A method, device and storage medium for optimizing target search

Abstract

The present application discloses an optimization method, device and storage medium for target search. The method includes: obtaining trajectory information of a target object to be searched, encoding the trajectory information based on a grid formed by area division, and determining a coding sequence associated with the trajectory information; determining a feature vector sequence corresponding to the trajectory information according to the coding sequence; determining a search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network; determining multiple candidate potential points at the edge of the search area; determining a target potential point corresponding to each drone from multiple candidate potential points using a genetic algorithm based on the positions and speeds of multiple drones; and dispatching the drone to reach the corresponding target potential point and search for the target object.

Description

Technical Field

[0001] The present application relates to the technical field of target search, and in particular to an optimization method, device and storage medium for target search. Background Art

[0002] Using drones for target search is a complex and delicate task, especially when multiple drones are used to search for targets in collaboration. It is not only necessary to predict the activity area of ​​the target object to be searched, but also to schedule and coordinate multiple drones.

[0003] Invention patent CN119247980A discloses a method and device for searching for moving targets of drone clusters, the method comprising: when M drones search for N targets in the search area, determining the probability expectation value and freshness value of the grid where the drone is located; determining the local probability expectation matrix and the global probability expectation matrix of the grid according to the probability expectation value, and determining the local freshness matrix and the global freshness matrix of the grid according to the freshness value; determining the action decision matrix according to the probability expectation decision matrix, the freshness decision matrix, the exploration task weight and the search task weight; determining the target grid that minimizes the freshness value and maximizes the probability value according to the action decision matrix, and determining the next movement direction of the drone according to the location of the target grid. This scheme utilizes both the target location probability information and the previously collected prior information, and can search for targets with known locations and targets with unknown locations, so that the search for targets in realistic scenarios can be more efficient and accurate.

[0004] Invention patent CN118625854A discloses a method for searching for target of drone clusters, so as to improve the search efficiency of drone clusters under the conditions of limited communication and energy. The method comprises the following steps: obtaining real-time data of drone clusters placed in the same local communication network; fusing the attraction pheromone and the repulsion pheromone and determining the search cost; determining the energy consumption cost and dispersion cost according to the drone coordinates, drone energy, and landing point coordinates; step S04, according to the search cost, energy consumption cost, dispersion cost and the flight alternative action of a drone in the local communication network at the next moment, using the dung beetle optimization algorithm to obtain the optimal solution of the local utility function; determining whether to update the flight action of the drone at the next moment according to the optimal solution of the local utility function; and executing step S04 in response to the fact that the flight action of the drone at the next moment does not need to be updated.

[0005] Therefore, in order to use drone swarms for target search more accurately and efficiently, it is necessary to optimize the prediction of the activity area of ​​the target object and the scheduling of drones. The present invention aims to propose an optimization scheme for target search, so as to more accurately predict the activity area of ​​the search target and more efficiently and accurately schedule the drone swarm. Summary of the invention

[0006] The embodiments of the present disclosure provide a method, device and storage medium for optimizing target search, so as to more accurately predict the activity area of ​​the search target and more efficiently and accurately schedule the drone cluster.

[0007] According to one aspect of an embodiment of the present disclosure, there is provided a method for optimizing target search, including: obtaining trajectory information of a target object to be searched, and encoding the trajectory information based on a grid formed by area division to determine a coding sequence associated with the trajectory information; determining a feature vector sequence corresponding to the trajectory information according to the coding sequence; determining a search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network; determining a plurality of candidate potential points at the edge of the search area; determining a target potential point corresponding to each drone from a plurality of candidate potential points using a genetic algorithm based on the positions and speeds of a plurality of drones; and dispatching the drone to reach the corresponding target potential point and search for the target object.

[0008] According to another aspect of the embodiments of the present disclosure, a storage medium is further provided. The storage medium includes a stored program, wherein the above method is executed by a processor when the program is running.

[0009] According to another aspect of an embodiment of the present disclosure, there is also provided an optimization device for target search, including: an encoding module, used to obtain trajectory information of a target object to be searched, and to encode the trajectory information based on a grid formed by area division, and to determine a coding sequence associated with the trajectory information; a feature vector determination module, used to determine a feature vector sequence corresponding to the trajectory information according to the coding sequence; a search area determination module, used to determine a search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network; a candidate potential point determination module, used to determine multiple candidate potential points at the edge of the search area; a target potential point determination module, used to determine a target potential point corresponding to each drone from multiple candidate potential points using a genetic algorithm based on the positions and speeds of multiple drones; and a scheduling module, used to schedule the drone to reach the corresponding target potential point and search for the target object.

[0010] According to another aspect of an embodiment of the present disclosure, there is also provided an optimization device for target search, comprising: a processor; and a memory connected to the processor, for providing the processor with instructions for processing the following processing steps: obtaining trajectory information of a target object to be searched, and encoding the trajectory information based on a grid formed by area division, and determining a coding sequence associated with the trajectory information; determining a feature vector sequence corresponding to the trajectory information according to the coding sequence; determining a search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network; determining multiple candidate potential points at the edge of the search area; determining target potential points corresponding to each drone from multiple candidate potential points using a genetic algorithm based on the positions and speeds of multiple drones; and dispatching the drone to the corresponding target potential point and searching for the target object.

[0011] Therefore, this embodiment uses the correspondence between the grid and the trajectory information of the target object for encoding, forms a coding sequence corresponding to the trajectory information and further converts the coding sequence into a feature vector sequence, and then uses a positioning prediction model based on a neural network to determine the search area corresponding to the target object. In this way, the search area of ​​the target object can be locked more accurately. Then, based on the current position and speed of the drone, a genetic algorithm is used to determine the target potential point corresponding to the drone at the edge of the search area, and then the drone is dispatched to the corresponding target potential point and the target object is searched, so that the drone group can be dispatched more efficiently and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present application. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0013] Figure 1 is a hardware structure block diagram of a computing device for implementing the method according to Embodiment 1 of the present disclosure;

[0014] Figure 2 is a schematic diagram of a target search system according to Embodiment 1 of the present disclosure;

[0015] Figure 3 is a flow chart of the target search optimization method according to the first aspect of Embodiment 1 of the present disclosure;

[0016] Figure 4 is a schematic diagram of dividing a sea area into multiple grids according to Embodiment 1 of the present disclosure;

[0017] Figure 5 is a schematic diagram of a target search area according to Embodiment 1 of the present disclosure;

[0018] Figure 6~Figure 8 is a schematic diagram of a positioning prediction model according to Embodiment 1 of the present disclosure;

[0019] Fig. 9 is a schematic diagram of the initial position distribution of multiple drones relative to the target search area according to Embodiment 1 of the present disclosure;

[0020] Fig.10 is a schematic diagram of a process of determining a target potential point corresponding to each drone from a plurality of candidate potential points using a genetic algorithm according to Embodiment 1 of the present disclosure;

[0021] Fig.11 is a schematic diagram of an optimization device for target search according to Embodiment 2 of the present disclosure; and

[0022] Fig.12 It is a schematic diagram of an optimization device for target search according to Example 3 of the present disclosure. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0025] Example 1

[0026] According to this embodiment, a method embodiment of an optimization method for target search is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] The method embodiment provided in this embodiment can be executed in a server or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computing device for implementing an optimization method for target search. Figure 1 As shown, the computing device may include one or more processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, the transmission device, and the input / output interface are connected to the processor via a bus. In addition, it may also include: a display, a keyboard, and a cursor control device connected to the input / output interface. A person skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0028] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computing device. As involved in the embodiments of the present disclosure, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0029] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the optimization method of target search in the embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the optimization method of target search of the above-mentioned application program is realized. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0030] The transmission device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computing device. In one example, the transmission device includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0031] The display may be, for example, a touch screen liquid crystal display (LCD) that may enable a user to interact with a user interface of the computing device.

[0032] It should be noted that, in some optional embodiments, the above Figure 1 The computing device shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware elements and software elements. Figure 1 This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computing devices described above.

[0033] Figure 2 is a schematic diagram of a target search system according to this embodiment. Figure 2 As shown, the system includes: a control center 100 and a plurality of drones 201-20N. The control center 100 communicates with the target object 300 in real time, for example, by radio, so as to obtain the movement trajectory information of the target object 300 in real time. When the control center 100 loses contact with the target object 300, the control center 100 can control the drones 201-20n by radio to search for the target object 300. In this embodiment, the target object 300 can be, for example, a ship sailing on the sea.

[0034] It should be noted that the control center 100 in the system can be applicable to the hardware structure described above.

[0035] Under the above operating environment, according to the first aspect of this embodiment, a method for optimizing target search is provided. The method comprises: Figure 2 The control center 100 shown in FIG. Figure 3 A schematic diagram showing the process of the method is shown in FIG. Figure 3 As shown, the method includes:

[0036] S302: Acquire trajectory information of a target object to be searched, encode the trajectory information based on a grid formed by area division, and determine a coding sequence associated with the trajectory information;

[0037] S304: Determine a feature vector sequence corresponding to the trajectory information according to the coding sequence;

[0038] S306: Determine a search area corresponding to the target object using a positioning prediction model based on a neural network according to the feature vector sequence;

[0039] S308: Determine multiple candidate potential points at the edge of the search area;

[0040] S310: Based on the positions and speeds of the multiple drones, using a genetic algorithm to determine a target potential point corresponding to each drone from the multiple candidate potential points; and

[0041] S312: dispatching the UAV to the corresponding target potential point and searching for the target object.

[0042] Specifically, refer to Figure 2 As shown, the control center 100 communicates with the target object 300 to be searched in real time, for example, by radio, so as to obtain the movement trajectory information of the target object 300 in real time. When the target object 300 loses contact with the control center 100, in order to search for the target object 300, the control center 100 first obtains the trajectory information of the target object. The trajectory information is composed of multiple nodes. For example, the trajectory information S can be expressed as: S={N1, N2, N3, ..., N S}. Where N s (s=1~S) is the node information contained in the trajectory information S.

[0043] Each node information includes an information pair consisting of time and location: N s =(T s , L s ). Where N s Represents the node information of the trajectory; T s is the time information corresponding to the node information; L s is the location information corresponding to the node information.

[0044] refer to Figure 4 As shown, in this embodiment, the sea area is divided into a plurality of grids G1 to G M . Thus, the control center 100 can encode the trajectory information according to the correspondence between the position information in each node information of the trajectory information of the target object 300 and the grid. Specifically, the control center 100 determines the grid corresponding to each node according to the position information of each node. And based on this, each node is encoded to obtain the encoding information corresponding to each node, so that the encoding information of each node is arranged in time sequence to form a coding sequence associated with the trajectory information (S302). The specific encoding operation is described in detail below.

[0045] Due to the sparsity of the coding sequence, the control center 100 processes the coding information in the coding sequence to obtain a feature vector corresponding to each coding information (S304). The specific processing process is described in detail below.

[0046] Then, the control center 100 uses a positioning prediction model based on a neural network to determine a search area corresponding to the target object 300 (S306). Figure 5 As shown, the control center 100 is from the grid G1~G M The grid Gx where the target object 300 may be active is predicted, thereby determining the search area Z of the target object 300 according to the predicted grid Gx. The specific process is described in detail below.

[0047] Then, the control center 100 determines a plurality of candidate potential points at the edge of the search area Z. The so-called potential points are the capture points that the drone needs to reach first when searching. Thus, for example, the control center 100 determines a plurality of candidate potential points at the edge of the search area in an evenly distributed manner (S308).

[0048] Then, the control center 100 determines the target potential point corresponding to each drone 201 - 20N from a plurality of candidate potential points using a genetic algorithm according to the position and speed of the drones 201 - 20N ( S310 ). The specific process is described in detail below.

[0049] Then, the control center 100 dispatches the drones 201 - 20N to reach the target potential point and searches for the target object 300 ( S312 ).

[0050] As described in the background technology, in order to be able to use drone swarms for target search more accurately and efficiently, it is necessary to optimize the prediction of the target's activity area and the scheduling of drones. In view of this, this embodiment uses the correspondence between the grid and the trajectory information of the target object for encoding, forms a coding sequence corresponding to the trajectory information and further converts the coding sequence into a feature vector sequence, and then uses a positioning prediction model based on a neural network to determine the search area corresponding to the target object. In this way, the search area of ​​the target object can be locked more accurately. Then, based on the current position and speed of the drone, a genetic algorithm is used to determine the target potential point corresponding to the drone at the edge of the search area, and then the drone is dispatched to the corresponding target potential point and the target object is searched, so that the drone swarm can be dispatched more efficiently and accurately.

[0051] Optionally, encoding the trajectory information based on the grid formed by the area division and determining the operation of the coding sequence associated with the trajectory information includes: determining the grid corresponding to the position information of each node in the trajectory information; encoding the position information by one-hot encoding according to the determined grid to determine the coding information corresponding to the position information; and arranging the coding information corresponding to each position information in time sequence to form a coding sequence.

[0052] Specifically, refer to Figure 4 As shown in FIG. 1 , the sea area is divided into M grids. Therefore, when the control center 100 encodes the trajectory information S, it first encodes the trajectory information S according to the nodes N. s Location information L s , determine which grid the node corresponds to.

[0053] For example, when according to node N s Location information L s , determine the location information L of the node s In grid G m , the position information can be encoded by one-hot encoding based on this. Specifically, the encoded information after the one-hot encoding is a vector with a dimension of M. When the position information L s In grid G m When , the mth element of the coded information takes the value "1" and the remaining elements take the value "0".

[0054] Thus, in this way, the encoding operation of the trajectory information S is realized. Thus, the encoding sequence {I1, I2, I3,..., I S}. Among them, I s is related to node information N s The corresponding encoding information.

[0055] Further optionally, the operation of determining the feature vector sequence corresponding to the trajectory information according to the coding sequence includes: performing word embedding processing on the coding information using a pre-trained word embedding model to determine the feature vector corresponding to the coding information.

[0056] Specifically, a pre-trained word embedding model, such as CBOW or skip-gram, can be used to generate the corresponding I according to the encoded information. s The eigenvector F s Among them, word embedding models such as CBOW or skip-gram can be trained using the massive trajectory data formed by the actual movement trajectories of various ships at sea. I will not go into details here.

[0057] Thus, the control center 100 obtains the feature vector sequence {F1, F2, F3, ..., F S}.

[0058] Further optionally, according to the feature vector sequence, using a positioning prediction model based on a neural network, an operation of determining the search area corresponding to the target object includes: inputting the feature vector sequence into the positioning prediction model; using the fully connected layer of the positioning prediction model to output integral information corresponding to each grid, wherein every two elements of the fully connected layer correspond to a grid, and outputting the integral of the corresponding grid belonging to the search area and the integral of the corresponding grid not belonging to the search area, respectively; and inputting the integral information of the fully connected layer into a softmax classification layer, wherein the softmax classification layer includes a softmax classifier corresponding to each grid, and is used to output the probability that the corresponding grid belongs to the search area and the probability that the corresponding grid does not belong to the search area, respectively.

[0059] Specifically, refer to Figure 6 As shown, the positioning prediction model based on the neural network may include, for example, an input layer, a hidden layer, an output layer and a softmax layer, wherein the output layer may be used as a fully connected layer.

[0060] Then further refer to Figure 6 As shown, every two neurons in the fully connected layer correspond to one grid. One neuron is used to output the integral of the grid belonging to the search area, and the other neuron is used to output the integral of the grid not belonging to the search area. Thus, the fully connected layer can output the integral of each grid belonging to the search area and the integral of each grid not belonging to the search area.

[0061] Then further refer to Figure 6 As shown, the fully connected layer includes softmax classifiers Q1~Q corresponding to each grid M Among them, each softmax classifier outputs the probability that the corresponding grid belongs to the search area and the probability that it does not belong to the search area.

[0062] It should be noted that more than one softmax classifier may output the result that the corresponding grid belongs to the search area, so that more grids that may be the search area can be found in this way to avoid omissions in the search. Therefore, through this embodiment, the prediction model based on the neural network can be used to accurately determine the search area of ​​the target object 300.

[0063] also, Figure 6 The structure of the prediction model shown can also have other variations. Figure 7 As shown, the feature vector sequence {F1, F2, F3, ..., F S}You can first input the BERT module, and then input the fully connected layer and softmax layer. For example, refer to Figure 8As shown, the feature vector sequence {F1, F2, F3, ..., F S}You can first input a convolutional layer containing multiple convolution kernels, and then input a fully connected layer.

[0064] Further preferably, the operation of determining the search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network also includes: determining a circular area as the search area, wherein the circular area envelope determines a grid belonging to the search area.

[0065] Specifically, refer to Figure 5 As shown, after determining the grids Gx belonging to the search area, the control center 100 determines a circular area including all grids Gx as the search area.

[0066] Optionally, the operation of determining a plurality of candidate potential points at the edge of the search area includes: determining a plurality of candidate potential points on a contour line of the circular area according to a preset center angle of the circle.

[0067] Specifically, for example, 3.6° may be used as the central angle, so that 100 candidate points may be determined on the circular contour of the search area. For ease of explanation, in this embodiment, it is assumed that J candidate potential points are determined on the circular contour of the search area.

[0068] So refer to Fig. 9 As shown, N drones 201 to 20N are deployed around the circular search area Z. Fig. 9 Not shown in FIG. 1 , J candidate potential points are also determined on the circular contour line of the circular search area Z. FIG.

[0069] Therefore, in this embodiment, the control center 100 needs to determine N target potential points corresponding to the N drones from the J candidate potential points, and then pair each drone with the corresponding target potential point.

[0070] Optionally, based on the positions and speeds of multiple drones, an operation of determining a target potential point corresponding to each drone from the multiple candidate potential points using a genetic algorithm includes: defining a chromosome X, wherein the number of genes in the chromosome corresponds to the number of candidate potential points, and when the gene value of the chromosome is 1, it indicates that the candidate potential point corresponding to the gene is the target potential point, and when the gene value of the chromosome is 0, it indicates that the candidate potential point corresponding to the gene is not the target potential point; randomly initializing a population according to a constraint condition to obtain a population that meets the constraint condition, wherein the constraint condition indicates that the number of chromosomes whose gene value is 1 is equal to the number of drones; performing selection, crossover and mutation operations on the population according to a pre-set fitness function; iterating the selection, crossover and mutation operations, and determining the target potential point according to the chromosome with the largest fitness; and pairing the drones and the target potential points using the Hungarian algorithm to determine the target potential points corresponding to each drone.

[0071] Among them, the fitness function is:

[0072] .

[0073] Among them, w1~w3 are weights, satisfying w1~w3>0 and the sum of w1~w3 is equal to 1, Nor( ) is the normalization function, arccot ​​is the inverse cotangent function; D represents the sum of the distances between each UAV and the corresponding target potential point; represents the variance of the center angle of the arc defined by each adjacent target potential point; and Represents the variance of the flight time of each UAV to reach the corresponding target potential point.

[0074] Specifically, Fig.10 FIG. 2 shows a flow chart of determining a target potential point corresponding to each drone from multiple candidate potential points using a genetic algorithm in this embodiment. Fig.10 As shown, the method includes:

[0075] S1002: Define chromosomes.

[0076] In order to determine the target potential point corresponding to the UAV from the candidate potential points through genetic algorithm, define chromosome X=[x1, x2, ..., x J ]. The genes x of the chromosome j Corresponding to the jth candidate potential point. j Takes "0" or "1". j = 1, it means that the jth candidate potential point is the target potential point; and when x j =0, it means that the jth candidate potential point is not the target potential point.

[0077] S1004: Define fitness function:

[0078] First, the first constraint condition is determined. For the target potential points represented by the gene with a value of "1" in chromosome X, the UAVs are paired with the target potential points using the Hungarian algorithm. The sum of the distances between each UAV and the corresponding target potential point is minimized, that is:

[0079]

[0080] Among them, D n represents the distance between the nth UAV and the corresponding target potential point. Therefore, under the first constraint, D takes the minimum value.

[0081] Secondly, the second constraint condition is determined. Each target potential point is evenly distributed on the circular contour of the search area, that is, the center angles of the arcs defined by each adjacent target potential point are as close as possible. The specific formula is expressed as follows:

[0082]

[0083] in, Indicates the center angle of the arc defined by the adjacent target potential points; Indicates the central angles of each circle The average value of Indicates the central angles of each circle The variance of . Among them, the central angle It can be determined by establishing a polar coordinate system based on the center of the search area and then determining the polar coordinates of each target potential point, which will not be described in detail here.

[0084] Next, determine the third constraint condition. Under the third constraint condition, each UAV should reach the corresponding target potential point at the same time as much as possible. The specific formula is:

[0085]

[0086] in, represents the flight time of the nth UAV to reach the corresponding target potential point; Represents the average flight time of each drone; Indicates the flight time of each drone The variance of Based on the above distance D n and the flight speed v of the nth drone n The calculation is obtained and will not be repeated here.

[0087] Therefore, the fitness function can be constructed according to the above constraints:

[0088]

[0089] Where w1~w3 are weights, satisfying w1~w3>0 and the sum of w1~w3 is equal to 1. Nor( ) is the normalization function. arccot ​​is the inverse cotangent function.

[0090] S1006: Initialize the population

[0091] Before initializing the population, we first determine the constraints. Considering that there are N drones to be dispatched, the number of target potential points that need to be determined is also N, so we have:

[0092] ;as well as

[0093] x j =0 or 1.

[0094] According to the above constraints, the population is randomly initialized to obtain the following population that meets the above constraints: {X1, X2, X3, ..., X K}.

[0095] S1008: Using the above fitness function, determine X1~X K The corresponding fitness value H1~H K .

[0096] S1010: According to fitness H1~H K Perform selection operation on the population to obtain the selected population: {Y1, Y2,Y3, ..., Y K}.

[0097] S1012: Perform crossover and mutation operations on the selected population to obtain the population: {Z ’ 1, Z ’ 2, Z ’ 3,..., Z ’ K}.

[0098] S1014: Normalize the population after crossover and mutation

[0099] Specifically, for the population {Z after crossover and mutation ’ 1, Z ’ 2, Z ’ 3, ..., Z ’ K}, the number of genes with a value of 1 is greater than the number of chromosomes in N, and the following operations are performed:

[0100] Using the Hungarian algorithm, the drones are paired with candidate potential points corresponding to genes with a value of 1 in the chromosome, so that the value of the gene corresponding to the potential point paired with the drone remains unchanged at 1, and the other genes with a value of 1 are changed to "0". Thus, the population is obtained: {Z1, Z2, Z3, ..., Z K}.

[0101] In this way, the population after crossover and mutation is normalized so that the chromosomes in the population meet the constraints: .

[0102] Then the population {Z1, Z2, Z3, ..., Z K} as the new round of iteration {X1, X2, X3, ..., X K}.

[0103] S1016: Repeat the operations of S1008 to S1014 until the condition for iteration termination is met.

[0104] Specifically, the conditions for terminating the iteration include: determining the chromosome with the largest fitness value; or the number of iterations reaches a predetermined threshold.

[0105] S1018: Using the Hungarian algorithm, the drones and target potential points are paired to determine the target potential points corresponding to each drone.

[0106] Thus, through the above method, the target potential point can be determined by using the genetic algorithm. The total distance of the drone to the search area can be minimized, thereby reducing the total consumption of the drone; in addition, after the drone arrives at the search area, it can be evenly distributed around the search area to avoid missing the target object; and the drones can arrive at the search area at the same time.

[0107] In addition, reference Figure 1 As shown, according to the second aspect of this embodiment, a storage medium is provided, wherein the storage medium includes a stored program, wherein when the program is run, a processor executes any one of the above methods.

[0108] Therefore, this embodiment uses the correspondence between the grid and the trajectory information of the target object for encoding, forms a coding sequence corresponding to the trajectory information and further converts the coding sequence into a feature vector sequence, and then uses a positioning prediction model based on a neural network to determine the search area corresponding to the target object. In this way, the search area of ​​the target object can be locked more accurately. Then, based on the current position and speed of the drone, a genetic algorithm is used to determine the target potential point corresponding to the drone at the edge of the search area, and then the drone is dispatched to the corresponding target potential point and the target object is searched, so that the drone group can be dispatched more efficiently and accurately.

[0109] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0110] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0111] Example 2

[0112] Fig.11 FIG. 1 shows an optimization device 1100 for target search according to this embodiment, which corresponds to the method according to the first aspect of embodiment 1. Fig.11As shown, the device 1100 includes: an encoding module 1102, which is used to obtain trajectory information of a target object to be searched, and encode the trajectory information based on a grid formed by area division, and determine a coding sequence associated with the trajectory information; a feature vector determination module 1104, which is used to determine a feature vector sequence corresponding to the trajectory information according to the coding sequence; a search area determination module 1106, which is used to determine a search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network; a candidate potential point determination module 1108, which is used to determine multiple candidate potential points at the edge of the search area; a target potential point determination module 1110, which is used to determine the target potential point corresponding to each drone from multiple candidate potential points based on the positions and speeds of multiple drones using a genetic algorithm; and a scheduling module 1112, which is used to schedule the drone to reach the corresponding target potential point and search for the target object.

[0113] Optionally, the encoding module 1102 includes: a grid determination submodule, used to determine the grid corresponding to the position information of each node in the trajectory information; a coding information determination submodule, used to encode the position information according to the determined grid, and determine the coding information corresponding to the position information; and a sequence construction submodule, used to arrange the coding information corresponding to each position information in time sequence to form a coding sequence.

[0114] Optionally, the feature vector determination module 1104 includes: a feature vector determination submodule, which is used to perform word embedding processing on the encoded information using a pre-trained word embedding model to determine a feature vector corresponding to the encoded information.

[0115] Optionally, the search area determination module 1106 includes: an input submodule, used to input the feature vector sequence into the positioning prediction model; a fully connected layer calculation submodule, used to use the fully connected layer of the positioning prediction model to output the integral information corresponding to each grid, wherein every two neurons of the fully connected layer correspond to a grid, and output the integral of the corresponding grid belonging to the search area and the integral of the corresponding grid not belonging to the search area, respectively; and a classification submodule, used to input the integral information of the fully connected layer into the softmax classification layer, wherein the softmax classification layer contains a softmax classifier corresponding to each grid, and outputs the probability that the corresponding grid belongs to the search area and the probability that the corresponding grid does not belong to the search area, respectively.

[0116] Optionally, the search area determination module 1106 further includes: a search area determination submodule, configured to determine a circular area as the search area, wherein the circular area envelops the grid determined to belong to the search area.

[0117] Optionally, the candidate potential point determination module 1108 includes: a candidate potential point determination submodule, configured to determine a plurality of candidate potential points on the contour line of the circular area according to a preset center angle of the circle.

[0118] Optionally, the target potential point determination module 1110 includes: a chromosome definition submodule, which is used to define a chromosome, wherein the number of genes in the chromosome corresponds to the number of candidate potential points, and when the gene value of the chromosome is 1, it indicates that the candidate potential point corresponding to the gene is the target potential point, and when the gene value of the chromosome is 0, it indicates that the candidate potential point corresponding to the gene is not the target potential point; a population initialization submodule, which is used to randomly initialize the population according to the constraint conditions to obtain a population that meets the constraint conditions, wherein the constraint conditions indicate that the number of chromosomes with gene values ​​of 1 is equal to the number of drones; a selection, crossover and mutation submodule, which is used to select, crossover and mutate the population according to a pre-set fitness function; an iteration submodule, which is used to iterate the selection, crossover and mutation operations, and determine the target potential point according to the chromosome with the largest fitness; and a pairing submodule, which is used to pair the drones and the target potential points using the Hungarian algorithm, so as to determine the target potential points corresponding to each drone.

[0119] Among them, the fitness function is: . Among them, w1~w3 are weights, satisfying w1~w3>0 and the sum of w1~w3 is equal to 1, Nor( ) is the normalization function, arccot ​​is the inverse cotangent function; D represents the sum of the distances between each UAV and the corresponding target potential point; represents the variance of the center angle of the arc defined by each adjacent target potential point; and Represents the variance of the flight time of each UAV to reach the corresponding target potential point.

[0120] Therefore, this embodiment uses the correspondence between the grid and the trajectory information of the target object for encoding, forms a coding sequence corresponding to the trajectory information and further converts the coding sequence into a feature vector sequence, and then uses a positioning prediction model based on a neural network to determine the search area corresponding to the target object. In this way, the search area of ​​the target object can be locked more accurately. Then, based on the current position and speed of the drone, a genetic algorithm is used to determine the target potential point corresponding to the drone at the edge of the search area, and then the drone is dispatched to the corresponding target potential point and the target object is searched, so that the drone group can be dispatched more efficiently and accurately.

[0121] Example 3

[0122] Fig.12 FIG. 1 shows an optimization device 1200 for target search according to this embodiment, and the device 1200 corresponds to the method according to the first aspect of embodiment 1. Fig.12As shown, the device 1200 includes: a processor; and a memory, connected to the processor, for providing the processor with instructions for processing the following processing steps: obtaining trajectory information of a target object to be searched, and encoding the trajectory information based on a grid formed by area division, and determining a coding sequence associated with the trajectory information; determining a feature vector sequence corresponding to the trajectory information according to the coding sequence; determining a search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network; determining multiple candidate potential points at the edge of the search area; determining target potential points corresponding to each drone from multiple candidate potential points using a genetic algorithm based on the positions and speeds of multiple drones; and dispatching the drone to the corresponding target potential point and searching for the target object.

[0123] Optionally, encoding the trajectory information based on the grid formed by the area division and determining the operation of the coding sequence associated with the trajectory information includes: determining the grid corresponding to the position information of each node in the trajectory information; encoding the position information by one-hot encoding according to the determined grid to determine the coding information corresponding to the position information; and arranging the coding information corresponding to each position information in time sequence to form a coding sequence.

[0124] Optionally, the operation of determining a feature vector sequence corresponding to the trajectory information according to the coding sequence includes: performing word embedding processing on the coding information using a pre-trained word embedding model to determine the feature vector corresponding to the coding information.

[0125] Optionally, an operation of determining a search area corresponding to a target object according to a feature vector sequence using a positioning prediction model based on a neural network includes: inputting the feature vector sequence into the positioning prediction model; using a fully connected layer of the positioning prediction model to output integral information corresponding to each grid, wherein every two neurons of the fully connected layer correspond to a grid, and output integrals of the corresponding grid belonging to the search area and integrals of the corresponding grid not belonging to the search area, respectively; and inputting the integral information of the fully connected layer into a softmax classification layer, wherein the softmax classification layer includes a softmax classifier corresponding to each grid, and is used to output probabilities of the corresponding grid belonging to the search area and probabilities of the corresponding grid not belonging to the search area, respectively.

[0126] Optionally, the operation of determining the search area corresponding to the target object according to the feature vector sequence using a neural network-based positioning prediction model also includes: determining a circular area as the search area, wherein the circular area envelope determines a grid belonging to the search area.

[0127] Optionally, the operation of determining a plurality of candidate potential points at the edge of the search area includes: determining a plurality of candidate potential points on a contour line of the circular area according to a preset center angle of the circle.

[0128] Optionally, based on the positions and speeds of multiple drones, an operation of determining a target potential point corresponding to each drone from multiple candidate potential points using a genetic algorithm includes: defining a chromosome, wherein the number of genes in the chromosome corresponds to the number of candidate potential points, and when the gene value of the chromosome is 1, it indicates that the candidate potential point corresponding to the gene is the target potential point, and when the gene value of the chromosome is 0, it indicates that the candidate potential point corresponding to the gene is not the target potential point; randomly initializing a population according to a constraint condition to obtain a population that meets the constraint condition, wherein the constraint condition indicates that the number of chromosomes with a gene value of 1 is equal to the number of drones; performing selection, crossover and mutation operations on the population according to a pre-set fitness function; iterating the selection, crossover and mutation operations, and determining the target potential point according to the chromosome with the largest fitness; and pairing the drones and the target potential points using the Hungarian algorithm to determine the target potential points corresponding to each drone.

[0129] Among them, the fitness function is: . Among them, w1~w3 are weights, satisfying w1~w3>0 and the sum of w1~w3 is equal to 1, Nor( ) is the normalization function, arccot ​​is the inverse cotangent function; D represents the sum of the distances between each UAV and the corresponding target potential point; represents the variance of the center angle of the arc defined by each adjacent target potential point; and Represents the variance of the flight time of each UAV to reach the corresponding target potential point.

[0130] Therefore, this embodiment uses the correspondence between the grid and the trajectory information of the target object for encoding, forms a coding sequence corresponding to the trajectory information and further converts the coding sequence into a feature vector sequence, and then uses a positioning prediction model based on a neural network to determine the search area corresponding to the target object. In this way, the search area of ​​the target object can be locked more accurately. Then, based on the current position and speed of the drone, a genetic algorithm is used to determine the target potential point corresponding to the drone at the edge of the search area, and then the drone is dispatched to the corresponding target potential point and the target object is searched, so that the drone group can be dispatched more efficiently and accurately.

[0131] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0132] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0133] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0134] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0135] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0136] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0137] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for optimizing target search, characterized in that: include: Acquire trajectory information of a target object to be searched, encode the trajectory information based on a grid formed by area division, and determine a coding sequence associated with the trajectory information; Determine a feature vector sequence corresponding to the trajectory information according to the coding sequence; According to the feature vector sequence, a positioning prediction model based on a neural network is used to determine a search area corresponding to the target object; Determining a plurality of candidate potential points at the edge of the search area; Based on the positions and speeds of the plurality of UAVs, a target potential point corresponding to each UAV is determined from the plurality of candidate potential points using a genetic algorithm; as well as The UAV is dispatched to reach the corresponding target potential point and search for the target object, and wherein: The operation of determining the search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network also includes: determining a circular area as the search area, wherein the circular area envelops the grid determined to belong to the search area, and wherein, The operation of determining a plurality of candidate potential points at the edge of the search area includes: determining a plurality of candidate potential points on the contour line of the circular area according to a preset center angle of the circle.

2. The method according to claim 1, characterized in that: The operation of encoding the trajectory information based on the grid formed by the area division and determining the coding sequence associated with the trajectory information includes: Determine the grid corresponding to the position information of each node in the trajectory information; One-hot encoding the position information according to the determined grid to determine encoding information corresponding to the position information; and The coding information corresponding to each position information is arranged in time sequence to form the coding sequence.

3. The method according to claim 2, characterized in that The operation of determining a feature vector sequence corresponding to the trajectory information according to the coding sequence includes: performing word embedding processing on the coding information using a pre-trained word embedding model to determine a feature vector corresponding to the coding information.

4. The method according to claim 3, characterized in that The operation of determining a search area corresponding to the target object according to the feature vector sequence and using a positioning prediction model based on a neural network includes: Inputting the feature vector sequence into the positioning prediction model; Utilizing the fully connected layer of the positioning prediction model to output integral information corresponding to each grid, wherein every two neurons of the fully connected layer correspond to one grid, and output the integral of the corresponding grid belonging to the search area and the integral of the corresponding grid not belonging to the search area respectively; and The integral information of the fully connected layer is input into the softmax classification layer, wherein the softmax classification layer comprises a softmax classifier corresponding to each grid, which is used to output the probability that the corresponding grid belongs to the search area and the probability that the corresponding grid does not belong to the search area.

5. The method according to claim 4, characterized in that The operation of determining a target potential point corresponding to each drone from the plurality of candidate potential points using a genetic algorithm based on the positions and speeds of the plurality of drones includes: Define a chromosome, where the number of genes in the chromosome corresponds to the number of candidate potential points, and when the gene value of the chromosome is 1, it indicates that the candidate potential point corresponding to the gene is the target potential point, and when the gene value of the chromosome is 0, it indicates that the candidate potential point corresponding to the gene is not the target potential point; According to the constraint condition, the population is randomly initialized to obtain a population that meets the constraint condition, wherein the constraint condition indicates that the number of chromosomes whose gene value is 1 is equal to the number of drones; According to a preset fitness function, selection, crossover and mutation operations are performed on the population; Iterating the selection, crossover and mutation operations, and determining a target potential point according to the chromosome with the maximum fitness; and The UAVs and the target potential points are paired using the Hungarian algorithm to determine the target potential points corresponding to each UAV, where The fitness function is: ,in, w1~w3 are weights, satisfying w1~w3>0 and the sum of w1~w3 is equal to 1, Nor() is the normalization function, and arccot ​​is the inverse cotangent function; D represents the sum of the distances between each UAV and the corresponding target potential point; represents the variance of the center angle of the arc defined by each adjacent target potential point; and Represents the variance of the flight time of each UAV to reach the corresponding target potential point.

6. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is run, the processor executes the method according to any one of claims 1 to 5.

7. An optimization device for target search, characterized in that: include: An encoding module, used to obtain trajectory information of a target object to be searched, encode the trajectory information based on a grid formed by area division, and determine a coding sequence associated with the trajectory information; A feature vector determination module, used to determine a feature vector sequence corresponding to the trajectory information according to the coding sequence; A search area determination module, used to determine a search area corresponding to the target object according to the feature vector sequence and using a positioning prediction model based on a neural network; A candidate potential point determination module, used to determine a plurality of candidate potential points at the edge of the search area; A target potential point determination module is used to determine a target potential point corresponding to each UAV from the plurality of candidate potential points using a genetic algorithm based on the positions and speeds of the plurality of UAVs; as well as A scheduling module is used to schedule the drone to reach the corresponding target potential point and search for the target object, and wherein: The search area determination module further comprises a search area determination submodule, which is used to determine a circular area as the search area, wherein the circular area envelops the determined grid belonging to the search area, and wherein, The candidate potential point determination module includes a candidate potential point determination submodule, which is used to determine a plurality of candidate potential points on the contour line of the circular area according to a preset center angle of the circle.

8. An optimization device for target search, characterized in that: include: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: Acquire trajectory information of a target object to be searched, encode the trajectory information based on a grid formed by area division, and determine a coding sequence associated with the trajectory information; Determine a feature vector sequence corresponding to the trajectory information according to the coding sequence; According to the feature vector sequence, a positioning prediction model based on a neural network is used to determine a search area corresponding to the target object; Determining a plurality of candidate potential points at the edge of the search area; Based on the positions and speeds of the plurality of UAVs, a target potential point corresponding to each UAV is determined from the plurality of candidate potential points using a genetic algorithm; as well as The UAV is dispatched to reach the corresponding target potential point and search for the target object, and wherein: The operation of determining the search area corresponding to the target object according to the feature vector sequence using a positioning prediction model based on a neural network also includes: determining a circular area as the search area, wherein the circular area envelops the grid determined to belong to the search area, and wherein, The operation of determining a plurality of candidate potential points at the edge of the search area includes: determining a plurality of candidate potential points on the contour line of the circular area according to a preset center angle of the circle.

Images