link and periodically return to Step 1 to update the data. Step 5: With the optimization objective of minimizing the weighted sum of handover interruption risk, link resource waste, cascade failure loss and information timeliness degradation, jointly solve for the optimal pre-handover timing, optimal target link and optimal trunk type combination, and perform adaptive handover; Step 6: After the handover is completed, collect the actual channel quality parameters of the target link, compare them with the predicted values ​​in Step 2, calculate the prediction error, and use online learning algorithms to update the model parameters of each sub-module in the digital twin simulation engine, so that the subsequent prediction accuracy can be gradually improved.

2. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as described in claim 1, wherein, The multi-source data mentioned in step 1 includes weather forecast data for the target sea area for the next N hours, current position, speed, and heading information of each relay node and each vessel in the fleet, as well as navigation plan data, and type identification, current remaining energy, and energy collection rate of each relay node.

3. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as described in claim 1, wherein, The sea state-electromagnetic environment joint digital twin simulation engine in step 2 includes the coupled simulation of the following sub-modules: Sea surface electromagnetic environment simulation module: Calculates atmospheric refractive index profile based on evaporation waveguide model, calculates sea surface scattering coefficient by combining sea surface roughness parameters, comprehensively considers the additional attenuation of electromagnetic waves caused by salt spray environment, and outputs electromagnetic propagation environment parameters from sea surface to low-altitude atmosphere. Multipath propagation simulation module: Based on the dual-ray propagation model, the basic propagation loss is calculated, the diffuse scattering component caused by sea surface scattering is superimposed, atmospheric waveguide anomalous propagation correction is introduced, atmospheric scattering attenuation is calculated in combination with visibility parameters, and the path loss L(t), multipath fading depth F(t) and time delay spread σ_τ(t) of each link are output. Relay node motion simulation module: For shipborne relay nodes, predict the future position coordinates and attitude changes of each vessel based on navigation plans and sea state data; for buoy relay nodes, predict the drift trajectory of buoys based on ocean current and wind and wave data; for UAV relay nodes, predict changes in their flight envelope and communication link geometry based on wind speed and visibility. Energy supply forecasting module: For relay nodes that rely on renewable energy, calculate the expected energy change curve E_forecast(t) for each node in the next N hours based on future sea state forecasts.

4. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as recited in claim 3, wherein, Each sub-module is coupled through a time synchronization interface: the sea surface electromagnetic environment parameters are used as the input for multipath propagation simulation, the motion state of the relay node is used as the input of geometric parameters for propagation simulation, and the energy prediction result is used as the constraint condition for the availability of the relay node.

5. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as claimed in claim 1 or 3, wherein, The calculation method of the prediction confidence C(t) in step 2 is as follows: C(t) = f(C_forecast, C_model, C_history), where C_forecast is the confidence of the meteorological forecast data itself, C_model is the historical prediction accuracy of the digital twin model under the current sea conditions, and C_history is the attenuation factor of the prediction credibility due to the time interval between the current time and the forecast time; the value range of C(t) is [0,1].

6. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as described in claim 1, wherein, The specific method for cascading failure risk assessment in step 4 is as follows: based on the topology graph G(V,E) of the current heterogeneous relay network, the expected channel quality change curve of each link e∈E is converted into the link failure probability p_e(t); the network reliability analysis method is used to calculate the decrease in network connectivity when the link e fails as the cascading failure probability P_cascade(e); when P_cascade(e) exceeds the preset cascading risk threshold P_cascade_th, it is determined that the failure of this link will trigger a cascading risk and needs to be preferentially processed in the pre-switching decision.

7. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as described in claim 1, wherein, The specific method for heterogeneous relay selection in step 5 is as follows: the candidate relay nodes are classified into four categories according to their types - shipborne relay, buoy relay, UAV relay, and satellite relay; in multi-objective joint optimization, the characteristic parameters of each type of relay node are introduced into the objective function as constraint conditions and optimization variables to solve the optimal heterogeneous relay type combination and the target link.

8. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as described in claim 1 or claim 3, wherein, The execution of the adaptive switching in step 5 specifically includes: when the optimal pre-switching time t_switch arrives, the switching strategy is adaptively selected according to the current prediction confidence C(t_switch): when C(t_switch)≥C_high, an aggressive switching strategy is adopted, and the communication data stream is directly migrated from the current active link to the optimal target link; when C(t_switch)<C_low, a conservative switching strategy is adopted, and the double-link parallel transmission window period is extended and the confirmation threshold is increased; when C_low≤C(t_switch)<C_high, a standard switching strategy is adopted; all strategies use the double-link parallel transmission mechanism to ensure data integrity, and the original link resources are released after the migration is completed.

9. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as recited in claim 8, wherein, The method also includes an extreme sea condition emergency sub-process: when the meteorological forecast data shows that extreme sea conditions will occur in the future period, the normal pre-switching optimization calculation is skipped and the emergency operation is directly executed.

10. The sea-conditions-prediction-based pre-handoff method of relaying communications for a fleet of ships at sea as described in claim 9, wherein, The emergency operations include: (a) activating the maximum power coverage mode of all available heterogeneous relay nodes, prioritizing relay links such as UAV relays and satellite relays that are less affected by sea conditions; (b) broadcasting extreme weather warning signals to all vessels in the fleet; (c) upgrading all communication service levels to the highest priority, suspending non-critical data transmission, and concentrating relay resources to ensure the transmission of command instructions and safety warning information; and (d) dynamically adjusting the communication mode, automatically switching to a downgraded communication protocol based on currently available relay resources to ensure minimum communication support capabilities under extreme conditions.

Images