Maritime search and rescue resource scheduling method based on improved A-Star algorithm
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
- Filing Date
- 2025-12-26
- Publication Date
- 2026-05-08
AI Technical Summary
[0006]a)海上险情如船舶倾覆等往往于恶劣气象状况相联系，现有方法对此类状况考虑较少难以满足恶劣气象状况下的搜救资源调度工作
Smart Images
Figure CN121998302A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for scheduling maritime search and rescue resources, and more particularly to a method for scheduling maritime search and rescue resources based on an improved A-Star algorithm. Background Technology
[0002] This section provides only background information relevant to this disclosure and is not necessarily prior art.
[0003] With the increasing frequency of marine economic activities and the continuous expansion of maritime transport, maritime safety has become a key issue in the shipping industry. Maritime search and rescue, as a crucial task for ensuring the safety of life and property at sea, is highly complex and requires timely action.
[0004] Traditional dispatching methods can provide effective support for command and decision-making in open seas or simple scenarios. However, in adverse sea conditions, poor visibility, or significant target drift, dynamic changes in environmental factors often severely reduce the integrity and reliability of available information, leading to resource dispatching schemes deviating from actual needs. Furthermore, real-world constraints such as variable weather and complex sea conditions can easily result in imbalanced resource allocation, response delays, or insufficient coverage, making the final dispatching scheme unable to meet the actual requirements of efficient search and rescue.
[0005] Currently, existing technologies for maritime search and rescue resource allocation have the following shortcomings:
[0006] a) Maritime emergencies such as ship capsizing are often associated with severe weather conditions. Existing methods take such conditions into account little and are insufficient to meet the needs of search and rescue resource allocation under severe weather conditions.
[0007] b) Existing methods are mostly based on heuristic learning, and the computational efficiency of such methods is usually related to parameters. In the complex sea conditions and weather scenarios of actual rescue operations, it is difficult to determine an optimal set of parameters. Incorrect parameter settings can lead to slow algorithm convergence, resulting in getting stuck in local optima and excluding the most efficient rescue path.
[0008] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0009] Purpose of the invention: The technical problem to be solved by the present invention is to provide a method for scheduling maritime search and rescue resources based on an improved A-Star algorithm, which addresses the shortcomings of the existing technology.
[0010] To address the aforementioned technical problems, this invention discloses a maritime search and rescue resource scheduling method based on an improved A-Star algorithm, comprising the following steps:
[0011] Step 1: Collect relevant data on the target area for maritime search and rescue;
[0012] Step 2: Rasterize the target area for maritime search and rescue to obtain nodes and node data;
[0013] Step 3: Based on the nodes and node data, generate search and rescue routes using the improved A-Star algorithm and distribute them in real time;
[0014] Step 4: Update the relevant data of the maritime search and rescue target area according to the preset cycle, and repeat Step 3 to update the search and rescue route.
[0015] Furthermore, the collection of relevant data on the maritime search and rescue target area mentioned in step 1 includes:
[0016] Information on meteorological conditions, status of rescue vessels, ocean currents, search and rescue area, and operational risk level.
[0017] Furthermore, step 2, which involves rasterizing the target area for maritime search and rescue, includes:
[0018] Step 2-1, Set the meridional resolution and latitudinal resolution And based on this, the target area is rasterized;
[0019] Step 2-2, treat grid points as nodes The node's attribute is set to the job risk level. ;
[0020] Steps 2-3 involve constructing an adjacency matrix, which means establishing connectivity between each node and its neighboring nodes, and calculating the basic movement cost based on the voyage distance, the rescue vessel's speed, and the external environmental influence vector. Specifically, it is expressed as follows:
[0021]
[0022] in, Indicates the number is nodes With number nodes The basic movement cost between them The distance between the two nodes. To control the speed of the rescue vessels, This is a vector representing the influence of the external environment, including ocean current speed and wind speed.
[0023] Furthermore, step 3, which involves generating a search and rescue route using the improved A-Star algorithm, includes:
[0024] Step 3-1: Set up the Open table as a list of nodes to be expanded, which is used to store all discovered but not yet processed nodes, and sort them by priority according to the magnitude of the expected cost function value of the nodes;
[0025] Step 3-2: Set up the Close table as a list of expanded nodes. Take the node with the smallest expected cost function value from the Open table, expand the adjacent nodes of the node, and put the node into the Close table.
[0026] Step 3-3: Finally, based on the nodes in the Close table, sort them according to the connection order to obtain the search and rescue route.
[0027] Furthermore, the expected cost function value of the node mentioned in step 3-1 is calculated as follows:
[0028] Based on the travel time, design the expected cost function for each node and calculate the total cost. The details are as follows:
[0029]
[0030] Among them, the actual cost From the starting point to the node Sailing time, estimated cost For nodes Time to reach the destination.
[0031] Furthermore, the aforementioned expected cost The values are obtained by predicting historical sea state data using a random forest model, as shown below:
[0032]
[0033] in, For nodes To the target point The straight-line distance The cost of sea state impact predicted by machine learning models. For nodes The level of operational risk. and These are the weighting coefficients.
[0034] Furthermore, the expansion of the neighboring nodes of the node described in step 3-2 includes:
[0035] Step 3-2-1: Calculate the total cost of all nodes adjacent to the current node, and select the node with the smallest total cost to expand into an adjacent node;
[0036] Step 3-2-2: When the number of nodes with the minimum total cost is greater than 1, select the dynamic weighting factor. The smallest node, where the dynamic weight factor is... Calculated based on wind and wave data, as shown below:
[0037]
[0038] in, for Standard deviation of wind and waves at all times The standard deviation of historical fluctuations.
[0039] Furthermore, the cost of the aforementioned sea state impact The specific prediction method is as follows:
[0040] Construct a training dataset, in which the data sources include historical meteorological records, real-time meteorological elements, ocean current distribution characteristics, and wind and wave fluctuation parameters;
[0041] By statistically analyzing massive historical navigation trajectories, the deviation between the actual and theoretical speeds of ships under different weather, ocean current, and wave intensity conditions is calculated, and this speed deviation is used as a training label for supervised learning.
[0042] The random forest model is trained using the training dataset and training labels to obtain the prediction model;
[0043] Real-time collected meteorological, ocean current, and wave intensity data are used as input features and fed into a pre-trained prediction model to obtain the cost of sea state impact under the current environment. .
[0044] Furthermore, the real-time delivery mentioned in step 3 includes:
[0045] The search and rescue route obtained in step 3, as well as the estimated time and risk level information, are sent to the ship terminal carrying out the rescue mission via satellite communication link.
[0046] Furthermore, step 4, which involves updating the relevant data for the maritime search and rescue target area, includes:
[0047] Step 4-1: Receive real-time updates on weather conditions, rescue vessel status, ocean currents, search and rescue area, and operational risk level.
[0048] Step 4-2: Update the node attributes based on the update data received in Step 4-1;
[0049] Step 4-3: Using the ship's current position as the new starting point, repeat step 3 to obtain a new search and rescue route; if the total cost of the new search and rescue route is reduced by more than the set threshold ΔC, then the current search and rescue route is replaced and sent out in real time.
[0050] Beneficial effects:
[0051] 1. This invention employs a search and rescue resource scheduling method with the ability to dynamically update meteorological conditions, thus solving the problem of interference from severe weather conditions on search and rescue resource scheduling.
[0052] 2. This invention adopts an improved A-Star algorithm based on Dijstra's algorithm and combined with machine learning, which addresses the difficulty of parameter setting in heuristic learning methods and solves the problem of easy local optima. Attached Figure Description
[0053] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.
[0054] Figure 1 This is a schematic diagram of the process of the present invention.
[0055] Figure 2 This is a schematic diagram illustrating the implementation process of the A-Star algorithm in one embodiment.
[0056] Figure 3 This is a schematic diagram of path comparison and sea state field in one embodiment.
[0057] Figure 4 This is a diagram illustrating the comparison of performance metrics in one embodiment. Detailed Implementation
[0058] The "Maritime Search and Rescue Resource Scheduling Method Based on Improved A-Star Algorithm" provided by this invention solves the above-mentioned technical defects:
[0059] a) Regarding the problem of severe weather conditions interfering with the scheduling of search and rescue resources, the present invention adopts a search and rescue resource scheduling method with the ability to dynamically update weather conditions.
[0060] b) To address the problem of difficult parameter setting and easy local optima in heuristic learning methods, this invention adopts an improved A-Star algorithm based on Dijstra's algorithm and combined with machine learning.
[0061] The core of this invention is a method for scheduling maritime search and rescue resources based on an improved A-Star algorithm, the overall flowchart of which is shown below:
[0062] Figure 1 The flowchart of this invention is as follows:
[0063] Step 1, Algorithm Initialization:
[0064] The algorithm begins by determining the coordinates of the search and rescue center and the approximate location of the target to be rescued, based on the accident alarm information received by the search and rescue center and using GNSS data. Then, using historical meteorological data, real-time weather fields, and ocean current observation data, combined with the search and rescue mission level and equipment performance, the search and rescue area parameters and vessel status parameters are initialized. The initialization phase includes the following:
[0065] Meteorological condition analysis: Based on meteorological elements such as wind speed, wave height, ocean current speed and visibility, a model of areas with severe search and rescue conditions is established, and a distribution map of search and rescue restricted areas and areas where special vessels can operate is generated.
[0066] Initialize rescue vessel status information: including the vessel's current location, speed, and type.
[0067] Ocean current information construction: Based on real-time ocean current data and historical average current field data, an ocean current velocity field is constructed for path cost correction.
[0068] Search and rescue area determination: Based on the accident center point and the drift caused by ocean currents, the initial search and rescue area is determined. Areas with severe search and rescue conditions are identified based on meteorological information and marked as "priority areas for special search and rescue vessels".
[0069] Step 2, Rasterization of Marine Route Data:
[0070] Complete the gridded preprocessing of the search and rescue area and establish a node database. Node information includes coordinates, grid number, ocean current direction and velocity, and operational risk level (refer to: Guidelines for Formal Safety Assessment (FSA) for use in the IMO rule-making process, etc.).
[0071] The sea route data in the accident area was rasterized for use in subsequent path optimization algorithms. The latitude and longitude resolution was first set. , Then, the grid points are used as nodes n in the A-Star algorithm, and the node attribute is the job risk level. , A value of 1 indicates that only special search and rescue vessels are permitted to pass through this node; External environmental influence vector This includes ocean current speed and wind speed. Finally, an adjacency matrix is constructed for each node. Establish connectivity with other adjacent nodes and calculate the basic movement cost. :
[0072]
[0073] in, The distance between the two nodes. Speed of rescue vessels.
[0074] Step 3: Solving the high-efficiency route using the A-Star algorithm
[0075] The implementation process of the A-Star algorithm is as follows: Figure 2 As shown, the details are as follows:
[0076] The Open table is a list of nodes to be expanded. It stores all discovered but not yet processed nodes and prioritizes them according to the expected cost function size of the nodes.
[0077] The Close list is a list of expanded nodes. The node with the lowest cost is taken from the Open list, and its neighboring nodes are expanded. Then, the node is added to the Close list.
[0078] Node cost function design: The total cost function of the A-Star algorithm is defined as follows:
[0079]
[0080] Wherein, the actual cost G(n) is the travel time from the starting point to node n, and the expected cost H(n) is the travel time from node n to the target point.
[0081] The estimated cost H(n) is obtained by predicting historical sea state data using a random forest model to improve the estimation accuracy under complex sea states. The cost calculation formula is as follows:
[0082]
[0083] in, Let n be the straight-line distance from node n to the target point. R(n) represents the cost of sea state impact predicted by the machine learning model, and R(n) represents the operational risk level of node n. These are weighting factors used to balance the effects of distance and sea state.
[0084] Cost of Sea State Impact The determination of this depends on machine learning methods driven by multi-source data, as follows:
[0085] First, a high-quality training dataset is constructed, with data sources covering historical meteorological records, real-time meteorological elements, ocean current distribution characteristics, and wind and wave fluctuation parameters.
[0086] During the data processing phase, by statistically analyzing massive historical navigation trajectories, the deviation between the actual and theoretical speeds of ships under different weather, ocean current, and wave intensity conditions is calculated, and this speed deviation is used as a training label for supervised learning.
[0087] Subsequently, the random forest algorithm was trained using the aforementioned multi-source environmental feature data and corresponding bias labels to establish a mapping model between environmental parameters and navigation impacts.
[0088] Finally, after determining At the same time, real-time collected environmental data such as wind speed, wave height, and ocean currents are used as input features and fed into a pre-trained random forest model. The predicted value output by the model is the cost of the impact of sea state under the current environment. .
[0089] During node expansion, the algorithm prioritizes expanding the node with the smallest F(n). When multiple candidate nodes exist, a dynamic weighting factor is introduced. (t) Select candidate nodes.
[0090]
[0091] in, Let t be the standard deviation of the wind and waves. The standard deviation of historical wind and wave patterns is used. The algorithm tends to choose stable paths rather than the shortest paths.
[0092] When the number of nodes with the minimum total cost is greater than 1, select the dynamic weight factor. The smallest node, A larger value indicates more violent fluctuations in the current environment, while a smaller value indicates a more stable environment.
[0093] Step 4, Send rescue route
[0094] Once the optimal route is calculated, the system transmits the route coordinate sequence to the corresponding rescue vessel terminal via satellite communication link. Each route includes node coordinates, estimated travel time, risk level, and adjustable waypoint information. The vessel automatically loads and displays the route and enters search and rescue execution mode.
[0095] Step 5, Rescue Route Update
[0096] To address changes in sea state and the drift of search and rescue targets, this method employs a dynamic replanning mechanism. The system updates environmental data and recalculates the locally optimal route every 5 minutes.
[0097] The update process is as follows: receive real-time ocean current and meteorological data updates; update node attributes; re-execute the A-Star search with the ship's current position as the new starting point; if the total cost of the new route is reduced by more than the set threshold ΔC, replace the currently executed route; finally, send the new route to the corresponding rescue ship.
[0098] Example:
[0099] To verify the effectiveness of the "maritime search and rescue resource scheduling method based on the improved A-Star algorithm" proposed in this invention, a high-fidelity maritime search and rescue simulation verification environment was constructed based on historical meteorological and hydrological datasets of a certain sea area.
[0100] The regional grid resolution was set to 1 nm × 1 nm, and the environmental data (wind speed, wave height, ocean current speed) update cycle was set to 5 minutes to simulate the dynamic changes of real sea conditions.
[0101] Performance verification of path planning under complex sea conditions: Under simulated severe weather conditions (average wave height exceeding 3 meters, local wind force reaching level 8), the improved A-Star algorithm based on machine learning cost prediction and dynamic weights proposed in this invention is compared with the traditional A-Star algorithm based on Euclidean distance. The results are as follows: Figure 3 and Figure 4 As shown.
[0102] Path avoidance effect such as Figure 3 As shown: the black areas represent areas inaccessible to search and rescue vessels, such as islands and reefs. The heatmap in the image displays predictions generated by the random forest model. (Sea state impact cost) distribution, with dark red areas representing high-risk areas with high winds and waves. Traditional algorithm (blue dashed line): Because it only aims for the shortest distance, the planned route directly passes through the core area with the worst sea conditions. Although the theoretical voyage is the shortest, in actual operation, it is very easy to cause the ship to stall or even capsize.
[0103] The method of this invention (solid red line): by introducing the cost of sea state impact. and operational risk level During the search process, the algorithm keenly identifies high-risk environments ahead and automatically generates a smooth trajectory that detours towards low sea state areas (light blue areas).
[0104] Comprehensive performance statistics, such as Figure 4 As shown, quantitative analysis reveals that although the path generated by this invention increases the physical distance by approximately 20.3% compared to traditional methods, the average comprehensive risk index of the area traversed by the path decreases by 37.9%. The average comprehensive risk index is calculated by averaging the cost of sea state impact on the path. This means that in actual search and rescue operations, rescue vessels can maintain a more stable speed, avoiding additional time and equipment losses caused by wind and waves, fully verifying the algorithm's safety and robustness in complex and harsh environments.
[0105] The effectiveness of the dynamic weight control strategy was validated in simulating rapid changes in sea state, such as sudden gusts causing wind and wave standard deviation. In scenarios with sudden increases, the dynamic weighting factor proposed in this invention is applied. The adjustment mechanism and experimental results are as follows: When the monitoring system detects the standard deviation of environmental wind and waves... Significantly higher than the historical average At that time, the system automatically increases the weighting factor. (In this experiment, the version was increased from 1.0 to 1.8).
[0106] The path demonstrated by the method of this invention shows that, after introducing dynamic weights, the algorithm tends to select deterministic gains due to stable sea conditions, rather than simply pursuing theoretical gains in distance. This achieves a balance between search and rescue efficiency and operational safety. The simulation data above shows that the method proposed in this invention can not only avoid static geographical obstacles but also make intelligent responses to dynamic meteorological and hydrological environments, significantly improving the intelligence level and reliability of maritime search and rescue resource scheduling compared to existing technologies.
[0107] In its specific implementation, this application provides a computer storage medium and a corresponding data processing unit. The computer storage medium is capable of storing a computer program, which, when executed by the data processing unit, can run the invention's content regarding a maritime search and rescue resource scheduling method based on an improved A-Star algorithm, as well as some or all of the steps in various embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0108] Those skilled in the art will clearly understand that the technical solutions in the embodiments of the present invention can be implemented using computer programs and their corresponding general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of computer programs, i.e., software products. These computer program software products can be stored in a storage medium and include several instructions to cause a device containing a data processing unit (which may be a personal computer, server, microcontroller, MCU, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.
[0109] This invention provides an idea and method for scheduling maritime search and rescue resources based on an improved A-Star algorithm. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.
Claims
1. A method for scheduling maritime search and rescue resources based on an improved A-Star algorithm, characterized in that, Includes the following steps: Step 1: Collect relevant data on the target area for maritime search and rescue; Step 2: Rasterize the target area for maritime search and rescue to obtain nodes and node data; Step 3: Based on the nodes and node data, generate search and rescue routes using the improved A-Star algorithm and distribute them in real time; Step 4: Update the relevant data of the maritime search and rescue target area according to the preset cycle, and repeat Step 3 to update the search and rescue route.
2. The maritime search and rescue resource scheduling method based on the improved A-Star algorithm according to claim 1, characterized in that, The collection of relevant data on the maritime search and rescue target area mentioned in step 1 includes: Information on meteorological conditions, status of rescue vessels, ocean currents, search and rescue area, and operational risk level.
3. A maritime search and rescue resource scheduling method based on an improved A-Star algorithm according to claim 2, characterized in that, Step 2, which involves rasterizing the target area for maritime search and rescue, includes: Step 2-1, Set the meridional resolution and latitudinal resolution And based on this, the target area is rasterized; Step 2-2, treat grid points as nodes The node's attribute is set to the job risk level. ; Steps 2-3 involve constructing an adjacency matrix, which means establishing connectivity between each node and its neighboring nodes, and calculating the basic movement cost based on the voyage distance, the rescue vessel's speed, and the external environmental influence vector. .
4. A maritime search and rescue resource scheduling method based on an improved A-Star algorithm according to claim 3, characterized in that, Step 3, which describes generating search and rescue routes using the improved A-Star algorithm, includes: Step 3-1: Set up the Open table as a list of nodes to be expanded, which is used to store all discovered but not yet processed nodes, and sort them by priority according to the magnitude of the expected cost function value of the nodes; Step 3-2: Set up the Close table as a list of expanded nodes. Take the node with the smallest expected cost function value from the Open table, expand the adjacent nodes of the node, and put the node into the Close table. Step 3-3: Finally, based on the nodes in the Close table, sort them according to the connection order to obtain the search and rescue route.
5. A maritime search and rescue resource scheduling method based on an improved A-Star algorithm according to claim 4, characterized in that, The expected cost function value of the node mentioned in step 3-1 is calculated as follows: Based on the travel time, design the expected cost function for each node and calculate the total cost. The details are as follows: ； Among them, the actual cost From the starting point to the node Sailing time, estimated cost For nodes Time to reach the destination.
6. A method for scheduling maritime search and rescue resources based on an improved A-Star algorithm according to claim 5, characterized in that, The estimated cost It is obtained by predicting historical sea state data using a random forest model.
7. A method for scheduling maritime search and rescue resources based on an improved A-Star algorithm according to claim 6, characterized in that, The expansion of the neighboring nodes of the node described in step 3-2 includes: Step 3-2-1: Calculate the total cost of all nodes adjacent to the current node, and select the node with the smallest total cost to expand into an adjacent node; Step 3-2-2: When the number of nodes with the minimum total cost is greater than 1, select the dynamic weighting factor. The smallest node, where the dynamic weight factor is... Calculated based on wind and wave data.
8. A method for scheduling maritime search and rescue resources based on an improved A-Star algorithm according to claim 7, characterized in that, The cost of sea state impact The specific prediction method is as follows: Construct a training dataset, in which the data sources include historical meteorological records, real-time meteorological elements, ocean current distribution characteristics, and wind and wave fluctuation parameters; By statistically analyzing massive historical navigation trajectories, the deviation between the actual and theoretical speeds of ships under different weather, ocean current, and wave intensity conditions is calculated, and this speed deviation is used as a training label for supervised learning. The random forest model is trained using the training dataset and training labels to obtain the prediction model; Real-time collected meteorological, ocean current, and wave intensity data are used as input features and fed into a pre-trained prediction model to obtain the cost of sea state impact under the current environment. .
9. A method for scheduling maritime search and rescue resources based on an improved A-Star algorithm according to claim 8, characterized in that, The real-time delivery mentioned in step 3 includes: The search and rescue route obtained in step 3, as well as the estimated time and risk level information, are sent to the ship terminal carrying out the rescue mission via satellite communication link.
10. A method for scheduling maritime search and rescue resources based on an improved A-Star algorithm according to claim 9, characterized in that, Step 4, which involves updating the relevant data for the maritime search and rescue target area, includes: Step 4-1: Receive real-time updates on weather conditions, rescue vessel status, ocean currents, search and rescue area, and operational risk level. Step 4-2: Update the node attributes based on the update data received in Step 4-1; Step 4-3: Using the ship's current position as the new starting point, repeat step 3 to obtain a new search and rescue route; if the total cost of the new search and rescue route is reduced by more than the set threshold ΔC, then the current search and rescue route is replaced and sent out in real time.