Multi-unmanned aerial vehicle target coverage path planning method based on cluster pooling graph neural network

Abstract

The invention discloses a multi-unmanned aerial vehicle target coverage path planning method based on a cluster pooling graph neural network, and belongs to the field of path planning. The method comprises the following steps: obtaining an explored sub-graph of each unmanned aerial vehicle at the current moment based on position information of each unmanned aerial vehicle and explored area information of each unmanned aerial vehicle at the current moment; under the condition that the total coverage area of the explored sub-graphs corresponding to all the unmanned aerial vehicles is smaller than the target area, inputting the explored sub-graphs corresponding to the at least one unmanned aerial vehicle into a graph neural network model to obtain a target motion path corresponding to the at least one unmanned aerial vehicle; controlling the at least one unmanned aerial vehicle to move based on the target motion path and acquiring position information and explored area information after the at least one unmanned aerial vehicle moves so as to update an explored sub-graph of the at least one unmanned aerial vehicle; and controlling each unmanned aerial vehicle to stop moving under the condition that the total coverage area of the updated explored sub-graphs corresponding to all unmanned aerial vehicles coincides with the target area. The coverage task completion efficiency is high.

Description

Technical Field

[0001] The present application belongs to the field of path planning technology, and in particular to a multi-UAV target coverage path planning method based on cluster pooling graph neural network. Background Art

[0002] In the inspection of volatile organic compounds (VOCs) in industrial parks, it is necessary to inspect the gas leakage of pipeline pollution sources and extract gas samples from suspected leaks to bring them back to the laboratory for chemical analysis. With the continuous improvement of drone technology, the use of multiple drones to conduct VOCs inspections has become an emerging technical means. In the related technology, there is a method based on manually controlled drone flight for inspection. This method requires a lot of time and manpower costs, and is prone to missed inspections or repeated inspections, and the inspection efficiency is low. Summary of the invention

[0003] This application aims to solve at least one of the technical problems existing in the related technology. To this end, this application proposes a multi-UAV target coverage path planning method based on cluster pooling graph neural network, which can automatically collect environmental information for path planning without consuming a lot of manpower costs, and promotes coordination between multiple UAVs, reduces missed inspections or repeated inspections, and improves the efficiency of completing coverage tasks.

[0004] In a first aspect, the present application provides a multi-UAV target coverage path planning method based on a cluster pooling graph neural network, the method comprising:

[0005] Based on the position information of each of the drones at the current moment and the explored area information of each of the drones in the target area, acquiring the explored subgraph of each of the drones at the current moment;

[0006] When the total coverage area of ​​the explored subgraphs corresponding to all the drones is smaller than the target area, the explored subgraphs corresponding to at least one of the drones among all the drones are respectively input into the graph neural network model to obtain the target motion path corresponding to at least one of the drones output by the graph neural network model;

[0007] Controlling at least one of the drones to move based on the target motion path, and obtaining position information and explored area information of at least one of the drones after the movement, so as to update the explored subgraph of at least one of the drones;

[0008] When the total coverage area of ​​the updated explored subgraphs corresponding to all the drones coincides with the target area, each of the drones is controlled to stop moving.

[0009] According to the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application, local path planning is performed on the UAV through environmental information such as the explored area information and location information corresponding to the UAV, and a new local path planning is generated based on the environmental information after the UAV movement, and it is iterated until the inspection task is finally completed. The environmental information is effectively utilized in the inspection process, and the environmental information can be automatically collected for path planning without consuming a lot of manpower costs. It also promotes coordination between multiple UAVs, reduces missed inspections or repeated inspections, and improves the efficiency of completing coverage tasks.

[0010] In one embodiment of the present application, a multi-UAV target coverage path planning method based on a cluster pooling graph neural network is provided, wherein the graph neural network model is trained based on the following steps:

[0011] Obtain multiple sample explored subgraphs corresponding to the sample drone in one collection cycle;

[0012] Based on the map information of the complete grid map, determining the sample motion paths corresponding to the sample explored sub-maps; at least part of the sample motion paths constitute a coherent motion path;

[0013] The graph neural network model is trained using the multiple sample explored subgraphs as samples and the sample motion paths corresponding to the sample explored subgraphs as sample labels.

[0014] In one embodiment of the present application, a multi-UAV target coverage path planning method based on a cluster pooling graph neural network is provided, wherein one acquisition cycle includes multiple sample first moments, and the multiple sample explored subgraphs corresponding to the sample UAV in one acquisition cycle are obtained, including:

[0015] Based on each grid point in the complete grid map, obtaining an incomplete obstacle grid set from the complete grid map;

[0016] Based on the sample position information of the sample UAV in the complete grid map at the first moment of the sample and the distance between each of the incomplete obstacle grid points in the incomplete obstacle grid set, the incomplete obstacle grid set is processed to obtain the sample explored subgraph corresponding to the sample UAV at the first moment of the sample.

[0017] In one embodiment of the present application, a multi-UAV target coverage path planning method based on a cluster pooling graph neural network is provided, wherein the map information based on the complete grid map is used to determine the sample motion path corresponding to each sample explored sub-graph, including:

[0018] Based on the area ratio of the interest region corresponding to each grid point in the complete grid map, obtaining the interest grid set;

[0019] Acquire a first grid set based on the interest grid set, the incomplete obstacle grid set, and the perception range parameter of the sample UAV;

[0020] Constructing a global key point grid set based on the viewpoint grid set corresponding to the first grid set and the interest grid set;

[0021] Based on the node set in the sample explored subgraph corresponding to the global key point grid set, path planning is performed on the sample UAV to obtain a sample motion path corresponding to the sample explored subgraph.

[0022] A multi-UAV target coverage path planning method based on a cluster pooling graph neural network in one embodiment of the present application, wherein the plurality of sample explored subgraphs are used as samples, and the sample motion paths corresponding to each of the sample explored subgraphs are used as sample labels, and the graph neural network model is trained, including:

[0023] Input the multiple sample explored subgraphs into the graph neural network model, and obtain the predicted motion path corresponding to each of the sample explored subgraphs output by the graph neural network model;

[0024] constructing a target loss function based on the predicted motion path and the sample motion path corresponding to the sample explored sub-graph;

[0025] The graph neural network model is trained based on the target loss function.

[0026] A multi-UAV target coverage path planning method based on a cluster pooling graph neural network according to an embodiment of the present application, wherein the graph neural network model includes a dead zone cluster pooling module, a graph aggregation module, and a strategy module connected in sequence, and the explored subgraphs corresponding to at least one of the UAVs among all the UAVs are respectively input into the graph neural network model, and a target motion path corresponding to at least one UAV output by the graph neural network model is obtained, including:

[0027] Performing a cluster pooling operation on the explored subgraph based on the dead zone cluster pooling module, obtaining a first subgraph corresponding to the explored subgraph output by the dead zone cluster pooling module;

[0028] Processing the first subgraph based on the graph aggregation module to obtain a plurality of second subgraphs output by the graph aggregation module;

[0029] The plurality of second sub-images are cascaded based on the strategy module to obtain the target motion path output by the strategy module.

[0030] In one embodiment of the present application, a multi-UAV target coverage path planning method based on a cluster pooling graph neural network is provided, wherein the cluster pooling operation is performed on the explored subgraph based on the dead zone cluster pooling module, and a first subgraph corresponding to the explored subgraph output by the dead zone cluster pooling module is obtained, including:

[0031] When there is no other drone, interest area and boundary area in at least one of the drones except the target drone in the receptive field corresponding to the target drone in at least one of the drones, a target unit is obtained based on a grid of adjacent targets in the target area; the boundary area is a grid adjacent to an unexplored area and an explored area;

[0032] Based on the connected nodes in the target unit, acquiring at least one cluster node corresponding to the target unit;

[0033] Processing the connected cluster nodes in the adjacent target units to obtain edge features between the connected cluster nodes;

[0034] Based on the at least one cluster node and each of the edge features, the first subgraph is obtained.

[0035] In one embodiment of the present application, a multi-UAV target coverage path planning method based on a cluster pooling graph neural network is provided, wherein the graph aggregation module includes a graph encoding module, a plurality of sequentially connected graph aggregation operation modules and a plurality of graph decoding modules, wherein the output end of the graph encoding module is respectively connected to the graph aggregation operation module and the graph decoding module, and the output end of each graph aggregation module is connected to the graph decoding module; the first subgraph is processed based on the graph aggregation module to obtain a plurality of second subgraphs output by the graph aggregation module, including:

[0036] Based on the graph encoding module, nonlinearly transform the cluster nodes and edge features corresponding to the first subgraph to obtain a third subgraph;

[0037] Perform multiple aggregation operations on the first subgraph based on the multiple graph aggregation operation modules to obtain multiple fourth subgraphs;

[0038] The third sub-image and the plurality of fourth sub-images are decoded based on the plurality of image decoding modules to obtain the plurality of second sub-images.

[0039] In an embodiment of the present application, a multi-UAV target coverage path planning method based on a cluster pooling graph neural network, the strategy module includes an activation function layer; the strategy module is used to perform cascade processing on the multiple second sub-graphs to obtain the target motion path output by the strategy module, including:

[0040] Performing cascade processing on the multiple second sub-images to obtain a fifth sub-image;

[0041] Based on the activation function layer, probability values ​​corresponding to the candidate motion paths in the fifth sub-image are obtained, and the candidate motion path corresponding to the largest probability value among multiple probability values ​​is determined as the target motion path.

[0042] In a second aspect, the present application provides a multi-UAV target coverage path planning device based on a cluster pooling graph neural network, comprising:

[0043] A first processing module, configured to obtain an explored subgraph of each drone at the current moment based on the position information of each drone at the current moment and the explored area information of each drone in the target area;

[0044] A second processing module is used to input the explored subgraphs corresponding to at least one of the drones into the graph neural network model to obtain the target motion path corresponding to at least one of the drones output by the graph neural network model when the total coverage area of ​​the explored subgraphs corresponding to all the drones is smaller than the target area;

[0045] a third processing module, configured to control at least one of the drones to move based on the target motion path, and obtain position information and explored area information of at least one of the drones after the movement, so as to update an explored subgraph of at least one of the drones;

[0046] The fourth processing module is used to control each of the drones to stop moving when the total coverage area of ​​the updated explored sub-graphs corresponding to all the drones coincides with the target area.

[0047] According to the multi-UAV target coverage path planning device based on cluster pooling graph neural network provided in the embodiment of the present application, local path planning is performed on the UAV through environmental information such as the explored area information and location information corresponding to the UAV, and a new local path planning is generated based on the environmental information after the movement of the UAV, and it is iterated until the inspection task is finally completed. The environmental information is effectively utilized in the inspection process, and the environmental information can be automatically collected for path planning without consuming a lot of manpower costs. It also promotes coordination between multiple UAVs, reduces missed inspections or repeated inspections, and improves the efficiency of completing coverage tasks.

[0048] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the multi-UAV target coverage path planning method based on cluster pooling graph neural network as described in the first aspect above is implemented.

[0049] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-UAV target coverage path planning method based on cluster pooling graph neural network as described in the first aspect above.

[0050] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the multi-UAV target coverage path planning method based on cluster pooling graph neural network as described in the first aspect above.

[0051] The above one or more technical solutions in the embodiments of the present application have at least one of the following technical effects:

[0052] The local path planning of the UAV is carried out through the environmental information such as the explored area information and the location information corresponding to the UAV, and a new local path planning is generated based on the environmental information after the UAV movement, and it is iterated until the inspection task is finally completed. The environmental information is effectively utilized in the inspection process, and the environmental information can be automatically collected for path planning without consuming a lot of manpower costs. It also promotes coordination between multiple UAVs, reduces missed inspections or repeated inspections, and improves the efficiency of completing coverage tasks.

[0053] Furthermore, by processing the explored subgraphs based on the cluster pooling method, environmental information can be used more effectively when generating the planned path, which promotes the coordination between multiple drones and improves the efficiency of completing the inspection tasks. By constructing a graph neural network model, drones can also autonomously plan paths in complex environments, reducing labor costs and improving the accuracy and precision of the generated motion paths.

[0054] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0056] Figure 1 This is one of the flow diagrams of the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application;

[0057] Figure 2 It is a schematic diagram of the coverage path planning of a drone according to the multi-drone target coverage path planning method based on the cluster pooling graph neural network provided in an embodiment of the present application;

[0058] Figure 3It is a structural schematic diagram of a graph neural network model of a multi-UAV target coverage path planning method based on a cluster pooling graph neural network provided in an embodiment of the present application;

[0059] Figure 4 It is a schematic diagram of the principle of cluster pooling operation of the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in an embodiment of the present application;

[0060] Figure 5 It is a schematic diagram of the generation principle of the target motion path of the multi-UAV target coverage path planning method based on the cluster pooling graph neural network provided in an embodiment of the present application;

[0061] Figure 6 This is the second flow chart of the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application;

[0062] Figure 7 It is a structural schematic diagram of a multi-UAV target coverage path planning device based on a cluster pooling graph neural network provided in an embodiment of the present application;

[0063] Figure 8 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0065] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0066] In combination with the accompanying drawings, the multi-UAV target coverage path planning method based on cluster pooling graph neural network, the multi-UAV target coverage path planning device based on cluster pooling graph neural network, the electronic device and the readable storage medium provided in the embodiments of the present application are described in detail through specific embodiments and their application scenarios.

[0067] Among them, the multi-UAV target coverage path planning method based on cluster pooling graph neural network can be applied to the terminal, and can be specifically executed by hardware or software in the terminal.

[0068] The terminal includes, but is not limited to, a portable communication device such as a mobile phone or tablet computer with a touch-sensitive surface (e.g., a touch screen display and / or a touch pad). It should also be understood that in some embodiments, the terminal may not be a portable communication device, but a desktop computer with a touch-sensitive surface (e.g., a touch screen display and / or a touch pad).

[0069] In the following various embodiments, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, a mouse and a joystick.

[0070] The embodiment of the present application provides a multi-UAV target coverage path planning method based on a cluster pooling graph neural network. The execution subject of the multi-UAV target coverage path planning method based on a cluster pooling graph neural network can be an electronic device or a functional module or functional entity in the electronic device that can implement the multi-UAV target coverage path planning method based on a cluster pooling graph neural network. The electronic devices mentioned in the embodiment of the present application include but are not limited to mobile phones, tablet computers, computers, cameras, wearable devices, etc. The multi-UAV target coverage path planning method based on a cluster pooling graph neural network provided in the embodiment of the present application is explained below using an electronic device as an example of the execution subject.

[0071] like Figure 1 As shown, the multi-UAV target coverage path planning method based on cluster pooling graph neural network includes: step 110, step 120, step 130 and step 140.

[0072] Step 110: based on the position information of each drone at the current moment and the explored area information of each drone in the target area, obtain the explored subgraph of each drone at the current moment;

[0073] In this step, the current moment is the current collection moment, and the location information of each drone at the current moment can be collected.

[0074] The location information may include coordinate information and the like.

[0075] The target area is the area where multiple drones need to inspect.

[0076] like Figure 2 As shown, a grid map corresponding to the target area can be obtained, and the grid map includes multiple grids. If the drone has visited a grid, the grid can be determined as the explored area of ​​the drone.

[0077] The explored area information may include explored grid points, wherein the explored grid points may include non-obstacle grids in the explored area adjacent to the unexplored area, and the like.

[0078] The explored area is the area that has been detected by the drone's perception camera, and the unexplored area is the area that has not been detected by the drone's perception camera.

[0079] The set of grids that the drone has explored is advancing over time, and the set of grids contained in the drone camera perception area is constantly expanding.

[0080] The area of ​​interest information of the drone can be obtained, wherein the area of ​​interest is an area containing potential pollution sources, and the area of ​​interest information may include the ratio of the area of ​​the pollution source in the interest grid to the area of ​​the interest grid.

[0081] Based on the location information of each drone, the ratio of the number of drones in each grid to the area of ​​the grid can be obtained.

[0082] Based on the explored area information, the location information of each drone and the area of ​​interest information, the explored subgraph corresponding to each drone can be obtained.

[0083] The explored subgraph may include graph structure data of explored grid points and edges, etc. The edges are used to represent the connection relationship between each grid point.

[0084] The explored subgraph can also include features related to grid points and edges, such as gas concentration and obstacle location information.

[0085] Step 120: When the total coverage area of ​​the explored subgraphs corresponding to all the drones is smaller than the target area, the explored subgraphs corresponding to at least one drone among all the drones are respectively input into the graph neural network model to obtain the target motion path corresponding to at least one drone output by the graph neural network model;

[0086] In this step, each drone can communicate with each other, and each drone can obtain the location information and explored subgraphs of other drones.

[0087] The coverage areas of the explored subgraphs of each drone may overlap. The explored subgraphs of each drone can be deduplicated to obtain the total coverage area of ​​the explored subgraphs corresponding to all drones.

[0088] When the total coverage area is smaller than the target area, that is, there are still areas in the target area that have not been explored by the drone, path planning can continue to be performed on at least one drone to control the at least one drone to continue moving.

[0089] The input of the graph neural network model is the explored subgraph of the drone, and the output is the target motion path corresponding to the drone.

[0090] The graph neural network model can be trained by using the sample explored subgraph as the sample and the sample motion path corresponding to the sample explored subgraph as the sample label.

[0091] The target motion path may include the motion steps of the UAV in four directions. For example, the candidate motion steps of the UAV in each direction may be calculated, and each candidate motion step has a corresponding probability value. The candidate motion step corresponding to the largest probability value among multiple probability values ​​may be determined as the target motion path of the UAV.

[0092] For example, candidate motion steps may include: moving forward 1 grid (corresponding probability value is 10%), moving backward 1 grid (corresponding probability value is 50%), moving left 1 grid (corresponding probability value is 30%), and moving right 1 grid (corresponding probability value is 10%), then the target motion path can be determined as moving backward 1 grid.

[0093] Step 130: Control at least one UAV to move based on the target motion path, and obtain position information and explored area information of at least one UAV after the movement, so as to update the explored subgraph of at least one UAV;

[0094] In this step, after the target motion path is generated, the drone can be controlled to move based on its corresponding target motion path.

[0095] After the UAV moves based on the target motion path, the current location information of the UAV and the corresponding explored area information of the UAV can be obtained.

[0096] Then, the explored subgraph of the drone can be updated based on the updated position information and the explored area information.

[0097] Step 140: When the total coverage area of ​​the updated explored sub-graphs corresponding to all the drones coincides with the target area, control each drone to stop moving.

[0098] In this step, when the total coverage area of ​​the explored subgraphs corresponding to all drones coincides with the target area, that is, when there is no unexplored area of ​​the drone in the target area, the drone can be controlled to stop moving.

[0099] like Figure 2 As shown, Figure 2The interest grid is illustrated. During the movement, the drone can inspect the gas leakage of the pipeline pollution source and extract the polluted gas in the interest grid, such as volatile organic compounds (VOCs).

[0100] In this application, if Figure 5 As shown, for a drone, the explored subgraph of the drone at the current acquisition moment can be obtained. When it is determined that there are unexplored areas in the target area, the target motion path with a step length of t can be generated based on the explored subgraph of the drone, and then the drone is controlled to move based on the target motion path, and the explored subgraph of the drone after the movement is updated. When there are still unexplored areas in the target area, the target motion path with a step length of t+1 can be generated based on the updated explored subgraph of the drone, and then the drone is controlled to move based on the target motion path until the drone has visited all areas of interest and the drone's perception camera has detected all non-obstacle grids. It can be determined that the drone has completed the inspection task and the drone is controlled to stop moving.

[0101] According to the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application, local path planning is performed on the UAV through environmental information such as the explored area information and location information corresponding to the UAV, and a new local path planning is generated based on the environmental information after the UAV movement, and it is iterated until the inspection task is finally completed. The environmental information is effectively utilized in the inspection process, and the environmental information can be automatically collected for path planning without consuming a lot of manpower costs. It also promotes coordination between multiple UAVs, reduces missed inspections or repeated inspections, and improves the efficiency of completing coverage tasks.

[0102] In some embodiments, the graph neural network model can be trained based on the following steps:

[0103] Obtain multiple sample explored subgraphs corresponding to the sample drone in one collection cycle;

[0104] Based on the map information of the complete grid map, determine the sample movement path corresponding to the explored sub-map of each sample;

[0105] Taking multiple sample explored subgraphs as samples and the sample motion paths corresponding to each sample explored subgraph as sample labels, the graph neural network model is trained.

[0106] In this embodiment, for a sample drone, multiple sample explored subgraphs corresponding to the sample drone may be collected in one collection cycle.

[0107] For example, a collection cycle may include multiple collection moments, and at each collection moment, the sample explored subgraph corresponding to the sample drone may be collected.

[0108] When the locations of the sample drones are different, the sample explored subgraphs corresponding to the sample drones are different.

[0109] Based on the map information of the complete raster map, multiple sample motion paths can be constructed. The time steps corresponding to the multiple sample motion paths are different. The sample explored subgraph at the target time t corresponds to the sample motion path at the time step t~t+1 corresponding to the target time.

[0110] For example, the sample explored subgraph corresponding to the sample drone collected at time t0 is x1, and the corresponding sample motion path y1 in the step length of t0 to t1 can be determined from multiple sample motion paths; the sample explored subgraph corresponding to the sample drone collected at time t1 is x2, and the corresponding sample motion path y2 in the step length of t1 to t2 can be determined from multiple sample motion paths.

[0111] At least some of the multiple sample motion paths may constitute a continuous motion path. For example, the end point of the sample motion path y1 is the starting point of the sample motion path y2.

[0112] like Figure 6 As shown, in some embodiments, obtaining multiple sample explored subgraphs corresponding to the sample drone in one collection cycle may include:

[0113] Based on each grid point in the complete grid map, an incomplete obstacle grid set is obtained from the complete grid map;

[0114] Based on the sample position information of the sample UAV in the complete grid map at the first moment of the sample and the distance between each incomplete obstacle grid point in the incomplete obstacle grid set, the incomplete obstacle grid set is processed to obtain the sample explored subgraph corresponding to the sample UAV at the first moment of the sample.

[0115] In this embodiment, a simulation environment can be built to create a complete grid map on the inspection flight altitude plane, such as Figure 2 shown.

[0116] The incomplete obstacle grid set may include multiple incomplete obstacle grid points.

[0117] Non-complete obstacles can include partial occlusions or obstacles that the drone can circumvent.

[0118] An acquisition cycle is the duration from the moment when acquisition starts to the moment when acquisition ends. An acquisition cycle may include multiple acquisition moments, that is, an acquisition cycle includes multiple sample first moments. For example, the acquisition moment t to the acquisition moment t+n (where n is a natural number greater than 0) can be set as an acquisition cycle, and an acquisition cycle can include n+1 sample first moments.

[0119] The distance between each incomplete obstacle grid point can be obtained, and then based on the sample position information of the sample drone at the first moment of the sample and the distance between each incomplete obstacle grid point, the sample explored subgraph can be constructed. That is, the node features and edge features corresponding to the sample explored subgraph can be constructed to construct the sample explored subgraph.

[0120] In the actual execution process, a simulation environment can be built to determine the obstacle space and flyable space in the map, and then a complete grid map can be established on the inspection flight altitude plane, such as Figure 2 As shown, all incomplete obstacle grid points are marked as incomplete obstacle grid set C0, and the grid points in C0 are respectively constructed into nodes Where i = 1, 2, ..., m, is the number of drones in the grid / grid area, is the proportion of unexplored area in the grid, is the area ratio of the region of interest (if the drone has visited the grid, the area of ​​the region of interest degenerates to 0), is the proportion of boundary grids (the boundary refers to the non-obstacle grids adjacent to the unexplored area in the explored area at time t).

[0121] When there is a passage for drones to pass between adjacent grid points i and j in C0, the edge features corresponding to grid points i and j can be constructed. Edge Features is the path distance between the center positions of grids i and j.

[0122] Get the grid where the sample drone is located (i.e., the sample location information), and then construct the graph structure data of the sample covering the mission environment at the first moment t:

[0123]

[0124] Then we can use G at time t t The number of nodes S explored by the drone t The sample explored subgraph is represented as G t [S t ].

[0125] In some embodiments, determining the sample motion path corresponding to each sample explored sub-map based on the map information of the complete grid map may include:

[0126] Based on the area ratio of the interest area corresponding to each grid point in the complete grid map, obtain the interest grid set;

[0127] Based on the interest grid set, the incomplete obstacle grid set and the perception range parameters of the sample UAV, a first grid set is obtained;

[0128] Constructing a global key point grid set based on the viewpoint grid set and the interest grid set corresponding to the first grid set;

[0129] Based on the node set in the sample explored subgraph corresponding to the global key point grid set, the sample UAV is path planned to obtain the sample motion path corresponding to the sample explored subgraph.

[0130] In this embodiment, the area ratio of the region of interest is used to characterize the access status of the grid point.

[0131] The perception range parameters of the sample drone can be determined based on the configuration of the sample drone. For example, the perception range parameters can be 9 grids, 16 grids, or 25 grids, etc., which is not limited in this application.

[0132] like Figure 2 As shown, the perception range of the camera of the sample drone can cover grids 1 to 25. When the perception camera of the drone has detected grids 1 to 25, it can be determined that grids 1 to 25 have been explored by the drone.

[0133] During the detection process of the drone, the drone may reach a grid (eg, grid 13) at a central position among multiple grids (grid 1 to grid 25), and grid 13 may be determined as a grid that has been visited by the drone.

[0134] For grid points that have been visited by the sample drone, the area ratio of the region of interest corresponding to the grid point is 0.

[0135] The grid set of interest includes at least one grid point, wherein the grid point is a grid that contains a potential pollution source and has not been visited by the drone.

[0136] The coverage of the grid point includes that the sample drone has visited the grid point or that the sample drone has not visited the grid point. Based on the perception range parameters of the sample drone and the grid set of interest, a grid set that can be completely covered by the perception range of the sample drone can be obtained, and then based on the grid set, the first grid set is obtained from the incomplete obstacle grid set.

[0137] The corresponding viewpoint grid set can be generated at equal intervals in the area corresponding to the first grid set according to the perception range parameter, and then the global key point grid set is constructed based on the sum of the viewpoint grid set and the interest grid set.

[0138] Based on the node set in the sample explored subgraph corresponding to the global key point grid set, the sample UAV can be path planned to obtain the motion path corresponding to the sample explored subgraph.

[0139] In the actual implementation process, Figure 6 As shown, the grid points whose area ratio of the interest area is greater than 0 can be marked as the interest grid set C1. Assuming that there are |C1| drones located at the center of each grid point in C1, and the grids that can be completely covered by the perception range are set C2, the first grid set C3 = C0-C2 can be obtained.

[0140] According to the perception range parameter, the viewpoint grid set C can be generated in the C3 area at equal intervals. A , so that the distribution in C A The perception range of multiple sample drones at the center of each grid point can completely cover the C3 area.

[0141] Construct a global key point grid set C = C1 + C4. The set C corresponds to the node set V in the sample explored subgraph G0. In G0, a path planning algorithm (such as A * Algorithm) Calculate the shortest path between two elements in V, and use the path length as the edge length connecting the nodes in V to construct a key point graph Then you can see The problem of grouping n sample drones through |V| key points is modeled as an operations research VRP (Vehicle Routing Problem). For example, the OR-tools tool can be used to calculate the paths of n drones and map the paths to the graph G0 to obtain the expert solution sample labels. The samples in the expert solution dataset are the following graph G at each time step t. t The explored subgraph G t [S t ], where S t is the set of nodes explored at time t, and the sample labels are the path selection actions of each drone in four directions (i.e., sample motion paths) in, like Figure 5 middle shown.

[0142] According to the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application, path planning is performed for sample UAVs based on global grid map information, and sample labels are generated for training the graph neural network model. This allows the graph neural network model trained based on the sample explored subgraph and the sample motion path to generate a more accurate motion path based on the local explored subgraph information in actual applications, and the path planning has high accuracy and precision.

[0143] like Figure 6 As shown, in some embodiments, taking multiple sample explored subgraphs as samples and taking sample motion paths corresponding to each sample explored subgraph as sample labels, training a graph neural network model may include:

[0144] Input multiple sample explored subgraphs into the graph neural network model, and obtain the predicted motion path corresponding to each sample explored subgraph output by the graph neural network model;

[0145] Construct a target loss function based on the predicted motion path and the sample motion path corresponding to the sample explored sub-graph;

[0146] Train the graph neural network model based on the objective loss function.

[0147] In this embodiment, the objective loss function is used to measure the difference between the predicted results of the model and the actual results.

[0148] The objective loss function may include a cross entropy loss function, a sum of squares loss function, or a logarithmic loss function, etc.

[0149] The back-propagation algorithm can be used to update the weights of the graph neural network model to minimize the target loss function until the performance of the graph neural network model on the validation set meets the requirements.

[0150] In the actual implementation process, Figure 6 As shown, an expert dataset B = {(G t [S t ],A t )} t＝1:T , where G t [S t ] is G at time t t The number of nodes S explored by the drone t The sample explored subgraph A t is the set of drone actions (sample motion paths) in the expert solution.

[0151] Model training can be performed based on imitation learning methods:

[0152]

[0153] Among them, G t [S t ] is G at time t t The number of nodes S explored by the drone t The sample explored subgraph A t is the set of drone actions (sample motion paths) in the expert solution, π is the controller of the graph neural network model, is the cross entropy loss between the predicted motion path obtained by the controller based on the explored subgraph and the sample motion path.

[0154] In some embodiments, step 120 may include:

[0155] Performing cluster pooling operation on the explored subgraph based on the dead zone cluster pooling module, obtaining a first subgraph corresponding to the explored subgraph output by the dead zone cluster pooling module;

[0156] Processing the first subgraph based on the graph aggregation module to obtain multiple second subgraphs output by the graph aggregation module;

[0157] Based on the strategy module, a plurality of second sub-images are cascaded to obtain a target motion path output by the strategy module.

[0158] In this embodiment, if Figure 3 As shown, the graph neural network model may include a dead zone cluster pooling module (DTCModule), a graph aggregation module (K-layer AGG Module) and a policy module (Policy Module) connected sequentially.

[0159] The dead zone cluster pooling module can perform cluster pooling operation on the explored sub-graph and output the first sub-graph corresponding to the explored sub-graph.

[0160] The size of the first subgraph is smaller than the explored subgraph.

[0161] The graph aggregation module can encode, aggregate, and decode the first subgraph, and can output multiple second subgraphs at different aggregation stages.

[0162] The strategy module can perform cascade processing on multiple second sub-images to output the target motion path.

[0163] In some embodiments, performing a cluster pooling operation on the explored subgraph based on the dead zone cluster pooling module to obtain a first subgraph corresponding to the explored subgraph output by the dead zone cluster pooling module may include:

[0164] When there are no other drones, regions of interest, and boundary regions in the receptive field corresponding to the target drone in the at least one drone, except the target drone, a target unit is acquired based on a grid of the number of adjacent targets in the target region;

[0165] Based on the connected nodes in the target unit, obtaining at least one cluster node corresponding to the target unit;

[0166] Process the connected cluster nodes in adjacent target units to obtain edge features between the connected cluster nodes;

[0167] A first subgraph is obtained based on at least one cluster node and each edge feature.

[0168] In this embodiment, the boundary area is a grid where the unexplored area is adjacent to the explored area.

[0169] When the K-order receptive field of drone node i does not contain other drones, interest areas and boundary areas, the receptive field area can be called the K-order dead region of drone node i, and cluster pooling operation can be performed on the explored subgraph.

[0170] The target unit includes one or more cluster nodes.

[0171] The position of a cluster node is the average of the positions of the nodes in the cluster, and the edge feature between cluster nodes can be one-third of the distance between cluster nodes.

[0172] In the actual implementation process, Figure 4 As shown in the figure, when there are no other drones, interest areas, and boundary area nodes in the K-order neighbors corresponding to the target drone, the Clupool cluster pooling operation is performed on the explored subgraph, and the adjacent 3x3 grid areas can be aggregated into a target unit, and the connected node set is determined as a cluster, and a 1-order cluster node is aggregated to construct a cluster. When there are multiple clusters in the target unit, multiple cluster nodes can be constructed. The calculation formula of cluster node characteristics is:

[0173]

[0174] Among them, c vi is the grid point set contained in node i, c vp,1 is the set of grid points contained in cluster node p obtained by first-order cluster pooling.

[0175] Based on the connected cluster nodes in adjacent cells, the edges between the two can be constructed.

[0176] The cluster node position is the average of the positions of the nodes in the cluster. The edge feature between the constructed cluster nodes is the distance between the cluster nodes ÷ 3. t After CluPool treatment, G t can be compressed into a smaller first subgraph consisting of cluster nodes

[0177] And if there are still drone nodes in the K-order dead zone in the first subgraph after pooling, the CluPool operation can be performed again.

[0178] In some embodiments, processing the first subgraph based on the graph aggregation module to obtain multiple second subgraphs output by the graph aggregation module includes:

[0179] Based on the graph encoding module, nonlinearly transform the cluster nodes and edge features corresponding to the first subgraph to obtain a third subgraph;

[0180] Perform multiple aggregation operations on the first subgraph based on multiple graph aggregation operation modules to obtain multiple fourth subgraphs;

[0181] The third sub-image and the plurality of fourth sub-images are decoded based on the plurality of image decoding modules to obtain a plurality of second sub-images.

[0182] In this embodiment, if Figure 3 As shown, the graph aggregation module may include a graph encoding module (encoder), a plurality of sequentially connected graph aggregation operation modules (GN-block) and a plurality of graph decoding modules (decoders).

[0183] The output ends of the graph encoding modules are respectively connected to the graph aggregation operation module and the graph decoding module, and the output ends of each graph aggregation module are connected to the graph decoding module.

[0184] The graph encoding module can embed the cluster nodes and edge features corresponding to the first subgraph, and can use two multilayer perceptrons (MLP) to perform nonlinear transformation on the cluster nodes and edge features respectively, and output them to the specified dimension to obtain the third subgraph.

[0185] The graph decoding module can decode the encoded graph and the graphs after each order of aggregation operation.

[0186] In the actual execution process, the graph encoding module encoder can embed the graph node features and edge features in the first subgraph, and can use two multi-layer perceptrons MLP to respectively realize the nonlinear transformation of node features and edge features, and output them to the specified dimension.

[0187] The graph aggregation module GN-block is the backbone for performing graph aggregation operations. Each time a GN operation is performed, the current node can aggregate the neighbor node information once. The input graph data is: G = {{v i},{e k}}, after one update, the graph data is G′={{v i ′},{e k ′}}, the GN-block operation feature update formula is as follows:

[0188]

[0189] Among them, MLP v and MLP e We take a 3-layer 16-node MLP network, e k is the original edge feature, and is the feature of the two end nodes of the edge, e′ k is the updated edge feature, v i is the original node feature, v′ i is the updated node feature, is the average value of the neighbor edge features of node i, E′ i is the updated set of edges connected to node i.

[0190] K aggregation operations can be stacked and performed. The graph decoding module decoder can decode the encoded graph and the graphs after each order of aggregation operations. The graph decoding module decoder can be a 3-layer 16-node MLP network, which is output to the specified dimension to obtain

[0191] In some embodiments, performing cascade processing on multiple second sub-images based on the strategy module to obtain the target motion path output by the strategy module may include:

[0192] Cascading multiple second sub-images to obtain a fifth sub-image;

[0193] Based on the activation function layer, probability values ​​corresponding to the candidate motion paths in the fifth sub-image are obtained, and the candidate motion path corresponding to the largest probability value among the multiple probability values ​​is determined as the target motion path.

[0194] In this embodiment, the policy module includes an activation function (softmax) layer.

[0195] In the actual implementation process, A concat cascade is performed, and then based on the nonlinear MLP module, the feature values ​​of the edges connected to the node where the drone is located are sampled. Then, based on the softmax layer, the probability value of selecting each candidate motion path is obtained, and the largest one is taken as the target motion path.

[0196] According to the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application, by processing the explored subgraph based on the cluster pooling method, it is possible to more effectively utilize environmental information when generating a planned path, promote coordination between multiple UAVs, and thus improve the efficiency of completing inspection tasks. By constructing a graph neural network model, UAVs can also autonomously plan paths in complex environments, reducing labor costs and improving the accuracy and precision of the generated motion path.

[0197] The multi-UAV target coverage path planning device based on cluster pooling graph neural network provided in the present application is described below. The multi-UAV target coverage path planning device based on cluster pooling graph neural network described below and the multi-UAV target coverage path planning method based on cluster pooling graph neural network described above can be referenced to each other.

[0198] The multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application can be executed by a multi-UAV target coverage path planning device based on cluster pooling graph neural network. In the embodiment of the present application, the multi-UAV target coverage path planning method based on cluster pooling graph neural network is executed by a multi-UAV target coverage path planning device based on cluster pooling graph neural network as an example to illustrate the multi-UAV target coverage path planning method based on cluster pooling graph neural network provided in the embodiment of the present application.

[0199] An embodiment of the present application also provides a multi-UAV target coverage path planning device based on a cluster pooling graph neural network.

[0200] like Figure 7 As shown, the multi-UAV target coverage path planning device based on cluster pooling graph neural network includes: a first processing module 710, a second processing module 720, a third processing module 730 and a fourth processing module 740.

[0201] The first processing module 710 is used to obtain the explored subgraph of each drone at the current moment based on the position information of each drone at the current moment and the explored area information of each drone in the target area;

[0202] The second processing module 720 is used to input the explored subgraphs corresponding to at least one drone among all the drones into the graph neural network model respectively, and obtain the target motion path corresponding to at least one drone output by the graph neural network model when the total coverage area of ​​the explored subgraphs corresponding to all the drones is smaller than the target area;

[0203] The third processing module 730 is used to control at least one UAV to move based on the target motion path, and obtain the position information and the explored area information of at least one UAV after the movement, so as to update the explored subgraph of at least one UAV;

[0204] The fourth processing module 740 is used to control each drone to stop moving when the total coverage area of ​​the updated explored sub-graphs corresponding to all drones coincides with the target area.

[0205] According to the multi-UAV target coverage path planning device based on cluster pooling graph neural network provided in the embodiment of the present application, local path planning is performed on the UAV through environmental information such as the explored area information and location information corresponding to the UAV, and a new local path planning is generated based on the environmental information after the movement of the UAV, and it is iterated until the inspection task is finally completed. The environmental information is effectively utilized in the inspection process, and the environmental information can be automatically collected for path planning without consuming a lot of manpower costs. It also promotes coordination between multiple UAVs, reduces missed inspections or repeated inspections, and improves the efficiency of completing coverage tasks.

[0206] In some embodiments, the multi-UAV target coverage path planning device based on cluster pooling graph neural network may further include a fifth processing module for training a graph neural network model based on the following steps:

[0207] Obtain multiple sample explored subgraphs corresponding to the sample drone in one collection cycle;

[0208] Based on the map information of the complete grid map, determine the sample motion paths corresponding to the explored sub-maps of each sample; at least part of the sample motion paths constitute a coherent motion path;

[0209] Taking multiple sample explored subgraphs as samples and the sample motion paths corresponding to each sample explored subgraph as sample labels, the graph neural network model is trained.

[0210] In some embodiments, one acquisition cycle includes multiple sample first moments, and the fifth processing module may also be used for:

[0211] Based on each grid point in the complete grid map, an incomplete obstacle grid set is obtained from the complete grid map;

[0212] Based on the sample position information of the sample UAV in the complete grid map at the first moment of the sample and the distance between each incomplete obstacle grid point in the incomplete obstacle grid set, the incomplete obstacle grid set is processed to obtain the sample explored subgraph corresponding to the sample UAV at the first moment of the sample.

[0213] In some embodiments, the fifth processing module may also be used to:

[0214] Based on the area ratio of the interest area corresponding to each grid point in the complete grid map, obtain the interest grid set;

[0215] Based on the interest grid set, the incomplete obstacle grid set and the perception range parameters of the sample UAV, a first grid set is obtained;

[0216] Constructing a global key point grid set based on the viewpoint grid set and the interest grid set corresponding to the first grid set;

[0217] Based on the node set in the sample explored subgraph corresponding to the global key point grid set, the sample UAV is path planned to obtain the sample motion path corresponding to the sample explored subgraph.

[0218] In some embodiments, the fifth processing module may also be used to:

[0219] Input multiple sample explored subgraphs into the graph neural network model, and obtain the predicted motion path corresponding to each sample explored subgraph output by the graph neural network model;

[0220] Construct a target loss function based on the predicted motion path and the sample motion path corresponding to the sample explored sub-graph;

[0221] Train the graph neural network model based on the objective loss function.

[0222] In some embodiments, the graph neural network model includes a dead zone cluster pooling module, a graph aggregation module, and a strategy module connected in sequence, and the second processing module 720 can also be used to:

[0223] Performing cluster pooling operation on the explored subgraph based on the dead zone cluster pooling module, obtaining a first subgraph corresponding to the explored subgraph output by the dead zone cluster pooling module;

[0224] Processing the first subgraph based on the graph aggregation module to obtain multiple second subgraphs output by the graph aggregation module;

[0225] Based on the strategy module, a plurality of second sub-images are cascaded to obtain a target motion path output by the strategy module.

[0226] In some embodiments, the second processing module 720 may also be used to:

[0227] When there are no other drones, interest areas, and boundary areas in the receptive field corresponding to the target drone in at least one drone, the target unit is obtained based on the grids of the number of adjacent targets in the target area; the boundary area is the grid adjacent to the unexplored area and the explored area;

[0228] Based on the connected nodes in the target unit, obtaining at least one cluster node corresponding to the target unit;

[0229] Process the connected cluster nodes in adjacent target units to obtain edge features between the connected cluster nodes;

[0230] A first subgraph is obtained based on at least one cluster node and each edge feature.

[0231] In some embodiments, the graph aggregation module includes a graph encoding module, a plurality of sequentially connected graph aggregation operation modules and a plurality of graph decoding modules, the output end of the graph encoding module is respectively connected to the graph aggregation operation module and the graph decoding module, and the output end of each graph aggregation module is connected to the graph decoding module; the second processing module 720 can also be used for:

[0232] Based on the graph encoding module, nonlinearly transform the cluster nodes and edge features corresponding to the first subgraph to obtain a third subgraph;

[0233] Perform multiple aggregation operations on the first subgraph based on multiple graph aggregation operation modules to obtain multiple fourth subgraphs;

[0234] The third sub-image and the plurality of fourth sub-images are decoded based on the plurality of image decoding modules to obtain a plurality of second sub-images.

[0235] In some embodiments, the policy module includes an activation function layer; the second processing module 720 may also be used to:

[0236] Cascading multiple second sub-images to obtain a fifth sub-image;

[0237] Based on the activation function layer, probability values ​​corresponding to the candidate motion paths in the fifth sub-image are obtained, and the candidate motion path corresponding to the largest probability value among the multiple probability values ​​is determined as the target motion path.

[0238] The multi-UAV target coverage path planning device based on cluster pooling graph neural network in the embodiment of the present application can be an electronic device, or a component in an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or it can be other devices other than a terminal. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc. It can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., and the embodiment of the present application is not specifically limited.

[0239] The multi-UAV target coverage path planning device based on cluster pooling graph neural network in the embodiment of the present application can be a device with an operating system. The operating system can be an Android operating system, an IOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0240] The multi-UAV target coverage path planning device based on cluster pooling graph neural network provided in the embodiment of the present application can achieve Figures 1 to 6 To avoid repetition, the various processes implemented by the method embodiment are not described here.

[0241] In some embodiments, Figure 8 As shown, an embodiment of the present application also provides an electronic device 800, including a processor 801, a memory 802, and a computer program stored in the memory 802 and executable on the processor 801. When the program is executed by the processor 801, each process of the embodiment of the multi-UAV target coverage path planning method based on the cluster pooling graph neural network is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0242] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0243] On the other hand, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the various processes of the above-mentioned multi-UAV target coverage path planning method embodiment based on cluster pooling graph neural network, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0244] On the other hand, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the various processes of the above-mentioned multi-UAV target coverage path planning method embodiment based on cluster pooling graph neural network, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0245] On the other hand, an embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned multi-UAV target coverage path planning method embodiment based on cluster pooling graph neural network, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0246] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0247] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0248] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution can be essentially or in other words, the part that contributes to the relevant technology can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0249] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-UAV target coverage path planning method based on cluster pooling graph neural network, characterized in that: include: Based on the position information of each of the drones at the current moment and the explored area information of each of the drones in the target area, acquiring the explored subgraph of each of the drones at the current moment; When the total coverage area of ​​the explored subgraphs corresponding to all the drones is smaller than the target area, the explored subgraphs corresponding to at least one of the drones among all the drones are respectively input into the graph neural network model to obtain the target motion path corresponding to at least one of the drones output by the graph neural network model; Controlling at least one of the drones to move based on the target motion path, and obtaining position information and explored area information of at least one of the drones after the movement, so as to update the explored subgraph of at least one of the drones; When the total coverage area of ​​the updated explored subgraphs corresponding to all the drones coincides with the target area, each of the drones is controlled to stop moving.

2. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to claim 1 is characterized in that: The graph neural network model is trained based on the following steps: Obtain multiple sample explored subgraphs corresponding to the sample drone in one collection cycle; Based on the map information of the complete grid map, determining the sample motion paths corresponding to the sample explored sub-maps; at least part of the sample motion paths constitute a coherent motion path; The graph neural network model is trained using the multiple sample explored subgraphs as samples and the sample motion paths corresponding to the sample explored subgraphs as sample labels.

3. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to claim 2 is characterized in that: The one collection cycle includes multiple sample first moments, and the multiple sample explored subgraphs corresponding to the sample acquisition drone in the one collection cycle include: Based on each grid point in the complete grid map, obtaining an incomplete obstacle grid set from the complete grid map; Based on the sample position information of the sample UAV in the complete grid map at the first moment of the sample and the distance between each of the incomplete obstacle grid points in the incomplete obstacle grid set, the incomplete obstacle grid set is processed to obtain the sample explored subgraph corresponding to the sample UAV at the first moment of the sample.

4. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to claim 3 is characterized in that: The determining of the sample motion path corresponding to each sample explored sub-graph based on the map information of the complete grid map includes: Based on the area ratio of the interest region corresponding to each grid point in the complete grid map, obtaining the interest grid set; Acquire a first grid set based on the interest grid set, the incomplete obstacle grid set, and the perception range parameter of the sample UAV; Constructing a global key point grid set based on the viewpoint grid set corresponding to the first grid set and the interest grid set; Based on the node set in the sample explored subgraph corresponding to the global key point grid set, path planning is performed on the sample UAV to obtain a sample motion path corresponding to the sample explored subgraph.

5. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to claim 2 is characterized in that: The method of training the graph neural network model by taking the plurality of sample explored subgraphs as samples and the sample motion paths corresponding to the sample explored subgraphs as sample labels includes: Input the multiple sample explored subgraphs into the graph neural network model, and obtain the predicted motion path corresponding to each of the sample explored subgraphs output by the graph neural network model; constructing a target loss function based on the predicted motion path and the sample motion path corresponding to the sample explored sub-graph; The graph neural network model is trained based on the target loss function.

6. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to any one of claims 1 to 5 is characterized in that: The graph neural network model includes a dead zone cluster pooling module, a graph aggregation module and a strategy module connected in sequence, and the explored subgraphs corresponding to at least one of the drones among all the drones are respectively input into the graph neural network model to obtain a target motion path corresponding to at least one of the drones output by the graph neural network model, including: Performing a cluster pooling operation on the explored subgraph based on the dead zone cluster pooling module, obtaining a first subgraph corresponding to the explored subgraph output by the dead zone cluster pooling module; Processing the first subgraph based on the graph aggregation module to obtain a plurality of second subgraphs output by the graph aggregation module; The plurality of second sub-images are cascaded based on the strategy module to obtain the target motion path output by the strategy module.

7. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to claim 6 is characterized in that: The performing a cluster pooling operation on the explored subgraph based on the dead zone cluster pooling module to obtain a first subgraph corresponding to the explored subgraph output by the dead zone cluster pooling module includes: When there is no other drone, interest area and boundary area in at least one of the drones except the target drone in the receptive field corresponding to the target drone in at least one of the drones, a target unit is obtained based on a grid of adjacent targets in the target area; the boundary area is a grid adjacent to an unexplored area and an explored area; Based on the connected nodes in the target unit, acquiring at least one cluster node corresponding to the target unit; Processing the connected cluster nodes in the adjacent target units to obtain edge features between the connected cluster nodes; Based on the at least one cluster node and each of the edge features, the first subgraph is obtained.

8. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to claim 6 is characterized in that: The graph aggregation module includes a graph encoding module, a plurality of sequentially connected graph aggregation operation modules and a plurality of graph decoding modules, wherein the output end of the graph encoding module is respectively connected to the graph aggregation operation module and the graph decoding module, and the output end of each graph aggregation module is connected to the graph decoding module; the processing of the first subgraph based on the graph aggregation module to obtain a plurality of second subgraphs output by the graph aggregation module includes: Based on the graph encoding module, nonlinearly transform the cluster nodes and edge features corresponding to the first subgraph to obtain a third subgraph; Perform multiple aggregation operations on the first subgraph based on the multiple graph aggregation operation modules to obtain multiple fourth subgraphs; The third sub-image and the plurality of fourth sub-images are decoded based on the plurality of image decoding modules to obtain the plurality of second sub-images.

9. The multi-UAV target coverage path planning method based on cluster pooling graph neural network according to claim 6 is characterized in that: The strategy module includes an activation function layer; the cascading processing of the plurality of second sub-images based on the strategy module to obtain the target motion path output by the strategy module includes: Performing cascade processing on the multiple second sub-images to obtain a fifth sub-image; Based on the activation function layer, probability values ​​corresponding to the candidate motion paths in the fifth sub-image are obtained, and the candidate motion path corresponding to the largest probability value among multiple probability values ​​is determined as the target motion path.

10. A multi-UAV target coverage path planning device based on cluster pooling graph neural network, characterized in that: include: A first processing module, configured to obtain an explored subgraph of each drone at the current moment based on the position information of each drone at the current moment and the explored area information of each drone in the target area; A second processing module is used to input the explored subgraphs corresponding to at least one of the drones into the graph neural network model to obtain the target motion path corresponding to at least one of the drones output by the graph neural network model when the total coverage area of ​​the explored subgraphs corresponding to all the drones is smaller than the target area; a third processing module, configured to control at least one of the drones to move based on the target motion path, and obtain position information and explored area information of at least one of the drones after the movement, so as to update an explored subgraph of at least one of the drones; The fourth processing module is used to control each of the drones to stop moving when the total coverage area of ​​the updated explored sub-graphs corresponding to all the drones coincides with the target area.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the multi-UAV target coverage path planning method based on cluster pooling graph neural network is implemented as described in any one of claims 1-9.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the multi-UAV target coverage path planning method based on cluster pooling graph neural network as described in any one of claims 1 to 9 is implemented.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the multi-UAV target coverage path planning method based on cluster pooling graph neural network is implemented as described in any one of claims 1 to 9.

Images