DCBGA-based Heterogeneous UAV Cluster ACK Construction Method, System and Device

Abstract

The present application relates to a method, system, and device for constructing an ACK for a heterogeneous UAV cluster based on DCBGA. The method includes: dividing the overall task process in a dynamic adversarial environment into discrete time series and determining each time step; in the current time step, each UAV in the heterogeneous UAV cluster uses the constructed DCBGA model for target selection; in the current time step, each UAV uses the DCBGA model for a consensus process; when the number of rounds completed in the target selection and consensus process in the current time step reaches a set number of rounds, each UAV completes the update of its respective local information, switches its own task execution status and motion state, and enters the next time step; following the target selection and consensus process of each time step, the constructed ACK network of the heterogeneous UAV cluster is output and updated. Significantly improves the efficiency of task network construction for heterogeneous UAV clusters in large-scale dynamic environments.

Description

Technical Field

[0001] The present invention belongs to the technical field of UAV swarm control, and relates to a method, a system and a device for constructing an ACK for a heterogeneous UAV swarm based on DCBGA. Background Art

[0002] The models and solution methods for the classical multi-agent task allocation problem are already very mature. Its essence is to model this problem as a combinatorial optimization problem and realize the mapping relationship between UAV nodes and task nodes under complex constraint conditions. In terms of solution methods, they can be mainly divided into two categories: centralized methods and distributed methods. There is a global information control and decision-making center in the centralized task allocation architecture. During the allocation process, each UAV platform reports the situation information to the central node, and finally a globally unified task allocation scheme is generated. However, this method is very sensitive to the scale effect. Especially under dynamic conditions, the situation information is updated frequently, which is extremely likely to cause communication congestion and difficult decision-making. In addition, there is a risk of "single point failure" of the central node under adversarial conditions, and the robustness of the system is poor. Most current UAV swarms adopt platforms with low-cost modular design. The platforms achieve distributed deployment of task resources by carrying heterogeneous payloads. The large-scale, dynamic and distributed characteristics of the swarm make it difficult to form a unified information interaction node globally. Therefore, the applicability of the solution methods for distributed task allocation is better.

[0003] Under the distributed task allocation architecture, each UAV makes decisions based on its own local information as an independent functional node, and realizes consistent situation awareness and decision conflict resolution through information interaction with other nodes in the domain. Existing research results include contract net auction methods based on market mechanisms, CBAA methods for single node - single task allocation, CBBA methods and CBGA methods for single node - multi task allocation, etc. However, the foregoing traditional UAV swarm technologies have the technical problem of low efficiency in constructing task networks for heterogeneous UAV swarms in large-scale and dynamic environments. Summary of the Invention

[0004] In view of the problems existing in the above traditional methods, the present invention proposes a method for constructing an ACK for a heterogeneous UAV swarm based on DCBGA, a system for constructing an ACK for a heterogeneous UAV swarm based on DCBGA, and a computer device, which can significantly improve the efficiency and applicability of constructing task networks for heterogeneous UAV swarms in large-scale and dynamic environments.

[0005] To achieve the above object, the embodiments of the present invention adopt the following technical solutions:

[0006] On the one hand, a method for constructing an ACK for a heterogeneous UAV swarm based on DCBGA is provided, including the steps:

[0007] Divide the overall task process in a dynamic adversarial environment into discrete time series and determine each time step;

[0008] Within the current time step, each unmanned aerial vehicle (UAV) in the heterogeneous UAV cluster uses the constructed DCBGA model for target selection; during the target selection process, each UAV selects the target with the highest marginal benefit according to the latest situation information within its own perspective.

[0009] Within the current time step, each UAV uses the DCBGA model for the consensus process; during the consensus process, each UAV respectively exchanges information with adjacent UAVs directly connected by communication to complete the sharing of situation information and conflict resolution.

[0010] When the number of rounds completed in the target selection and consensus process within the current time step reaches the set number of rounds, each UAV completes the update of its own local information, the switching of its own task execution state and motion state, and enters the next time step.

[0011] Following the target selection and consensus process of each time step, output and update the constructed ACK network of the heterogeneous UAV cluster; the ACK network is used to indicate that each UAV executes the task process for the selected target.

[0012] On the other hand, there is also provided a heterogeneous UAV cluster ACK construction system based on DCBGA, including:

[0013] A task division module, which is used to divide the overall task process in a dynamic adversarial environment into discrete time series and determine each time step;

[0014] A target selection module, which is used to indicate that each UAV in the heterogeneous UAV cluster uses the constructed DCBGA model for target selection within the current time step; each UAV selects the target with the highest marginal benefit according to the latest situation information within its own perspective.

[0015] A consensus module, which is used to indicate that each UAV uses the DCBGA model for the consensus process within the current time step; during the consensus process, each UAV respectively exchanges information with adjacent UAVs directly connected by communication to complete the sharing of situation information and conflict resolution.

[0016] A time switching module, which is used to indicate that each UAV completes the update of its own local information, the switching of its own task execution state and motion state, and enters the next time step when the number of rounds completed in the target selection and consensus process within the current time step reaches the set number of rounds.

[0017] A network output module, which is used to output and update the constructed ACK network of the heterogeneous UAV cluster following the target selection and consensus process of each time step; the ACK network is used to indicate that each UAV executes the task process for the selected target.

[0018] On the other hand, a computer device is also provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA are implemented.

[0019] One of the above technical solutions has the following advantages and beneficial effects:

[0020] In the above-mentioned method, system and device for constructing an ACK for a heterogeneous UAV cluster based on DCBGA, in the context of a large-scale heterogeneous UAV cluster in a dynamic confrontation environment, by dividing the overall task process into discrete time series and determining each time step, and using the constructed DCBGA model, each UAV in the heterogeneous UAV cluster repeats multiple rounds of target selection processes and consensus processes within each time step to determine the global task allocation plan for the next time step, and finally implements the global-level target selection plan into the construction of the cluster task execution network. As the task process evolves, the above-mentioned processing flow of the DCBGA model is repeatedly executed to reconstruct the task execution network, continuously promoting the progress of the cluster task completion. Compared with traditional technologies, the above solution realizes the construction of a task network for a heterogeneous UAV cluster in a large-scale and dynamic environment, with high efficiency and strong applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0022] Figure 1 It is a schematic flowchart of a method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA in one embodiment;

[0023] Figure 2 It is a schematic diagram of the motion state and transfer process of a UAV in one embodiment;

[0024] Figure 3 It is a schematic flowchart of a method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA in another embodiment;

[0025] Figure 4 It is a schematic diagram of the module structure of a system for constructing an ACK for a heterogeneous UAV cluster based on DCBGA in one embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] To make the objectives, technical solutions and advantages of this application more clear and understandable, the following further elaborates on this application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining this application and are not intended to limit this application.

[0027] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0028] It should be noted that the mention of "embodiment" in this application means that the specific features, structures or characteristics described in connection with the embodiment may be included in at least one embodiment of the present invention. The display of this phrase at various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments.

[0029] Those skilled in the art can understand that the embodiments described in this application can be combined with other embodiments. The term "and / or" used in the specification and claims of the present invention refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0030] Through the modeling and analysis of the dynamic UAV cluster system, it is known that the communication network of the UAV cluster in the mission scenario is dynamically evolved by the self-organization process of the cluster, and the task allocation situation of each mission platform directly determines the operation direction of the cluster self-organization, that is, the construction of the execution network (also known as ACK) of the UAV cluster system is ultimately transformed into a self-organization process of system efficiency release based on the task allocation scheme. Therefore, the problem of constructing the execution network of the UAV cluster under dynamic confrontation conditions is abstracted into a dynamic task allocation problem of multi-agent.

[0031] The models and solution methods of the classical multi-agent task allocation problem are already very mature. Its essence is to model this problem as a combinatorial optimization problem and realize the mapping relationship between UAV nodes and task nodes under complex constraint conditions. In terms of solution methods, it can be mainly divided into two categories: centralized methods (solution) and distributed methods. There is a global information control and decision-making center in the centralized task allocation architecture. During the allocation process, each UAV platform reports the situation information to the central node, and finally a globally unified task allocation scheme is generated. However, this method is very sensitive to the scale effect. Especially under dynamic conditions, the situation information is updated frequently, which is extremely likely to cause communication congestion and decision-making and solution difficulties. In addition, there is a risk of "single point failure" of the central node under confrontation conditions, and the robustness of the system is poor.

[0032] Most current UAV swarms adopt platforms with low-cost modular designs. The platforms achieve distributed deployment of combat resources by carrying heterogeneous payloads. The large-scale, dynamic, and distributed characteristics of the swarm make it difficult to form a unified information interaction node globally. Therefore, the applicability of the solution method for distributed task allocation is better. Under the distributed task allocation architecture, each UAV, as an independent functional node, makes decisions based on its local information and achieves consistent situation awareness and decision conflict resolution through information interaction with other nodes in the domain, thus making this method exhibit better robustness, adaptability, and scalability.

[0033] For example, the contract net auction method based on the market mechanism is widely used in the field of multi-agent task allocation. This process includes three steps: solicitation of bids, submission of bids, and winning the bid. To further weaken the central node status of the auctioneer and enhance the distributed characteristics of the method, some researchers have proposed the CBAA method for single-node - single-task allocation and the CBBA method for single-node - multi-task allocation based on the contract net mechanism, realizing a "fully" distributed consistency negotiation process, which is more in line with the scenario setting of the emergence of swarm intelligence in the process of cluster combat. However, the method itself is based on the scenario of homogeneous multi-agent task allocation under static conditions. Therefore, many scholars have extended the problem complexity and correspondingly carried out various method improvements on this basis.

[0034] First, the author team of the CBBA method has made a series of improvements to enhance the applicability of the method in actual task scenarios. Some scholars have introduced a partial replanning mechanism on the basis of the CBBA method to deal with the situation of new dynamic targets emerging during the task execution process. Some scholars have combined the CBBA method with the Dijkstra method to achieve path planning under obstacle avoidance constraints while performing task allocation. Some scholars have further considered the maintenance of multi-UAV communication connectivity during the task execution process on the basis of task allocation, enhancing the multi-aircraft cooperation ability under dynamic communication constraints. It should be noted that the task background of this method is the communication maintenance of a small-scale multi-UAV facing the information feedback requirement of the ground control center, which is different from the scenario of this application where a large-scale heterogeneous UAV swarm "operates offshore" autonomously to perform tasks and maintain internal communication connections within the swarm. Some scholars have further extended the heterogeneous attributes of tasks and agents, introducing a heterogeneous agent cooperation mechanism, broadening the application scenario of the method.

[0035] In addition, some scholars have added temporal coupling constraints during the task execution process to the method model, making it more in line with the actual application requirements. The task execution process is divided into discrete time series, and multi-UAV task allocation in dynamic scenarios is achieved through iterative greedy optimization. Inspired by the cognitive behavior of gregarious animals, some scholars have proposed the CBGA method, which takes heterogeneous UAV coalitions as agent individuals in the CBBA method. The coalition collaboratively completes the assigned tasks through the joint cooperation of heterogeneous units, which is very close to the task scenario of multiple UAVs launching simultaneous attacks on enemy targets.

[0036] However, most of the above existing methods have the following problems: (1) Assuming that multi-agents have prior knowledge of global target information, lacking the process of consistent perception of the unknown environment by the cluster, especially the update of the "pop-up" targets and the operation situation of UAV battle damage under dynamic confrontation conditions; (2) The agent model is too simplified, lacking the consideration of kinematic constraints, resulting in inaccurate calculation of the travel cost during task transfer; (3) For the combination of heterogeneous agents, the superposition of non-linear effectiveness is not considered; (4) The scale of multi-agents is small, and the real-time requirement for task solving is not high, which is different from the actual application scenario of large-scale UAV cluster operations.

[0037] Therefore, this application makes improvements to the CBGA method in terms of dynamic confrontation, scale scalability, and non-linear heterogeneous effectiveness superposition in an unknown environment, and proposes the DCBGA (Dynamic Consensus-Based Group Algorithm) method for multi-UAV - multi-target task allocation under dynamic confrontation conditions, which is more in line with the task scenario of heterogeneous UAV cluster execution network construction.

[0038] Based on the CBBA method, some scholars considered the constraint that a single task requires multiple UAVs to be matched simultaneously and accordingly proposed the CBGA method. This application makes dynamic improvements based on this method and proposes the DCBGA method to drive the UAV individuals in the cluster to select the target with the maximum marginal benefit, thereby promoting multiple heterogeneous UAVs to form an execution network for each target at the global level.

[0039] In one embodiment, as Figure 1 shown, a method for constructing an ACK of a heterogeneous UAV cluster based on DCBGA is provided, including steps S11 to S15:

[0040] S11, dividing the overall task process in a dynamic confrontation environment into discrete time series and determining each time step;

[0041] S12. During the current time step, each UAV in the heterogeneous UAV cluster uses the constructed DCBGA model for target selection. During the target selection process, each UAV selects the target with the highest marginal benefit according to the latest situation information within its own perspective.

[0042] S13. During the current time step, each UAV uses the DCBGA model for the consistency process. During the consistency process, each UAV exchanges information with adjacent UAVs directly connected by communication to complete the sharing of situation information and conflict resolution.

[0043] S14. When the number of rounds completed in the target selection and consistency process in the current time step reaches the set number of rounds, each UAV completes the update of its own local information, switches its own task execution status and motion status, and enters the next time step.

[0044] S15. Following the target selection and consistency process of each time step, the constructed ACK network of the heterogeneous UAV cluster is output and updated. The ACK network is used to indicate each UAV to execute the task process for the selected target.

[0045] Key information about the DCBGA (model) method: According to the requirements of the CBGA method, during the entire task execution process, each UAV locally stores two pieces of decision-making information, the successful bidder matrix and the tender price matrix, for target decision-making. Considering the dynamic and unknown characteristics of the real operation environment, the task process needs to be divided into multiple discrete time steps, and the global dynamic task allocation is completed in a rolling manner within the time domain. The time step Δt here can be any defined period of time, and the specific detail granularity depends on the actual situation change of the operation. After each time step Δt, each UAV U i communicates with neighbor UAVs in the connected node set C i in a multi-round communication to complete the update of the above two pieces of decision-making information. Each UAV makes a target selection in the way of joining each task alliance based on the latest decision-making information, so that a dynamic execution network is formed in the task space for the entire cluster.

[0046] Among them, the successful bidder matrix is denoted as Z i , with a size of N U being the total number of UAVs in the heterogeneous UAV cluster, and i being the number of targets currently known to UAV U. At the initial moment, all elements in Zi are set to zero. Under dynamic conditions, UAVs often cannot achieve timely situation awareness of all UAVs and targets in the entire domain. Therefore, it is defined that UAV U iThe dead UAV list is DL i = {U i,k}(k=1,2,...), where U i,k For UAV i The kth known failed drone at present; define drone U i The target list is TL i ={T i,j |T i,j =(x i,j ,ts i,j ,q i,j , tr i,j ),j=1,2,...}, used to record the current drone U i Known target information, easy to know That is, the length of the target list is consistent with the number of columns in the winning bidder matrix. i,j For UAV i The jth target known from its own perspective; x i,j Target T i,j Location information; ts i,j Target T i,j The state flag is set to 1 at the initial moment. i,j Set to -1 after being eliminated; q i,j For the drone to target T i,j The quotation of the target T is the same as the value of the drone below. i,j Bid value The value of tr is consistent; i,j For UAV i Fly from current location to destination T i,j The flight track of the mission execution position is first calculated according to formula (1) to obtain the UAV motion state sequence with fine time granularity, and then the state information of the corresponding time is intercepted from it in units of time step Δt to form the flight track, which is recorded as Element tr i,j (t) represents the state information of the drone at the time (x i (t),v i (t)). Considering the cluster scale effect, a lightweight optimal control model is used to plan the trajectory for a single UAV, and the optimal control input at any time is calculated as:

[0047]

[0048] Among them, u i For UAV i The control quantity, v i For UAVi The speed, x i For the drone U i The current position, t is the time, Is the final arrival at the target T j The task execution position, i.e., the optimal entry point; Is the estimated time to reach the optimal entry point position of the target. Element Indicates that for the drone U i It seems that the drone U k Is the winner of the target T i,j That is, the drone U i Considers that the drone U k Chooses to join the task coalition of the target T i,j Otherwise Correspondingly, in the winning bid price matrix Y i Among them, the element The value of is the drone U i Learns that the drone U k The bid value for the target T i,j The winning bid price matrix and the winner matrix are of the same size and the elements correspond one by one. At the initial moment, all elements in Y i Are set to negative infinity (-∞). Based on the non-linear description of the heterogeneous task efficiency, the expected task revenue that can be obtained by the current temporary task coalition is:

[0049]

[0050] Among them, Is the nominal return of the target T j ; Is the non-linear task efficiency of the temporary task coalition G(T j ) constructed for the target, Is the target T j The total non-linear efficiency formed by the demand D(T j ) for each type of drone. If Take λ is the adjustment coefficient; Is the formation of each type of drone The estimated time to reach its designated task execution position, which takes the arrival time of the drone that arrives at the task position the latest in the formation. If the formation scale of this type of drone is 0, then set To a relatively large value Q.

[0051] According to the formula for calculating the expected task revenue of the temporary task coalition in Equation (2), it can be seen that under the condition of the same type and scale of coalition member combinations, the formation of each type of drone The estimated time to reach its designated task execution position The smaller it is, the higher the expected return of the current task, that is, under the condition of fixing the composition of the UAV formation of this type , the task return is a monotonic function of , and that is, the task return is also a monotonic function of the expected arrival time of any member at the target T i,j time . Then, using other monotonic functions of as the bid value of the UAV U i for the target T i,j has the same utility as using the task return. For convenience, this application uses the moment when the UAV U i arrives at the task execution position of the target T i,j with the opposite number as its bid value for the target T . In this way, when UAVs of the same type compete to join the task coalition of the same target, the two only need to compare their respective i,j bid values values.

[0052] In addition, when UAVs compete for targets, they all give bid values for each target based on their own perspectives. Due to communication limitations, the formation of the temporary task coalition of the target in their respective perspectives G(T i,j ) may be different, which will lead to deviations in the calculation results of the task return. And the value ofis independent of the coalition formation situation. Therefore, using as the bid value (quotation) of the UAV U i for each target is more accurate, the internal consistency recognition of this information in the cluster is faster, and it is also convenient for the UAV to directly record the arrival time of each member in the coalition at the task execution position for calculating the task return

[0053] In addition, in the DCBGA method, each UAV also needs to record some key information of the method locally, including the task package vector b i , the execution path vector p i , and the communication timestamp vector S i . Among them, the task package vector b i contains all the target tasks to be executed by the UAV U i and is arranged in the order of addition; the content of the execution path vector p i is the same as that of the task package vector, and the difference is that the arrangement order of the target tasks is the optimal execution order, that is, after executing all tasks in this order, the return of the UAV U i is the largest. The communication timestamp vector Si for recording the UAV U i the latest moment when it receives information from adjacent UAVs. The specific usage will be described in detail this afternoon.

[0054] The starting point for constructing a task package vector containing multiple tasks to be executed for a single UAV is that when multiple target positions are close, the swarm only needs to assign a single UAV to execute multiple targets in the contiguous area instead of assigning multiple UAVs to execute these target tasks separately. However, under dynamic confrontation conditions, the battlefield situation changes rapidly, and new targets may pop up and UAVs may fail at any time. The large-length task packages constructed by UAVs will face frequent task reallocations, and at the same time, generating new optimal execution paths will be very time-consuming. And due to the limited characteristics of swarm communication, the perception speed of each UAV for updating the battlefield situation is slow, and the credibility of the constructed large-length task execution paths is in doubt. For this reason, this application enhances the greedy characteristics of UAV individuals during the task execution process. In the DCBGA method, the task package vector b i and the execution path vector p i are degraded into a key piece of information, that is, the target choice list c i of the current UAV U i , c i = j > 0 indicates that the UAV U i currently selects the execution target T i,j , otherwise c i = 0. In addition, based on the current typical UAV swarm task idea, there are often a large number of redundancies in the swarm. The number of UAVs is much larger than the number of task targets. When UAV individuals execute tasks, they more often consider real-time requirements rather than optimality requirements. And starting from the idea of system destruction, under high-dynamic confrontation conditions, the swarm should launch simultaneous attacks on multiple targets rather than sequential attacks. In summary, the greedy characteristics of UAV individuals designed in this application are more in line with the task scenario setting of constructing an execution network under high-dynamic confrontation conditions.

[0055] Specifically, in the context of dynamic confrontation, the entire task process is divided into discrete time series. At each time step, the operation of the DCBGA (model) method is divided into two stages: the target selection process and the consistency process. Each UAV individual selects the target with the highest marginal benefit according to the latest situation information within its own perspective, and then conducts information interaction with adjacent UAVs directly connected by communication to complete the sharing of situation information and achieve conflict resolution. The swarm will repeat multiple rounds of the two-stage DCBGA operation process at each time step, and finally determine the global task allocation plan for the next time step (according to the need for information exchange, the number of rounds of method operation within a single time step Δt is set to c tAfter that, each UAV completes the update of its own local information and the switching of its own task execution status and motion status, and enters the next time step, and finally implements the global-level target selection scheme to the construction of the cluster execution network. As the task process evolves, the DCBGA method is repeatedly run to reconstruct the execution network, continuously promoting the completion progress of the cluster task.

[0056] The above heterogeneous UAV cluster ACK construction method based on DCBGA, in the context of a large-scale heterogeneous UAV cluster in a dynamic adversarial environment, divides the overall task process into discrete time series and determines each time step. Using the constructed DCBGA model, each UAV in the heterogeneous UAV cluster repeats multiple rounds of target selection process and consensus process within each time step, determines the global task allocation scheme in the next time step, and finally implements the global-level target selection scheme to the construction of the cluster task execution network. As the task process evolves, the processing flow of the above DCBGA model is repeatedly executed to reconstruct the task execution network, continuously promoting the completion progress of the cluster task. Compared with traditional technologies, the above scheme realizes the construction of the task network of heterogeneous UAV clusters in a large-scale and dynamic environment, with high efficiency and strong applicability.

[0057] In one embodiment, the task execution status of the UAVs in the heterogeneous UAV cluster includes no target traction, target traction, and fixed target selection, and the motion status of the UAVs includes serpentine maneuver, track flight, circumferential waiting, synchronous attack, and node failure.

[0058] It can be understood that under the action of the behavior-driven module in this embodiment, the various motion states and transition rules of the UAVs are as follows Figure 2 As shown. It is set that the task execution status es (task engagement state) of the UAVs during the task execution process is divided into three types, namely uncommitted, committed, and sticky. The motion state ms is divided into five types, namely serpentine maneuver, track flight, circumferential waiting, synchronous attack, and node failure.

[0059] Among them, no target traction means that the UAV does not currently respond to the needs of any target and randomly wanders in the task space, being in the serpentine maneuver state. Target traction means that the UAV currently responds to the task requirements of a certain target and is in the track flight state. During the process of flying to the task execution location, the UAV may discover a target with higher benefits and switch to the flight track to the task execution location of that target. In addition, the UAV may find that there are already UAVs better than itself going to execute this task, then it is considered to have failed in the "bid" for this task and switches back to the serpentine maneuver state.

[0060] The fixed target selection means that the UAV "sticks" to the currently selected target task without task switching until it reaches a specific task execution position around the target and enters the surrounding waiting state, waiting to form a temporary task alliance for the target with the remaining heterogeneous UAVs. After the alliance is formed, the heterogeneous UAVs in the alliance will launch a synchronous attack on the target. After the attack is completed, the alliance will dissolve, and each UAV will return to the snake-shaped maneuver state. In some cases, due to the remote location of the target or the low priority of the target, the UAV in the surrounding waiting state may not be able to welcome newly joined UAVs in the alliance for a long time, thus falling into a "stuck" state. At this time, the UAV can actively break this state (such as setting a waiting countdown) and re-enter the snake-shaped maneuver state to respond to the needs of other targets. In addition, under dynamic confrontation conditions, the UAV may be attacked by the opponent at any time during the entire task execution process and enter the node failure state (dead).

[0061] In one embodiment, the process of target selection in step S12 above may specifically include the following process:

[0062] Each UAV selects and executes the subtask of the target with the largest marginal benefit increment for itself based on the latest situation information within its own perspective and adds the target to its own target selection list.

[0063] It can be understood that in the target selection stage, each UAV selects and executes the subtask of the target with the largest marginal benefit increment for itself based on the latest situation information within its own perspective and adds the target to its target selection list c i If UAV U i is currently in the fixed target selection state (sticky), it will directly jump out of the target selection process and continue to maintain the target selection of the previous time step. UAVs in the other two states will actually enter the target selection process.

[0064] The core of the target selection process is to calculate the marginal benefit r i (T i,j ) brought by each UAV selecting and executing the subtasks of each target in the current situation. The specific process has been described in detail in the key information of the foregoing DCBGA (model) method. The main situation information relied on is the non-linear task effectiveness of each target's temporary task alliance and the formation of each type of UAV The estimated time to reach its best task execution position UAV U i makes target selection in the way of joining the task alliance and will obtain and different values. The solution of can refer to the calculation process of the execution network global task performance.

[0065] However, the drone U i cannot know the communication connection situation of each task unit in the temporary task alliance based on its own perspective, that is, it cannot calculate the response speed between each task node. Therefore, it is necessary to estimate the non-linear task performance of the alliance. It is assumed that after all members of the alliance enter the encirclement waiting stage, a coordinated silent attack can be launched based on short-range communication. Then, the ACK performance transfer efficiency between nodes can be ignored in this process, that is, the distance sensitivity coefficient Δ = 1 is taken. At this time, the alliance task performance is only related to the scale of each type of drone grouping (scale). Therefore, in the task selection process, the non-linear task performance of the temporary task alliance of the target T i,j is:

[0066]

[0067]

[0068] Among them, μ is the scale positive factor, which is used to attract drones to initiate alliance formation in the initial stage; I(·) is the indicator function, which takes the value of 1 if the criterion is true, otherwise 0. In this application, the initial arrangement order of each drone in the cluster is used as its exclusive identity identifier, that is, ID(U i ) = i, and type(U i ) ∈ {S, A, J, D} represents the type of the drone U i , which are reconnaissance type, attack type, electronic warfare type, and decoy type respectively.

[0069] To obtain the value of the expected time , it is necessary to first calculate the optimal control sequence of the drone U i arriving at different discrete points of the circular path of the target T i,j , and then obtain the best entry point, the best entry direction, and the corresponding arrival time If then the value of is Otherwise, its value is the same as If the scale of the alliance is saturated before the drone joins the temporary task alliance of the target, that is then it is necessary to examine and If i then the drone U l will replace the drone U i in the alliance, otherwise U i

[0070] In one embodiment, further, the process of target selection in step S12 may specifically further include the following process:

[0071] During the process of calculating the marginal benefit, record the optimal task coalition combinations of each target in the current UAV's perspective.

[0072] If the members of the optimal task coalition combination change, the current UAV performs target selection; otherwise, skip the execution of target selection.

[0073] Specifically, for a large-scale UAV cluster, there are a large number of "idle" UAVs in the cluster as redundant task nodes, and these UAVs will be in an uncommitted state for a long time before launching an attack. Each time entering the target selection process, these UAVs need to calculate the marginal benefit for all known targets within the entire domain. As can be seen from the above analysis, the solution of the optimal control sequence in this process involves numerical integration calculations, which will consume a large amount of computing resources and take a long time, and it can be ultimately predicted that these UAVs will still be in an uncommitted state, that is, it can be considered that there are a large number of ineffective calculations in the operation of the method in this process. It can also be understood that the task unit has an intelligent defect of being "forgetful", always constantly forgetting the experiences of competing and failing with other UAVs during the target selection process.

[0074] Therefore, it is necessary to record the optimal coalition combination BG i of each target in the current perspective of UAV U i (T i,j ) during the process of calculating the marginal benefit. Only when the members of the optimal coalition combination change, will UAV U i enter the target selection process for this target; otherwise, set the status flag ts i,j of target T i,j to 0.5 and directly skip this process, thereby improving the efficiency of target selection processing and reducing the computational overhead.

[0075] In one embodiment, further, the process of target selection in step S12 may specifically further include the following process:

[0076] After the UAV enters the first set range from the target, the UAV sets its task execution status to the fixed target selection state; the target selection range of the UAV is the targets within the second set range from the UAV itself.

[0077] It can be understood that UAV U i depends on the formation of the target coalition within its own perspective G(T i,j), and large-scale clusters under dynamic confrontation conditions have problems such as delayed local information updates and slow global situation consistency due to limited communication distance and the number of exchanges within a single time step. It is difficult for each drone to reach a consensus on the formation of each target alliance in a short period of time. At the same time, a certain drone switching target selection will trigger a large-scale chain reaction at the cluster level, and the reaction speed will lag behind the consistency process and further increase the difficulty of reaching a situation consensus. The first setting range and the second setting range can be different distance ranges set according to actual needs, and the first setting range can be smaller than the second setting range.

[0078] In the process of target selection, based on the pursuit of optimality, individual drones will fall into a "confused" situation, constantly switching target selection for a long period of time and spreading this "symptom" on a large scale at the cluster level. At certain moments, especially in the initial stage of dynamically injecting "pop-up" targets into the task space, members of the task alliance of some targets are missing, and global execution network vulnerabilities appear, affecting the performance of cluster tasks. At the same time, the switching of drone target selection will correspondingly trigger repeated switching of movement modes, which will hinder the drone from quickly reaching the task execution position of the selected target and delay the progress of the task.

[0079] Specifically, in order to further accelerate the convergence speed of the DCBGA method and realize the rapid and complete construction of the heterogeneous UAV cluster execution network, this embodiment further enhances the greedy characteristics of the UAV in the target selection process to alleviate the "confusion" symptoms of the UAV in the initial stage of target selection. After the UAV enters the first set range (sticky_limit) from the target, it changes its task execution state to es i (task engagement state) is set to fixed target selection (sticky), that is, set es i = 2, after which the drone will no longer switch the current target selection from its own perspective. The drones in the cluster that are better than themselves will be "squeezed out" of the task alliance of the current target selection through the consistency process, thereby indirectly achieving the pursuit of optimality of the execution network.

[0080] In addition, to avoid unnecessary attempts by drones to select targets at very long distances, that is, to prevent drones from making unnecessary attempts to select targets at very long distances, that is, to prevent drones from making unnecessary attempts to select targets at very long distances, which is considered to be inconsistent with the dynamic characteristics of the fast closure of the execution network and which will most likely be “squeezed out” of the temporary mission alliance by other drones, reflecting the “reckless” nature of drones in target selection, individual drones are set to be able to select targets within the second set range (select_limit) from themselves.

[0081] In one embodiment, further, the consistency process in step S13 may specifically include the following processing procedures:

[0082] Each unmanned aerial vehicle (UAV) updates the latest situation information by exchanging the target list and the list of failed UAVs; the latest situation information includes pop-up target information, target status changes, and UAV status changes.

[0083] When each UAV learns of new target information, the UAVs in the fixed target selection state are released and their task execution status is reset to zero.

[0084] It can be understood that under dynamic conditions, individual UAVs within the cluster exchange the latest battlefield situation and decision-making information with adjacent communication nodes within one-hop range, ultimately achieving the consistency of situation information and the conflict resolution of the global task assignment scheme. Before performing the conflict resolution of the task assignment scheme, the UAVs need to share the latest battlefield situation information with adjacent nodes to unify the situation awareness level to the same level. The update of the latest battlefield situation information including "pop-up" target information, target status changes, and UAV status changes is achieved by exchanging the target list TL and the list of failed UAVs (deadUAV list) DL. When a UAV learns of new target information through communication, to prevent it from ignoring the possibility of optimizing the target selection plan and blindly adhering to the previous target selection, thus showing a "stubborn" characteristic, the UAVs in the fixed target selection state need to be released and their task execution status is reset to zero.

[0085] In one embodiment, further, the consistency process in step S13 may specifically further include the following processing procedures:

[0086] Each UAV respectively performs information interaction with adjacent UAVs directly connected by communication based on the communication timestamp vector to complete the conflict resolution of the target selection plan.

[0087] It can be understood that after the situation awareness level reaches the same level, the UAVs continue to exchange decision-making information with adjacent UAVs directly connected by communication for the conflict resolution of the target selection plan. During the process of information exchange with adjacent UAVs, UAV i needs to determine the newness and oldness of various received information to avoid being interfered by outdated and invalid error information in decision-making. This requires the use of the communication timestamp vector whose size is 1×N U , and the element s ik is used to record UAV i receives UAV kThe latest moment of information. This application divides the entire task process into multiple discrete time steps, and sets that within a single time step Δt, the UAV exchanges information with adjacent UAVs in direct communication multiple times. To describe the granularity of this process accurately to each communication, in this embodiment, the latest cumulative round s of communication between two UAVs is used ik to indirectly reflect the newness and oldness of messages by characterizing the concept of "latest moment" above. Every time they communicate, s ik needs to be updated once. Its update rule under dynamic discrete time steps is as follows:

[0088]

[0089] where c t is the number of times each UAV exchanges information with adjacent UAVs within each time step Δt; τ r is the time when UAV U i receives the information of UAV U k . Here, "time" is understood as the round of information exchange within the current time step. For example, if each time step is set to 2 seconds and the value of c t within each time step is 3, and if UAV U i and UAV U k are in direct communication, then after the first round of communication in the 4th time step, the value of s ik is s ik = [(4 - 1)×2] / 2×3 + 1 = 10. It should be noted that in the CBGA method and the CBBA method, it is necessary to set the upper limit of the number of inter-aircraft communications as the number of steps for the method to converge, so as to ensure that the method globally converges to a conflict-free allocation scheme. For the CBGA method, its upper limit of the number of convergence steps is N min D. Where D is the diameter of the cluster communication network, defined as the longest shortest path between nodes in the cluster communication network, that is:

[0090]

[0091] For a certain complete allocation scheme:

[0092]

[0093] Under dynamic adversarial conditions, for a large-scale UAV cluster, the order of magnitude of N min D can easily reach 10 2 . Obviously, setting the number of inter-aircraft communications within a unit time step as the value N minD is too demanding and does not meet the real-time requirements. Moreover, for redundant clusters, a sufficient number of heterogeneous UAVs can often be evenly configured around the target, eliminating the need to globally propagate the target response requirements within a short time step. Therefore, the number of inter-UAV communications within each time step is set to a small positive integer c t , when a new target appears, there will be an "over-allocation" phenomenon in the initial stage. However, after a limited number of subsequent time steps, the cluster can achieve conflict resolution and form a good global allocation scheme, meeting the real-time requirements of dynamic execution network construction at the cost of sacrificing some optimality.

[0094] Still taking UAV U i receiving the message sent by UAV U k as an example, analyze the conflict resolution process of the target selection scheme based on the timestamp vector. When performing conflict resolution on the task coalition formation scheme G(T j ), it is necessary to resolve conflicts for various types of heterogeneous UAV combinations j one by one, that is, compare the reconnaissance UAV combinations attack UAV combinations i and UAV U k in the perspective of UAV U j to select the target T attack UAV combinations electronic warfare UAV combinations and decoy UAV combinations to specifically determine which UAVs each consists of.

[0095] When the number of UAVs of a certain type required by the target T j is 1, the conflict resolution process is the same as the CBBA method. When the demand is greater than 1, multi-UAV conflict resolution for the UAV combination scheme is required. This process can be divided into two steps: Update Outdated Info and Add Latest Info, ultimately obtaining a consistency process based on the communication timestamp vector and the same situation awareness level. Update Outdated Info means that compared with UAV U

[0096] , the information in the perspective of UAV U k regarding UAV U i selecting the target T m is outdated. UAV U k,j will replace this part of the information with that from UAV U i ​k Replace and update the corresponding information; Add Latest Info indicates that the drone U i does not know the drone U m Select the target T k,j In the case of, it examines its own situation regarding the target T k,j The situation of coalition formation and combines the information from the drone U k and decides whether to add the drone U m to the task coalition of the target T k,j According to the definition of the timestamp in Equation (5), after each information exchange between drones within a time step, the belonging moment of the information will change from (∑Δt + τ r Δt / c t ) to (∑Δt + (τ r +1)Δt / c t ).

[0097] In one embodiment, as Figure 3 shown, further, the above method for constructing the heterogeneous drone cluster ACK based on DCBGA may further include step S145:

[0098] S145, after the target selection and consensus process in each time step are completed, each drone respectively resets the failure information and updates the valid information; the reset of the failure information includes the reset of the quotes for all targets and the planned flight trajectories of the drones, and the update of the valid information includes the update of the quotes and flight trajectories of the selected targets of the drones.

[0099] It can be understood that after the two-stage information decision of each drone based on the DCBGA method, the target selection of itself within the next time step is finally determined. During this process, each drone also simultaneously makes corresponding adjustments to the task execution status. To finally implement the construction of the cluster execution network in the task space, each drone updates its own position x i and speed v i i i i i i i status information based on the target selection c

[0100] Subsequently, based on the latest task space status information, the UAV conducts a new round of battlefield situation awareness, which can specifically include: detecting "pop-up" targets, judging failed UAV nodes, and sensing the task execution status. Assume that the "pop-up" target can be immediately detected by reconnaissance UAVs and various target information calculations including target value, firepower requirements, and the best attack position can be achieved. Other types of UAVs sense the newly emerged target information through the sharing of situation information in the consistency process. Assume that after a UAV is attacked and fails, the adjacent UAVs directly connected to it in communication can sense its state change and spread this situation information within the cluster network.

[0101] After the UAV enters the surrounding waiting stage, it begins to closely monitor the task execution status of the other members in the coalition. When it is found that the coalition size reaches the target requirement and all members enter the surrounding waiting stage, the entire coalition switches to the cooperative operation state, effectively executes the target and removes the target from the task space, and the message that the target has been eliminated is also spread within the cluster. When it is found that all members in the coalition have reached the best attack position but the coalition size has not reached the target requirement, the UAV immediately starts state timing. If this situation persists for more than a certain time threshold, it is determined that the task coalition has entered the "stuck" state, and the task status ts j is marked as 0.

[0102] Since the heterogeneous UAV cluster can achieve high-intensity and networked synchronous operations on multiple targets in the task space, rather than the strategy of attacking one by one under low-confrontation conditions. When it is found that a certain target task coalition has fallen into the "stuck" state, it indicates that the heterogeneous UAV cluster does not have the ability to break through the joint system under this condition. Therefore, the cluster should appropriately concentrate task resources to launch a thorough operation on fewer targets, rather than launching ineffective operations on a large area of targets. For this reason, the temporary task coalition of UAVs in the "stuck" state will actively jump out of this state to facilitate responding to the task requirements of the remaining targets.

[0103] To fully study the cluster execution network construction ability under different task resource densities, it is necessary to eliminate the influence of the target distribution condition on the experimental results. Therefore, after any target is eliminated, a new target will be randomly generated in the task space to replace it, achieving the effect of generating "pop-up" targets without changing the task resource density. Except for the change in position information, the other target information of the new target is the same as that of the replaced target. And the targets that have fallen into the "stuck" state before will continue to be retained in the task space. At the same time, to avoid repeatedly falling into this situation, no new target will be generated in the task space to replace it and the cluster will no longer construct an execution network for this target.

[0104] In addition to perceiving and updating the situation information of the mission environment, the UAV needs to adjust its locally stored decision information to prepare for entering the next time step for the DCBGA method process, including resetting invalid information and updating valid information. The invalid information reset operation is mainly for UAVs that are not currently in the fixed target selection state, that is, es i For UAVs with a target size of less than 2, since these UAVs need to make a new target selection decision for the global targets in the next time step, the multiple target information calculated in the previous time step is invalid. i = 0, the failure information includes its bids for all targets and planned flight paths, i.e. Before re-entering the DCBGA method, you need to reset this part of information to the initial value. i = 1, the failure information includes the quotes and planned flight paths of all targets except the target selected in the previous time step, that is, The same reset operation is required. The effective information update operation is mainly for the quotation and flight track of the target selected by the drone, that is, {q i,j ,tr i,j}(c i =j), and the winning bid price matrix Y i By default, all drones will firmly stick to the target selection in the previous time step, and the effective information update method is:

[0105]

[0106] At this point, each drone has completed the update and adjustment of all local information, and entered a new round of DCBGA method flow with a new spatial position, speed and communication topology structure, improving the construction of the global execution network and promoting the completion progress of the cluster task.

[0107] It should be understood that although Figures 1 to 3 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in the application, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Figures 1 to 3 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequentially, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0108] In one embodiment, Figure 4As shown in the figure, a heterogeneous UAV cluster ACK construction system 100 based on DCBGA is also provided, which includes a task division module 11, a target selection module 13, a consistency module 15, a time switching module 17, and a network output module 19. Among them, the task division module 11 is used to divide the overall task process in a dynamic adversarial environment into discrete time series and determine each time step. The target selection module 13 is used to instruct each UAV in the heterogeneous UAV cluster to perform target selection using the constructed DCBGA model within the current time step; each UAV selects the target with the highest marginal benefit according to the latest situation information within its own perspective. The consistency module 15 is used to instruct each UAV to perform a consistency process using the DCBGA model within the current time step; during the consistency process, each UAV respectively exchanges information with adjacent UAVs directly connected by communication to complete the sharing of situation information and conflict resolution. The time switching module 17 is used to instruct each UAV to complete the update of its own local information, the switching of its own task execution status and motion status and enter the next time step when the number of rounds completed in the target selection and consistency process within the current time step reaches a set number of rounds. The network output module 19 is used to output and update the constructed ACK network of the heterogeneous UAV cluster following the target selection and consistency process of each time step; the ACK network is used to instruct each UAV to execute the task process for the selected target.

[0109] The above-mentioned heterogeneous UAV cluster ACK construction system 100 based on DCBGA, in the context of a large-scale heterogeneous UAV cluster in a dynamic adversarial environment, by dividing the overall task process into discrete time series and determining each time step, using the constructed DCBGA model, each UAV in the heterogeneous UAV cluster repeats multiple rounds of target selection process and consistency process within each time step, determines the global task allocation plan for the next time step, and finally implements the global-level target selection plan to the construction of the cluster task execution network. As the task process evolves continuously, the above-mentioned processing flow of the DCBGA model is repeatedly executed to realize the reconstruction of the task execution network, and continuously promotes the progress of the cluster task completion. Compared with traditional technologies, the above solution realizes the construction of the task network of a large-scale and dynamic environment heterogeneous UAV cluster, with high efficiency and strong applicability.

[0110] In one embodiment, the above-mentioned heterogeneous UAV cluster ACK construction system 100 based on DCBGA can also be used to implement other steps or processes in each embodiment of the above-mentioned heterogeneous UAV cluster ACK construction method.

[0111] For the specific limitations of the heterogeneous UAV cluster ACK construction system 100 based on DCBGA, reference can be made to the corresponding limitations of the heterogeneous UAV cluster ACK construction method based on DCBGA in the above text, which will not be elaborated here. Each module in the above-mentioned heterogeneous UAV cluster ACK construction system 100 can be implemented in whole or in part by software, hardware, and their combination. The above-mentioned modules can be embedded in the form of hardware in or independent of devices with specific data processing functions, or stored in the memory of the aforementioned devices in the form of software, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules. The aforementioned devices can be, but are not limited to, various types of portable, vehicle-mounted or shipborne UAV centralized control devices existing in the art.

[0112] In one embodiment, a computer device is further provided, including a memory and a processor. When the processor executes the computer program, the following processing steps are implemented: dividing the overall task process in a dynamic confrontation environment into discrete time series and determining each time step; within the current time step, each UAV in the heterogeneous UAV cluster uses the constructed DCBGA model for target selection; during the target selection process, each UAV selects the target with the highest marginal benefit according to the latest situation information within its own perspective; within the current time step, each UAV uses the DCBGA model for a consistency process; during the consistency process, each UAV respectively conducts information interaction with adjacent UAVs directly connected by communication to complete the sharing of situation information and conflict resolution; when the number of rounds completed in the target selection and consistency process within the current time step reaches the set number of rounds, each UAV completes the update of its own local information, the switching of its own task execution status and motion state, and enters the next time step; following the target selection and consistency process of each time step, the constructed ACK network of the heterogeneous UAV cluster is output and updated; the ACK network is used to instruct each UAV to execute the task process for the selected target.

[0113] It can be understood that in addition to the above-mentioned memory and processor, the above-mentioned computer device also includes other software and hardware components not listed in this specification, which can be specifically determined according to the model of the specific data processing and control device in different application scenarios, and will not be listed and elaborated one by one in this specification.

[0114] In one embodiment, when the processor executes the computer program, it can also implement the additional steps or sub-steps in each embodiment of the above-mentioned heterogeneous UAV cluster ACK construction method based on DCBGA.

[0115] In one embodiment, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the following processing steps are implemented: dividing the overall task process in a dynamic adversarial environment into discrete time series and determining each time step; within the current time step, each unmanned aerial vehicle (UAV) in the heterogeneous UAV cluster uses the constructed DCBGA model for target selection; during the target selection process, each UAV selects the target with the highest marginal benefit according to the latest situation information within its own perspective; within the current time step, each UAV uses the DCBGA model for a consensus process; during the consensus process, each UAV respectively conducts information interaction with adjacent UAVs directly connected by communication to complete situation information sharing and conflict resolution; when the number of rounds completed in the target selection and consensus process within the current time step reaches a set number of rounds, each UAV completes the update of its own local information, switches its own task execution state and motion state, and enters the next time step; following the target selection and consensus process of each time step, the constructed ACK network of the heterogeneous UAV cluster is output and updated; the ACK network is used to instruct each UAV to execute the task process for the selected target.

[0116] In one embodiment, when the computer program is executed by a processor, the steps or sub-steps added in each of the above embodiments of the method for constructing the ACK of the heterogeneous UAV cluster based on DCBGA can also be implemented.

[0117] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. 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 methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), memory bus dynamic random access memory (Rambus DRAM, abbreviated as RDRAM), and interface dynamic random access memory (DRDRAM), etc.

[0118] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, 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, it should be considered as within the scope described in this specification.

[0119] The above embodiments only represent several implementation manners of the present application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can still be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for constructing ACK for a heterogeneous UAV cluster based on DCBGA, characterized in that Including the steps: Dividing the overall task process in a dynamic adversarial environment into discrete time series and determining each time step; Within the current time step, each unmanned aerial vehicle (UAV) in the heterogeneous UAV cluster uses the constructed DCBGA model for target selection; During the target selection process, each of the UAVs selects the target with the highest marginal benefit according to the latest situation information within its own perspective; Within the current time step, each of the UAVs uses the DCBGA model for a consensus process; During the consensus process, each of the UAVs respectively conducts information interaction with adjacent UAVs directly connected by communication to complete situation information sharing and conflict resolution; When the number of rounds completed in the target selection and the consensus process within the current time step reaches a set number of rounds, each of the UAVs completes the update of its own local information, switches its own task execution state and motion state, and enters the next time step; Following the target selection and the consensus process of each time step, output and update the constructed ACK network of the heterogeneous UAV cluster; The ACK network is used to instruct each of the UAVs to execute the task process for the selected target.

2. The method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA according to claim 1, wherein The task execution states of the UAVs in the heterogeneous UAV cluster include no target traction, having target traction, and fixed target selection, and the motion states of the UAVs include serpentine maneuver, track flight, circumferential waiting, synchronous attack, and node failure.

3. The method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA according to claim 1 or 2, characterized in that, The process of the target selection includes: Each UAV selects and executes the subtask of the target with the largest marginal benefit increment for itself according to the latest situation information within its own perspective and adds the target to its own target selection list.

4. The method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA according to claim 3, wherein, The process of the target selection further includes: During the calculation of the marginal benefit, record the optimal task alliance combination of each target in the current UAV perspective; If the members of the optimal task alliance combination change, the current UAV conducts target selection, otherwise skip the execution of target selection.

5. The method for constructing ACK of heterogeneous UAV clusters based on DCBGA according to claim 3, characterized in that, The process of the target selection further includes: After the UAV enters the first set range from the target, the UAV sets its own task execution state to the fixed target selection state; the target selection range of the UAV is the targets within the second set range from the UAV itself.

6. The method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA according to claim 1, wherein The consensus process includes: Each of the UAVs updates the latest situation information by exchanging the target list and the list of failed UAVs; the latest situation information includes pop-up target information, target state changes, and UAV state changes; When each of the UAVs learns new target information, release the UAVs in the fixed target selection state and reset the task execution state to zero.

7. The method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA according to claim 6, wherein, The consensus process further includes: Each of the UAVs respectively conducts information interaction with adjacent UAVs directly connected by communication based on the communication timestamp vector to complete the conflict resolution of the target selection scheme.

8. The method for constructing an ACK for a heterogeneous UAV cluster based on DCBGA according to claim 1, wherein The method further includes: After the target selection and the consistency process at each of the time steps are completed, each of the drones separately resets the failure information and updates the valid information; the failure information reset includes the reset of the quotes for all targets and the planned flight trajectories by the drones, and the valid information update includes the update of the quotes and flight trajectories of the targets selected by the drones.

9. A heterogeneous UAV cluster ACK construction system based on DCBGA, characterized in that, It includes: A task division module, configured to divide the overall task process in a dynamic adversarial environment into discrete time series and determine each time step; A target selection module, configured to, within the current time step, instruct each drone in the heterogeneous drone cluster to perform target selection by using the constructed DCBGA model; each of the drones selects the target with the highest marginal benefit according to the latest situation information within its own perspective; A consistency module, configured to, within the current time step, instruct each of the drones to perform a consistency process by using the DCBGA model; During the consistency process, each of the drones separately exchanges information with adjacent drones connected directly through communication to complete the sharing of situation information and conflict resolution; A time switching module, configured to, when the number of rounds completed in the target selection and the consistency process within the current time step reaches a set number of rounds, instruct each of the drones to complete the update of their respective local information, the switching of their own task execution status and motion status, and enter the next time step; A network output module, configured to follow the target selection and the consistency process at each of the time steps, output and update the constructed ACK network of the heterogeneous drone cluster; The ACK network is used to instruct each of the drones to execute the task process for the selected target.

10. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for constructing an ACK of a heterogeneous drone cluster based on DCBGA according to any one of claims 1 to 8.

Images