Aerial maneuvering cluster target aggregation display method

Abstract

The invention discloses an aggregation display method for air maneuvering cluster targets, and belongs to the technical field of artificial intelligence. In the method, a low-altitude moving target control end and a central processing device respectively run a client and a service navigation program and respectively display through a client screen interface and a service screen interface, and the client screen interface and the service screen interface respectively comprise a three-dimensional coordinate system established by taking a map as an XY plane and a digital navigation area; an operator inputs the initial position, the target position and the model of the low-altitude moving target through the input assembly according to an instruction of a client screen interface and sends the initial position, the target position and the model of the low-altitude moving target to the central processing device; the central processing device processes information sent by a plurality of low-altitude moving targets to generate navigation functions respectively, grouping is carried out according to the navigation functions, the digital twin bodies are generated by taking the groups as units, and the navigation functions of the digital twin bodies of the low-altitude moving targets in the groups are further optimized through reinforcement learning. According to the invention, a visual four-dimensional navigation track can be provided.

Description

Technical Field

[0001] This invention provides a method for aggregated display of aerial maneuvering swarm targets, belonging to the field of reinforcement learning technology. Background Technology

[0002] Chinese invention patent application CN120725566A discloses a joint optimization method for UAV swarm path planning and communication resource allocation for logistics delivery. The method includes: constructing a scenario model for UAV swarm logistics delivery, including a UAV swarm delivery model and a UAV swarm communication model; performing cluster analysis on ground customers (target points) in the UAV swarm delivery model, dividing the digital twin points into Kclus clusters, ensuring that the digital twin points within the same cluster are relatively concentrated geographically, while density constraints ensure a balanced distribution of the number of digital twin points in each cluster, thus optimizing the initial logistics delivery task; using simulated annealing algorithm for UAV swarm path planning to obtain the optimal flight path of the UAV swarm; and based on the position of the UAVs determined by the optimal flight path of the UAV swarm in each future time slot t, using a competitive architecture deep Q-network to perform distributed communication resource allocation for the UAV swarm during dynamic flight.

[0003] There is currently no publicly available technical solution for visualizing four-dimensional navigation, nor is there a publicly available technical solution for optimizing navigation parameters for low-altitude moving targets through reinforcement learning. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention provides a method for aggregated display of aerial maneuvering swarm targets, which can provide visualized four-dimensional navigation and optimize the navigation parameters of low-altitude moving targets through reinforcement learning without the need for pre-setting.

[0005] To achieve the aforementioned objective, this invention provides a method for aggregated display of aerial maneuvering swarm targets, comprising: The low-altitude moving target control terminal and the central processing unit are connected via a network. They can run client navigation programs and service navigation programs respectively, and are displayed on client screen interfaces and service screen interfaces respectively. Both the client screen interface and the service screen interface include a three-dimensional coordinate system established with the map as the XY plane, and a digital navigation area established on the three-dimensional coordinate system. The digital navigation area is a simulation of the real three-dimensional navigation area. The operator inputs the starting position, destination position, and model of the low-altitude moving target through the input component according to the instructions on the client screen interface and sends it to the central processing unit. The central processing unit generates navigation functions based on information sent from multiple low-altitude moving targets, groups the targets according to the navigation functions, and generates digital twins of the low-altitude moving targets in groups. The navigation functions of the digital twins of the low-altitude moving targets in the groups are further optimized through reinforcement learning.

[0006] Compared with existing technologies, the aggregation display method for aerial maneuvering swarm targets provided by this invention has the following advantages: it can provide visualized four-dimensional navigation and optimize the navigation parameters of low-altitude moving targets through reinforcement learning, without the need for prior setting. Attached Figure Description

[0007] Figure 1 This is a block diagram of the aggregated display system for aerial maneuvering cluster targets provided in the first embodiment of the present invention.

[0008] Figure 2A This is the first auxiliary function curve of the pipe boundary repulsion function provided in the first embodiment of the present invention.

[0009] Figure 2B This is the second auxiliary function curve of the pipe boundary repulsion function provided in the first embodiment of the present invention.

[0010] Figure 3A This is the first auxiliary function curve of the collision avoidance function provided in the first embodiment of the present invention.

[0011] Figure 3B This is the second auxiliary function curve of the collision avoidance function provided in the first embodiment of the present invention.

[0012] Figure 4A This is the first auxiliary function curve of the group cohesion attraction function provided in the first embodiment of the present invention.

[0013] Figure 4B This is the second auxiliary function curve of the group cohesion attraction function provided in the first embodiment of the present invention. Detailed Implementation

[0014] It should be noted that, below, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The advantages and features of the present invention, as well as the methods for achieving these advantages and features, will become clear from the accompanying drawings and the detailed embodiments described below.

[0015] However, the present invention is not limited to the embodiments disclosed below, and can be implemented in many different forms. This embodiment is only used to make the disclosure of the present invention more complete and to fully inform those skilled in the art of the present invention of the scope of the invention. The present invention is defined only by the scope of the claims.

[0016] While terms such as "first," "second," etc., are used to describe various elements, components, and / or parts, these elements, components, and / or parts are not limited by these terms. These terms are used only to distinguish one element, component, or part from other elements, components, or parts. Therefore, it is apparent that, within the technical spirit of this disclosure, the first element, first component, or first part mentioned below may also be a second element, second component, or second part, and the terminology used in this specification is for describing embodiments only and is not intended to limit this disclosure.

[0017] In this specification, unless otherwise specified in the text, the singular includes the plural. The use of "comprising" and / or "consisting of" in this specification does not exclude the presence or addition of one or more other structural elements, steps, actions, and / or components mentioned.

[0018] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Furthermore, terms defined in commonly used dictionaries shall not be interpreted ideally or excessively unless explicitly and specifically defined.

[0019] First Embodiment

[0020] Figure 1 This is a block diagram of the aggregated display system for aerial maneuvering cluster targets provided in the first embodiment of the present invention. Figure 1As shown, the first embodiment of the present invention provides an aggregated display system for aerial maneuvering cluster targets, comprising: a low-altitude moving target control terminal and a central processing unit connected via a network; the control terminal includes a first display module, an input component, a first processor, and a first communication unit; the first processor calls a client navigation program and displays it on the first display module in the form of a client screen interface; the client screen interface includes a three-dimensional coordinate system established with a map as the XY plane, and a digital navigation area established on the three-dimensional coordinate system, wherein the digital navigation area is a simulation of a real three-dimensional navigation area; the operator inputs the starting position, destination position, and static parameters of the low-altitude moving target through the input component according to the instructions of the client screen interface and sends them to the central processing unit through the first communication unit; the central processing unit includes a second display module, a second processor, and a second... The system includes a communication unit and an input unit. The second processor invokes a service navigation program and displays it on a second display module as a service screen interface. The service screen interface includes a three-dimensional coordinate system established with the map as the XY plane, and a digital navigation area established on this three-dimensional coordinate system. This digital navigation area is a simulation of a real three-dimensional navigation area. The second processor processes information received from multiple low-altitude moving targets via the second communication unit, generates navigation functions for each target based on this information, groups the targets according to the navigation functions, and generates digital twins for each group of low-altitude moving targets. The navigation functions of the digital twins of the low-altitude moving targets in the group are further optimized using reinforcement learning. The optimized navigation functions are then sent to the control terminal and flight control device of the low-altitude moving targets via the second communication unit. The input unit is used to input commands and data. Each low-altitude moving target control terminal generates a digital twin or icon of the low-altitude moving target based on the optimized navigation function and the target's static parameters, which is then superimposed on the digital navigation area. Each low-altitude moving target flight control device includes a flight controller and a third communication unit. The flight controller receives optimized navigation functions from the central control unit and commands from the control terminal of the low-altitude moving target via the third communication unit. The flight controller drives the low-altitude moving target to fly in a real three-dimensional navigation area according to the optimized navigation functions. In the first embodiment, the central processing unit can be a cloud server. In the first embodiment, the central processing unit, the control terminal of the low-altitude moving target, and the flight control device all include a storage module, which is used to store programs and data.

[0021] In the first embodiment, a low-altitude moving target generally refers to an object with motion characteristics that flies in low-altitude airspace (such as below 100 meters), such as a drone, an airship, or a small aircraft.

[0022] In the first embodiment, grouping based on information sent from multiple low-altitude moving targets includes: S01: Based on the navigation function of the low-altitude moving target, the low-altitude moving target is clustered into multiple groups to obtain a group set; S02: Calculate the maneuvering low-altitude moving target u that requires navigation. NEW The similarity between the navigation function of the given group and the navigation function of each group in the group set; S03: If the similarity is less than the threshold, create a new group and add the new group to the group set; if the similarity is greater than or equal to the threshold, then the maneuvering low-altitude moving target u... NEW Place it in the group with the highest similarity.

[0023] In the first embodiment, digital twins are generated from low-altitude moving targets that are generated in groups.

[0024] In the first embodiment, each group CL k The control strategy for the digital twin of a low-altitude moving target is as follows: ,in, , These refer to the guiding curve function, the tube boundary repulsion function, the collision avoidance function, and the group cohesion attraction function, respectively.

[0025] In the first embodiment, , In the formula, Represents the constraint function. Indicates low-altitude moving targets Maximum permissible speed; Indicates low-altitude moving targets Location; Indicates coefficient; For low-altitude moving targets At the position of the guide curve The unit tangent vector at that point; For low-altitude moving targets Speed ​​command; Indicates the time period Including location The arc length of the guiding curve.

[0026] In the first embodiment, , In the formula, and Simulations were performed using the first and second largest language models. The radius of the virtual tube; For drones To the vertical projection point of the virtual tube centered on the guide curve; , The coefficients are used. The large language model is, for example, the Transformer model.

[0027] The training process of the first major language model includes: from Figure 2A Obtain a series of data sets of independent and dependent variables from the curve. ,Will The input is fed into the embedding layer of the first large language model. Specifically, the embedding layer of the first large language model handles the independent variables, such as... Encode each element in the string, such as encoding each element in the string. "", "", " "", "、"-" "", " "", "", "Encoding is performed; the positional encoding layer of the first large language model encodes the position of each element and then inserts it before and after the element's encoding. The encoding is to encode the element into binary code, and the positional encoding is the encoding of the element's position in the independent variable. The positional encoding also includes the encoding of the subscript element; the fully connected layer at the output of the large language model outputs the estimated value of the dependent variable:" , The first language model parameter vector to be optimized, optionally; according to and Calculate the loss function Determine the loss function Is it the smallest? If it is the smallest, then from... Figure 2A Take one or more sets of data from the curve for verification. If the verification is successful, output the prioritized parameter vector. Otherwise, based on the loss By updating the parameter vector Each element in, and then from Figure 2A Take a data set from the curve and continue training the first language model.

[0028] In the first embodiment of the present invention, if a large language model cannot completely simulate... Figure 2A When calculating the curve, multiple large language simulations can be used to simulate the segmented process.

[0029] The training process of the second major language model includes: from Figure 2B Obtain a series of data sets of independent and dependent variables from the curve. ,Will The input is fed into the embedding layer of the second largest language model, which then processes the dependent variable " Encode each element in the string, for example, for " "", "", "、 "", "、 "-" "", “、 “,” "", "", "", The symbols “” and “s” are entered into the encoding; the positional encoding layer of the second language model encodes the position of each element and then inserts it before and after the element encoding. The encoding is to encode the elements into binary codes, and the positional encoding is the encoding of the position of the element in the independent variable. The positional encoding also includes the encoding of subscript elements and their positional encoding; the fully connected layer at the output of the second language model outputs the estimated value of the dependent variable: , The second largest language model parameter vector to be optimized; according to and Calculate the loss function Determine the loss function Is it the smallest? If it is the smallest, then from... Figure 2B Take one or more sets of data from the curve for verification. If the verification is successful, output the prioritized parameter vector. Otherwise, based on the loss By updating the parameter vector Each element in, and then from Figure 2B We take a data set from the curve and continue training the second language model.

[0030] In the first embodiment of the present invention, if a large language model cannot completely simulate... Figure 2B When calculating the curve, multiple large language simulations can be used to simulate the segmented process.

[0031] In the first embodiment of the present invention, , In the formula, and Simulations were performed using the third and fourth largest language models. The radius of the virtual tube; The location of the obstacle; , is a coefficient.

[0032] The training process of the third major language model includes: from Figure 3A Obtain a series of data sets of independent and dependent variables from the curve. ,Will The input is fed into the input layer of the third language model. Specifically, the embedding layer of the third language model processes the dependent variable... Encode each element in the string, for example, for " "", "", " "", ", ", " "", “、 “,” "", "The third language model's positional encoding layer encodes the positions of each element and inserts them before and after the element's encoding. This encoding involves converting the elements into binary codes. The positional encoding is the encoding of the element's position within the independent variable, and it also includes encoding the subscript element. The full transition layer at the output of the third language model outputs an estimated value for the dependent variable:" , The parameter vector to be optimized for the third largest language model; according to and Calculate the loss function Determine the loss function Is it the smallest? If it is the smallest, then from... Figure 3A Take one or more sets of data from the curve for verification. If the verification is successful, output the prioritized parameter vector. Otherwise, based on the loss By updating the parameter vector Each element in, and then from Figure 3A Take a data set from the curve and continue training the third language model.

[0033] In the first embodiment of the present invention, if a large language model cannot completely simulate... Figure 3A When calculating the curve, multiple large language simulations can be used to simulate the segmented process.

[0034] The training process of the fourth major language model includes: from Figure 3B Obtain a series of data sets of independent and dependent variables from the curve. ,Will The input is fed into the embedding layer of the fourth language model, which then processes the dependent variable. Encode each element in the string, for example, for " "", "", ", " "、 "-" "", “、 “,” "", "", “2” "", The symbols “” and “s” are entered into the encoding; the positional encoding layer of the fourth language model encodes the position of each element and then inserts it before and after the element encoding. The encoding is to encode the elements into binary code, and the positional encoding is the encoding of the position of the element in the independent variable. The positional encoding also includes the encoding of subscript elements and their positional encoding; the fully connected layer at the output of the fourth language model outputs the estimated value of the dependent variable: , The fourth language model parameter vector to be optimized; according to and Calculate the loss function Determine the loss function Is it the smallest? If it is the smallest, then from... Figure 3B Take one or more sets of data from the curve for verification. If the verification is successful, output the prioritized parameter vector. Otherwise, based on the loss By updating the parameter vector Each element in, and then from Figure 3B We take a data set from the curve and continue training the fourth language model.

[0035] In the first embodiment of the present invention, if a large language model cannot completely simulate... Figure 3B When calculating the curve, multiple large language simulations can be used to simulate the segmented process.

[0036] In the first embodiment of the present invention, the cluster navigation function includes a cluster cohesion attraction function: , In the formula, and Simulation using a large language model; In order to deal with low-altitude moving targets Low-altitude moving targets belonging to the same group Location; , is a coefficient.

[0037] The training process of the fifth language model includes: from Figure 3A Obtain a series of data sets of independent and dependent variables from the curve. ,Will The input is fed into the input layer of the fifth language model, and the embedding layer of the fifth language model processes the dependent variable. Encode each element in the string, for example, for " "", "", " "", ", ", " "", " "", "", "The fifth language model's positional encoding layer encodes the positions of each element and inserts them before and after the element's encoding. This encoding involves converting the elements into binary codes. The positional encoding is the encoding of the element's position within the independent variable, and it also includes encoding the subscript element. The full transition layer at the output of the fifth language model outputs an estimated value for the dependent variable:" , The fifth language model parameter vector to be optimized; according to and Calculate the loss function Determine the loss function Is it the smallest? If it is the smallest, then from... Figure 4A Take one or more sets of data from the curve for verification. If the verification is successful, output the prioritized parameter vector. Otherwise, according to the loss function By updating the parameter vector Each element in, and then from Figure 4A Take a data set from the curve and continue training the fifth language model.

[0038] In the first embodiment of the present invention, if a large language model cannot completely simulate... Figure 4A When calculating the curve, multiple large language simulations can be used to simulate the segmented process.

[0039] The training process of the sixth language model includes: from Figure 3B Obtain a series of data sets of independent and dependent variables from the curve. ,Will The input is fed into the embedding layer of the sixth language model, which then processes the dependent variable. Encode each element in the string, for example, for " "", "", ", " "、 "-" "", "、 "", "", "", "", The symbols “” and “s” are entered into the encoding; the positional encoding layer of the sixth language model encodes the position of each element and then inserts it before and after the element encoding. The encoding is to encode the elements into binary codes, and the positional encoding is the encoding of the position of the element in the independent variable. The positional encoding also includes the encoding of subscript elements and their positional encoding; the fully connected layer at the output of the sixth language model outputs the estimated value of the dependent variable: , The sixth language model parameter vector to be optimized; according to and Calculate the loss function Determine the loss function Is it the smallest? If it is the smallest, then from... Figure 4B Take one or more sets of data from the curve for verification. If the verification is successful, output the prioritized parameter vector. Otherwise, based on the loss By updating the parameter vector Each element in, and then from Figure 4B Take a data set from the curve and continue training the sixth language model.

[0040] In the first embodiment of the present invention, if a large language model cannot completely simulate... Figure 4B When calculating the curve, multiple large language simulations can be used to simulate the segmented process.

[0041] In the first embodiment of the present invention, the parameters of the navigation function of a low-altitude moving target are optimized through reinforcement learning. The specific process includes: S1-01 Low-altitude moving target The digital twin's flight status at current time t At that time, based on low-altitude moving targets Digital twin flight strategy Obtain the navigation function: The flight state at the next time t+1 is obtained by performing flight according to the navigation function. k=1,2,…,K For low-altitude moving targets The current parameter vector of the navigation function of the digital twin; S1-02: Assess Flight Status And award flight prizes ; The current flight status of digital twins of K low-altitude moving targets. Flight rewards and the next flight status : , , K is a positive integer greater than or equal to 2; S1-03: Repeat the above process to obtain T data sets, and calculate the approximate regression value according to the following formula:

[0042] S1-04: Approximate regression values The broadcast is sent to digital twins of all low-flying moving targets in the same group. It is the discount factor; It is a state-value function; These are the parameters of the state value function; S1-05: Low-altitude moving targets The digital twin is based on the received approximate regression values. Update flight strategy Network parameters : , For learning rate, To The gradient function.

[0043] Specifically, , , , , , , , , In the formula, , , , , , , , , This is for adjusting the coefficient.

[0044] Repeat the above process until the set number of training iterations or termination conditions are met.

[0045] After the central processing unit optimizes the navigation function based on the digital twin of the low-altitude moving target, it sends the optimized navigation function to the control terminal and flight control device of the low-altitude moving target. The flight controller of the low-altitude moving target's flight control device drives the low-altitude moving target to fly according to the optimized navigation function. The control terminal of the low-altitude moving target generates dynamic digital icons based on the optimized navigation function. The digital icons are dynamically superimposed on a three-dimensional coordinate system established with the map as the XY plane. K flight icons can be reduced to display as a group icon, or they can be enlarged to display the icon of each low-altitude moving target in the group. When the operator clicks on the icon, the static parameters of each low-altitude moving target can be displayed, such as volume, shape, type, etc.; the dynamic parameters of each low-altitude moving target can also be displayed, such as flight speed, acceleration, position, attitude, etc.

[0046] Optionally, the control terminal of the low-altitude moving target generates a dynamic digital twin based on the optimized navigation function and the parameters of the low-altitude moving target. The digital twin is dynamically superimposed on a three-dimensional coordinate system established with the map as the XY plane. K digital twins can be scaled down to be displayed as a group icon, or they can be scaled up to display the digital twin of each low-altitude moving target in the group. The operator can display the digital twin of the low-altitude moving target at any angle as needed. In this way, the operator can understand the static and dynamic parameters of any low-altitude moving target in the same group.

[0047] Optionally, the control terminal of a low-altitude moving target displays only its own digital twin, while other low-altitude moving targets in the group are displayed as icons.

[0048] Compared with existing technologies, this invention can provide visualized four-dimensional navigation and optimizes the navigation parameters of low-altitude moving targets through reinforcement learning, without the need for pre-setting.

[0049] Second Embodiment

[0050] The second embodiment of the present invention only describes the contents that are different from those of the first embodiment; the contents that are the same will not be described again.

[0051] In the second embodiment of the present invention, the parameters of the navigation function of a low-altitude moving target are optimized through reinforcement learning. The specific process includes: S2-01 Low-altitude moving target The digital twin's current flight state at time t is At that time, based on low-altitude moving targets Digital twin flight strategy Obtain the navigation function: The flight state at the next time t+1 is obtained by performing flight according to the navigation function. k=1,2,…,K; For low-altitude moving targets The current parameter vector of the navigation function of the digital twin; S2-02: Assessing Low-Altitude Moving Targets The flight status of the digital twin And award flight prizes ; Summarize the current flight status of K low-altitude moving target digital twins. Flight rewards and the next flight status : , , K is a positive integer greater than or equal to 2; S2-03: The state is calculated according to the following formula. State value function value at time: , In the formula, for Action strategy, It is a low-altitude moving target The state of the digital twin at the current time t. It is a low-altitude moving target The actions of the digital twin at the current time t; For low-altitude moving targets The digital twin is performing an action at the current time t. From state Transition to state The probability of; For low-altitude moving targets The digital twin is performing an action at the current time t. From state Transition to state The reward received; The state is State value function; Indicates low-altitude moving targets All actions of the digital twin; This is the discount factor; Indicates low-altitude moving targets All states of the digital twin; S2-04: Judgment Is it less than or equal to the threshold? If yes, proceed to step S1-08; otherwise, proceed to step S1-05. S2-05: Calculate the state-action function according to the following formula: , S2-06: Improve the operating strategy according to the following formula .

[0052] This invention optimizes the navigation parameters of low-altitude moving targets through reinforcement learning via the above steps. It does not require prior setting, and the low-altitude moving targets execute flight through the optimized navigation function of the digital twin.

[0053] Third Embodiment

[0054] The third embodiment of the present invention only describes the content that is different from the first embodiment; the same content will not be described again.

[0055] The third embodiment of the present invention provides a computer program product, which includes computer program code that can be called by a processor to execute the methods of the first and second embodiments.

[0056] Fourth embodiment

[0057] The fourth embodiment of the present invention only describes the content that is different from the fourth embodiment; the same content will not be described again.

[0058] The fourth embodiment of the present invention provides a storage medium for storing computer program code from the third embodiment.

[0059] Additionally, it should be noted that the central control unit, the low-altitude moving target control terminal, and the low-altitude moving target flight control unit all include storage devices, which include at least ROM and RAM to store navigation programs and data.

[0060] The preferred embodiments of the present invention disclosed herein are merely for the purpose of illustrating the present invention. The preferred embodiments do not describe all the details exhaustively, nor do they limit the invention to specific implementation methods. Obviously, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for aggregated display of aerial maneuvering cluster targets, characterized in that, include: The low-altitude moving target control terminal and the central processing unit are connected via a network. They run client navigation programs and service navigation programs respectively, and are displayed on client screen interfaces and service screen interfaces respectively. Both the client screen interface and the service screen interface include a three-dimensional coordinate system established with the map as the XY plane. A digital navigation area is established on the three-dimensional coordinate system. The digital navigation area is a simulation of the real three-dimensional navigation area. The operator inputs the starting position, destination position, and model of the low-altitude moving target through the input component according to the instructions on the client screen interface and sends it to the central processing unit. The central processing unit generates navigation functions based on information sent from multiple low-altitude moving target control terminals, groups the targets according to the navigation functions, and generates digital twins of low-altitude moving targets in groups. The navigation functions of the digital twins of low-altitude moving targets in the groups are further optimized through reinforcement learning.

2. The method for aggregated display of aerial maneuvering cluster targets according to claim 1, characterized in that, Grouping based on navigation functions includes: S01: Based on the navigation function of the low-altitude moving target, the low-altitude moving target is clustered into multiple groups to obtain a group set; S02: Calculate the maneuvering low-altitude moving target u that requires navigation. NEW The similarity between the navigation function of the given group and the navigation function of each group in the group set; S03: If the similarity is less than the threshold, create a new group and add the new group to the group set; if the similarity is greater than or equal to the threshold, then the maneuvering low-altitude moving target u... NEW Place it in the group with the highest similarity.

3. The method for aggregated display of aerial maneuvering cluster targets according to claim 2, characterized in that, Low-altitude moving targets in the kth group The control strategy for the digital twin is as follows: ,in, , These refer to the guiding curve function, the tube boundary repulsion function, the collision avoidance function, and the group cohesion attraction function, respectively.

4. The method for aggregated display of aerial maneuvering cluster targets according to claim 3, characterized in that, ， In the formula, Represents the constraint function. Indicates low-altitude moving targets Maximum permissible speed; Indicates low-altitude moving targets Location; Indicates coefficient; For low-altitude moving targets At the position of the guide curve The unit tangent vector at that point; For low-altitude moving targets Speed ​​command; Indicates the time period Including location The arc length of the guiding curve.

5. The method for aggregated display of aerial maneuvering cluster targets according to claim 4, characterized in that, ， In the formula, and The simulations will be conducted using the first and second largest language models, respectively. The radius of the virtual tube; For drones To the vertical projection point of the virtual tube centered on the guide curve; , is a coefficient.

6. The method for aggregated display of aerial maneuvering cluster targets according to claim 5, characterized in that, , In the formula, and Simulations were performed using the third and fourth language models, respectively. The radius of the virtual tube; The location of the obstacle; , is a coefficient.

7. The method for aggregated display of aerial maneuvering cluster targets according to claim 3, characterized in that, Cluster navigation functions all include a group cohesion attraction function: , In the formula, and Simulations were performed using the fifth and sixth language models. In order to deal with low-altitude moving targets Low-altitude moving targets belonging to the same group Location; , is a coefficient.

8. The method for aggregated display of aerial maneuvering cluster targets according to any one of claims 4-7, characterized in that, Low-altitude moving targets The digital twin's current flight state at time t is At that time, based on low-altitude moving targets Digital twin flight strategy Obtain the navigation function: The flight state at the next time t+1 is obtained by performing flight according to the navigation function. k=1,2,…,K For low-altitude moving targets The current parameter vector of the navigation function of the digital twin; Assess flight status And award flight prizes ; Summarize the current flight status of K low-altitude moving targets. Flight rewards and the next flight status K is a positive integer greater than or equal to 2. The first approximate regression value is calculated according to the following formula: ， pseudo-regression value The broadcast is sent to digital twins of all low-flying moving targets in the same group. It is the discount factor; It is a state-value function; These are the parameters of the state value function; Low-altitude moving targets The digital twin is based on the received approximate regression values. Update flight strategy Network parameters for .

9. The method for aggregated display of aerial maneuvering cluster targets according to claim 8, characterized in that, ， This is the learning rate.

10. The method for aggregated display of aerial maneuvering cluster targets according to claim 9, characterized in that, Digital twins of K low-altitude moving targets in the same group generate flight tracks based on T flight actions and flight states, and superimpose them on a three-dimensional coordinate system with the map as the XY plane. The K flight tracks can be scaled down to be displayed as a group icon, or they can be scaled up to display the track of each low-altitude moving target in the group.

Images