Intelligent navigation control method and system for combined fleet

Abstract

The invention relates to the technical field of navigation control, and discloses an intelligent navigation control method and system for a combined fleet, and the method comprises the steps: carrying out the abstraction of N types of established formations of a target fleet, and obtaining a feature matrix corresponding to each type of established formation; based on the feature matrix, analyzing a conversion path for converting each set formation into the remaining set formation and conversion time corresponding to each conversion path; performing calculation between any two feature matrixes to obtain a correlation value; and obtaining the distance time of the target sea condition, and carrying out dynamic switching on the established formation within the distance time based on the conversion paths, the conversion time corresponding to each conversion path and the associated value. According to the method, the negative influence of sudden sea condition change on the safety of the fleet is reduced to the greatest extent.

Description

Technical Field

[0001] The present invention relates to the field of navigation control technology, and more specifically, to an intelligent navigation control method and system for a combined fleet. Background Art

[0002] Existing unmanned fleets usually cluster sea conditions to obtain a variety of typical sea conditions, and then establish a set formation of unmanned ship fleets for each typical sea condition to meet the navigation needs under different sea conditions. However, in practical applications, when changing formations, it is often done by directly switching from one set formation to another. However, there is a situation where the time spent switching the set formation is greater than the remaining time to enter the corresponding sea condition, resulting in the unmanned ship fleet having entered the corresponding sea condition, but the formation switching has not yet been completed. As a result, the safe operation of the unmanned ship fleet is negatively affected. Summary of the invention

[0003] The present invention provides a combined fleet intelligent navigation control system to solve the technical problems raised in the background technology.

[0004] The present invention provides a combined fleet intelligent navigation control system, comprising:

[0005] The feature construction module is used to abstractly obtain the feature matrix corresponding to each of the N predetermined formations of the target fleet;

[0006] A matrix conversion module, used for analyzing the conversion paths of each predetermined formation to the remaining predetermined formations and the conversion time corresponding to each conversion path based on the characteristic matrix;

[0007] Matrix association module, used to calculate the association value between any two feature matrices;

[0008] The fleet management module is used to obtain the distance time of the target sea condition and dynamically switch the established formation within the distance time based on the conversion path, the conversion time corresponding to each conversion path and the associated value.

[0009] Furthermore, the characteristic matrix corresponding to each given formation is obtained, including:

[0010] Obtain the plane geometric structure of each given formation and the corresponding remote sensing image;

[0011] Extract the center of gravity of the plane geometric structure;

[0012] Extracting the plane geometric structure of the corresponding predetermined formation from the remote sensing image, and marking the center of gravity of the plane geometric structure in the remote sensing image;

[0013] Aligning N remote sensing images according to the centroid point to obtain overlapping rectangles of the N remote sensing images;

[0014] Rasterize each overlapping rectangle, including:

[0015] Initialize several individuals that meet the constraints, each of which represents the number of rows and columns for rasterizing overlapping rectangles;

[0016] The constraints include: for any overlapping rectangles, there is at most one unmanned ship in each of the cells obtained by rasterizing them;

[0017] The fitness function includes: maximizing the number of rows and columns;

[0018] Based on the constraints and fitness function, the number of rows and columns of each overlapping rectangle rasterization is obtained through a genetic algorithm, and the corresponding feature matrix is ​​obtained; wherein the feature matrix is ​​a multi-value matrix, which includes P+1 elements, P is the number of unmanned ships in the target fleet, each unmanned ship corresponds to an element, and the remaining elements in the multi-value matrix are all 0.

[0019] Furthermore, the conversion paths and the conversion time corresponding to each conversion path include:

[0020] Get the conversion path and conversion time between the i-th established formation and the j-th established formation, i≠j, specifically:

[0021] Get the corresponding i-th feature matrix and j-th feature matrix;

[0022] Establish movement constraints: for any non-zero element of a feature matrix, obtain the corresponding eight-neighborhood element values; within a preset time interval, if the neighborhood element is 0, the non-zero element is allowed to move to the corresponding neighborhood; if the neighborhood element is not 0, the non-zero element is not allowed to move to the corresponding neighborhood;

[0023] Based on the movement constraint, the i-th feature matrix is ​​converted to the j-th feature matrix, and the simulation conversion is repeated several times to obtain the sum of the preset time intervals of each simulation conversion;

[0024] Selecting the simulation conversion with the smallest sum of preset time intervals as the conversion path from the i-th feature matrix to the j-th feature matrix, and taking the sum of the preset time intervals as the conversion time;

[0025] Among them, the conversion path represents the intermediate state matrix and the jth characteristic matrix corresponding to several preset time intervals in the process of converting the i-th characteristic matrix to the j-th characteristic matrix, and the intermediate state matrix is ​​obtained based on the corresponding simulation conversion collection.

[0026] Furthermore, the correlation value between any two feature matrices is calculated, including:

[0027] Based on the established formations corresponding to the i-th feature matrix and the j-th feature matrix, the head and tail of the j-th feature matrix are determined respectively; wherein the head represents the head of the established formation, and the tail represents the tail of the established formation;

[0028] The feature matrix corresponding to the given formation is classified. The classification types include: forward type, and descend to stable type according to a fixed gradient, a total of K types, and a first weight is assigned to each type. The value range of the first weight is , and the first The value of the first weight of each type is 1;

[0029] Based on the first weight of the j-th feature matrix, the first calculation, the second calculation or the third calculation is performed to obtain the correlation value between the i-th feature matrix and the j-th feature matrix .

[0030] Furthermore, if the first weight of the j-th feature matrix is ​​greater than 1, the first calculation is performed; for the i-th feature matrix and the j-th feature matrix, matrix segmentation is performed based on the parallel lines connecting the head and the tail of the team, and M feature vectors are obtained. The second weight is assigned to the feature vector based on the Euclidean distance between each feature vector and the center of the corresponding feature matrix. The value range of the second weight is , and the sum of several second weights is 1; where the Euclidean distance of the eigenvector is inversely proportional to the second weight, and if the Euclidean distances of any two eigenvectors are equal, the corresponding second weights are equal; calculate the association value of the i-th eigenmatrix and the j-th eigenmatrix ,as follows: , ;in, represents the first weight of the j-th feature matrix, represents the number of eigenvectors obtained by matrix partitioning, express The index of represents the Euclidean distance between the mth eigenvector and the center of the feature matrix, represents the Euclidean distance between the mth eigenvector in the i-th feature matrix and the mth eigenvector in the j-th feature matrix, represents the distance weight;

[0031] The Euclidean distance between the eigenvector and the center of the eigenmatrix represents the Euclidean distance between the coordinates of the central element of the eigenvector in the eigenmatrix and the coordinates of the central element of the eigenmatrix.

[0032] Furthermore, if the first weight of the j-th feature matrix is ​​equal to 1, the second calculation is performed; the Hamming distance is calculated for the i-th feature matrix and the j-th feature matrix, and the Hamming distance is used as the correlation value between the i-th feature matrix and the j-th feature matrix. .

[0033] Furthermore, if the first weight of the j-th feature matrix is ​​less than 1, a third calculation is performed; for the i-th feature matrix and the j-th feature matrix, matrix segmentation is performed based on the direction from the head to the tail, and N feature vectors are obtained. The value range of the third weight assigned to the N feature vectors based on the direction from the head to the tail is , and the sum of several third weights is 1; wherein the third weight decreases in the direction from the head to the tail; calculate the correlation value between the i-th feature matrix and the j-th feature matrix ,as follows: , ;in, represents the number of eigenvectors obtained by matrix partitioning, express The index of represents the gradient of the nth eigenvector, represents the Euclidean distance between the nth eigenvector in the i-th feature matrix and the nth eigenvector in the j-th feature matrix, represents the gradient weight.

[0034] Furthermore, the distance and time of the target sea conditions are obtained, including:

[0035] The remote sensing image of the sea area at a distance L from the target fleet is obtained based on the ship-borne UAV, and the remote sensing image of the sea area is recognized based on the recognition model to obtain the target sea conditions;

[0036] If the target sea condition is the same as the current sea condition, the dynamic switching of the established formation will not be performed;

[0037] If the target sea condition is inconsistent with the current sea condition, the distance time is calculated, including:

[0038] Get the current moving speed V of the target fleet, then the distance time is .

[0039] Furthermore, dynamic switching is performed, including:

[0040] Obtain the target predetermined formation corresponding to the target sea condition, and obtain the target conversion time based on the current feature matrix s corresponding to the current predetermined formation and the target feature matrix g corresponding to the target predetermined formation;

[0041] If the target conversion time ≤ the distance time, a first switching is performed, specifically including: when the distance time is 0, the current feature matrix is ​​just converted to the target feature matrix;

[0042] If the target conversion time is greater than the distance time, the second switching is performed, which specifically includes:

[0043] Get the candidate feature matrix b of the target feature matrix; where the correlation value between the current feature matrix and the candidate feature matrix , , represents the proportionality coefficient, ;

[0044] based on Sort the candidate feature matrices from large to small to obtain feature sorting;

[0045] Based on the order of feature sorting, judge one by one from front to back:

[0046] If the candidate conversion time between the current feature matrix and the candidate feature matrix ranked r in the feature ranking is greater than the distance time, the candidate feature matrix ranked r is abandoned, and the candidate feature matrix ranked r+1 is judged;

[0047] If the alternative conversion time between the current feature matrix and the alternative feature matrix ranked r in the feature sorting is ≤ the distance time, the alternative feature matrix ranked r replaces the target feature matrix, and when the distance time is 0, the current feature matrix is ​​exactly converted to the alternative feature matrix ranked r.

[0048] Further, the target fleet is controlled to switch at a plurality of preset time intervals based on the switching path corresponding to the first switching or the second switching.

[0049] A combined fleet intelligent navigation control method, the method comprising:

[0050] Based on the N predetermined formations of the target fleet, the feature matrix corresponding to each predetermined formation is abstractly obtained;

[0051] Analyze the conversion paths of each given formation to the remaining given formations and the conversion time corresponding to each conversion path based on the characteristic matrix;

[0052] Calculate the correlation value between any two feature matrices;

[0053] The distance time of the target sea condition is obtained, and the established formation is dynamically switched within the distance time based on the conversion path, the conversion time corresponding to each conversion path and the associated value.

[0054] The beneficial effect of the present invention is that even if the conversion time required to switch to the target predetermined formation exceeds the remaining time to enter the new sea condition, it can intelligently select and quickly switch to an alternate predetermined formation, so that the unmanned ship fleet is in a relatively safe and adaptable state before entering the sea condition. Specifically, by calculating the correlation of the characteristic matrix, the system can determine an alternate formation that can be switched within a limited time under time pressure, thereby minimizing the negative impact of sudden changes in sea conditions on fleet safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a module diagram of a combined fleet intelligent navigation control system of the present invention. DETAILED DESCRIPTION

[0056] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0057] like Figure 1 As shown, a combined fleet intelligent navigation control system comprises:

[0058] The feature construction module is used to abstractly obtain the feature matrix corresponding to each of the N predetermined formations of the target fleet;

[0059] A matrix conversion module, used for analyzing the conversion paths of each predetermined formation to the remaining predetermined formations and the conversion time corresponding to each conversion path based on the characteristic matrix;

[0060] Matrix association module, used to calculate the association value between any two feature matrices;

[0061] The fleet management module is used to obtain the distance time of the target sea condition and dynamically switch the established formation within the distance time based on the conversion path, the conversion time corresponding to each conversion path and the associated value.

[0062] In one embodiment of the present invention, the characteristic matrix corresponding to each predetermined formation is obtained, including:

[0063] Obtain the plane geometric structure of each given formation and the corresponding remote sensing image;

[0064] Extract the center of gravity of the plane geometric structure;

[0065] Extracting the plane geometric structure of the corresponding predetermined formation from the remote sensing image, and marking the center of gravity of the plane geometric structure in the remote sensing image;

[0066] Aligning N remote sensing images according to the centroid point to obtain overlapping rectangles of the N remote sensing images;

[0067] Rasterize each overlapping rectangle, including:

[0068] Initialize several individuals that meet the constraints, each of which represents the number of rows and columns for rasterizing overlapping rectangles;

[0069] The constraints include: for any overlapping rectangles, there is at most one unmanned ship in each of the cells obtained by rasterizing them;

[0070] The fitness function includes: maximizing the number of rows and columns;

[0071] Based on the constraints and fitness function, the number of rows and columns of each overlapping rectangle rasterization is obtained through a genetic algorithm, and the corresponding feature matrix is ​​obtained; wherein the feature matrix is ​​a multi-value matrix, which includes P+1 elements, P is the number of unmanned ships in the target fleet, each unmanned ship corresponds to an element, and the remaining elements in the multi-value matrix are all 0.

[0072] Specifically, by acquiring a remote sensing image of the target fleet, the remote sensing image is similar to a two-dimensional image, so the plane geometric structure of the target fleet in the two-dimensional image can be extracted. The center of gravity of the plane geometric structure is extracted, and a plurality of remote sensing images are aligned based on the center of gravity. It should be noted that the extraction of the center of gravity includes but is not limited to geometric construction, or establishing a coordinate system and calculating it through the centroid method.

[0073] In one embodiment of the present invention, a plurality of remote sensing images are aligned, specifically:

[0074] The acquired remote sensing images are cut into rectangular images, and the lines connecting the head and tail of the target fleet are parallel to any set of parallel image edges of the rectangular image. Then, the center of gravity of each remote sensing image is used as the alignment point, and the directions of the head and tail of all rectangular images are kept consistent. At this time, a maximum overlapping rectangle can be obtained, and only the images within the overlapping rectangle are retained for each rectangular image. It should be noted that the overlapping rectangle also includes P unmanned ships.

[0075] Specifically, the overlapping rectangles are rasterized based on the genetic algorithm to obtain a rasterization process. Thus, each given formation is abstractly processed into a feature matrix. The feature matrix includes at least P+1 elements, where the P elements correspond to an unmanned ship respectively, and the element 0 represents a blank sea area, so as to facilitate the movement path of each ship in the subsequent conversion path.

[0076] In one embodiment of the present invention, the conversion path and the conversion time corresponding to each conversion path include:

[0077] Get the conversion path and conversion time between the i-th established formation and the j-th established formation, i≠j, specifically:

[0078] Get the corresponding i-th feature matrix and j-th feature matrix;

[0079] Establish movement constraints: for any non-zero element of a feature matrix, obtain the corresponding eight-neighborhood element values; within a preset time interval, if the neighborhood element is 0, the non-zero element is allowed to move to the corresponding neighborhood; if the neighborhood element is not 0, the non-zero element is not allowed to move to the corresponding neighborhood;

[0080] Based on the movement constraint, the i-th feature matrix is ​​converted to the j-th feature matrix, and the simulation conversion is repeated several times to obtain the sum of the preset time intervals of each simulation conversion;

[0081] Selecting the simulation conversion with the smallest sum of preset time intervals as the conversion path from the i-th feature matrix to the j-th feature matrix, and taking the sum of the preset time intervals as the conversion time;

[0082] Among them, the conversion path represents the intermediate state matrix and the jth characteristic matrix corresponding to several preset time intervals in the process of converting the i-th characteristic matrix to the j-th characteristic matrix, and the intermediate state matrix is ​​obtained based on the corresponding simulation conversion collection.

[0083] In one embodiment of the present invention, each time a conversion is simulated, a random movement is performed on the movement of each non-zero element based on the movement constraint to obtain different conversion paths and conversion times, wherein the conversion paths and conversion times are used to guide the conversion of the actual predetermined formation.

[0084] In one embodiment of the present invention, all ships of the target fleet maintain the same moving speed, that is, taking any unmanned ship as the reference system, the other ships are stationary. Therefore, a preset interval time can be set, within which the unmanned ship can move a distance corresponding to a cell in the relative reference system.

[0085] Specifically, the movement constraint includes: the system examines the eight neighborhoods (up and down, left and right, and diagonal directions) of each non-zero element. Within a preset time interval, if a neighborhood position is 0 (that is, the position is idle), the non-zero element is allowed to move to this position. If the neighborhood position is not 0, it means that there are other unmanned ships at this position, and the element is not allowed to move there. This constraint ensures that the positions of the unmanned ships will not overlap during the conversion process and meet the actual movement restrictions.

[0086] In one embodiment of the present invention, calculating the correlation value between any two feature matrices includes:

[0087] Based on the established formations corresponding to the i-th feature matrix and the j-th feature matrix, the head and tail of the j-th feature matrix are determined respectively; wherein the head represents the head of the established formation, and the tail represents the tail of the established formation;

[0088] The feature matrix corresponding to the given formation is classified. The classification types include: forward type, and descend to stable type according to a fixed gradient, a total of K types, and a first weight is assigned to each type. The value range of the first weight is , and the first The value of the first weight of each type is 1;

[0089] Based on the first weight of the j-th feature matrix, the first calculation, the second calculation or the third calculation is performed to obtain the correlation value between the i-th feature matrix and the j-th feature matrix .

[0090] Specifically, the K types from the forward type to the stable type are classified based on the moving speed of the target fleet, where the forward type is the maximum speed of the target fleet's given formation, and the stable type is the minimum speed. The forward type and the stable type are different in structure, including: the forward type tends to be narrow and long from the front to the back of the team, and the stable type tends to be wide and short from the front to the back of the team. Therefore, the j-th feature matrix is ​​first classified, including: the first weight is greater than 1, equal to 1 and less than 1, and the first calculation, the second calculation and the third calculation are performed respectively.

[0091] In one embodiment of the present invention, if the first weight of the j-th feature matrix is ​​greater than 1, a first calculation is performed; for the i-th feature matrix and the j-th feature matrix, matrix segmentation is performed based on the parallel lines connecting the head and the tail of the team, and M feature vectors are obtained. A second weight is assigned to the feature vector based on the Euclidean distance between each feature vector and the center of the corresponding feature matrix. The value range of the second weight is , and the sum of several second weights is 1; where the Euclidean distance of the eigenvector is inversely proportional to the second weight, and if the Euclidean distances of any two eigenvectors are equal, the corresponding second weights are equal; calculate the association value of the i-th eigenmatrix and the j-th eigenmatrix ,as follows: , ;in, represents the first weight of the j-th feature matrix, represents the number of eigenvectors obtained by matrix partitioning, express The index of represents the Euclidean distance between the mth eigenvector and the center of the feature matrix, represents the Euclidean distance between the mth eigenvector in the i-th feature matrix and the mth eigenvector in the j-th feature matrix, represents the distance weight;

[0092] The Euclidean distance between the eigenvector and the center of the eigenmatrix represents the Euclidean distance between the coordinates of the central element of the eigenvector in the eigenmatrix and the coordinates of the central element of the eigenmatrix.

[0093] In one embodiment of the present invention, the j-th characteristic matrix is ​​biased towards the advancing type, and the shape is narrow and long. Therefore, after matrix segmentation, the characteristic vector is obtained, and most of the non-zero elements are concentrated in the central part of the characteristic matrix. Therefore, the second weight is assigned to the center from high to low on both sides. This weight allocation method reflects the dominant role of the central part in the overall formation, and thus can better reflect the key characteristics of the advancing formation. Then the Euclidean distance of the characteristic vectors of the i-th characteristic matrix and the j-th characteristic matrix are compared one by one. The product of the Euclidean distance and the second weight is calculated and accumulated to obtain the sum value, and the sum value is multiplied by the first weight again to obtain the correlation between the i-th characteristic matrix and the j-th characteristic matrix. The correlation represents: the degree of morphological similarity between the i-th characteristic matrix and the j-th characteristic matrix, that is, the bearing capacity of the established formation corresponding to the i-th characteristic matrix when encountering the corresponding sea conditions of the j-th characteristic matrix.

[0094] In one embodiment of the present invention, if the first weight of the j-th feature matrix is ​​equal to 1, the second calculation is performed; the Hamming distance is calculated for the i-th feature matrix and the j-th feature matrix, and the Hamming distance is used as the correlation value between the i-th feature matrix and the j-th feature matrix. .

[0095] Specifically, for wide and short stable type formations, their overall structure is balanced and their distribution is relatively regular, so their first weight is set to 1. For this type of formation, there is no need to emphasize the importance of the central area through complex gradient or distance weighting, because the entire formation structure is evenly distributed in the horizontal direction. For stable, wide and short formations, their feature matrix usually shows a relatively clear distribution of 0 and non-0, and the Hamming distance can quickly and effectively capture the difference between the two. The smaller the Hamming distance value, the less similar the two formations are. On the contrary, the more obvious the difference in form.

[0096] In one embodiment of the present invention, if the first weight of the j-th feature matrix is ​​less than 1, a third calculation is performed; for the i-th feature matrix and the j-th feature matrix, matrix segmentation is performed based on the direction from the head to the tail, and N feature vectors are obtained. The value range of the third weight assigned to the N feature vectors based on the direction from the head to the tail is , and the sum of several third weights is 1; wherein the third weight decreases in the direction from the head to the tail; calculate the correlation value between the i-th feature matrix and the j-th feature matrix ,as follows: , ;in, represents the number of eigenvectors obtained by matrix partitioning, express The index of represents the gradient of the nth eigenvector, represents the Euclidean distance between the nth eigenvector in the i-th feature matrix and the nth eigenvector in the j-th feature matrix, represents the gradient weight.

[0097] Specifically, if the first weight of the jth feature matrix is ​​less than 1, the morphology is wide and short, and the number of unmanned boats decreases from the head to the tail, so there is a gradient feature in the direction from the head to the tail. Therefore, the feature matrix is ​​segmented to obtain the feature vector, and the third weight is assigned according to the gradient.

[0098] In one embodiment of the present invention, obtaining the distance and time of the target sea condition includes:

[0099] The remote sensing image of the sea area at a distance L from the target fleet is obtained based on the ship-borne UAV, and the remote sensing image of the sea area is recognized based on the recognition model to obtain the target sea conditions;

[0100] If the target sea condition is the same as the current sea condition, the dynamic switching of the established formation will not be performed;

[0101] If the target sea condition is inconsistent with the current sea condition, the distance time is calculated, including:

[0102] Get the current moving speed V of the target fleet, then the distance time is .

[0103] It should be noted that the recognition model includes but is not limited to open source models with image recognition capabilities such as Chatgpt or Tongyi Qianwen.

[0104] In one embodiment of the present invention, dynamic switching is performed, including:

[0105] Obtain the target predetermined formation corresponding to the target sea condition, and obtain the target conversion time based on the current feature matrix s corresponding to the current predetermined formation and the target feature matrix g corresponding to the target predetermined formation;

[0106] If the target conversion time ≤ the distance time, a first switching is performed, specifically including: when the distance time is 0, the current feature matrix is ​​just converted to the target feature matrix;

[0107] If the target conversion time is greater than the distance time, the second switching is performed, which specifically includes:

[0108] Get the candidate feature matrix b of the target feature matrix; where the correlation value between the current feature matrix and the candidate feature matrix , , represents the proportionality coefficient, ;

[0109] based on Sort the candidate feature matrices from large to small to obtain feature sorting;

[0110] Based on the order of feature sorting, judge one by one from front to back:

[0111] If the candidate conversion time between the current feature matrix and the candidate feature matrix ranked r in the feature ranking is greater than the distance time, the candidate feature matrix ranked r is abandoned, and the candidate feature matrix ranked r+1 is judged;

[0112] If the alternative conversion time between the current feature matrix and the alternative feature matrix ranked r in the feature sorting is ≤ the distance time, the alternative feature matrix ranked r replaces the target feature matrix, and when the distance time is 0, the current feature matrix is ​​exactly converted to the alternative feature matrix ranked r.

[0113] Specifically, according to the current sea conditions and the target sea conditions, the system first determines the target formation corresponding to the target sea conditions. Using the characteristic matrix corresponding to the current formation (denoted as s) and the characteristic matrix corresponding to the target formation (denoted as g), the conversion time required to switch from the current formation to the target formation is calculated, which is called the "target conversion time". The target conversion time reflects the time required to switch from the current formation to the formation that best adapts to the target sea conditions, and serves as a key indicator for subsequent decision-making.

[0114] Compare the target conversion time with the current remaining distance time. The remaining distance time refers to the time left before entering the new sea state. If the target conversion time is less than or equal to this time, it means that the direct conversion to the target formation can be completed within the time, thus ensuring that the fleet is in a safe state when entering the new sea state. ≤Distance Time , then the first switch is performed: when the distance time is 0, the current feature matrix is ​​at the distance time The conversion is started when the waiting distance time is 0, and the conversion is successful. Through this method, the dual influence of the current sea state and the target sea state can be minimized.

[0115] Specifically, when the target conversion time is greater than the distance time, it means that the direct conversion to the target formation cannot be completed within the remaining time. At this time, the second switching scheme is initiated: extract the set of alternative feature matrices from the target feature matrix. The alternative matrix represents some candidate formations that are similar to the target formation but have a shorter conversion time. The system calculates the correlation value between the current feature matrix and each alternative feature matrix. The alternative feature matrices are sorted from large to small according to the correlation value to obtain a feature ranking. The sorted order reflects the matching degree and conversion feasibility of each alternative formation with the target formation. Judge and select the appropriate alternative formations one by one, and judge one by one from the ranking according to the rank: if the alternative conversion time between the current feature matrix and the alternative feature matrix ranked r is still greater than the distance time, then abandon the alternative formation and judge the next one (r+1). If it is found that the conversion time of a certain alternative formation is ≤ the distance time, then select the alternative feature matrix to replace the original target feature matrix; when the distance time is 0, the current feature matrix will be exactly converted to the selected alternative formation. In this method, when the target formation conversion time cannot meet the urgent requirements, a "compromise" given formation can be selected within a limited time through alternative options. Sorting and position-by-position judgment ensure that the selection process takes into account both the time required for conversion and the morphological matching and relevance, thereby ensuring safety and adaptability.

[0116] In one embodiment of the present invention, the target fleet is controlled to switch at a plurality of preset time intervals based on the switching path corresponding to the first switching or the second switching.

[0117] Specifically, based on the determined switching scheme, whether it is direct switching (first switching) or alternative switching (second switching), an optimal switching path has been calculated. In each preset time interval, the system gradually instructs the fleet to change its position and formation according to the intermediate state matrix predetermined in the optimal switching path until the switching is finally completed.

[0118] A combined fleet intelligent navigation control method, the method comprising:

[0119] Based on the N predetermined formations of the target fleet, the feature matrix corresponding to each predetermined formation is abstractly obtained;

[0120] Analyze the conversion paths of each given formation to the remaining given formations and the conversion time corresponding to each conversion path based on the characteristic matrix;

[0121] Calculate the correlation value between any two feature matrices;

[0122] The distance time of the target sea condition is obtained, and the established formation is dynamically switched within the distance time based on the conversion path, the conversion time corresponding to each conversion path and the associated value.

[0123] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A combined fleet intelligent navigation control system, characterized in that: include: The feature construction module is used to abstractly obtain the feature matrix corresponding to each predetermined formation based on the N predetermined formations of the target fleet; A matrix conversion module, used for analyzing the conversion paths of each predetermined formation to the remaining predetermined formations and the conversion time corresponding to each conversion path based on the characteristic matrix; Matrix association module, used to calculate the association value between any two feature matrices; The fleet management module is used to obtain the distance time of the target sea condition and dynamically switch the established formation within the distance time based on the conversion path, the conversion time corresponding to each conversion path and the associated value.

2. The combined fleet intelligent navigation control system according to claim 1, characterized in that: Get the feature matrix corresponding to each given formation, including: Obtain the plane geometric structure of each given formation and the corresponding remote sensing image; Extract the center of gravity of the plane geometric structure; Extracting the plane geometric structure of the corresponding predetermined formation from the remote sensing image, and marking the center of gravity of the plane geometric structure in the remote sensing image; Aligning N remote sensing images according to the centroid point to obtain overlapping rectangles of the N remote sensing images; Rasterize each overlapping rectangle, including: Initialize several individuals that meet the constraints, each of which represents the number of rows and columns for rasterizing overlapping rectangles; The constraints include: for any overlapping rectangles, there is at most one unmanned ship in each of the cells obtained by rasterizing them; The fitness function includes: maximizing the number of rows and columns; Based on the constraints and fitness function, the number of rows and columns of each overlapping rectangle rasterization is obtained through a genetic algorithm, and the corresponding feature matrix is ​​obtained; wherein the feature matrix is ​​a multi-value matrix, which includes P+1 elements, P is the number of unmanned ships in the target fleet, each unmanned ship corresponds to an element, and the remaining elements in the multi-value matrix are all 0.

3. The combined fleet intelligent navigation control system according to claim 2 is characterized in that: The conversion paths and the conversion time corresponding to each conversion path include: Get the conversion path and conversion time between the i-th established formation and the j-th established formation, i≠j, specifically: Get the corresponding i-th feature matrix and j-th feature matrix; Establish movement constraints: for any non-zero element of a feature matrix, obtain the corresponding eight-neighborhood element values; within a preset time interval, if the neighborhood element is 0, the non-zero element is allowed to move to the corresponding neighborhood; if the neighborhood element is not 0, the non-zero element is not allowed to move to the corresponding neighborhood; Based on the movement constraint, the i-th feature matrix is ​​converted to the j-th feature matrix, and the simulation conversion is repeated several times to obtain the sum of the preset time intervals of each simulation conversion; Selecting the simulation conversion with the smallest sum of preset time intervals as the conversion path from the i-th feature matrix to the j-th feature matrix, and taking the sum of the preset time intervals as the conversion time; Among them, the conversion path represents the intermediate state matrix and the jth characteristic matrix corresponding to several preset time intervals in the process of converting the i-th characteristic matrix to the j-th characteristic matrix, and the intermediate state matrix is ​​obtained based on the corresponding simulation conversion collection.

4. The combined fleet intelligent navigation control system according to claim 3 is characterized in that: The correlation value between any two feature matrices is calculated, including: Based on the established formations corresponding to the i-th feature matrix and the j-th feature matrix, the head and tail of the j-th feature matrix are determined respectively; wherein the head represents the head of the established formation, and the tail represents the tail of the established formation; The feature matrix corresponding to the given formation is classified. The classification types include: forward type, and descend to stable type according to a fixed gradient, a total of K types, and a first weight is assigned to each type. The value range of the first weight is , and the first The value of the first weight of each type is 1; Based on the first weight of the j-th feature matrix, the first calculation, the second calculation or the third calculation is performed to obtain the correlation value between the i-th feature matrix and the j-th feature matrix .

5. The combined fleet intelligent navigation control system according to claim 4, characterized in that: If the first weight of the j-th feature matrix is ​​greater than 1, the first calculation is performed; for the i-th feature matrix and the j-th feature matrix, the matrix is ​​split based on the parallel lines connecting the head and the tail of the team, and M feature vectors are obtained. The second weight is assigned to the feature vector based on the Euclidean distance between each feature vector and the center of the corresponding feature matrix. The value range of the second weight is , and the sum of several second weights is 1; where the Euclidean distance of the eigenvector is inversely proportional to the second weight, and if the Euclidean distances of any two eigenvectors are equal, the corresponding second weights are equal; calculate the association value of the i-th eigenmatrix and the j-th eigenmatrix , that is, the association value is equal to the first weight of the j-th matrix multiplied by the weighted sum of the normalized distances of each pair of eigenvectors.

6. The combined fleet intelligent navigation control system according to claim 4, characterized in that: If the first weight of the j-th feature matrix is ​​equal to 1, the second calculation is performed; the Hamming distance is calculated for the i-th feature matrix and the j-th feature matrix, and the Hamming distance is used as the correlation value between the i-th feature matrix and the j-th feature matrix .

7. The combined fleet intelligent navigation control system according to claim 4, characterized in that: If the first weight of the j-th feature matrix is ​​less than 1, the third calculation is performed; for the i-th feature matrix and the j-th feature matrix, matrix segmentation is performed based on the direction from the head to the tail, and N feature vectors are obtained. The value range of the third weight assigned to the N feature vectors based on the direction from the head to the tail is , and the sum of several third weights is 1; wherein the third weight decreases in the direction from the head to the tail; calculate the correlation value between the i-th feature matrix and the j-th feature matrix , that is, the association value is equal to the first weight of the j-th matrix multiplied by the weighted sum of the gradients of each pair of eigenvectors.

8. The combined fleet intelligent navigation control system according to claim 4, characterized in that: Get the distance and time of target sea conditions, including: The remote sensing image of the sea area at a distance L from the target fleet is obtained based on the ship-borne UAV, and the remote sensing image of the sea area is recognized based on the recognition model to obtain the target sea conditions; If the target sea condition is the same as the current sea condition, the dynamic switching of the established formation will not be performed; If the target sea condition is inconsistent with the current sea condition, the distance time is calculated, including: Get the current moving speed V of the target fleet, then the distance time is .

9. The combined fleet intelligent navigation control system according to claim 8, characterized in that: Perform dynamic switching, including: Obtain the target predetermined formation corresponding to the target sea condition, and obtain the target conversion time based on the current feature matrix s corresponding to the current predetermined formation and the target feature matrix g corresponding to the target predetermined formation; If the target conversion time ≤ the distance time, a first switching is performed, specifically including: when the distance time is 0, the current feature matrix is ​​just converted to the target feature matrix; If the target conversion time is greater than the distance time, the second switching is performed, which specifically includes: Get the candidate feature matrix b of the target feature matrix; where the correlation value between the current feature matrix and the candidate feature matrix ; based on Sort the candidate feature matrices from large to small to obtain feature sorting; Based on the order of feature sorting, judge one by one from front to back: If the candidate conversion time between the current feature matrix and the candidate feature matrix ranked r in the feature ranking is greater than the distance time, the candidate feature matrix ranked r is abandoned, and the candidate feature matrix ranked r+1 is judged; If the alternative conversion time between the current feature matrix and the alternative feature matrix ranked r in the feature sorting is ≤ the distance time, the alternative feature matrix ranked r replaces the target feature matrix, and when the distance time is 0, the current feature matrix is ​​exactly converted to the alternative feature matrix ranked r.

10. A combined fleet intelligent navigation control method, characterized in that: The method is used to execute the system according to any one of claims 1 to 9, and the method comprises: Based on the N predetermined formations of the target fleet, the feature matrix corresponding to each predetermined formation is abstractly obtained; Analyze the conversion paths of each given formation to the remaining given formations and the conversion time corresponding to each conversion path based on the characteristic matrix; Calculate the correlation value between any two feature matrices; The distance time of the target sea condition is obtained, and the established formation is dynamically switched within the distance time based on the conversion path, the conversion time corresponding to each conversion path and the associated value.

Images