A data transmission method and system based on marine multi-element environment monitoring networking

Abstract

The embodiment of the present application provides a data transmission method and system based on marine multi-element environment monitoring networking, the method obtains first data through an underwater sensing layer, selects a first optimal transmission link according to an underwater sound-light dual-communication adaptive switching algorithm, and transmits the first data to a sea surface relay layer through the first optimal transmission link; after the sea surface relay layer processes the first data into second data, adopts a dual-link adaptive cross-domain communication algorithm to transmit the second data to a shore-based management and control layer, so that the shore-based management and control layer adopts a multi-source heterogeneous data fusion algorithm to process the second data into third data. Through the implementation of the present application, efficient and reliable transmission of marine monitoring data from the seabed to the shore base is realized, and the transmission efficiency and reliability of the marine monitoring data are improved.

Description

Technical Field

[0001] This invention relates to the field of marine data transmission technology, and in particular to a data transmission method and system based on a marine multi-element environmental monitoring network. Background Technology

[0002] In existing technologies, underwater monitoring systems mostly rely on underwater acoustic or optical communication alone. These systems cannot adaptively switch transmission links based on real-time channel conditions and sea state changes, making it difficult to balance transmission distance and rate performance. Data transmission in complex aquatic environments is prone to jitter, disconnections, and poor reliability. Surface buoy relay platforms often use a single link for cross-domain data transmission, lacking a redundant link adaptive scheduling mechanism. When the link fails due to marine environmental interference, data interruption and loss are highly likely. Traditional marine data fusion technologies do not consider the spatiotemporal timeliness of data and differences in equipment reliability. They lack a mechanism for coupling and fusing spatiotemporal attenuation weights and reliability weights, resulting in poor adaptability to heterogeneous marine data from multiple sources, time series, and precision levels. Data gaps and poor continuity in gridded data are common in sparsely populated areas. Furthermore, the lack of network-wide scheduling and insufficient overall network automation fail to meet the demands for comprehensive, efficient, and highly reliable marine environmental monitoring. Summary of the Invention

[0003] This invention provides a data transmission method and system based on a multi-element marine environmental monitoring network, which solves the problems of low data transmission efficiency and insufficient security in marine monitoring data transmission, and realizes high-efficiency and high-reliability transmission of marine monitoring data from the seabed to the shore, thereby improving the transmission efficiency and reliability of marine monitoring data.

[0004] An embodiment of the present invention provides a data transmission method based on a marine multi-element environmental monitoring network, comprising the following steps: The first data is acquired through an underwater sensing layer. A first optimal transmission link is selected based on an underwater acoustic-optical dual-communication adaptive switching algorithm, and the first data is transmitted to the surface relay layer via this first optimal transmission link. The first data consists of raw observation data of multi-source marine environment, channel, and equipment operating conditions. The underwater sensing layer is used for multi-element marine environment full-domain acquisition, local edge preprocessing and caching, underwater short-range node adaptive Mesh self-organizing network data transmission, and on-site anomaly autonomous alarm. The underwater sensing layer includes a multi-element integrated sensor group, seabed fixed mooring nodes, and underwater mobile monitoring nodes. Both the seabed fixed mooring nodes and the underwater mobile monitoring nodes are equipped with underwater acoustic-optical dual-communication adaptive switching communication modules. The underwater acoustic-optical dual-communication adaptive switching algorithm is constructed by establishing an underwater channel switching model based on a Markov decision process and iteratively solving the underwater channel switching model using a lightweight Q-Learning sub-algorithm. After processing the first data into second data, the sea surface relay layer uses a dual-link adaptive cross-domain communication algorithm to transmit the second data to the shore-based control layer, so that the shore-based control layer can use a multi-source heterogeneous data fusion algorithm to process the second data into third data. The dual-link adaptive cross-domain communication algorithm is used for dual-link dynamic switching to adapt to sea area transmission and to stably transmit marine monitoring data across domains. The multi-source heterogeneous data fusion algorithm is used to couple spatiotemporal and reliability weights and to fuse the second data in a grid-like manner. The sea surface relay layer is used for cross-media signal docking, multi-source heterogeneous data preprocessing and aggregation, dual-link adaptive cross-domain intelligent scheduling, and bidirectional relay and local fault-tolerant caching. The shore-based control layer is used for multi-source heterogeneous data fusion processing, intelligent scheduling and communication strategy control of the entire network, full-domain visualization control, global network self-healing decision-making, business applications, and alarms.

[0005] Furthermore, the establishment of the underwater channel handover model based on the Markov decision process includes: The optimization objectives of the underwater channel handover model are to maximize transmission throughput, minimize bit error rate and packet loss, reduce link handover power consumption, and reduce local data backlog. The state space of the underwater channel switching model is defined based on hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data. Determine the action space of the underwater channel switching model, which includes pure underwater acoustic communication mode, pure blue-green optical communication mode, and acoustic-optical parallel hybrid transmission mode. Define the state transition probability and reward function of the underwater channel switching model; The underwater channel handover model is iteratively solved using a lightweight Q-Learning sub-algorithm, specifically as follows: A lightweight Q-table adapted to underwater communication handover scenarios is constructed to store the value estimates of each state-action pair of the underwater channel handover model. The initialized Q-table is serialized and written to the non-volatile storage unit of the underwater node to achieve parameter persistence. At the same time, the hyperparameters of the Q-Learning algorithm are configured, including the learning rate, discount factor and exploration rate. use The strategy outputs a pre-selected action based on the current state, and then combines hysteresis constraints to determine whether link switching is allowed, in order to determine the final selected action.

[0006] Furthermore, the reward function of the underwater channel handover model is:

[0007] in, Represents the current state vector. Represents the state vector at the next moment; Indicates the currently executing switching action. This indicates the historical actions performed in the previous cycle; This represents the positive reward for transmission rate, and its value is a positive number, belonging to the revenue gain term. Indicates hardware power consumption loss. Indicates link reliability loss. This represents the loss from frequent switching, and all three are positive penalty bases. The preset weighted penalty coefficient can be adjusted according to the needs of underwater node tasks. The larger the coefficient value, the stronger the constraint on the corresponding indicator.

[0008] Furthermore, the step of selecting the first optimal transmission link based on the underwater acoustic-optical dual-communication adaptive switching algorithm includes the following steps: The current status is obtained based on the hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data in the first data. For the current state, adopt The strategy retrieves the pre-selected actions from the Q-table, including pure underwater acoustic communication mode, pure blue-green light-optical communication mode, and acoustic-optical parallel hybrid transmission mode. Hysteresis-based anti-ping-pong handover logic is used to constrain action handover. When the channel amplitude exceeds the hysteresis threshold, the current action is locked as the selected action and no action handover is performed; when the channel amplitude exceeds the hysteresis threshold, the pre-selected action is set as the selected action. The first optimal transmission link is determined based on the selected action, and the first data is transmitted to the sea surface relay layer via the first optimal transmission link.

[0009] Furthermore, the dual-link adaptive cross-domain communication algorithm is an intelligent scheduling method for cross-domain transmission in the sea area, based on dual communication links, real-time monitoring of link transmission quality, and adaptive switching of the optimal link according to sea conditions and channel status. The method of transmitting the second data to the shore-based control layer using a dual-link adaptive cross-domain communication algorithm includes the following steps: According to the preset period, the dual-link communication parameters, equipment power parameters, buoy attitude parameters and ocean wave environment parameters are collected in parallel. The collected Beidou Tiantong dual-link communication parameters, equipment power parameters, buoy attitude parameters and ocean wave environment parameters are normalized and cleaned, and abnormal sampled values ​​are removed to generate a standardized input dataset. The standardized input dataset is input into the link comprehensive scoring model to calculate the comprehensive scores of the first satellite link and the second satellite link respectively; The second optimal transmission link is determined based on the comprehensive score and the hierarchical decision rule; the hierarchical decision rule refers to the scheduling logic of dividing priority levels according to data urgency and determining the transmission link hierarchically. Predict short-term buoy fluctuations using wave fitting models to pinpoint effective communication windows; When the effective communication window period is reached and the link status is normal, a differentiated compression strategy is executed according to the data level of the second data, and the compressed second data is forwarded to the shore-based platform through the second optimal transmission link and satellite domain, so that the shore-based platform can clear its local cache after responding and confirming; when the effective communication window period is reached and a link interruption and signal attenuation failure occur, the DTN storage and carry mechanism is activated to cache all untransmitted second data, and after the link is restored to normal, the breakpoint location is automatically identified, and the untransmitted second data is fragmented and resumed.

[0010] Furthermore, the comprehensive link scoring model is as follows:

[0011] in, This indicates the overall link score; This represents the normalized signal-to-noise ratio quality score; This represents the normalized value of the remaining battery capacity. This represents the inverse score of time delay normalization. This indicates the normalized score for the business urgency level. This represents the inverse score of the historical packet loss rate. This represents a five-dimensional adaptive weight that satisfies the weight normalization constraint.

[0012] Furthermore, the multi-source heterogeneous data fusion algorithm is constructed based on the multi-source heterogeneous original dataset, the spatiotemporal benchmark unified model, the triple dynamic credibility weight model, the spatiotemporal joint distance decay model, the global multi-source data fusion master calculation model, and the online adaptive parameter iterative optimization mechanism. The multi-source heterogeneous raw dataset is a set of structured samples formed after the second data has been purified. Each sample in the multi-source heterogeneous raw dataset corresponds to a single valid observation information of a monitoring node. Each sample includes marine element observation values, spatial location coordinates, local acquisition time of the equipment, real-time link quality score, equipment type identifier, data transmission delay and unique node number. The unified spatiotemporal benchmark model is used for unified calibration of the global spatiotemporal benchmark of the multi-source heterogeneous original dataset; the triple dynamic credibility weight model is used to calculate the dynamic credibility weight of samples in the multi-source heterogeneous original dataset; the spatiotemporal joint distance decay model is used to calculate the spatiotemporal joint decay weight of samples in the multi-source heterogeneous original dataset; and the global multi-source data fusion master calculation model is used to perform global gridded fusion of the multi-source heterogeneous original dataset to generate third data. The online adaptive parameter iterative optimization mechanism calculates the fusion error of the third data based on real-time measured data, and adjusts the spatiotemporal weight control parameters online according to the fusion error to continuously optimize the fusion accuracy of the third data. Finally, the optimized third data is standardized and packaged, and output to the digital twin visualization module, intelligent scheduling algorithm module, and business early warning and data storage module, respectively.

[0013] Furthermore, the unified spatiotemporal reference model is used for the unified calibration of the global spatiotemporal reference of the multi-source heterogeneous original dataset. Specifically, it uses BeiDou UTC standard time and CGCS2000 geodetic coordinate system as the only global reference to correct the local clock offset and local coordinate deviation problems of each node in the multi-source heterogeneous original dataset, and constructs a uniform gridded calculation area for the monitored sea area to determine the standard spatiotemporal coordinates of the center point of each grid. The triple dynamic credibility weight model is used to calculate the dynamic credibility weight of samples in the multi-source heterogeneous original dataset. Specifically, for each sample in the multi-source heterogeneous original dataset, a dynamic comprehensive credibility weight is calculated based on three dimensions: inherent device precision, real-time communication link quality, and data transmission latency. The weight of low-precision devices, poor-quality communication link data, and data with time delays is automatically reduced, and failed link data is forcibly removed, thus completing the intelligent screening and credibility classification of the multi-source heterogeneous original dataset. The spatiotemporal joint distance attenuation model is used to calculate the spatiotemporal joint attenuation weight of samples in the multi-source heterogeneous original dataset. Specifically, with the center point of each grid as the reference, the spatial Euclidean distance and time difference between the effective samples in the multi-source heterogeneous original dataset and the grid points are calculated respectively. The spatiotemporal contribution weight of the effective samples is solved by the spatiotemporal joint Gaussian attenuation model, thereby increasing the weight of the effective samples whose spatiotemporal position is closer to the grid and whose collection time is more recent. The full-domain multi-source data fusion master calculation model is used to perform full-domain gridded fusion of the multi-source heterogeneous original dataset to generate third data. Specifically, it fuses sample confidence weights and spatiotemporal decay weights, adopts a normalized weighted fusion algorithm, calculates the element fusion value for each grid point in the entire domain, and uses discrete multi-source heterogeneous original dataset sample interpolation to fill the blank areas of marine observation, eliminating the problems of uneven distribution of data measurement points and local missing data, and generating third data. The third data is standardized gridded fusion data that is continuous, regular, uniform, and without data gaps throughout the domain.

[0014] Furthermore, the main computing model for the full-domain multi-source data fusion is as follows:

[0015] in, This represents the output value of the fusion of gridded marine elements across the entire region. This represents the overall weight of credibility. Indicates the first The contribution weight of each sample to the grid point This represents the set of valid samples with non-zero weights within the current spatiotemporal window. Indicates the first Multidimensional ocean element observation vectors for each sample; express The first in One sample.

[0016] Based on the above-described method embodiments, the present invention provides another embodiment; Another embodiment of the present invention provides a data transmission system based on a marine multi-element environmental monitoring network, including an underwater sensing layer subsystem, a sea surface relay subsystem, and a shore-based control subsystem; The underwater sensing layer subsystem is used for multi-element marine environment full-domain acquisition, local edge preprocessing and caching, underwater short-range node adaptive Mesh self-organizing network data transmission, and on-site anomaly autonomous alarm; The sea surface relay subsystem is used for cross-medium signal docking, multi-source heterogeneous data preprocessing and aggregation, dual-link adaptive cross-domain intelligent scheduling, and bidirectional relay and local fault-tolerant caching. The shore-based control subsystem is used for unified fusion processing of multi-source heterogeneous data, intelligent scheduling and communication strategy control of the entire network, full-domain visual control, global network self-healing decision-making, business applications and alarms. The first data is acquired through the underwater sensing layer subsystem, and the first optimal transmission link is selected according to the underwater acoustic-optical dual-communication adaptive switching algorithm. The first data is then transmitted to the sea surface relay subsystem through the first optimal transmission link. The first data is the raw observation data of multi-source marine environment, channel and equipment operating conditions. The underwater acoustic-optical dual-communication adaptive switching algorithm is constructed by establishing an underwater channel switching model based on the Markov decision process and iteratively solving the underwater channel switching model using a lightweight Q-Learning sub-algorithm. After processing the first data into second data, the surface relay subsystem uses a dual-link adaptive cross-domain communication algorithm to transmit the second data to the shore-based control subsystem. The shore-based control subsystem then uses a multi-source heterogeneous data fusion algorithm to process the second data into third data. The dual-link adaptive cross-domain communication algorithm is used for dynamic switching between dual links to adapt to sea area transmission and to stably transmit marine monitoring data across domains. The multi-source heterogeneous data fusion algorithm is used to couple spatiotemporal and reliability weights and to fuse the second data in a grid-like manner.

[0017] The embodiments of the present invention have the following beneficial effects: This invention provides a data transmission method and system based on a multi-element marine environmental monitoring network. The method constructs a three-layer collaborative monitoring network integrating underwater sensing, surface relay, and shore-based control, breaking down technical barriers at each level and forming an integrated closed-loop system encompassing data acquisition, transmission, fusion, scheduling, and early warning. The system integrates intelligent scheduling across the entire network, network self-healing, and full-domain visualized control, enabling autonomous operation and adaptive optimization of the network. This effectively improves the intelligence and autonomy of the marine monitoring network, thereby enhancing the efficiency and reliability of marine data transmission. The underwater sensing layer overcomes the limitations of traditional single underwater communication by integrating underwater acoustic and optical dual communication channels. It can dynamically select the optimal transmission link based on underwater channel quality and sea conditions, balancing transmission distance and rate advantages. The surface layer innovates a dual-link adaptive cross-domain relay transmission mode, abandoning the traditional single-link backhaul scheme and establishing a dual-link redundant intelligent scheduling mechanism to support dynamic link switching. Furthermore, this method proposes a spatiotemporal weighted multi-source heterogeneous data fusion algorithm that integrates spatiotemporal joint distance attenuation weights and equipment reliability weights to perform gridded weighted fusion of multi-time series and multi-type marine data, accurately filling data gaps in sparse areas of measurement points. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of a data transmission method based on a marine multi-element environmental monitoring network provided in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the process for selecting the first optimal transmission link in a data transmission method based on a marine multi-element environmental monitoring network according to an embodiment of the present invention.

[0020] Figure 3 This is a schematic diagram of the data transmission system based on a marine multi-element environmental monitoring network provided in an embodiment of the present invention. Detailed Implementation

[0021] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] See Figure 1 This is a schematic diagram of a data transmission method based on a marine multi-element environmental monitoring network according to an embodiment of the present invention, including the following steps: Step S101: Acquire first data through the underwater sensing layer, select the first optimal transmission link according to the underwater acoustic-optical dual-communication adaptive switching algorithm, and transmit the first data to the sea surface relay layer through the first optimal transmission link; the first data is the raw observation data of multi-source marine environment, channel and equipment operating conditions. Specifically, the first data includes basic environmental data, underwater real-time channel data and node equipment status data. The basic environmental data includes hydrological data, water quality data, meteorological data and dynamic environment data; the underwater sensing layer is used for multi-element marine environment full-domain acquisition, local edge preprocessing and caching, underwater short-range node adaptive Mesh self-organizing network data transmission and on-site anomaly autonomous alarm; the underwater sensing layer includes a multi-element integrated sensor group, a fixed underwater mooring node and an underwater mobile monitoring node. The fixed underwater mooring node and the underwater mobile monitoring node are equipped with an underwater acoustic-optical dual-communication adaptive switching communication module. The acoustic-optical dual-communication adaptive switching communication module has a built-in underwater acoustic-optical dual-communication adaptive switching algorithm. The underwater acoustic-optical dual-communication adaptive switching algorithm is constructed by establishing an underwater channel switching model based on a Markov decision process, and then using a lightweight Q-Learning sub-algorithm to iteratively solve the underwater channel switching model. The underwater short-range node adaptive mesh self-organizing network data transmission is a distributed, decentralized underwater acoustic networking transmission mechanism deployed in the underwater sensing layer. Each underwater sensing node uses short-range underwater acoustic communication as its link foundation, autonomously completing neighbor discovery, topology self-construction, and dynamic routing networking to form a mesh topology structure. The system can adapt to dynamic environmental changes such as underwater acoustic attenuation, multipath interference, node dormancy / failure, and ocean current displacement.

[0023] The underwater acoustic-optical dual-communication adaptive switching algorithm is a lightweight Q-Learning reinforcement learning algorithm. It integrates the first data as the state input and autonomously outputs three types of actions: pure underwater acoustic communication mode, pure blue-green light-optical communication mode, and acoustic-optical parallel hybrid transmission mode. The optimization objectives are to maximize transmission throughput, minimize bit error rate and packet loss, reduce link switching power consumption, and reduce local data backlog. It has built-in hysteresis anti-ping-pong switching logic and is compatible with low-computing-power edge hardware such as fixed underwater moorings and AUV mobile monitoring nodes.

[0024] As one embodiment, an underwater channel handover model is established based on a Markov decision process, and a lightweight Q-Learning algorithm is used to iteratively solve the underwater channel handover model to construct the underwater acoustic-optical dual-communication adaptive handover algorithm. The establishment of the underwater channel handover model based on the Markov decision process includes: The optimization objectives of the underwater channel handover model are to maximize transmission throughput, minimize bit error rate and packet loss, reduce link handover power consumption, and reduce local data backlog. The state space of the underwater channel switching model is defined based on hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data. Specifically, since the original sensing data is continuous, to adapt to embedded low computing power and small-capacity Q-tables, each type of indicator data is segmented and discretized. For a finite number of data points, a single set of discrete observation vectors constitutes a state vector:

[0025] in, This indicates the hydrological status level, including but not limited to water temperature, salinity, and water depth stratification. This indicates the water quality level, including but not limited to turbidity and suspended particulate matter concentration; This indicates the weather and sea surface condition level, including but not limited to wind speed, wave height, and sea surface bubble layer intensity; This indicates the dynamic environment setting, including ocean current velocity and Doppler frequency deviation; This indicates the underwater acoustic channel status level, including but not limited to underwater acoustic SNR, multipath delay, and transmission loss; This indicates the optical channel status level, including but not limited to optical receive SNR and optical attenuation coefficient; The remaining battery level of a node is represented by S; S is the complete state space set, generated by the Cartesian product of discrete levels in each dimension. Considering the inherent temporal correlation in the underwater environment, this embodiment of the invention performs smoothing preprocessing on the original continuous data through a short-time sliding window, and then implements discrete quantization of the levels to weaken the influence of long-period historical states, so that the system approximately satisfies the Markov assumption of no aftereffect: the state vector at the next moment is determined only by the current state vector and the currently executed switching action, and is not related to the earlier historical state sequence.

[0026] The action space of the underwater channel switching model is determined, including pure underwater acoustic communication mode, pure blue-green optical communication mode, and parallel acoustic-optical hybrid transmission mode; specifically, the set of discrete acoustic-optical channel switching actions that can be executed by each node of the underwater sensing layer is defined. ;in, This indicates a pure underwater acoustic communication mode, such as in long-distance or high-turbidity scenarios, where only control commands and compressed multi-element average small packets are transmitted; the optical emission module is turned off to reduce power consumption. It represents a pure blue-green light optical communication mode, such as in close-range low-turbidity scenarios, for high-speed transmission of large-capacity raw sensor profile data stored locally. This invention represents a hybrid acoustic-optical parallel transmission mode, where dual channels work independently and collaboratively during transition intervals. The optical channel carries high-capacity service data transmission, while the underwater acoustic channel independently transmits heartbeat messages and scheduling signaling. The two channels do not compete for time slot resources, achieving reliable parallel transmission of data and signaling. Simultaneously, a hysteresis-prevented ping-pong handover logic is employed to constrain action switching. When the channel amplitude exceeds the hysteresis threshold, the current action is locked, and action switching is not performed; otherwise, action switching is executed. This invention breaks away from the simple mutually exclusive switching mode of acoustic and optical channels, innovatively designing three adaptive transmission modes: pure underwater acoustic, pure optical, and acoustic-optical parallel, which can accurately match service scenarios based on sea turbidity and transmission distance. Low-power pure underwater acoustic transmission is used in harsh, long-distance scenarios, while high-speed pure optical transmission is used in favorable, short-distance scenarios. Dual-channel parallel operation is employed during transition intervals. This solution achieves decoupled transmission of service data and scheduling signaling. The optical channel carries high-capacity data, and the underwater acoustic channel independently transmits scheduling signaling, avoiding time slot resource contention and improving channel utilization. Simultaneously, it integrates reinforcement learning decision-making and hysteresis threshold constraint mechanisms to perform secondary verification and locking of channel state, suppress ping-pong handover errors caused by short-term interference, balance adaptive optimization capability and system stability, and reduce power consumption and link oscillation risk of invalid handover.

[0027] Define the state transition probability and reward function of the underwater channel switching model.

[0028] As one embodiment, the state transition probability is defined as follows:

[0029] Represents the current state vector. Indicates the currently executing switching action. This is the state vector at the next moment after the action is performed; Indicates the current state Execute the switching action below The state evolves to the next time step due to random factors such as the marine environment, channel disturbances, and node energy consumption. The conditional probability; Represents a probability measure; express The action random variable at time. Simultaneously, any set of... All corresponding subsequent state transition probabilities satisfy the normalization constraint: That is, starting from the same state and the same action, the sum of the probabilities of all possible next states is 1.

[0030] As one embodiment, the reward function is defined as follows:

[0032] in, Indicates an immediate reward. This indicates the historical actions performed in the previous cycle; This represents the positive reward for transmission rate, and its value is a positive number, belonging to the revenue gain term. When the channel quality is good, the positive reward ranking is: parallel transmission > pure optical communication > pure underwater acoustic communication. The better the channel quality, the higher the rate reward. This indicates hardware power consumption loss; optical communication transmission power consumption is greater than underwater acoustic power consumption, with dual-channel power consumption being the highest; automatic amplification occurs at low battery levels. Weighting, forcing energy conservation; Indicates link reliability loss, when When the mid-channel SNR is too low and the packet loss rate exceeds the standard, apply a large negative reward to avoid selecting a failed channel; This indicates the loss from frequent switching. , , All three are positive penalty base numbers; The preset weighted penalty coefficient can be adjusted according to the underwater node task requirements; the larger the coefficient value, the stronger the constraint on the corresponding indicator. This invention constructs a dedicated reward and punishment rule system tailored to underwater communication scenarios. It applies large negative rewards to failed channel actions with excessively low SNR and excessive packet loss rate to suppress the selection of invalid channels; and provides positive rewards for high-stability, low-latency, and low-power transmission actions to continuously strengthen the optimal strategy.

[0033] The underwater channel handover model is solved iteratively using a lightweight Q-Learning algorithm, specifically as follows: A lightweight Q-table adapted to underwater communication handover scenarios is constructed to store the value estimates of each state-action pair in the underwater channel handover model. .

[0034] The initialized Q-table is serialized and written to the underwater node's non-volatile storage unit to achieve parameter persistence. Simultaneously, the hyperparameters of the Q-Learning algorithm are configured, including the learning rate, discount factor, and exploration rate; the discount factor... The range of values ​​is Preferably, the learning rate A small learning rate adapts to the slowly changing marine environment, avoiding drastic strategy oscillations; discount factor Emphasis should be placed on short-term transmission gains; exploration rate , Strategy. Preferably, the Q-table is subjected to lightweight processing, including performing tiered discrete compression on the first data to significantly reduce the total number of states. The Q-value is stored using 16-bit fixed-point integer quantization; the Q-table is serialized and written to Flash non-volatile memory, and the initial estimate is directly loaded after power-on without repeated initialization. The strategy outputs a pre-selected action based on the current state, and then combines this with hysteresis constraints to determine whether link switching is allowed, thus determining the final selected action. Preferably, a hysteresis-based anti-ping-pong switching logic is used to constrain action switching. When the channel amplitude exceeds the hysteresis threshold, the current action is locked as the selected action, and no action switching is performed; when the channel amplitude exceeds the hysteresis threshold, the action switching is performed, i.e., the pre-selected action becomes the selected action. This invention differs from traditional fixed threshold decision-making and heavy offline machine learning schemes by constructing a lightweight online iterative learning framework without dataset dependence. It relies on real-time interaction between nodes and the marine environment for autonomous convergence, requiring no offline samples or pre-trained models. The lightweight Q-table estimation is corrected through temporal difference updates, combined with... The linear decay mechanism of coefficients balances the exploration of the environment and the reuse of the optimal strategy, dynamically iteratively optimizes the communication switching strategy, and is adapted to low-computing-power embedded devices underwater, solving the drawbacks of traditional algorithms such as redundant computing power and inability to iterate in real time.

[0035] As one embodiment, the underwater acoustic-optical dual-communication adaptive switching algorithm is trained according to the following steps: Step S10: Obtain the current status based on the hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data.

[0036] Step S11: For the current state, adopt... The strategy outputs pre-selected actions, which are then combined with hysteresis constraints to determine whether link switching is allowed, in order to determine the final selected action.

[0037] Step S12: After executing the selected action, collect the hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data for the next moment to obtain the status of the next moment.

[0038] Step S13: Input the current state, selected action, next moment state, historical actions executed in the previous cycle, positive transmission rate reward, hardware power consumption loss, link reliability loss, and frequent handover loss into the reward function to calculate the instant reward. The instant reward is used to quantify the quality of this action and provide a basis for Q-value updates.

[0039] Step S14: Update the value estimate of the state-action pair in the Q-table using the temporal difference update formula, so that the current state is updated to the successor state, i.e. The process repeats steps S11 to S14 until the underwater acoustic-optical dual-communication adaptive switching algorithm converges. Specifically, when the estimated update magnitude of all state-action pairs in the Q-table is less than the preset convergence threshold, training is considered complete and convergence has been achieved. The temporal difference update formula must rely on the current instantaneous reward and the estimated value of the next state to calculate the temporal difference error, thereby correcting and updating the state-action estimates corresponding to the lightweight Q-table. The temporal difference update formula is as follows:

[0040] in, This indicates the pre-selected action to be performed at the next moment. This indicates a one-step, immediate reward.

[0041] See Figure 2 This is a flowchart illustrating the selection of the first optimal transmission link in a data transmission method for a marine multi-element environmental monitoring network according to an embodiment of the present invention. As one embodiment, the selection of the first optimal transmission link based on the underwater acoustic-optical dual-communication adaptive switching algorithm includes the following steps: Step S1011: Based on the hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data in the first data, obtain the current status; Step S1012: For the current state, adopt... The strategy retrieves the pre-selected actions from the Q-table, including pure underwater acoustic communication mode, pure blue-green light-optical communication mode, and acoustic-optical parallel hybrid transmission mode. Step S1013: Use hysteresis anti-ping-pong handover logic to constrain action handover. When the channel amplitude exceeds the hysteresis threshold, lock the current action as the selected action and do not perform action handover; when the channel amplitude exceeds the hysteresis threshold, set the pre-selected action as the selected action. Step S1014: Determine the first optimal transmission link according to the selected action, and transmit the first data to the sea surface relay layer via the first optimal transmission link.

[0042] As one embodiment, step S1014 specifically involves: accurately matching the corresponding first optimal transmission link according to the selected action: if the selected action is a pure underwater acoustic communication mode, then the underwater acoustic communication channel is activated, the optical transmission module is turned off, and a low-power command and small packet data transmission link is constructed; if the selected action is a pure optical communication mode, then the blue-green optical communication channel is activated, and a high-speed, high-capacity raw data transmission link is constructed; if the selected action is an acoustic-optical parallel communication mode, then the dual channels are simultaneously enabled to work independently, and a parallel transmission link of the optical channel and the underwater acoustic channel is constructed, including the optical channel transmitting service data and the underwater acoustic channel transmitting signaling heartbeats.

[0043] Step S102: The surface relay layer performs time synchronization, noise reduction, format standardization, and data classification on the first data to obtain the second data. A dual-link adaptive cross-domain communication algorithm is then used to transmit the second data to the shore-based control layer, which then uses a multi-source heterogeneous data fusion algorithm to process the second data into the third data. The surface relay layer is used for cross-medium signal docking, multi-source heterogeneous data preprocessing, dual-link adaptive cross-domain intelligent scheduling, and bidirectional relay and local fault-tolerant buffering. The dual-link adaptive cross-domain intelligent scheduling refers to real-time sensing of the dual-link status and adaptive optimal switching to achieve reliable intelligent cross-domain transmission on the surface. The bidirectional relay and local fault-tolerant buffering refers to bidirectional transmission and reception of command data, with local buffering ensuring the integrity and reliability of cross-domain transmission. The surface relay layer uses surface buoys as carriers and includes a main control processing module, an underwater cross-medium communication module, a dual-link long-distance communication module, and a multi-parameter parallel sensing module. The data tiered processing specifically involves classifying data according to business priorities: Level 1 is for hazard alarms and anomaly monitoring data, i.e., the highest priority data; Level 2 is for routine timed observation data; and Level 3 is for raw high-frequency waveforms and massive log data, i.e., low priority data and large-capacity data. The shore-based control layer is used for unified fusion processing of multi-source heterogeneous data, intelligent scheduling and communication strategy control across the entire network, full-domain visual control, global network self-healing decision-making, business applications, and alarms. The intelligent scheduling and communication strategy control across the entire network refers to intelligent optimal scheduling of all links, adaptively controlling communication transmission strategies and power consumption. The global network self-healing decision-making refers to dynamically reconstructing transmission topology links and autonomously repairing network faults to ensure communication self-healing. The shore-based control layer includes a multi-mode satellite communication access module, a central data processing server, a large-capacity distributed storage unit, a human-machine interactive terminal, a full-domain network scheduling module, an underwater node operation and maintenance control module, a marine data intelligent analysis module, and a disaster early warning decision-making module. The shore-based control layer is deployed in the land-based computer room, monitoring center, and aquaculture control center.

[0044] The underwater cross-medium communication module is used to receive the first data uploaded by the underwater sensing layer, complete the conversion of underwater acoustic signals to electrical signals, and transmit the multi-source heterogeneous raw data to the main control processing module. The main control processing module is used to perform time synchronization processing, noise reduction processing, and format standardization processing on the received first data, and to perform data hierarchical processing according to the urgency of the business. At the same time, it controls the multi-parameter parallel sensing submodule to synchronously collect dual-link communication parameters, equipment power parameters, buoy attitude parameters, and ocean wave environment parameters. By running a five-dimensional weighted link decision algorithm, an ocean wave fitting window prediction algorithm, a hierarchical differential compression algorithm, and online weight iterative training logic, it outputs the second optimal transmission link and data scheduling control commands. The multi-parameter parallel sensing module collects four types of scheduling parameters in parallel at a fixed period, normalizes them, and then sends them to the main control processing submodule. The knowledge module includes a six-axis attitude sensor, a voltage and power acquisition unit, and a dual-communication module status reading unit. The dual-link long-distance communication module is used to transmit processed data packets across the satellite domain to the shore-based control layer during the effective communication window, according to the scheduling instructions issued by the master controller. At the same time, it receives control instructions issued by the shore-based control layer and forwards them to the master controller. The dual-link long-distance communication sub-module includes a Beidou short message communication unit and a Tiantong-1 satellite communication unit. The local cache storage module is used to cache data packets to be transmitted when there is no effective communication window or when the satellite link is interrupted. It also relies on the DTN storage carrying mechanism to realize the interruption resumption after the link is restored. At the same time, it stores algorithm weights and historical transmission statistics. The power supply management module is used to connect the solar photovoltaic panel and the energy storage battery, receive the master controller's low-power scheduling signal, manage the power consumption of the whole machine, and ensure the long-term unattended operation of the marine buoy.

[0045] As one embodiment, the dual-link adaptive cross-domain communication algorithm is an intelligent scheduling method for cross-domain transmission in the sea area, based on dual communication links, real-time monitoring of link transmission quality, and adaptive switching of the optimal link according to sea conditions and channel status. The step of transmitting the second data to the shore-based control layer using the dual-link adaptive cross-domain communication algorithm includes the following steps: Step S1021: Parallel acquisition of dual-link communication parameters, device power parameters, buoy attitude parameters, and ocean wave environment parameters is performed according to a preset period. The acquired BeiDou-TianTong dual-link communication parameters, device power parameters, buoy attitude parameters, and ocean wave environment parameters are then normalized and cleaned, and outlier values ​​are removed to generate a standardized input dataset. Specifically, BeiDou and TianTong dual-link communication parameters, device power parameters, buoy attitude parameters, and ocean wave environment parameters are acquired in parallel at a 100ms period. All raw indicators are normalized and cleaned, outlier values ​​are removed, and the standardized input dataset is generated to provide accurate support for intelligent decision-making.

[0046] Step S1022: Input the standardized input dataset into the link comprehensive scoring model to calculate the comprehensive scores of the first satellite link and the second satellite link respectively. Preferably, the first satellite link is a BeiDou link and the second satellite link is a TianTong link.

[0047] Step S1023: Determine the second optimal transmission link based on the comprehensive score and the hierarchical decision rule. Specifically, the second optimal transmission link is determined by simultaneously considering the comprehensive score and the hierarchical decision rule. The hierarchical decision rule refers to the scheduling logic of dividing priority levels according to the urgency level of the data and determining the transmission link accordingly; when the second data is a level 1 emergency alarm data, the first satellite link is selected; when the second data is level 2 or 3 regular data, or level 2 or 3 large-capacity data, the link with the highest comprehensive score is automatically selected; when the signal is stable and the bandwidth is sufficient, the second satellite link is selected; when in a weak signal scenario, the first satellite link is automatically switched for backup transmission.

[0048] Step S1024: Predict short-term buoy fluctuations using a wave fitting model to lock in the effective communication window. Only during the peak antenna's emergence period, the selected communication module is activated to transmit data; during the trough antenna's submersion period, the RF module is immediately shut down and enters sleep mode, with the data to be transmitted cached in local Flash memory to avoid transmission failures and unnecessary power consumption caused by seawater obstruction. The wave fitting model uses a sine periodic function to fit the sea surface wave fluctuation pattern based on the high-frequency vertical height time-series data collected from the buoy, predicting the buoy height at any given time.

[0049] Step S1025: When the effective communication window period is reached and the link status is normal, a differentiated compression strategy is executed according to the data level of the second data, and the compressed second data is forwarded to the shore-based platform through the second optimal transmission link and satellite domain, so that the shore-based platform can clear its local cache after responding and confirming. When the effective communication window period is reached and a link interruption and signal attenuation failure occur, the DTN storage and carrying mechanism is activated to cache all untransmitted second data, and after the link is restored to normal, the breakpoint location is automatically identified, and the untransmitted second data is fragmented and resumed, completing the full-domain cross-domain reliable transmission from the sea surface to the satellite to the shore-based platform. The differentiated compression strategy is specifically executed according to the data level. Level 1 alarm data is not compressed, only CRC check is performed; Level 2 regular observation data adopts second-order differential compression and LZW lossless hybrid compression; Level 3 massive raw data adopts high-rate compression and segmented storage to adapt to satellite link bandwidth limitations and improve transmission efficiency.

[0050] As one embodiment, the online weight iterative self-training optimization of the dual-link adaptive cross-domain communication algorithm specifically involves: statistically analyzing core indicators such as transmission success rate, average latency, average power consumption, and packet loss rate within an hourly statistical period, substituting them into the loss function and weight update formula, and automatically iteratively correcting the weight parameters of the five-dimensional evaluation model to achieve adaptive optimization of the algorithm and continuously adapt to the dynamic changes in sea area, season, and sea state.

[0051] As one embodiment, the dual-link adaptive cross-domain communication algorithm is constructed based on a link comprehensive scoring model, a wave communication window prediction model, a DTN cross-domain transmission utility judgment model, a hierarchical differential data compression sub-algorithm, and an adaptive weight iterative update sub-algorithm. It integrates five dimensions—channel quality, remaining battery power, transmission delay, data priority, and link stability—to construct a quantifiable and iterative link comprehensive scoring model, accurately characterizing the real-time performance of the two communication links. The link comprehensive scoring model is as follows:

[0052] in, This represents the overall link score, with a value range of [0,1]. A higher score indicates better link transmission performance. It represents the normalized signal-to-noise ratio quality score, characterizing the channel signal quality and the ability to resist sea clutter interference; This represents the normalized value of the remaining battery capacity, characterizing the device's battery life. This represents the normalized score for the business urgency level, with higher scores corresponding to Level 1 alarms, enabling priority scheduling of emergency data. This represents the inverse score of historical packet loss rate; the lower the packet loss rate, the higher the link stability score. , This refers to the packet loss rate of the link. This represents a five-dimensional adaptive weight that satisfies the weight normalization constraint. .

[0053] As one embodiment, the BeiDou link score is calculated respectively. Tiantong Link Score In normal scenarios, it automatically selects high-resolution links for transmission; in emergency alarm scenarios, it forces priority to use BeiDou short message links, realizing a scheduling logic that combines rule-based and intelligent approaches.

[0054] As one embodiment, the wave communication window prediction model is based on high-frequency sampling data of buoy attitude, and uses a sine fitting algorithm to reconstruct the periodic motion pattern of waves, accurately predicting the effective communication period of the antenna. The wave communication window prediction model is as follows:

[0055] in, Indicates the amplitude of ocean waves. Indicates the angular frequency of ocean waves. Indicates the initial phase. Indicates the buoy's reference height. Sets the antenna's exit threshold. When the fitting height If the window is valid, radio frequency transmission is allowed to start; otherwise, it is considered an invalid communication window, the communication module is shut down, and the cached data is put into sleep mode.

[0056] As one embodiment, this invention designs a hierarchical differential hybrid compression sub-algorithm, addressing the inherent characteristics of strong continuity and gentle fluctuations in ocean time-series observation data, and adapting to the narrow bandwidth transmission constraints of sea surface buoy satellites. The algorithm executes differentiated compression strategies according to the urgency level of the data. Regular monitoring data adopts hybrid compression using second-order linear predictive differential combined with LZW lossless compression, significantly reducing the number of transmitted bits. Level 1 alarm data skips the compression process and is transmitted directly, balancing bandwidth utilization with the low latency requirements of emergency transmission. The core formula of the second-order differential predictive compression is as follows:

[0057] in, This represents the original sampled data at the current moment. This represents the predicted value at the current moment obtained by linear extrapolation based on the time series data of the previous two frames. This represents the residual difference between the original value and the predicted value. These represent the historical sampled values ​​from the previous two time moments, respectively; LZW lossless compression algorithm is also used to achieve hybrid compression. Level 1 alarm data is not compressed to ensure millisecond-level low-latency transmission.

[0058] As one embodiment, the DTN cross-domain transmission utility determination model is constructed by quantifying the overall utility of the link, accurately determining the link's working status, and triggering caching and resuming mechanisms.

[0059] in, Indicates link utility, This represents the normalized link quality score. Indicates the reverse normalized link wait time. This represents the normalized transmission power consumption cost. This represents a positive, non-negative weighting coefficient, which supports dynamic adaptive adjustment based on the node's remaining power and the urgency level of the service, controlling the constraint strength of three indicators: link quality, latency, and power consumption cost. When the link effectiveness falls below a preset threshold, the link is determined to have failed, and the DTN storage and carrying mechanism is immediately activated to ensure no data loss.

[0060] As one embodiment, to address the issue of fixed weights failing to adapt to dynamic sea conditions, an embedded online gradient update mechanism is employed. This mechanism adjusts the weights of each dimension in real time based on historical transmission success rates, constructing an adaptive weight iterative update sub-algorithm. This achieves adaptive iterative optimization of the algorithm without the need for manual parameter tuning. The adaptive weight iterative update sub-algorithm is as follows:

[0061] in, Indicates the first The weight of the dimension factor at the current time step; Indicates the weight at the next time step after the iterative update; This represents the learning rate, which is fixed in engineering and ranges from 0.01 to 0.05, balancing iterative stability and optimization efficiency. This represents the real-time transmission success rate of the current link, calculated by the percentage of successfully transmitted data packets within a fixed-length sliding statistical window, with a value range of [0, 1]. This represents the system's long-term average transmission success rate. Indicates the transmission success rate for the first... i The gradient of the weights represents the direction of the impact of weight changes on transmission success rate, and is used to automatically control the direction of weight iteration. .

[0062] This invention employs a five-dimensional adaptive weighted link decision model, abandoning the traditional fixed-priority scheduling mode. Relying on a multi-factor quantitative evaluation and adaptive iterative system, it dynamically adapts to changes in sea conditions, equipment power consumption, and data types, solving the problems of rigid traditional link scheduling and poor scenario adaptability. A wave window matching low-power transmission mechanism is introduced, combining the buoy's wave motion patterns to achieve precise packet transmission at wave peaks and sleep buffering at wave troughs, eliminating unnecessary power consumption and effectively improving equipment endurance. A tiered compression and DTN cross-domain transmission fusion architecture is constructed, addressing issues such as limited satellite bandwidth, intermittent ocean links, and data loss through bandwidth adaptation and breakpoint resumption mechanisms. Simultaneously, an embedded online lightweight self-training mechanism is adopted, eliminating the need for massive data and high-end computing power. The device can autonomously iterate and optimize, freeing it from manual parameter tuning and adapting to long-term unattended marine monitoring scenarios.

[0063] As one embodiment, the multi-source heterogeneous data fusion algorithm is constructed based on a multi-source heterogeneous original dataset, a unified spatiotemporal benchmark model, a triple dynamic credibility weight model, a spatiotemporal joint distance attenuation model, a global multi-dimensional data fusion master calculation model, and an online adaptive parameter iterative optimization mechanism. Specifically, a multi-source heterogeneous original dataset is first constructed. Relying on the triple dynamic credibility weight model, the credibility of the samples is calculated from the perspectives of device accuracy, link quality, and transmission delay, and low-quality samples are filtered out. Then, the spatiotemporal contribution weight of the samples to each grid is calculated through the spatiotemporal joint Gaussian attenuation model. Using the global multi-dimensional fusion model, global gridded fusion data is generated by joint normalized weighted interpolation of credibility and spatiotemporal weights. Finally, an online adaptive iterative mechanism is introduced, which uses RMSE error to fine-tune the spatial scale, temporal scale, and backup basic weights. The parameters are synchronously written back to the preceding model for iterative optimization, forming an end-to-end closed-loop multi-source heterogeneous data fusion algorithm.

[0064] The multi-source heterogeneous raw dataset is a structured sample set formed after the second data has been cleaned. Each sample in the multi-source heterogeneous raw dataset corresponds to a single valid observation information of a monitoring node. Each sample includes marine element observation values, spatial location coordinates, local acquisition time of the equipment, real-time link quality score, equipment type identifier, data transmission latency, and a unique node number. Preferably, the multi-source heterogeneous raw dataset is:

[0065] in, This represents a multi-source heterogeneous original dataset, loaded in batches using a fixed-length time-series sliding window. Indicates the first The sample multidimensional marine element observation vector, including indicators such as water temperature, salinity, water depth, and turbidity, is obtained by receiving compressed data from sea surface buoys and restoring it through LZW and differential decoding; This represents a globally unique identifier for the monitoring node, used to distinguish different data acquisition terminals; Indicates the original coordinates of the sample collection point; Indicates the device's local time; Indicates real-time link quality; Indicates the device type; Indicates data latency; This represents the total number of valid observation samples within the current time sliding window. The constructed multi-source heterogeneous original dataset provides a standardized and highly purified basic data source for subsequent spatiotemporal calibration, dynamic weighting, and global gridded fusion calculation.

[0066] As one embodiment, the unified spatiotemporal reference model is used for the unified spatiotemporal reference calibration of the multi-source heterogeneous raw dataset. Preferably, the unified spatiotemporal reference model uses BeiDou UTC standard time and CGCS2000 geodetic coordinate system as the sole global reference, correcting the local clock offset and local coordinate deviation problems of each node in the multi-source heterogeneous raw dataset, and achieving accurate spatiotemporal alignment of all heterogeneous data; it constructs a uniform gridded computing area for the monitored sea area, determines the standard spatiotemporal coordinates of the center point of each grid, and provides a unified computing carrier for subsequent interpolation fusion and data completion. Therefore, the unified spatiotemporal reference model is used to complete the full-domain spatiotemporal reference calibration of the multi-source heterogeneous raw dataset. By unifying BeiDou UTC time and CGCS2000 spatial reference and constructing a uniform gridded computing base for the sea area, it achieves spatiotemporal alignment and standardization of multi-source data, providing accurate pre-processing data support for dual-link adaptive cross-domain communication scheduling.

[0067] As one embodiment, the triple dynamic credibility weight model is used to dynamically calculate the credibility weight of samples in the multi-source heterogeneous original dataset. Specifically, for each sample in the multi-source heterogeneous original dataset, a dynamic comprehensive credibility weight is calculated based on three dimensions: inherent device accuracy, real-time communication link quality, and data transmission latency. The weights of low-precision devices, data from poor-quality communication links, and data with time delays are automatically reduced, and failed link data is forcibly removed, thus completing the intelligent screening and credibility classification of the multi-source heterogeneous original dataset. Preferably, the basic accuracy weight for high-precision AUV configuration devices is 0.9, the basic accuracy weight for fixed-point underwater mooring configuration devices is 0.8, the basic accuracy weight for surface buoy configuration devices is 0.6, and the basic accuracy weight for simple sensor configuration devices is 0.4. The triple dynamic credibility weight model is as follows:

[0068] in, Indicates the first The overall credibility weight of each sample This indicates the basic accuracy weight of the equipment. Indicates the communication link quality correction weight. This represents the time delay decay weight.

[0069] As one embodiment, the spatiotemporal joint distance attenuation model is used to calculate the spatiotemporal joint attenuation weights for samples in the multi-source heterogeneous original dataset. Preferably, using the center point of each grid as a reference, the spatial Euclidean distance and temporal difference between the effective samples in the multi-source heterogeneous original dataset and the grid points are calculated respectively. The spatiotemporal contribution weight of the effective samples is solved by the spatiotemporal joint Gaussian attenuation model, increasing the weight of effective samples whose spatiotemporal location is closer to the grid and whose collection time is more recent. The spatiotemporal joint distance attenuation model addresses the problem of sparse and unevenly distributed ocean measurement points by constructing a two-dimensional spatiotemporal Gaussian attenuation weight, achieving smooth transition of spatial interpolation and dynamic updating of the temporal dimension. Preferably, the spatiotemporal joint distance attenuation model is as follows:

[0070] in, This indicates the contribution weight of the current sample to the grid points. This represents the spatial Euclidean distance between the sample and the center of the grid. Indicates time distance, The parameter representing the spatial influence radius scale. This indicates the time-dependent decay scale parameter.

[0071] As one embodiment, the full-domain multi-source data fusion master calculation model is used to perform full-domain gridded fusion of the multi-source heterogeneous original dataset to generate third data. Specifically, it involves fusing sample reliability weights and spatiotemporal attenuation weights, employing a normalized weighted fusion algorithm, calculating the element fusion value for each grid point across the entire domain, and using discrete interpolation of the multi-source heterogeneous original dataset samples to fill in the blank areas of marine observation, eliminating the problems of uneven distribution of data measurement points and local missing data, thereby generating third data. The third data is standardized gridded fusion data that is continuous, regular, uniform, and without data gaps across the entire domain. The full-domain multi-source data fusion master calculation model, through dual weight joint constraints, ensures both data reliability and spatial continuity and temporal real-time performance. Preferably, the full-domain multi-source data fusion master calculation model is as follows:

[0072] in, This represents the fused output value of marine elements at the target grid point, i.e., the final interpolated and fused values ​​of marine environmental elements (such as sea temperature, salinity, current velocity, etc.) at the target grid point. This represents the set of all valid samples with non-zero weights within the current spatiotemporal window; Represents a set Inner One valid observation sample.

[0073] In one embodiment, the online adaptive parameter iterative optimization mechanism calculates the fusion error of the third data based on real-time measured data, and adaptively fine-tunes the spatiotemporal weight control parameters online according to the fusion error to continuously optimize the fusion accuracy of the third data. Finally, the optimized third data is standardized and encapsulated, and output to the digital twin visualization module, intelligent scheduling algorithm module, and business early warning and data storage module, respectively, completing the complete transformation from second data to third data. The spatiotemporal weight control parameters include spatial influence radius, temporal influence scale, and equipment baseline weight. Preferably, the fusion error is calculated according to the following formula:

[0074] in, This represents the error evaluation index of the fusion result. The larger the value, the greater the deviation between the data fusion result and the true new observation under the current parameters. It is the loss function of this optimization mechanism and is used to guide the optimization process. , Iterative adjustments; This represents the total number of independent validation samples that participated in the error verification. Indicates the first A sample used for error verification; Indicates the first Fusion prediction of marine elements for each sample; Indicates the first Each sample contains measured values. The core hyperparameters are dynamically fine-tuned based on error changes. For example, when measurement points are sparse and fusion errors are large, the spatial influence radius is adaptively increased; when sea state changes abruptly or factors change drastically, the time scale is adaptively reduced, and the weight of new data is strengthened; when a certain type of equipment has consistently large errors, the basic weight of the corresponding equipment is automatically reduced. The parameter fine-tuning range is limited to ±10% to ensure model stability and prevent oscillations, achieving a closed-loop capability of running, optimizing, and adapting simultaneously.

[0075] In one embodiment, the shore-based control layer receives the second data output by the sea surface relay layer, and uses the multi-source heterogeneous data fusion algorithm to integrate and optimize the second data across the entire domain, transforming it into standardized third data that is spatiotemporally unified, globally continuous, and highly reliable. The process of enabling the shore-based control layer to use the multi-source heterogeneous data fusion algorithm to process the second data into the third data includes the following steps: Step S1026: Second Data Reception and Dirty Data Cleaning. The shore-based control system receives the second data uploaded by each sea surface relay layer in a laminar flow manner, completes the packet unpacking and parsing, and extracts core information such as marine element observation values, equipment parameters, spatiotemporal coordinates, communication link quality, and transmission delay of each monitoring node; empty packets, duplicate packets, and garbled data are removed, and abnormal jump data are removed by combining marine physical thresholds, completing the primary cleaning of the second data and constructing an effective original observation sample set.

[0076] Step S1027: Unified calibration of the spatiotemporal reference for the second data. Using BeiDou UTC standard time and CGCS2000 geodetic coordinate system as the sole global reference, the local clock offset and local coordinate deviation of each node in the second data are corrected to achieve accurate spatiotemporal alignment of all heterogeneous data; a uniform gridded calculation area is constructed for the monitored sea area, and the standard spatiotemporal coordinates of the center point of each grid are determined to provide a unified calculation carrier for subsequent interpolation fusion and data completion.

[0077] Step S1028: Calculation of Dynamic Credibility Weights for Samples. For each sample of the second data after cleaning and calibration, a dynamic comprehensive credibility weight is calculated based on three dimensions: inherent device accuracy, real-time communication link quality, and data transmission latency. The weights of low-precision devices, poor-quality communication links, and low-quality data with time delays are automatically weakened, and data from failed links is forcibly removed, thus completing the intelligent screening and credibility grading of the second data.

[0078] Step S1029: Calculation of Spatiotemporal Joint Attenuation Weights. Using the center point of each grid as a reference, calculate the spatial Euclidean distance and time difference between the effective second data sample and the grid point. Solve the spatiotemporal contribution weight of the sample through the spatiotemporal joint Gaussian attenuation model, so that the sample of the second data that is closer to the grid in terms of spatiotemporal location and has a more recent collection time has a higher proportion of fusion contribution.

[0079] Step S1030: Generate third data through full-domain grid fusion. The reliability weight and spatiotemporal attenuation weight of the fusion samples are combined, and a normalized weighted fusion algorithm is used to calculate the fusion value of each element for all grid points across the entire domain. Discrete second data samples are interpolated to fill in the blank areas of the marine observation, eliminating the defects of uneven distribution and local missing points in the second data. This generates standardized grid fusion data that is continuous, regular, and without data gaps across the entire domain, thus obtaining the third data.

[0080] Step S1031: Adaptive Data Optimization and Standardized Output. Based on real-time measured data, the fusion error is calculated, and core parameters such as spatial influence radius, temporal influence scale, and equipment baseline weight are adaptively fine-tuned online to continuously optimize the output accuracy of the third data. Finally, the optimized third data is standardized and packaged, and output to the digital twin visualization module, intelligent scheduling algorithm module, and business early warning and data storage module, respectively, completing the complete transformation from the second data to the third data.

[0081] This invention proposes a triple dynamic reliability joint weighting mechanism, overcoming the shortcomings of traditional fixed weights. It integrates three heterogeneous factors—equipment hardware precision, real-time communication link quality, and data latency attenuation—to construct a unified reliability evaluation system, effectively suppressing marine noise, faulty and delayed data, and improving the accuracy of multi-source data fusion. Simultaneously, it constructs a spatiotemporal joint Gaussian attenuation fusion model, breaking the limitations of processing in a single spatial or temporal dimension. Relying on spatiotemporal two-dimensional weight constraints, it accurately maps sparse discrete measurement point data into a global continuous gridded field, adapting to the non-uniform distribution characteristics of marine measurement points.

[0082] Based on the above method embodiment, another system embodiment is provided; see [link to system embodiment]. Figure 3 This is a schematic diagram of a data transmission system based on a marine multi-element environmental monitoring network provided in one embodiment of the present invention. Another embodiment of the present invention provides a data transmission system 200 based on a marine multi-element environmental monitoring network. The system is used to execute the data transmission method based on a marine multi-element environmental monitoring network provided in the above embodiment, including an underwater sensing layer subsystem 2001, a sea surface relay subsystem 2002, and a shore-based control subsystem 2003. The underwater sensing layer subsystem is used for multi-element marine environment full-domain acquisition, local edge preprocessing and caching, underwater short-range node adaptive Mesh self-organizing network data transmission, and on-site anomaly autonomous alarm; the underwater sensing layer subsystem includes a multi-element integrated sensor group 20011, a fixed underwater mooring node 20012, and an underwater mobile monitoring node 20013. Both the fixed underwater mooring node and the underwater mobile monitoring node are equipped with an underwater acoustic and optical dual-communication adaptive switching communication module; The surface relay subsystem is used for cross-medium signal docking, multi-source heterogeneous data preprocessing and aggregation, dual-link adaptive cross-domain intelligent scheduling, and bidirectional relay and local fault-tolerant caching. The surface relay layer uses surface buoys as carriers. The surface relay subsystem includes a main control processing module 20021, an underwater cross-medium communication module 20022, a dual-link long-distance communication module 20023, a multi-parameter parallel sensing module 20024, and a local cache storage module 2005. The surface relay subsystem includes a main control processing module 20021 for coordinating the scheduling of the entire subsystem, completing the preprocessing and aggregation of the underwater raw first data, generating standardized second data, and controlling the coordinated operation of all other submodules; an underwater cross-medium communication module 20022 for interfacing with the underwater sensing layer's acoustic and optical dual communication signals, receiving the first data uploaded by underwater nodes, and realizing bidirectional transmission and reception and signal adaptation of underwater-surface cross-medium signals; a dual-link long-distance communication module 20023 for carrying a dual-link adaptive cross-domain communication algorithm, adaptively scheduling and transmitting the second data back to the shore-based control layer through primary and backup dual links, ensuring stable long-distance cross-domain transmission; a multi-parameter parallel sensing module 20024 for synchronously collecting various environmental parameters such as sea surface hydrology, meteorology, channel, and buoy operating conditions, supplementing and improving multi-dimensional observation data of marine monitoring; and a local cache storage module 2005 for implementing data fault tolerance caching, temporarily storing data to be transmitted when the link is interrupted, resuming transmission after communication is restored, and retaining historical monitoring data for local verification.

[0083] The shore-based control subsystem is used for unified fusion processing of multi-source heterogeneous data, intelligent scheduling and communication strategy management across the entire network, full-domain visual control, global network self-healing decision-making, business applications, and alarms. The shore-based control subsystem includes a multi-mode satellite communication access module 20031, a central data processing server 20032, a large-capacity distributed storage unit 20033, a human-machine interface terminal 20034, an underwater node operation and maintenance management module 20035, a marine data intelligent analysis module 20036, a full-domain network scheduling module 20037, and a disaster early warning decision-making module 20038. The multi-mode satellite communication access module is used to connect to the dual-link backhaul data of the sea surface relay layer, enabling multi-standard satellite signal compatibility access and completing two-way communication between shore-based and offshore buoys. The central data processing server is used to carry multi-source heterogeneous data fusion and online adaptive parameter optimization algorithms to generate gridded third fused data from the second data backhauled from the sea surface. The large-capacity distributed storage unit is used for hierarchical storage of raw observation data and fused product data. The system includes equipment operation and maintenance logs and historical sea state archives, supporting efficient reading, writing, and retrieval of massive amounts of marine data; a human-machine interface providing a visual operation interface to display all marine elements, node online status, and communication link status, supporting manual issuance of scheduling and control commands; an underwater node operation and maintenance management module 20035 for real-time monitoring of underwater moorings, mobile node power, equipment faults, and communication connectivity, issuing parameter adjustment, wake-up, and hibernation operation and maintenance control commands; a marine data intelligent analysis module 20036 for performing time-series statistics and spatial feature analysis on gridded fused marine elements, extracting environmental change patterns such as ocean currents, temperature, and salinity; a global network scheduling module 20037 for coordinating underwater Mesh networking and surface dual-link communication resources, adaptively issuing communication strategies based on fusion errors, and achieving self-healing regulation of the entire network; and a disaster early warning decision module 20038 for judging the risk level of marine disasters such as red tides and storm surges based on real-time marine monitoring data, automatically outputting early warning information and generating emergency response decision plans.

[0084] The first data is acquired through the underwater sensing layer subsystem, and the first optimal transmission link is selected according to the underwater acoustic-optical dual-communication adaptive switching algorithm. The first data is then transmitted to the sea surface relay subsystem through the first optimal transmission link. The first data is the raw observation data of multi-source marine environment, channel and equipment operating conditions. The underwater acoustic-optical dual-communication adaptive switching algorithm is constructed by establishing an underwater channel switching model based on the Markov decision process and iteratively solving the underwater channel switching model using a lightweight Q-Learning sub-algorithm. After processing the first data into second data, the surface relay subsystem uses a dual-link adaptive cross-domain communication algorithm to transmit the second data to the shore-based control subsystem. The shore-based control subsystem then uses a multi-source heterogeneous data fusion algorithm to process the second data into third data. The dual-link adaptive cross-domain communication algorithm is used for dynamic switching between dual links to adapt to sea area transmission and to stably transmit marine monitoring data across domains. The multi-source heterogeneous data fusion algorithm is used to couple spatiotemporal and reliability weights and to fuse the second data in a grid-like manner.

[0085] The data transmission system based on a marine multi-element environmental monitoring network provided in this embodiment of the invention is used to execute the data transmission method based on a marine multi-element environmental monitoring network provided in the above embodiments, and has the same or corresponding technical features and effects. Contents not described in detail in this embodiment can be referred to the above embodiments, and will not be repeated here.

[0086] It should be noted that the terminology used in this invention is for describing specific embodiments only and is not intended to limit the scope of this application. As shown in this specification, unless the context clearly indicates otherwise, words such as "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. The terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, 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, or apparatus.

[0087] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A data transmission method based on a marine multi-element environmental monitoring network, characterized in that, Includes the following steps: The first data is acquired through an underwater sensing layer. A first optimal transmission link is selected based on an underwater acoustic-optical dual-communication adaptive switching algorithm, and the first data is transmitted to the surface relay layer via this first optimal transmission link. The first data consists of raw observation data of multi-source marine environment, channel, and equipment operating conditions. The underwater sensing layer is used for multi-element marine environment full-domain acquisition, local edge preprocessing and caching, underwater short-range node adaptive Mesh self-organizing network data transmission, and on-site anomaly autonomous alarm. The underwater sensing layer includes a multi-element integrated sensor group, seabed fixed mooring nodes, and underwater mobile monitoring nodes. Both the seabed fixed mooring nodes and the underwater mobile monitoring nodes are equipped with underwater acoustic-optical dual-communication adaptive switching communication modules. The underwater acoustic-optical dual-communication adaptive switching algorithm is constructed by establishing an underwater channel switching model based on a Markov decision process and iteratively solving the underwater channel switching model using a lightweight Q-Learning sub-algorithm. After processing the first data into second data, the sea surface relay layer uses a dual-link adaptive cross-domain communication algorithm to transmit the second data to the shore-based control layer, so that the shore-based control layer can use a multi-source heterogeneous data fusion algorithm to process the second data into third data. The dual-link adaptive cross-domain communication algorithm is used for dynamic switching of dual links to adapt to sea area transmission and stable cross-domain transmission of marine monitoring data. The multi-source heterogeneous data fusion algorithm is used to couple spatiotemporal and reliability weights and fused the second data in a grid-like manner. The sea surface relay layer is used for cross-medium signal docking, multi-source heterogeneous data preprocessing, dual-link adaptive cross-domain intelligent scheduling, and bidirectional relay and local fault-tolerant caching. The shore-based control layer is used for multi-source heterogeneous data fusion processing, intelligent scheduling and communication strategy control of the entire network, full-domain visualization control, global network self-healing decision-making, business applications, and alarms.

2. The data transmission method based on marine multi-element environmental monitoring network as described in claim 1, characterized in that, The underwater channel handover model established based on the Markov decision process includes: The optimization objectives of the underwater channel handover model are to maximize transmission throughput, minimize bit error rate and packet loss, reduce link handover power consumption, and reduce local data backlog. The state space of the underwater channel switching model is defined based on hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data. Determine the action space of the underwater channel switching model, which includes pure underwater acoustic communication mode, pure blue-green optical communication mode, and acoustic-optical parallel hybrid transmission mode. Define the state transition probability and reward function of the underwater channel switching model; The underwater channel handover model is iteratively solved using a lightweight Q-Learning sub-algorithm, specifically as follows: A lightweight Q-table adapted to underwater communication handover scenarios is constructed to store the value estimates of each state-action pair of the underwater channel handover model. The initialized Q-table is serialized and written to the non-volatile storage unit of the underwater node to achieve parameter persistence. At the same time, the hyperparameters of the Q-Learning algorithm are configured, including the learning rate, discount factor and exploration rate. use The strategy outputs a pre-selected action based on the current state, and then combines hysteresis constraints to determine whether link switching is allowed, in order to determine the final selected action.

3. The data transmission method based on marine multi-element environmental monitoring network as described in claim 2, characterized in that, The reward function of the underwater channel handover model is: ; in, Represents the current state vector. Represents the state vector at the next moment; Indicates the currently executing switching action. This indicates the historical actions performed in the previous cycle; This represents the positive reward for transmission rate, and its value is a positive number, belonging to the revenue gain term. Indicates hardware power consumption loss. Indicates link reliability loss. This represents the loss from frequent switching, and all three are positive penalty bases. This is the preset weighted penalty coefficient.

4. The data transmission method based on marine multi-element environmental monitoring network as described in claim 3, characterized in that, The step of selecting the first optimal transmission link based on the underwater acoustic-optical dual-communication adaptive switching algorithm includes the following steps: The current status is obtained based on the hydrological status data, water quality status data, meteorological and sea surface status data, dynamic environment data, underwater acoustic channel status data, optical channel status data, and node remaining battery power data in the first data. For the current state, adopt The strategy retrieves the pre-selected actions from the Q-table, including pure underwater acoustic communication mode, pure blue-green light-optical communication mode, and acoustic-optical parallel hybrid transmission mode. Hysteresis-based anti-ping-pong handover logic is used to constrain action handover. When the channel amplitude exceeds the hysteresis threshold, the current action is locked as the selected action and no action handover is performed; when the channel amplitude exceeds the hysteresis threshold, the pre-selected action is set as the selected action. The first optimal transmission link is determined based on the selected action, and the first data is transmitted to the sea surface relay layer via the first optimal transmission link.

5. The data transmission method based on marine multi-element environmental monitoring network as described in claim 4, characterized in that, The dual-link adaptive cross-domain communication algorithm is an intelligent scheduling method for cross-domain transmission in the sea area, based on dual communication links, real-time monitoring of link transmission quality, and adaptive switching of the optimal link according to sea conditions and channel status. The method of transmitting the second data to the shore-based control layer using a dual-link adaptive cross-domain communication algorithm includes the following steps: According to the preset period, the dual-link communication parameters, equipment power parameters, buoy attitude parameters and ocean wave environment parameters are collected in parallel. The collected Beidou Tiantong dual-link communication parameters, equipment power parameters, buoy attitude parameters and ocean wave environment parameters are normalized and cleaned, and abnormal sampled values ​​are removed to generate a standardized input dataset. The standardized input dataset is input into the link comprehensive scoring model to calculate the comprehensive scores of the first satellite link and the second satellite link respectively; The second optimal transmission link is determined based on the comprehensive score and the hierarchical decision rule; the hierarchical decision rule refers to the scheduling logic of dividing priority levels according to data urgency level and determining the transmission link hierarchically. Predict short-term buoy fluctuations using wave fitting models to pinpoint effective communication windows; When the effective communication window period is reached and the link status is normal, a differentiated compression strategy is executed according to the data level of the second data, and the compressed second data is forwarded to the shore-based platform through the second optimal transmission link and satellite domain, so that the shore-based platform can clear its local cache after responding and confirming; when the effective communication window period is reached and a link interruption and signal attenuation failure occur, the DTN storage and carry mechanism is activated to cache all untransmitted second data, and after the link is restored to normal, the breakpoint location is automatically identified, and the untransmitted second data is fragmented and resumed.

6. The data transmission method based on a marine multi-element environmental monitoring network as described in claim 5, characterized in that, The comprehensive scoring model for the link is as follows: ； in, This indicates the overall link score; This represents the normalized signal-to-noise ratio quality score; This represents the normalized value of the remaining battery capacity. This represents the inverse score of time delay normalization. This indicates the normalized score for the business urgency level. This represents the inverse score of the historical packet loss rate. This represents a five-dimensional adaptive weight that satisfies the weight normalization constraint.

7. The data transmission method based on a marine multi-element environmental monitoring network as described in claim 6, characterized in that, The multi-source heterogeneous data fusion algorithm is constructed based on a multi-source heterogeneous original dataset, a unified spatiotemporal benchmark model, a triple dynamic credibility weight model, a spatiotemporal joint distance decay model, a global multi-source data fusion master calculation model, and an online adaptive parameter iterative optimization mechanism. The multi-source heterogeneous raw dataset is a set of structured samples formed after the second data has been purified. Each sample in the multi-source heterogeneous raw dataset corresponds to a single valid observation information of a monitoring node. Each sample includes marine element observation values, spatial location coordinates, local acquisition time of the equipment, real-time link quality score, equipment type identifier, data transmission delay and unique node number. The unified spatiotemporal benchmark model is used for unified calibration of the global spatiotemporal benchmark of the multi-source heterogeneous original dataset; the triple dynamic credibility weight model is used to calculate the dynamic credibility weight of samples in the multi-source heterogeneous original dataset; the spatiotemporal joint distance decay model is used to calculate the spatiotemporal joint decay weight of samples in the multi-source heterogeneous original dataset; and the global multi-source data fusion master calculation model is used to perform global gridded fusion of the multi-source heterogeneous original dataset to generate third data. The online adaptive parameter iterative optimization mechanism calculates the fusion error of the third data based on real-time measured data, and adjusts the spatiotemporal weight control parameters online according to the fusion error to continuously optimize the fusion accuracy of the third data. Finally, the optimized third data is standardized and packaged, and output to the digital twin visualization module, intelligent scheduling algorithm module, and business early warning and data storage module, respectively.

8. The data transmission method based on a marine multi-element environmental monitoring network as described in claim 7, characterized in that, The unified spatiotemporal reference model is used for the unified calibration of the global spatiotemporal reference of the multi-source heterogeneous original dataset. Specifically, it uses BeiDou UTC standard time and CGCS2000 geodetic coordinate system as the only global reference to correct the local clock offset and local coordinate deviation problems of each node in the multi-source heterogeneous original dataset, and constructs a uniform gridded calculation area for the monitored sea area to determine the standard spatiotemporal coordinates of the center point of each grid. The triple dynamic credibility weight model is used to calculate the dynamic credibility weight of samples in the multi-source heterogeneous original dataset. Specifically, for each sample in the multi-source heterogeneous original dataset, a dynamic comprehensive credibility weight is calculated based on three dimensions: inherent device precision, real-time communication link quality, and data transmission latency. The weight of low-precision devices, poor-quality communication link data, and data with time delays is automatically reduced, and failed link data is forcibly removed, thus completing the intelligent screening and credibility classification of the multi-source heterogeneous original dataset. The spatiotemporal joint distance attenuation model is used to calculate the spatiotemporal joint attenuation weight of samples in the multi-source heterogeneous original dataset. Specifically, with the center point of each grid as the reference, the spatial Euclidean distance and time difference between the effective samples in the multi-source heterogeneous original dataset and the grid points are calculated respectively. The spatiotemporal contribution weight of the effective samples is solved by the spatiotemporal joint Gaussian attenuation model, thereby increasing the weight of the effective samples whose spatiotemporal position is closer to the grid and whose collection time is more recent. The full-domain multi-source data fusion master calculation model is used to perform full-domain gridded fusion of the multi-source heterogeneous original dataset to generate third data. Specifically, it fuses sample confidence weights and spatiotemporal decay weights, adopts a normalized weighted fusion algorithm, calculates the element fusion value for each grid point in the entire domain, and uses discrete multi-source heterogeneous original dataset sample interpolation to fill the blank areas of marine observation, eliminating the problems of uneven distribution of data measurement points and local missing data, and generating third data. The third data is standardized gridded fusion data that is continuous, regular, uniform, and without data gaps throughout the domain.

9. The data transmission method based on a marine multi-element environmental monitoring network as described in any one of claims 1 to 8, characterized in that, The main computational model for full-domain multi-source data fusion is as follows: ； in, This represents the output value of the fusion of gridded marine elements across the entire region. This represents the overall weight of credibility. Indicates the first The contribution weight of each sample to the grid point This represents the set of valid samples with non-zero weights within the current spatiotemporal window. Indicates the first Multidimensional ocean element observation vectors for each sample; express The first in One sample.

10. A data transmission system based on a marine multi-element environmental monitoring network, characterized in that, It includes an underwater sensing layer subsystem, a sea surface relay subsystem, and a shore-based control subsystem; The underwater sensing layer subsystem is used for multi-element marine environment full-domain acquisition, local edge preprocessing and caching, underwater short-range node adaptive Mesh self-organizing network data transmission, and on-site anomaly autonomous alarm; The sea surface relay subsystem is used for cross-medium signal docking, multi-source heterogeneous data preprocessing and aggregation, dual-link adaptive cross-domain intelligent scheduling, and bidirectional relay and local fault-tolerant caching. The shore-based control subsystem is used for unified fusion processing of multi-source heterogeneous data, intelligent scheduling and communication strategy control of the entire network, full-domain visual control, global network self-healing decision-making, business applications and alarms. The first data is acquired through the underwater sensing layer subsystem, and the first optimal transmission link is selected according to the underwater acoustic-optical dual-communication adaptive switching algorithm. The first data is then transmitted to the sea surface relay subsystem through the first optimal transmission link. The first data is the raw observation data of multi-source marine environment, channel and equipment operating conditions. The underwater acoustic-optical dual-communication adaptive switching algorithm is constructed by establishing an underwater channel switching model based on the Markov decision process and iteratively solving the underwater channel switching model using a lightweight Q-Learning sub-algorithm. After processing the first data into second data, the surface relay subsystem uses a dual-link adaptive cross-domain communication algorithm to transmit the second data to the shore-based control subsystem. The shore-based control subsystem then uses a multi-source heterogeneous data fusion algorithm to process the second data into third data. The dual-link adaptive cross-domain communication algorithm is used for dynamic switching between dual links to adapt to sea area transmission and to stably transmit marine monitoring data across domains. The multi-source heterogeneous data fusion algorithm is used to couple spatiotemporal and reliability weights and to fuse the second data in a grid-like manner.

Images