Target search methods, apparatus, devices, and storage media

Abstract

This disclosure provides a target search method, apparatus, device, and storage medium, relating to the field of computer technology, and particularly to the fields of unmanned aerial vehicles (UAVs), target search, and cooperative communication. The specific implementation involves: updating the feature values ​​of the grids traversed by the agent in the search area based on the target search results, wherein the feature values ​​of the grids are related to the probability of a target appearing in that grid; and predicting the motion information required by the agent to search for the target in the search area based on the feature values ​​of the grids in the search area. By employing the embodiments of this disclosure, the feature values ​​of the grids can be updated based on the search results of the agent in the search area, thereby improving the efficiency of the agent's target search.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to the fields of unmanned aerial vehicles, target search, and collaborative communication technology. Background Technology

[0002] In scenarios involving multiple UAVs collaboratively searching for targets, various search methods can be used. For example, a multi-colony ant colony algorithm can be employed. Ants from the same colony guide each other, while ants from different colonies repel each other. The optimal path in each generation is selected for pheromone updates, encouraging the algorithm to move towards a better path. Another approach is to use a grid method to divide and number the search area, creating an environmental map model; then, a multi-particle swarm optimization algorithm is used to optimize the collaborative path design for multiple UAVs, updating the highly dynamic environment through inter-UAV communication; finally, task allocation is achieved through information exchange. However, these algorithms are primarily applicable to scenarios where the prior information about the target is completely unknown, resulting in relatively low search efficiency. Summary of the Invention

[0003] This disclosure provides a target search method, apparatus, device, and storage medium.

[0004] According to one aspect of this disclosure, a target search method is provided, comprising:

[0005] Based on the target search results of the grids traversed by the agent in the search area, update the feature values ​​of the grids traversed by the agent, where the feature values ​​of the grids are related to the probability of the target appearing in the grid.

[0006] Based on the feature values ​​of the grid in the search area, predict the motion information required for the agent to search for the target in the search area.

[0007] According to another aspect of this disclosure, a target search apparatus is provided, comprising:

[0008] The update module is used to update the feature values ​​of the grids traversed by the agent based on the target search results of the grids traversed by the agent in the search area, wherein the feature values ​​of the grids are related to the probability of the target appearing in the grid.

[0009] The prediction module is used to predict the motion information required by the agent to search for the target in the search area based on the feature values ​​of the grid in the search area.

[0010] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0011] At least one processor; and

[0012] The memory is communicatively connected to the at least one processor; wherein,

[0013] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described in the present disclosure.

[0014] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform any of the methods according to embodiments of this disclosure.

[0015] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements any of the methods according to embodiments of this disclosure.

[0016] By employing embodiments of this disclosure, the feature values ​​of the grid can be updated based on the search results of the agent in the search area, thereby improving the efficiency of the agent in target search.

[0017] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0018] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0019] Figure 1 This is a schematic flowchart of a target search method according to an embodiment of the present disclosure;

[0020] Figure 2 This is a schematic flowchart of a target search method according to another embodiment of the present disclosure;

[0021] Figure 3 This is a schematic diagram illustrating an example of mesh division according to an embodiment of this disclosure;

[0022] Figure 4 This is a schematic diagram of another example of mesh division according to embodiments of this disclosure;

[0023] Figure 5 This is a schematic diagram of the set reference value according to the embodiments of this disclosure;

[0024] Figure 6 This is a schematic flowchart of a target search method according to another embodiment of the present disclosure;

[0025] Figure 7 This is a schematic diagram of the width ratio according to the embodiments of this disclosure;

[0026] Figure 8 This is a schematic diagram of path selection according to the embodiments of this disclosure;

[0027] Figure 9 This is a schematic diagram of steering direction selection according to an embodiment of the present disclosure;

[0028] Figure 10 This is a schematic flowchart of a target search method according to another embodiment of the present disclosure;

[0029] Figure 11 This is a schematic diagram of the structure of a target search device according to an embodiment of the present disclosure;

[0030] Figure 12 This is a schematic diagram of the structure of a target search device according to another embodiment of the present disclosure;

[0031] Figure 13 This is a block diagram of an electronic device used to implement embodiments of the present disclosure. Detailed Implementation

[0032] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0033] Figure 1 This is a flowchart illustrating a target search method according to an embodiment of the present disclosure, the method comprising:

[0034] S101. Based on the target search results of the grids traversed by the agent in the search area, update the feature values ​​of the grids traversed by the agent, wherein the feature values ​​of the grids are related to the probability of the target appearing in the grid.

[0035] S102. Based on the feature value of the grid in the search area, predict the motion information required for the agent to search for the target in the search area.

[0036] In this embodiment, the intelligent agent may include drones, satellites, cameras, sensors, intelligent robots, etc., without limitation. The number of intelligent agents can be single or multiple. Multiple intelligent agents can form an intelligent agent cluster when cooperating. The search area may include areas where the target may appear. The search area can be obtained by mapping actual geographical locations to a certain scale, or it can be independent of geographical location, for example, by setting the search area according to relative positions within a closed space. The search area may include multiple grids, which can be the same or different in size. The search area can be two-dimensional or three-dimensional. Correspondingly, the grids in the search area can be two-dimensional or three-dimensional. The feature value of the grid can be equal to the probability of the target appearing in that grid, or it can be a value calculated based on the probability of the target appearing in that grid. The feature value can also be called a score, parameter value, comparison value, possible value, etc., without limitation to specific names. The feature value of the grid is mainly used to represent the probability of the target appearing in that grid.

[0037] In this embodiment, the target is an object that the intelligent agent can recognize. The type of target can be various, such as images and / or signals of people, animals, or objects. Specific targets can be flexibly set according to the type of intelligent agent, the search scenario, etc. For example, in a search and rescue scenario, the target might be an image of a person or animal awaiting rescue. In a tracking scenario, the target might be an image and radar signal of a vehicle, person, or animal that needs to be tracked. Target search results can include whether the target exists, whether the target does not exist, the target location, the grid where the target is located, the target's movement speed, the target's movement direction, the search width percentage, and the attenuation coefficient. When the intelligent agent passes through a grid in the search area, it can determine whether the target to be searched is in that grid, and then update the feature values. For example, if the target is in the grid, the feature value of that grid is increased. If the target is not in the grid, the feature value of that grid is decreased.

[0038] In this embodiment, the motion information of the agent may include various parameters required for the agent's movement, such as movement distance, speed, direction, path, turning angle, duration, or turning interval. Furthermore, S102 can be executed under certain conditions. For example, if it is determined that an agent needs to change its current motion information, such as needing to turn, S102 is executed to predict the agent's motion information. If the agent does not need to turn, S102 may not be executed.

[0039] By employing embodiments of this disclosure, the feature values ​​of the grid can be updated based on the search results of the agent in the search area, thereby improving the efficiency of the agent in target search.

[0040] In one implementation, the search area may contain one or more agents. When multiple agents exist in the search area, they share the feature values ​​of the grids within the search area. In this case, in S101, the feature values ​​of the grids traversed by the multiple agents are updated based on the target search results of the grids they pass through in the search area, where the feature values ​​are related to the probability of a target appearing in that grid. In S102, the motion information required by one or more agents to search for the target in the search area is predicted based on the feature values ​​of the grids in the search area. That is, if multiple agents exist in the search area, the agent in S101 and the agent in S102 may not be the same agent. For example, the feature values ​​of the grids traversed by the first agent can be updated based on the target search results of the grids they pass through in the search area; then, the motion information required by a second agent to search for the target in the search area is predicted based on the feature values ​​of one or more grids in the search area.

[0041] The target search method of this disclosure can be executed by an intelligent agent or by a server in the background and / or middleware. Alternatively, some steps can be executed by an intelligent agent and some steps by a server. For example, an intelligent agent updates the feature values ​​of the grids it has traversed and uploads the updated feature values ​​to the server, which then predicts the subsequent motion information of one or more intelligent agents. Another example is that an intelligent agent uploads the target search results of the grids it has traversed to the server, which then updates the feature values ​​of the grids it has traversed and predicts the subsequent motion information of one or more intelligent agents. Yet another example is that after an intelligent agent updates the feature values ​​of the grids it has traversed, it combines this information with shared information received from the server or other intelligent agents, such as feature values ​​of other grids, to predict the subsequent motion information of the intelligent agent.

[0042] Taking a drone swarm as an example, a given search area is divided into multiple grids. After a drone completes a search of one or more grids, it can upload the search results to the swarm network's server. The server updates the feature values ​​of these grids and can use these updated feature values ​​to control the motion information of each drone in the swarm. For example, the server can plan a route for each drone in the swarm in real time. Alternatively, after updating the feature values ​​of each grid, the server can adjust the motion direction of each drone in real time. Furthermore, when a drone's motion time reaches a set duration or turning interval, the server uses the currently shared feature values ​​of each grid to plan a route or adjust its motion direction. The swarm network server can control the drones through a backend or middleware platform. The server can uniformly process the received search results, feature values ​​from multiple grids, and motion information from multiple agents, and issue commands. Alternatively, the computing, processing, and communication modules within a drone in the swarm, such as the master drone, can replace the functions of the server.

[0043] By employing the embodiments of this disclosure, feature values ​​of the grid in the search area can be shared by multiple agents, thereby improving the efficiency and accuracy of multi-agent collaborative search.

[0044] Figure 2 This is a flowchart illustrating a target search method according to another embodiment of the present disclosure. The target search method may include one or more features of the target search method described above. In one possible implementation, the method further includes:

[0045] S201. Divide the search area into multiple grids, where each grid has an initial reference value.

[0046] In this embodiment, if the search area is two-dimensional, the grid can be rectangular, square, or other shapes. If the search area is three-dimensional, the grid can be cube, cuboid, or other shapes. The specific number and size of the grid can be flexibly set according to the actual situation, and the size of the grid within the same search area can be uniform or non-uniform. Grids within the same search area can be distinguished and labeled by number or by grid coordinates. For example, the coordinates of a two-dimensional grid are (x, y), and the coordinates of a three-dimensional grid are (x, y, z).

[0047] For example, taking the sea surface as the search area, a swarm of drones can be used to search for a specific vessel on a sea surface. Figure 3As shown, if the search area includes a square sea surface with sides of 1000m, dividing the sea surface into a 10*10 grid using a uniform division method results in each grid being a 100m*100m square. Figure 4 As shown, if the search area includes a rectangle with a length of 1000m and a width of 500m, dividing the sea surface into a 10*10 grid and using a uniform division method, each grid will be a rectangle with a length of 100m and a width of 50m.

[0048] In this embodiment of the disclosure, after dividing the search area into multiple grids, a reference value can be set for each grid. This reference value can serve as an upper limit for the feature values ​​of the grids used during the search process. This reference value can be determined based on a probability distribution function or a default value.

[0049] In this embodiment of the disclosure, a large and complex search area is divided into multiple smaller and simpler grids. This allows for more accurate prediction of the possible location of the target at the grid level, thereby improving the efficiency and accuracy of the agent in searching for the target within the search area.

[0050] In many scenarios, methods can be used to infer prior information such as the probability distribution of a target's location. However, using traditional methods that assume the target's location probability distribution is unknown in such cases results in all locations within the area having the same search importance, leading to low search efficiency. In this embodiment, prior information can be used to assign values ​​to the grid, facilitating faster target location and thus improving search efficiency.

[0051] In one implementation, if a prior probability exists for the grid, a reference value for the grid is set based on that prior probability.

[0052] In one implementation, if there is no prior probability for the grid, the initial reference value of the grid is set to a default value.

[0053] In this embodiment, the prior probability of a grid may include the target distribution probability obtained through a pre-trained model, the distribution probability predicted through historical data, etc. A reference value for a grid can be understood as the upper limit of its feature values. For example, if the feature value of a grid is a score, the reference value can be understood as the upper limit of that score. In some examples, the feature value of a grid can be set to be equal to its prior probability. Alternatively, the feature value of a grid can be set to other values ​​calculated based on the prior probability.

[0054] In this embodiment of the disclosure, an upper limit for the feature value of each grid in the search area is assigned in advance based on the prior probability. The feature value of the grid can be obtained by making reasonable use of the prior probability of the grid, which is beneficial to improving the efficiency and accuracy of target search.

[0055] In one implementation, the reference value for the grid is a normalized value. For example, the reference value for the grid can be the normalized value of the prior probability of the grid. The prior probability of a grid at time t can be based on the probability distribution function f. t (x,y) is obtained. f t The specific formula for (x, y) is not limited and can be constructed based on the characteristics of the target, such as velocity and position. An example of the normalization formula is as follows:

[0056]

[0057] In this formula, (x,y) can represent the coordinates of the currently calculated grid, and (x',y') can represent the coordinates of the grid with the largest reference value; g t (x, y) can represent the normalized value. It can represent the maximum value of the reference values ​​of all grids in the search region at time t.

[0058] As time changes, the upper limit of the eigenvalues ​​of the mesh may also change; for example, the upper limit of the eigenvalues ​​of the mesh at time t1. With at time t1 Different, and correspondingly, the calculated and It may also be different. The above method of using the normalized prior probability value as the upper bound of the mesh's eigenvalues ​​is merely an example, not a limitation. Other methods can also be used to calculate the upper bound of the mesh's eigenvalues. For example, the above g can be used... t (x,y) multiplied by 10 or 100, etc., is used as the upper limit of the eigenvalues ​​of the grid.

[0059] In this embodiment, if the grid has no prior probability, the default values ​​for the upper limits of feature values ​​of different grids can be the same or different, depending on the actual situation. For example, assuming that the probability of a target appearing is lower closer to the edge of the search area, the default value can be set to increase gradually from the edge grid to the center grid. Alternatively, if the probability of a target appearing is the same at different locations within the search area, the default value can be kept consistent; for example, the default value can be 1, 5, 10, 100, etc. The default value can remain unchanged over time or change according to a certain time pattern, depending on the needs of the application scenario.

[0060] For example, such as Figure 5As shown, after dividing the search area into multiple grids, the upper limit of the eigenvalues ​​of each grid, such as the upper limit of the score, can be determined based on the coordinates of each grid. If the prior probability of grid (2,3) is 0.6, the prior probability of grid (5,5) is 0.8, and the prior probability of grid (6,3) is 0.4, normalized values ​​can be calculated based on the prior probabilities, and the upper limit of the eigenvalues ​​of these grids can be set as the normalized values. For example, the normalized value of grid (2,3) is 0.6 / 0.8 = 0.75, the normalized value of grid (5,5) is 0.8 / 0.8 = 1, and the normalized value of grid (6,3) is 0.4 / 0.8 = 0.5. The prior probabilities of the remaining grids are unknown, and the upper limit of their eigenvalues ​​can be a default value, such as 1.

[0061] In this embodiment, the value normalized according to prior probability is used to assign an upper limit to the feature value of each grid in the search area. This allows for flexible setting of the grid feature values ​​using prior information, which is beneficial for improving the efficiency and accuracy of target search. Of course, normalization is only an example and not a limitation; other values ​​can also be used as the upper limit of the grid feature values. For example, the upper limit of the feature values ​​of all grids can be a fixed value, but the upper limits of the feature values ​​of all grids in the search area can not be exactly the same.

[0062] Figure 6 This is a flowchart illustrating a target search method according to another embodiment of the present disclosure. The target search method may include one or more features of the target search method described above. In one implementation, S101 updates the feature values ​​of the grids traversed by the agent based on the target search results of the grids traversed by the agent in the search area, including:

[0063] S601. Update the feature values ​​of the grids traversed by the agent based on the search width ratio and / or attenuation parameter of the grids traversed by the agent.

[0064] The search width ratio of the agent is the ratio of the area of ​​the grid covered by the agent's search width to the total area of ​​the grid; the attenuation parameter is obtained based on whether the agent finds the target within the grid covered by the search width.

[0065] In embodiments of this disclosure, the search width of the agent can include the range within which the agent detects signals or information about a target object. For example, for a drone equipped with a camera, its search width is the field of view of its camera; for radar, its search width is the radar's coverage area. The agent's search width may cover the entire grid or only a portion of it. See also... Figure 7If grid A is within the agent's search width, then grid A's search width percentage is 1. If half of grid B is within the agent's search width, then grid B's search width percentage is 0.5. If one-quarter of grid C is within the agent's search width, then grid C's search width percentage is 0.25. Figure 7 This is for illustrative purposes only and not an exact scale.

[0066] In this embodiment, the attenuation parameter can be obtained based on the target search result and / or time interval, but is not limited to the above method and can be set according to the actual situation. For example, if the agent's search result in the current grid indicates that the target does not exist, the feature value of the grid can be reduced by the attenuation parameter. If the agent's search result in the current grid indicates that the target exists, the feature value of the grid can be increased by the attenuation parameter.

[0067] In this embodiment of the disclosure, the feature values ​​of the grid can be updated in real time according to the search width ratio and / or attenuation coefficient, which can provide data support for the subsequent accurate prediction of the agent's motion information, thereby facilitating the faster and more accurate search for the target.

[0068] In one implementation, the target occurrence probability of the grid traversed by the agent is updated based on the search width ratio and / or attenuation parameter of the grid traversed by the agent, including at least one of the following:

[0069] Method 1: Obtain the feature value of the grid at the first time step based on the first search width ratio of the grid and the first reference value of the grid. Here, the first search width ratio of the grid is the search width ratio of the agent in the grid at the first time step, and the first reference value of the grid is the reference value of the grid at that first time step.

[0070] In this embodiment of the disclosure, the first moment can be the initial moment when the agent begins to search for the target, or it can be a moment after the agent has been searching for the target for a period of time.

[0071] In this embodiment of the disclosure, the calculation formula for the candidate feature values ​​of the mesh can be preset, as shown in the following example:

[0072]

[0073] For example, a swarm of drones searches for a ship in a search area, such as a sea surface. A drone begins searching for a target based on initial motion information. At the first moment of starting the search, the first reference value of a grid on the sea surface is 0.8, and the first search width α is 0.5. The characteristic value of the grid at the first moment can be calculated as 0.5 × 0.8 = 0.4.

[0074] Method 2: Based on at least one of the following: the first search width proportion of the grid, the second search width proportion of the grid, the attenuation coefficient, the first reference value of the grid, and the second reference value of the grid, candidate feature values ​​for the grid are obtained. These candidate feature values ​​are then compared with the second reference value of the grid to obtain the feature value of the grid at the second time step. Here, the first search width proportion of the grid is the search width proportion of the agent within the grid at the first time step, and the first reference value of the grid is the reference value of the grid at the first time step. The second search width proportion of the grid is the search width proportion of the agent within the grid at the second time step, and the second reference value of the grid is the reference value of the grid at the second time step. The second time step is a time step following the first time step.

[0075] In this embodiment of the disclosure, the second moment can be the moment after the first moment, or it can be a moment after a certain period of time has elapsed since the first moment. The specific time can be set according to the needs of the actual application scenario.

[0076] In this embodiment of the disclosure, the calculation formula for the candidate feature values ​​of the mesh can be preset, as shown in the following example:

[0077]

[0078] Where α is the proportion of the first search width of the grid, β is the proportion of the second search width of the grid, and ρ is the attenuation coefficient. The first reference value of the grid at time t1. This is the second reference value for the grid at the second time t2.

[0079] For example, suppose a grid has a first reference value of 0.4 at the first time step and a second reference value of 0.6 at the second time step. Based on a UAV, the search width ratio α of this grid at the first time step is 0.5, and the search width ratio β at the first time step is 0.8.

[0080] If the drone search result indicates that the target does not exist and the attenuation coefficient is 0.2, then referring to the above formula (3), we can calculate 0.5*0.4*0.2+0.8*0.6 to obtain the candidate feature value of the grid as 0.52. Since the candidate feature value of the grid is less than the second reference value of the grid at the second time step of 0.6, the feature value of the grid can be updated to the candidate feature value.

[0081] If the drone search result indicates the existence of the target and the attenuation coefficient is 2, then referring to the above formula (3), we can calculate 0.5*0.4*2+0.8*0.6 to obtain the candidate feature value of the grid as 0.88. Since the candidate feature value of the grid is greater than the second reference value of the grid at the second time step of 0.6, the feature value of the grid can be updated to the second reference value of the grid.

[0082] In this embodiment of the disclosure, the feature value of the grid can be reasonably updated based on the agent's search width ratio and attenuation coefficient. For example, if the agent's search result indicates that there is no target in the current grid, the feature value of the current grid can be reduced using the attenuation coefficient; if there is a target in the current grid, the feature value of the current grid can be increased using the attenuation coefficient.

[0083] In one implementation, the method further includes: determining, under a first condition, that the agent needs to change its current motion information; wherein the first condition includes at least one of the following:

[0084] The agent's search width exceeds the search area;

[0085] The agent entered a sensitive area;

[0086] The agent received a collision warning;

[0087] The agent can travel in a straight line for a maximum of 100 seconds.

[0088] In this embodiment of the disclosure, if it is necessary to change the current motion information of the intelligent agent, the motion distance, motion speed, motion direction, motion path, turning angle, motion duration, or turning interval of the intelligent agent can be changed. Taking changing the motion direction of the intelligent agent, such as turning, as an example, when the intelligent agent meets the first condition, the turning direction of the intelligent agent can be predicted and the intelligent agent can be controlled to turn.

[0089] For example, if the drone's search width exceeds the current target sea surface area, it is determined to meet the first condition. Similarly, if the drone's search width does not exceed the current target sea surface area, but the drone's straight-line travel time reaches the upper limit, such as 1 minute, it can be determined to meet the first condition. Furthermore, if a portion of the drone's search width is within a sensitive area, it is determined to meet the first condition. Finally, if the drone detects an impending collision with reefs, mountains, walls, ships, or other drones, it can be determined to meet the first condition.

[0090] In this embodiment of the disclosure, the server can detect whether the agent meets the first condition, the agent can detect whether it meets the first condition, or the agent can report to the server that it meets the first condition. For example, after receiving a collision warning from the server or another agent S2, agent S1 can send its own collision warning to the server, or agent S1 can determine that it meets the first condition and notify the server.

[0091] In this embodiment of the disclosure, situations such as the intelligent agent leaving the search area, entering a sensitive area, colliding with other objects, or failing to turn for a long time can be avoided, which not only improves search efficiency but also enhances the security of the overall search process.

[0092] In one implementation, S102 predicts the motion information required by the agent to search for the target in the search area based on the feature values ​​of the grid in the search area, including one of the following:

[0093] Method 1: When the agent needs to change its current motion information, obtain the feature values ​​of multiple paths based on a set distance and the agent's current grid. The feature value of each path includes the sum of the feature values ​​of multiple grids covered by the set distance from the agent's current grid. Based on these multiple path feature values, select the agent's motion path from among them.

[0094] In the disclosed embodiments, if the feature value of each grid is understood as the score of that grid, the feature value of a path can also be understood as the reward of that path, and the reward of a path can be equal to the sum of the feature values, such as the scores, of all grids on that path. If the feature values ​​of multiple paths are equal, a path can be randomly selected as the movement path of the agent.

[0095] For example, if a drone is detected needing to turn, multiple grids extending outwards from grid A can be obtained based on a set distance L (e.g., 10m, 100m, 1000m) and the drone's current location on grid A. These grids may form multiple paths L1, L2, and L3, each path comprising grids with potentially different sums of eigenvalues. See also... Figure 8 Path L1 includes grids B1, B2, and B3, with a total eigenvalue of 1.5. Path L2 includes grids C1, C2, and C3, with a total eigenvalue of 2.5. Path L3 includes grids D1, D2, and D3, with a total eigenvalue of 2. Therefore, path L2, which has the largest total eigenvalue, can be selected as the path for the agent's subsequent operation.

[0096] In this embodiment of the disclosure, if the distance from the UAV to the boundary of the search area on a certain path is less than a given distance, or the travel time to the boundary of the search area is less than a given travel time, the path may be disregarded when calculating the path. Furthermore, besides comparing the sum of the path's eigenvalues, the path can also be selected by calculating the mean, variance, or other methods of the grid's eigenvalues; this embodiment of the disclosure does not impose any limitations on this approach.

[0097] Method 2: When the agent needs to change its current motion information, at least one grid is selected from the multiple candidate grids adjacent to the agent's current grid to predict the agent's motion direction based on the feature values ​​of the candidate grids.

[0098] For example, the drone is currently in grid A0, and the adjacent grids surrounding grid A0 include A1 to A8, see [link to relevant documentation]. Figure 9 If the eigenvalue of grid A3 is the largest, then grid A3 can be used as the grid to be rotated.

[0099] In this embodiment of the disclosure, the direction and / or path of subsequent movement of the agent can be reasonably planned based on the feature values ​​of the grid, thereby controlling the agent to perform target search more accurately, reducing repeated searches, thereby reducing search time and improving search efficiency.

[0100] This disclosure provides a target search method, which may include a search area map gridding strategy, a grid reward cap allocation strategy, a grid score update strategy, a turning direction reward strategy, and a turning strategy. For example... Figure 10 As shown, the target search method may specifically include:

[0101] S1001, Gridding Strategy: The search area is divided into rectangular grids, and the length and width of the grids can be freely set (for example, they can be set to 50% or less of the drone's search radius).

[0102] S1002. Grid Score Upper Limit Allocation Strategy: The grid score upper limit differs depending on whether the target information is known or unknown. For example, in target search without target information, since the probability of the target appearing at all locations is the same, the upper limit of the score for each grid is 1. Conversely, in target search with target information, after knowing the prior information—the probability distribution of the target appearing at various locations—we can assume that the probability distribution function at grid coordinates (x, y) at time t is f. t (x,y). After normalization, the upper bound g of the score for the grid (x,y) is... t (x,y) is In this formula, g t (x,y) can represent the upper bound of the normalized score at time t. It can represent the maximum value of the probability distribution of all grids in the search region at time t.

[0103] S1003, Grid Score Update Strategy: Each grid has a score. For example, the score ranges from 0 to 1. Initially, each grid has a score of 0, indicating that the UAV's frame width has not covered the current grid. For any grid coordinate (x, y), when the UAV moves, at time t1, the area of ​​the grid covered by the UAV's search frame width is α (between 0 and 1), then the score of that grid is updated as follows: Furthermore, the grid's fraction decays exponentially by ρ over time. When the UAV again covers the area β of the grid at time t2, the grid's fraction is updated to min. The scores of each grid can be shared with all or some drones in the search area, and the goal of collaboration can be achieved automatically through grid score updates and rewards.

[0104] S1004. Turning Strategy: Turning conditions may include exceeding the search area, entering a sensitive area, drone collision warning, and drone straight-ahead time interval T. When any one or more of these conditions are met, the drone can be controlled to perform a turning operation. Within the allowable turning angle range, the drone will select the direction with the highest benefit through a turning direction benefit strategy. If the turning conditions are not met, the drone can continue to fly straight.

[0105] S1005, Turning Direction Gain Strategy: Set a forward distance L, such as 1000 meters or 10000 meters, to prevent the drone from making frequent turns. Calculate the total score gain of all grids along the path covered by the swath width when the drone travels a distance L, and use this as the gain for the current direction. The drone can be controlled to tend to move in the direction with the greater gain.

[0106] The target search method disclosed in this embodiment is a heuristic target search method based on multi-agent cooperation. It achieves automatic multi-agent cooperation through gridded score updates and reward calculations. The grid score upper limit allocation strategy can be utilized to improve search efficiency when prior target information is available.

[0107] Figure 11 This is a schematic diagram of a target search device according to an embodiment of the present disclosure, the device comprising:

[0108] The update module 1101 is used to update the feature value of the grid that the agent has passed through in the search area based on the target search result of the grid in the search area, wherein the feature value of the grid is related to the probability of the target appearing in the grid.

[0109] The prediction module 1102 is used to predict the motion information required by the agent to search for the target in the search area based on the feature values ​​of the grid in the search area.

[0110] Figure 12 This is a schematic diagram of the structure of a target search device according to another embodiment of the present disclosure. The target search method may include one or more features of the target search method described above. In one embodiment, the device further includes: a division module 1201, used to divide the search area into multiple grids, wherein each grid has a reference value, which is used to determine the feature value of the grid.

[0111] In one implementation, such as Figure 12 As shown, the device also includes a reference value setting module 1202, used for at least one of the following:

[0112] If a prior probability exists for the grid, set the reference value for the grid according to that prior probability;

[0113] If there is no prior probability for this grid, the reference value for this grid is set to the default value.

[0114] In one implementation, the reference value for the grid is a normalized value.

[0115] In one implementation, the update module 1101 is further configured to update the feature values ​​of the grid traversed by the agent based on the search width ratio and / or attenuation parameters of the grid traversed by the agent.

[0116] The search width ratio of the agent is the ratio of the area of ​​the grid covered by the agent's search width to the total area of ​​the grid; the attenuation parameter is obtained based on whether the agent finds the target within the grid covered by the search width.

[0117] In one implementation, updating the target occurrence probability of the grid traversed by the agent based on the search width ratio and / or attenuation parameter of the grid traversed by the agent is further used for at least one of the following:

[0118] Based on the first search width ratio of the grid and the first reference value of the grid, the feature value of the grid at the first time is obtained; wherein, the first search width ratio of the grid is the search width ratio of the agent in the grid at the first time, and the first reference value of the grid is the reference value of the grid at the first time.

[0119] Candidate feature values ​​of the grid are obtained based on at least one of the first search width proportion of the grid, the second search width proportion of the grid, the attenuation coefficient, the first reference value of the grid, and the second reference value of the grid. The candidate feature values ​​are compared with the second reference value of the grid to obtain the feature value of the grid at the second time. The second search width proportion of the grid is the search width proportion of the agent in the grid at the second time, the second reference value of the grid is the reference value of the grid at the second time, and the second time is the time after the first time.

[0120] In one implementation, such as Figure 12 As shown, the device further includes a determination module 1203, configured to determine, under a first condition, that the intelligent agent needs to change its current motion information. The first condition includes at least one of the following:

[0121] The agent's search width exceeds the search area;

[0122] The agent entered a sensitive area;

[0123] The agent received a collision warning;

[0124] The agent has reached its maximum straight-line travel time.

[0125] In one implementation, such as Figure 12 As shown, the prediction module 1102 is also used for one of the following:

[0126] When the agent needs to change its current motion information, feature values ​​of multiple paths are obtained based on a set distance and the current grid where the agent is located. The feature value of a path includes the sum of feature values ​​of multiple grids covered by the set distance from the current grid where the agent is located. Based on the feature values ​​of the multiple paths, the motion path of the agent is selected from the multiple paths.

[0127] When the agent needs to change its current motion information, at least one grid is selected from multiple candidate grids adjacent to the agent's current grid to predict the agent's motion direction based on the feature values ​​of these candidate grids.

[0128] In one implementation, such as Figure 12 As shown, the device also includes:

[0129] The sharing module 1204 is used to share the feature values ​​of the grid in the search area with the multiple agents when there are multiple agents in the search area.

[0130] The specific functions and examples of each module and submodule of the apparatus in this disclosure can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0131] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0132] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0133] Figure 13 A schematic block diagram of an example electronic device 1300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0134] like Figure 13 As shown, device 1300 includes a computing unit 1301, which can perform various appropriate actions and processes according to a computer program stored in read-only memory (ROM) 1302 or a computer program loaded from storage unit 1308 into random access memory (RAM) 1303. The RAM 1303 may also store various programs and data required for the operation of device 1300. The computing unit 1301, ROM 1302, and RAM 1303 are interconnected via bus 1304. Input / output (I / O) interface 1305 is also connected to bus 1304.

[0135] Multiple components in device 1300 are connected to I / O interface 1305, including: input unit 1306, such as keyboard, mouse, etc.; output unit 1307, such as various types of monitors, speakers, etc.; storage unit 1308, such as disk, optical disk, etc.; and communication unit 1309, such as network card, modem, wireless transceiver, etc. Communication unit 1309 allows device 1300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0136] The computing unit 1301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1301 performs the various methods and processes described above, such as the target search method. For example, in some embodiments, the target search method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1308. In some embodiments, part or all of the computer program may be loaded and / or installed on device 1300 via ROM 1302 and / or communication unit 1309. When the computer program is loaded into RAM 1303 and executed by the computing unit 1301, one or more steps of the target search method described above may be performed. Alternatively, in other embodiments, the computing unit 1301 may be configured to perform the target search method by any other suitable means (e.g., by means of firmware).

[0137] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0138] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0139] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0140] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0141] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0142] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0143] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0144] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A target search method, comprising: Based on the target search results of the grids traversed by the agent in the search area, update the feature values ​​of the grids traversed by the agent, wherein the feature values ​​of the grids are related to the probability of the target appearing in the grid; Based on the feature values ​​of the grid in the search area, predict the motion information required for the agent to search for the target in the search area; Specifically, updating the feature values ​​of the grids traversed by the agent based on the target search results of the grids traversed in the search area includes: Update the feature values ​​of the grids traversed by the agent based on the search width ratio and / or attenuation parameters of the grids traversed by the agent. Wherein, the search width ratio of the agent is the ratio of the area of ​​the grid covered by the agent's search width to the total area of ​​the grid; The attenuation parameter is obtained based on whether the agent finds the target within the grid covered by the search width.

2. The method according to claim 1, wherein, Also includes: The search area is divided into multiple grids, each grid having a reference value used to determine the feature value of the grid.

3. The method according to claim 2, wherein, It also includes at least one of the following: If the grid has a prior probability, set the reference value of the grid according to the prior probability; If there is no prior probability for the grid, the reference value of the grid is set to the default value.

4. The method according to claim 3, wherein, The reference value for the grid is a normalized value.

5. The method according to claim 1, wherein, The target occurrence probability of the grid traversed by the agent is updated based on the search width ratio and / or attenuation parameter of the grid traversed by the agent, including at least one of the following: The feature value of the grid at the first time moment is obtained based on the first search width ratio of the grid and the first reference value of the grid; wherein, the first search width ratio of the grid is the search width ratio of the agent in the grid at the first time moment, and the first reference value of the grid is the reference value of the grid at the first time moment; Candidate feature values ​​of the grid are obtained based on at least one of the first search width proportion of the grid, the second search width proportion of the grid, the attenuation coefficient, the first reference value of the grid, and the second reference value of the grid; the candidate feature values ​​are compared with the second reference value of the grid to obtain the feature values ​​of the grid at the second time; wherein, the second search width proportion of the grid is the search width proportion of the agent in the grid at the second time, the second reference value of the grid is the reference value of the grid at the second time, and the second time is a time after the first time.

6. The method according to any one of claims 1 to 5, wherein, Also includes: Under a first condition, it is determined that the agent needs to change its current motion information; wherein the first condition includes at least one of the following: The search width of the intelligent agent exceeds the search area; The intelligent agent enters the sensitive area; The intelligent agent received a collision warning; The agent's straight-line travel time has reached its maximum limit.

7. The method according to any one of claims 1 to 5, wherein, Based on the feature values ​​of the grid in the search area, predict the motion information required by the agent to search for the target in the search area, including one of the following: When the agent needs to change its current motion information, feature values ​​of multiple paths are obtained based on a set distance and the current grid where the agent is located. The feature value of a path includes the sum of feature values ​​of multiple grids covered by the set distance from the current grid where the agent is located. Based on the feature values ​​of the multiple paths, the motion path of the agent is selected from the multiple paths. When the agent needs to change its current motion information, at least one grid is selected from multiple candidate grids adjacent to the agent's current grid to predict the agent's motion direction based on the feature values ​​of the candidate grids.

8. The method according to any one of claims 1 to 5, further comprising: When multiple agents exist in the search area, the feature values ​​of the grid in the search area are shared among the multiple agents.

9. A target search device, comprising: An update module is used to update the feature values ​​of the grids traversed by the agent based on the target search results of the grids traversed by the agent in the search area, wherein the feature values ​​of the grids are related to the probability of the target appearing in the grid; The prediction module is used to predict the motion information required by the agent to search for the target in the search area based on the feature values ​​of the grid in the search area. The update module is also used for: Update the feature values ​​of the grids traversed by the agent based on the search width ratio and / or attenuation parameters of the grids traversed by the agent. Wherein, the search width ratio of the agent is the ratio of the area of ​​the grid covered by the agent's search width to the total area of ​​the grid; The attenuation parameter is obtained based on whether the agent finds the target within the grid covered by the search width.

10. The apparatus according to claim 9, wherein, Also includes: A partitioning module is used to divide the search area into multiple grids, wherein each grid has a reference value, and the reference value is used to determine the feature value of the grid.

11. The apparatus according to claim 10, wherein, It also includes a reference value setting module for at least one of the following: If the grid has a prior probability, set the reference value of the grid according to the prior probability; If there is no prior probability for the grid, the reference value of the grid is set to the default value.

12. The apparatus according to claim 11, wherein, The reference value for the grid is a normalized value.

13. The apparatus according to claim 9, wherein, The target occurrence probability of the grid traversed by the agent is updated based on the search width ratio and / or attenuation parameter of the grid traversed by the agent, including at least one of the following: The feature value of the grid at the first time moment is obtained based on the first search width ratio of the grid and the first reference value of the grid; wherein, the first search width ratio of the grid is the search width ratio of the agent in the grid at the first time moment, and the first reference value of the grid is the reference value of the grid at the first time moment; Candidate feature values ​​of the grid are obtained based on at least one of the first search width proportion of the grid, the second search width proportion of the grid, the attenuation coefficient, the first reference value of the grid, and the second reference value of the grid; the candidate feature values ​​are compared with the second reference value of the grid to obtain the feature values ​​of the grid at the second time; wherein, the second search width proportion of the grid is the search width proportion of the agent in the grid at the second time, the second reference value of the grid is the reference value of the grid at the second time, and the second time is a time after the first time.

14. The apparatus according to any one of claims 9 to 13, wherein, Also includes: The determination module is configured to determine, under a first condition, that the agent needs to change its current motion information; wherein the first condition includes at least one of the following: The search width of the intelligent agent exceeds the search area; The intelligent agent enters the sensitive area; The intelligent agent received a collision warning; The agent's straight-line travel time has reached its maximum limit.

15. The apparatus according to any one of claims 9 to 13, wherein, The prediction module is also used for one of the following: When the agent needs to change its current motion information, feature values ​​of multiple paths are obtained based on a set distance and the current grid where the agent is located. The feature value of a path includes the sum of feature values ​​of multiple grids covered by the set distance from the current grid where the agent is located. Based on the feature values ​​of the multiple paths, the motion path of the agent is selected from the multiple paths. When the agent needs to change its current motion information, at least one grid is selected from multiple candidate grids adjacent to the agent's current grid to predict the agent's motion direction based on the feature values ​​of the candidate grids.

16. The apparatus according to any one of claims 9 to 13, further comprising: A sharing module is used to share feature values ​​of a grid in the search area for multiple agents when multiple agents exist in the search area.

17. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-8.

19. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-8.

20. An intelligent agent comprising the electronic device of claim 17 or the readable storage medium of claim 18, or a computer program product capable of running the computer program of claim 19.

Images