A ship ecological route planning method and system based on marine mammal prediction

Abstract

The application discloses a ship ecological route planning method and system based on marine mammal prediction, relates to the technical field of intelligent navigation of ships, and obtains multi-source space-time data; the multi-source space-time data are subjected to space-time alignment and fusion, and are unified to a regular grid to form an environment variable field consistent in space-time; a marine biological risk prediction model is constructed in advance, marine mammal habitat suitability index is output, and the habitat suitability index is mapped to an initial navigation biological risk cost; the initial navigation biological risk cost is dynamically corrected to obtain an individual navigation biological risk cost; a multi-objective comprehensive cost function is constructed; and an optimal route is planned with the lowest total cost as the target. The application engineering standardizes the prediction information of dynamic large marine mammals into a standardized navigation risk field, establishes a coupling correction mechanism of the ship maneuvering state and biological risk, and automatically solves the optimal route considering economy and ecological protection through a multi-objective optimization algorithm.

Description

Technical Field

[0001] This invention relates to the field of intelligent ship navigation technology, and more specifically to a method and system for ship ecological route planning based on marine mammal prediction. Background Technology

[0002] Currently, the core logic of mainstream ship route planning systems (such as Electronic Chart Display and Information System (ECDIS) and meteorological navigation services) is still based on static geographic information and meteorological and hydrological data. Although fixed obstacles such as reefs and shoals, as well as predictable meteorological risks such as typhoons and ocean currents, have been incorporated into the conventional constraint system, large marine mammals such as whales, as "dynamic obstacles" that pose a substantial threat to navigation safety, are still in a significant "perception blind spot" in the current engineered route planning system.

[0003] Most existing data treat marine mammals as uncertain "environmental background noise" rather than quantifiable "dynamic risk factors." For example, over 92% of the global marine mammal activity ranges highly overlap with shipping routes, and effective management measures have been implemented in less than 7% of identified high-risk ship-whale collision hotspots. However, existing navigation algorithms cannot handle the high spatiotemporal heterogeneity of biological distribution. Marine mammal habitats are driven by environmental factors such as sea surface temperature, chlorophyll a concentration, and mesoscale eddies, exhibiting significant seasonal migration and diurnal fluctuations. Existing systems lack the technical means to transform multi-source biological observation data into a standardized, continuous navigation risk field, resulting in ships lacking both proactive early warning capabilities and automated avoidance decision support when facing sudden or migrating biological aggregation areas. The root cause of this deficiency lies in the fact that existing systems follow a binary collision avoidance logic of "spatial isolation," i.e., mandatory avoidance through the designation of no-navigation zones; while marine mammal risk is essentially a continuous field problem requiring "risk probability response," necessitating systems capable of making refined dynamic responses such as detours and speed reductions based on the level of risk. Current technology lacks the engineering capability to transform a continuous probabilistic risk field into executable ship control commands.

[0004] To address collision risks, current methods primarily rely on establishing static "avoidance zones" or seasonal speed-limited zones. However, this extensive management approach, based on historical average distributions, suffers from two major drawbacks: firstly, static boundaries cannot keep pace with real-time changes in marine mammal habitats, forcing vessels to perform ineffective detours in areas where risks have shifted or encountering missed warnings in newly emerging high-risk areas; secondly, to ensure coverage, static zones are often over-established, compelling vessels to make unnecessary long-distance detours, significantly increasing fuel consumption and carbon emissions. Studies show that only about 1.25% of the world's existing marine protected areas truly bear high-intensity shipping pressure, and the vast majority of static protected areas have limited effectiveness in mitigating collision risks due to temporal and spatial misalignment. This "one-size-fits-all" approach reduces shipowners' willingness to proactively comply with regulations. Current technologies lack a multi-objective optimization framework that couples dynamic biological risk fields with ship maneuvering parameters, necessitating a dynamic optimization mechanism that can finely balance between low-risk and high-risk areas.

[0005] Therefore, in view of the shortcomings of the existing technology, how to provide a ship ecological route planning method and system based on marine mammal prediction, engineer marine ecological prediction information into a dynamic navigation risk field, and couple it with ship maneuvering parameters for multi-objective route optimization is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a method and system for ship ecological route planning based on marine mammal prediction. It engineered the dynamic large marine mammal habitat prediction information into a standardized navigation risk field, established a coupled correction mechanism between ship maneuvering status and biological risk, and automatically solved the optimal route that takes into account both economy and ecological protection through a multi-objective optimization algorithm, so as to fill the technical gap in the existing ship navigation system in dynamic biological obstacle perception and response.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a method for ship ecological route planning based on marine mammal prediction, comprising: Acquire multi-source spatiotemporal data on ships and sea areas; The multi-source spatiotemporal data are spatiotemporally aligned and fused, and unified into a regular grid to form a spatiotemporally consistent environmental variable field; Using a pre-built marine biological risk prediction model, with the environmental variable field as input, the model outputs the marine mammal habitat suitability index for each grid within a future preset time period, and maps the habitat suitability index to the initial navigation biological risk cost. Based on the current speed and underwater noise radiation level, the initial navigation biological risk cost is dynamically corrected to obtain an individualized navigation biological risk cost. Construct a multi-objective comprehensive cost function based on time cost, fuel economy cost, and the individualized biological risk cost of navigation; Based on the multi-objective integrated cost function, the optimal route from the current location to the destination is planned with the goal of minimizing the total cost.

[0008] Preferably, the marine biological risk prediction model is constructed based on a species distribution modeling method, which is used to transform multidimensional environmental element representations into standardized spatial distribution probability assessment results of marine mammals; Then, through a nonlinear signal conditioning function, the habitat suitability index is mapped to the initial navigation biological risk cost for adapting to the ship's path search.

[0009] Preferably, the nonlinear signal conditioning function employs an exponential mapping, and the mapping process is as follows: ; in, This represents the initial biological risk cost of the voyage, where k is the normalized scaling factor for the time cost equivalent. HSI This indicates the habitat suitability index.

[0010] Preferably, based on the current speed and underwater noise radiation level, the initial navigation biological risk cost is dynamically corrected to obtain an individualized navigation biological risk cost, including: ; in, V 0 , N 0 Here, V represents the reference speed and reference noise level, and N represents the current underwater noise radiation level. p、q This is an empirical correction factor. This represents the initial biological risk cost of the voyage. This indicates the biological risks and costs associated with individualized navigation.

[0011] Preferably, the multi-objective comprehensive cost function is expressed as: ； in, To start from the current node i to adjacent nodes j The single-step comprehensive cost; The time cost required to traverse this grid segment; The fuel economy cost required to traverse this grid segment; For the target node j The biological risks and costs of individualized navigation at various locations; a、b、c These are non-negative adjustable weighting coefficients.

[0012] Preferably, based on the multi-objective comprehensive cost function, with the goal of minimizing the total cost, the optimal route from the current location to the destination is planned, including: With the goal of minimizing the total cumulative cost from the origin to the destination, a path search algorithm is used to solve the optimal route from the current location to the destination on a comprehensive cost grid map superimposed with individualized biological risk costs of navigation.

[0013] Preferably, the multi-source spatiotemporal data is spatiotemporally aligned and fused, and unified to a regular grid to form a spatiotemporally consistent environmental variable field, including: Multi-source spatiotemporal data with different spatiotemporal resolutions are unified into a regular grid using a weighted averaging method with a time decay window, forming a spatiotemporally consistent environmental variable field. The regular grid covers the target sea area, and each grid node carries the fused multidimensional environmental variable values.

[0014] Preferably, it further includes: graphically displaying the optimal route and outputting auxiliary evaluation indicators for the route; During navigation, a hybrid mechanism combining periodic and event-triggered events is employed to re-predict and plan. When a confirmed real-time alarm for the presence of marine mammals is received, the data is fed back to the model for incremental learning and to update the model parameters. For the planned candidate routes, the cumulative exposure is calculated as an auxiliary assessment indicator of ecological impact, used to characterize the cumulative impact of the ship on marine mammal habitats along the entire route.

[0015] Preferably, a ship ecological route planning system based on marine mammal prediction includes: The data acquisition module is used to acquire multi-source spatiotemporal data of ships and sea areas; The spatiotemporal alignment module is used to perform spatiotemporal alignment and fusion of the multi-source spatiotemporal data and unify it into a regular grid to form a spatiotemporally consistent environmental variable field. The risk prediction module is used to utilize a pre-built marine biological risk prediction model, with the environmental variable field as input, to output the marine mammal habitat suitability index for each grid within a future preset time period, and to map the habitat suitability index to the initial navigation biological risk cost. The dynamic correction module is used to dynamically correct the initial navigation biological risk cost based on the current speed and underwater noise radiation level, so as to obtain an individualized navigation biological risk cost. The cost layer construction module is used to construct a multi-objective comprehensive cost function based on time cost, fuel economy cost, and the individualized biological risk cost of navigation; The optimal route planning module is used to plan the optimal route from the current location to the destination based on the multi-objective comprehensive cost function, with the goal of minimizing the total cost.

[0016] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for ship ecological route planning based on marine mammal prediction, which has the following technical effects: 1. This invention overcomes the shortcomings of existing technologies in lacking engineering methods for representing dynamic biological obstacles, achieving standardized perception and quantification of navigation biological risks. Through spatiotemporal alignment and fusion, and ecological risk cost mapping, this invention transforms marine mammal habitat prediction results, which rely on marine ecology, into standardized navigation biological risk costs that can be directly utilized by ship navigation systems. By using a nonlinear signal conditioning function to map the habitat suitability index into a cost equivalent comparable to time and fuel consumption, marine mammals are elevated from "environmental background noise" to "dynamic risk factors" in route planning, filling the gap in dynamic biological obstacle information on electronic nautical charts.

[0017] 2. This invention overcomes the severe mismatch between existing static control measures and dynamic ecological realities, achieving refined synergistic optimization of spatial avoidance and speed management. By correcting ship maneuvering states, this invention introduces speed and underwater noise radiation levels to individually adjust risk costs, allowing ships in different maneuvering states within the same sea area to receive differentiated risk costs. Combined with an adjustable-weight multi-objective comprehensive cost function, the system can generate a continuous family of optimal routes from purely economical routes to strictly ecological routes, supporting refined trade-offs between slightly risky and high-risk areas for "detours" or "reduced speed passages," avoiding the ineffective detours or underreporting problems caused by the "one-size-fits-all" approach of static protection zones.

[0018] 3. This invention overcomes the shortcomings of existing technologies, such as the lack of an optimization framework and the static open-loop nature of the model, and establishes an evolutionary multi-objective joint optimization system. Through a constructed comprehensive cost function, this invention unifies time cost, fuel economy cost, and individualized navigation biological risk costs under the same mathematical framework, enabling direct comparison and weighted optimization of three different dimensional costs. By employing a hybrid periodic and event-based triggering mechanism and incremental learning, the risk prediction model can continuously evolve using real-time observation data, solving the problem of accumulated prediction bias in sparse high seas data areas. Furthermore, it introduces the cumulative exposure total as a global assessment indicator of the ecological impact of shipping routes, compensating for the inability of single-step optimization to characterize cumulative effects. At the system implementation level, a cloud-edge collaborative architecture is adopted, with shore-based systems handling large-scale computations and ship-based systems responsible for local decision-making. An offline degradation strategy is included to ensure continuous availability of the system under complex sea conditions. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a ship ecological route planning method based on marine mammal prediction, provided as an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of a ship ecological route planning system based on marine mammal prediction, provided as an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention discloses a method for ship ecological route planning based on marine mammal prediction, such as... Figure 1 As shown, it includes: Acquire multi-source spatiotemporal data of ships and sea areas, including the ship's current position, destination, current speed, underwater noise radiation characteristic parameters of the ship, as well as real-time and forecast marine environmental variable data of the target sea area and historical marine mammal appearance records; The multi-source spatiotemporal data are spatiotemporally aligned and fused, and unified into a regular grid to form a spatiotemporally consistent environmental variable field; Using a pre-built marine biological risk prediction model, with the environmental variable field as input, the model outputs the marine mammal habitat suitability index for each grid within a future preset time period, and maps the habitat suitability index to the initial navigation biological risk cost R0. Specifically, using a pre-built marine biological risk prediction model, with the aligned environmental variable field as input, the marine mammal habitat suitability index (HSI) of each grid in the future preset time period is output. Based on the current speed and underwater noise radiation level, the initial navigation biological risk cost is dynamically corrected to obtain an individualized navigation biological risk cost. Specifically, risk correction is based on the ship's maneuvering status. The initial navigation biological risk cost R0 is dynamically corrected based at least on the current speed V and underwater noise radiation level N, resulting in an individualized navigation biological risk cost R. Here, R characterizes the real-time ecological impact of a specific ship on marine mammals under a specific maneuvering state, distinct from the initial navigation biological risk cost R0, which only characterizes the inherent ecological sensitivity of the marine area.

[0024] Construct a multi-objective comprehensive cost function based on time cost, fuel economy cost, and the individualized biological risk cost of navigation; Specifically, a comprehensive cost function is constructed. A multi-objective comprehensive cost function is built, comprising time cost T, fuel economy cost F, and the individualized navigation biological risk cost R, with weight coefficients assigned to each cost. Based on the established rule grid, a single-step transfer cost from the current node i to the adjacent node j is defined, where time cost and fuel cost are path-related process costs, and the individualized navigation biological risk cost is a node attribute cost bound to the target node. By adjusting the weight coefficients, continuous switching between economic priority and ecological priority can be achieved, providing a quantified optimization objective for path search, and generating a comprehensive cost raster map accordingly.

[0025] Based on the multi-objective integrated cost function, the optimal route from the current location to the destination is planned with the goal of minimizing the total cost.

[0026] Specifically, the marine mammals described in this embodiment of the invention are large marine mammals, such as whales.

[0027] Specifically, the marine biological risk prediction model is constructed based on a species distribution modeling method, which is used to transform multidimensional environmental element representations into standardized spatial distribution probability assessment results for marine mammals. Then, through a nonlinear signal conditioning function, the habitat suitability index is mapped to the initial navigation biological risk cost for adapting to the ship's path search.

[0028] Specifically, the nonlinear signal conditioning function employs an exponential mapping, and the mapping process is as follows: ; in, This represents the initial biological risk cost of the voyage, where k is the normalized scaling factor for the time cost equivalent. HSI This represents the habitat suitability index. This transformation occurs at high... HSI The area generates a magnification effect, ensuring that high-risk grids are forced to avoid collisions during flight path searches.

[0029] Other nonlinear functions with monotonically increasing properties (such as the Sigmoid function) can also be used to implement the above mapping.

[0030] Specifically, based on the current speed and underwater noise radiation level, the initial navigation biological risk cost is dynamically adjusted to obtain an individualized navigation biological risk cost, including: ; in, V 0 , N 0 Here, V represents the reference speed and reference noise level, and N represents the current underwater noise radiation level. p、q This is an empirical correction factor. This represents the initial biological risk cost of the voyage. This represents the biological risk cost of individualized navigation. p and q are determined by logistic regression fitting of historical ship-whale collision accident data. p reflects the elasticity coefficient of speed with respect to collision probability, and q reflects the elasticity coefficient of noise attenuation of whales' perception ability.

[0031] This correction reflects the amplified effect of high-speed, high-noise vessels on the potential impact on marine mammals. This correction is based on the following mechanism: Firstly, the higher the vessel speed, the shorter the effective time window for marine mammals to perceive the vessel and react evasively, and the longer the vessel's turning and braking distances, leading to a significantly increased probability of collision. Secondly, the higher the underwater noise radiation level of the vessel, the stronger the auditory masking effect on marine mammals, making it difficult for them to accurately perceive the vessel's position and approach speed, thus reducing their ability to effectively avoid collisions. These two effects are superimposed, equivalent to a simultaneous amplification of collision risk. Therefore, the individualized correction function applies both speed and noise factors to the initial biological risk cost of navigation. R 0 This allows ships in different maneuvering states within the same risk field to receive differentiated risk costs commensurate with their actual collision risks. R .

[0032] Specifically, the multi-objective comprehensive cost function is expressed as: ； in, To start from the current node i to adjacent nodes j The single-step comprehensive cost; The time cost required to traverse this grid segment; The fuel economy cost required to traverse this grid segment; For the target node j The biological risks and costs of individualized navigation at various locations; a、b、c These are non-negative adjustable weight coefficients; the multi-objective integrated cost function is used to calculate node costs during path search.

[0033] Existing route optimization algorithms are limited to physical / economic dimensions such as time, fuel consumption, and weather safety in their objective functions, lacking a method to quantify biological risks into continuous cost parameters that can be uniformly calculated with the aforementioned indicators. Current achievements mostly remain at the level of scientific early warning or information display, failing to be embedded in the core solvers of route planning. Furthermore, most are static open-loop models trained on historical data, lacking the closed-loop learning capability to receive real-time observation data and perform online incremental updates, making it difficult to adapt to the continuous correction needs of prediction biases in high seas data-sparse areas. Due to the lack of a mathematical framework for joint optimization of "biological risk and economic cost," computer systems cannot automatically calculate the optimal route that balances navigation efficiency and ecological protection. Pilots are forced to make a binary choice between "completely ignoring risk" and "mechanically enforcing static prohibitions" based on experience, failing to achieve a paradigm shift from "static geographical safety" to "dynamic ecological safety."

[0034] The embodiments of the present invention adjust c It is possible to use the "pure economic model" ( c =0) and "strict ecological model" ( α =0, β =0, c Switch continuously between =1).

[0035] Specifically, based on the aforementioned multi-objective integrated cost function, with the goal of minimizing the total cost, the optimal route from the current location to the destination is planned, including: With the goal of minimizing the total cumulative cost from the origin to the destination, a path search algorithm is used to solve the optimal route from the current location to the destination on a comprehensive cost grid map superimposed with individualized biological risk costs of navigation.

[0036] Specifically, the path search algorithm is an improved fast traversal method or an improved A algorithm. The algorithm searches for the optimal path on the regular grid, using the overall cost as the transfer cost between nodes.

[0037] Specifically, the multi-source spatiotemporal data is spatiotemporally aligned and fused, and unified into a regular grid to form a spatiotemporally consistent environmental variable field, including: Multi-source spatiotemporal data with different spatiotemporal resolutions are unified into a regular grid using a weighted averaging method with a time decay window, forming a spatiotemporally consistent environmental variable field. The regular grid covers the target sea area, and each grid node carries the fused multidimensional environmental variable values, providing a unified spatial computing framework for subsequent steps and serving as the search space for state nodes in path planning; Specifically, the weighted average method with time decay window is as follows: for each target grid, at the current time... t Based on this, all valid observations from the past 72 hours are collected, and their fusion weights are... for: ; in, t For the current moment, λ represents the historical occurrence time of this observation data, and λ is the attenuation coefficient.

[0038] Specifically, the spatial resolution of the regular grid is set to 0.1°×0.1° or 1 / 12°×1 / 12°, covering a buffer zone of a predetermined width centered on the ship's planned route. Each grid node stores at least the fused sea surface temperature, chlorophyll a concentration, sea surface height anomaly, and water depth values. The spatial resolution setting must satisfy the following conditions: sufficient to resolve the spatial heterogeneity of environmental variables in the target whale species' habitat, and the grid side length must match the typical maneuvering response scale of the ship.

[0039] Specifically, it also includes: graphically displaying the optimal route and outputting auxiliary evaluation indicators for the route; During navigation, a hybrid mechanism combining periodic triggering and event triggering is adopted to re-predict and plan. When a confirmed real-time alarm of marine mammals is received, the data is fed back to the model for incremental learning and to update the model parameters. For the planned candidate routes, the cumulative exposure total E is calculated as an auxiliary assessment index of ecological impact. The cumulative exposure total E represents the degree of cumulative impact of the ship on marine mammal habitats along the entire route.

[0040] Specifically, the hybrid mechanism combining periodic triggering and event triggering is as follows: a basic update cycle is set, and the gradient of the risk field of the waypoint ahead and shore-based sudden alarms are monitored in real time. When the gradient exceeds the limit or an alarm is received, replanning is triggered immediately. The incremental learning method involves adjusting the local weights or fine-tuning the online risk prediction model. The cumulative exposure total E is calculated as follows: ; in, For navigation grid k The biological risks and costs of individualized navigation E represents the distance (nautical miles) a ship travels through this grid; E serves as an ecological impact assessment indicator independent of economic costs, used to assist crew decision-making or as one of the reference conditions for triggering replanning.

[0041] In one specific embodiment of the present invention, a ship ecological route planning system based on marine mammal prediction, such as... Figure 2 As shown, it includes: The data acquisition module is used to acquire multi-source spatiotemporal data of ships and sea areas, including ship dynamic data, underwater noise characteristic parameters, real-time and forecast marine environmental variable data of the target sea area, and historical marine mammal appearance records.

[0042] The spatiotemporal alignment module is used to perform spatiotemporal alignment and fusion of the multi-source spatiotemporal data and unify it into a regular grid to form a spatiotemporally consistent environmental variable field; the embodiment of the present invention performs time decay window weighted fusion of multi-source spatiotemporal data; The risk prediction module is used to utilize a pre-built marine biological risk prediction model, with the environmental variable field as input, to output the marine mammal habitat suitability index for each grid within a future preset time period, and to map the habitat suitability index to the initial navigation biological risk cost. The dynamic correction module is used to dynamically correct the initial navigation biological risk cost based on the current speed and underwater noise radiation level to obtain an individualized navigation biological risk cost. In this embodiment of the invention, a pre-trained marine biological risk prediction model is built in to generate the initial navigation biological risk cost based on the aligned environmental variable field, and to perform individual correction based on the ship's real-time speed and underwater noise radiation level to obtain the corrected navigation biological risk cost grid. The cost layer construction module is used to construct a multi-objective comprehensive cost function based on time cost, fuel economy cost, and the individualized navigation biological risk cost; in this embodiment of the invention, time cost, fuel economy cost, and the modified navigation biological risk cost are superimposed according to weights; The optimal route planning module is used to plan the optimal route from the current location to the destination based on the multi-objective comprehensive cost function, with the goal of minimizing the total cost. In this embodiment of the invention, the goal is to minimize the comprehensive cost function, and a path search algorithm is used to solve the optimal route that takes into account both biological risks and economic costs on the comprehensive cost grid map.

[0043] Specifically, it also includes: a dynamic update and self-learning module, which is used to execute a hybrid mechanism that combines periodic triggering and event triggering. When the triggering conditions are met, it triggers the re-prediction of the risk field and the replanning of the route. After receiving a confirmed real-time alarm of the presence of marine mammals, it uses the alarm data to incrementally update the risk prediction model. The route assessment and display module is used to calculate the cumulative exposure of the planned route as an auxiliary assessment indicator, and to overlay the optimal route with the comprehensive cost grid on the electronic nautical chart, while providing heading guidance and yaw alarm.

[0044] Specifically, the system adopts a cloud-edge collaborative architecture: the shore-based server deploys the model inference and computation in the risk prediction and correction modules, as well as the model incremental update in the dynamic update and self-learning modules; the shipboard client deploys the data acquisition module, spatiotemporal alignment module, cost layer construction module, route planning module, and route evaluation and display module; the shore-based and shipboard clients interact through standardized message exchanges, the messages of which at least include risk field data packets, route request packets, and route result packets; when ship-shore communication is interrupted, the shipboard client uses the most recently cached risk field data and the local simplified model to maintain the route planning and evaluation functions, and automatically synchronizes the latest data after communication is restored.

[0045] Example 1: This invention takes a container ship sailing from Shanghai Yangshan Port to Los Angeles Port, USA, as an example. The ship has a design speed of 18 knots and an underwater noise radiation level of approximately 165 dB re 1μPa@1m. The specific steps are as follows: S1: Acquisition of Multi-Source Spatiotemporal Data The ship's current position (30.5°N, 122.5°E), destination (33.7°N, 118.2°W), and current speed of 18 knots are read from the ship's navigation system. Propeller speed and hull parameters are obtained from the engine room monitoring system to estimate the underwater noise radiation level (N). The ship's marine environmental forecast for the Northeast Pacific region, including sea surface temperature (SST), chlorophyll a concentration (CHL), sea surface height anomaly (SLA), and water depth (BAT), is retrieved daily from the Copernicus Marine Service Center via shipborne broadband satellite communication, with a raw spatial resolution of approximately 1 / 12°. Simultaneously, a database of historical sightings of fin whales and blue whales in this area over the past 10 years is acquired as a training basis for the risk prediction model.

[0046] S2: Spatiotemporal alignment and fusion of multi-source heterogeneous data

[0047] Forecast data from different sources were uniformly resampled to a regular grid of 0.1° × 0.1°. For sparse point-source observations such as historical cetacean sightings, a weighted average method with a time decay window was used for fusion processing: for each target grid, using the current time t as a baseline, all valid observations from the past 72 hours were collected, and their fusion weights were... ,in, The historical occurrence time of the observation data is represented by λ, which is the attenuation coefficient; in this embodiment, λ = 0.05. After this processing, the weight of observation data from 24 hours ago is approximately 0.3, the weight of observation data from 72 hours ago is approximately 0.03, and observation data older than 72 hours is no longer included in the calculation. This method effectively solves the time scale mismatch problem between daily average satellite remote sensing products and instantaneous cetacean sighting data, forming a spatiotemporally consistent environmental variable field and providing a unified spatial calculation framework for subsequent steps.

[0048] S3: Ecological Risk Cost Mapping

[0049] This step uses a pre-built cetacean risk prediction model, taking the aligned environmental variable field as input, and outputs the cetacean habitat suitability index (HSI) for each grid in the future preset time period. Then, through a nonlinear signal conditioning function, the HSI is mapped to the initial navigation biological risk cost R0 to adapt to the ship path search.

[0050] In this embodiment, the marine biological risk prediction model is preferably implemented using a gradient boosting tree model (XGBoost). Its construction process is as follows: (1) Data preparation and feature engineering: The historical cetacean occurrence records of the target sea area over the past 10 years are used as positive samples, and an equal number of background points are randomly generated in the study area as negative samples. Four environmental variables, sea surface temperature (SST), chlorophyll a concentration (CHL), sea surface height anomaly (SLA), and water depth (BAT), are extracted from the synchronous historical marine environment reanalysis data. After resampling to a uniform spatial resolution (0.1°×0.1°), they are spatiotemporally matched with the positive and negative samples to construct a training dataset.

[0051] (2) Model training: Input the training dataset into the XGBoost model for binary classification training. The search range of key hyperparameters is: learning rate 0.01 to 0.3, maximum tree depth 3 to 10, subsampling ratio 0.6 to 1.0. The optimal combination is determined by 5-fold cross-validation to obtain the habitat suitability prediction function.

[0052] (3) Model evaluation: The area under the receiver operating characteristic curve (AUC) is used to evaluate the model performance. When the AUC value is greater than the preset threshold (e.g., 0.8), the model is deemed qualified; otherwise, the model is returned to adjust the feature combination or hyperparameters.

[0053] In the prediction application phase, the forecast field generated by S2 for the next 0-24 hours is input into the trained model. The model outputs the probability of cetacean occurrence for each grid, which is then calibrated by probability (such as by Platt Scaling) and used as the Habitat Suitability Index (HSI), with a value range of 0 to 1.

[0054] In this embodiment of the invention, the HSI output is engineered into a standardized navigation biological risk cost that can be embedded in the route planning solver through subsequent nonlinear signal conditioning and ship maneuvering state correction.

[0055] The nonlinear signal conditioning function employs an exponential mapping: ; Where k is the normalized scaling factor for the time cost equivalent, and in this embodiment, k = 100 nautical miles equivalent is taken. This transformation produces an amplification effect in the high HSI region. When HSI = 0.7, R0 ≈ 101 nautical miles equivalent, which means that the cost of crossing this grid is equivalent to sailing an additional 101 nautical miles, ensuring that high-risk grids are forcibly avoided by the algorithm in subsequent path search.

[0056] S4: Risk Correction Based on Ship Maneuvering Status

[0057] Within the same sea area, different ships and under different navigation conditions face significantly different collision risks. This step introduces the ship's maneuvering status to modify the initial navigation biological risk cost. Based at least on the current speed V and underwater noise radiation level N, R0 is dynamically amplified to obtain the individualized navigation biological risk cost R.

[0058] The individualized correction function is defined as follows: ; Where V0 is the reference speed (12 knots), N0 is the reference noise level (160 dB re 1μPa@1m), and p and q are empirical correction coefficients. In this embodiment, p=0.8 and q=0.5.

[0059] Substituting the ship parameters in this embodiment: V=18 knots, N=165 dB, calculate the correction factor: = (18 / 12)^0.8 ≈ 1.38; = (165 / 160)^0.5 ≈ 1.03; Correction factor = 1.38 × 1.03 ≈ 1.42; With this correction, R = 1.42 × R0, the biological risk cost of individualized navigation at 18 knots and 165 dB noise level is increased by about 42% compared to the initial value.

[0060] This correction is based on the following mechanism: On the one hand, the higher the ship's speed, the shorter the effective time window for cetaceans to perceive the ship and make an evasive response, and the longer the ship's own turning and braking distances, leading to a significant increase in the probability of a collision; on the other hand, the higher the underwater noise radiation level of the ship, the stronger the shielding effect on the cetaceans' hearing, making it difficult for cetaceans to accurately perceive the ship's position and approaching speed, thus reducing their ability to effectively avoid collisions. The two effects are superimposed, which is equivalent to amplifying the collision risk simultaneously.

[0061] R represents the real-time ecological impact of a specific vessel on cetaceans under a specific maneuvering condition, distinct from the initial navigational biological risk cost R0, which only represents the inherent ecological sensitivity of the marine area. Through this correction, vessels in different maneuvering conditions within the same risk field obtain differentiated risk costs that match their actual collision risks, supporting the synergistic optimization of two risk control strategies: spatial avoidance and speed management.

[0062] S5: Construction of the Comprehensive Cost Function

[0063] A multi-objective comprehensive cost function is constructed, comprising time cost T, fuel economy cost F, and the individualized navigation biological risk cost R, with weight coefficients set for each cost. Based on the regular grid established in S2, the single-step transfer cost from the current node i to the adjacent node j is defined as follows: ; in, To start from the current node i to adjacent nodes j The single-step comprehensive cost; The time cost required to traverse this grid segment; The fuel economy cost required to traverse this grid segment; For the target node j The biological risks and costs of individualized navigation at various locations; a、b、c These are non-negative adjustable weighting coefficients.

[0064] In this embodiment, the ship selects the "balanced" mode, with corresponding weights α=1, β=1, and γ=3. By adjusting γ, the system can continuously switch between the "pure economic mode" (γ=0) and the "strict ecological mode" (α=0, β=0, γ=1), generating an optimal route family that balances ecological risks and economic costs.

[0065] Will Defined as a node attribute rather than a path attribute, this allows the pathfinding algorithm to evaluate both the cost of a "detour" and the cost of "slowing down" during the exploration process. The reduced state costs enable the coordinated optimization of two risk control strategies: space avoidance and velocity management.

[0066] S6: Optimal Route Search

[0067] Based on the comprehensive cost function constructed in S5, an improved Fast Journey Method (FMM) is used as the path search algorithm to progressively explore from the starting point to the destination on the regular grid described in S2. During the search process, the comprehensive cost of the explored nodes and their adjacent nodes is calculated as needed, with the goal of minimizing the cumulative total cost from the starting point to the destination, thus solving for the optimal route that balances ecological risk and economic cost. In this embodiment, the calculation takes approximately 2.8 minutes to obtain the optimal route.

[0068] It should be noted that the comprehensive cost raster map is not pre-calculated for all grids in the entire domain before the path search begins. Instead, it is gradually filled in by the algorithm as it calculates the comprehensive cost of the explored nodes during the search process. For grids not explored by the algorithm, their cost data is not included in the calculation, thus avoiding unnecessary pre-calculation overhead for the entire domain.

[0069] In this embodiment of the invention, the comprehensive cost function constructed in S3-S5, which includes the individualized biological risk cost R of navigation, is used as the cost input for the path search algorithm, rather than the traditional single cost of time or fuel consumption; the comprehensive cost function contains... The node attribute cost differs fundamentally from the traditional edge cost (such as distance and time) in its computational logic. The path search algorithm processes both node cost and edge cost simultaneously to adapt to the synergistic optimization of two risk control strategies: spatial avoidance and speed management.

[0070] S7: Dynamic Updates and Self-Learning

[0071] The system outputs the optimal route and displays it graphically. This optimal route is overlaid with a comprehensive cost grid map on the electronic nautical chart—high-risk areas are presented as red heatmaps, and safe areas as green, allowing the navigator to visually see how the route avoids the red high-risk areas. It also provides heading guidance and yaw warnings. When a vessel deviates from the recommended route, the system can trigger a replanning process, forming a safety loop of "deviation → warning → replanning".

[0072] The navigation employs a hybrid mechanism combining periodic and event-triggered approaches: a basic update cycle of 6 hours is set, periodically re-acquiring the latest forecast data and executing S3-S6; simultaneously, the risk field gradient at the waypoint ahead and shore-based emergency alarms are monitored in real time. When the gradient exceeds the limit or an alarm is received, replanning is immediately triggered. For example, midway through the voyage, the shipborne acoustic buoy network suddenly reports the appearance of a group of blue whales 30 nautical miles ahead of the planned route. The system detects a sudden increase in the risk gradient and immediately triggers event-based replanning, generating a new detour route in approximately 1.5 minutes.

[0073] When a confirmed real-time alert of a whale sighting is received, the alert data is fed back to the model for incremental learning. The ship uploads the coordinates (with timestamp) of the confirmed blue whale sighting to the shore-based server. The shore-based server uses this point as a new positive sample, adjusts the local weights or fine-tunes the risk prediction model online, and distributes the updated model parameters to the ship and other vessels in the same area, achieving collective intelligent evolution.

[0074] For the planned candidate flight routes, the cumulative total exposure E is calculated as a supplementary indicator for assessing ecological impact. E is calculated as follows: ,in, The biological risk cost of individualized navigation through grid k, This represents the distance (in nautical miles) the ship travels through this grid. E characterizes the cumulative ecological impact of the ship on cetacean habitats along the entire route; a higher value indicates a greater potential disturbance to cetaceans. As a global ecological assessment indicator independent of economic costs, E is used to assist crew decision-making—when E exceeds a preset threshold, the system can suggest adjusting the weighting coefficients and trigger replanning, forming a closed loop of "optimization-evaluation-re-optimization".

[0075] Comparison of proportions and effects: To verify the technical effectiveness of the embodiments of the present invention, Monte Carlo simulations (e.g., 200 random encounters) were conducted on a ship maneuvering simulation platform using scaled-down experimental data; three sets of comparative schemes were set up under the same scenario: Option A (purely economical route): γ=0, only optimizes time and fuel consumption, without considering ecological risks.

[0076] Option B (Static Avoidance Route): Designate a large static cetacean sanctuary established by the IMO as a no-navigation zone and automatically detour.

[0077] Option C (Dynamic Ecological Route of the Invention Embodiment): The method of the present invention is adopted, with weights α=1, β=1, and γ=3, enabling all risk corrections and hybrid trigger updates.

[0078] The experimental results are shown in Table 1: Table 1 Comparative Example and Effect Verification Results

[0079] Option C reduces the proportion of high-probability risk zone crossings from 18.2% to 1.8% at the minimal cost of increasing the flight range by only 3.2%, reducing the average risk cost by 85.8%, and achieving minute-level response to sudden risks. Its effect is far superior to the static detour option B.

[0080] Model validity verification

[0081] To verify the accuracy of HSI prediction in this embodiment, the 2020 shipping window was used as a natural experiment. Environmental variables before and after the sharp drop in ship AIS density during this period were extracted, input into the model, and the high-probability distribution area was output. This was then compared spatiotemporally with cetacean distribution data from an independent survey conducted during the same period. The results show that the spatial correlation between the model-predicted high-probability area and the actual cetacean density area exceeds 0.85 (P<0.01), demonstrating that the model can realistically respond to environmental changes and human activity disturbances, providing empirical support for the credibility of the risk field.

[0082] Example 2: This embodiment provides a ship ecological route planning system that implements the above method, and adopts a cloud-edge collaborative architecture.

[0083] The shore-based server deploys a data acquisition module (interfacing with the Marine Forecasting Center), a spatiotemporal alignment module, a model inference computation module within the risk prediction and correction module, and incremental model updates within the dynamic update and self-learning module. Prediction runs daily at 00 UTC, generating two versions of the risk field for the next 0-48 hours and 48-96 hours, which are distributed via satellite multicast. The data message includes a compressed raster matrix, timestamp, model version number, and validity period.

[0084] The shipboard client deploys a data acquisition module (reading data from the ship's own sensors), a cost layer construction module, a route planning module, and a route evaluation and display module. The shipboard client is responsible for: caching and verifying the latest risk field messages; collecting the ship's V and N data and performing individualized corrections for S4; receiving the destination and weighting mode input by the pilot and completing local route planning; calculating the cumulative exposure total E and overlaying the optimal route with the comprehensive cost raster map.

[0085] The shore-based and shipboard systems communicate via standardized messages, which include at least: (1) Risk field data package: raster coordinates, HSI value, timestamp, model version number; (2) Route request packet: vessel position, destination, speed, noise parameters, weighting coefficient; (3) Route result package: waypoint sequence, estimated arrival time.

[0086] When ship-to-shore communication is interrupted, the shipboard client automatically switches to offline mode, using the most recently cached risk field data and the local emergency simplified model (which approximates the HSI from the forecast data based on preset regression coefficients) to maintain the route planning function. Once communication is restored, the latest data will be automatically synchronized.

[0087] This invention addresses the problem that existing ship navigation systems lack the ability to handle the risks of dynamic biological aggregation, cannot map the spatiotemporal dynamic changes of large cetacean habitats into engineering parameters that can be used for route optimization, resulting in delayed collision avoidance decisions for ships in ecologically sensitive waters, and the inability to achieve a quantifiable balance between biological risks and economic costs.

[0088] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0089] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for ship ecological route planning based on marine mammal prediction, characterized in that, include: Acquire multi-source spatiotemporal data on ships and sea areas; The multi-source spatiotemporal data are spatiotemporally aligned and fused, and unified into a regular grid to form a spatiotemporally consistent environmental variable field; Using a pre-built marine biological risk prediction model, with the environmental variable field as input, the model outputs the marine mammal habitat suitability index for each grid within a future preset time period, and maps the habitat suitability index to the initial navigation biological risk cost. Based on the current speed and underwater noise radiation level, the initial navigation biological risk cost is dynamically corrected to obtain an individualized navigation biological risk cost. Construct a multi-objective comprehensive cost function based on time cost, fuel economy cost, and the individualized biological risk cost of navigation; Based on the multi-objective integrated cost function, the optimal route from the current location to the destination is planned with the goal of minimizing the total cost.

2. The method for ship ecological route planning based on marine mammal prediction according to claim 1, characterized in that, The marine biological risk prediction model is constructed based on the species distribution modeling method and is used to transform the multidimensional environmental element characterization into standardized spatial distribution probability assessment results of marine mammals. Then, through a nonlinear signal conditioning function, the habitat suitability index is mapped to the initial navigation biological risk cost for adapting to the ship's path search.

3. The method for ship ecological route planning based on marine mammal prediction according to claim 2, characterized in that, The nonlinear signal conditioning function employs an exponential mapping, and the mapping process is as follows: ； in, This represents the initial biological risk cost of the voyage, where k is the normalized scaling factor for the time cost equivalent. HSI This indicates the habitat suitability index.

4. The method for ship ecological route planning based on marine mammal prediction according to claim 1, characterized in that, Based on the current speed and underwater noise radiation level, the initial navigation biological risk cost is dynamically adjusted to obtain an individualized navigation biological risk cost, including: ； in, V 0 , N 0 Here, V represents the reference speed and reference noise level, and N represents the current underwater noise radiation level. p, q This is an empirical correction factor. This represents the initial biological risk cost of the voyage. This indicates the biological risks and costs associated with individualized navigation.

5. The method for ship ecological route planning based on marine mammal prediction according to claim 1, characterized in that, The multi-objective comprehensive cost function is expressed as follows: ； in, To start from the current node i to adjacent nodes j The single-step comprehensive cost; The time cost required to traverse this grid segment; The fuel economy cost required to traverse this grid segment; For the target node j The biological risks and costs of individualized navigation at various locations; α, β, γ These are non-negative adjustable weighting coefficients.

6. The method for ship ecological route planning based on marine mammal prediction according to claim 1, characterized in that, Based on the aforementioned multi-objective comprehensive cost function, with the goal of minimizing the total cost, the optimal route from the current location to the destination is planned, including: With the goal of minimizing the total cumulative cost from the origin to the destination, a path search algorithm is used to solve the optimal route from the current location to the destination on a comprehensive cost grid map superimposed with individualized biological risk costs of navigation.

7. The method for ship ecological route planning based on marine mammal prediction according to claim 1, characterized in that, The multi-source spatiotemporal data is spatiotemporally aligned and fused, and unified to a regular grid to form a spatiotemporally consistent environmental variable field, including: Multi-source spatiotemporal data with different spatiotemporal resolutions are unified into a regular grid using a weighted averaging method with a time decay window, forming a spatiotemporally consistent environmental variable field. The regular grid covers the target sea area, and each grid node carries the fused multidimensional environmental variable values.

8. The method for ship ecological route planning based on marine mammal prediction according to claim 1, characterized in that, Also includes: The optimal route is displayed graphically, and auxiliary evaluation indicators for the route are output. During navigation, a hybrid mechanism combining periodic and event-triggered events is employed to re-predict and plan. When a confirmed real-time alarm for the presence of marine mammals is received, the data is fed back to the model for incremental learning and to update the model parameters. For the planned candidate routes, the cumulative exposure is calculated as an auxiliary assessment indicator of ecological impact, used to characterize the cumulative impact of the ship on marine mammal habitats along the entire route.

9. A ship ecological route planning system based on marine mammal prediction, employing the ship ecological route planning method based on marine mammal prediction as described in any one of claims 1-8, characterized in that, include: The data acquisition module is used to acquire multi-source spatiotemporal data of ships and sea areas; The spatiotemporal alignment module is used to perform spatiotemporal alignment and fusion of the multi-source spatiotemporal data and unify it into a regular grid to form a spatiotemporally consistent environmental variable field. The risk prediction module is used to utilize a pre-built marine biological risk prediction model, with the environmental variable field as input, to output the marine mammal habitat suitability index for each grid within a future preset time period, and to map the habitat suitability index to the initial navigation biological risk cost. The dynamic correction module is used to dynamically correct the initial navigation biological risk cost based on the current speed and underwater noise radiation level, so as to obtain an individualized navigation biological risk cost. The cost layer construction module is used to construct a multi-objective comprehensive cost function based on time cost, fuel economy cost, and the individualized biological risk cost of navigation; The optimal route planning module is used to plan the optimal route from the current location to the destination based on the multi-objective comprehensive cost function, with the goal of minimizing the total cost.

Images