A system and method for dynamic collision avoidance and path optimization of coastal intelligent waterway

Abstract

The application discloses a kind of coastal wisdom channel dynamic collision avoidance and path optimization system and method thereof.System includes multi-source perception module, wind-wave-current coupling simulation module, path optimization module, multi-ship cooperative collision avoidance module, shore-ship interaction module and core processing unit.Through multi-source perception module, wind, wave, flow and ship data are collected in real time;Wind-wave-current coupling simulation module realizes dynamic environment accurate prediction using WRF+GAN+WTGAN multi-model fusion technology;Path optimization module generates and corrects optimal route in real time based on improved Dijkstra algorithm combined with dynamic environment parameters;Multi-ship cooperative collision avoidance module generates differentiated collision avoidance instructions based on four-dimensional risk assessment and priority sorting;Shore-ship interaction module realizes low-latency two-way information interaction;Core processing unit realizes computing power optimization and redundant backup.The application solves the problems of route mismatch, collision avoidance lag, large computing power overhead and module fragmentation in the prior art, significantly improves the navigation safety and efficiency of coastal wisdom channel.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary fields of waterway engineering, intelligent management and control of ship traffic, and coupled simulation of marine environment. It is applicable to multi-ship navigation scenarios such as complex open coastal waters and port channels. Specifically, it relates to a system and method for dynamic collision avoidance and path optimization in coastal smart waterways. Background Technology

[0002] As a core hub of water transport, the safety and efficiency of coastal smart waterways directly impact coastal economic development and the progress of shipping. However, coastal waterways are significantly affected by the open marine environment, with strong coupling and dynamic changes in marine elements such as wind, waves, and currents significantly interfering with ship navigation attitude, track, and energy consumption, becoming a key bottleneck restricting the safety and efficiency of waterway navigation.

[0003] Currently, existing coastal shipping route optimization and collision avoidance technologies have the following obvious limitations: 1. Route planning level: Most methods only consider basic geometric factors such as ship position and static channel boundary, without integrating the dynamic hydrological and meteorological influences of wind-wave-current multi-field coupling in real time. This leads to a mismatch between the planned route and the actual complex sea conditions along the coast, which can easily result in problems such as route deviation, increased navigation energy consumption, and reduced navigation efficiency.

[0004] 2. In terms of collision avoidance control, existing collision avoidance technologies mostly rely on single AIS or radar monitoring data, and have not established an accurate predictive model of the ship's navigation status in a wind-wave-current coupled environment. The response to collision avoidance commands is delayed and the prediction deviation is large. Especially in the scenario of mixed navigation of large ships and small and medium-sized ships, problems such as missed or misjudged collision risks and conflicting collision avoidance commands are likely to occur, resulting in insufficient navigation safety assurance capabilities.

[0005] 3. Data and computing power: The uneven deployment of monitoring stations in coastal waters and the scarcity of effective monitoring data in some areas lead to a decrease in simulation and planning accuracy due to insufficient data integrity; existing algorithms have not been optimized for lightweight coastal dynamic scenarios, resulting in high computing power consumption, poor real-time performance, and inability to meet the needs of second-level dynamic management and control of waterways.

[0006] 4. System linkage level: Existing technologies have not formed an integrated linkage mechanism of "accurate environmental prediction → dynamic optimization of navigation track → multi-vehicle collaborative collision avoidance → shore-vehicle closed-loop interaction". The functional modules are independent of each other, data are not shared, and decision-making is fragmented, making it impossible to achieve closed-loop management of the entire process and failing to meet the actual needs of high-quality development of coastal smart waterways.

[0007] In summary, developing an integrated technology that combines wind-wave-current coupling fields, high-precision prediction, low computational overhead, multi-ship collaborative collision avoidance, and dynamic path optimization has become an urgent technical challenge to be solved in the field of coastal smart waterways. Summary of the Invention

[0008] The purpose of this invention is to address the aforementioned deficiencies and shortcomings of existing technologies by providing a system and method for dynamic collision avoidance and path optimization in coastal smart waterways. This aims to solve core technical problems in existing technologies such as mismatch between shipping routes and actual sea conditions, delayed or conflicting collision avoidance commands, high computational overhead, and fragmented system modules. Ultimately, it achieves dynamic optimization of vessel navigation paths, early warning of collision risks, efficient multi-vehicle collaborative collision avoidance, and low-latency synchronization of shore-vehicle information in coastal smart waterways, thereby comprehensively improving waterway navigation safety and efficiency.

[0009] To achieve the above objectives, the present invention provides the following technical solution: A system for dynamic collision avoidance and path optimization in coastal smart waterways, the system comprising: Multi-source sensing module, wind-wave-current coupling simulation module, path optimization module, multi-ship cooperative collision avoidance module, shore-ship interaction module, and core processing unit; The multi-source sensing module is used to collect wind-wave-current environmental data, ship operation data, and waterway basic data in real time. The wind-wave-current coupling simulation module is used to acquire wind-wave-current environmental data, ship operation data and waterway basic data in real time. Based on the deep fusion model of WRF atmospheric model and generative adversarial network, and combined with Wasserstein temporal generative adversarial network for data augmentation, it is used to realize real-time wind-wave-current coupling simulation and accurate environmental prediction for the next 1 to 6 hours. The path optimization module is based on the improved Dijkstra algorithm, introduces a heuristic function that includes distance, environment and risk weights, and constructs a path optimization objective function that can dynamically adjust the weights to generate and correct the dynamic optimal route in real time. The multi-ship collaborative collision avoidance module calculates the comprehensive risk value based on a four-dimensional risk assessment system and the analytic hierarchy process, combines the estimated collision time, classifies the collision risk level, and establishes a ship avoidance priority ranking mechanism to generate differentiated collision avoidance instructions. The shore-to-ship interaction module adopts a fusion communication of 5G edge computing and VHF data link to realize two-way low-latency information interaction between the shore-based control center and the ship navigation terminal. The core processing unit connects to each module via dual links of gigabit Ethernet and 5G wireless communication. It has a built-in computing power optimization unit and a redundant backup module for data aggregation, analysis and processing and unified instruction distribution. The above six modules work together to construct a closed-loop intelligent control system that covers the entire process: "multi-source perception → coupled simulation and environmental prediction → dynamic path optimization → multi-ship collaborative collision avoidance → shore-ship two-way interaction → real-time closed-loop adjustment".

[0010] Furthermore, the multi-source sensing module includes a shore-based X-band radar, a shipborne meteorological instrument, a satellite remote sensing receiving device, a marine monitoring buoy, and a ship's AIS terminal; the multi-source sensing module has a built-in data cleaning unit that uses the 3σ criterion to remove outliers and uses linear interpolation to fill in missing values; and it also has a built-in AES-256 data encryption unit that transmits data through the NETCDF4.0 standard interface.

[0011] Furthermore, the wind-wave-current coupling simulation module uses the finite volume method to calculate the flow field gradient force and the Coriolis force, with a calculation step size of 10 minutes and an overall calculation error ≤3%; in the data augmentation step of the Wasserstein temporal generative adversarial network, the generator and discriminator iterate 1000-2000 times with a learning rate of 0.001, and the similarity between the generated data and the measured data is ≥95%; the environmental prediction is based on an LSTM neural network, with a prediction accuracy ≥85%. The wind-wave-current coupling simulation module includes: The physical model driving unit is used to run the WRF atmospheric model, the SWAN wave model, and the FVCOM ocean model, and output a low-resolution wind-wave-current mesh field. The time series data enhancement unit uses the Wasserstein temporal generative adversarial network to generate synthetic time series data for data-sparse regions. The GAN super-resolution unit uses a conditional generative adversarial network to take the low-resolution grid field as input and output a high-resolution wind-wave-current coupled field. The environmental prediction unit, based on an LSTM neural network, is used to predict environmental changes in the next 1 to 6 hours based on a high-resolution coupled field time series sequence. The output of the physical model driving unit is connected to the input of the GAN super-resolution unit; the output of the temporal data augmentation unit is connected to the input of the physical model driving unit, and is used to supplement the physical model driving unit with synthetic data; the output of the GAN super-resolution unit is connected to the input of the environment prediction unit; the output of the environment prediction unit is also connected to the input of the physical model driving unit, and is used to feed back the prediction result as a dynamic boundary condition to the physical model driving unit.

[0012] Furthermore, the path optimization module includes: The data preprocessing unit is used to perform Kalman filtering repair on historical AIS trajectory data and discretize the waterway area into a raster map. The parallel scheduling unit is used to group multiple ships according to their tonnage and speed, and allocate computing resources accordingly. The route generation unit, based on the improved Dijkstra algorithm, combines the raster map output by the data preprocessing unit with dynamic environmental parameters to generate the initial optimal route. The real-time correction unit acquires the latest environmental data every 5 to 10 minutes, compensates and corrects the initial optimal route, and outputs the final dynamic optimal route. The output of the data preprocessing unit is connected to the input of the route generation unit and the real-time correction unit, respectively; the control terminal of the parallel scheduling unit is connected to the scheduling input of the route generation unit; the output of the route generation unit is connected to the input of the real-time correction unit; and the output of the real-time correction unit is also connected to the trigger terminal of the route generation unit through a feedback loop to form a closed-loop correction.

[0013] Furthermore, in the path optimization module, the heuristic function of the improved Dijkstra algorithm is: h(n) = α × distance weight + β × environment weight + γ × risk weight, where α = 0.3, β = 0.4, and γ = 0.3; the path optimization objective function is: minF(x) = ω1 × distance (x) + ω2 × time (x) + ω3 × energy consumption (x) + ω4 × risk (x), and ω1 + ω2 + ω3 + ω4 = 1. Each weight coefficient is dynamically and adaptively adjusted according to the ship type and sea state level; the algorithm adopts a "grouping in parallel first, then serial" process for optimization, with the grouping based on the ship's tonnage and sailing speed; and a Kalman filter algorithm is used to repair historical AIS trajectory data, with a repair error ≤ 2%; the optimized path is compensated and corrected in real time at a period of 5 to 10 minutes.

[0014] Furthermore, in the multi-vessel collaborative collision avoidance module, the four-dimensional risk assessment system includes four dimensions: water conditions, vessel parameters, environmental factors, and collision distance. The Analytic Hierarchy Process (AHP) is used to determine the weight coefficients of each sub-evaluation indicator. First, each evaluation indicator is dimensionless, and then the comprehensive risk value is calculated by weighted summation of the indicator weights. Combining the comprehensive risk value and the estimated collision time, the collision risk is divided into three levels: Level 1 risk is a collision time ≤ 5 minutes and a comprehensive risk value ≥ 0.8; Level 2 risk is a collision time 5–15 minutes and a comprehensive risk value 0.4–0.8; and Level 3 risk is a collision time > 15 minutes and a comprehensive risk value < 0.4. The vessel avoidance priority is ranked as follows: vessels with emergency risks > large vessels > medium and small vessels > non-navigable vessels. Differentiated collision avoidance commands include course adjustments of 5°–30°, speed adjustments to reduce speed by 10%–50%, and emergency avoidance deviations from the original course ≥ 0.5 nautical miles.

[0015] Furthermore, the data transmission latency of the shore-ship interaction module is ≤1 second; the computing power optimization unit built into the core processing unit adopts an edge computing architecture, with edge node latency ≤50ms, reducing overall computing power overhead by more than 60%, and the overall system response time ≤1 second; the built-in redundant backup module adopts a dual-machine hot standby mode, and the backup server seamlessly switches over and takes over the work within 30 seconds when the main server fails.

[0016] Furthermore, this invention also provides a dynamic collision avoidance and path optimization method for coastal smart waterways that integrates wind-wave-current coupling, implemented based on the above system, and includes the following steps: Step 1: Multi-source data acquisition and standardized preprocessing; Step 2: Wind-Wave-Current Coupling Simulation and Accurate Environmental Prediction; Step 3: Dynamic optimization and real-time correction of the flight path; Step 4: Multi-ship linkage analysis and collaborative collision avoidance decision-making; Step 5: Two-way interaction between shore and ship and closed-loop dynamic adjustment; Step 6: System operation and maintenance and optimization.

[0017] Further, step 1 specifically includes: collecting wind-wave-current environment data, ship operation data and basic channel data of the coastal waterway through the multi-source sensing module; removing outliers using the 3σ criterion and filling in missing values ​​using linear interpolation; and transmitting the data to the core processing unit after formatting according to the NETCDF4.0 standard.

[0018] Furthermore, step 2 specifically includes: the core processing unit inputs standardized environmental data into the wind-wave-current coupling simulation module; real-time coupling calculation is completed through the WRF and GAN fusion model; data augmentation is performed on data-scarce areas using Wasserstein temporal generative adversarial network technology to generate high-quality temporal synthetic data; and dynamic environmental parameters and extreme environment early warning information are output based on LSTM neural network to accurately predict environmental changes in the next 1 to 6 hours.

[0019] Furthermore, step 3 specifically includes: the path optimization module calls the improved Dijkstra algorithm, introduces a heuristic function that includes distance, environment and risk weights, and combines ship operation data, waterway basic data and dynamic environmental parameters to construct a path optimization objective function with dynamically adjustable weights to generate an initial optimal route; based on the AIS trajectory data repaired by Kalman filtering, the route compensation correction is completed at a period of 5 to 10 minutes, and the final dynamic optimal route is output.

[0020] Further, step 4 specifically includes: the collaborative collision avoidance module analyzes the navigation status of multiple vessels at a frequency of 30 seconds / time, and combines the environmental prediction results with the four-dimensional risk assessment system and the analytic hierarchy process, specifically as follows: 1) Layered construction of the evaluation system: the top layer is the comprehensive risk target layer for ship collisions, the middle criterion layer is divided into four categories: water conditions, ship parameters, environmental factors, and others (such as collision distance), and the bottom indicator layer is the specific quantitative parameters corresponding to each dimension; 2) Construction of pairwise comparison matrices: based on the experience of the waterway navigation industry and measured data, the influence of indicators at the same level on ship collisions is compared, and scaled values ​​are assigned to form a judgment matrix; 3) Weight solution: the eigenvectors of the judgment matrix are solved and... 4) Consistency check: Perform consistency check on the judgment matrix. If the check is successful, determine the final weight of the indicator. If the consistency is not satisfied, readjust the matrix scale and repeat the calculation until the check is successful. 5) Calculate the comprehensive risk value by weighted summation using the final weight of the indicator. Combine the comprehensive risk value and the estimated collision time as two parameters to perform a collision and complete the risk classification assessment. Based on the priority of avoiding collisions of emergency risk vessels > large vessels > medium and small vessels > non-navigable vessels, generate differentiated collision avoidance instructions including course adjustment, speed adjustment and emergency avoidance, and issue them to the corresponding vessels. At the same time, detect and avoid conflicts between multiple collision avoidance instructions.

[0021] Furthermore, step 5 specifically includes: the shore-to-ship interaction module uses a 5G edge computing and VHF data link converged communication method to push the dynamic optimal route, collision risk warning and differentiated collision avoidance instructions to the ship navigation terminal with low latency, and simultaneously receives real-time navigation status feedback data of the ship, with data transmission latency ≤1 second; the core processing unit adjusts the route and collision avoidance strategy in real time according to the feedback data and real-time environmental changes, forming a closed-loop management of the entire process.

[0022] Furthermore, step 6 specifically includes: regularly calibrating and maintaining the equipment of the multi-source sensing module; dynamically optimizing the weight coefficients in the wind-wave-current coupling simulation model, the path optimization algorithm, and the risk assessment model parameters in the multi-ship cooperative collision avoidance module based on ship navigation feedback and changes in the navigation water environment.

[0023] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention breaks through the technical bottleneck of existing technologies that do not integrate wind-wave-current coupling fields. It is the first to create a multi-model fusion technology of WRF+GAN+WTGAN, which realizes real-time simulation and accurate medium- and long-term prediction of wind-wave-current multi-field coupling. At the same time, it combines the improved Dijkstra algorithm to realize efficient optimization of the trajectory, effectively solving the problems of unreasonable path optimization, collision avoidance lag and high computing power consumption in existing technologies. It fills the technical gap in the field of integrated intelligent management and control under complex coastal sea conditions and has significant creativity.

[0024] 2. The system of the present invention relies on existing mature hardware such as shore-based monitoring, AIS, and 5G edge computing, without the need for large-scale modification of existing navigation facilities, resulting in low implementation costs and a short implementation cycle. The method steps are specific and operable, and can be directly promoted and applied in typical coastal waterways such as the Bohai Bay and Zhoushan Islands. It is adaptable to the navigation needs of different types of ships and different sea conditions, and can effectively improve the safety and efficiency of waterway navigation, with broad application prospects.

[0025] 3. Data level: WTGAN data augmentation technology effectively solves the problem of scarce monitoring data and significantly improves data reliability; Computing power level: The combination of edge computing and lightweight algorithms reduces computing power consumption by 60% and response time is ≤1 second; Efficiency level: Path planning efficiency is improved by more than 30%; Collision avoidance level: A collision risk classification assessment system and a multi-ship collaborative collision avoidance mechanism have been built, improving collision avoidance accuracy by 40%; Energy consumption level: The optimal route can reduce ship navigation energy consumption by 15% and shorten navigation time by 12%.

[0026] 4. This invention realizes a closed-loop management of the entire process of "environmental prediction - path optimization - collision avoidance control - shore-ship interaction". It is not only applicable to conventional scenarios such as open coastal waters and port channels, but also adaptable to ship navigation and safety assurance in extreme environments such as typhoons and freezing. It can provide core technical support for the construction of coastal smart waterways and shipping safety management. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall architecture of the coastal smart waterway dynamic collision avoidance and path optimization system that integrates wind-wave-current coupling as described in this invention.

[0028] Figure 2 This is a logical schematic diagram of the wind-wave-current coupling simulation module in this invention.

[0029] Figure 3 This is a flowchart illustrating the path optimization module in this invention.

[0030] Figure 4 This is a logical schematic diagram of the multi-ship cooperative collision avoidance module in this invention. Detailed Implementation

[0031] To make the technical solutions, advantages, and features of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0032] Example 1: Please read Figures 1-4This embodiment provides a system for dynamic collision avoidance and path optimization in coastal smart waterways, with the Bohai Bay coastal waterway as a pilot area for application verification. This area is a typical open sea waterway with significant wind-wave-current coupling characteristics and dense ship traffic, encompassing various ship types such as large cargo ships and medium and small passenger ships, making it suitable for application verification of this invention.

[0033] Multi-source sensing equipment was deployed in the pilot waterway, including: 3 shore-based X-band radars (model: JMA-9200, detection range 0.1~60km, azimuth accuracy ±0.1°), covering the entire length of the waterway; 10 FB-100 marine buoys (endurance ≥365 days, data reporting frequency 1 time / minute), deployed at key nodes in the waterway; 5 satellite remote sensing receiving devices (preferably Gaofen-6 satellite, spatial resolution 10m); and shipborne meteorological instruments (model: WS500, wind speed 0~60m / s, wave height 0~20m, current speed 0~5m / s, accuracy ±1%) and AIS terminals (compliant with the SOLAS Convention, positioning accuracy ±1m) for 20 vessels in the pilot area.

[0034] The shore-based control center deploys a core processing server (Dell PowerEdge R750, CPU: Intel Xeon Gold5318Y, Memory: 64GB, Hard Drive: 2TB SSD) to build a visual control platform and construct a wind-wave-current coupled simulation model and path optimization algorithm model. Two-way interconnection and communication between the shore-based system and the ship is achieved through a 5G edge computing network (base station coverage radius 1.5km) and a VHF data link (frequency 156-162MHz).

[0035] The multi-source sensing module collects the following three types of data in real time: (1) Wind-wave-current environmental data: wind speed (0~30m / s), wind direction (0~360°), wave height (0~5m), wave direction, current speed (0~2m / s), current direction; (2) Ship operation data: position (accuracy ±1m), heading (0~360°), speed over land (0~20 knots), draft (2~15m), tonnage (500~150000 tons), and ship type (cargo ship, passenger ship, oil tanker, fishing boat) of 20 ships; (3) Basic data of the waterway: multibeam bathymetry data (depth accuracy ±0.1m, grid resolution 50m×50m), waterway boundary (electronic chart S-57 format), distribution of obstructions (shipwrecks, reefs, aquaculture areas).

[0036] The data cleaning unit employs the 3σ criterion to remove outliers—for example, if the wind speed data reported by a buoy suddenly jumps to 100 m / s, the system determines it as an outlier and removes it. For data with no more than five consecutive missing time points, linear interpolation is used to complete the data, with the interpolation error controlled to within 5% through cross-validation. The cleaned data is formatted according to the NETCDF 4.0 standard, encrypted with AES-256, and transmitted to the core processing unit via gigabit fiber optic cable, forming a high-quality standardized dataset.

[0037] like Figure 1 As shown, the core processing unit inputs the preprocessed environmental data into the wind-wave-current coupling simulation module, and performs the following operations: (1) Initialization: The initial field is the temperature-salinity-topography field of Bohai Bay (from the reanalysis data of the HYCOM model, with a spatial resolution of 1 / 12°).

[0038] (2) Atmospheric drive: The WRF v4.5 atmospheric data assimilation model uses the NCEP GFS 0.25° forecast field as the boundary condition, runs a triple nested grid (innermost layer resolution 1.5km), and outputs the sea surface 10m wind speed and wind direction.

[0039] (3) Wave simulation: The third-generation wave model SWAN is driven by the wind speed output of WRF to calculate the significant wave height, average wave direction and spectral peak period.

[0040] (4) Ocean current calculation: Based on the FVCOM ocean model, the three-dimensional Reynolds-averaged Navier-Stokes equations are solved using the finite volume method, taking into account wind stress, tides and baroclinic gradient forces. The calculation step is 10 minutes, and the surface and stratified flow velocities and directions are output.

[0041] (5) GAN fusion: The simulation results of WRF, SWAN and FVCOM (time resolution of 1 hour) are used as input to the GAN model to train a generator composed of convolutional and deconvolutional layers, and output high-resolution (5 minutes) wind-wave-current coupling field; the discriminator is used to judge the consistency between the simulation results and the measured data, and iterates 2000 times until the discrimination accuracy is lower than 60%, and the overall calculation error is ≤3%.

[0042] For areas with sparse buoy deployment (such as 15 nautical miles southeast of the middle of the channel where no measured data is available), Wasserstein Temporal Generative Adversarial Network (WTGAN) was used for data augmentation. Specifically, the sparse observation sequence of the past 72 hours was input, the generator (based on a Temporal Convolutional Network TCN) outputs synthetic data for the next 24 hours, and the discriminator (based on a Recurrent Neural Network RNN) underwent adversarial training. The generator and discriminator iterated 1000 times with a learning rate of 0.001, and the similarity between the generated data and the real observations (verified through temporary encrypted observations) reached 96.2%.

[0043] The environmental prediction unit constructs a prediction model based on an LSTM neural network. The model structure is a two-layer LSTM with 128 units per layer, a dropout rate of 0.2, and uses the Adam optimizer. The wind-wave-current coupled time series (10-minute time step) of the past 6 hours is input into the model, and the predicted values ​​for the next 1-6 hours are output, with a prediction accuracy of ≥85%. During a winter cold wave, the system predicted 3 hours in advance that the wind speed would suddenly increase from 8 m / s to 18 m / s and the wave height from 1.2 m to 3.5 m, automatically triggering a red extreme environment warning (the warning threshold is set at wind speed ≥15 m / s or wave height ≥3 m), providing sufficient response time for ships to enter port for shelter.

[0044] In this embodiment, the path optimization module is further divided into four sub-units: The data preprocessing unit integrates ship operation data, waterway basic data, and dynamic environmental parameters from the core processing unit. It uses the Kalman filter algorithm (parameters: process noise covariance Q=0.01, measurement noise covariance R=0.1) to repair historical AIS trajectories with a repair error ≤2%. At the same time, the waterway area is discretized into a 100m×100m grid map, with each grid including water depth, obstacles, and real-time wind-wave-current attributes.

[0045] The parallel scheduling unit divides the 20 ships into "large ship group", "medium ship group", "small ship group" and "dangerous goods ship group" according to the ship tonnage and sailing speed. Each group is assigned an independent CPU core, and the processing within the group follows the strategy of "first grouping in parallel, then serial".

[0046] For each group, the route generation unit calls the improved Dijkstra algorithm in parallel, taking the raster map output by the data preprocessing unit and dynamic environmental parameters as input, to calculate the heuristic function h(n) = 0.3 × distance + 0.4 × environmental resistance + 0.3 × risk value, and solve the objective function minF(x) = 0.2 × distance + 0.2 × time + 0.3 × energy consumption + 0.3 × risk, and outputs the initial optimal route for each ship.

[0047] Every 10 minutes, the real-time correction unit obtains the latest environmental prediction data and the ship's real-time position from the data preprocessing unit and performs local replanning of the initial route being executed. When anomalies such as a sudden increase in wave height are detected, route correction is automatically triggered, and the corrected route is pushed to the ship terminal through the shore-to-ship interaction module. At the same time, if the correction magnitude exceeds a threshold (e.g., offset > 1 nautical mile), the real-time correction unit notifies the route generation unit through a feedback loop to perform global replanning of the remaining voyage.

[0048] The aforementioned units work together to achieve efficient parallelism and dynamic adaptive correction in path planning.

[0049] like Figure 3 As shown, the path optimization module calls the improved Dijkstra algorithm. Taking the navigation mission of a 50,000-ton bulk carrier (225m in length, 10m in draft, and 12 knots) from Tianjin Port (38°59'N, 117°43'E) to Qinhuangdao Port (39°56'N, 119°36'E) as an example, the specific execution process of the algorithm is explained.

[0050] (1) Channel gridding: The channel area is discretized into a 100m×100m grid map. Each grid contains water depth, obstacle attributes and real-time wind-wave-flow field parameters (wind speed, flow direction, wave height) provided by the coupled simulation module.

[0051] (2) Objective function weight setting: Based on the ship type (bulk carrier) and the current sea state (wind speed 12m / s, wave height 2m, current speed 0.8m / s), an adaptive weight adjustment strategy is adopted, and the weight coefficients in the objective function are set as follows: distance weight ω1=0.2, time weight ω2=0.2, energy consumption weight ω3=0.3, and risk weight ω4=0.3. Among them, energy consumption is estimated based on the ship resistance model, and resistance is proportional to the square of the speed, wave height, and current speed; risk is calculated by normalization based on the ratio of wave height in the grid to the ship's seakeeping limit.

[0052] (3) Heuristic function calculation: h(n) = 0.3 × Euclidean distance + 0.4 × average environmental resistance of the path + 0.3 × maximum risk value of the path.

[0053] (4) Parallel computing: The 50,000-ton ships are grouped into the “large ships group” and the planning tasks of the other three ships in the group are processed in parallel. Each group is assigned an independent CPU core to execute the Dijkstra algorithm.

[0054] (5) Initial route generation: The algorithm outputs the initial optimal route, with a total distance of 145 nautical miles and an estimated sailing time of 12.5 hours.

[0055] (6) Real-time correction: Every 10 minutes, the remaining voyage is replanned using the historical AIS trajectory repaired by Kalman filtering (filter parameters: process noise covariance Q=0.01, measurement noise covariance R=0.1) and the latest environmental prediction data. When the ship sails to the Laotieshan Channel (where wind-wave-current coupling is strong), the system detects a sudden increase in wave height to 3.2m, immediately triggering a route correction to guide the ship to deviate 1.2 nautical miles southwest to avoid the high wave area. The final actual voyage distance is 147 nautical miles, and the sailing time is 12.1 hours. Compared with the traditional shortest path route (without considering environmental factors), the sailing time is reduced by 12%, fuel consumption is reduced from 24 tons to 20.4 tons (a reduction of 15%), and the maximum roll angle of the ship is reduced from 15° to 8°, significantly improving sailing comfort and safety.

[0056] Comparative experiments show that under the same sea conditions, the traditional Dijkstra algorithm (without environmental weights, parallel computing, and real-time correction) takes 12.6 seconds to plan, while the improved algorithm of this invention takes only 8.3 seconds, improving planning efficiency by about 34%.

[0057] The collaborative collision avoidance module monitors the navigation status of 20 vessels within the pilot area at a frequency of 30 seconds per instance. At a certain moment, the following scenario occurs in the channel intersection area (38°42'N, 118°15'E): Vessel A is a large bulk carrier (80,000 tons, speed 10 knots, heading 090°), and Vessel B is a medium-sized passenger ship (500 tons, speed 18 knots, heading 180°). The initial closest encounter distance (DCPA) is 0.3 nautical miles, and the minimum encounter time (TCPA) is 4.5 minutes.

[0058] (1) Risk assessment: such as Figure 4 As shown, the system calculates based on a four-dimensional risk assessment framework—water conditions: channel width 500 meters, water depth 15 meters; vessel parameters: vessel A draft 11 meters, vessel B draft 3 meters; environmental factors: current wind speed 15 m / s (southeast wind), wave height 2.5 m, current speed 1.0 m / s (southwest current); collision distance: DCPA = 0.3 nautical miles, TCPA = 4.5 minutes. The Analytic Hierarchy Process (AHP) is used to determine the weights of each indicator. After a consistency test (CR = 0.04 < 0.1), the calculated comprehensive risk value is 0.85 ≥ 0.8, and TCPA = 4.5 minutes ≤ 5 minutes, classifying it as a Level 1 risk (emergency collision).

[0059] (2) Priority ranking for collision avoidance: According to the priority ranking mechanism of “emergency risk vessel > large vessel > medium and small vessel > non-navigable vessel”, both vessel A (large vessel) and vessel B (medium and small vessel) are emergency risk vessels and need to take collision avoidance actions simultaneously.

[0060] (3) Generation of differentiated collision avoidance commands: The system generates the following differentiated collision avoidance commands based on the wind-wave-current coupling prediction results: For ship A: "Immediately adjust the course to the right by 20° to 110° and reduce the speed by 20% to 8 knots"; For ship B: "Immediately adjust the course to the left by 15° to 165° and reduce the speed by 30% to 12.6 knots". When generating the above commands, the system also considers the current southwest current and current speed of 1.0 m / s, and pre-adds a 2° current compensation angle to the course adjustment of ship A to counteract the track drift caused by the water flow.

[0061] (4) Command conflict detection: The system synchronously checks the collision avoidance commands of other vessels (C, D, E...) in the area to ensure that there will be no command conflict, such as vessel A turning right and vessel C turning left, which would lead to a new collision point. In this scenario, there is no command conflict. The command is sent to the corresponding vessel through the 5G network with an end-to-end delay of 0.7 seconds.

[0062] (5) Execution result: After ships A and B sailed at the new course and speed, DCPA increased to 1.2 nautical miles and TCPA became infinite, successfully avoiding a collision. Post-event simulation verification showed that without the intervention of the system of this invention, the two ships would have collided after 4 minutes and 20 seconds.

[0063] The shore-to-ship interaction module uses 5G edge computing technology (deployed at three base stations along the waterway, each equipped with an MEC server) to push dynamically optimized routes, collision risk warnings, and differentiated collision avoidance commands to the ship's navigation terminal (a 10.1-inch rugged tablet running Android). Actual transmission latency: 0.6 seconds for P95, maximum latency 0.9 seconds, meeting the design requirement of ≤1 second.

[0064] The ship's terminal provides real-time feedback on navigation status data (position, heading, speed, rudder angle, main engine speed) at a frequency of once per second. Based on the feedback data and the latest environmental predictions—such as a sudden gust of wind increasing from 15 m / s to 18 m / s—the core processing unit immediately adjusts the collision avoidance strategy and issues a fine-tuning instruction to ship A: "Due to the strengthening gust, it is recommended to reduce speed by another 5% to 7.6 knots and increase the port rudder by 5° to resist lateral drift."

[0065] The shore-based visualization and control platform, developed based on WebGIS, overlays and displays the following information in real time: real-time wind field (vector arrows), wave height (color-coded map), flow field (streamlines), real-time positions and predicted trajectories of 20 vessels (dashed lines), optimized routes (solid blue lines), risk warning areas (flashing red borders), and collision avoidance instructions (pop-up windows). On-duty personnel can click on any vessel through the platform to view detailed information and send voice or text commands for manual intervention, achieving closed-loop control of the entire process between shore-based operations and vessels.

[0066] To ensure the long-term stable operation of the system, this embodiment establishes a comprehensive operation, maintenance, and optimization mechanism: Equipment maintenance: Weekly angle calibration of 3 shore-based radars (using GPS timing and known azimuth calibration), and cleaning of biological deposits and sensor zero point and range calibration of 10 buoys.

[0067] Model optimization: Each month, based on the navigation feedback data and measured data from the past 30 days, the WTGAN and LSTM prediction models are retrained using an incremental learning method (retaining 80% of the old model weights); the weight matrix for AHP risk assessment is optimized using a genetic algorithm; the default values ​​of the heuristic function weights in the Dijkstra algorithm are adjusted and improved according to seasonal sea state changes—for example, during the summer peak tourist season, when passenger ship traffic increases, the risk weight γ is increased from 0.3 to 0.35.

[0068] Data Updates: The basic geographic database is updated quarterly based on changes such as channel dredging and the installation of new navigation marks.

[0069] After six months of continuous operation, the system successfully issued warnings in 15 out of 16 potential collision events, achieving a collision avoidance success rate of 93.75%; the average response time was 0.85 seconds; average ship energy consumption decreased by 14.2%, and sailing time was shortened by 11.8%; the user satisfaction survey score (out of 10) was 9.2. These results demonstrate that the present invention has good practicality and scalability in the field of coastal smart waterways.

[0070] Example 2: This example is basically the same as Example 1, the main difference being the system application performance under extreme weather scenarios during the passage of Typhoon Haishen through the Bohai Bay.

[0071] Scenario: The typhoon's center is located 80 nautical miles southeast of the channel entrance, with a central pressure of 950 hPa, maximum wind speed of 40 m / s, and a moving speed of 15 km / h. It is expected to make landfall in 12 hours. Currently, several ships are still entering or leaving ports within the channel, and environmental conditions are extremely harsh.

[0072] System response and characteristics: (1) Data augmentation: Due to the approaching typhoon, some marine buoys were damaged or had abnormal data. WTGAN generated synthetic environmental data for the next 12 hours in the waterway based on typhoon path forecast data and the wind-wave-current response model during historical typhoon passage. The simulation results showed that the maximum wind speed in the waterway could reach 35 m / s and the maximum wave height could reach 8 m, providing key data support for subsequent decision-making.

[0073] (2) Environmental prediction: The LSTM prediction model outputs a red extreme warning (triggering conditions: wind speed > 25m / s or wave height > 5m), and predicts that there is a "sheltered window area" with relatively low wave height (3-4m) on the west side of the channel entrance, which provides an important basis for ship avoidance.

[0074] (3) Route optimization: The system replans the optimal route for all vessels stranded in the channel, guiding them to the sheltered window area or nearby sheltered anchorage. The risk weight ω4 in the objective function is dynamically adjusted to 0.6, and the distance weight ω1 is reduced to 0.1, fully reflecting the "safety first" extreme situation control strategy.

[0075] (4) Collision avoidance: The system automatically switches to "Typhoon Emergency Collision Avoidance Mode". In this mode, the first-level risk threshold of all ships is reduced from 0.8 to 0.6, and the avoidance priority is temporarily adjusted to "small ships take priority over large ships" (based on the safety consideration that small ships are more likely to capsize in extreme sea conditions). At the same time, all ships are required to reduce their speed to below 5 knots and maintain a safe distance of more than 5 nautical miles.

[0076] (5) Shore-to-ship interaction: The system broadcasts typhoon dynamics and the latest emergency instructions to all vessels via the VHF forced broadcast system, and pushes detailed shelter paths and anchorage locations via 4G or 5G networks. The shore-based visualization platform activates the emergency command interface, integrating meteorological monitoring, marine forecasts and search and rescue resource information to provide comprehensive decision support for emergency command.

[0077] Performance Verification: In the simulated application of this embodiment, all 12 vessels safely entered the sheltered area 3 hours before the typhoon made landfall, with no collisions or groundings occurring, successfully averting a major shipping disaster during the typhoon's passage. This embodiment fully demonstrates the reliability and emergency response capabilities of the present invention under extremely severe sea conditions.

[0078] Example 3: This example is basically the same as Example 1, the main difference being that the scenario is set as an important waterway in the Zhoushan Islands. The waterway has an extremely high ship density (50 ships pass through per hour) and frequent convergence of multiple ships, which places higher demands on the system's real-time performance and collaborative capabilities.

[0079] System optimization and feature performance: (1) Parallel computing extension: The path optimization module divides the ships into four parallel computing groups: "large merchant ship group", "medium cargo ship group", "small passenger and fishing ship group" and "dangerous goods ship group". Each group is assigned a dedicated computing node. The process strategy of "grouping in parallel first and then serial" is adopted within the group. The routes are coordinated between groups through message queues, which effectively avoids route conflicts between ships of different groups.

[0080] (2) Collision Avoidance Module Upgrade: Based on the original Analytic Hierarchy Process (AHP) risk assessment model, a multi-agent cooperative collision avoidance algorithm based on Deep Q-Network (DQN) is further introduced. Each ship in the channel is regarded as an agent, with its state space including its own motion state and environmental parameters, and its action space consisting of adjustments to its heading and speed. The reward function is set as a weighted sum of negative collision risk values ​​and energy consumption values. The cooperative collision avoidance strategy is obtained through offline training (using historical AIS data and a high-precision simulation environment, training for 1 million steps). When deployed online, multiple ships can collaboratively decide on a globally conflict-free collision avoidance strategy within 0.5 seconds.

[0081] (3) Edge computing deployment: MEC servers are deployed at shore base stations on both sides of the waterway. Each MEC server is responsible for ship collision avoidance calculation tasks within its coverage radius of 5 nautical miles. This distributed architecture significantly reduces the load and communication latency of the core processing server, and the local decision latency is reduced to less than 20ms.

[0082] Performance Verification: In a simulated stress test with a 200% increase in ship density compared to traditional periods, the average response time of the system in this embodiment remained ≤0.6 seconds, and the collision avoidance accuracy remained above 90%. In contrast, the traditional centralized architecture solution had a response time exceeding 2 seconds and a collision avoidance accuracy below 60% under the same ship density. This embodiment fully demonstrates the dynamic control capability of the present invention in high-density navigation scenarios, successfully achieving the technical goal of "smooth passage without congestion, and high speed without collisions."

[0083] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A system for dynamic collision avoidance and path optimization in coastal smart waterways, characterized in that, The system includes: Multi-source sensing module, wind-wave-current coupling simulation module, path optimization module, multi-ship cooperative collision avoidance module, shore-ship interaction module, and core processing unit; The multi-source sensing module is used to collect wind-wave-current environmental data, ship operation data, and waterway basic data in real time. The input end of the wind-wave-current coupling simulation module is connected to the output end of the multi-source sensing module to acquire the data collected by the multi-source sensing module in real time. Based on the deep fusion model of WRF atmospheric model and generative adversarial network, and combined with Wasserstein temporal generative adversarial network for data augmentation, the module realizes real-time wind-wave-current coupling simulation and accurate environmental prediction for the next 1 to 6 hours. The input of the path optimization module is connected to the output of the multi-source sensing module and the wind-wave-current coupling simulation module, respectively. It is used to introduce a heuristic function containing distance, environment and risk weights based on the improved Dijkstra algorithm, and to construct a path optimization objective function with dynamically adjustable weights to generate and correct the dynamic optimal route in real time. The input of the multi-ship collaborative collision avoidance module is connected to the output of the multi-source sensing module and the wind-wave-current coupling simulation module, respectively. It is used to classify the collision risk level based on the four-dimensional risk assessment system and the hierarchical analysis method, establish a ship avoidance priority ranking mechanism, and generate differentiated collision avoidance instructions. The input end of the shore-to-ship interaction module is connected to the output end of the path optimization module and the multi-ship collaborative collision avoidance module, respectively, and is used to realize two-way low-latency information interaction between the shore-based control center and the ship navigation terminal by using 5G edge computing and VHF data link fusion communication. The core processing unit connects to each module via dual links of gigabit Ethernet and 5G wireless communication. It has a built-in computing power optimization unit and a redundant backup module for data aggregation, analysis and processing and unified instruction distribution. The output of the shore-ship interaction module is also connected to the input of the core processing unit, which is used to transmit the feedback data from the ship terminal back to the core processing unit. The core processing unit adjusts the operating parameters of the wind-wave-current coupling simulation module, the path optimization module, and the multi-ship cooperative collision avoidance module according to the feedback data, forming a closed-loop control of the entire process.

2. The system according to claim 1, characterized in that, The multi-source sensing module includes a shore-based X-band radar, a shipborne meteorological instrument, a satellite remote sensing receiving device, a marine monitoring buoy, and a ship's AIS terminal. The multi-source sensing module has a built-in data cleaning unit that uses the 3σ criterion to remove outliers and linear interpolation to fill in missing values. It also has a built-in AES-256 data encryption unit that transmits data through the NETCDF4.0 standard interface.

3. The system according to claim 1, characterized in that, The wind-wave-current coupling simulation module includes: The physical model driving unit is used to run the WRF atmospheric model, the SWAN wave model, and the FVCOM ocean model, and output a low-resolution wind-wave-current mesh field. The time series data enhancement unit uses the Wasserstein temporal generative adversarial network to generate synthetic time series data for data-sparse regions. The GAN super-resolution unit uses a conditional generative adversarial network to take the low-resolution grid field as input and output a high-resolution wind-wave-current coupled field. The environmental prediction unit, based on an LSTM neural network, is used to predict environmental changes in the next 1 to 6 hours based on a high-resolution coupled field time series sequence. The output of the physical model driving unit is connected to the input of the GAN super-resolution unit; the output of the temporal data augmentation unit is connected to the input of the physical model driving unit, and is used to supplement the physical model driving unit with synthetic data; the output of the GAN super-resolution unit is connected to the input of the environment prediction unit; the output of the environment prediction unit is also connected to the input of the physical model driving unit, and is used to feed back the prediction result as a dynamic boundary condition to the physical model driving unit.

4. The system according to claim 1, characterized in that, The path optimization module includes: The data preprocessing unit is used to perform Kalman filtering repair on historical AIS trajectory data and discretize the waterway area into a raster map. The parallel scheduling unit is used to group multiple ships according to their tonnage and speed, and allocate computing resources accordingly. The route generation unit, based on the improved Dijkstra algorithm, combines the raster map output by the data preprocessing unit with dynamic environmental parameters to generate the initial optimal route. The real-time correction unit acquires the latest environmental data at a preset frequency, compensates and corrects the initial optimal route, and outputs the final dynamic optimal route. The output of the data preprocessing unit is connected to the input of the route generation unit and the real-time correction unit, respectively; the control terminal of the parallel scheduling unit is connected to the scheduling input of the route generation unit; the output of the route generation unit is connected to the input of the real-time correction unit; and the output of the real-time correction unit is also connected to the trigger terminal of the route generation unit through a feedback loop to form a closed-loop correction.

5. The system according to claim 1 or 4, characterized in that, In the path optimization module, the heuristic function of the improved Dijkstra algorithm is: h(n) = α × distance weight + β × environment weight + γ × risk weight, where α = 0.3, β = 0.4, and γ = 0.3; the path optimization objective function is: minF(x) = ω1 × distance (x) + ω2 × time (x) + ω3 × energy consumption (x) + ω4 × risk (x), and ω1 + ω2 + ω3 + ω4 = 1. Each weight coefficient is dynamically and adaptively adjusted according to the ship type and sea state level. The path optimization module adopts a "grouping in parallel first, then serial" optimization process, with grouping based on ship tonnage and sailing speed. It also uses a Kalman filter algorithm to repair historical AIS trajectory data, with a repair error ≤ 2%. The optimized path is compensated and corrected in real time at a period of 5 to 10 minutes.

6. The system according to claim 1, characterized in that, In the multi-vehicle collaborative collision avoidance module, the four-dimensional risk assessment system includes four dimensions: water conditions, vessel parameters, environmental factors, and collision distance. The weight coefficients of each sub-evaluation indicator are determined using the analytic hierarchy process (AHP). First, each evaluation indicator is dimensionless, and then the comprehensive risk value is calculated by weighted summation of the indicator weights. Combining the comprehensive risk value and the estimated collision time, the collision risk is divided into three levels: Level 1 risk is a collision time ≤ 5 minutes and a comprehensive risk value ≥ 0.8; Level 2 risk is a collision time 5–15 minutes and a comprehensive risk value 0.4–0.8; and Level 3 risk is a collision time > 15 minutes and a comprehensive risk value < 0.

4. The vessel avoidance priority is ranked as follows: vessels with emergency risks > large vessels > medium and small vessels > non-navigable vessels. Differentiated collision avoidance commands include course adjustment, speed adjustment, and emergency avoidance deviation from the original course.

7. The system according to claim 1, characterized in that, The data transmission latency of the shore-ship interaction module is ≤1 second; the computing power optimization unit built into the core processing unit adopts an edge computing architecture with an edge node latency of ≤50ms; the built-in redundant backup module adopts a dual-machine hot standby mode, and the backup server seamlessly switches over and takes over the work within 30 seconds when the main server fails.

8. A method for dynamic collision avoidance and path optimization in coastal smart waterways, implemented based on the system described in any one of claims 1 to 7, characterized in that, Includes the following steps: Step 1: Multi-source data acquisition and standardized preprocessing; Step 2: Wind-Wave-Current Coupling Simulation and Accurate Environmental Prediction; Step 3: Dynamic optimization and real-time correction of the flight path; Step 4: Multi-ship linkage analysis and collaborative collision avoidance decision-making; Step 5: Two-way interaction between shore and ship and closed-loop dynamic adjustment; Step 6: System operation and maintenance and optimization.

9. The method according to claim 8, characterized in that, Step 1 specifically includes: collecting wind-wave-current environment data, ship operation data, and basic waterway data of the coastal waterway through a multi-source sensing module; removing outliers using the 3σ criterion and filling in missing values ​​using linear interpolation; and transmitting the data to the core processing unit after formatting according to the NETCDF4.0 standard.

10. The method according to claim 8, characterized in that, Step 2 specifically includes: the core processing unit inputs standardized environmental data into the wind-wave-current coupling simulation module; the physical model driving unit in the wind-wave-current coupling simulation module runs WRF, SWAN, and FVCOM models to output a low-resolution grid field; the time series data enhancement unit generates synthetic data to supplement the physical model driving unit; the GAN super-resolution unit reconstructs the low-resolution field into a high-resolution coupling field; and the environmental prediction unit predicts environmental changes in the next 1 to 6 hours based on the high-resolution time series, outputting dynamic environmental parameters and extreme environment early warning information.

11. The method according to claim 8, characterized in that, Step 3 specifically includes: the path optimization module calls the improved Dijkstra algorithm, introduces a heuristic function that includes distance, environment and risk weights, and combines ship operation data, waterway basic data and dynamic environmental parameters to construct a path optimization objective function with dynamically adjustable weights to generate the initial optimal route; based on the AIS trajectory data repaired by Kalman filtering, the route compensation correction is completed at a period of 5 to 10 minutes, and the final dynamic optimal route is output.

12. The method according to claim 8, characterized in that, In step 4, the collaborative collision avoidance module analyzes the navigation status of multiple vessels at a preset frequency, and combines the environmental prediction results with the four-dimensional risk assessment system and the analytic hierarchy process, as detailed below: The evaluation system is constructed in layers: the top layer is the comprehensive risk target layer for ship collisions, the middle criterion layer is divided into four categories: water conditions, ship parameters, environmental factors, and others, and the bottom indicator layer consists of the specific quantitative parameters corresponding to each dimension. Construct a pairwise comparison matrix: Based on the experience and measured data of the waterway navigation industry, benchmark the impact of indicators of the same level on ship collisions, and assign scale values ​​to form a judgment matrix; Weight calculation: Solve for the eigenvectors of the judgment matrix and normalize them to obtain the initial weights of each evaluation index; Consistency check: Perform consistency verification on the judgment matrix. If the verification is successful, determine the final weight of the indicator; if the consistency is not satisfied, readjust the matrix scale and repeat the calculation until the verification is successful. The comprehensive risk value is calculated by weighting and summing the final weights of the indicators; the collision risk is graded and assessed by combining the comprehensive risk value and the estimated collision time; based on the avoidance priority ranking of "emergency risk vessels > large vessels > medium and small vessels > non-navigable vessels", differentiated collision avoidance instructions including course adjustment, speed adjustment and emergency avoidance are generated and issued to the corresponding vessels, while detecting and avoiding conflicts between multiple vessels' collision avoidance instructions.

13. The method according to claim 8, characterized in that, Step 5 specifically includes: the shore-to-ship interaction module uses a converged communication method of 5G edge computing and VHF data link to push dynamic optimal routes, collision risk warnings and differentiated collision avoidance instructions to the ship navigation terminal with low latency, and simultaneously receives real-time navigation status feedback data of the ship; the core processing unit adjusts the route and collision avoidance strategy in real time according to the feedback data and real-time environmental changes, forming a closed-loop management of the entire process.

14. The method according to claim 8, characterized in that, Step 6 specifically includes: regularly calibrating and maintaining the equipment of the multi-source sensing module; dynamically optimizing the weight coefficients in the wind-wave-current coupling simulation model, the path optimization algorithm, and the risk assessment model parameters in the multi-ship cooperative collision avoidance module based on ship navigation feedback and changes in the navigation water environment.

Images