Mine area ship route planning method

Abstract

The application discloses a mine area ship route planning method and belongs to the field of ship route planning. The method comprises the following steps: establishing a physical field characteristic model of a ship; constructing a first risk field model containing a predicted position sequence and a fuse trigger rule for each proven mine; constructing a navigable grid map; performing global path search based on the physical field characteristic model and the first risk field model to minimize the comprehensive navigation cost, and generating an initial global route; detecting a newly appeared mine in real time during navigation, predicting a movement trajectory of the mine, and constructing a second risk field model containing a predicted position sequence and a fuse trigger rule of the mine; performing space-time correlation analysis on the initial global route and the second risk field model to mark a dangerous navigation section; and performing path re-planning to obtain a local avoidance path. The application realizes accurate evaluation and dynamic avoidance of navigation risks in a mine area, and improves the safety and task completion capability of a ship during navigation in a mine area.

Description

Technical Field

[0001] This invention relates to the field of ship route planning. More specifically, this invention relates to a method for ship route planning in minefields. Background Technology

[0002] Maritime salvage and rescue operations are characterized by high risk and difficulty, often requiring rescue forces to operate in complex sea conditions and dangerous environments. When a distressed vessel sinks or runs aground in waters potentially containing mines, the safety of subsequent salvage vessels faces severe challenges. Unlike military vessels, salvage vessels cannot deploy specialized minesweeping teams to clear mines beforehand; therefore, they must rely on precise route planning to avoid threats and venture deep into minefields to conduct rescue operations, based on information about mine distribution.

[0003] In actual salvage operations, equipment on salvage vessels such as side-scan sonar, multibeam echo sounders, and magnetometers are already capable of effectively detecting and identifying suspicious underwater targets, allowing salvage vessels to obtain approximate mine distribution information before operations begin. However, mine clearance requires the intervention of specialized mine countermeasures forces, which involves a long processing cycle and high operational risks. Meanwhile, the location of distressed vessels or the drifting area of ​​distressed personnel often overlaps with minefields. Salvage vessels cannot completely avoid dangerous waters by making significant detours and must venture deep into these areas to conduct effective rescue operations, placing higher demands on the route planning methods of salvage vessels.

[0004] Existing ship route planning methods, when dealing with mine threats, typically simplify mines into geometric obstacles with a fixed danger radius. For example, the method proposed by Zhang Qi et al. in "Research on Ship Avoidance Areas Based on Mine Countermeasures Navigation Operations" calculates the radius of the mine's danger circle and designs avoidance routes and areas based on the ship's maneuverability. This method treats a mine as a static circular danger zone; once a ship enters this radius, it is considered to have triggered a mine, and avoidance paths are designed based on tangential geometry. This simplification has some practicality in mine countermeasures navigation operations. However, mine triggering is not simply affected by the blast radius, but depends on whether the ship's physical field characteristics (magnetic field, acoustic field, water pressure field) reach the set threshold of the mine fuse. The actual threat range of the same mine may be completely different for salvage vessels of different types and tonnages. Existing methods fail to reflect this personalized threat relationship of "mine-ship coupling." For dynamic moving targets such as drifting mines, the current general approach is to delineate an expanded area based on experience, resulting in low initial prediction accuracy and failing to meet the high-precision planning requirements of salvage operations.

[0005] In summary, existing methods for ship route planning in minefields suffer from the following technical shortcomings: First, mine threat modeling is overly simplistic, treating mines as geometric circles with fixed radii and failing to consider the coupling effect between mine fuses and the ship's physical field; second, prediction of dynamic drifting mines is insufficient; and third, avoidance decisions only consider geometric safety, lacking utilization of the ship's physical field characteristics. Therefore, there is an urgent need for a ship route planning method in minefields that integrates mine fuse characteristics, ship physical field characteristics, and hydrodynamic conditions to improve the safety of salvage and rescue forces navigating in minefields. Summary of the Invention

[0006] This invention provides a method for ship route planning in minefield areas. It performs global path planning by constructing a multi-physics risk field coupled with the physical field characteristics of the ship, detects newly appearing mines in real time during navigation and predicts their trajectory, constructs a local dynamic risk field to identify dangerous sections, and performs local replanning with mine avoidance rules as constraints to generate an updated safe route.

[0007] To achieve these objectives and other advantages according to the present invention, the present invention provides a method for planning ship routes in minefields, comprising the following steps: S1. Establish a physical field characteristic model of the ship. The physical field characteristic model includes at least the spatial distribution of magnetic field, sound field and water pressure field at a preset speed. The physical field characteristic model is stored in the ship's own coordinate system and is rotated and transformed according to the ship's course when used. S2. Construct a first risk field model for all known mines. The first risk field model includes the predicted location sequence of each known mine during the ship's navigation period and the fuse triggering rules. S3. Construct a navigable grid map of the target sea area, build a comprehensive navigation cost function based on the physical field feature model and the first risk field model, and use a heuristic optimization algorithm to search for the path that minimizes the comprehensive navigation cost function, and generate an initial global route from the route start point to the route end point. S4. During the ship's navigation along the initial global route, the information of newly appearing mines in the target sea area is detected and updated in real time. The predicted location sequence of the newly appearing mines in the subsequent navigation period is predicted by combining real-time wind and sea state data. A second risk field model is constructed, which includes the predicted location sequence of each newly appearing mine in the ship's navigation period and the fuse triggering rules. S5. Perform a spatiotemporal correlation analysis between the initial global route and the second risk field model. Combine the predicted position sequence and fuse triggering rules in the physical field characteristic model and the second risk field model to calculate the mine-trapping risk cost of each segment in the initial global route at the ship's expected arrival time. Segments with mine-trapping risk costs exceeding the risk threshold are recorded as dangerous segments. S6. Based on the dangerous route segment, expand the area to form a local replanning area. Perform route replanning within the local replanning area to obtain a local avoidance route. Replace the initial global route within the local replanning area with the local avoidance route.

[0008] Preferably, in step S2, the historical position sequence of the known mines is obtained. For drifting mines, the predicted position sequence during the ship's navigation period is predicted based on their historical position sequence. For anchored mines and bottom mines, their fixed positions are regarded as their predicted position sequences. The predicted position sequences of all known mines and their corresponding fuse triggering rules are integrated to construct the first risk field model.

[0009] Preferably, for drifting mines with existing historical location sequences, a pre-trained PSO-Attention-LSTM model is used to obtain the predicted location sequence. The PSO-Attention-LSTM model includes: The Long Short-Term Memory (LSTM) network layer is used to receive the historical location sequence of drifting mines, sea surface wind field data, and surface current data, extract the time-dependent features of drift motion, and output the hidden state at each time step. The attention mechanism layer, connected to the long short-term memory network layer, is used to weight the hidden states at each time step and assign higher attention weights to key time steps. The particle swarm optimization module is used to globally optimize the hyperparameters of the long short-term memory network layer and the attention mechanism layer. The output layer is used to output the predicted location sequence of the drifting mines.

[0010] Preferably, step S3 includes the following steps: S31. On the navigable grid map, define a set of nodes and a set of analysis edges. Each analysis edge corresponds to a flight segment formed by connecting two adjacent nodes. S32. The comprehensive travel cost of each analysis edge is obtained through weighted calculation, wherein the comprehensive travel cost includes: a) The time cost of navigation is determined based on the geometric length of the analysis edge and the preset speed; b) The cost of mine strike risk, determined based on the physical field characteristic model, the predicted location sequence of mines, and the fuse triggering rules, represents the maximum risk of a ship striking a mine when it passes through the analysis edge at a preset speed at a specific time; S33. Ant colony optimization is used to find the path. Heuristic information is determined based on the comprehensive travel cost. The next node is selected based on the pheromone concentration of each analysis edge traversed by each ant and the heuristic information, until the destination is reached. After one iteration, the pheromone concentration of each analysis edge is updated based on the total comprehensive travel cost of the complete path traversed by each ant, so that the path with the smaller total comprehensive travel cost receives a higher pheromone increment. The iteration is repeated until the termination condition is met, and the path with the smallest total comprehensive travel cost is output as the initial global route.

[0011] Preferably, the formula for calculating the comprehensive navigation cost is as follows: in, i and j To analyze the node index of the edge, d ij To analyze the geometric length of the side, L The reference length used for normalizing the distance cost. v ij To analyze the preset speed on the side, R ij To analyze the risks and costs of stepping on landmines on the side, M ij (t) For ships in analysis edge ij The moment t A group of mines in which the distance between the ship and the mine is less than the mine's preset radius of influence. m This is the index for the mine's serial number; For distance penalty terms, the value is 0 if the ship is at least 1 away from any mine on the analysis side; otherwise, the value is 1. As a speed penalty term, it is set to 0 if the ship's speed on the analysis edge is not greater than the maximum safe speed, and 1 otherwise. β , γ These are the weighting coefficients for each rule's penalty item. w d 、w t 、w r 、w p These are the weighting coefficients for each cost item.

[0012] Preferably, the formula for calculating the cost of lightning strike risk is as follows: in, F m (t) For ships at time t This analysis examines the physical field intensity generated at the location of the mine. Tm This is the fuse triggering threshold for a naval mine. a This is for the safety factor.

[0013] Preferably, in step S4, the wind-induced drift coefficient and lateral offset probability corresponding to the newly appearing mine are matched from the mine hydrodynamic parameter library, and the predicted location sequence of the newly appearing mine is obtained by using a drift prediction model based on physical mechanism in combination with real-time sea surface wind field data and surface current data.

[0014] Preferably, step S5 includes: S51. Discretize the initial global route into several segments, each segment corresponding to the ship's expected arrival time, preset speed and expected course; S52. Combine the physical field characteristic model and the second risk field model to calculate the risk cost of a ship hitting a mine when passing through; S53. If the risk of striking a mine exceeds the preset risk threshold, the flight segment is determined to be covered by the second risk field model and marked as a dangerous flight segment.

[0015] Preferably, obtaining the local avoidance path includes the following steps: A1. Based on the dangerous flight segment, a preset buffer length is extended along both ends of the flight segment as the longitudinal boundary of the replanning area, and a preset buffer width is extended along both sides of the flight segment as the lateral boundary of the replanning area, forming a local replanning area that includes the dangerous flight segment and its surrounding area. A2. Discretize the local replanning region into a local grid map, which consists of multiple nodes and analysis edges connecting adjacent nodes, with the grid spacing being smaller than that of the global grid map. A3. Spatiotemporally overlay the portions of the first risk field model and the second risk field model located within the local replanning region to form a comprehensive risk field within that region; A4. Within the local grid map, taking the boundary points where ships enter and leave the local replanning area as the local start point and local end point, the K-short-circuit algorithm is used for path search. During the search process, the total comprehensive navigation cost on each candidate path is calculated, and the paths are sorted from smallest to largest according to the total comprehensive navigation cost. The top n paths are selected as the candidate path set. A5. If the candidate path set is not empty, then candidate paths are selected in order of increasing total comprehensive navigation cost for ship turnaround feasibility verification, and the first candidate path that passes the verification is taken as the local avoidance path; if the candidate path set is empty, or all candidate paths fail the turnaround feasibility verification, then the buffer length and buffer width are increased, and steps A2 to A4 are repeated.

[0016] Preferably, step S6 is followed by: S7. Based on the time offset caused by sailing along the local avoidance path, determine the updated estimated arrival time of the ship at the local endpoint; S8. Using the updated estimated arrival time as the time base, perform spatiotemporal correlation analysis on the remaining initial global route between the local endpoint and the route endpoint with the first risk field model and the second risk field model to verify whether a new dangerous route segment is formed. If it is verified that there are no new dangerous segments, the local avoidance path in step S6 is directly spliced ​​with the remaining initial global route as the secondary planning route. If a new dangerous segment is verified, the local endpoint is used as the new route starting point, and the updated estimated arrival time is used as the starting time. Step S3 is performed on the remaining voyage after the local avoidance path to obtain the adjusted route. The initial route segment that has not yet been traversed before the local avoidance path, the local avoidance path, and the adjusted route are sequentially spliced ​​together to form the secondary planned route.

[0017] The present invention has at least the following beneficial effects: First, this invention constructs a multi-physics coupled risk field that couples the physical field characteristics of the ship, linking the characteristics of the mine fuse with the ship's own magnetic field, acoustic field, and water pressure field. This overcomes the limitations of traditional methods that simplify mines into geometric circles with a fixed radius. This technique can dynamically assess the actual threat range of mines based on the actual physical field characteristics of salvage vessels of different types and tonnages, significantly improving the accuracy and personalization of risk modeling, and making the planned routes more aligned with actual safety requirements.

[0018] Secondly, this invention addresses the problem of traditional time-series prediction models failing to function effectively when data accumulation is insufficient by detecting newly discovered mines in real time during navigation and predicting their trajectories using real-time sea state data. This technique fully utilizes prior knowledge such as the wind-induced drift coefficient and lateral offset probability corresponding to different mine types, enabling relatively accurate initial predictions even for newly discovered drifting mines with only one location, providing a reliable basis for subsequent avoidance decisions.

[0019] Third, this invention uses mine avoidance rules as constraints for local replanning, integrating avoidance requirements into the path search process, overcoming the shortcomings of existing methods that only consider geometric avoidance. By identifying dangerous sections through spatiotemporal correlation analysis and dynamically adjusting the planning area, it can identify the threat of newly emerging mines in real time, generating updated routes that meet both safety requirements and operational specifications, thus improving the navigation safety and mission completion capabilities of salvage vessels in complex minefield environments.

[0020] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating one technical solution of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0023] It should be understood that terms such as “having,” “including,” and “comprising” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0024] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0025] It should be noted that, unless otherwise specified, the experimental methods described in the following embodiments are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified. In the description of this invention, the orientation or positional relationship indicated by the terms is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this invention and simplifying the description. It does not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0026] like Figure 1 As shown, the present invention provides a method for planning ship routes in minefields, comprising the following steps: S1. Establish a physical field characteristic model of the ship. The physical field characteristic model includes at least the spatial distribution of magnetic field, sound field and water pressure field at a preset speed. The physical field characteristic model is stored in the ship's own coordinate system and is rotated and transformed according to the ship's course when used. Modern naval mines primarily employ three triggering methods: magnetic fuses, acoustic fuses, and hydraulic fuses. These triggers detect magnetic field anomalies, radiated noise, and hydraulic pressure changes caused by a ship's passage, respectively. Therefore, to accurately assess the risk of a ship striking a mine while navigating in a minefield, it is necessary to understand the ship's characteristic data across these three physical dimensions.

[0027] The spatial distribution of the magnetic field refers to the variation of the magnetic field strength around a ship with its position. As a ferromagnetic object, a ship generates an additional magnetic field after being magnetized in the Earth's magnetic field. The strength and direction of this magnetic field vary with distance and azimuth. Typically, the magnetic field is stronger at the bow and stern and weaker on the sides, and it decreases cubically with increasing distance. The spatial distribution of the sound field refers to the variation of the noise level radiated by a ship during navigation with its position. This mainly originates from propeller cavitation noise, mechanical vibration, and hydrodynamic noise. Its intensity exhibits a clear directionality (strongest at the stern) and frequency characteristics. The spatial distribution of the water pressure field refers to the variation of the pressure disturbance generated by a ship on the surrounding water with its position. It is closely related to the ship's draft, speed, and displacement, typically forming a high-pressure area at the bow and low-pressure areas on both sides of the midships.

[0028] The spatial distribution of the three physical fields mentioned above is closely related to the ship's speed: the higher the speed, the more significant the changes in the magnetic field due to hull vibration and eddies; the sound source level increases exponentially with speed; and the amplitude of water pressure disturbance is also proportional to the square of the speed. Therefore, this step establishes a physical field characteristic model at a preset speed. This preset speed can be the ship's planned cruising speed or multiple typical speed ranges, so that the corresponding field distribution data can be retrieved as needed under different navigation conditions.

[0029] By acquiring data on the ship's magnetic field, acoustic field, and hydraulic pressure field at a preset speed through actual measurement or simulation, and storing this data in a three-dimensional mesh format within the ship's own coordinate system, a physical field characteristic model of the ship can be obtained. In subsequent risk field construction and mine-trapping risk calculation, the physical field characteristic model is subjected to coordinate rotation transformation based on the ship's current course, converting the physical field distribution to a geodetic coordinate system. This allows for accurate calculation of the physical field intensity distribution in each direction under actual navigation conditions, providing a fundamental input for subsequent risk field construction and mine-trapping risk calculation.

[0030] S2. Construct a first risk field model for all known mines. The first risk field model includes the predicted position sequence and fuse triggering rules of each known mine during the ship's navigation period. The predicted position sequence requires obtaining the historical position sequence of the known mines. For drifting mines, the predicted position sequence during the ship's navigation period is predicted based on their historical position sequence. For moored mines and bottom mines, their fixed positions are regarded as their predicted position sequences. Integrate the predicted position sequences of all known mines and their corresponding fuse triggering rules to construct the first risk field model.

[0031] Information on known mines typically comes from historical intelligence, preliminary marine survey data, detection results from UAVs or unmanned underwater vehicles, and reports from passing vessels. For each known mine, its type (drifting mine, moored mine, or bottom mine) and its historical position sequence need to be identified. Different types of mines require different position processing methods: For drifting mines, since they move with ocean currents and wind fields, their future position sequence during the ship's navigation period can be predicted based on their historical position sequence. This prediction can employ mature time-series prediction models, combining historical position data with corresponding surface wind field and surface current data to output the predicted position coordinates of the drifting mine at various future times. For moored and bottom mines, their positions are relatively fixed, therefore no prediction is needed; the known fixed positions are directly used as their predicted position sequence during the ship's navigation period.

[0032] The fuze triggering rules describe the fuze type of the mine (magnetic fuze, acoustic fuze, hydraulic fuze, or a combination thereof) and the corresponding triggering threshold. This information can be obtained from a pre-established mine fuze characteristic database. For mines of known types, the corresponding fuze parameters in the database can be directly matched; for mines of unknown types, conservative principles are followed, such as using the maximum triggering range of the same type of mine or considering the most unfavorable combination of the three fuzes simultaneously to ensure a safety margin.

[0033] After obtaining the predicted location sequence of each known mine and its corresponding fuse triggering rule, all information is integrated to construct a first risk field model. The first risk field model is a database containing dynamic information for each mine. Given a ship's position, time, physical field characteristic model, speed, and heading, the set of threatening mines and their fuse triggering rules within the ship's spatiotemporal range can be retrieved from the first risk field model. The first risk field model records the predicted location of each known mine at every moment during the ship's navigation period, as well as the fuse triggering rule for that mine, providing a data foundation for calculating the mine-trapping risk cost in subsequent global path searches. In this way, the first risk field model can accurately reflect the dynamic threat of known mines over time, at a preset speed, and with a heading, and combined with the ship's physical field characteristic model, achieve an accurate assessment of navigation risks in minefield areas.

[0034] In one technical solution, a pre-trained PSO-Attention-LSTM model is used to obtain the future position sequence of drifting mines. The PSO-Attention-LSTM model is based on in-depth research in recent years on the prediction of marine floating object trajectories. Extensive experiments have demonstrated that the combined model, integrating particle swarm optimization, attention mechanisms, and long short-term memory networks, has significant advantages in handling temporal prediction problems in the marine environment. Drifting mines, as a type of marine floating object, exhibit similar motion patterns to those of people falling into the water and drifting buoys, all being dominated by surface wind fields and surface currents. Therefore, this model is also applicable to the trajectory prediction of drifting mines.

[0035] The PSO-Attention-LSTM model includes: The Long Short-Term Memory (LSTM) network layer receives historical location sequences of drifting mines, sea surface wind field data, and surface current data. It extracts the time-dependent features of the drift motion and outputs the hidden states at each time step. The model's input data consists of three parts: the historical location sequence of the drifting mines, the corresponding sea surface wind field data, and the surface current data. The historical location sequence typically comprises longitude and latitude coordinates at multiple time points, such as location points recorded every hour over the past 12 hours. Wind field and current data can be obtained from meteorological and oceanographic forecasting centers, such as ERA5 reanalysis data from the European Centre for Medium-Range Weather Forecasts (ECMWF) or regional ocean numerical forecast products, and aligned with the location sequence after spatiotemporal matching. To improve data stability and prediction accuracy, the historical location sequence is first processed using a differential moving average before being input into the model. This calculates the position deviation sequence between adjacent time points to eliminate noise and short-term fluctuations in the original data, preserving the long-term trend of the drift motion. The LSM network layer, through its forget gate, input gate, and output gate structure, effectively extracts the time-dependent features of the drift motion, overcoming the gradient vanishing or gradient exploding problems that traditional recurrent neural networks easily encounter when processing long sequences. This layer outputs the hidden states at each time step, which encode the historical patterns of the drifting mine movement, including periodic and trend characteristics.

[0036] The attention mechanism layer, connected to the Long Short-Term Memory (LSTM) network layer, is used to weight the hidden states at each time step, assigning higher attention weights to key time steps. The attention mechanism layer calculates the attention score for each time step, normalizes it using Softmax to obtain the attention weights, and then assigns higher weights to key time steps that have a greater impact on the prediction results. This mechanism enables the model to focus on important historical information in the drift trajectory, further enhancing its ability to represent motion patterns and improving prediction accuracy.

[0037] The Particle Swarm Optimization (PSO) module is used to globally optimize the hyperparameters of the Long Short-Term Memory (LSTM) network layers and the attention mechanism layers. Hyperparameters include at least the learning rate, hidden layer dimension, and number of hidden layers. The PSO algorithm simulates the foraging behavior of a flock of birds, iteratively searching for the optimal solution in the parameter space: first, a swarm of particles is initialized, each representing a combination of hyperparameters; then, the particle quality is evaluated based on a fitness function (such as the mean absolute error on the validation set), updating the individual optimal and the global optimal; next, the particle speed and position are adjusted, and the iteration is repeated until the termination condition is met. This module avoids the blindness of manual parameter tuning and effectively improves the model's prediction accuracy and generalization ability.

[0038] The output layer outputs the predicted position sequence of the drifting mines. It maps the attention-weighted features to the predicted position deviations for future times, then superimposes these deviations with the actual positions at the current time to obtain the predicted position sequence of the drifting mines at future times. The prediction results typically output positions at multiple future times, such as the coordinates of positions at various points in the future (10 minutes, 20 minutes, 30 minutes, up to the time period during which the ship passes).

[0039] The model training process employs supervised learning. First, a large amount of historical drifting mine trajectory data and corresponding wind and flow field data are collected as training samples, and then proportionally divided into training, validation, and test sets. Mean squared error is used as the loss function during training, and the Adam optimizer is selected. The particle swarm optimization module alternates with model training: in each iteration, the model is first trained based on the current particle swarm position, the mean absolute error on the validation set is calculated as the fitness value, and then the particle swarm is updated based on the fitness value until the preset maximum number of iterations is reached or the convergence condition is met. In this way, the model can automatically find the optimal combination of hyperparameters, maximizing prediction performance.

[0040] Compared to traditional time-series prediction models such as LSTM and GRU, the PSO-Attention-LSTM model exhibits higher prediction accuracy and stability when dealing with the problem of predicting the trajectory of floating objects at sea. Experiments have shown that this model can effectively capture the periodic and trend-based motion patterns in drift trajectories and maintain low positional errors even in long-term time-series prediction scenarios. Therefore, it is particularly suitable for applications such as drifting mines that require long-term predictions during the ship's navigation period.

[0041] S3. Construct a navigable grid map of the target sea area, build a comprehensive navigation cost function based on the physical field feature model and the first risk field model, and use a heuristic optimization algorithm to search for the path that minimizes the comprehensive navigation cost function, and generate an initial global route from the route start point to the route end point. Step S3 includes the following steps: S31. On the navigable grid map, define a set of nodes and a set of analysis edges. Each analysis edge corresponds to a navigable segment formed by connecting two adjacent nodes. The navigable grid map is constructed based on electronic nautical chart data, discretizing the continuous target sea area into a network structure composed of multiple nodes and analysis edges connecting the nodes. Each node corresponds to a geographic coordinate, and each analysis edge represents a navigable straight segment between two adjacent nodes. Its geometric length is calculated using the great circle distance formula to ensure accuracy. The grid spacing can be set according to the sea area range and planning accuracy requirements. For example, a sparser grid is used in open sea areas to reduce computation, while a denser grid is used in complex areas such as near islands or narrow waterways to improve path quality. Nodes and analysis edges corresponding to non-navigable areas such as islands, shoals, and reefs within the sea area are removed from the search space to ensure the physical feasibility of the planned path.

[0042] S32. The comprehensive travel cost of each analysis edge is obtained through weighted calculation, wherein the comprehensive travel cost includes: a) The navigation time cost, determined based on the geometric length of the analysis edge and the preset speed, reflects the time cost required for the ship to traverse that analysis edge. Its calculation is based on the geometric length of the analysis edge and the ship's preset speed on that segment; the shorter the path and the higher the speed, the lower the time cost. This component ensures that the planning results do not involve meaningless detours, guaranteeing basic navigation efficiency.

[0043] b) The cost of mine contact risk, determined based on a physical field characteristic model, the predicted mine location sequence, and the fuze triggering rules, represents the maximum risk of a ship contacting a mine when passing through the analysis edge at a preset speed at a specific time. For each analysis edge, based on the ship's expected arrival time at that segment, the predicted locations of all mines within the ship's influence radius at that moment and their corresponding fuze triggering rules are obtained from the first risk field model. Then, based on the physical field intensity (magnetic field strength, sound pressure level, or water pressure change) generated at the mine location when the ship passes through the segment at a preset speed and expected course, it is compared with the mine fuze triggering threshold to quantify the mine contact risk using a ratio. The larger the ratio, the higher the probability of the ship triggering the mine fuze on that segment, and the greater the cost of mine contact risk.

[0044] The formula for calculating the comprehensive navigation cost is as follows: in, i and j To analyze the node index of the edge, d ij To analyze the geometric length of the side, the formula for the distance of a great circle on a sphere is used for calculation; LFor the reference length used to normalize the distance cost, the characteristic scale of the target sea area (such as the typical voyage length to and from the minefield) can be taken as the benchmark value. v ij To analyze the preset speed on the side, R ij To analyze the cost of triggering landmines on the side.

[0045] M ij (t) For ships in analysis edge ij The moment t A group of mines in which the distance between the ship and the mine is less than the mine's preset radius of influence. m This serves as the index for the mine's serial number; in the time dimension, it represents the moment. t The estimated time when the ship arrives at a point on the analysis edge after departing from the starting point of the route is determined. Since the mine positions recorded in the first risk field model are predicted position sequences that change over time, it is necessary to determine the specific time based on the actual location. t The predicted position coordinates of each mine at that moment are extracted from the model. Spatially, this is based on the ship's position at time... t Centered on the predicted location, and using the preset influence radius of the mine as the search radius, a circular search area is delineated to initially screen the range of mines that may pose a threat to ships. The value of this area follows a conservative principle and can be directly set to a fixed conservative value.

[0046] Rule Penalty Items P ij The rules for mine avoidance are designed to reflect the hard constraints that mine avoidance imposes on navigation safety, and include at least two items: minimum safe distance and maximum safe speed.

[0047] The distance penalty term indicates the distance between the ship and the first edge in the analysis. m Whether the distance between the mines meets the minimum safe distance requirement is determined by setting the value to 0 if the distance between the ship and any mine on the analysis side is not less than the minimum safe distance, otherwise setting the value to 1. The minimum safe distance can be set according to the type of mine and the ship's maneuverability, and is usually not less than 50 meters.

[0048] This is a speed penalty term, indicating whether the ship's speed on the analysis edge meets the maximum safe speed requirement. It is set to 0 if the ship's speed on the analysis edge is not greater than the maximum safe speed, and 1 otherwise. The maximum safe speed can be set according to the ship's physical characteristics, and is usually no higher than 10 knots.

[0049] β , γ These are the weighting coefficients for each rule's penalty item. w d 、w t、w r 、w p These are the weighting coefficients for each cost item.

[0050] The formula for calculating the cost of mine strike risk is as follows: in, F m (t) For ships at time t This analysis examines the physical field intensity generated at the location of the mine. T m This is the fuse triggering threshold for a naval mine. a For safety reasons, F m (t) Based on the physical field characteristic model of this ship, and in conjunction with the mines... m The fuze type matches: for magnetic fuze mines, F m (t) For acoustic fuze mines, the magnetic field strength is... F m (t) The sound pressure level; for water pressure fuze mines, F m (t) This represents the change in water pressure. T m For sea mines m The trigger threshold for the fuze can be obtained from the mine fuze characteristic database; a As a safety factor, it can be set according to the operational requirements (e.g., a value between 1.2 and 2.0) to provide additional safety margin.

[0051] In the above weighting process, for an analysis edge ij Ships from node i sail to the node j The process involves a period of time during which the ship's spatiotemporal position continuously changes, and the corresponding predicted mine locations and ship physical field distribution also change accordingly. Therefore, the risk of mine contact and rule violations vary at different points along the analysis edge. The highest risk level and the maximum rule violation level along the analysis edge are used as representative values ​​for that segment, ensuring that any segment with localized high risk or rule violations is penalized in the comprehensive navigation cost function. The maximum value operation is a design choice based on a conservative safety perspective; it ensures the robustness of risk assessment and rule checking without significantly increasing computational complexity, making the path search results more reliable.

[0052] S33. Ant colony optimization is used to find the path. Heuristic information is determined based on the comprehensive travel cost. The next node is selected based on the pheromone concentration of each analysis edge traversed by each ant and the heuristic information, until the destination is reached. After one iteration, the pheromone concentration of each analysis edge is updated based on the total comprehensive travel cost of the complete path traversed by each ant, so that the path with the smaller total comprehensive travel cost receives a higher pheromone increment. The iteration is repeated until the termination condition is met, and the path with the smallest total comprehensive travel cost is output as the initial global route.

[0053] Ant colony optimization (ACO) is an intelligent optimization algorithm that simulates the foraging behavior of ant colonies. Its core idea is to achieve indirect communication and cooperation among ants by releasing pheromones along their paths and sensing pheromone concentrations. In nature, ants secrete pheromones during their journeys, and subsequent ants tend to follow paths with higher pheromone concentrations. Shorter paths have higher ant turnover rates and faster pheromone accumulation, attracting more ants and creating a positive feedback effect that ultimately leads the entire colony to converge on the optimal path. This invention introduces this mechanism into route planning in minefields, using comprehensive navigation costs as the criterion for path quality, guiding the ant colony to search for a global route that achieves the optimal balance between navigation efficiency and safety. The specific implementation process is as follows: C1. Parameter Initialization and Initial Pheromone Distribution Setting. Specifically, before starting the ant colony search, the parameters required for the algorithm to run are first initialized. The main parameters that need to be set include: the number of ants. K Typically, the number of grid nodes can be taken as 1.5 to 2 times to ensure sufficient search; pheromone volatile factor ρ The value typically ranges from 0.1 to 0.5 to control the decay rate of pheromones with iteration; information heuristic factor α The relative importance of pheromone concentration in transfer decisions; expected heuristic factor. β The relative importance of heuristic information in the transfer decision; pheromone intensity constant. Q This is used to adjust the overall magnitude of the pheromone increment in each iteration; the maximum number of iterations. N max This serves as a hard upper limit for terminating the iteration. The values ​​of the above parameters can be adjusted appropriately based on the size of the target sea area and the required planning accuracy. After the parameters are set, the initial pheromone concentration of all analysis edges in the network graph is set. τ ij Set it to a small positive number (e.g., 1.0 or a slightly lower value) to ensure that the ants have enough randomness in choosing directions during the initial search phase, and to prevent the algorithm from focusing on a certain path too early.

[0054] C2. Determine the heuristic information based on the comprehensive navigation costs; specifically, the heuristic information... ηij Defined as the comprehensive navigation cost of this analysis edge C ij The reciprocal of the metric. The lower the overall travel cost, the better the analytical edge is after comprehensively weighing distance, time, mine risk, and rule compliance; correspondingly, its heuristic information is greater, and its attractiveness to ants is stronger. When an ant needs to evaluate analytical edges... ij At that time, along the entire path of the analysis edge, with a set risk assessment step size. Δd The time intervals are gradually increased, the mine situation is updated segment by segment, and the risk is calculated segment by segment. Finally, the comprehensive navigation cost of the analysis edge is obtained by summarizing the results.

[0055] For example, the ant arrives at the node i At that time, the accumulated sailing time T acc Analyze the edges ij The total geometric length is d ij The preset speed is v ij Along the navigation direction of this analysis edge, starting from the starting point at each interval... Δd Take a risk assessment sampling point, and for each sampling time... t n Perform the following risk calculations: t n To index the first risk field model, obtain the predicted locations of all known mines at that moment; based on the ship's position... t n Using the predicted location at a given time as the center and a preset influence radius as the search radius, all mines within the influence range are selected to form the set of threatening mines corresponding to that sampling point. M ij (t n ) ;against M ij (t n ) For each mine in the system, the corresponding physical field strength is extracted from the physical field feature model after coordinate rotation transformation, based on its fuze type. F m (t n ) Combined with the fuse trigger threshold T m Calculate the location of the mine at the sampling point. t n place The maximum value among all mines is taken as the mine risk ratio for that sampling point; at the same time, based on the distance between the ship and each mine and the current speed, the rule penalty value for that sampling point is calculated according to the definition of the rule penalty item in step S32.

[0056] After calculating all N sampling points, the maximum value of the mine-trapping risk ratio for all sampling points on the analysis edge is taken to obtain the mine-trapping risk cost of the analysis edge. R ij The maximum value of the rule penalty for all sampling points is taken to obtain the rule penalty term for the analyzed edge. P ij .

[0057] C3. Based on the pheromone concentration of each edge traversed by the ants and this heuristic information, select the next node until the destination is reached. (When an ant is at the current node...) i At that time, for the sake of i Adjacent candidate nodes j First, calculate each analysis edge according to C2. ij Comprehensive navigation costs under current spatiotemporal conditions C ij and corresponding inspirational information η ij Then, the current pheromone concentration on each analysis edge is combined. τ ij Calculate the number of ants from the node i Select Move to Node j State transition probability: Each ant starts from the beginning of the route, with an initial cumulative travel time. T acc = 0. At each step, calculate the selection probability of each candidate direction according to the above formula, and randomly select the next node using a roulette wheel strategy based on probability. Move to the selected node and record the analyzed edges and their combined travel costs in the path. Repeat this process until the destination is reached, constructing a complete path from the starting point to the end point. Simultaneously, obtain the sum of the combined travel costs of all analyzed edges on the path, i.e., the total combined travel cost of the path. C total .

[0058] C4. After one iteration, update the pheromone concentration of each analysis edge based on the total comprehensive travel cost of the complete path traversed by each ant, so that paths with lower total comprehensive travel costs receive higher pheromone increments. Iteration refers to all ants completing the construction of a complete path from the starting point to the ending point. The update process consists of two stages: pheromone evaporation and pheromone enhancement. Pheromones evaporation weakens historical pheromone accumulation, preventing the algorithm from prematurely falling into local optima. Pheromones enhancement involves adding pheromone to the analysis edges along the ant's path, based on the reciprocal of the path's total comprehensive travel cost. The better the path (the lower the total cost), the more pheromone is added, gradually increasing the attractiveness of high-quality paths to ants in subsequent iterations.

[0059] C5. Repeat the iteration until the termination condition is met, and output the path with the minimum total comprehensive navigation cost as the initial global route. After completing the pheromone update, clear the path record, reset the ants to the starting point, and start a new round of iteration. Repeat this process until the maximum number of iterations is reached or the optimal path does not change significantly for several consecutive generations. Extract the path with the minimum total comprehensive navigation cost from all iteration records as the initial global route, and form a complete route from the starting point to the ending point, with the route starting point as the first node and the route ending point as the last node, providing a safe reference route for ships to enter minefields to perform missions.

[0060] S4. During the ship's navigation along the initial global route, the information of newly appearing mines in the target sea area is detected and updated in real time. The predicted location sequence of the newly appearing mines in the subsequent navigation period is predicted by combining real-time wind and sea state data. A second risk field model is constructed, which includes the predicted location sequence of each newly appearing mine in the ship's navigation period and the fuse triggering rules. The second risk field model includes the predicted location sequence and fuse triggering rules for each newly discovered mine during the ship's navigation period. As the ship navigates along the initial global route, the location information of newly discovered mines in the target sea area needs to be detected and updated in real time. "Detected mines" refers to mines whose location information has been confirmed through historical intelligence or preliminary detection before global path planning; "newly discovered mines" refers to mines first discovered by real-time detection equipment during the ship's navigation. Real-time detection can be accomplished through a combination of technologies. For example, mine-detecting sonar mounted on helicopters can use ultrasonic waves to detect, locate, and identify mines, providing distance and bearing information for newly discovered mines. Mine-detecting sonar can detect moored mines and exposed bottom mines. In addition, airborne laser detection systems, unmanned underwater vehicles, and shipborne multibeam sonar can also be used to acquire real-time location information of newly discovered mines. For underwater targets suspected of being mines but whose type cannot be accurately identified during detection, for safety reasons, they are all handled according to the worst-case scenario of simultaneously possessing magnetic, acoustic, and hydraulic fuses, and the smallest empirical value among the triggering thresholds of the three types of fuses is used as the basis for calculation.

[0061] After obtaining the location information of newly discovered mines, it is necessary to predict their trajectory during subsequent navigation periods. Unlike known mines, newly discovered mines typically lack sufficient historical location sequences, making it impossible to directly apply deep learning-based time-series prediction models. Therefore, a physics-based drift prediction model is required. The wind-induced drift coefficient and lateral offset probability corresponding to the newly discovered mine are matched from a mine hydrodynamic parameter database. Combined with real-time sea surface wind field data and surface current data, a physics-based drift prediction model is used to predict the predicted location sequence of the newly discovered mine during subsequent navigation periods.

[0062] Currently, the most widely used model internationally, and validated by numerous search and rescue forecasting systems, is the AP98 (Automated Prediction 98) model and its derivatives developed by the U.S. Coast Guard. This model has a clear physical meaning, simplifying the motion of floating objects at sea into a vector superposition of two independent processes: current drag and wind-induced drift. It is highly suitable for rapidly estimating the initial trajectory of newly appearing targets. This invention preferably uses the AP98 model as the basis for prediction, and the specific implementation steps are as follows: First, it is necessary to determine the hydrodynamic characteristic parameters of newly discovered mines, namely the wind-induced drift coefficient and the lateral drift probability. These parameters can be automatically matched from a pre-established database of mine hydrodynamic parameters based on the detected mine type. The wind-induced drift coefficient describes the drift response characteristics of a mine under wind force and is defined as the ratio of the mine's wind-induced drift velocity to the sea surface wind speed. This coefficient is closely related to factors such as the mine's surface structure and windward area. The lateral drift probability refers to the statistical distribution characteristics of the mine's deflection motion caused by the irregularity of the mine's shape and the complexity of sea conditions. For mines whose types can be clearly identified, their corresponding precise parameters are directly read from the database; for unknown mines whose types cannot be matched, a conservative approach is adopted, such as using a larger wind-induced drift coefficient among similar mines and setting the lateral drift probability to 0.5 (i.e., equal left and right deflection probabilities) to ensure that the prediction results have sufficient safety margin.

[0063] After obtaining the hydrodynamic parameters of the mines, these parameters, along with real-time sea surface wind field data (wind speed and direction) and surface current data (current velocity and direction) acquired from the meteorological and oceanographic forecasting center, are input into the AP98 drift prediction model. The core of the model's calculation is to decompose the drift motion into two components: drift with the current and wind-induced drift. Finally, after obtaining the predicted position sequence of each newly appearing mine and its corresponding fuse triggering rule, a second risk field model can be constructed for each newly appearing mine. This model's data structure is completely consistent with the first risk field model, containing the dynamic position sequence of newly appearing mines and their fuse triggering rules during the ship's navigation period. Its establishment provides an accurate and dynamic data foundation for subsequent spatiotemporal correlation analysis of routes and local path replanning, enabling ships to perceive and respond to new threats in real time during navigation.

[0064] S5. Perform spatiotemporal correlation analysis between the initial global route and the second risk field model. Combining the predicted position sequence and fuse triggering rules in the physical field characteristic model and the second risk field model, calculate the mine-trapping risk cost for each segment of the initial global route at the ship's expected arrival time. Segments with mine-trapping risk costs exceeding the risk threshold are marked as dangerous segments. Since the second risk field model contains the predicted position sequence and fuse triggering rules for each newly appearing mine during the ship's navigation period, and the initial global route provides the time information for the ship's expected passage through each position, spatiotemporal correlation analysis can accurately determine whether the ship will enter the risk area of ​​a newly appearing mine during navigation. This analysis combines the time and space dimensions, considering not only the geographical location of each point on the route but also whether the specific time the ship arrives at that point is within the same spatiotemporal interval as the mine. This is specifically achieved through the following sub-steps: S51. Discretize the initial global route into several segments. Each segment corresponds to the ship's expected arrival time, preset speed, and expected course. Specifically, the initial global route, output in step S3, is represented as a sequence of analytical edges connected end-to-end from the route's starting point to its ending point. To facilitate segment-by-segment assessment of the threat of newly emerging mines, this continuous route is first reconstructed into several segments, each corresponding to an analytical edge in the navigable grid map. For each segment, based on the information recorded during the path search in step S3, extract the following spatiotemporal parameters: the geometric length of the segment. d ij and preset speed v ij This is used to calculate the time required for the ship to traverse the segment; the direction of the segment, serving as the ship's expected course as it traverses; and the expected time when the ship arrives at the starting point of the segment, which has been accumulated along with the ant colony's travel time in step S3 during the ant colony search. T accThe estimated arrival time of a ship at any point on the route is determined by the progress of the journey. Based on the above parameters, the estimated arrival time of the ship at any point on the route can be further calculated, providing a time reference for subsequent calculation of mine-trapping risk costs. Taking the starting time of the initial global route (usually the current time or the planned departure time) as zero, the estimated arrival time sequence for each route segment can be obtained by accumulating the travel time segment by segment.

[0065] S52. Calculate the mine-trapping risk cost of the ship during passage by combining the physical field characteristic model and the second risk field model. Specifically, for each segment discretized in S51, assess the degree of threat posed by newly appearing mines to the ship during passage segment by segment using the same method as the mine-trapping risk cost calculation in step S32. The calculation process is as follows: S521. Determine the time interval for a vessel to pass through the segment. The estimated time when the vessel arrives at the starting point of the segment is taken as the start time, and the sum of the start time and the travel time of the segment is taken as the end time. This time interval is the period during which the vessel is on the segment.

[0066] S522. Obtain the predicted locations of newly appearing mines within this time period. Using several characteristic moments within this time interval (e.g., the moment corresponding to the midpoint of the navigation segment, or multiple moments sampled along the navigation segment with a fixed step size), query the second risk field model to obtain the instantaneous predicted locations of all newly appearing mines at each characteristic moment.

[0067] S523. Determine the set of threatening mines. For each characteristic moment, using the predicted position of the ship at that moment as the center and the preset influence radius as the search radius, mines within the influence range are selected from the instantaneous predicted positions of newly appearing mines to form the set of threatening mines for that segment at each characteristic moment.

[0068] S524. Calculate the risk cost of striking a mine. For each newly appearing mine in the threatening mine set, based on the ship's expected course and preset speed for this segment, obtain the physical field intensity generated by the ship at the mine's location from the physical field characteristic model established in step S1, and obtain the mine's fuze triggering threshold from the mine fuze characteristic database. Calculate the mine-striking risk ratio at each characteristic moment. Take the maximum mine-striking risk ratio among all characteristic moments as the mine-striking risk cost for this segment. R ij S53. If the mine-trapping risk cost exceeds a preset risk threshold, the flight segment is determined to be covered by the second risk field model and marked as a dangerous flight segment. Specifically, the mine-trapping risk cost for each flight segment calculated in step S52 is... R ijThe value is compared with a preset risk threshold. The risk threshold can be set according to operational safety requirements, such as 0.8 or 1.0. When the risk threshold is 1.0, it means that a danger is only determined when the physical field strength generated by the ship at the mine location reaches or exceeds the mine fuse triggering threshold; when the risk threshold is less than 1.0 (such as 0.8), it means that a certain safety margin is allowed, and even if the physical field strength has not fully reached the triggering threshold, it is considered a threat as long as it is close to a certain level.

[0069] When one or more consecutive segments are identified as hazardous segments, they are merged into a single hazardous segment interval, and its start and end nodes on the initial global route are recorded. This hazardous segment interval represents the area where the vessel cannot safely pass as originally planned, and a local replanning area needs to be designated around it. Through route replanning, a safe detour path can be generated.

[0070] S6. Using the dangerous section as a baseline, expand the area outward to form a local replanning area. Perform path replanning within this area to obtain a local avoidance path. Replace the initial global route within the local replanning area with the local avoidance path. Following the spatiotemporal correlation analysis in step S5, the sections of the initial global route covered by the second risk field model have been marked as dangerous sections. To generate a safe alternative route, this step performs local path replanning on the dangerous section. The basic process is as follows: First, using the dangerous section as a baseline, expand outward along the route direction and vertically to form a local replanning area; then, perform path search within this area to generate a local avoidance path that can safely avoid all mine threats and meets the ship's maneuverability requirements; finally, replace the original sections of the initial global route within this area with the local avoidance path, thus obtaining the updated planned route, enabling the ship to safely pass through the area where new mines have appeared.

[0071] The acquisition of the local avoidance path includes the following steps: A1. Using the dangerous section as a reference, a predetermined buffer length is extended along both ends of the section as the longitudinal boundary of the replanning area, and a predetermined buffer width is extended along both sides of the section as the lateral boundary of the replanning area, forming a local replanning area that includes the dangerous section and its surrounding area. Specifically, using the start and end positions of the dangerous section marked in step S5 on the initial global route as a reference, a predetermined buffer length is extended forward and backward along the route direction to form the longitudinal range of the replanning area; then, a predetermined buffer width is extended to both sides perpendicular to the route direction to form the lateral range of the replanning area. The buffer length and buffer width can be predetermined according to the actual sea conditions, mine distribution density, and ship maneuverability. For example, the buffer length can be 1 to 2 nautical miles, and the buffer width can be 0.5 to 1 nautical mile.

[0072] Longitudinal extension ensures that the replanning area covers a sufficiently long range before and after the dangerous section, allowing ample space for new routes to smoothly diverge from and rejoin the original route; lateral extension provides sufficient space for lateral detours, enabling routes to safely bypass the risk area from the side. In this way, the local replanning area encompasses all the risky waters that need to be avoided while limiting the search range to a reasonable interval, balancing planning quality and computational efficiency.

[0073] A2. Discretize the local replanning region into a local grid map. The local grid map consists of multiple nodes and analysis edges connecting adjacent nodes. The grid spacing is smaller than that of the global grid map. Specifically, the construction method of the local grid map is the same as that of the global navigable grid map in step S31, but the grid precision is higher to meet the needs of local fine-grained planning. The grid spacing can be set according to the local area range and the density of mine distribution, usually taking 1 / 2 to 1 / 5 of the global grid spacing. Grid nodes are evenly distributed within the local replanning region. Each node corresponds to a geographical location coordinate. Adjacent nodes are connected by analysis edges. Each analysis edge represents a straight navigable segment of the ship, and its geometric length is calculated using the great circle distance formula. Nodes and analysis edges corresponding to non-navigable areas (such as islands, shoals, and reefs) within the local replanning region are also removed from the search space.

[0074] A3. Spatiotemporally superimpose the portions of the first risk field model and the second risk field model located within the local replanning area to form a comprehensive risk field within the area. Specifically, the first risk field model contains the dynamic risk information of all known mines constructed in step S2, and the second risk field model contains the dynamic risk information of all newly discovered mines constructed in step S4. Together, they constitute the complete risk situation within the local area.

[0075] The spatiotemporal overlay process merges the risk information of two risk field models at the same time and space coordinates: for any time and any location, if the spatiotemporal point is determined to be threatened by either model, then the point is included in the comprehensive risk field and marked as a risk area; if multiple mines (including known mines and newly discovered mines) pose a threat at the same spatiotemporal point, then the largest mine-touch risk ratio is taken as the mine-touch risk cost of that point.

[0076] A4. Within the local grid map, using the boundary points where the ship enters and leaves the local replanning area as the local start and local end points, the K-short-circuit algorithm is used for path search. During the search process, the total comprehensive navigation cost on each candidate path is calculated, and the paths are sorted from smallest to largest according to the total comprehensive navigation cost. The top n paths are selected as the candidate path set. Specifically, the start and end points of the path replanning are determined based on the boundary points where the ship enters and leaves the local replanning area along the initial global route: the local start point is the first intersection point between the initial global route and the boundary of the local replanning area, and the local end point is the last intersection point.

[0077] The search algorithm employs the K-short-circuit algorithm, which, based on heuristic search, successively deviates from the already found shortest path to generate the first few shortest paths. During the search process, the total comprehensive travel cost of each candidate path is calculated according to the formula defined in step S32, i.e., the comprehensive travel cost of each analyzed edge on the path. C ij Summation, where the weighting coefficients of each cost component are calculated. w d , w t , w r , w p Adjustments can be made based on the safety priority strategy during the local replanning phase. The specific calculation methods and weighting principles for each cost item in the comprehensive navigation cost have been explained in detail in step S32 and will not be repeated here.

[0078] During the search process, only paths where the risk of hitting a mine at each point is below a preset risk threshold are retained. After sorting the paths by total comprehensive navigation cost from smallest to largest, several top-ranked paths are selected as a candidate path set.

[0079] A5. If the candidate path set is not empty, then candidate paths are selected in order of increasing total comprehensive navigation cost for ship turnaround feasibility verification, and the first candidate path that passes the verification is taken as the local avoidance path; if the candidate path set is empty, or all candidate paths fail the turnaround feasibility verification, then the buffer length and buffer width are increased, and steps A2 to A4 are repeated.

[0080] The vessel turning feasibility check ensures that each turning point on the candidate path meets the vessel's actual maneuverability constraints. First, all turning points requiring a course change are extracted from the candidate path. For each turning point, based on the vessel's current speed and planned rudder angle, the corresponding turning lead and minimum turning radius are retrieved from the vessel turning element table. The actual turning radius of the candidate path at that turning point is compared with the vessel's minimum turning radius: if the actual turning radius is less than the minimum turning radius, the turning point is determined not to meet the turning feasibility requirements. Simultaneously, based on the vessel's expected arrival time and speed at the turning point, the turning lead is verified to match the path geometry. Only when all turning points on the path simultaneously meet the minimum turning radius and turning lead requirements does the candidate path pass the turning feasibility check. If the current candidate path passes the check, it is directly output as a local avoidance path.

[0081] If the initial path fails, the next candidate path is selected for verification in ascending order of total comprehensive navigation cost. If the candidate path set is empty (i.e., no path satisfying the risk threshold and avoidance rule constraints was found in step A4), or all candidate paths fail the loop feasibility check, it indicates that the current boundary setting of the local replanning area may be too small, making it impossible to generate a safe and practically executable detour path within a limited space. In this case, the buffer length and buffer width are increased, and the step size can be set to 50% of the original value. The maximum number of attempts should not exceed 3 to avoid infinite loops. The process returns to step A2 to re-discretize the mesh and perform subsequent searches. Through this iterative feedback mechanism, the scope of the replanning area is continuously adjusted until a local avoidance path that satisfies all safety constraints and maneuverability requirements is found.

[0082] In practice, local avoidance paths typically involve a small detour near the hazardous section, with a geometric length not significantly different from the original section replaced in the initial global route. Consequently, the resulting time offset is generally short. Considering that the predicted mine locations in both the first and second risk field models typically change only slightly over a short period, the probability of new mine-touch risks in subsequent sections due to time offset is relatively low. After local replanning, the remaining original route can be quickly verified at the new time offset. Replanning is only triggered when a new threat is confirmed, thus minimizing computational overhead while ensuring safety.

[0083] The specific steps are as follows: Step S6 includes the following: S7. Based on the time offset generated by navigating along the local avoidance path, determine the updated estimated arrival time of the ship at the local endpoint. Since the local avoidance path differs from the replaced hazardous segment in the initial global route in terms of geometric length and preset speed, the actual time the ship takes to pass through the local replanning area will differ from the original planned time. Based on the actual length and preset speed of the local avoidance path, calculate the actual time the ship takes to travel along this path from the local starting point to the local endpoint, compare this time with the originally planned time to pass through the replaced segment, obtain the time offset, and then update the estimated arrival time of the ship at the local endpoint.

[0084] S8. Using the updated estimated arrival time as the time base, perform spatiotemporal correlation analysis on the remaining initial global route between the local endpoint and the route endpoint with the first risk field model and the second risk field model to verify whether a new dangerous route segment is formed. If no new dangerous segments are verified, the local avoidance path from step S6 is directly concatenated with the remaining initial global route to form the secondary planned route. The updated estimated arrival time obtained in S7 is used as the starting time for the ship to enter the remaining route. For the remaining initial global route between the local endpoint and the route endpoint, the spatiotemporal correlation analysis is re-performed according to the method in step S5. If the verification results show that there are no new dangerous segments on the remaining route where the risk of mine strikes exceeds the preset risk threshold, indicating that the sailing time offset has not affected the subsequent voyage, the local avoidance path generated in step S6 is directly concatenated with the remaining initial global route to form the secondary planned route.

[0085] If a new dangerous segment is verified, the local endpoint is used as the new route starting point, and the updated estimated arrival time is used as the starting time. Step S3 is performed on the remaining voyage after the local avoidance path to obtain the adjusted route. The initial route segment that has not yet been traversed before the local avoidance path, the local avoidance path, and the adjusted route are sequentially spliced ​​together to form the secondary planned route.

[0086] It should be noted that although the steps are described in a specific order above, this does not mean that they must be performed in that order. In fact, some of these steps can be executed concurrently, or even in a different order, as long as the required functionality is achieved. The number of devices and processing scale described herein are for simplification of the invention; applications, modifications, and variations of this invention will be readily apparent to those skilled in the art.

[0087] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A minefield area ship route planning method, characterized by, Includes the following steps: S1. Establish a physical field characteristic model of the ship. The physical field characteristic model includes at least the spatial distribution of magnetic field, sound field and water pressure field at a preset speed. The physical field characteristic model is stored in the ship's own coordinate system and is rotated and transformed according to the ship's course when used. S2. Construct a first risk field model for all known mines. The first risk field model includes the predicted location sequence of each known mine during the ship's navigation period and the fuse triggering rules. S3. Construct a navigable grid map of the target sea area, build a comprehensive navigation cost function based on the physical field feature model and the first risk field model, and use a heuristic optimization algorithm to search for the path that minimizes the comprehensive navigation cost function, and generate an initial global route from the route start point to the route end point. S4. During the ship's navigation along the initial global route, the information of newly appearing mines in the target sea area is detected and updated in real time. The predicted location sequence of the newly appearing mines in the subsequent navigation period is predicted by combining real-time wind and sea state data. A second risk field model is constructed, which includes the predicted location sequence of each newly appearing mine in the ship's navigation period and the fuse triggering rules. S5. Perform a spatiotemporal correlation analysis between the initial global route and the second risk field model. Combine the predicted position sequence and fuse triggering rules in the physical field characteristic model and the second risk field model to calculate the mine-trapping risk cost of each segment in the initial global route at the ship's expected arrival time. Segments with mine-trapping risk costs exceeding the risk threshold are recorded as dangerous segments. S6. Based on the dangerous route segment, expand the area to form a local replanning area. Perform route replanning within the local replanning area to obtain a local avoidance route. Replace the initial global route within the local replanning area with the local avoidance route.

2. The minefield ship route planning method of claim 1, wherein, In step S2, the historical position sequence of the known mines is obtained. For drifting mines, the predicted position sequence during the ship's navigation period is predicted based on their historical position sequence. For anchored mines and bottom mines, their fixed positions are regarded as their predicted position sequences. The predicted position sequences of all known mines and their corresponding fuse triggering rules are integrated to construct the first risk field model.

3. The minefield ship route planning method of claim 2, wherein, For drifting mines with existing historical location sequences, a pre-trained PSO-Attention-LSTM model is used to obtain predicted location sequences. The PSO-Attention-LSTM model includes: The Long Short-Term Memory (LSTM) network layer is used to receive the historical location sequence of drifting mines, sea surface wind field data, and surface current data, extract the time-dependent features of drift motion, and output the hidden state at each time step. The attention mechanism layer, connected to the long short-term memory network layer, is used to weight the hidden states at each time step and assign higher attention weights to key time steps. The particle swarm optimization module is used to globally optimize the hyperparameters of the long short-term memory network layer and the attention mechanism layer. The output layer is used to output the predicted location sequence of the drifting mines.

4. The minefield ship route planning method of claim 1, wherein, Step S3 includes the following steps: S31. On the navigable grid map, define a set of nodes and a set of analysis edges. Each analysis edge corresponds to a flight segment formed by connecting two adjacent nodes. S32. The comprehensive travel cost of each analysis edge is obtained through weighted calculation, wherein the comprehensive travel cost includes: a) The cost of travel time is determined based on the geometric length of the analysis edge and the preset travel speed; b) The cost of mine strike risk, determined based on the physical field characteristic model, the predicted location sequence of mines, and the fuse triggering rules, represents the maximum risk of a ship striking a mine when it passes through the analysis edge at a preset speed at a specific time; S33. Ant colony optimization is used to find the path. Heuristic information is determined based on the comprehensive travel cost. The next node is selected based on the pheromone concentration of each analysis edge traversed by each ant and the heuristic information, until the destination is reached. After one iteration, the pheromone concentration of each analysis edge is updated based on the total comprehensive travel cost of the complete path traversed by each ant, so that the path with the smaller total comprehensive travel cost receives a higher pheromone increment. The iteration is repeated until the termination condition is met, and the path with the smallest total comprehensive travel cost is output as the initial global route.

5. The minefield ship route planning method of claim 4, wherein, The formula for calculating the comprehensive navigation cost is as follows: in, i and j To analyze the node index of the edge, d ij To analyze the geometric length of the side, L The reference length used for normalizing the distance cost. v ij To analyze the preset speed on the side, R ij To analyze the risks and costs of stepping on landmines on the side, M ij (t) For ships in analysis edge ij The moment t A group of mines in which the distance between the ship and the mine is less than the mine's preset radius of influence. m This is the index for the mine's serial number; For distance penalty terms, the value is 0 if the ship is at least 1 away from any mine on the analysis side; otherwise, the value is 1. As a speed penalty term, it is set to 0 if the ship's speed on the analysis edge is not greater than the maximum safe speed, and 1 otherwise. β , γ These are the weighting coefficients for each rule's penalty item. w d 、w t 、w r 、w p These are the weighting coefficients for each cost item.

6. The minefield ship route planning method of claim 5, wherein, The formula for calculating the cost of mine strike risk is as follows: wherein, F m (t) is the time at which the ship is at the position of the mine, t By this analysis the intensity of the physical field generated at the position of the mine at the time t, T m is the trigger threshold of the fuze of the mine, a is the safety factor.

7. The minefield ship route planning method of claim 4, wherein, In step S4, the wind-induced drift coefficient and lateral offset probability corresponding to the newly appearing mine are matched from the mine hydrodynamic parameter library. Combined with real-time sea surface wind field data and surface current data, a drift prediction model based on physical mechanism is used to obtain the predicted location sequence of the newly appearing mine.

8. The minefield ship route planning method of claim 4, wherein, Step S5 includes: S51. Discretize the initial global route into several segments, each segment corresponding to the ship's expected arrival time, preset speed and expected course; S52. Combine the physical field characteristic model and the second risk field model to calculate the risk cost of a ship hitting a mine when passing through; S53. If the risk of striking a mine exceeds the preset risk threshold, the flight segment is determined to be covered by the second risk field model and marked as a dangerous flight segment.

9. The minefield ship route planning method of claim 8, wherein, The acquisition of the local avoidance path includes the following steps: A1. Based on the dangerous flight segment, a preset buffer length is extended along both ends of the flight segment as the longitudinal boundary of the replanning area, and a preset buffer width is extended along both sides of the flight segment as the lateral boundary of the replanning area, forming a local replanning area that includes the dangerous flight segment and its surrounding area. A2. Discretize the local replanning region into a local grid map, which consists of multiple nodes and analysis edges connecting adjacent nodes, with the grid spacing being smaller than that of the global grid map. A3. Spatiotemporally overlay the portions of the first risk field model and the second risk field model located within the local replanning region to form a comprehensive risk field within that region; A4. Within the local grid map, taking the boundary points where ships enter and leave the local replanning area as the local start point and local end point, the K-short-circuit algorithm is used for path search. During the search process, the total comprehensive navigation cost on each candidate path is calculated, and the paths are sorted from smallest to largest according to the total comprehensive navigation cost. The top n paths are selected as the candidate path set. A5. If the candidate path set is not empty, then candidate paths are selected in order of increasing total comprehensive navigation cost for ship turnaround feasibility verification, and the first candidate path that passes the verification is taken as the local avoidance path; if the candidate path set is empty, or all candidate paths fail the turnaround feasibility verification, then the buffer length and buffer width are increased, and steps A2 to A4 are repeated.

10. The minefield ship route planning method of claim 9, wherein, Step S6 is followed by: S7. Based on the time offset caused by sailing along the local avoidance path, determine the updated estimated arrival time of the ship at the local endpoint; S8. Using the updated estimated arrival time as the time base, perform spatiotemporal correlation analysis on the remaining initial global route between the local endpoint and the route endpoint with the first risk field model and the second risk field model to verify whether a new dangerous route segment is formed. If it is verified that there are no new dangerous segments, the local avoidance path in step S6 is directly spliced ​​with the remaining initial global route as the secondary planning route. If a new dangerous segment is verified, the local endpoint is used as the new route starting point, and the updated estimated arrival time is used as the starting time. Step S3 is performed on the remaining voyage after the local avoidance path to obtain the adjusted route. The initial route segment that has not yet been traversed before the local avoidance path, the local avoidance path, and the adjusted route are sequentially spliced ​​together to form the secondary planned route.

Images