An Adaptive Aircraft Cluster Control System and Method

Abstract

This invention relates to an adaptive aircraft swarm control system and method. Each aircraft in the swarm includes: an intelligent master control scheduling module, a tacit consensus module, a communication adaptive interaction module, a multi-source situational awareness module, and a multi-task coupled decision-making module. This invention employs a swarm control system mounted on each aircraft, enabling compatibility with both centralized and distributed collaborative models and achieving closed-loop collaboration of swarm tasks such as perception, decision-making, and communication. Through the offline construction and online updating of multi-source consensus information by the tacit consensus module, the consistency of swarm behavior and task collaboration are maintained even in weak communication environments such as localized, intermittent, and delayed communication. This invention effectively solves the challenges of collaborative decision-making, situational awareness sharing, and task execution in intelligent swarm systems under extreme communication interference conditions.

Description

Technical Field

[0001] This invention relates to the field of control system technology, and in particular to an adaptive aircraft swarm control system and method. Background Technology

[0002] With the deep integration of unmanned systems and artificial intelligence technologies, drone / vehicle swarm collaborative operations have become an important form of future intelligent applications. These swarm systems need to perform highly challenging tasks such as collaborative detection, path planning, and target allocation in complex, dynamic, and highly interference-prone environments. However, factors such as complex electromagnetic environments, communication interference and suppression, and high-speed platform maneuvers in real-world conditions can easily lead to deterioration in the quality of communication within the swarm, resulting in "weak communication" conditions such as sudden bandwidth drops, dramatic increases in latency, and even communication outages.

[0003] Traditional cluster collaborative control systems often employ a "centralized" or "pre-defined rule-based" architecture. Centralized architectures heavily rely on high-bandwidth, low-latency, and highly reliable communication links; if the central node fails or communication is disrupted, the overall cluster's collaborative capability will significantly decrease or even collapse. While pre-defined rule-based architectures have lower requirements for real-time communication, their rigid behavior makes them difficult to adapt to highly dynamic and uncertain environments, resulting in limited collaborative effectiveness.

[0004] In recent years, algorithms such as multi-agent reinforcement learning have provided new ideas for swarm intelligence collaboration. However, existing research faces systemic challenges under real-world weak communication conditions, such as incomplete sharing of perceptual information, inconsistent decision-making criteria, and mismatch between communication load and channel capacity. These challenges make it difficult to guarantee the robustness, adaptability, and task success rate of swarm collaboration.

[0005] Therefore, it is necessary to improve one or more of the problems existing in the above-mentioned related technical solutions.

[0006] It should be noted that this section is intended to provide background or context for the technical solutions of this disclosure as set forth in the claims. The description herein does not constitute an admission that it is prior art simply because it is included in this section. Summary of the Invention

[0007] The purpose of this invention is to provide an adaptive aircraft swarm control system and method, thereby overcoming, at least to some extent, one or more problems caused by the limitations and defects of related technologies.

[0008] This invention first provides an adaptive aircraft swarm control system, in which each aircraft in the swarm includes: an intelligent master control scheduling module, a tacit consensus module, a communication adaptive interaction module, a multi-source situational awareness module, and a multi-task coupled decision-making module;

[0009] Among them, the intelligent master control scheduling module is used to control the runtime sequence of the other four modules and schedule the information flow, and connect each module to the simulation platform to carry out offline training;

[0010] The tacit consensus module is used to provide knowledge data retrieval, model parameter initialization and knowledge injection support for the communication adaptive interaction module, multi-source situational awareness module and multi-task coupled decision module, while receiving updated data from the above modules to complete the knowledge base iteration;

[0011] The multi-source situational awareness module is used to fuse and detect multi-source detection information and identify targets. It transmits the perception results to the multi-task coupled decision module as the basis for decision-making, sends the perception results to the communication adaptive interaction module for inter-group interaction, and synchronizes the perception results to the tacit consensus module to complete the knowledge update.

[0012] The multi-task coupled decision module is used to generate the optimal task execution plan and action instructions based on the perception results of the multi-source situational awareness module, the knowledge data of the tacit consensus module, and the communication status parsed by the communication adaptive interaction module. The action instructions are then sent to the aircraft's actuators, and the decision information is sent to the communication adaptive interaction module for inter-group interaction. The decision information is also synchronized to the tacit consensus module to complete the knowledge update.

[0013] The adaptive communication interaction module is used to adaptively analyze the inter-group communication status under dynamic communication interference conditions, and transmit the communication status to the multi-task coupled decision module as decision constraints. At the same time, the communication status is synchronized to the tacit consensus module. The module also processes the received perception results and role information to complete the inter-group information interaction, and merges the peer information obtained from the interaction and synchronizes it to the tacit consensus module to complete the knowledge update.

[0014] The present invention further provides an adaptive aircraft swarm control method, which utilizes the swarm control system for control, and the control method includes the following steps:

[0015] S1. Pre-launch initialization: The intelligent master control scheduling module binds the basic knowledge related to the mission and the knowledge of cluster collaboration into the knowledge base of the tacit consensus module. Each module retrieves the basic parameters from the knowledge base to complete the initialization.

[0016] S2, Aircraft Takeoff: The intelligent main control scheduling module switches to inference mode, triggering the communication adaptive interaction module to run;

[0017] S3. Communication information interaction and knowledge update: The communication adaptive interaction module determines whether it has received peer interaction information. If it has received it, it will decode, reconstruct, and fuse the peer information and synchronize it to the knowledge base. If it has not received it, it will update its own data from the previous moment to the knowledge base, and at the same time parse the current communication status and transmit it to the multi-task coupling decision module.

[0018] S4. Coupled Task Decision: The multi-task coupled decision module generates the optimal decision scheme and action instructions based on knowledge base knowledge, communication state constraints and current situation, sends them to the aircraft actuators and synchronizes the decision information to the knowledge base, and sends the decision information to the communication adaptive interaction module.

[0019] S5. Multi-source situational awareness: If a search task is performed, the multi-source situational awareness module performs fusion detection and target recognition on multi-source detection information, synchronizes the perception results to the knowledge base, and sends the perception results to the communication adaptive interaction module.

[0020] S6. Inter-group information transmission: The communication adaptive interaction module determines the information to be interacted based on the current environment's effective communication bandwidth constraints, perception results, and decision information. It then calls upon the knowledge base to retrieve task-related knowledge, filters, compresses, and encodes it before sending it to other aircraft.

[0021] S7. Cyclic Execution: Under the scheduling of the intelligent master control scheduling module, S3-S6 are executed cyclically until the collaborative task is completed.

[0022] The technical solution provided by this invention may include the following beneficial effects:

[0023] The adaptive aircraft swarm control system and method of this invention adopts a configuration where one swarm control system is mounted on each aircraft, which is compatible with both centralized and distributed collaborative models, achieving closed-loop collaboration of swarm tasks such as perception, decision-making, and communication. Through offline construction and online updating of multi-source consensus information via a tacit consensus module, it maintains consistency in swarm behavior and task coordination even in weak communication environments such as local, intermittent, and delayed communication. This invention effectively solves the challenges of collaborative decision-making, situational awareness sharing, and task execution in swarm intelligent systems under extreme communication interference conditions. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0025] Figure 1 This is a diagram of the adaptive flight vehicle cluster control system architecture of the present invention;

[0026] Figure 2 This is a schematic diagram of the training process for the adaptive aircraft swarm control system of the present invention.

[0027] Figure 3This is a diagram showing the interface relationship between the adaptive flight vehicle swarm control system and the UAV system of this invention;

[0028] Figure 4 This is a flowchart illustrating the operational reasoning of the adaptive flight vehicle cluster control system of this invention.

[0029] Figure 5 This is a flowchart illustrating the internal operation of the multi-task coupled decision-making module of the present invention.

[0030] Figure 6 This is a flowchart illustrating the internal operation of the multi-source situational awareness module of the present invention.

[0031] Figure 7 The training loss convergence curve and reward value curve are for normal communication applications of this invention;

[0032] Figure 8 The training loss convergence curve and reward value curve for the communication degradation application of this invention are shown below.

[0033] Figure 9 The training loss convergence curve and reward value curve are for the communication intermittent application of this invention. Detailed Implementation

[0034] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0035] Furthermore, the accompanying drawings are merely illustrative diagrams of embodiments of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0036] This example implementation first provides an adaptive aircraft swarm control system, such as... Figure 1 and Figure 2 As shown, each aircraft in the cluster is equipped with a cluster control system, which includes: an intelligent master control scheduling module, a tacit consensus module, a communication adaptive interaction module, a multi-source situational awareness module, and a multi-task coupled decision-making module.

[0037] Among them, the intelligent master control scheduling module is used to control the runtime sequence of the other four modules and schedule the information flow, and connect each module to the simulation platform to carry out offline training;

[0038] The tacit consensus module is used to provide knowledge data retrieval, model parameter initialization and knowledge injection support for the communication adaptive interaction module, multi-source situational awareness module and multi-task coupled decision module, while receiving updated data from the above modules to complete the knowledge base iteration;

[0039] The multi-source situational awareness module is used to fuse and detect multi-source detection information and identify targets. It transmits the perception results to the multi-task coupled decision module as the basis for decision-making, sends the perception results to the communication adaptive interaction module for inter-group interaction, and synchronizes the perception results to the tacit consensus module to complete the knowledge update.

[0040] The multi-task coupled decision module is used to generate the optimal task execution plan and action instructions based on the perception results of the multi-source situational awareness module, the knowledge data of the tacit consensus module, and the communication status parsed by the communication adaptive interaction module. The action instructions are then sent to the aircraft's actuators, and the decision information is sent to the communication adaptive interaction module for inter-group interaction. The decision information is also synchronized to the tacit consensus module to complete the knowledge update.

[0041] The adaptive communication interaction module is used to adaptively analyze the inter-group communication status under dynamic communication interference conditions, and transmit the communication status to the multi-task coupled decision module as decision constraints. At the same time, the communication status is synchronized to the tacit consensus module. The module also processes the received perception results and role information to complete the inter-group information interaction, and merges the peer information obtained from the interaction and synchronizes it to the tacit consensus module to complete the knowledge update.

[0042] in, Figure 2 The "intelligent agent" in this context refers to the swarm control system of aircraft.

[0043] In this embodiment, a cluster control system is mounted on a single aircraft, which is compatible with both centralized and distributed collaborative models, enabling closed-loop collaboration of cluster tasks such as perception, decision-making, and communication. Through offline construction and online updating of multi-source consensus information via a tacit consensus module, consistency of cluster behavior and task coordination are maintained even in weak communication environments such as localized, intermittent, and delayed communication. This invention effectively solves the challenges of collaborative decision-making, situational awareness sharing, and task execution in cluster intelligent systems under extreme communication interference conditions.

[0044] The specific composition and function of each module in the above embodiments are described below.

[0045] (1) The intelligent master control scheduling module serves as the "scheduling center" of the cluster control system. It is responsible for the timing control, information flow scheduling and distribution of each functional module during online task execution, and supports the algorithm models within each module to conduct offline training through the simulation platform.

[0046] The intelligent master control scheduling module has the following two working modes:

[0047] Training Mode: This mode distributes current situational information from the simulation platform to the cluster control system and sends control commands for the cluster algorithm model's perception / decision / communication platforms to the simulation platform. In training mode, the cluster control systems of all aircraft are connected to the simulation platform, and all intelligent master control scheduling modules undergo unified training.

[0048] Inference mode: Used to perform real-time scheduling and information flow distribution of each module of the cluster control system according to the environmental situation during task execution, accurately allocating situational awareness information, communication interaction information, decision-making information and prior knowledge information to each functional module.

[0049] (2) The tacit consensus module serves as the system's "memory and common sense center," used to store, update, and extract multi-source, multi-task knowledge data, including basic mission knowledge of aircraft and targets, parameters of collaborative algorithm models under different communication levels, historical situation characteristics, pre-trained action plans, etc., providing knowledge injection and model initialization support for other modules.

[0050] The tacit consensus module includes: knowledge base, communication consensus model, perception consensus model, and decision consensus model.

[0051] Specifically, the knowledge base is used to store basic knowledge of aircraft missions and cluster collaboration knowledge, receive updated data from the communication consensus model, perception consensus model and decision consensus model, and complete online iteration of knowledge across all dimensions.

[0052] The communication consensus model is used for real-time parsing of inter-group communication rates and extracts communication-related knowledge and algorithm model parameters from the knowledge base according to the current communication status and the data to be transmitted. It supports communication compression / reconstruction model initialization and knowledge injection, and transmits communication rate data to the communication adaptive interaction module and the multi-task coupled decision module. It also receives communication interaction data from the communication adaptive interaction module to complete the communication knowledge update of the knowledge base.

[0053] The perception consensus model is used to retrieve target knowledge information and perception model parameters from the knowledge base based on the detection area, provide perception model initialization and knowledge injection for the multi-source situational awareness module, and receive the perception results from the multi-source situational awareness module to complete the perception knowledge update of the knowledge base.

[0054] The decision consensus model is used to retrieve decision algorithm model parameters and task-related knowledge from the knowledge base according to the current communication status, providing decision model initialization and knowledge injection for the multi-task coupled decision module, and simultaneously receiving decision information from the multi-task coupled decision module to update the decision knowledge in the knowledge base.

[0055] (3) The multi-source situational awareness module, as the "eyes" of the system, is responsible for the fusion detection and processing of multi-source information acquired by heterogeneous sensors such as infrared, visible light, and synthetic aperture radar (SAR) on the simulation platform, and for target identification, thereby improving the detection accuracy and robustness in complex environments.

[0056] (4) The multi-task coupling decision module serves as the "control center" of the system. It is responsible for integrating the current situation, communication conditions and knowledge base information to make integrated decisions on coupled tasks such as trajectory planning, search, target allocation and formation reconstruction, and generate the optimal action instruction sequence.

[0057] (5) The adaptive communication interaction module serves as the "adaptive throat" for the system's interaction with the outside world. Under dynamically changing communication conditions, it is responsible for intelligently filtering, hierarchically compressing, reliably encoding / decoding, high-fidelity reconstruction, and multi-source fusion of information that needs to be shared among groups, ensuring that key information is transmitted preferentially and reliably under limited bandwidth.

[0058] The adaptive communication interaction module includes:

[0059] The sending end includes an information filtering model, an information compression model, and a communication coding model, which are responsible for prioritizing, compressing, and coding the information to be transmitted according to the current communication status.

[0060] The receiving end, including the communication decoding model, information reconstruction model, and information fusion model, is responsible for decoding and reconstructing the received data, and updating it into the knowledge base after filtering and fusion, based on its own previous observation and action information.

[0061] This example embodiment further provides an adaptive aircraft swarm control method, which utilizes the swarm control system for control. The control method includes the following steps:

[0062] S1. Pre-launch initialization: The intelligent master control scheduling module binds the basic knowledge related to the mission and the knowledge of cluster collaboration into the knowledge base of the tacit consensus module. Each module retrieves the basic parameters from the knowledge base to complete the initialization.

[0063] S2, Aircraft Takeoff: The intelligent main control scheduling module switches to inference mode, triggering the communication adaptive interaction module to run;

[0064] S3. Communication information interaction and knowledge update: The communication adaptive interaction module determines whether it has received peer interaction information. If it has received it, it will decode, reconstruct, and fuse the peer information and synchronize it to the knowledge base. If it has not received it, it will update its own data from the previous moment to the knowledge base, and at the same time parse the current communication status and transmit it to the multi-task coupling decision module.

[0065] S4. Coupled Task Decision: The multi-task coupled decision module generates the optimal decision scheme and action instructions based on knowledge base knowledge, communication state constraints and current situation, sends them to the aircraft actuators and synchronizes the decision information to the knowledge base, and sends the decision information to the communication adaptive interaction module.

[0066] S5. Multi-source situational awareness: If a search task is performed, the multi-source situational awareness module performs fusion detection and target recognition on multi-source detection information, synchronizes the perception results to the knowledge base, and sends the perception results to the communication adaptive interaction module.

[0067] S6. Inter-group information transmission: The communication adaptive interaction module determines the information to be interacted based on the current environment's effective communication bandwidth constraints, perception results, and decision information. It then calls upon the knowledge base to retrieve task-related knowledge, filters, compresses, and encodes it before sending it to other aircraft.

[0068] S7. Cyclic Execution: Under the scheduling of the intelligent master control scheduling module, S3-S6 are executed cyclically until the collaborative task is completed.

[0069] The data processing procedure of the cluster control system for multi-source fusion identification tasks includes:

[0070] T1, the multi-source situational awareness module determines in real time whether it has acquired the detection image of the aircraft. If so, the perception process of the multi-source situational awareness module is started.

[0071] T2: The perception consensus model queries the knowledge base based on the location of the detection area to determine whether there is relevant information in the same area. If it exists, proceed to T3; otherwise, proceed to T4.

[0072] T3 extracts multi-source fusion detection and recognition model parameters and knowledge information of the same detection area from the knowledge base. After initializing the network, it combines the knowledge information with the real-time detection images of the aircraft to perform fusion recognition and obtain the target type and location information of the image in the detection area.

[0073] T4 extracts target detection and recognition model parameters from the knowledge base, initializes the network, and independently identifies the real-time detection images of the aircraft to obtain target type and location information.

[0074] T5 sorts the target information by importance, analyzes the current effective communication rate and determines the data to be transmitted, and performs data compression.

[0075] T6, after communication encoding, sends the target information to other aircraft, repeating the aforementioned steps until the perception mission stops.

[0076] The cluster control system's data processing procedure for the integrated decision-making task of target search and allocation includes:

[0077] J1, the multi-task coupled decision-making module starts the decision-making process after obtaining the current situation information;

[0078] J2 determines the distance between the aircraft's position and the target area or the target recognition status. If the conditions are met, the decision consensus model is invoked. Based on the current communication rate, the decision model parameters and related knowledge are retrieved from the knowledge base and allocated for initialization.

[0079] J3, the decision consensus model outputs the type of decision task to be executed. If it is a search task, then execute J4; if it is an assignment task, then execute J5.

[0080] J4 runs the collaborative search model to generate search action instructions, sends all action instructions to the execution mechanism, sends them to the information filtering model of the communication adaptive interaction module, and synchronizes them to the knowledge base;

[0081] J5 runs the target allocation model, combines mission knowledge, generates target allocation results online, and controls the aircraft to execute action commands;

[0082] J6 sorts the action commands by importance, parses the current communication rate and determines the data to be transmitted, and performs data compression;

[0083] J7, after communication encoding, sends action commands to other aircraft, repeating the aforementioned steps until the decision-making task stops.

[0084] In this embodiment, a consensus-driven mechanism is used to pre-launch and online update the cluster's shared knowledge base (including environmental cognition, task understanding, and collaborative strategies) through a tacit consensus module. This enables each cluster control system (agent) to "predict" or "understand" the intentions of its peers and the overall task situation based on local consensus when communication is not smooth, thereby making coordinated individual decisions.

[0085] Utilizing an integrated sensing, control, and transmission mechanism, the three stages of perception, decision-making (control), and communication are deeply coupled and operate in a closed loop. Perception results drive decision-making and determine communication content; communication status constrains decision-making options and influences information transmission strategies; decision-making results are then synchronized through communication and guide the next stage of perception focus. The adaptive communication interaction module, multi-source situational awareness module, and multi-task coupled decision-making module all receive knowledge guidance from the tacit consensus module and can adaptively adjust their working modes based on real-time communication bandwidth (communication consensus model parsing) (e.g., calling models with different compression ratios, selecting feature-level transmission instead of image-level transmission, etc.).

[0086] The training and inference of the cluster control system adopts a "centralized training-distributed execution" paradigm. During the offline training phase, a simulation platform is used for cluster simulation, accessed through an intelligent master control scheduling module. Deep learning and multi-agent reinforcement learning methods are employed to optimize the model parameters and strategies in the tacit consensus module. During the inference (execution) phase, each aircraft cluster control system interacts in real time based on its local consensus library, independently and in parallel generating decisions to achieve distributed collaboration.

[0087] The adaptive flight vehicle swarm control system and method of this application will be further described below through specific embodiments.

[0088] This embodiment describes a cluster control system applied to collaborative material transportation tasks using drone swarms. It illustrates a specific implementation of the system, using the scenario of a drone swarm performing a "discovery-identification-allocation" collaborative material transportation task for stranded disaster victims in a disaster area. The task faces environmental communication interference, and the internal communication bandwidth of the swarm is limited and dynamically changing.

[0089] 1. System Construction and Initialization

[0090] 1.1 Each UAV participating in the mission is equipped with a cluster control system as described in this application. In terms of hardware, the system runs on an onboard embedded intelligent computing payload (integrating CPU, GPU, etc.) software, such as... Figure 1 As shown, the five core functional modules are integrated in the manner described above.

[0091] 1.2 During the mission preparation phase, the cluster control system is connected to the UAV cluster collaborative simulation platform via the "training mode" interface of the intelligent master control scheduling module, such as... Figure 2 As shown. Using historical task data and simulation-generated data, deep learning and multi-agent deep reinforcement learning algorithms are employed to train the system offline. Training objectives include:

[0092] 1) Optimize the initialization content of the knowledge base in the tacit consensus module: such as disaster victim feature images, communication compression / reconstruction neural network parameters under different communication levels, and strategy network parameters for collaborative search and resource allocation, etc.

[0093] 2) Optimize the knowledge scheduling logic of the communication consensus model, perception consensus model, and decision-making consensus model;

[0094] 3) Optimize the information filtering model and the information priority evaluation criteria under different communication rates.

[0095] 1.3 After training, the trained model parameters, strategies, feature maps, etc. are used as cluster collaborative knowledge, and the UAV performance parameters, detection payload indicators, and disaster victims' facial / body features are used as basic mission knowledge. These are then bound to the tacit consensus module of each UAV cluster control system.

[0096] 1.4. The cluster control system (referred to as the "intelligent brain" in this application) is integrated into the UAV system, connected to the airborne flight control unit, airborne detection unit, data link, etc., to receive external information in real time and issue control commands to the detection payload and airborne control system based on autonomous decision-making results, and send inter-cluster interactive data to the airborne data link unit. The interface relationships are as follows: Figure 3 As shown.

[0097] 2. Online task execution (taking a drone as an example, its intelligent brain reasoning mode operation process is as follows) Figure 4 (As shown).

[0098] 2.1 Pre-launch planning: At the start of the mission, based on the initially set coordinates and range of the trapped disaster area, the multi-task coupling decision module is run to obtain the takeoff landing point, takeoff time, initial flight path and formation relationship. The information is then loaded into the tacit consensus module, and each functional module is activated. The UAV takes off as planned.

[0099] 2.2 Information Reception and Situation Update: The communication adaptive interaction module determines in real time whether it has received information from its companions via the data link.

[0100] 2.2.1 If an interactive message is received, the receiving end of the communication adaptive interaction module starts working, the communication decoding model verifies the data, the communication consensus model obtains the corresponding reconstruction model parameters and prior knowledge (features of similar targets) from the tacit consensus module according to the data type (such as compressed infrared feature map of disaster victims), the information reconstruction model parameters are initialized and high-fidelity data reconstruction is performed, and the information fusion model merges the reconstructed companion information with its own navigation, sensor status and other information from the previous moment to form a more complete local situation and updates it to the knowledge base;

[0101] 2.2.2 If no interaction information is received, the system will directly update its own navigation, sensor status and other information from the previous moment to the knowledge base.

[0102] 2.3 Collaborative Decision Generation: The multi-task coupled decision module runs continuously. The decision consensus model integrates the current position, target information in the knowledge base, and the current effective inter-group communication rate parsed by the communication consensus model to decide which control tasks to execute in the current stage, and generates action commands online to control the aircraft. For example... Figure 5 As shown, the internal operation flow of the multi-task coupled decision module is as follows:

[0103] 2.3.1 Real-time determination of whether the conditions for starting the cluster collaborative decision-making task are met at the current moment:

[0104] 2.3.1.1 The search mission is initiated when the distance between the spacecraft's location and the center of the disaster area does not exceed the maximum effective range of the detection payload.

[0105] 2.3.1.2 The condition for initiating the material distribution task is that the location of disaster victims has been identified;

[0106] 2.3.1.3. The conditions for initiating the trajectory planning mission are when an abnormal situation occurs, such as entering a communication-restricted area, discovering a new target or obstacle area, etc.

[0107] 2.3.1.4. The formation reconfiguration task is initiated when a drone in the cluster experiences a malfunction and returns to base.

[0108] 2.3.2 If any one of 2.3.1.1-2.3.1.4 is satisfied, the multi-task coupling decision module starts working, the decision consensus model determines which control tasks to execute at the current moment, and obtains the corresponding task model parameters and task-related knowledge from the tacit consensus module according to the current effective inter-group communication rate;

[0109] 2.3.3 Initialize the parameters of the corresponding control task model and generate action commands in sequence to control the UAV to execute;

[0110] 2.3.4 Send all action instructions to the information filtering model and update the knowledge base.

[0111] 2.4 Autonomous Sensing and Recognition: Once the local infrared detector begins imaging the designated area, the multi-source situational awareness module is activated. For example... Figure 6 As shown, the internal operating flow is as follows:

[0112] 2.4.1 Real-time determination of whether a local infrared detection image has been acquired at the current moment;

[0113] 2.4.2 If obtained, the perception consensus model queries the knowledge base to determine whether this detection area has its own historical acquisition or that of its peers has transmitted image-level, feature-level, or target-level information;

[0114] 2.4.3 If applicable, extract the multi-source fusion detection and recognition model parameters and knowledge information of the same detection area from the knowledge base, initialize the network, and then combine the knowledge with the real-time infrared detection image of the local machine to perform fusion recognition and obtain the target type and location information of the image in the detection area.

[0115] 2.4.4 If not, extract the target detection and recognition model parameters from the knowledge base, initialize the network, and then independently identify the real-time infrared detection image of the local machine to obtain the target type and location information;

[0116] 2.4.5. Send the infrared detection image and target type / location to the information filtering model and update the knowledge base.

[0117] 2.5 Information Transmission: The adaptive communication module determines in real time whether there are any updates to situation / decision information (such as recognition results, decision instructions, and its own status).

[0118] 2.5.1 If there is an information update, the communication adaptive interaction module starts working, the communication consensus model parses the current effective communication rate between groups, and the information filtering model dynamically prioritizes and prunes the data to be sent according to the communication rate constraint (for example, when the bandwidth is extremely low, only parameter-level data such as "target type + coordinates" or key instruction codes are sent, and when the bandwidth allows, image-level data can be sent). The communication consensus model obtains the corresponding compression model parameters and prior knowledge from the tacit consensus module according to the data type. The information compression model parameters are initialized and high-density data compression is performed. Finally, the data is encoded by the communication encoding model and sent.

[0119] 2.5.2 If there is no information update, skip 2.5.1.

[0120] 2.6. Cyclic Execution: Under the scheduling of the intelligent main control scheduling module, the system cyclically executes steps 2.2-2.5 until the task is completed or a termination instruction is received.

[0121] A typical communication-constrained scenario was constructed on a drone swarm collaborative simulation platform. Comparative experiments were conducted using a traditional centralized command system, a distributed multi-agent reinforcement learning system without a consensus database, and the consensus-driven integrated sensing, control, and transmission intelligent brain system described in this invention. The results show that, under scenarios with intermittent and strong communication interference, the system of this invention significantly outperforms the former two in terms of task completion rate, resource allocation coverage, and swarm survival rate, effectively verifying its ability to maintain efficient collaboration under weak communication conditions.

[0122] The cluster control system of this application was verified in a scenario. By conducting online collaborative flight tests of UAV clusters under communication degradation conditions, the tacit collaborative capability of the cluster was verified under conditions of communication rate reduction from the Kb level to the bit level and partial communication interruption.

[0123] The cluster control system of this application was tested at the flight test site, and four flight test site tests were completed under different communication bandwidths: normal / restricted, restricted / extremely low, and extremely low / zero. All tests were successful.

[0124] A performance analysis of the swarm control system (hereinafter referred to as the tacit intelligent brain) of this application was conducted under normal communication application conditions. The verification scenario was an online collaborative flight test of the UAV swarm under normal communication bandwidth (10Kbps). The current UAV swarm control system (hereinafter referred to as the communication collaborative brain), which relies on fully real-time communication, was used as the baseline for comparison. The analysis results are as follows: Figure 7As shown in Table 1. It should be noted that... Figure 7 , Figure 8 and Figure 9 The tacit cooperation model curve in the text represents the data results of the communication cooperation brain, and the communication cooperation model represents the data results of the communication cooperation brain.

[0125] Table 1. Results of Performance Analysis of Normal Communication Applications of the Synergistic Intelligent Brain and the Communication Collaboration Brain

[0126]

[0127] Figure 7 (a) shows the convergence curve of the training loss for normal communication applications; Figure 7 (b) shows the reward value curve for normal communication applications. Figure 7 It can be seen that both systems mentioned above can achieve model convergence under sufficient communication bandwidth. The cluster control system of this application has a faster training convergence speed with the assistance of cooperative task knowledge. As shown in Table 1, the cluster control system of this application successfully completed the predetermined test requirements, and the completion rate of the three cooperative tasks under normal communication reached 100%.

[0128] A communication degradation application performance analysis was conducted on the tacit intelligent brain of this application. The verification scenario was an online collaborative flight test of a UAV swarm under dynamic changes in communication bandwidth: normal / limited (10-1Kbps), limited / extremely low (1-0.1Kbps), and extremely low / zero (0.1-0Kbps). The communication collaborative brain was used as a baseline for comparison. The analysis results are as follows: Figure 8 As shown in Table 2.

[0129] Table 2. Results of Performance Analysis of Degraded Communication Applications by the Tacit Intelligence Brain and the Communication Collaboration Brain

[0130]

[0131] Figure 8 (a) shows the convergence curve of the training loss for communication degradation applications; Figure 8 (b) shows the reward value curve for communication degradation applications. Figure 8 As can be seen, the cluster control system of this application can still achieve model training convergence under the condition of dynamic degradation of communication bandwidth, while the communication collaboration brain cannot cope with the communication degradation environment. As shown in Table 2, the collaborative task completion rate of the cluster control system of this application has increased from 26.7% to 93.3%.

[0132] The intermittent communication application performance of the tacit intelligent brain in this application was analyzed. The verification scenario was an online collaborative flight test of a UAV swarm under intermittent communication (interval periods of 3s, 5s, and 10s). The communication collaborative brain was used as a baseline for comparison. The analysis results are as follows: Figure 9 As shown in Table 3.

[0133] Table 3. Results of Performance Analysis of Intermittent Communication Applications of the Tacit Intelligence Brain and the Communication Collaboration Brain

[0134]

[0135] Figure 9 (a) shows the convergence curve of the training loss for intermittent communication applications; Figure 9 (b) shows the reward value curve for intermittent communication applications. Figure 9 It can be seen that the cluster control system of this application can still achieve model training convergence under the condition of periodic communication interruption, while the communication collaboration brain cannot cope with intermittent communication environment. As shown in Table 3, the collaborative task completion rate of the cluster control system of this application has increased from 32.6% to 86.7%.

[0136] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0137] 1. Fundamental Enhancement of Knowledge-Driven Collaborative Robustness: This invention introduces a "tacit consensus module" as a distributed knowledge center. Through offline training and online updates, it encodes cluster collaboration strategies, environmental characteristics, and adversary models into shareable and invokeable knowledge. This enables each agent in the cluster to perform "mimicry inference" and "intent prediction" based on a local consensus library when facing communication interference. It autonomously generates individual behaviors consistent with the global task objective, achieving a paradigm shift from "communication-dependent collaboration" to "knowledge-driven collaboration," fundamentally improving the cluster's survivability and task resilience under extreme communication interference conditions.

[0138] 2. Deeply Integrated Closed-Loop Adaptive Capability of Sensing, Control, and Transmission: This invention deeply integrates sensing, decision-making, and communication into a closed-loop adaptive system through "consensus-driven" mechanisms. Sensing results, enhanced by consensus knowledge, drive decision-making; the multi-task coupled decision module considers the channel capacity constraints resolved by the communication consensus model in real time, generating matching instructions and communication requirements; the adaptive communication interaction module dynamically invokes the optimal compression and encoding strategies from the knowledge base for adaptive transmission based on decision priorities and the characteristics of the sensed information. This real-time bidirectional coupling and closed-loop optimization of "sensing-decision-communication" achieves dynamic optimal allocation of system resources, achieving an integrated intelligent effect of "transmitting the right information and driving the right action at the right time, with the right bandwidth."

[0139] 3. Intelligent and Value-Maximizing Communication Resource Utilization: Unlike traditional fixed compression or priority queues, the adaptive communication interaction module of this invention achieves content-aware intelligent communication under the guidance of consensus knowledge. It can dynamically decide the content and form of transmission based on task stage, target value, information novelty, and real-time channel conditions, ensuring that limited communication resources are always used to transmit information that maximizes global collaborative utility, thereby minimizing the impact of communication bottlenecks at the physical level.

[0140] 4. System Architecture Compatibility and Evolution: The distributed design of "one machine, one brain" and the modular consensus knowledge base enable the system to be compatible with various collaborative forms, from centralized guidance to distributed autonomy. Simultaneously, the consensus base supports offline training and online incremental updates, allowing the entire intelligent brain system to continuously optimize its knowledge model and policy base through continuous data accumulation and task execution, achieving autonomous evolution of collaborative capabilities and possessing long-term adaptability to cope with complex scenarios.

[0141] It should be noted that although several modules of the system for executing actions are mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more modules described above can be embodied in one module. Conversely, the features and functions of one module described above can be further divided into multiple modules for embodiment. Components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without any inventive effort.

[0142] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

[0143] It should be noted that the installation of image acquisition and personal identification equipment in public places involved in this application is necessary for maintaining public safety, complies with relevant national regulations, and is accompanied by prominent warning signs. The collected personal images and identification information can only be used for the purpose of maintaining public safety and will not be used for other purposes; or the images, personal identification data, etc. in this application are all legally and compliantly obtained or collected with the individual's separate consent.

Claims

1. An adaptive aircraft swarm control system, characterized by, Each aircraft in the cluster includes: an intelligent master control and scheduling module, a tacit consensus module, a communication adaptive interaction module, a multi-source situational awareness module, and a multi-task coupled decision-making module; Among them, the intelligent master control scheduling module is used to control the runtime sequence of the other four modules and schedule the information flow, and connect each module to the simulation platform to carry out offline training; The tacit consensus module is used to provide knowledge data retrieval, model parameter initialization and knowledge injection support for the communication adaptive interaction module, multi-source situational awareness module and multi-task coupled decision module, while receiving updated data from the above modules to complete the knowledge base iteration; The multi-source situational awareness module is used to fuse and detect multi-source detection information and identify targets. It transmits the perception results to the multi-task coupled decision module as the basis for decision-making, sends the perception results to the communication adaptive interaction module for inter-group interaction, and synchronizes the perception results to the tacit consensus module to complete the knowledge update. The multi-task coupled decision module is used to generate the optimal task execution plan and action instructions based on the perception results of the multi-source situational awareness module, the knowledge data of the tacit consensus module, and the communication status parsed by the communication adaptive interaction module. The action instructions are then sent to the aircraft's actuators, and the decision information is sent to the communication adaptive interaction module for inter-group interaction. The decision information is also synchronized to the tacit consensus module to complete the knowledge update. The adaptive communication interaction module is used to adaptively analyze the inter-group communication status under dynamic communication interference conditions, and transmit the communication status to the multi-task coupled decision module as decision constraints. At the same time, the communication status is synchronized to the tacit consensus module. The module also processes the received perception results and role information to complete the inter-group information interaction, and merges the peer information obtained from the interaction and synchronizes it to the tacit consensus module to complete the knowledge update.

2. The adaptive aircraft formation control system in accordance with claim 1, wherein, The tacit consensus module includes: a knowledge base, a communication consensus model, a perception consensus model, and a decision-making consensus model. The knowledge base is used to store basic knowledge of aircraft missions and cluster collaboration knowledge, receive updated data from the communication consensus model, perception consensus model and decision consensus model, and complete online iteration of knowledge across all dimensions. The communication consensus model is used for real-time parsing of inter-group communication rates and extracts communication-related knowledge and algorithm model parameters from the knowledge base according to the current communication status and the data to be transmitted. It supports communication compression / reconstruction model initialization and knowledge injection, and transmits communication rate data to the communication adaptive interaction module and the multi-task coupled decision module. It also receives communication interaction data from the communication adaptive interaction module to complete the communication knowledge update of the knowledge base. The perception consensus model is used to retrieve target knowledge information and perception model parameters from the knowledge base based on the detection area, provide perception model initialization and knowledge injection for the multi-source situational awareness module, and receive the perception results from the multi-source situational awareness module to complete the perception knowledge update of the knowledge base. The decision consensus model is used to retrieve decision algorithm model parameters and task-related knowledge from the knowledge base according to the current communication status, providing decision model initialization and knowledge injection for the multi-task coupled decision module, and simultaneously receiving decision information from the multi-task coupled decision module to update the decision knowledge in the knowledge base.

3. The adaptive aircraft formation control system of claim 2, wherein, The adaptive communication interaction module includes: The sending end includes an information filtering model, an information compression model, and a communication coding model, which are responsible for prioritizing, compressing, and coding the information to be transmitted according to the current communication status. The receiving end, including the communication decoding model, information reconstruction model, and information fusion model, is responsible for decoding and reconstructing the received data, and updating it into the knowledge base after filtering and fusion, based on its own previous observation and action information.

4. The adaptive aircraft formation control system in accordance with claim 3, wherein, The intelligent master control scheduling module has the following two working modes: Training mode: Used to distribute the current situation information sent by the simulation platform to the cluster control system, and send the cluster algorithm model's perception / decision / communication platform control commands to the simulation platform; Inference mode: Used to perform real-time scheduling and information flow distribution of each module of the cluster control system according to the environmental situation during task execution, accurately allocating situational awareness information, communication interaction information, decision-making information and prior knowledge information to each functional module.

5. A method of adaptive aircraft swarm control, characterized in that, Control is performed using the cluster control system described in claim 4, and the control method includes the following steps: S1. Pre-launch initialization: The intelligent master control scheduling module binds the basic knowledge related to the mission and the knowledge of cluster collaboration into the knowledge base of the tacit consensus module. Each module retrieves the basic parameters from the knowledge base to complete the initialization. S2, Aircraft Takeoff: The intelligent main control scheduling module switches to inference mode, triggering the communication adaptive interaction module to run; S3. Communication information interaction and knowledge update: The communication adaptive interaction module determines whether it has received peer interaction information. If it has received it, it will decode, reconstruct, and fuse the peer information and synchronize it to the knowledge base. If it has not received it, it will update its own data from the previous moment to the knowledge base, and at the same time parse the current communication status and transmit it to the multi-task coupling decision module. S4. Coupled Task Decision: The multi-task coupled decision module generates the optimal decision scheme and action instructions based on knowledge base knowledge, communication state constraints and current situation, sends them to the aircraft actuators and synchronizes the decision information to the knowledge base, and sends the decision information to the communication adaptive interaction module. S5. Multi-source situational awareness: If a search task is performed, the multi-source situational awareness module performs fusion detection and target recognition on multi-source detection information, synchronizes the perception results to the knowledge base, and sends the perception results to the communication adaptive interaction module. S6. Inter-group information transmission: The communication adaptive interaction module determines the information to be interacted based on the current environment's effective communication bandwidth constraints, perception results, and decision information. It then calls upon the knowledge base to retrieve task-related knowledge, filters, compresses, and encodes it before sending it to other aircraft. S7. Cyclic Execution: Under the scheduling of the intelligent master control scheduling module, S3-S6 are executed cyclically until the collaborative task is completed.

6. The adaptive aircraft formation control method of claim 5, wherein, The data processing procedure of the cluster control system for multi-source fusion identification tasks includes: T1, the multi-source situational awareness module determines in real time whether it has acquired the detection image of the aircraft. If so, the perception process of the multi-source situational awareness module is started. T2: The perception consensus model queries the knowledge base based on the location of the detection area to determine whether there is relevant information in the same area. If it exists, proceed to T3; otherwise, proceed to T4. T3 extracts multi-source fusion detection and recognition model parameters and knowledge information of the same detection area from the knowledge base. After initializing the network, it combines the knowledge information with the real-time detection images of the aircraft to perform fusion recognition and obtain the target type and location information of the image in the detection area. T4 extracts target detection and recognition model parameters from the knowledge base, initializes the network, and independently identifies the real-time detection images of the aircraft to obtain target type and location information. T5 sorts the target information by importance, analyzes the current effective communication rate and determines the data to be transmitted, and performs data compression. T6, after communication encoding, sends the target information to other aircraft, repeating the aforementioned steps until the perception mission stops.

7. The adaptive flight vehicle cluster control method according to claim 6, characterized in that, The cluster control system's data processing procedure for the integrated decision-making task of target search and allocation includes: J1, the multi-task coupled decision-making module starts the decision-making process after obtaining the current situation information; J2 determines the distance between the aircraft's position and the target area or the target recognition status. If the conditions are met, the decision consensus model is invoked. Based on the current communication rate, the decision model parameters and related knowledge are retrieved from the knowledge base for initialization. J3, the decision consensus model outputs the type of decision task to be executed. If it is a search task, then execute J4; if it is an assignment task, then execute J5. J4 runs the collaborative search model to generate search action instructions, sends all action instructions to the execution mechanism, sends them to the information filtering model of the communication adaptive interaction module, and synchronizes them to the knowledge base; J5 runs the target allocation model, combines mission knowledge, generates target allocation results online, and controls the aircraft to execute action commands; J6 sorts the action commands by importance, parses the current communication rate and determines the data to be transmitted, and performs data compression; J7, after communication encoding, sends action commands to other aircraft, repeating the aforementioned steps until the decision-making task stops.

Images