A method for dynamic search probability estimation of underwater moving targets

Abstract

This invention proposes a dynamic search probability estimation method for underwater moving targets, belonging to the field of path planning technology for target search. The method includes: setting a search task, loading a map, and gridding the corresponding task area on the map; initializing the probability field and navigation parameters of the search platform; calculating the dynamically updated probability field diffusion probability of the moving target within the task area grid over time; and obtaining a navigation planning path for the search platform based on the updated probability field diffusion probability. This invention enables dynamic estimation of the probability of a moving target within the task area during the search process, thereby achieving dynamic path planning for the search of moving targets.

Description

Technical Field

[0001] This invention relates to the field of path planning technology for target search, and more particularly to a dynamic search probability estimation method for underwater moving targets. Background Technology

[0002] In marine production activities, there is a need to search for moving targets, such as searching for underwater unmanned vehicles (UAVs) that are still underway but have lost communication, and searching for and monitoring high-value fish schools in fisheries. However, under the large-scale conditions of the ocean and the underwater environment, the detection range of commonly used acoustic methods is very limited for small UAVs or marine organisms with inconspicuous target features, and searching for them requires a significant amount of time. Furthermore, for moving targets whose direction of movement is unknown, commonly used methods such as parallel area searches and arc searches based on searching for fixed targets cannot effectively assess the probability of the target reappearing in the already searched area due to movement during dynamic searches.

[0003] Therefore, based on the above situation, it is essential to provide a dynamic search probability estimation method for underwater moving targets to dynamically estimate the probability of marine moving targets in the mission area. Summary of the Invention

[0004] In view of this, the present invention proposes a dynamic search probability estimation method for underwater moving targets, which realizes dynamic estimation of the probability of marine moving targets in the mission area and provides a reference for path planning decisions for dynamic search of moving targets.

[0005] This invention provides a dynamic search probability estimation method for underwater moving targets, comprising the following:

[0006] Set up the search task, load the map, and grid the corresponding task area on the map;

[0007] Initialize the probability field and the navigation parameters of the search platform, and calculate the dynamically updated probability field diffusion probability of the moving target within the mission area grid over time.

[0008] The search platform obtains the navigation planning path based on the updated content of the probability field diffusion probability.

[0009] Based on the above technical solutions, preferably, the step of gridding the corresponding task area in the map involves dividing the task area into several planar grids, setting up several planar grid arrays, and ensuring that the size of the several grids is exactly the same.

[0010] Preferably, some planar grids are rectangular grids.

[0011] More preferably, the initial probability field is defined by using the boundary of the task area as the task boundary, within which a target in motion exists; the starting position of the search platform is set, and a detector is placed at one end of the search platform's navigation direction, with the detector having a fan-shaped detection range; during the search platform's navigation, the area swept by the detector forms the detected envelope boundary; the area currently within the detection range, since the search task is still ongoing and the detector has not yet confirmed the target, forms an area that has not probabilistically diffused; if a target within the task boundary but outside the detected envelope boundary enters the detected envelope boundary due to movement, a probability diffusion area is formed where the probability field diffuses towards the detected envelope boundary, forming a 0-probability envelope boundary between this area and the area where the probability has not yet diffused, and the 0-probability envelope boundary is located inside the detected envelope boundary.

[0012] More preferably, the calculation of the dynamically updated probability field diffusion probability of the moving target within the task region grid over time is based on the theory of matter diffusion probability. The moving target is treated as a single molecule, and the probability change during its movement is calculated using the flow process described by the LBM method. The line connecting the center of the task region grid to the midpoint of each side of the task region grid points to the adjacent positive grid, and the line connecting the center of the task region grid to each vertex of the task region grid points to the adjacent diagonal grid. C0 represents the molecule remaining in the current task region grid, C1, C2, C3, and C4 represent the molecule moving towards the adjacent positive task region grid, and C5, C6, C7, and C8 represent the diagonal task region grid directions adjacent to the vertex of the task region grid where the molecule is located. At each time step, the diffusion probability f of the molecule moving to adjacent grids is... c for in ∑f ci =1; The fundamental equation for the collision process to reach equilibrium is: f i (x+c i Δt,t+Δt)-f i (x,t)=Ω i +F i The left side of the equation is the flow term, f i The normalized concentration of the substance is represented by x, which represents the current grid position in the task region, and x+c. i Δt represents the direction c i The corresponding adjacent grid positions, Δt represents the time step, t is the current time, F i For external force, Ω i The collision term represents the degree of diffusion of the equilibrium state of molecular motion during the fluid motion process within time Δt. Where the equilibrium concentration u is the macroscopic velocity vector of the task region grid; the relaxation time τ satisfies v represents kinematic viscosity.

[0013] Furthermore, preferably, the search platform updates the content based on the probability field diffusion probability by calculating the derived result f′ of the probability field diffusion probability using the following formula. c : U is the maximum speed of the moving target, determined by the moving target itself; L is the length of the task area grid, and the intermediate variable s satisfies...

[0014] More preferably, the search platform updates the probability field diffusion probability by using numerical experiments to obtain the derived result of the probability field diffusion probability: Within a 3×3 task area grid, within the central task area grid and the maximum speed range, a point is randomly set in terms of position, speed, and direction. The position of the random point is calculated within one time step. This process is repeated multiple times, and the proportion of the point falling into the surrounding task area grid is counted until the proportion stabilizes. This proportion is the derived result f′ of the probability field diffusion probability. c .

[0015] In a further preferred embodiment, when the intermediate parameter is 0.1 < s < 1, the derivation result of the probability field diffusion probability is obtained by using the formula; when 0 < s < 0.1, the derivation result of the probability field diffusion probability is obtained by numerical experiment statistics.

[0016] More preferably, obtaining the navigation planning path involves iteratively updating the probability P of the task region grid with pointer i at time t+1 after obtaining the derivation result of the probability field diffusion probability. i，t+1 :P i，t+1 =P i,t -[f′ c ][P i，t -P inei,t ], P i，t f' represents the probability of the task region grid with pointer i at time t; c [P] is the vector corresponding to the derived results of the probability field diffusion probability of the task region grid with pointer i in all directions; inei，t This represents the adjacent task region grids of the task region grid with pointer i at time t; the dynamically updated probability field is obtained according to the formula, the navigation planning path optimization function is determined, and the final navigation planning path is obtained.

[0017] A further optimized function for the route planning path takes minimizing the time T required to find the moving target as its optimization objective: minT, satisfying... Where A represents the search platform's action, and the subscripts represent different times; Φ(·) is the strategy function; P iS represents the state of the task area grid at time i; S represents the actual physical state of the environment, with subscripts indicating different times; F represents the update of the probability field diffusion probability.

[0018] The present invention provides a dynamic search probability estimation method for underwater moving targets, which has the following advantages compared with the prior art:

[0019] (1) This invention provides a dynamic search probability estimation method for underwater moving targets, which can be applied to scenarios where both the task boundary and the target being searched are moving. It provides input for path planning decisions for dynamic search of moving targets and realizes dynamic estimation of the probability of moving targets in the task area during the search process. It can be applied to scenarios where both the task boundary and the target being searched are moving.

[0020] (2) The present invention uses a map gridding method, which can be extended to a planar or three-dimensional environment, and can also be further extended to water surface, land, air and even cross-domain application environment to perform dynamic and fast search of dynamic targets. Attached Figure Description

[0021] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram illustrating the working process of the dynamic search probability estimation method for underwater moving targets according to the present invention.

[0023] Figure 2 This is a schematic diagram of the probability field diffusion in the dynamic search probability estimation method for underwater moving targets according to the present invention;

[0024] Figure 3 This is a schematic diagram of the probability diffusion model of the reference LBM method for the dynamic search probability estimation method for underwater moving targets according to the present invention.

[0025] Figure 4 This is a schematic diagram of the route planning path for an embodiment of the dynamic search probability estimation method for underwater moving targets according to the present invention.

[0026] Figure 5 This is a schematic diagram illustrating the probability of a moving target being scanned in the dynamic search probability estimation method for underwater moving targets according to the present invention.

[0027] Figure 6This is a schematic diagram of the strategy function decision process of the dynamic search probability estimation method for underwater moving targets according to the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the implementation methods of this invention without creative effort are within the scope of protection of this invention.

[0029] In view of this, such as Figures 1-3 As shown, this invention provides a dynamic search probability estimation method for underwater moving targets, including the following:

[0030] S1: Set the search task, load the map, and grid the corresponding task area in the map;

[0031] The corresponding task area on the map is gridded. Specifically, the task area is divided into several planar grids, and these planar grids are arranged in an array with all grids having the same size. Preferably, the planar grids are rectangular grids.

[0032] S2: Initialize the probability field and the navigation parameters of the search platform, and calculate the dynamic updated probability field diffusion probability of the moving target within the mission area grid over time.

[0033] S21: Initialize the probability field, using the boundary of the task region as the task boundary, i.e., the task boundary is... Figure 2 The outermost rectangle in the context; a moving target exists within the mission boundary; the starting position of the search platform is set, and a detector with a fan-shaped detection range is placed at one end of the search platform's navigation direction; during the search platform's navigation, the area swept by the detector forms the detected envelope boundary; the area currently within the detection range, which has not yet been detected by the detector due to the ongoing search mission, forms a region that has not probabilistically diffused; a target within the mission boundary but outside the detected envelope boundary, if it enters the detected envelope boundary due to movement, forms a probability diffusion region where the probability field diffuses towards the detected envelope boundary, forming a 0-probability envelope boundary between this region and the region where the probability has not yet diffused. Figure 2 The dashed part in the diagram indicates that the 0-probability envelope boundary is located inside the already detected envelope boundary.

[0034] S22: Calculate the dynamically updated diffusion probability field of a moving target within the task region grid over time. This is based on the theory of matter diffusion probability, treating the moving target as a single molecule. Referring to the flow process of the LBM method, the probability change during the moving target's movement is calculated. The lines connecting the center of the task region grid to the midpoints of each side of the task region grid point to adjacent positive grids, and the lines connecting the center of the task region grid to each vertex of the task region grid point to adjacent diagonal grids. C0 represents the molecule remaining in the current task region grid, C1, C2, C3, and C4 represent the molecule moving towards adjacent positive task region grids, and C5, C6, C7, and C8 represent the diagonal task region grids adjacent to the vertex of the task region grid where the molecule is located. At each time step, the diffusion probability f of the molecule moving to adjacent grids is calculated. c for in ∑f ci =1; The fundamental equation for the collision process to reach equilibrium is: f i (x+c i Δt,t+Δt)-f i (x,t)=Ω i +F i The left side of the equation is the flow term, f i The normalized concentration of the substance is represented by x, which represents the current grid position in the task region, and x+c. i Δt represents the direction c i The corresponding adjacent grid positions, Δt represents the time step, t is the current time, F i For external force, Ω i The collision term represents the degree of diffusion of the equilibrium state of molecular motion during the fluid motion process within time Δt. Where the equilibrium concentration u is the macroscopic velocity vector of the task region grid; the relaxation time τ satisfies v represents the kinematic viscosity. It is assumed that the movement of molecules within the grid follows a Maxwell distribution, and the macroscopic average velocity is the speed of sound.

[0035] This method, referencing the Lattice Boltzmann Method (LBM), analogizes the search probability of moving targets within the task area. It treats each moving target as a separate element, using probability to characterize the statistical probability of its appearance in each grid cell, and employs the flow process of the LBM to analogize the probability changes during target movement. The LBM method is a commonly used technique for handling boundary conditions. This invention further improves upon this method. After setting the task area grid, starting from the right side, proceeding counter-clockwise through each boundary cell, the diffusion probability in the direction adjacent to the edge is... Starting from the top right vertex, the diagonal diffusion probability of each vertex in the task area grid, moving clockwise, is: The probability that a molecule remains in the current task region grid is:

[0036] S3: The search platform obtains the navigation planning path based on the updated content of the probability field diffusion probability.

[0037] S31: Due to the fundamental physical difference between f and the diffusion of matter within a fluid, c The updated content needs to be adjusted based on the parameters of the task and the search platform. The search platform can update the content based on the diffusion probability of the probability field in the following two ways:

[0038] Method 1: The derived result f′ of the probability field diffusion probability is obtained by calculating the following formula. c : U is the maximum speed of the moving target, determined by the moving target itself; L is the length of the task area grid, and the intermediate variable S satisfies...

[0039] Method 2: Obtain the derivation result of the probability field diffusion probability through numerical experiments and statistics: Within a 3×3 task area grid, within the central task area grid and the maximum velocity range, randomly set the position, velocity, and direction of a point. Calculate the position of the random point within one time step. Repeat this process multiple times and count the proportion of the point falling into the surrounding task area grid until the proportion stabilizes. This proportion is the derivation result f′ of the probability field diffusion probability. c .

[0040] When the intermediate parameter 0.1 < s < 1, the derivation result of the probability field diffusion probability is obtained by using the formula, i.e., method 1; when 0 < s < 0.1, the derivation result of the probability field diffusion probability is obtained by numerical experiment statistics, i.e., method 2.

[0041] S32: Obtaining the navigation planning path involves iteratively updating the probability P of the task region grid with pointer i at time t+1 after obtaining the derivation result of the probability field diffusion probability. i，t+1 :P i，t+1 =P i,t -[f′ c ][P i，t -P inei,t ], P i，t f' represents the probability of the task region grid with pointer i at time t; c [P] is the vector corresponding to the derived results of the probability field diffusion probability of the task region grid with pointer i in all directions; inei，t This represents the neighboring task region grids of the task region grid with pointer i at time t; the dynamically updated probability field is obtained according to the formula, the navigation planning path optimization function is determined, and the final navigation planning path is obtained. The probability P is calculated using the above formula.i，t+1 In order to improve the calculation speed, the kinematic viscosity v and the macroscopic velocity u of the task region mesh can both be set to 0.

[0042] The optimization function for the route planning path aims to minimize the time T required to find the moving target: minT, satisfying... Where A represents the search platform's actions, mainly including speed, heading information, or other physical quantities that can be converted into this information, and the subscripts represent different times; φ(·) is the strategy function; P i The state of the task area grid at time i is represented; S represents the actual physical state of the environment, including the vehicle's position, changes in obstacle positions on the map, etc., with subscripts indicating different times; F indicates that the probability field diffusion probability is updated using any of the methods in step 31. Figure 6 A flowchart illustrating the decision-making process of the policy function is shown.

[0043] It should be noted that for a given patrol mission, the route can also be planned in advance. The path optimization method and policy function for updating search strategies or navigation parameters can use conventional algorithms such as greedy algorithms, or artificial intelligence methods such as reinforcement learning, decision-making Transformers, Long Short-Term Memory networks, and generative networks. Regardless of the method used, the optimization function can be the route planning path optimization function proposed in this invention.

[0044] Example: A movable, ultra-large marine aquaculture cage, 4 km in diameter, is used to cultivate a high-value fish species. To ensure the quality of the fish and reduce construction costs, it is not advisable to use a concentrated large active sonar or a large number of small sonars for real-time detection within the cage. Therefore, two unmanned monitoring vessels, each carrying a small sonar, continuously patrol and detect within the cage area. The fish's cruising speed and the cage platform's movement speed are both 1 knot, while the patrolling unmanned vessels travel at 1.5 knots, with an effective detection range of 300 meters. Unmanned vessels 1 and 2 are assigned search areas, one to the left and one to the right of the main vessel's direction of movement. The probability field diffusion is dynamically calculated using this method. The length of the rectangular grid is L = 200 meters, and the time step Δt is 40 seconds. Without pre-training, a greedy algorithm is used for decision-making. The relative paths of the two unmanned monitoring vessels over a period of time, with the main vessel as the reference frame, are as follows: Figure 4 As shown, the probability of all marine targets being scanned once within this time range is as follows: Figure 5 As shown, the probability of detecting each high-value target in the cage after approaching after 1750 time steps is about 99%, demonstrating the effectiveness of this method.

[0045] 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, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for dynamically searching and estimating the probability of underwater moving targets, characterized in that, Includes the following: Set up the search task, load the map, and grid the corresponding task area on the map; Initialize the probability field and the navigation parameters of the search platform, and calculate the dynamic, time-updated probability field diffusion probability of the moving target within the mission area grid. The calculation of the diffusion probability of a moving target within the task region grid, dynamically updated over time, is based on the theory of matter diffusion probability. The moving target is treated as a single molecule, and the probability change during its movement is calculated using the LBM method's flow process. The lines connecting the center of the task region grid to the midpoints of each side of the task region grid point to adjacent positive grids, and the lines connecting the center of the task region grid to each vertex of the task region grid point to adjacent diagonal grids. C0 represents the molecule remaining in the current task region grid, C1, C2, C3, and C4 represent the molecule moving towards adjacent positive task region grids, and C5, C6, C7, and C8 represent the diagonal task region grids adjacent to the vertex of the task region grid where the molecule is located. At each time step, the diffusion probability f of the molecule moving to adjacent grids is calculated. c for in ∑f ci =1; The fundamental equation for the collision process to reach equilibrium is: f i (x+c i Δt,t+Δt)-f i (x,t)=Ω i +F i The left side of the equation is the flow term, f i The normalized concentration of the substance is represented by x, which represents the current grid position in the task region, and x+c. i Δt represents the direction c i The corresponding adjacent grid positions, Δt represents the time step, t is the current time, F i For external force, Ω i The collision term represents the degree of diffusion towards the equilibrium state of molecular motion during the fluid motion within a time interval Δt. Where the equilibrium concentration u is the macroscopic velocity vector of the task region grid; the relaxation time τ satisfies v is the kinematic viscosity; Then, obtain the diffusion probability of the probability field that changes dynamically over time, and define an intermediate variable s that satisfies... Where U is the maximum speed of the moving target, determined by the moving target; L is the length of the task area grid. When the intermediate variable 0.1 < s < 1, the updated content f′ of the probability field diffusion probability is calculated using the formula. c , When the intermediate variable 0 < s < 0.1, the updated content f′ of the probability field diffusion probability is obtained by numerical experiment statistics. c ; The search platform obtains the navigation planning path based on the updated content of the probability field diffusion probability.

2. The method for dynamic search probability estimation of underwater moving targets according to claim 1, characterized in that, The process of gridding the corresponding task area in the map involves dividing the task area into several planar grids, setting up an array of several planar grids, and ensuring that the size of the several grids is exactly the same.

3. The method for dynamic search probability estimation of underwater moving targets according to claim 2, characterized in that, Several planar grids are rectangular grids.

4. The method for dynamic search probability estimation of underwater moving targets according to claim 3, characterized in that, The initial probability field is defined by using the boundary of the task area as the task boundary, within which a moving target exists. The starting position of the search platform is set, and a detector with a fan-shaped detection range is placed at one end of the platform's navigation direction. During the platform's navigation, the area swept by the detector forms the detected envelope boundary. Areas currently within the detection range, but not yet confirmed by the detector due to the ongoing search, form areas where the probability has not diffused. Targets within the task boundary but outside the detected envelope boundary, if they move into the detected envelope boundary, form a probability diffusion region where the probability field diffuses towards the detected envelope boundary. This region, along with areas where the probability has not yet diffused, forms a 0-probability envelope boundary, located inside the detected envelope boundary.

5. The method for dynamic search probability estimation of underwater moving targets according to claim 1, characterized in that, The updated content f′ of the diffusion probability of the probability field is obtained by numerical experiment statistics. c Within a 3×3 task area grid, and within the central task area grid and the maximum velocity range, a point is randomly selected with a fixed position, velocity, and direction. The position of this random point is calculated within one time step. This process is repeated multiple times, and the proportion of the point falling into the surrounding task area grid is counted until this proportion stabilizes. This proportion is the updated content f′ of the probability field diffusion probability. c .

6. The method for dynamic search probability estimation of underwater moving targets according to claim 1, characterized in that, The process of obtaining the navigation planning path involves iteratively updating the probability P of the task region grid with pointer i at time t+1 after obtaining the updated content of the probability field diffusion probability. i，t+1 :P i，t+1 =P i，t -[f′ c ][P i，t -P inei，t ], P i，t f' represents the probability of the task region grid with pointer i at time t; c [P] is the vector corresponding to the update content of the probability field diffusion probability in all directions of the task region grid with pointer i; inei，t This represents the probability of adjacent task region grids of the task region grid with pointer i at time t; The dynamically updated probability field is obtained based on the formula, the navigation planning path optimization function is determined, and the final navigation planning path is obtained.

7. The method for dynamic search probability estimation of underwater moving targets according to claim 6, characterized in that, The optimization function for route planning aims to minimize the time T required to find the moving target: minT, satisfying... Where A represents the search platform's action, and the subscripts represent different times; Φ(·) is the strategy function; P t * S represents the state of the task region grid at time t; 0 The actual physical state of the environment is represented by the subscripts, which indicate different times. F indicates that the probability field diffusion probability is updated.

Images