A method and device for planning a maritime search and rescue path

By constructing marine dynamic environment model field and multi-agent collaborative algorithm, the maritime search and rescue path planning is optimized, and the problem of insufficient size and resource allocation in the existing technology is solved, thus achieving more efficient search and rescue path planning and faster Search success rate.

CN118670401BActive Publication Date: 2024-11-12SHANDONG UNIV OF SCI & TECH +1
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202411164393.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-12
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

The existing maritime search and rescue path planning theory is difficult to determine the better search task area size and search resource allocation ratio, resulting in the inability to obtain a better search path.

Method used

By building a marine dynamic environment model field, we determine the total number of available rescue ships and the size of the search task area, and use multi-agent collaborative algorithm to allocate search task points to optimize search path planning, including determining the search starting point and dividing the search task area , calculate the actual search time and Manhattan distance to allocate task points, and optimize the objective function to reassign task points.

Benefits of technology

It realizes global optimal search task point allocation and path planning, reduces the workload and search time of rescue ships, and improves search efficiency and success rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118670401B_ABST
    Figure CN118670401B_ABST
Patent Text Reader

Abstract

The present invention provides a method and device for planning a search and rescue path at sea, belonging to the field of maritime rescue technology, the method comprising: determining a search and rescue area according to a target in distress at sea, obtaining ocean dynamics environment data and image data of the search and rescue area, and constructing an environmental model field of the search and rescue area using the ocean dynamics environment data and the image data; determining the total amount and size of the search task area according to the total amount of available rescue ships, and then allocating each of the search task points to the search task area; and planning a search path for the corresponding available rescue ship according to each of the search task areas using a multi-agent collaborative algorithm. The present invention allocates search task points to each available rescue ship with a global optimal orientation by searching the relative position and distance between the starting point and the available rescue ship, and provides path planning to achieve global coverage.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of maritime rescue, and in particular, relates to a maritime search and rescue path planning method and device. Background Art

[0002] With the continuous development of information technology and artificial intelligence, the path planning technology for searching mobile targets at sea can also be continuously innovated and improved. This technology can be widely used in the fields of maritime rescue, marine environmental monitoring, maritime traffic control and maritime battlefield target search, and can bring more reliable and efficient protection for the safety and development of the maritime industry.

[0003] At present, maritime search and rescue path planning is mainly based on search planning theory, that is, maximizing the probability of successful search (POS), probability of containment (POC) and probability of detection (POD); the expression of search theory is POS = POC × POD .

[0004] However, in the process of path planning, search planning theory focuses on the optimal search path to improve search efficiency and save search time. However, considering the navigation speed of rescue ships and the time spent searching, search planning theory is relatively weak in determining the optimal search area and allocating search resources. It usually cannot obtain a better search task area size and the distribution ratio of search task points, and thus cannot obtain a better search path. Summary of the invention

[0005] The embodiment of the present invention provides a method and device for maritime search and rescue path planning, which aims to solve the problem that the existing search planning theory cannot obtain a more optimal search task area size and search task point distribution ratio.

[0006] In view of the above problems, the technical solution proposed by the present invention is:

[0007] In a first aspect, a method for maritime search and rescue path planning is provided, the method comprising:

[0008] Determine a search and rescue area according to a target in distress at sea, obtain ocean dynamics environment data and image data of the search and rescue area, and construct an environmental model field of the search and rescue area using the ocean dynamics environment data and the image data, wherein the environmental model field includes the total amount of available rescue ships, multiple grid areas, and multiple search task points;

[0009] Determine the total amount of available rescue ships according to the environmental model field, determine the total amount and size of the search task area according to the total amount of available rescue ships, and then allocate each of the search task points to the search task area;

[0010] A multi-agent collaborative algorithm is used to plan a search path for the corresponding available rescue ship according to each search task area.

[0011] Further, the total amount and size of the search mission area are determined according to the total amount of the available rescue ships, and then each of the search mission points is allocated to the search mission area, including:

[0012] Determining the search starting point for each of the available rescue vessels;

[0013] Dividing the size of the search mission area according to the proportional relationship between the total area of ​​the environmental model field and the speed of each of the available rescue ships;

[0014] The search task points are allocated to the search task area according to the distance relationship between each search task point and each search starting point.

[0015] Further, the determining of the search starting point of each available rescue vessel comprises:

[0016] Determining the total amount of the search mission area according to the total amount of available rescue ships;

[0017] The search task point closest to the environmental model field for each available rescue ship is used as the search starting point.

[0018] Furthermore, the size of the search mission area is divided according to the proportional relationship between the total area of ​​the environment model field and the speed of each of the available rescue ships, including:

[0019] Calculating the total theoretical search time consumed by each of the available rescue ships to search the environmental model field;

[0020] The calculation formula for the theoretical total search time is as follows:

[0021]

[0022] Where: T 总 Total time spent searching for theories;

[0023] A is the total area of ​​the environmental model field;

[0024] S is the speed of the available rescue vessel;

[0025] Calculate the actual search time of each available rescue vessel according to each theoretical total search time;

[0026] The calculation formula for the actual search time is as follows:

[0027]

[0028] Where: T 实 The actual search for available rescue vessels takes time;

[0029] T 行 The time it takes for available rescue vessels to travel from their original location to the search starting point;

[0030] According to the actual search time of each available rescue ship and the total amount of the search task area, the environment model field is divided into the total area of ​​each search task area in proportion and the task capacity is determined.

[0031] Furthermore, allocating the search task points to the search task area according to the distance relationship between each search task point and each search starting point includes:

[0032] Calculating the Manhattan distance between the remaining search task points and the search starting point of each available rescue ship;

[0033] The remaining search mission points are allocated to the search mission area closest to the available rescue ships according to the Manhattan distance of each available rescue ship.

[0034] Furthermore, the determining the total amount and size of the search mission area according to the total amount of the available rescue ships, and then allocating each of the search mission points to the search mission area, further includes:

[0035] Optimizing the search task points of each of the search task areas, including:

[0036] Calculating the objective function of each of the search task areas to obtain the sum of the distances of each of the search task areas;

[0037] When the sum of the distances of the search task area exceeds a preset threshold, the search task points of the search task area exceeding the preset threshold are reallocated to the nearest search task area.

[0038] Furthermore, the calculation formula of the objective function is as follows:

[0039]

[0040] Where: J is the objective function;

[0041] k is the total amount of the search mission area;

[0042] C i For the i Search mission area;

[0043] x Search mission points for searching mission areas;

[0044] p i is the search starting point;

[0045] d ( x , p i ) is the search task point to i The closest Manhattan distance to the search task area.

[0046] Furthermore, the multi-agent collaborative algorithm includes: a path planning model, an environment modeling and action space model, a reward function model and a DQN model.

[0047] Furthermore, the calculation formula of the reward function model is as follows:

[0048]

[0049] Where: R exp Exploration reward: refers to the exploration reward obtained when a rescue ship enters a new grid area;

[0050] R limit Restricted state penalty: refers to the corresponding penalty given when the available rescue ship crosses the boundary or enters the obstacle area;

[0051] R cov Duplicate coverage penalty: refers to the corresponding penalty given when a rescue ship enters the searched grid area;

[0052] R exsteps Termination state penalty: The available rescue ship's sailing distance exceeds the preset maximum learning sailing distance. N max When the rescue ship is available, it will be placed in the terminated state. S exsteps and impose corresponding punishment;

[0053] R unit Fixed reward for completing full coverage: refers to the fixed reward obtained when the rescue ship completes the search mission area;

[0054] R pre Efficient Path Reward: refers to the additional reward obtained when a rescue ship completes the search mission area with a number of turns less than the preset turn threshold.

[0055] In a second aspect, a maritime search and rescue path planning device is provided, the device comprising:

[0056] A model field construction module is configured to determine a search and rescue area according to a target in distress at sea, obtain ocean dynamics environment data and image data of the search and rescue area, and use the ocean dynamics environment data and the image data to construct an environmental model field of the search and rescue area, wherein the environmental model field includes the total amount of available rescue ships, multiple grid areas, and multiple search task points;

[0057] An allocation module is configured to determine the total amount and size of the search task area according to the total amount of the available rescue ships, and then allocate each of the search task points to the search task area;

[0058] The planning module is configured to plan a search path for the corresponding available rescue ship according to each search task area using a multi-agent collaborative algorithm.

[0059] The beneficial effects of the above technical solution provided by the embodiment of the present invention include at least: based on the created environment model field, by searching the relative position and distance between the starting point and the available rescue ships, the search task points are assigned to each available rescue ship with global optimization as the guide, and path planning is provided for each available rescue ship to achieve global coverage. This can reduce the workload of each available rescue ship and save search time.

[0060] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a flowchart of a method for planning a maritime search and rescue path disclosed in the present invention;

[0062] Figure 2 It is a flow chart of a method for maritime search and rescue path planning disclosed in the present invention;

[0063] Figure 3 is a flowchart of step S10 disclosed in the present invention;

[0064] Figure 4 is a flowchart of step S20 disclosed in the present invention;

[0065] Figure 5 is a flowchart of step S21 disclosed in the present invention;

[0066] Figure 6 is a flowchart of step S22 disclosed in the present invention;

[0067] Figure 7 It is a rendering of the total area of ​​the divided search task area disclosed in the present invention;

[0068] Figure 8 is a flowchart of step S23 disclosed in the present invention;

[0069] Fig. 9 is a flowchart of step S24 disclosed in the present invention;

[0070] Fig.10 It is a rendering of the planning search path disclosed in the present invention;

[0071] Fig.11 It is a structural schematic diagram of the multi-agent collaborative algorithm disclosed in the present invention;

[0072] Fig.12 It is a comparison diagram of multi-method simulation experiments disclosed in the present invention;

[0073] Fig.13 is a graph showing a change in search success rate disclosed in the present invention;

[0074] Fig.14 is a graph of search time cost variation disclosed in the present invention;

[0075] Fig.15 It is a structural schematic diagram of a marine search and rescue path planning device disclosed in the present invention.

[0076] Description of reference numerals: 100, model field construction module; 200, allocation module; 300, planning module. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0079] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0080] Please refer to Figure 1 , Figure 2 and Fig.10 As shown, the present invention provides a method for maritime search and rescue path planning, comprising the following steps:

[0081] S10, determining a search and rescue area according to the target in distress at sea, obtaining ocean dynamic environment data and image data of the search and rescue area, and constructing an environmental model field of the search and rescue area using the ocean dynamic environment data and the image data, wherein the environmental model field includes the total number of available rescue ships, multiple grid areas, and multiple search task points;

[0082] S20, determining the total amount and size of the search mission area according to the total amount of the available rescue ships, and then allocating each of the search mission points to the search mission area;

[0083] S30, using a multi-agent collaborative algorithm to plan a search path for the corresponding available rescue ship according to each search mission area.

[0084] Please refer to Figure 2 and Figure 3 As shown, in a possible implementation manner, S10 includes S11-S16.

[0085] S11, determining a search and rescue area according to the target in distress at sea, obtaining ocean dynamic environment data and image data of the search and rescue area, and using a particle swarm algorithm to analyze the drift motion of the target in distress at sea according to the ocean dynamic environment data to obtain a plurality of random particles.

[0086] The calculation formula for the drift motion of the target in distress at sea is as follows:

[0087]

[0088] Where: M Quality of targets for distress at sea;

[0089] The speed of drifting motion of a target in distress at sea;

[0090] The time of drifting motion of the target in distress at sea;

[0091] For the Koch force;

[0092] It is the force of sea surface wind on the distressed objects at sea;

[0093] It is the force of ocean current on objects in distress at sea;

[0094] It is the force exerted by waves on objects in distress at sea.

[0095] For example, the ocean dynamic environment data includes ocean wind field data and ocean current field data. The image data may be electronic chart data or remote sensing image data.

[0096] It should be noted that the drift motion of a target in distress at sea is a complex synthetic motion under the action of external forces such as wind, waves and currents. In order to simplify the force calculation of a target in distress at sea, the present invention only considers the drift motion of a target in distress at sea under the combined action of wind, current, waves and Coriolis force on the sea surface.

[0097] S12, using a Lagrangian pursuit algorithm to calculate the displacement of each of the random particles and determine the position of each of the random particles.

[0098] The calculation formula of the displacement of the random particles is as follows:

[0099]

[0100] Where: dX The moving position of the drifting motion of the target in distress at sea;

[0101] X t is the displacement of random particles;

[0102] A is the drift coefficient;

[0103] B is the diffusion coefficient;

[0104] Z n Is a random number.

[0105] S13, using a Monte Carlo algorithm to perform N repeated tests on each of the random particles according to the position of each of the random particles, to obtain a drift motion trajectory and a distribution range of each of the random particles.

[0106] S14, using the distribution range of each random particle as a search task point.

[0107] S15, obtaining available rescue ships in a preset radius unit of the search and rescue area, correcting the sweep width of each available rescue ship using a correction coefficient, and creating the grid area according to the sweep width of each available rescue ship.

[0108] For example, the following table 1 shows the search and rescue performance information that can be used by available rescue ships searching in preset radius units in the search and rescue area. The following table 2 shows the correction coefficients corresponding to the sweep width correction of each available rescue ship under different visibility.

[0109] Table 1: Search and rescue performance information of available rescue vessels

[0110]

[0111] Table 2: Correction factors for scan width at different visibility levels

[0112]

[0113] S16, calculating the inclusion probability value of the corresponding grid area according to each search task point to form an inclusion probability layer, and superimposing the inclusion probability layer with the image data to obtain an environment model field.

[0114] Please refer to Figure 4 and Figure 7 As shown, in a possible implementation manner, S20 includes S21-S24.

[0115] S21, determining a search starting point for each of the available rescue vessels.

[0116] S22, dividing the size of the search mission area according to the proportional relationship between the total area of ​​the environmental model field and the speed of each of the available rescue ships.

[0117] S23, allocating the search task points to the search task areas according to the distance relationship between each search task point and each search starting point.

[0118] S24, optimizing the search task points of each of the search task areas.

[0119] Specifically, by quantifying the spatial distances of all search task points, the similarities and differences of each search task point can be analyzed. At the same time, the search and rescue area is effectively discretized, which can not only facilitate the evaluation and distinction of the relative position relationship between the search task points during the search, but also accurately calculate the geometric distance between the search task points, providing accurate data support for path planning.

[0120] Please refer to Figure 5 and Figure 7As shown, in a possible implementation manner, S21 includes S211 - S212.

[0121] S211, determining the total number of the search mission area according to the total number of available rescue ships.

[0122] S212: taking the search task point of each available rescue ship that is closest to the environmental model field as the search starting point.

[0123] Specifically, by determining a reasonable search starting point, while ensuring the search quality, it is possible to improve search efficiency, save search time, and minimize losses.

[0124] Please refer to Figure 6 and Figure 7 As shown, in a possible implementation, S22 includes S221-S223.

[0125] S221, calculating the total theoretical search time consumed by each of the available rescue ships to search the environmental model field.

[0126] The calculation formula for the theoretical total search time is as follows:

[0127]

[0128] Where: T 总 Total time spent searching for theories;

[0129] A is the total area of ​​the environmental model field;

[0130] S is the speed of the available rescue vessel;

[0131] S222, calculating the actual search time of each available rescue vessel according to each theoretical total search time;

[0132] The calculation formula for the actual search time is as follows:

[0133]

[0134] Where: T 实 The actual search for available rescue vessels takes time;

[0135] T 行 The time it takes for available rescue vessels to travel from their original location to the search starting point;

[0136] S223, dividing the total area of ​​each search task area in proportion and determining the task capacity according to the actual search time of each available rescue ship and the total amount of the search task area.

[0137] Specifically, the size of the search mission area is divided according to the relationship between area, speed and time, so that the task capacity of each search mission area matches the actual search time, ensuring that each available rescue ship can complete the search of the search mission area within the allowed time, thereby achieving a balanced and comprehensive search of the entire search and rescue area.

[0138] Please refer to Figure 8 As shown, in a possible implementation, S23 includes S231 - S232.

[0139] S231, calculating the Manhattan distance between the remaining search task points and the search starting point of each available rescue ship;

[0140] S232, allocating the remaining search mission points to the nearest search mission area according to the Manhattan distance of each available rescue ship.

[0141] Specifically, allocating search mission points by Manhattan distance can ensure that each search mission point is effectively assigned to the search mission area and balance the workload of each available rescue ship. This not only optimizes the allocation of resources, but also helps to improve the efficiency and effectiveness of the entire rescue operation.

[0142] Please refer to Fig. 9 As shown, in a possible implementation, S24 includes S241 - S242.

[0143] S241, calculating the objective function of each of the search task areas to obtain the sum of the distances of each of the search task areas;

[0144] S242, when the sum of the distances of the search task area exceeds a preset threshold, reallocate the search task points of the search task area exceeding the preset threshold to the nearest search task area.

[0145] Specifically, by continuously updating the allocation of search task points, the allocation of search task points in the search task area can be gradually optimized, thereby improving the search efficiency and potentially increasing the success rate of rescue.

[0146] In a possible implementation, the calculation formula of the objective function is as follows:

[0147]

[0148] Where: J is the objective function;

[0149] k is the total amount of the search mission area;

[0150] C i For the i Search mission area;

[0151] x Search mission points for searching mission areas;

[0152] p i is the search starting point;

[0153] d ( x , p i ) is the search task point to i The closest Manhattan distance to the search task area.

[0154] In a possible implementation, the multi-agent collaborative algorithm includes: a path planning model, an action space model, a reward function model and a DQN model.

[0155] In a possible implementation, the calculation formula of the reward function model is as follows:

[0156]

[0157] Where: R exp Exploration reward: refers to the exploration reward obtained when a rescue ship enters a new grid area;

[0158] R limit Restricted state penalty: refers to the corresponding penalty given when the available rescue ship crosses the boundary or enters the obstacle area;

[0159] R cov Duplicate coverage penalty: refers to the corresponding penalty given when a rescue ship enters the searched grid area;

[0160] R exsteps Termination state penalty: The available rescue ship's sailing distance exceeds the preset maximum learning sailing distance. N max When the rescue ship is available, it will be placed in the terminated state. S exsteps and impose corresponding punishment;

[0161] R unitFixed reward for completing full coverage: refers to the fixed reward obtained when the rescue ship completes the search mission area;

[0162] R pre Efficient Path Reward: refers to the additional reward obtained when a rescue ship completes a mission area with a number of turns less than the preset turning threshold with a number of turns less than the preset turning threshold.

[0163] Please refer to Fig.11 As shown, in one possible implementation, the DQN model is constructed as follows:

[0164] (1) Determine the state space and action space:

[0165] State space: The environmental state of the available rescue ship, including position, speed, POC value and obstacle position, etc.

[0166] Action space: The actions that the available rescue ships can perform, such as up, down, left, and right.

[0167] (2) Design of deep neural network structure:

[0168] Input layer: corresponds to the dimension of the state space.

[0169] Hidden layer: The present invention designs a three-layer deep neural network, each layer has 128 neurons, and uses the ReLU activation function.

[0170] Output layer: corresponds to the size of the action space, that is, there are 4 output nodes, each of which outputs the expected Q value of an action.

[0171] (3) Set learning parameters:

[0172] Learning rate: The initial value is 0.01, which controls the step size of the neural network weight update.

[0173] Decay factor: 0.9, used to discount future rewards.

[0174] Exploration rate : 1, the probability that the agent chooses a random action at the beginning to achieve exploration.

[0175] (4) Experience replay mechanism:

[0176] Experience memory library: stores historical experience data; for example, the present invention is set to 10,000.

[0177] Sample size: the size of the experience data randomly extracted from the experience replay memory, for example, the present invention is set to 512.

[0178] (5) Parallel greedy random exploration structure:

[0179] Two DQN networks are used, one for selecting actions (primary network) and the other for calculating target Q values ​​(target network). After N iterations, the weights of the primary network are copied to the target network.

[0180] (6) Training the DQN model:

[0181] Through multiple rounds of environmental interaction, experience (state, action, reward, and new state) is collected. Experience is stored in the experience replay memory. During the training process, samples are randomly selected for learning and the weights of the main network are updated. The weights of the target network are regularly updated to stabilize the learning process.

[0182] (7) Optimization and evaluation:

[0183] Tune hyperparameters: learning rate, exploration rate, decay factor, etc. to improve performance.

[0184] Experimental simulation and comparison

[0185] See also Fig.12 As shown, by simulating a situation where a person falls into the water, according to the simulation environment, three ships are identified near the search area for search and rescue, and the distances from the search and rescue area are approximately 2 nautical miles, 3 nautical miles and 5 nautical miles respectively. The search and rescue path planning is performed using the Boustrophedon motion (BM) algorithm, the improved Boustrophedon motion algorithm, the particle swarm optimization algorithm (PSO) and the method of the present invention.

[0186] See also Fig.13 As shown in the figure, in the early stage of the search and rescue mission, because the rescue ship has just arrived at the edge of the search and rescue area, the search success rate is generally not high. As the search and rescue mission continues, the method proposed in the present invention shows its superiority. Compared with the other three algorithms, the method of the present invention can improve the search success rate faster and take the lead in achieving a search success rate of more than 80% in a short period of time. Specifically, compared with the traditional ox-ploughing algorithm, the method of the present invention has achieved an 8.9% lead in the search success rate.

[0187] See also Fig.14 As shown in Figure 2, the search time cost of the method of the present invention is only higher than that of the improved ox-farming algorithm, but lower than that of the traditional ox-farming algorithm and the particle swarm optimization algorithm. Compared with the ox-farming algorithm, the method of the present invention achieves a 5.3% reduction in search time cost.

[0188] In summary, the method proposed in the present invention not only performs well in improving the search success rate, but also shows good performance in controlling the search time cost.

[0189] Please refer to Fig.15 As shown, the present invention further provides a marine search and rescue path planning device, the device comprising:

[0190] A model field construction module is configured to determine a search and rescue area according to a target in distress at sea, obtain ocean dynamics environment data and image data of the search and rescue area, and use the ocean dynamics environment data and the image data to construct an environmental model field of the search and rescue area, wherein the environmental model field includes the total amount of available rescue ships, multiple grid areas, and multiple search task points;

[0191] An allocation module is configured to determine the total amount and size of the search task area according to the total amount of the available rescue ships, and then allocate each of the search task points to the search task area;

[0192] The planning module is configured to plan a search path for the corresponding available rescue ship according to each search task area using a multi-agent collaborative algorithm.

[0193] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for planning a maritime search and rescue path, characterized in that: The method comprises: Determine a search and rescue area according to a target in distress at sea, obtain ocean dynamics environment data and image data of the search and rescue area, and construct an environmental model field of the search and rescue area using the ocean dynamics environment data and the image data, wherein the environmental model field includes the total amount of available rescue ships, multiple grid areas, and multiple search task points; Determining the total amount and size of the search mission area according to the total amount of the available rescue ships, and then allocating each of the search mission points to the search mission area, includes: Determine the starting point for the search for each of the available rescue vessels, including: Determining the total amount of the search mission area according to the total amount of available rescue ships; Taking the search task point closest to the environmental model field for each available rescue ship as the search starting point; Dividing the size of the search mission area according to the proportional relationship between the total area of ​​the environmental model field and the speed of each of the available rescue ships comprises: Calculating the total theoretical search time consumed by each of the available rescue ships to search the environmental model field; The calculation formula for the theoretical total search time is as follows: , where: T 总 Total time spent searching for theories; A is the total area of ​​the environmental model field; S is the speed of the available rescue vessel; Calculate the actual search time of each available rescue vessel according to each theoretical total search time; The calculation formula for the actual search time is as follows: , where: T 实 The actual search for available rescue vessels takes time; T 行 The time it takes for available rescue vessels to travel from their original location to the search starting point; According to the actual search time of each available rescue ship and the total amount of the search task area, the environment model field is divided into the total area of ​​each search task area in proportion and the task capacity is determined; Allocating the search task points to the search task area according to the distance relationship between each search task point and each search starting point includes: Calculating the Manhattan distance between the remaining search task points and the search starting point of each available rescue ship; Allocating the remaining search mission points to the nearest search mission area according to the Manhattan distance of each available rescue ship; A multi-agent collaborative algorithm is used to plan a search path for the corresponding available rescue ship according to each search task area.

2. A method for maritime search and rescue path planning according to claim 1, characterized in that: The method of determining the total amount and size of the search mission area according to the total amount of the available rescue ships, and then allocating each of the search mission points to the search mission area, further comprises: Optimizing the search task points of each of the search task areas, including: Calculating the objective function of each of the search task areas to obtain the sum of the distances of each of the search task areas; When the sum of the distances of the search task area exceeds a preset threshold, the search task points of the search task area exceeding the preset threshold are reallocated to the nearest search task area.

3. A method for maritime search and rescue path planning according to claim 2, characterized in that: The calculation formula of the objective function is as follows: , where: J is the objective function; k is the total amount of the search mission area; C i For the i Search mission area; x Search mission points for searching mission areas; p i is the search starting point; d ( x , p i ) is the search task point to i The closest Manhattan distance to the search task area.

4. A method for maritime search and rescue path planning according to claim 1, characterized in that: The multi-agent collaborative algorithm includes: a path planning model, an environment modeling and action space model, a reward function model and a DQN model.

5. A method for maritime search and rescue path planning according to claim 4, characterized in that: The calculation formula of the reward function model is as follows: , where: R exp Exploration reward: refers to the exploration reward obtained when a rescue ship enters a new grid area; R limit Restricted state penalty: refers to the corresponding penalty given when the available rescue ship crosses the boundary or enters the obstacle area; R cov Duplicate coverage penalty: refers to the corresponding penalty given when a rescue ship enters the searched grid area; R exsteps Termination state penalty: The available rescue ship's sailing distance exceeds the preset maximum learning sailing distance. N max When the rescue ship is available, it will be placed in the terminated state. S exsteps and impose corresponding punishment; R unit Fixed reward for completing full coverage: refers to the fixed reward obtained when the rescue ship completes the search mission area; R pre Efficient Path Reward: refers to the additional reward obtained when a rescue ship completes the search mission area with a number of turns less than the preset turn threshold.

6. A maritime search and rescue path planning device, applied to a maritime search and rescue path planning method as claimed in any one of claims 1 to 5, characterized in that: The device comprises: A model field construction module is configured to determine a search and rescue area according to a target in distress at sea, obtain ocean dynamics environment data and image data of the search and rescue area, and use the ocean dynamics environment data and the image data to construct an environmental model field of the search and rescue area, wherein the environmental model field includes the total amount of available rescue ships, multiple grid areas, and multiple search task points; An allocation module is configured to determine the total amount and size of the search task area according to the total amount of the available rescue ships, and then allocate each of the search task points to the search task area; The planning module is configured to plan a search path for the corresponding available rescue ship according to each search task area using a multi-agent collaborative algorithm.

Citation Information

Patent Citations

  • Target to be searched and rescued identification method, system and device based on water robot

    CN111982120A

  • Formation path planning method and system for maritime search and rescue

    CN113867337A

  • Maritime search and rescue method and system

    CN115824217A

  • Unmanned aerial vehicle maritime search and rescue path planning method based on tabu bee colony algorithm

    CN116625376A

  • Rescue path planning method and system for rescue boat

    CN117029835A