Unmanned aerial vehicle cluster multi-dynamic target autonomous tracking method and device

Abstract

The invention relates to an unmanned aerial vehicle cluster multi-dynamic target autonomous tracking method and device. The method comprises the following steps: acquiring cluster information, environment information and behavior data information of an unmanned aerial vehicle cluster, constructing a scene model and a cluster behavior model of the unmanned aerial vehicle cluster, and generating path planning information of the cluster behavior model through a cluster path planning strategy based on the scene model; identifying distance information between the unmanned aerial vehicle cluster and each dynamic target cluster through a scene model based on the path planning information, and generating a dynamic tracking control strategy of the unmanned aerial vehicle cluster through a multi-dynamic target tracking algorithm based on the scene model when the distance information meets a preset distance condition; and based on the dynamic tracking control strategy, the unmanned aerial vehicle cluster is controlled, tracking processing is performed on each dynamic target cluster, and a multi-dynamic-target autonomous tracking task of the unmanned aerial vehicle cluster is completed. By adopting the method, the multi-dynamic target tracking efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method and device for autonomously tracking multiple dynamic targets in a drone cluster. Background Art

[0002] With the development of intelligent technology, autonomous unmanned swarm systems are gaining more and more attention. Unmanned Aerial Vehicle (UAV) is widely used due to its low cost, high mobility, and environmental adaptability. The multi-dynamic target tracking task of UAV swarm has the characteristics of high complexity and nonlinearity. Traditional path planning methods, such as the Artificial Potential Field (APF), can calculate a preliminary path, but it is difficult to make real-time state assessment and motion decisions with a high success rate, and it is easily affected by modeling accuracy, environmental noise, etc. Therefore, how to improve the accuracy of multi-dynamic target tracking of UAV swarms is the current research focus.

[0003] In existing research, end-to-end reinforcement learning methods are usually used to track multiple dynamic targets in UAV clusters. Due to the large range of accessible maps and the strong maneuverability of UAVs, end-to-end MARL has a very large state space, which makes it difficult for the algorithm to train a usable strategy, the reward function is difficult to converge, and it is difficult to generate a tracking strategy that reaches the success rate indicator within a limited time. As a result, the efficiency of tracking multiple dynamic targets is low. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for autonomous tracking of multiple dynamic targets in a drone swarm in response to the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a method for autonomously tracking multiple dynamic targets in a drone cluster, comprising:

[0006] Acquire cluster information of a drone cluster, environmental information of the drone cluster, and behavior data information of the drone cluster, and construct a scene model of the drone cluster based on the cluster information and the environmental information;

[0007] Based on the behavior data information of the drone cluster, a cluster behavior model of the drone cluster is constructed, and based on the scenario model, path planning information of the cluster behavior model is generated through a cluster path planning strategy;

[0008] Based on the path planning information, the distance information between the drone cluster and each dynamic target cluster is identified through the scene model, and when the distance information meets the preset distance condition, a dynamic tracking control strategy of the drone cluster is generated based on the scene model through a multi-dynamic target tracking algorithm;

[0009] Based on the dynamic tracking control strategy, the drone cluster is controlled to track each of the dynamic target clusters to complete the multi-dynamic target autonomous tracking task of the drone cluster.

[0010] Optionally, constructing the scene model of the drone cluster based on the cluster information and the environmental information includes:

[0011] Based on the environmental information, identify information about each restriction in the environment in which the drone cluster is located, and initial target position information of a dynamic target cluster in the environment in which the drone cluster is located, and based on the cluster information, identify initial drone position information of each drone in the drone cluster, and cluster target position information of cluster targets of the dynamic target cluster;

[0012] Based on the information of each of the restricting objects, a scene plane model of the drone cluster is constructed, and based on the initial target position information of the dynamic target cluster, the initial drone position information of each of the drones, and the cluster target position information of the cluster targets of the dynamic target cluster, drone position parameters of the drone cluster, target position parameters of the dynamic target cluster, and cluster target position parameters of the cluster targets are identified in the scene plane model;

[0013] The drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster targets are added to the scene plane model to obtain the scene model of the drone cluster.

[0014] Optionally, constructing a cluster behavior model of the drone cluster based on the behavior data information of the drone cluster includes:

[0015] For each drone of the drone cluster, based on the behavior data information of the drone, identifying the speed information of the drone and the location information of the drone;

[0016] Based on the speed information of the drone, a motion model of the drone is constructed through a motion model modeling strategy, and based on the position information of the drone, position state progressive information of the drone is generated through a position state algorithm;

[0017] The motion model of the drone and the progressive information of the position state of the drone are used as the drone behavior model of the drone, and all the drone behavior models are used as the cluster behavior model of the drone cluster.

[0018] Optionally, the generating the path planning information of the cluster behavior model based on the scenario model through a cluster path planning strategy includes:

[0019] Identify the drone position distribution information of the drone cluster in the scene model, the target position distribution information of the dynamic target cluster in the scene model, and the restricted position information of each restricted object in the scene model;

[0020] Based on the drone position distribution information, the target position distribution information, and each of the restricted position information, a cluster path planning strategy is used to generate path planning information of the cluster behavior model of each drone.

[0021] Optionally, the identifying, based on the path planning information and through the scene model, distance information between the drone cluster and each dynamic target cluster includes:

[0022] Based on the path planning information, the target tracking process of the cluster behavior model is simulated through the scenario model to obtain new drone position distribution information of the drone cluster and new target position distribution information of the dynamic target cluster;

[0023] Based on the new target position distribution information and the new drone position distribution information, the target center position information of the dynamic target cluster and the drone center position information of the drone cluster are identified through the center position division algorithm, and based on the target center position information and the drone center position information, the distance information between the drone cluster and each dynamic target cluster is calculated.

[0024] Optionally, when the distance information meets a preset distance condition, a dynamic tracking control strategy for the drone cluster is generated based on the scene model through a multi-dynamic target tracking algorithm, including:

[0025] When the distance information is lower than a preset distance threshold, determining that the distance information satisfies a preset distance condition, and updating the scene model based on the new drone position distribution information of each of the drones and the new target position distribution information of the dynamic target cluster to obtain a new scene model;

[0026] Based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, a new tracking route for each drone of the drone cluster is generated through a multi-dynamic target tracking algorithm;

[0027] Based on the new tracking routes of each of the drones, the new scene model is updated to obtain the current drone position distribution information of the drone cluster and the current target position distribution information of the dynamic target cluster, and when there is no current drone position information that overlaps with the current target position distribution information of the dynamic target cluster, the current drone position distribution information of the drone cluster is replaced by the new drone position distribution information of the drone cluster, and the current target position distribution information of the dynamic target cluster is replaced by the current target position distribution information of the dynamic target cluster;

[0028] Return to the step of generating a new tracking route for each drone of the drone cluster based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target through a multi-dynamic target tracking algorithm, until the current position information of the drone exists and overlaps with the current target position distribution information of the dynamic target cluster. Then, the new tracking route obtained in each iteration is summarized in the order of the generation time of each new tracking route to obtain the dynamic tracking control strategy of the drone cluster.

[0029] In the second aspect, the present application also provides a drone cluster multi-dynamic target autonomous tracking device, comprising:

[0030] An acquisition module, used to acquire cluster information of a drone cluster, environmental information of the drone cluster, and behavioral data information of the drone cluster, and to construct a scene model of the drone cluster based on the cluster information and the environmental information;

[0031] A generation module, configured to construct a cluster behavior model of the drone cluster based on the behavior data information of the drone cluster, and generate path planning information of the cluster behavior model through a cluster path planning strategy based on the scenario model;

[0032] An identification module, configured to identify the distance information between the drone cluster and each dynamic target cluster based on the path planning information and through the scene model, and generate a dynamic tracking control strategy for the drone cluster based on the scene model and through a multi-dynamic target tracking algorithm when the distance information meets a preset distance condition;

[0033] The control module is used to control the drone cluster based on the dynamic tracking control strategy, track each dynamic target cluster, and complete the multi-dynamic target autonomous tracking task of the drone cluster.

[0034] Optionally, the acquisition module is specifically used to:

[0035] Based on the environmental information, identify information about each restriction in the environment in which the drone cluster is located, and initial target position information of a dynamic target cluster in the environment in which the drone cluster is located, and based on the cluster information, identify initial drone position information of each drone in the drone cluster, and cluster target position information of cluster targets of the dynamic target cluster;

[0036] Based on the information of each of the restricting objects, a scene plane model of the drone cluster is constructed, and based on the initial target position information of the dynamic target cluster, the initial drone position information of each of the drones, and the cluster target position information of the cluster targets of the dynamic target cluster, drone position parameters of the drone cluster, target position parameters of the dynamic target cluster, and cluster target position parameters of the cluster targets are identified in the scene plane model;

[0037] The drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster targets are added to the scene plane model to obtain the scene model of the drone cluster.

[0038] Optionally, the generating module is specifically used for:

[0039] For each drone of the drone cluster, based on the behavior data information of the drone, identifying the speed information of the drone and the location information of the drone;

[0040] Based on the speed information of the drone, a motion model of the drone is constructed through a motion model modeling strategy, and based on the position information of the drone, position state progressive information of the drone is generated through a position state algorithm;

[0041] The motion model of the drone and the progressive information of the position state of the drone are used as the drone behavior model of the drone, and all the drone behavior models are used as the cluster behavior model of the drone cluster.

[0042] Optionally, the generating module is specifically used for:

[0043] Identify the drone position distribution information of the drone cluster in the scene model, the target position distribution information of the dynamic target cluster in the scene model, and the restricted position information of each restricted object in the scene model;

[0044] Based on the drone position distribution information, the target position distribution information, and each of the restricted position information, a cluster path planning strategy is used to generate path planning information of the cluster behavior model of each drone.

[0045] Optionally, the identification module is specifically used to:

[0046] Based on the path planning information, the target tracking process of the cluster behavior model is simulated through the scenario model to obtain new drone position distribution information of the drone cluster and new target position distribution information of the dynamic target cluster;

[0047] Based on the new target position distribution information and the new drone position distribution information, the target center position information of the dynamic target cluster and the drone center position information of the drone cluster are identified through the center position division algorithm, and based on the target center position information and the drone center position information, the distance information between the drone cluster and each dynamic target cluster is calculated.

[0048] Optionally, the identification module is specifically used to:

[0049] When the distance information is lower than a preset distance threshold, determining that the distance information satisfies a preset distance condition, and updating the scene model based on the new drone position distribution information of each of the drones and the new target position distribution information of the dynamic target cluster to obtain a new scene model;

[0050] Based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, a new tracking route for each drone of the drone cluster is generated through a multi-dynamic target tracking algorithm;

[0051] Based on the new tracking routes of each of the drones, the new scene model is updated to obtain the current drone position distribution information of the drone cluster and the current target position distribution information of the dynamic target cluster, and when there is no current drone position information that overlaps with the current target position distribution information of the dynamic target cluster, the current drone position distribution information of the drone cluster is replaced by the new drone position distribution information of the drone cluster, and the current target position distribution information of the dynamic target cluster is replaced by the current target position distribution information of the dynamic target cluster;

[0052] Return to the step of generating a new tracking route for each drone of the drone cluster based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target through a multi-dynamic target tracking algorithm, until the current position information of the drone exists and overlaps with the current target position distribution information of the dynamic target cluster. Then, the new tracking route obtained in each iteration is summarized in the order of the generation time of each new tracking route to obtain the dynamic tracking control strategy of the drone cluster.

[0053] In a third aspect, the present application provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the methods in the first aspect are implemented.

[0054] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of any one of the methods in the first aspect are implemented.

[0055] In a fifth aspect, the present application provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of any one of the methods in the first aspect are implemented.

[0056] The above-mentioned method and device for autonomous tracking of multiple dynamic targets in a drone cluster obtains cluster information of the drone cluster, environmental information of the drone cluster, and behavioral data information of the drone cluster, and constructs a scene model of the drone cluster based on the cluster information and the environmental information; constructs a cluster behavior model of the drone cluster based on the behavioral data information of the drone cluster, and generates path planning information of the cluster behavior model through a cluster path planning strategy based on the scene model; identifies the distance information between the drone cluster and each dynamic target cluster through the scene model based on the path planning information, and generates a dynamic tracking control strategy for the drone cluster through a multi-dynamic target tracking algorithm based on the scene model when the distance information meets a preset distance condition; controls the drone cluster based on the dynamic tracking control strategy, tracks each of the dynamic target clusters, and completes the autonomous tracking task of multiple dynamic targets of the drone cluster. This scheme generates the path planning information of the cluster behavior model according to the cluster path planning strategy, and then generates the dynamic tracking control strategy of the UAV cluster through the multi-dynamic target tracking algorithm. First, in the initial stage, the UAV autonomously explores the environment. Since the strategy is poor at this time, the environmental reward feedback obtained is very sparse. If the reinforcement learning method is used, it will take a lot of time from the beginning to the training to obtain a usable strategy. In order to avoid meaningless exploration in the initial stage, this scheme uses the cluster path planning strategy in the initial stage, which can enable the UAV cluster to quickly approach the mobile target cluster while avoiding obstacles. Then, when each UAV and the target cluster quickly approach to a certain distance threshold, the task difficulty is reduced. At this time, the multi-dynamic target tracking algorithm is used to fully explore the feasible strategy space, so that each UAV can accurately reach the target cluster. In this way, the tracking time of the UAV is greatly reduced in the early exploration stage, and the target cluster can be accurately tracked in the later stage of tracking the target, so that while ensuring the tracking accuracy of the target cluster, the early exploration time is greatly reduced, thereby ensuring that the success rate index is generated within a limited time, and comprehensively improving the efficiency of multi-dynamic target tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 A schematic diagram of a flow chart of a method for autonomously tracking multiple dynamic targets of a drone cluster in one embodiment;

[0059] Figure 2 is a visual schematic diagram of a scene model in one embodiment;

[0060] Figure 3 A schematic diagram of a search of a drone cluster in the first stage in one embodiment;

[0061] Figure 4 A convergence comparison diagram between the tracking method of the present solution and the traditional tracking method in an embodiment;

[0062] Figure 5 A flowchart of autonomous tracking of multiple dynamic targets in a drone cluster in one embodiment;

[0063] Figure 6 A schematic diagram of a process of autonomous tracking of multiple dynamic targets in a drone cluster in one embodiment;

[0064] Figure 7 It is a structural block diagram of a multi-dynamic target autonomous tracking device for a drone cluster in one embodiment;

[0065] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0067] The method for autonomous tracking of multiple dynamic targets in a drone cluster provided in an embodiment of the present application can be applied to an application environment of autonomous tracking of multiple dynamic targets in a drone cluster. The method can be applied to a terminal, a server, or a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, medium-sized computers, etc. The terminal generates the path planning information of the cluster behavior model according to the cluster path planning strategy, and then generates the dynamic tracking control strategy of the drone (Unmanned Aerial Vehicle, referred to as UAV) cluster through the multi-dynamic target tracking algorithm. First, in the initial stage, the drone autonomously explores the environment. Since the strategy is poor at this time, the environmental reward feedback obtained is very sparse. If the reinforcement learning method is used, it will take a lot of time from the beginning stage to the training to obtain a usable strategy. In order to avoid meaningless exploration in the initial stage, this solution uses the cluster path planning strategy in the initial stage, which can enable the drone cluster to quickly approach the mobile target cluster while avoiding obstacles. Then, when each drone and the target cluster quickly approach a certain distance threshold, the task difficulty is reduced. At this time, the multi-dynamic target tracking algorithm is used to fully explore the feasible strategy space, so that each drone can accurately reach the target cluster. This greatly reduces the tracking time of the drone in the early exploration stage, and enables the target cluster to be accurately tracked in the later target tracking stage. This ensures the tracking accuracy of the target cluster while greatly reducing the early exploration time, thereby ensuring that the success rate indicator is achieved within a limited time, and comprehensively improving the efficiency of multi-dynamic target tracking.

[0068] In an exemplary embodiment, Figure 1 As shown, a method for autonomous tracking of multiple dynamic targets in a drone cluster is provided, and the method is applied to a terminal as an example for explanation, including the following steps S101 to S104. Among them:

[0069] Step S101, obtaining cluster information of the drone cluster, environmental information of the drone cluster, and behavioral data information of the drone cluster, and constructing a scene model of the drone cluster based on the cluster information and environmental information.

[0070] In this embodiment, the terminal obtains the cluster information of the drone cluster, the environmental information of the drone cluster, and the behavioral data information of the drone cluster in response to the information upload operation of the staff. Among them, the cluster information includes the location distribution information of each drone in the drone cluster, and the environmental information includes the location distribution information of the target cluster, the location information of the target point that the target cluster needs to go to (that is, the cluster target corresponding to the target cluster), and the location information of the restriction of each obstacle in the environment, wherein the location information of the restriction includes the flight boundary range of the drone and each obstacle in the environment. The behavioral data information of the drone cluster is the speed information (angular velocity, linear velocity) of the drone and the orientation information of the drone. Then, the terminal constructs a scene model of the drone cluster based on the cluster information and the environmental information. The specific construction process will be described in detail later.

[0071] Step S102, based on the behavior data information of the drone cluster, a cluster behavior model of the drone cluster is constructed, and based on the scene model, path planning information of the cluster behavior model is generated through a cluster path planning strategy.

[0072] In this embodiment, the terminal constructs a cluster behavior model of the drone cluster based on the behavior data information of the drone cluster, and generates path planning information of the cluster behavior model based on the scene model through the cluster path planning strategy. Among them, the cluster behavior model includes the drone behavior model of each drone, and the drone behavior model includes the motion model of the drone and the position state progressive information of the drone. Among them, the position state progressive information is the prediction equation of the position of the drone at the next moment. The path planning information of the cluster behavior model is the exploration path planning information of each drone to explore and search for the target cluster during the exploration process of the first stage, wherein the cluster path planning strategy is a planning strategy based on the artificial potential field method (Artificial Potential Field, referred to as APF) path planning, which can calculate a preliminary path, but it is difficult to make a real-time state evaluation and motion decision with a high success rate. The specific generation process will be described in detail later.

[0073] Step S103, based on the path planning information, the distance information between the drone cluster and each dynamic target cluster is identified through the scene model, and when the distance information meets the preset distance condition, the dynamic tracking control strategy of the drone cluster is generated based on the scene model through the multi-dynamic target tracking algorithm.

[0074] In this embodiment, the terminal identifies the distance information between the drone cluster and each dynamic target cluster based on the path planning information and the scene model, and generates a dynamic tracking control strategy for the drone cluster based on the scene model and the multi-dynamic target tracking algorithm when the distance information meets the preset distance condition. Among them, the distance information is the distance information between the center position of the drone cluster and the center position of the dynamic target cluster, and the preset distance condition is that the distance information is lower than the distance threshold preset in the terminal. The specific judgment process of whether the distance information meets the preset distance condition will be described in detail later. The dynamic target tracking algorithm is a multi-drone reinforcement learning method (Multi-Agent Reinforcement Learning, referred to as MARL) combined with a neural network structure, which has certain feasibility for nonlinear large-scale multi-unmanned platform collaborative strategy generation, and has certain generalization ability in different random initializations. Among them, the dynamic tracking control strategy includes the dynamic tracking control strategy of the drone cluster in each iterative tracking process.

[0075] Step S104, based on the dynamic tracking control strategy, the drone cluster is controlled to track each dynamic target cluster to complete the multi-dynamic target autonomous tracking task of the drone cluster.

[0076] In this embodiment, the terminal uses a dynamic tracking control strategy to guide each drone in the drone cluster to iteratively track the dynamic target cluster until the dynamic target cluster is tracked, thereby determining that the multi-dynamic target autonomous tracking task of the drone cluster is completed.

[0077] Based on the above scheme, the path planning information of the cluster behavior model is first generated according to the cluster path planning strategy, and then the dynamic tracking control strategy of the UAV (Unmanned Aerial Vehicle, referred to as UAV) cluster is generated through the multi-dynamic target tracking algorithm. First, in the initial stage, the UAV autonomously explores the environment. Since the strategy is poor at this time, the environmental reward feedback obtained is very sparse. If the reinforcement learning method is used, it will take a lot of time from the beginning stage to the training to obtain a usable strategy. In order to avoid meaningless exploration in the initial stage, this scheme uses the cluster path planning strategy in the initial stage, which can enable the UAV cluster to quickly approach the mobile target cluster while avoiding obstacles. Then, when each UAV and the target cluster quickly approach to a certain distance threshold, the task difficulty is reduced. At this time, the multi-dynamic target tracking algorithm is used to fully explore the feasible strategy space, so that each UAV can accurately reach the target cluster. In this way, the tracking time of the UAV is greatly reduced in the early exploration stage, and the target cluster can be accurately tracked in the later stage of tracking the target, so that while ensuring the tracking accuracy of the target cluster, the early exploration time is greatly reduced, thereby ensuring that the success rate index is generated within a limited time, and comprehensively improving the efficiency of multi-dynamic target tracking.

[0078] Optionally, based on cluster information and environmental information, a scene model of a drone cluster is constructed, including: based on the environmental information, identifying information on various restrictions in the environment in which the drone cluster is located, and initial target position information of a dynamic target cluster in the environment in which the drone cluster is located, and based on the cluster information, identifying initial drone position information of each drone in the drone cluster, and cluster target position information of cluster targets of the dynamic target cluster; based on the information on various restrictions, a scene plane model of the drone cluster is constructed, and based on the initial target position information of the dynamic target cluster, the initial drone position information of each drone, and the cluster target position information of cluster targets of the dynamic target cluster, in the scene plane model, identifying drone position parameters of the drone cluster, target position parameters of the dynamic target cluster, and cluster target position parameters of the cluster targets; adding the drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster targets to the scene plane model to obtain a scene model of the drone cluster.

[0079] In this embodiment, the terminal identifies the information of each restriction in the environment where the drone cluster is located and the initial target position information of the dynamic target cluster in the environment where the drone cluster is located based on the environmental information, and identifies the initial drone position information of each drone in the drone cluster and the cluster target position information of the cluster target of the dynamic target cluster based on the cluster information. The information of each restriction includes the boundary range information of the drone and the obstacle position information of each obstacle.

[0080] Then, the terminal constructs a scene plane model of the drone cluster based on the information of each restriction, and identifies the drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster target in the scene plane model based on the initial target position information of the dynamic target cluster, the initial drone position information of each drone, and the cluster target position information of the cluster target of the dynamic target cluster. The scene plane model is a two-dimensional plane model, and each position parameter is two-dimensional position information in the two-dimensional plane model.

[0081] Finally, the terminal adds the drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster target to the scene plane model to obtain the scene model of the drone cluster.

[0082] Based on the above scheme, by first modeling the scene plane, and then adding the drone cluster, dynamic target cluster, and cluster target to the scene plane model, the scene model of the drone cluster is obtained, which improves the applicability and accuracy of the construction of the scene model.

[0083] Optionally, a cluster behavior model of the drone cluster is constructed based on the behavior data information of the drone cluster, including: for each drone in the drone cluster, identifying the speed information and the position information of the drone based on the behavior data information of the drone; based on the speed information of the drone, constructing a motion model of the drone through a motion model modeling strategy, and based on the position information of the drone, generating the position state progressive information of the drone through a position state algorithm; using the motion model of the drone and the position state progressive information of the drone as the drone behavior model of the drone, and using all drone behavior models as the cluster behavior model of the drone cluster.

[0084] In this embodiment, the terminal identifies the speed information and position information of each drone in the drone cluster based on the behavior data information of the drone. Then, based on the speed information of the drone, a motion model of the drone is constructed through a motion modeling strategy, and based on the position information of the drone, the position state progressive information of the drone is generated through a position state algorithm.

[0085] Specifically, a single UAV The motion model is shown below.

[0086]

[0087] in and UAV along The speed in the axis direction, is the angular velocity of the UAV at this moment, For its The angle of the positive axis, is the linear speed, is the angular velocity.

[0088] Then the position state equation of the UAV at the next moment (i.e., the progressive position state information) is as follows:

[0089]

[0090] Finally, the terminal uses the motion model of the drone and the progressive information of the position status of the drone as the drone behavior model of the drone, and uses all the drone behavior models as the cluster behavior model of the drone cluster.

[0091] Based on the above solution, by individually modeling each drone, it is ensured that the route of each drone can be accurately controlled when actually planning and tracking the route, thereby improving the accuracy of drone route planning.

[0092] Optionally, based on the scene model, path planning information of the cluster behavior model is generated through a cluster path planning strategy, including: identifying the drone position distribution information of the drone cluster in the scene model, the target position distribution information of the dynamic target cluster in the scene model, and the restricted position information of each restriction in the scene model; based on the drone position distribution information, the target position distribution information, and each restricted position information, the path planning information of the cluster behavior model of each drone is generated through a cluster path planning strategy.

[0093] In this embodiment, the terminal identifies the drone position distribution information of the drone cluster in the scene model, the target position distribution information of the dynamic target cluster in the scene model, and the restricted position information of each restriction in the scene model. Then, based on the drone position distribution information, the target position distribution information, and each restricted position information, the terminal generates the path planning information of the cluster behavior model of each drone through the cluster path planning strategy.

[0094] Specifically, the terminal constructs gravitational fields at the target point and obstacle positions respectively, based on the concepts of electric potential and electric potential field in electrostatic fields. With repulsive field ,in,

[0095] The drone will be attracted to move towards the target point. Helps drones avoid obstacles. In the potential field, the drone is regarded as a particle. It moves a specified unit step under the combined force of the current position. After reaching the next position, the potential field is updated in real time to plan the next path. One of the significant advantages of APF is its strong real-time performance, which can better cope with changes in targets and dynamic obstacles in the environment. The combined force on the drone in the virtual potential field obtained by the APF method is:

[0096]

[0097] in For the joint efforts, The attraction given to the drone by the target point, is the repulsive force exerted on the drone by obstacles in the environment. Each drone performs path planning in the initial stage under the action of its own combined force.

[0098] Based on the above scheme, the strategy design in the initial stage is carried out by directly introducing the expert experience of the APF strategy. Using the APF method in the initial stage can enable the UAV cluster to quickly approach the moving target cluster while avoiding obstacles. It improves the search efficiency of each UAV in the first stage.

[0099] Optionally, based on the path planning information, the distance information between the drone cluster and each dynamic target cluster is identified through the scene model, including: based on the path planning information, the target tracking process of the cluster behavior model is simulated through the scene model to obtain new drone position distribution information of the drone cluster and new target position distribution information of the dynamic target cluster; based on the new target position distribution information and the new drone position distribution information, the target center position information of the dynamic target cluster and the drone center position information of the drone cluster are identified through a center position division algorithm, and based on the target center position information and the drone center position information, the distance information between the drone cluster and each dynamic target cluster is calculated.

[0100] In this embodiment, the terminal simulates the target tracking process of the cluster behavior model through the scene model based on the path planning information to obtain new drone position distribution information of the drone cluster and new target position distribution information of the dynamic target cluster.

[0101] Specifically, when , that is, the distance information between the drone cluster and the dynamic target cluster is greater than a preset distance threshold When , it is considered to be in the first stage. In this stage, the rewards are sparse, and no reinforcement learning method is used. The drone uses the traditional path planning method to directly and weakly randomly approach the moving target quickly for subsequent tracking.

[0102] In this stage, the line connecting the virtual chaser and the virtual navigator divides the cluster into two parts. The drones in these two parts no longer chase the virtual navigator of the mobile target cluster, but instead chase an equivalent target point on a circle with it as the center and r as the radius. The drones on both sides chase the equivalent target point on the same side as themselves in this stage, and through such left and right "expansion" operations of the target point, an encirclement trend is formed.

[0103] The tracking strategy at this stage is as follows Figure 3 The adjustable parameters are , r.

[0104] Then, based on the new target location distribution information and the new drone location distribution information, the terminal uses the center location division algorithm to identify the target center location information of the dynamic target cluster and the drone center location information of the drone cluster, and calculates the distance information between the drone cluster and each dynamic target cluster based on the target center location information and the drone center location information. The specific process of calculating each center location information is as follows:

[0105] Terminal defines the location of the "virtual tracker" :

[0106] (4)

[0107] in For the The proposed “virtual tracker” can characterize the central position of the drone cluster to a certain extent.

[0108] Similarly, the observed center position of the moving target cluster can be defined as , called the "virtual navigator", represents the center position of the mobile target cluster. The relative distance between the coordinate centers of the "virtual tracker" and the "virtual navigator" clusters is calculated to obtain the distance information between the drone cluster and each dynamic target cluster.

[0109] Based on the above solution, by defining the center position of each cluster, the distance information between two clusters is calculated, thereby improving the calculation accuracy and adaptability of the distance information.

[0110] Optionally, when the distance information meets the preset distance condition, based on the scene model, a dynamic tracking control strategy for the drone cluster is generated through a multi-dynamic target tracking algorithm, including: when the distance information is lower than a preset distance threshold, determining that the distance information meets the preset distance condition, and based on the new drone position distribution information of each drone and the new target position distribution information of the dynamic target cluster, updating the scene model to obtain a new scene model; based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, generating a new tracking route for each drone in the drone cluster through a multi-dynamic target tracking algorithm; based on the new tracking route of each drone, updating the new scene model to obtain the current drone position distribution information of the drone cluster and the current target position distribution information of the dynamic target cluster, and in the absence of When the current position information of the drone overlaps with the current target position distribution information of the dynamic target cluster, the current drone position distribution information of the drone cluster replaces the new drone position distribution information of the drone cluster, and the current target position distribution information of the dynamic target cluster replaces the current target position distribution information of the dynamic target cluster; return to execute the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, and generate a new tracking route step for each drone in the drone cluster through a multi-dynamic target tracking algorithm until the current position information of the drone overlaps with the current target position distribution information of the dynamic target cluster, then the new tracking route obtained in each iteration is summarized in the order of the generation time of each new tracking route to obtain the dynamic tracking control strategy of the drone cluster.

[0111] In this embodiment, when the distance information is lower than a preset distance threshold, the terminal determines that the distance information meets the preset distance condition, and updates the scene model based on the new drone position distribution information of each drone and the new target position distribution information of the dynamic target cluster to obtain a new scene model.

[0112] Then, the terminal generates a new tracking route for each drone in the drone cluster through a multi-dynamic target tracking algorithm based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target.

[0113] Afterwards, the terminal updates the new scene model based on the new tracking routes of each drone, obtains the current drone position distribution information of the drone cluster, and the current target position distribution information of the dynamic target cluster, and when there is no current position information of the drone and it overlaps with the current target position distribution information of the dynamic target cluster, the current drone position distribution information of the drone cluster replaces the new drone position distribution information of the drone cluster, and the current target position distribution information of the dynamic target cluster replaces the current target position distribution information of the dynamic target cluster.

[0114] Finally, the terminal returns to execute the new drone position distribution information based on the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, and generates a new tracking route step for each drone in the drone cluster through a multi-dynamic target tracking algorithm until the current position information of the drone exists and overlaps with the current target position distribution information of the dynamic target cluster. The new tracking route obtained in each iteration is summarized in the order of the generation time of each new tracking route to obtain the dynamic tracking control strategy of the drone cluster.

[0115] Specifically, the terminal can combine any existing MARL algorithm through a phased framework. Such as Multi-Agent Deep Deterministic Policy Gradient (MADDPG) and Double Delay MADDPG (MATD3). The MADDPG algorithm is based on the Actor-Critic framework and is a multi-drone version of the DDPG algorithm based on the "centralized training + decentralized decision-making" architecture. It can handle continuous action spaces. The MATD3 algorithm is an improved algorithm of the MADDPG algorithm, which obtains a dynamic target tracking algorithm. In actual tracking, for drone clusters, considering that their mission focuses on tracking all targets as short as possible in the shortest mission time, and trying to keep the tracking position of the mobile target away from the target point of the mobile target, the reward function is not simply determined by whether the tracking is successful in the end. The phased rewards in the intermediate process can be designed through reward shaping to provide relatively dense reward information to speed up the training.

[0116] Drone The reward function is , considering collision rewards for successful tracking, distance rewards for dynamic threats (moving targets), distance rewards for other drones and static obstacles (obstacle avoidance rewards), etc.

[0117] Tracking Success Collision Reward: When the drone Distance to any moving target The distance is less than the set threshold When , the tracking is considered successful, and the drone and the target exit the environment at the same time. Get collision rewards as follows:

[0118] (5)

[0119] in For a larger completion reward, is an indicator function, and its value is 1 when the condition in the independent variable is true, otherwise it is 0. This reward item encourages the drone to track any moving target.

[0120] Distance negative reward: In order to make the drone complete the tracking step as few as possible, set a distance negative reward. ,as follows:

[0121] (6)

[0122] in, is the magnitude tuning parameter for the distance reward, Indicates distance The nearest moving target, for The geometric distance to the nearest moving target, The maximum distance between two points on the map. As the distance between the drone and its nearest moving target decreases, it gradually approaches - On the contrary, the longer the distance (the lower the tracking success rate), the closer This reward essentially encourages the drone to track the nearest moving target as quickly as possible, thereby generating a better fast tracking strategy.

[0123] Obstacle avoidance reward: In order to allow the drone to complete tracking safely, set an obstacle avoidance reward ,as follows:

[0124] (7)

[0125] in:

[0126] (8)

[0127] (9)

[0128] in, , is the magnitude adjustment parameter for the obstacle avoidance reward, for The geometric distance to the nearest other drone. The smaller this value is, the greater the probability of collision with its companions. The bigger; for The geometric distance to the nearest environmental obstacle. The smaller this value is, the greater the probability of collision with the obstacle. The larger the value, the greater the reward. This reward essentially encourages the drone to stay away from other drones and obstacles to ensure the safety of the tracking process.

[0129] Based on the above discussion, the total reward function designed for the task is as follows. The optimization goal of MARL is to maximize the cumulative reinforcement learning reward function.

[0130] (10)

[0131] In this stage, by using MARL for a sufficient number of rounds of training, a policy neural network suitable for the UAV cluster multi-dynamic target tracking task can be obtained in the simplified task formed in the first stage. In the reasoning stage of the actual task execution, it is only necessary to input the real-time observations of each UAV into the trained policy network to execute the feasible strategy obtained through MARL training.

[0132] like Figure 4 The following is the overall implementation process of this solution.

[0133] In specific implementation, Figure 5 As shown, the number of training rounds and the reward curve obtained from the environment in each training round are displayed, and the upper limit of the reward is about 60. The baseline algorithm is MATD3, and the phased algorithm uses APF in the first stage and MATD3 in the second stage. The boundary parameter between the two stages is 1.1. It can be seen that the MSRL method converges after 5000 rounds of training, while the baseline MATD3 algorithm is close to convergence after 7000 rounds of training.

[0134] Based on the above scheme, by combining the advantages of traditional methods and learning methods, the performance is significantly better than the single APF method and the end-to-end MARL method. Under the same task scale, the algorithm can be improved and converged significantly while ensuring the tracking success rate. By simply adjusting the distance threshold parameter of the MSRL method proposed in the present invention, the number of algorithm convergence rounds can be reduced to about 72% of the end-to-end MARL method in the scenario where there is a target (such as Figure 5 ), and upon final convergence, achieves reinforcement learning rewards and tracking task success rates comparable to those of end-to-end MARL.

[0135] This application also provides an example of autonomous tracking of multiple dynamic targets in a drone cluster, such as Figure 6 As shown, the specific processing process includes the following steps:

[0136] Step S601, obtaining the cluster information of the drone cluster, the environment information of the drone cluster, and the behavior data information of the drone cluster.

[0137] Step S602, based on the environmental information, identifies the information of various restrictions in the environment where the drone cluster is located, and the initial target position information of the dynamic target cluster in the environment where the drone cluster is located, and based on the cluster information, identifies the initial drone position information of each drone in the drone cluster, and the cluster target position information of the cluster target of the dynamic target cluster.

[0138] Step S603, based on the information of each restriction object, construct a scene plane model of the drone cluster, and based on the initial target position information of the dynamic target cluster, the initial drone position information of each drone, and the cluster target position information of the cluster target of the dynamic target cluster, identify the drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster target in the scene plane model.

[0139] Step S604, adding the drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster target to the scene plane model to obtain the scene model of the drone cluster.

[0140] Step S605, for each drone in the drone cluster, based on the behavior data information of the drone, identify the speed information and the location information of the drone.

[0141] Step S606, based on the speed information of the drone, a motion model of the drone is constructed through a motion modeling strategy, and based on the position information of the drone, a position state progressive information of the drone is generated through a position state algorithm.

[0142] Step S607, the motion model of the drone and the position state progressive information of the drone are used as the drone behavior model of the drone, and all the drone behavior models are used as the cluster behavior model of the drone cluster.

[0143] Step S608, identifying the drone position distribution information of the drone cluster in the scene model, the target position distribution information of the dynamic target cluster in the scene model, and the restricted position information of each restricted object in the scene model.

[0144] Step S609, based on the drone position distribution information, the target position distribution information, and each restricted position information, the path planning information of the cluster behavior model of each drone is generated through the cluster path planning strategy.

[0145] Step S610, based on the path planning information, the target tracking process of the cluster behavior model is simulated through the scene model to obtain new drone position distribution information of the drone cluster and new target position distribution information of the dynamic target cluster.

[0146] Step S611, based on the new target position distribution information and the new drone position distribution information, the target center position information of the dynamic target cluster and the drone center position information of the drone cluster are identified through the center position division algorithm, and based on the target center position information and the drone center position information, the distance information between the drone cluster and each dynamic target cluster is calculated.

[0147] Step S612, when the distance information is lower than the preset distance threshold, it is determined that the distance information meets the preset distance condition, and based on the new drone position distribution information of each drone and the new target position distribution information of the dynamic target cluster, the scene model is updated to obtain a new scene model.

[0148] Step S613, based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, a new tracking route for each drone in the drone cluster is generated through a multi-dynamic target tracking algorithm.

[0149] Step S614, based on the new tracking routes of each drone, update the new scene model to obtain the current drone position distribution information of the drone cluster and the current target position distribution information of the dynamic target cluster, and when there is no current position information of the drone and it overlaps with the current target position distribution information of the dynamic target cluster, replace the new drone position distribution information of the drone cluster with the current drone position distribution information of the drone cluster, and replace the current target position distribution information of the dynamic target cluster with the current target position distribution information of the dynamic target cluster.

[0150] Step S615, return to step S614, until the current position information of the drone exists and overlaps with the current target position distribution information of the dynamic target cluster, the new tracking route obtained in each iteration is summarized in the generation time sequence of each new tracking route to obtain the dynamic tracking control strategy of the drone cluster.

[0151] Step S616, based on the dynamic tracking control strategy, control the drone cluster to track each dynamic target cluster, and complete the multi-dynamic target autonomous tracking task of the drone cluster.

[0152] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0153] Based on the same inventive concept, the embodiment of the present application also provides a drone cluster multi-dynamic target autonomous tracking device for implementing the drone cluster multi-dynamic target autonomous tracking method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more drone cluster multi-dynamic target autonomous tracking device embodiments provided below can refer to the limitations of the drone cluster multi-dynamic target autonomous tracking method above, and will not be repeated here.

[0154] In an exemplary embodiment, Figure 7 As shown, a multi-dynamic target autonomous tracking device for a drone cluster is provided, including: an acquisition module 710, a generation module 720, an identification module 730 and a control module 740, wherein:

[0155] An acquisition module 710 is used to acquire cluster information of a drone cluster, environmental information of the drone cluster, and behavior data information of the drone cluster, and to construct a scene model of the drone cluster based on the cluster information and the environmental information;

[0156] A generation module 720 is used to construct a cluster behavior model of the drone cluster based on the behavior data information of the drone cluster, and generate path planning information of the cluster behavior model through a cluster path planning strategy based on the scenario model;

[0157] The identification module 730 is used to identify the distance information between the drone cluster and each dynamic target cluster based on the path planning information and through the scene model, and generate a dynamic tracking control strategy for the drone cluster based on the scene model and through a multi-dynamic target tracking algorithm when the distance information meets a preset distance condition;

[0158] The control module 740 is used to control the drone cluster based on the dynamic tracking control strategy, track each of the dynamic target clusters, and complete the multi-dynamic target autonomous tracking task of the drone cluster.

[0159] Optionally, the acquisition module 710 is specifically configured to:

[0160] Based on the environmental information, identify information about each restriction in the environment in which the drone cluster is located, and initial target position information of a dynamic target cluster in the environment in which the drone cluster is located, and based on the cluster information, identify initial drone position information of each drone in the drone cluster, and cluster target position information of cluster targets of the dynamic target cluster;

[0161] Based on the information of each of the restricting objects, a scene plane model of the drone cluster is constructed, and based on the initial target position information of the dynamic target cluster, the initial drone position information of each of the drones, and the cluster target position information of the cluster targets of the dynamic target cluster, drone position parameters of the drone cluster, target position parameters of the dynamic target cluster, and cluster target position parameters of the cluster targets are identified in the scene plane model;

[0162] The drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster targets are added to the scene plane model to obtain the scene model of the drone cluster.

[0163] Optionally, the generating module 720 is specifically configured to:

[0164] For each drone of the drone cluster, based on the behavior data information of the drone, identifying the speed information of the drone and the location information of the drone;

[0165] Based on the speed information of the drone, a motion model of the drone is constructed through a motion model modeling strategy, and based on the position information of the drone, position state progressive information of the drone is generated through a position state algorithm;

[0166] The motion model of the drone and the progressive information of the position state of the drone are used as the drone behavior model of the drone, and all the drone behavior models are used as the cluster behavior model of the drone cluster.

[0167] Optionally, the generating module 720 is specifically configured to:

[0168] Identify the drone position distribution information of the drone cluster in the scene model, the target position distribution information of the dynamic target cluster in the scene model, and the restricted position information of each restricted object in the scene model;

[0169] Based on the drone position distribution information, the target position distribution information, and each of the restricted position information, a cluster path planning strategy is used to generate path planning information of the cluster behavior model of each drone.

[0170] Optionally, the identification module 730 is specifically configured to:

[0171] Based on the path planning information, the target tracking process of the cluster behavior model is simulated through the scenario model to obtain new drone position distribution information of the drone cluster and new target position distribution information of the dynamic target cluster;

[0172] Based on the new target position distribution information and the new drone position distribution information, the target center position information of the dynamic target cluster and the drone center position information of the drone cluster are identified through the center position division algorithm, and based on the target center position information and the drone center position information, the distance information between the drone cluster and each dynamic target cluster is calculated.

[0173] Optionally, the identification module 730 is specifically configured to:

[0174] When the distance information is lower than a preset distance threshold, determining that the distance information satisfies a preset distance condition, and updating the scene model based on the new drone position distribution information of each of the drones and the new target position distribution information of the dynamic target cluster to obtain a new scene model;

[0175] Based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, a new tracking route for each drone of the drone cluster is generated through a multi-dynamic target tracking algorithm;

[0176] Based on the new tracking routes of each of the drones, the new scene model is updated to obtain the current drone position distribution information of the drone cluster and the current target position distribution information of the dynamic target cluster, and when there is no current drone position information that overlaps with the current target position distribution information of the dynamic target cluster, the current drone position distribution information of the drone cluster is replaced by the new drone position distribution information of the drone cluster, and the current target position distribution information of the dynamic target cluster is replaced by the current target position distribution information of the dynamic target cluster;

[0177] Return to the step of generating a new tracking route for each drone of the drone cluster based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target through a multi-dynamic target tracking algorithm, until the current position information of the drone exists and overlaps with the current target position distribution information of the dynamic target cluster. Then, the new tracking route obtained in each iteration is summarized in the order of the generation time of each new tracking route to obtain the dynamic tracking control strategy of the drone cluster.

[0178] Each module in the above-mentioned UAV swarm multi-dynamic target autonomous tracking device can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0179] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for autonomous tracking of multiple dynamic targets of a drone cluster is realized. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0180] Those skilled in the art will understand that Figure 8The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0181] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of any one of the methods in the first aspect are implemented.

[0182] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods in the first aspect are implemented.

[0183] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of any one of the methods in the first aspect.

[0184] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0185] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0186] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0187] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for autonomous tracking of multiple dynamic targets in a drone cluster, characterized in that: The method comprises: Acquire cluster information of a drone cluster, environmental information of the drone cluster, and behavior data information of the drone cluster, and construct a scene model of the drone cluster based on the cluster information and the environmental information; Based on the behavior data information of the drone cluster, a cluster behavior model of the drone cluster is constructed, and based on the scenario model, path planning information of the cluster behavior model is generated through a cluster path planning strategy; Based on the path planning information, the distance information between the drone cluster and each dynamic target cluster is identified through the scene model, and when the distance information meets the preset distance condition, a dynamic tracking control strategy of the drone cluster is generated based on the scene model through a multi-dynamic target tracking algorithm; Based on the dynamic tracking control strategy, the drone cluster is controlled to track each of the dynamic target clusters to complete the multi-dynamic target autonomous tracking task of the drone cluster.

2. The method according to claim 1, characterized in that The step of constructing a scene model of the drone cluster based on the cluster information and the environmental information includes: Based on the environmental information, identify information about each restriction in the environment in which the drone cluster is located, and initial target position information of a dynamic target cluster in the environment in which the drone cluster is located, and based on the cluster information, identify initial drone position information of each drone in the drone cluster, and cluster target position information of cluster targets of the dynamic target cluster; Based on the information of each of the restricting objects, a scene plane model of the drone cluster is constructed, and based on the initial target position information of the dynamic target cluster, the initial drone position information of each of the drones, and the cluster target position information of the cluster targets of the dynamic target cluster, drone position parameters of the drone cluster, target position parameters of the dynamic target cluster, and cluster target position parameters of the cluster targets are identified in the scene plane model; The drone position parameters of the drone cluster, the target position parameters of the dynamic target cluster, and the cluster target position parameters of the cluster targets are added to the scene plane model to obtain the scene model of the drone cluster.

3. The method according to claim 1, characterized in that The step of constructing a cluster behavior model of the drone cluster based on the behavior data information of the drone cluster includes: For each drone of the drone cluster, based on the behavior data information of the drone, identifying the speed information of the drone and the location information of the drone; Based on the speed information of the drone, a motion model of the drone is constructed through a motion model modeling strategy, and based on the position information of the drone, position state progressive information of the drone is generated through a position state algorithm; The motion model of the drone and the progressive information of the position state of the drone are used as the drone behavior model of the drone, and all the drone behavior models are used as the cluster behavior model of the drone cluster.

4. The method according to claim 2, characterized in that: The generating path planning information of the cluster behavior model based on the scenario model through a cluster path planning strategy includes: Identify the drone position distribution information of the drone cluster in the scene model, the target position distribution information of the dynamic target cluster in the scene model, and the restricted position information of each restricted object in the scene model; Based on the drone position distribution information, the target position distribution information, and each of the restricted position information, a cluster path planning strategy is used to generate path planning information of the cluster behavior model of each drone.

5. The method according to claim 4, characterized in that The identifying, based on the path planning information and through the scene model, distance information between the drone cluster and each dynamic target cluster includes: Based on the path planning information, the target tracking process of the cluster behavior model is simulated through the scenario model to obtain new drone position distribution information of the drone cluster and new target position distribution information of the dynamic target cluster; Based on the new target position distribution information and the new drone position distribution information, the target center position information of the dynamic target cluster and the drone center position information of the drone cluster are identified through the center position division algorithm, and based on the target center position information and the drone center position information, the distance information between the drone cluster and each dynamic target cluster is calculated.

6. The method according to claim 2, characterized in that When the distance information meets the preset distance condition, based on the scene model, a dynamic tracking control strategy of the drone cluster is generated through a multi-dynamic target tracking algorithm, including: When the distance information is lower than a preset distance threshold, determining that the distance information satisfies a preset distance condition, and updating the scene model based on the new drone position distribution information of each of the drones and the new target position distribution information of the dynamic target cluster to obtain a new scene model; Based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target, a new tracking route for each drone of the drone cluster is generated through a multi-dynamic target tracking algorithm; Based on the new tracking routes of each of the drones, the new scene model is updated to obtain the current drone position distribution information of the drone cluster and the current target position distribution information of the dynamic target cluster, and when there is no current drone position information that overlaps with the current target position distribution information of the dynamic target cluster, the current drone position distribution information of the drone cluster is replaced by the new drone position distribution information of the drone cluster, and the current target position distribution information of the dynamic target cluster is replaced by the current target position distribution information of the dynamic target cluster; Return to the step of generating a new tracking route for each drone of the drone cluster based on the new drone position distribution information of the drone cluster, the new target position distribution information of the dynamic target cluster, and the cluster target position information of the cluster target through a multi-dynamic target tracking algorithm, until the current position information of the drone exists and overlaps with the current target position distribution information of the dynamic target cluster. Then, the new tracking route obtained in each iteration is summarized in the order of the generation time of each new tracking route to obtain the dynamic tracking control strategy of the drone cluster.

7. An autonomous tracking device for multiple dynamic targets in a drone swarm, characterized in that: The device comprises: An acquisition module, used to acquire cluster information of a drone cluster, environmental information of the drone cluster, and behavioral data information of the drone cluster, and to construct a scene model of the drone cluster based on the cluster information and the environmental information; A generation module, configured to construct a cluster behavior model of the drone cluster based on the behavior data information of the drone cluster, and generate path planning information of the cluster behavior model through a cluster path planning strategy based on the scenario model; An identification module, configured to identify the distance information between the drone cluster and each dynamic target cluster based on the path planning information and through the scene model, and generate a dynamic tracking control strategy for the drone cluster based on the scene model and through a multi-dynamic target tracking algorithm when the distance information meets a preset distance condition; The control module is used to control the drone cluster based on the dynamic tracking control strategy, track each dynamic target cluster, and complete the multi-dynamic target autonomous tracking task of the drone cluster.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images