A multi-modal snorkeling flag integrated marine environment monitoring system

Abstract

The application discloses a kind of multimode snorkeling mark integrated ocean environment monitoring system, comprising: data acquisition and preprocessing module: real-time acquisition leakage source parameter, environmental parameter and the measured concentration data of gas detector arranged in field;Microfield construction module: single-point weather data is deduced and extended to microclimate field data covering entire three-dimensional space;Intelligent scene matching engine: automatically select and call pre-constructed diffusion model;Core calculation model library: generate preliminary three-dimensional concentration field prediction result;Data assimilation module: receive preliminary three-dimensional concentration field prediction result as prior information, dynamically correct model parameters and output more actual situation after correction three-dimensional concentration field;Dynamic risk representation and visualization module: according to corrected three-dimensional concentration field, real-time calculation and generation risk representation information, and output visual three-dimensional diffusion image to support emergency decision.

Description

Technical Field

[0001] This invention relates to the field of marine environmental monitoring technology, and in particular to an integrated marine environmental monitoring system with multimodal buoys. Background Technology

[0002] The multimodal buoy-moor integrated marine environmental monitoring system integrates multi-source heterogeneous data acquisition modules such as buoys, moorings, and satellite remote sensing, combined with artificial intelligence algorithms and intelligent data processing platforms, to achieve three-dimensional, all-weather, high-precision synchronous monitoring and dynamic analysis of environmental parameters from the sea surface to deep water. Its core purpose is to build an intelligent monitoring network covering the entire chain of "data acquisition-processing-transmission-analysis-application". It not only breaks through the limitations of traditional monitoring methods in terms of spatial coverage density, real-time response speed, and multi-parameter collaborative analysis, but also improves prediction accuracy and anomaly detection capabilities through multimodal data fusion technology, ultimately serving the multi-dimensional needs of marine scientific research, sustainable resource utilization, disaster early warning, and ecological protection.

[0003] Existing technologies typically employ a strict horizontal-to-vertical-to-in-situ cascaded judgment logic, treating horizontal and vertical displacements as mutually exclusive operating modes. In scenarios such as observing ocean fronts, which require simultaneously tracking horizontally moving isotherms and acquiring vertical profile data during their movement, traditional buoys can only choose one of the two. If horizontal tracking is selected, the changes in temperature, salinity, and depth (TDM) in the vertical direction will be lost; if vertical profile is selected, the buoy will deviate from the horizontal movement trajectory of the front. This mutually exclusive mode mechanism results in spatiotemporal fragmentation of observational data, failing to accurately reflect the true three-dimensional structure of ocean processes and causing a lack of crucial scientific data. Therefore, a multimodal buoy-submersible integrated marine environmental monitoring system is proposed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a multimodal buoy integrated marine environmental monitoring system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A multimodal buoy-integrated marine environmental monitoring system includes: Multimodal buoy: Equipped with a sensor array for acquiring marine environmental parameters, the multimodal buoy is configured with multiple sensors; One or more shore-based command and control centers are communicatively connected to the multimodal buoy group. The shore-based command and control centers delineate key observation areas based on remote sensing data or ocean models and issue macroscopic observation task instructions to the multimodal buoy group. Each of the multimodal buoys integrates the following: Group Cooperative Control Module: After receiving the macroscopic observation task, it exchanges status information, including real-time position and battery level, with other multimodal buoys in the neighborhood through low-power communication. Based on the consensus protocol of the market mechanism and the dynamic coverage evaluation model, it autonomously performs decentralized task allocation and path planning so that the entire buoy group can form optimized cooperative coverage of the key observation area. Multimodal motion control module: Decouples the traditionally mutually exclusive horizontal and vertical displacement control logic into action primitives that can be executed in parallel or interleaved. When a composite observation task that requires simultaneous horizontal tracking and vertical profile sampling is received, the multimodal motion control module generates and executes a composite motion trajectory that can take into account both horizontal and vertical displacements based on four-dimensional spatiotemporal coordinate decoupling control. Dynamic priority scheduling module: When multiple observation tasks or motion commands to be executed exist simultaneously inside the multimodal buoy, the module dynamically prioritizes each task according to its task value density and performs preemptive task scheduling, allowing urgent tasks with higher value density to be interrupted or to preempt the execution rights of lower priority tasks. Rapid Response Control Module: Real-time sensing of local marine environmental dynamic data, and based on a situation prediction-based feedforward control method, predicting the buoy position drift trend in the near future. When the prediction result is about to exceed the preset observation tolerance threshold, it triggers pre-action commands in advance, thereby achieving an advanced response to rapidly changing environments such as sudden strong currents.

[0006] The above technical solution further includes: Furthermore, the group collaborative control module includes a task bidding unit and a collaborative coverage optimization unit: When the sensor group of the multimodal buoy detects an abnormal event, the task bidding unit initiates a task bidding within the local buoy group through low-power underwater acoustic communication or radio communication, and introduces a virtual currency / reputation value mechanism as the basis for bidding and task assignment. The buoy with the highest bid or reputation value undertakes the encrypted observation task of the abnormal event. The collaborative coverage optimization unit employs an improved distributed ant colony algorithm, treating each multimodal buoy as an autonomous ant individual. Based on location information obtained from neighboring buoys and the regional pheromone concentration dynamically generated from remote sensing data or historical data, each ant individual autonomously adjusts its desired target observation position. This allows the entire multimodal buoy swarm to maintain a basic coverage density in conventional monitoring areas while automatically forming a higher buoy aggregation density in dynamic key areas with high pheromone concentrations.

[0007] Furthermore, the collaborative coverage optimization unit adopts a maximum-minimum ant colony system as the path planning algorithm to balance the weight between the group exploring new areas and utilizing known high-value areas. It introduces a virtual potential field method as a local collision avoidance and void filling mechanism. When any of the multimodal buoys exits operation due to failure or energy depletion, other multimodal buoys in its neighborhood sense the virtual potential field of the resulting coverage void and autonomously adjust their course and position under the influence of virtual repulsion and attraction forces to fill the coverage void, thus achieving self-healing of the entire buoy group's coverage configuration. The entire process does not require manual intervention from the shore-based command and control center.

[0008] Furthermore, the multimodal motion control module includes a task slicing engine and a priority arbiter: The task slicing engine resolves complex observation tasks, such as ocean front observation, into multiple sub-tasks that can be executed concurrently. Specifically, these include a horizontal drift sub-task representing horizontal tracking of isotherms of specific physical quantities and a vertical profile sub-task representing the collection of temperature, salinity, and depth profiles at fixed time intervals. The priority arbitrator has a built-in model predictive control-based solver, which is used to predict situations in complex dynamic environments such as strong currents where performing only the horizontal tracking subtask would result in missing the optimal window for vertical sampling. It adjusts the motion trajectory of the multimodal buoy into a sawtooth trajectory that integrates horizontal drift and vertical jitter, thereby enabling continuous horizontal tracking of ocean phenomena while also completing periodic sampling in the vertical direction, ensuring the spatiotemporal continuity of the observation data.

[0009] Furthermore, the multimodal motion control module is equipped with an energy optimization unit. When the multimodal buoy performs a horizontal displacement task, the energy optimization unit senses and utilizes the ocean current energy in the current sea area in real time. By adjusting the net buoyancy and attitude of the multimodal buoy, it controls the multimodal buoy to move horizontally with the current while simultaneously performing vertical displacement. This allows the multimodal buoy to complete the composite motion without consuming additional active drive energy, achieving zero-energy composite displacement under specific conditions. This significantly extends the buoy's continuous operating time and endurance in the marine environment.

[0010] Furthermore, the fast response control module includes a situational awareness unit, a lightweight prediction model, and a feedforward decision controller: The situational awareness unit is connected to a miniature acoustic Doppler current profiler mounted on a multimodal buoy, or reads flow field forecast data from a regional ocean dynamic model in real time via a communication link to obtain local flow velocity and direction information with high temporal resolution. The lightweight prediction model is internally deployed with a pre-trained long short-term memory network model. The long short-term memory network model takes the real-time position and measured flow velocity data of the multimodal buoy in the past preset time period as input and outputs the prediction of the position drift trajectory of the multimodal buoy in the future preset time period. The feedforward decision controller receives the prediction results from the lightweight prediction model. When the prediction results show that the multimodal buoy will exceed the preset mission position deviation tolerance threshold in a very short time if no measures are taken, the feedforward decision controller triggers pre-action commands, including adjusting the pitch angle or starting the drainage pump.

[0011] Furthermore, the dynamic priority scheduling module performs preemptive task scheduling, including the following steps: A unified model is performed on all observation tasks to be executed within the multimodal buoy. An attribute set is defined for each task. The attribute set includes at least the data value representing the expected scientific value of the task, the remaining time window before which the task must be completed to be effective, the energy consumption expected to be required to execute the task (i.e., the expected energy cost), and the execution status of the task on the time axis. A ready task queue is initialized, and all tasks to be executed are inserted into the ready task queue according to their initial priority and wait for scheduling. Within each scheduling cycle, the dynamic priority scheduling module traverses each task in the ready task queue and dynamically calculates its current task value density according to a preset value function. The task value density is defined as a nonlinear function of the data value, the remaining time window, and the expected energy consumption cost. This function ensures that the task value density of the task exhibits a nonlinear growth trend as the remaining time window shrinks or the data value increases. The currently executing task is recorded as the current task, and the task with the highest task value density in the ready task queue is recorded as the candidate task. When the candidate task is not the same as the current task, and the task value density of the candidate task is significantly higher than that of the current task, and the difference exceeds a preset preemption threshold, the dynamic priority scheduling module determines that the preemption condition is met and immediately executes the preemptive task scheduling operation. In response to the preemption condition, the dynamic priority scheduling module saves all execution context information of the current task, including the target position, execution time, current depth and attitude data of the current task, marks the status of the current task as suspended and re-inserts it into the ready task queue, the scheduling module releases control over the multimodal buoy motion control system and energy system and hands them all over to the candidate task, updates the status of the candidate task to executing, and starts executing high-priority emergency tasks; Once the candidate tasks, i.e. high-priority emergency tasks, are completed, the dynamic priority scheduling module searches for all tasks in the suspended state from the ready task queue. Based on the length of time the tasks have been suspended and their attributes, it generates a fast return trajectory using a feedback linearization control method. This enables the multimodal buoy to overcome ocean current interference and quickly and accurately return to the target position of the original suspended task, resuming the execution of the interrupted task. Simultaneously, the dynamic priority scheduling module starts a recovery timer. In subsequent buoy operation cycles, by appropriately shortening the duration of the predetermined sleep period, the module gradually compensates for the observation time lost due to task preemption, thereby ensuring that the long-term cumulative completion rate of all suspended tasks meets the initial requirements of the task design.

[0012] A monitoring method for an integrated multimodal buoy marine environmental monitoring system includes the following steps: The shore-based command and control center delineates key areas based on anomalies detected by satellite remote sensing images, and issues instructions to the multimodal buoy cluster in the target sea area to conduct vertically encrypted observations via satellite communication links; After receiving the instruction, multiple multimodal buoys in the cluster autonomously negotiate and reach the target area through the group cooperative control module, and then dive. During the tracking and observation process, the biosensors mounted on the multimodal buoy confirmed the specific depth and thickness of the red tide layer, the chemical sensors simultaneously detected the hypoxia and acidification phenomena in the water column, and the physical sensors recorded the local flow field structure that affected the movement of the red tide mass. After completing the profile observation, the multimodal buoy rises to the surface and transmits the complete, preliminarily fused profile data back to the shore-based command and control center via satellite; The data assimilation module of the shore-based command and control center assimilates real-time observation data into a preset ecodynamic model to predict the spread path and evolution trend of red tides within a certain period of time in the future. The system automatically generates early warning reports containing risk levels based on forecast results and pushes them to relevant marine management departments and aquaculture farmers in affected areas, providing a scientific basis for subsequent prevention and control decisions.

[0013] The present invention has the following beneficial effects: In this invention, the task slicing engine decouples the originally conflicting requirements of tracking isotherms and obtaining vertical profiles into concurrent action primitives. Combined with model predictive control, the buoy can plan a composite motion path with a zigzag trajectory, achieving continuous spatiotemporal observation with uninterrupted horizontal drift and no omissions in vertical sampling. Attached Figure Description

[0014] Figure 1 This is a system block diagram of an integrated multimodal buoy marine environmental monitoring system proposed in this invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figure 1 As shown, this invention is an integrated multimodal buoy marine environmental monitoring system, comprising: Multimodal buoy: Equipped with a sensor array for acquiring marine environmental parameters, the multimodal buoy is configured with multiple sensors; One or more shore-based command and control centers are communicatively connected to the multimodal buoy group. The shore-based command and control centers delineate key observation areas based on remote sensing data or ocean models and issue macroscopic observation task instructions to the multimodal buoy group. Each of the multimodal buoys integrates the following: Group Cooperative Control Module: After receiving the macroscopic observation task, it exchanges status information, including real-time position and battery level, with other multimodal buoys in the neighborhood through low-power communication. Based on the consensus protocol of the market mechanism and the dynamic coverage evaluation model, it autonomously performs decentralized task allocation and path planning so that the entire buoy group can form optimized cooperative coverage of the key observation area. Multimodal motion control module: Decouples the traditionally mutually exclusive horizontal and vertical displacement control logic into action primitives that can be executed in parallel or interleaved. When a composite observation task that requires simultaneous horizontal tracking and vertical profile sampling is received, the multimodal motion control module generates and executes a composite motion trajectory that can take into account both horizontal and vertical displacements based on four-dimensional spatiotemporal coordinate decoupling control. Dynamic priority scheduling module: When multiple observation tasks or motion commands to be executed exist simultaneously inside the multimodal buoy, the module dynamically prioritizes each task according to its task value density and performs preemptive task scheduling, allowing urgent tasks with higher value density to be interrupted or to preempt the execution rights of lower priority tasks. Rapid Response Control Module: Real-time sensing of local marine environmental dynamic data, and based on a situation prediction-based feedforward control method, predicting the buoy position drift trend in the near future. When the prediction result is about to exceed the preset observation tolerance threshold, it triggers pre-action commands in advance, thereby achieving an advanced response to rapidly changing environments such as sudden strong currents.

[0017] In one embodiment, the group collaborative control module includes a task bidding unit and a collaborative coverage optimization unit: When the sensor group of the multimodal buoy detects an abnormal event, the task bidding unit initiates a task bidding within the local buoy group through low-power underwater acoustic communication or radio communication, and introduces a virtual currency / reputation value mechanism as the basis for bidding and task assignment. The buoy with the highest bid or reputation value undertakes the encrypted observation task of the abnormal event. It should be noted that the specific analysis process for the task bidding unit is as follows: When the real-time monitoring data of the sensor group of any of the multimodal buoys exceeds the preset abnormal threshold and is determined to be an abnormal event, the buoy, as the initiating node, broadcasts a task bidding request message containing the event type, geographical location, suggested observation duration and initial bidding price to the local buoy group within the communication range through low-power underwater acoustic communication or radio communication. Other multimodal buoys in the local buoy group that receive the bidding request independently calculate their expected bid for this auction based on their current remaining battery power, relative distance to the buoy's real-time position, current task load, and virtual currency / reputation value accumulated from historical task completion quality. They then reply to the initiating node with a bidding message containing their respective bid and current reputation value. The initiating node collects all bidding messages within the preset bidding time limit, and comprehensively compares the bids and reputation values ​​of each bidder according to the pre-set evaluation rules, and selects the multimodal buoy with the highest comprehensive score as the winner. The reputation value is the core basis for task assignment, allowing buoys with high reputation values ​​to obtain task assignments with lower bids. The initiating node sends a task assignment confirmation message to the winning multimodal buoy and broadcasts the end of the auction and the information of the winner to other participating buoys. The winning buoy then adjusts its own task queue and goes to the area where the abnormal event occurred to perform encrypted observation tasks. After the winning buoy completes the encrypted observation task and sends back high-quality observation data, its reputation value is positively accumulated based on the quality of task completion. At the same time, the initiating node transfers a corresponding amount of virtual currency from its virtual account to the winning buoy based on the final transaction price. The collaborative coverage optimization unit adopts an improved distributed ant colony algorithm, treating each multimodal buoy as an autonomous ant individual. Based on the location information obtained from neighboring buoys and the regional pheromone concentration dynamically generated from remote sensing data or historical data, the ant individual autonomously adjusts the target observation position it wants to reach. This allows the entire multimodal buoy swarm to maintain a basic coverage density in the regular monitoring area while automatically forming a higher buoy aggregation density in the dynamic key area with high pheromone concentration. It should be noted that the specific analysis process for the collaborative coverage optimization unit is as follows: Initialization phase: Each of the multimodal buoys, as an autonomous ant individual, obtains initial map information of the mission area from the shore-based command and control center or through local communication, including the boundary of the regular monitoring area, the initialized regional pheromone concentration field, and the preset basic coverage density requirements; Information interaction and perception stage: Each ant periodically broadcasts its own real-time location information and receives location information from neighboring buoys within the communication range to construct a local neighboring buoy distribution map. At the same time, each ant obtains the pheromone concentration value of its current location grid from remote sensing data sent by the shore-based command and control center or dynamically updated from the locally stored historical database. This concentration value represents the real-time observation value of the area. Target location decision stage: Each individual ant, based on the improved distributed ant colony algorithm state transition rules, comprehensively weighs the following three factors to decide on the next target observation location it hopes to reach: first, the pheromone concentration of its current grid, tending to move towards areas with higher concentrations; second, the relative distance to neighboring buoys, avoiding excessive clustering with neighbors and resulting in spatial redundancy; and third, the distance to the boundary of the regular monitoring area, ensuring that the coverage does not produce large gaps. Autonomous movement and coverage execution phase: Each individual ant autonomously controls its propulsion system to move to the target observation location based on its decision, and continuously collects environmental parameters during the movement to maintain the basic coverage density of the area it is in; Pheromone update and aggregation effect formation stage: When multiple ants discover that the pheromone concentration in a certain area is continuously increasing and forming a dynamic key area during their movement, the ants in that area will reduce the weight of learning from their neighbors and increase the weight of moving to areas with high pheromone concentration, thereby attracting more individuals to that area. This automatically forms a higher buoy aggregation density in the dynamic key area, enabling enhanced observation of the key area. At the same time, the neighbor avoidance rule is automatically adjusted to avoid resource waste caused by excessive aggregation.

[0018] In one embodiment, the collaborative coverage optimization unit employs a maximum-minimum ant colony system as the path planning algorithm to balance the weight between the group exploring new areas and utilizing known high-value areas. It introduces a virtual potential field method as a local collision avoidance and void filling mechanism. When any of the multimodal buoys exits operation due to malfunction or energy depletion, other multimodal buoys in its neighborhood sense the virtual potential field of the resulting coverage void and autonomously adjust their course and position under the influence of virtual repulsion and attraction forces to fill the coverage void, thus achieving self-healing of the entire buoy swarm's coverage configuration. The entire process requires no manual intervention from the shore-based command and control center.

[0019] In one embodiment, the multimodal motion control module includes a task slicing engine and a priority arbitrator: The task slicing engine resolves complex observation tasks, such as ocean front observation, into multiple sub-tasks that can be executed concurrently. Specifically, these include a horizontal drift sub-task representing horizontal tracking of isotherms of specific physical quantities and a vertical profile sub-task representing the collection of temperature, salinity, and depth profiles at fixed time intervals. The priority arbitrator has a built-in model predictive control-based solver, which is used to predict, in complex dynamic environments such as strong currents, that performing only the horizontal tracking subtask would result in missing the optimal window for vertical sampling. It adjusts the motion trajectory of the multimodal buoy into a sawtooth trajectory that integrates horizontal drift and vertical jitter, thereby enabling continuous horizontal tracking of ocean phenomena while also completing periodic sampling in the vertical direction, ensuring the spatiotemporal continuity of the observation data. It should be noted that the specific analysis process for the priority arbitrator is as follows: System state-space modeling and prediction equation construction: The priority arbitrator first establishes the dynamic model of the buoy, decomposing its motion in three-dimensional space into two coupled yet relatively independent control channels: horizontal displacement and vertical displacement. The state variable x(k) of the multimodal buoy at time t is defined as: Where x(k) and y(k) are horizontal geodetic coordinates, z(k) is depth coordinate (vertical direction), u(k) and v(k) are horizontal flow velocities, and w(k) is vertical flow velocity. The control input is defined as follows. for: ,in, (Represents the attitude angle, which affects the horizontal gliding direction). Representing the net buoyancy adjustment, which directly affects the vertical rise and fall speed, a discrete-time prediction model (state equation) is constructed to predict the future. System state within each time domain: , i = 0, 1, ..., ,in, Let A represent the predicted state at time k+i+1, where A is the state transition matrix describing the buoy's kinematic characteristics, and B is the control input matrix. This refers to the external disturbance term, specifically the predicted value of ocean current disturbance obtained through acoustic Doppler current profiler measurements or regional flow field prediction models. Multi-objective cost function design and prediction evaluation: A composite cost function J was designed, which quantifies the originally abstract task objective (horizontal tracking isotherm + vertical profile sampling) into an optimizable mathematical index. ; Among them, the horizontal tracking item: The deviation between the buoy's real-time horizontal position and the expected position of the target ocean phenomenon was quantified, and the weight matrix was used. Ensure continuity of horizontal tracking; vertical sampling urgency term: , This represents the ideal sampling depth that should be achieved at a future time. These are time-varying weighting coefficients, representing the urgency of vertical sampling, and employ an exponential decay function: When the remaining time window is predicted When approaching zero, The increase is so sharp that if the solver predicts that performing only horizontal tracing will lead to a situation at some point in the future... If the cost corresponding to an item is infinitely large (i.e., the sampling window is missed), then it is determined that the trajectory needs to be adjusted, and the energy loss item is: Soft constraints and relaxation factors: Used to handle extreme situations such as strong currents; Sawtooth trajectory generation and optimization solution: After constructing the prediction model and cost function, the solver based on model predictive control solves the following finite-time optimization problem online in each control cycle: It satisfies the following constraints: Dynamic constraints: ; Mechanical execution constraints: Attitude angle limit Buoyancy adjustment rate limit ; Safety collision avoidance constraints: Avoid touching the bottom or leaking water. ; The solver calculates the optimal control sequence. ; When the predictive model indicates that the horizontal tracking deviation is about to increase, the solver prioritizes output adjustment. The control variable performs horizontal drift when an internal timer or external sensor indicates that the vertical sampling time point is approaching, and the cost function contains... When the weight of a term rises rapidly and exceeds the weight of the level term, the solver switches quickly at the next time step, adjusting the output. The control input executes rapid vertical jitter (diving or surfacing). The alternating action of this series of control commands makes the actual navigation trajectory of the multimodal buoy appear as a jagged trajectory with periodic vertical undulations superimposed on a background of continuous horizontal movement. Rolling optimization and feedback correction: After performing the above solution, the priority arbitrator only processes the first control variable in the optimal control sequence. The data is then sent to the actuator. At the next sampling time (k+1), the buoy collects the latest GPS / depth meter data again, updates the state x(k+1), and resolves the above optimization problem.

[0020] In one embodiment, the multimodal motion control module is equipped with an energy optimization unit. When the multimodal buoy performs a horizontal displacement task, the energy optimization unit senses and utilizes the ocean current energy in the current sea area in real time. By adjusting the net buoyancy and attitude of the multimodal buoy, it controls the multimodal buoy to move horizontally with the current while simultaneously performing vertical displacement. This allows the multimodal buoy to complete the composite motion without consuming additional active drive energy, achieving zero-energy composite displacement under specific conditions. This significantly extends the buoy's continuous operating time and endurance in the marine environment.

[0021] In one embodiment, the fast response control module includes a situational awareness unit, a lightweight predictive model, and a feedforward decision controller. The situational awareness unit is connected to a miniature acoustic Doppler current profiler mounted on a multimodal buoy, or reads flow field forecast data from a regional ocean dynamic model in real time via a communication link to obtain local flow velocity and direction information with high temporal resolution. The lightweight prediction model is internally deployed with a pre-trained long short-term memory network model. The long short-term memory network model takes the real-time position and measured flow velocity data of the multimodal buoy in the past preset time period as input and outputs the prediction of the position drift trajectory of the multimodal buoy in the future preset time period. It should be noted that the specific analysis process of the lightweight prediction model is as follows: LSTM time series model architecture: Based on historical indicator data and previous weights, it predicts the indicator weights for the current quarter. Model input: time series data matrix. , including recent Real-time positions and measured current velocities of n multimodal buoys for each quarter; Model output: positions of multimodal buoys within a future preset time period; LSTM core computation: LSTM captures temporal dependencies through a gating mechanism, as shown in the following formula: Forgotten Gate: Deciding which historical information to discard. ; in, For the Sigmoid function, This is the weight matrix. This is the hidden state from the previous moment. For the current input, For the bias term of the forget gate; Input gate: Determine which new information to update: ; ; in, The output of the input gate, Here is the weight matrix of the input gate. For the bias term of the input gate, The candidate cell states are generated using the tanh function, representing new information, and their range is between -1 and 1. To generate the weight matrix for candidate cell states, Bias terms for generating candidate cell states; Cell state update: fusing historical and new information: ; in, This represents the current cell state, and ⊙ represents element-wise multiplication. This represents the cell state at the previous moment; Output gate: Generates the current hidden state: ; ; in, For the output of the output gate, Here is the weight matrix of the output gate. This is the bias term for the output gate. The current hidden state; Weight mapping: mapping hidden states Mapped to a weight vector: , The weight matrix is ​​used to map the hidden states to the weight vectors. This is the bias term that maps the hidden state to the weight vector; The feedforward decision controller receives the prediction results from the lightweight prediction model. When the prediction results show that the multimodal buoy will exceed the preset mission position deviation tolerance threshold in a very short time if no measures are taken, the feedforward decision controller triggers pre-action commands, including adjusting the pitch angle or starting the drainage pump.

[0022] In one embodiment, the dynamic priority scheduling module performs preemptive task scheduling, including the following steps: A unified model is performed on all observation tasks to be executed within the multimodal buoy. An attribute set is defined for each task. The attribute set includes at least the data value representing the expected scientific value of the task, the remaining time window before which the task must be completed to be effective, the energy consumption expected to be required to execute the task (i.e., the expected energy cost), and the execution status of the task on the time axis. A ready task queue is initialized, and all tasks to be executed are inserted into the ready task queue according to their initial priority and wait for scheduling. Within each scheduling cycle, the dynamic priority scheduling module traverses each task in the ready task queue and dynamically calculates its current task value density according to a preset value function. The task value density is defined as a non-linear function of the data value, remaining time window, and expected energy consumption cost. This function ensures that the task value density exhibits a non-linear growth trend as the remaining time window decreases or the data value increases. For any task... Its task value density At time t, it can be defined as: ,in, The value of data, i.e., the basic static value coefficient of the task, for example: , The estimated energy cost is the energy expected to be consumed in performing the task. This is a function of the urgency of the remaining time window. This represents the remaining time window for the task at time t. ,in, It is the attenuation constant. It is the absolute deadline of the task, and it increases as time t approaches the deadline. , →0, →1, as time t moves further away from the deadline, Larger Approaching 0; The currently executing task is recorded as the current task, and the task with the highest task value density in the ready task queue is recorded as the candidate task. When the candidate task is not the same as the current task, and the task value density of the candidate task is significantly higher than that of the current task, and the difference exceeds a preset preemption threshold, the dynamic priority scheduling module determines that the preemption condition is met and immediately executes the preemptive task scheduling operation. In response to the preemption condition, the dynamic priority scheduling module saves all execution context information of the current task, including the target position, execution time, current depth and attitude data of the current task, marks the status of the current task as suspended and re-inserts it into the ready task queue, the scheduling module releases control over the multimodal buoy motion control system and energy system and hands them all over to the candidate task, updates the status of the candidate task to executing, and starts executing high-priority emergency tasks; Once the candidate task, i.e. the high-priority emergency task, is completed, the dynamic priority scheduling module searches for all tasks in the suspended state from the ready task queue. Based on the length of time the task has been suspended and its attributes, it generates a fast return trajectory using a feedback linearization control method. This enables the multimodal buoy to overcome ocean current interference and quickly and accurately return to the target position of the original suspended task, resuming the execution of the interrupted task. Simultaneously, the dynamic priority scheduling module starts a recovery timer. In subsequent buoy operation cycles, by appropriately shortening the duration of the predetermined sleep period, the observation time lost due to task preemption is gradually compensated back, thereby ensuring that the long-term cumulative completion rate of all suspended tasks meets the initial requirements of the task design. Specific implementation methods for generating fast return trajectories using feedback linearization control include: The dynamic priority scheduling module calls the underlying multimodal motion control module, taking the target position of the originally suspended task as the control endpoint, and using the buoy's current real-time position and velocity, as well as the real-time sensed ocean current velocity, as state feedback quantities to construct a nonlinear controller based on a precise linearized model. That is, the buoy's dynamic model underwater can be simplified to a system subjected to an external force (ocean current). The second-order system affected by this. Let the buoy target position be... The real-time position is x(t), and the control input (thrust) is... , represented as ,in It is a nonlinear time-varying disturbance generated by ocean currents, where m is the mass, c is the damping coefficient, and the design control input is... for: ,in, These are ocean current disturbance terms estimated by models or measured by sensors, used to accurately offset actual ocean currents. , This is the acceleration we expect, determined by the positional deviation. , It is a proportionality coefficient. These are the differential coefficients. Substituting into the second-order system, we get If the estimate is accurate, that is ,but This leads to the linear error dynamics equation: ,in It is a positional error; The controller dynamically counteracts the effects of nonlinear disturbances such as ocean currents, generates smooth and time-optimal return path control commands, drives the buoy to move along the path, and automatically switches back to the original suspended task after reaching the target position to continue the unfinished observation work.

[0023] A monitoring method for an integrated multimodal buoy marine environmental monitoring system includes the following steps: The shore-based command and control center delineates key areas based on anomalies detected by satellite remote sensing images, and issues instructions to the multimodal buoy cluster in the target sea area to conduct vertically encrypted observations via satellite communication links; After receiving the instruction, multiple multimodal buoys in the cluster autonomously negotiate and reach the target area through the group cooperative control module, and then dive. During the tracking and observation process, the biosensors mounted on the multimodal buoy confirmed the specific depth and thickness of the red tide layer, the chemical sensors simultaneously detected the hypoxia and acidification phenomena in the water column, and the physical sensors recorded the local flow field structure that affected the movement of the red tide mass. After completing the profile observation, the multimodal buoy rises to the surface and transmits the complete, preliminarily fused profile data back to the shore-based command and control center via satellite; The data assimilation module of the shore-based command and control center assimilates real-time observation data into a preset ecodynamic model to predict the spread path and evolution trend of red tides within a certain period of time in the future. The system automatically generates early warning reports containing risk levels based on forecast results and pushes them to relevant marine management departments and aquaculture farmers in affected areas, providing a scientific basis for subsequent prevention and control decisions.

[0024] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multimodal buoy-integrated marine environmental monitoring system, characterized in that, include: Multimodal buoy: Equipped with a sensor array for acquiring marine environmental parameters; Shore-based command and control center: Delineates key observation areas based on remote sensing data or ocean models, and issues macroscopic observation task instructions to the multimodal buoy group; Each of the multimodal buoys integrates the following: Group collaborative control module: After receiving the macroscopic observation task, it exchanges status information with other multimodal buoys in the neighborhood, and autonomously performs decentralized task allocation and path planning based on the consensus protocol of the market mechanism and the dynamic coverage evaluation model. Multimodal motion control module: When receiving a composite observation task that requires simultaneous horizontal tracking and vertical profile sampling, the multimodal motion control module generates and executes a composite motion trajectory that takes into account both horizontal and vertical displacements based on four-dimensional spatiotemporal coordinate decoupling control. Dynamic priority scheduling module: When multiple observation tasks or motion commands to be executed exist simultaneously within the multimodal buoy, the module dynamically prioritizes each task based on its task value density and performs preemptive task scheduling. Rapid Response Control Module: Real-time sensing of local marine environmental dynamic data, and based on a situation prediction-based feedforward control method, predicting the buoy position drift trend in the near future. When the prediction result is about to exceed the preset observation tolerance threshold, it triggers a pre-action command in advance.

2. The integrated marine environmental monitoring system for multimodal buoys according to claim 1, characterized in that, The group collaborative control module includes a task bidding unit and a collaborative coverage optimization unit: When the sensor group of the multimodal buoy detects an abnormal event, the task bidding unit initiates a task bidding within the local buoy group, and the buoy with the highest bid or reputation value undertakes the encrypted observation task of the abnormal event. The collaborative coverage optimization unit adopts an improved distributed ant colony algorithm, treating each multimodal buoy as an individual ant. Based on the location information obtained from neighboring buoys and the regional pheromone concentration dynamically generated from remote sensing data or historical data, the individual ant autonomously adjusts the target observation position it wants to reach. This allows the entire multimodal buoy swarm to maintain a basic coverage density in the regular monitoring area while automatically forming a high buoy aggregation density in the dynamic key area with high pheromone concentration.

3. The integrated marine environmental monitoring system for multimodal buoys according to claim 2, characterized in that, The collaborative coverage optimization unit uses a maximum-minimum ant colony system as the path planning algorithm and introduces a virtual potential field method as a local collision avoidance and hole filling mechanism. When any of the multimodal buoys stops working due to failure or energy depletion, other multimodal buoys in its neighborhood can autonomously adjust their course and position to fill the coverage hole by sensing the virtual potential field generated by the coverage hole.

4. The integrated marine environmental monitoring system for multimodal buoys according to claim 1, characterized in that, The multimodal motion control module includes a task slicing engine and a priority arbitrator: The task slicing engine parses complex observation tasks into multiple subtasks that can be executed concurrently; The priority arbiter has a built-in model predictive control-based solver, which is used to predict, in complex dynamic environments, that performing only the horizontal tracking subtask would result in missing the optimal window for vertical sampling, and adjust the motion trajectory of the multimodal buoy into a zigzag trajectory that integrates horizontal drift and vertical jitter.

5. The multimodal buoy integrated marine environmental monitoring system according to claim 4, characterized in that, The multimodal motion control module is equipped with an energy optimization unit. When the multimodal buoy performs a horizontal displacement task, the energy optimization unit senses and utilizes the current energy in the current sea area in real time. By adjusting the net buoyancy and attitude of the multimodal buoy, it controls the multimodal buoy to move horizontally with the current while simultaneously performing vertical displacement.

6. The integrated marine environmental monitoring system for multimodal buoys according to claim 1, characterized in that, The fast response control module includes a situational awareness unit, a lightweight prediction model, and a feedforward decision controller. The situational awareness unit acquires local flow velocity and flow direction information; The lightweight prediction model is internally deployed with a pre-trained long short-term memory network model. The long short-term memory network model takes the real-time position and measured flow velocity data of the multimodal buoy in the past preset time period as input and outputs the prediction of the position drift trajectory of the multimodal buoy in the future preset time period. The feedforward decision controller receives the prediction results from the lightweight prediction model. When the prediction results show that the multimodal buoy will exceed the preset mission position deviation tolerance threshold in a very short time if no measures are taken, the feedforward decision controller triggers pre-action commands, including adjusting the pitch angle or starting the drainage pump.

7. The integrated marine environmental monitoring system for multimodal buoys according to claim 1, characterized in that, The dynamic priority scheduling module performs preemptive task scheduling, including the following steps: All observation tasks to be executed inside the multimodal buoy are modeled in a unified manner, the attribute set of each task is defined, a ready task queue is initialized, and all tasks to be executed are inserted into the ready task queue according to their initial priority and wait for scheduling. Within each scheduling cycle, the dynamic priority scheduling module traverses each task in the ready task queue and dynamically calculates its current task value density according to a preset value function. The currently executing task is recorded as the current task, and the task with the highest task value density in the ready task queue is recorded as the candidate task. When the candidate task is not the same as the current task, and the task value density of the candidate task is higher than that of the current task, and the difference exceeds a preset preemption threshold, the dynamic priority scheduling module determines that the preemption condition is met and immediately executes the preemptive task scheduling operation. In response to the preemption condition, the dynamic priority scheduling module saves all execution context information of the current task, marks the status of the current task as suspended and re-inserts it into the ready task queue, the scheduling module releases control over the multimodal buoy motion control system and energy system and hands them all over to the candidate task, updates the status of the candidate task to executing, and starts executing high-priority emergency tasks. Once the candidate task, i.e. the high-priority emergency task, has been completed, the dynamic priority scheduling module searches for all tasks in the suspended state from the ready task queue, and generates a fast return trajectory based on the length of time the task has been suspended and the task attributes, so that the multimodal buoy returns to the target position of the original suspended task and resumes the execution of the interrupted task.

8. The monitoring method of the integrated marine environmental monitoring system for multimodal buoys according to claim 1, characterized in that, Includes the following steps: The shore-based command and control center delineates key areas based on anomalies detected by satellite remote sensing images, and issues instructions to the multimodal buoy cluster in the target sea area to conduct vertically encrypted observations via satellite communication links; After receiving the instruction, the multiple multimodal buoys in the cluster autonomously negotiate and reach the target area through the group cooperative control module, and then dive. During the tracking and observation process, the biosensors mounted on the multimodal buoy confirmed the specific depth and thickness of the red tide layer, the chemical sensors simultaneously detected the hypoxia and acidification phenomena in the water column, and the physical sensors recorded the local flow field structure that affected the movement of the red tide mass. After completing the profile observation, the multimodal buoy rises to the surface and transmits the profile data back to the shore-based command and control center via satellite; The shore-based command and control center assimilates real-time observation data into a preset ecodynamic model to predict the spread path and evolution trend of red tides within a certain period of time in the future. The system automatically generates early warning reports that include risk levels based on forecast results.

Images