A distributed unmanned aerial vehicle formation intelligent flight control cooperative scheduling system

Abstract

The application discloses a kind of distributed unmanned aerial vehicle formation intelligent flight control cooperative scheduling systems, including distributed ad hoc network communication module, global state perception module, cooperative scheduling decision module, flight control execution module, formation dynamic reconstruction module and fault protection module, whereby, the application is by distributed ad hoc network communication module to build centerless, multi-hop redundant formation communication network, can adapt formation topology dynamic change, node increase and decrease and complex electromagnetic interference airspace scene, rely on multi-agent reinforcement learning scheduling model, can be combined with global perception data, task demand and single machine performance parameter, each unmanned aerial vehicle terminal can be consistent by iterative communication to complete scheduling result confirmation, greatly improve the reliability of formation cooperative decision, flight control execution module built-in model predictive controller, in multi-machine collaborative operation process, guarantee the orderliness of multi-unmanned aerial vehicle formation cooperative flight.

Description

Technical Field

[0001] This application relates to the technical field of unmanned aerial vehicles (UAVs), and more particularly to a distributed UAV formation intelligent flight control and collaborative scheduling system. Background Technology

[0002] Unmanned aerial vehicle (UAV) swarm collaborative operations refer to multiple UAVs maintaining an orderly formation, communicating with each other, and cooperating under unified scheduling to jointly complete a practical task. Multiple UAV swarms can complete complex, large-scale, and highly challenging composite tasks through multi-node collaboration, effectively overcoming the limitations of single-UAV payload, endurance, and operational perspective. However, existing UAV swarm flight control and collaborative scheduling systems still have many shortcomings and are difficult to adapt to the needs of large-scale autonomous swarm operations in complex and dynamic airspace.

[0003] First, traditional drone formations often adopt centralized networking or simple wireless networking modes, which are highly dependent on the central control node. When the formation topology changes dynamically with the flight status, drone nodes are added or removed, or there is electromagnetic interference in the airspace, problems such as routing failure, communication interruption, and data transmission delay are very likely to occur.

[0004] Secondly, existing drone swarm scheduling mostly adopts fixed logic rules or centralized decision-making methods, which are difficult to adapt to the dynamic task scheduling needs of large-scale swarms. They are also difficult to autonomously complete distributed task decomposition and dynamic allocation. Furthermore, the lack of a decision consensus mechanism among drone nodes can easily lead to problems such as multi-drone task execution logic conflicts and inconsistent scheduling results.

[0005] Furthermore, traditional UAV flight control systems rely on precise single-unit control, adjusting attitude, speed, and flight path solely based on the individual UAV's own status, without incorporating constraints for coordinated formation flight. During multi-UAV collaborative operations, this can easily lead to significant deviations between the individual UAV's trajectory and the overall planned formation trajectory, resulting in chaotic formation configurations.

[0006] Application content

[0007] This application aims to address, at least to some extent, the technical problems in the related art.

[0008] To achieve the above objectives, this application proposes a distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system, including a distributed self-organizing network communication module, a global state perception module, a collaborative scheduling decision module, a flight control execution module, a formation dynamic reconstruction module, and a fault-tolerant protection module.

[0009] The distributed self-organizing network communication module enables point-to-point real-time data interaction between various UAV terminals, autonomously constructs a decentralized, multi-hop redundant formation communication network, and completes bidirectional synchronous transmission of flight status data, mission instructions, and environmental perception data.

[0010] The global state perception module collects the flight attitude, position coordinates, power status, and equipment operating condition data of the drone in real time. At the same time, it receives shared state data from other drone terminals in the formation, as well as external environmental obstacles and airspace situation data, to complete the formation's global state fusion perception and state anomaly identification.

[0011] The collaborative scheduling decision module includes a multi-agent reinforcement learning scheduling model, which autonomously completes distributed task decomposition, dynamic task allocation, flight trajectory collaborative planning, and formation timing scheduling based on the formation's global perception data, preset task requirements, and the performance parameters of each UAV terminal, thereby achieving autonomous collaborative decision-making for the formation.

[0012] The flight control execution module receives scheduling instructions and trajectory planning instructions output by the collaborative scheduling decision module, precisely controls the flight attitude, flight speed and flight path of the aircraft, synchronously matches the formation collaborative flight constraints, and completes the flight control execution operation.

[0013] The formation dynamic reconstruction module monitors abnormal operating conditions such as offline, fault, and airspace obstacle interference of the formation UAVs in real time. Based on the global status data, it quickly calculates the optimal formation reconstruction strategy and dynamically adjusts the formation topology, UAV flight positions and task assignments to achieve adaptive formation reconstruction.

[0014] The fault-tolerant protection module performs real-time monitoring and fault-tolerant processing of issues such as communication anomalies, data packet loss, single-machine failures, and scheduling command conflicts. Through data redundancy verification, command backup and resending, and fault node isolation mechanisms, it maintains the continuous execution of formation flight control scheduling operations.

[0015] In addition, the application may also include the following additional technical features:

[0016] Specifically, the distributed self-organizing network communication module adopts a wireless self-organizing network based on a dynamic source routing protocol, which supports UAV terminals to discover and maintain multi-hop redundant routes on demand, realizes fast route repair when the network topology changes, and adopts a hybrid access method of frequency division multiplexing and time division multiplexing to support parallel and conflict-free transmission of multiple nodes.

[0017] Specifically, the global state perception module includes an airborne differential GPS positioning unit, a nine-axis inertial measurement unit, a binocular vision sensor, and a millimeter-wave radar, and completes the spatiotemporal synchronous fusion of multi-source heterogeneous data through a lossless Kalman filter algorithm.

[0018] Specifically, the multi-agent reinforcement learning scheduling model adopts a deep reinforcement learning framework based on the multi-agent near-end policy optimization algorithm. Its state space includes the relative geometric configuration of the formation, task priority and energy constraints, and its action space includes speed increment, heading angle offset and task switching instructions.

[0019] Specifically, the collaborative scheduling decision module also has a built-in distributed consensus mechanism, in which each UAV terminal reaches a consensus on the task allocation result through a limited number of iterations of communication, and uses blockchain smart contracts to record scheduling decision logs.

[0020] Specifically, the formation dynamic reconstruction module includes a topology evaluation submodule and a reconstruction strategy generation submodule. The topology evaluation submodule quantifies the formation health in real time based on graph connectivity and elastic redundancy indicators. The reconstruction strategy generation submodule uses a mixed integer programming algorithm to calculate the optimal reconstruction scheme with the minimum adjustment cost.

[0021] Specifically, the fault-tolerant protection module includes a heartbeat monitoring unit, a data packet verification unit, and a fault isolation unit. The heartbeat monitoring unit sends a liveness signal to neighboring nodes at a dynamic adaptive cycle. The data packet verification unit uses cyclic redundancy check combined with forward error correction coding. The fault isolation unit dynamically blocks the control process of the faulty node through virtualization container technology.

[0022] Specifically, the flight control execution module has a built-in model predictive controller. Based on the deviation between the expected trajectory output by the cooperative scheduling decision module and the real-time state of the machine, it continuously optimizes and generates motor speed control quantities, and introduces formation cooperative constraints as hard constraints for model predictive control.

[0023] Specifically, the global state perception module also includes a dynamic obstacle behavior prediction submodule, which performs time-series prediction of the motion trajectory of nearby moving obstacles based on a long short-term memory network, and inputs the prediction results into the collaborative scheduling decision module for collision avoidance trajectory planning.

[0024] The beneficial effects of the distributed UAV formation intelligent flight control and collaborative scheduling system proposed in this application are as follows:

[0025] 1. A decentralized, multi-hop redundant formation communication network is constructed through a distributed self-organizing network communication module. Relying on the dynamic source routing protocol, it realizes autonomous route discovery, maintenance and rapid repair, and can adapt to dynamic changes in formation topology, addition and removal of nodes and complex electromagnetic interference airspace scenarios.

[0026] 2. Relying on a multi-agent reinforcement learning scheduling model, it can autonomously complete distributed task decomposition and dynamic task allocation by combining global perception data, task requirements and single-machine performance parameters, accurately adapting to large-scale formation dynamic scheduling operation scenarios. At the same time, it is equipped with a distributed consensus mechanism, and each UAV terminal can complete the consistency confirmation of scheduling results through iterative communication, which greatly improves the reliability of formation collaborative decision-making.

[0027] 3. The flight control execution module has a built-in model predictive controller, which can perform rolling optimization control based on the deviation between the scheduling instructions, the desired trajectory and the real-time state of the single drone. It incorporates the formation coordination constraints as hard constraints into the control algorithm. During multi-drone collaborative operations, it can correct the flight attitude, speed and flight path deviation of the single drone in real time, ensuring the orderly flight of multi-drone formations. Attached Figure Description

[0028] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0029] Figure 1 This is a flowchart of a distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to this application;

[0030] Figure 2 This is a flowchart illustrating the networking perception process of a distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to this application.

[0031] Figure 3 This is a flowchart illustrating the fault-tolerant reconfiguration process of a distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to this application. Detailed Implementation

[0032] To make the technical means, inventive features, objectives, and effects of this application easier to understand, the application is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0033] The present application will now be described in further detail with reference to the accompanying drawings.

[0034] like Figures 1-3 As shown in the figure, a distributed UAV formation intelligent flight control and collaborative scheduling system according to an embodiment of this application includes a distributed self-organizing network communication module, a global state perception module, a collaborative scheduling decision module, a flight control execution module, a formation dynamic reconstruction module, and a fault tolerance protection module.

[0035] Among them, the distributed self-organizing network communication module realizes point-to-point real-time data interaction between various UAV terminals, autonomously constructs a decentralized, multi-hop redundant formation communication network, and completes bidirectional synchronous transmission of flight status data, mission instructions, and environmental perception data.

[0036] It should be noted that each drone in the formation is defined as an independent and equal network node. In response to issues such as obstructed direct communication between two drones, signal attenuation, and link interruption caused by factors such as forest cover, building barriers, long-distance flight, and electromagnetic interference, the system can automatically search for other normally online drones in the formation as relay nodes. Through a multi-hop forwarding mechanism, data is relayed and multiple redundant communication links are dynamically constructed to prevent communication interruptions caused by single-point link failures.

[0037] The network supports dynamic UAV network entry and exit, and adaptive adaptation to dynamic location changes. No manual configuration is required when adding or removing nodes or adjusting formations; the network topology can be updated automatically. All flight statuses, mission commands, environmental information, and equipment data can be synchronously transmitted bidirectionally across the entire network, enabling real-time data sharing within the formation and effectively ensuring the autonomy, resilience, and stability of communication within the formation.

[0038] The global state perception module collects real-time data on the flight attitude, position coordinates, battery status, and equipment operating conditions of the drone itself. At the same time, it receives shared state data from other drone terminals in the formation, as well as data on external environmental obstacles and airspace situation, to complete the formation's global state fusion perception and state anomaly identification.

[0039] It should be noted that this system is based on a multi-source data fusion logic that integrates local data collection, cross-machine data sharing, and external environment monitoring. The drone collects its own basic operational data through attitude sensors, positioning modules, power detection units, and equipment condition sensors. It also receives shared data from all drones in the formation via a communication module, and simultaneously accesses airspace and obstacle detection data. All data is aggregated, compared, and analyzed to achieve comprehensive status integration. Based on preset thresholds, it identifies issues such as abnormal attitudes, equipment malfunctions, and dangerous airspaces, thus achieving comprehensive perception and initial anomaly assessment.

[0040] The collaborative scheduling decision module includes a multi-agent reinforcement learning scheduling model. Based on the formation's global perception data, preset task requirements, and the performance parameters of each UAV terminal, it autonomously completes distributed task decomposition, dynamic task allocation, collaborative flight trajectory planning, and formation timing scheduling, thereby achieving autonomous collaborative decision-making for the formation.

[0041] It should be noted that a multi-agent collaborative decision-making system is constructed, treating each drone as an independent intelligent agent. The model takes as input data dimensional from the formation's overall situational awareness, overall mission objectives, individual drone flight performance, remaining battery power, equipment status, and airspace environmental constraints. It then uses deep reinforcement learning algorithms for real-time iterative deduction and global optimal solution calculation. For large-scale, complex, and all-domain formation operation tasks, the system can autonomously decompose the overall macro-task into multiple sub-tasks adapted to different drones.

[0042] By dynamically allocating differentiated tasks based on the real-time status and performance advantages of each drone, the system avoids assigning high-load tasks to drones that are malfunctioning, low in battery, or in abnormal operating conditions. Simultaneously, it precisely plans a dedicated flight path for each drone, strictly avoiding airspace obstacles and no-fly zones. The system uniformly schedules the mission execution sequence, flight rhythm, and task allocation of all drones. Without any human intervention, it achieves efficient, orderly, and collaborative intelligent decision-making and scheduling for the entire formation, maximizing the overall operational efficiency of the formation and improving mission execution efficiency and completion quality.

[0043] The flight control execution module receives scheduling instructions and trajectory planning instructions output by the collaborative scheduling decision module, precisely controls the flight attitude, flight speed and flight path of the aircraft, synchronously matches the formation collaborative flight constraints, and completes the flight control execution operation.

[0044] It should be noted that after receiving trajectory parameters, scheduling instructions, timing requirements, and task instructions from the upper-level decision-making module, the module first completes instruction decoding, format verification, and compliance judgment, eliminating erroneous, invalid, and conflicting instructions. Then, it converts the standardized instructions into high-precision flight control electrical signals to precisely drive the UAV servo motor, power motor, and attitude adjustment unit, and adjust the UAV's pitch, roll, yaw, and other flight attitudes in real time to match the preset flight speed and planned route.

[0045] Throughout the flight, the module continuously links with the global status perception data to verify in real time the matching degree between the drone's flight actions and the overall formation, flight rhythm, coordination spacing, and operation sequence. It dynamically fine-tunes flight parameters to effectively avoid formation chaos and coordination failure caused by single-drone flight deviations, ensuring that the flight control actions of each drone are highly consistent with the overall operation requirements of the formation, achieving precise execution of individual drones and overall coordinated synchronization of the formation.

[0046] The formation dynamic reconfiguration module monitors abnormal operating conditions such as offline, fault, and airspace obstacle interference of formation drones in real time. Based on the global status data, it quickly calculates the optimal formation reconfiguration strategy and dynamically adjusts the formation topology, drone flight position and task division to achieve adaptive formation reconfiguration.

[0047] It should be noted that when abnormal nodes or environmental interference are detected, the system immediately integrates all UAV performance, location, and mission status data, and uses iterative optimization algorithms to quickly calculate the globally optimal formation adjustment scheme, simultaneously completing the following reconstruction actions: First, it dynamically updates the formation communication network topology, removes network links from faulty nodes, and reconstructs the communication network of normal nodes; second, it adjusts the flight positions and spatial relative positions of each normal UAV to quickly restore the formation; and third, it re-distributes the unfinished sub-tasks of the faulty nodes, evenly distributing the remaining tasks to each normal UAV. Through comprehensive adaptive adjustments, the overall formation structure, network mode, and task allocation automatically adapt to sudden changes in the field, thus maintaining the normal operation of the formation and significantly improving the stability of formation operations in complex dynamic scenarios.

[0048] The fault-tolerant protection module monitors and handles issues such as communication anomalies, data packet loss, single-machine failures, and scheduling command conflicts in real time. Through data redundancy verification, command backup and resending, and fault node isolation mechanisms, it maintains the continuous execution of formation flight control scheduling operations.

[0049] It should be noted that real-time inspections detect operational risks such as communication interruptions, data loss, single-machine failures, and command conflicts. Corresponding fault-tolerance mechanisms are activated for different problems. Error data is corrected through data redundancy verification, command loss is resolved by command backup and resending, and the impact of faulty equipment on the entire formation is cut off by isolating faulty nodes. This avoids the spread of faults at the communication, data, equipment, and command levels, ensuring continuous system operation.

[0050] In one embodiment of this application, the distributed ad hoc network communication module adopts a wireless ad hoc network based on a dynamic source routing protocol, which supports UAV terminals to discover and maintain multi-hop redundant routes on demand, realizes fast route repair when the network topology changes, and adopts a hybrid access method of frequency division multiplexing and time division multiplexing to support parallel and conflict-free transmission of multiple nodes.

[0051] It should be noted that this network architecture can support each UAV terminal to autonomously complete neighbor node detection and link quality assessment, realize on-demand route discovery and dynamic maintenance of multi-hop redundant routing links. Compared with traditional fixed routing protocols, it can perfectly adapt to complex scenarios such as frequent changes in formation topology, temporary wireless link obstruction and dynamic entry and exit of nodes in the network during UAV flight. It can complete millisecond-level fast route repair under abnormal conditions such as link interruption and node offline, ensuring the continuity and stability of formation communication links.

[0052] The module innovatively adopts a hybrid access method combining frequency division multiplexing (FDM) and time division multiplexing (TDM). Combining the channel isolation advantages of FDM with the time slot scheduling advantages of TDM, it performs dual fine-grained division of the communication channel in terms of both frequency and time slot. This allocates independent communication resources to UAV nodes in different locations and for different tasks, effectively avoiding signal interference, data conflicts, and channel congestion caused by simultaneous transmission from multiple nodes. It fully supports parallel conflict-free data transmission from multiple UAV nodes, significantly improving the overall communication throughput and real-time performance of the formation, and meeting the high-frequency communication requirements for formation collaborative perception, task interaction, and command transmission.

[0053] In one embodiment of this application, the global state perception module includes an airborne differential GPS positioning unit, a nine-axis inertial measurement unit, a binocular vision sensor, and a millimeter-wave radar, and completes the spatiotemporal synchronous fusion of multi-source heterogeneous data through a lossless Kalman filter algorithm.

[0054] It should be noted that the airborne differential GPS positioning unit can achieve centimeter-level absolute positioning of the UAV, correcting the positioning deviation of traditional civilian GPS and ensuring the accuracy of the overall position of the formation; the nine-axis inertial measurement unit can collect the UAV's three-axis acceleration, three-axis angular velocity, and three-axis attitude angle data at high frequency, and provide real-time feedback on the UAV's own flight attitude and motion status, making up for the deficiency of low satellite positioning update frequency; the binocular vision sensor has the ability to acquire environmental images, identify obstacles at close range, and calculate relative positions, and can acquire scene texture and spatial three-dimensional information; the millimeter-wave radar is not affected by adverse environments such as light, rain, snow, and fog, and has all-weather ranging, speed measurement, and angle measurement capabilities, and can accurately perceive the status of distant obstacles and surrounding targets.

[0055] This module employs a lossless Kalman filter algorithm to perform time synchronization calibration, spatial coordinate unification, noise filtering, and data fusion optimization on multi-source heterogeneous sensing data, maximizing the retention of effective sensing information and achieving high-precision spatiotemporal synchronization fusion of state data across the entire domain.

[0056] In one embodiment of this application, the multi-agent reinforcement learning scheduling model adopts a deep reinforcement learning framework based on the multi-agent proximal policy optimization algorithm. Its state space includes the relative geometric configuration of the formation, task priority and energy constraints, and its action space includes speed increment, heading angle offset and task switching instructions.

[0057] It should be noted that, leveraging the advantages of multi-agent near-end policy optimization algorithms, collaborative training, distributed decision-making, and real-time policy iteration among multiple UAV agents can be achieved, balancing individual autonomous decision-making with global optimization of the formation. The model's state space deeply integrates formation operation constraints, primarily including relative geometric configuration, task priority, and energy constraints. The relative geometric configuration encompasses spatial parameters such as UAV spacing, formation topology, and relative positional deviation. Task priority includes operational parameters such as task urgency, task weight, and task timing requirements. Energy constraints include state parameters such as remaining battery power, flight time, and energy consumption rate for each UAV, comprehensively covering the real-time status of the environment, tasks, and equipment for formation scheduling.

[0058] The corresponding action space precisely matches the formation control requirements, including three types of actions: speed increment, heading angle offset, and mission switching commands. This enables fine-tuning of UAV flight status, dynamic adaptation of formation, and dynamic switching of multi-UAV missions. Through continuous interaction with the environment and iterative optimization of strategies, the model can autonomously adapt to dynamic scenarios such as mission changes, topology adjustments, and energy differences, achieving intelligent and optimal scheduling of formation missions.

[0059] In one embodiment of this application, the collaborative scheduling decision module also incorporates a distributed consensus mechanism, whereby each UAV terminal reaches a consensus on the task allocation result through a limited number of iterations of communication, and uses a blockchain smart contract to record the scheduling decision log.

[0060] It should be noted that during the formation coordinated scheduling process, each UAV terminal acts as an independent decision-making node. Based on local perception data and global scheduling rules, after completing the initial task allocation scheme calculation, it completes data interaction, scheme comparison and deviation correction between nodes through a limited number of iterative communications, and quickly achieves consensus confirmation on the global task allocation results, which greatly improves decision-making efficiency and system fault tolerance.

[0061] Meanwhile, the module innovatively introduces blockchain smart contract technology to record log data such as task allocation results, decision time, participating nodes, and execution instructions for each formation scheduling decision in real time on the blockchain. Leveraging the decentralized, tamper-proof, and fully traceable characteristics of blockchain, it achieves visualized traceability and compliant retention of the scheduling decision process. This provides data support for post-operation task review, fault diagnosis, and responsibility determination, and also ensures the stable execution of scheduling decisions through the automatic fulfillment mechanism of smart contracts.

[0062] In one embodiment of this application, the formation dynamic reconstruction module includes a topology evaluation submodule and a reconstruction strategy generation submodule. The topology evaluation submodule quantifies the formation health in real time based on graph connectivity and elastic redundancy indicators, and the reconstruction strategy generation submodule uses a mixed integer programming algorithm to calculate the optimal reconstruction scheme with the minimum adjustment cost.

[0063] It should be noted that the topology assessment submodule is based on graph theory models, abstracting UAV nodes as network vertices and communication links as network edges. Based on two quantitative indicators, graph connectivity and elastic redundancy, it models and analyzes the connectivity integrity, link redundancy, and topology resilience of the formation network in real time, accurately quantifies the overall health of the formation, and can identify potential risks such as formation topology breaks, local node disconnection, and formation imbalance in real time, providing a quantitative basis for reconstruction decisions.

[0064] The reconfiguration strategy generation submodule aims to minimize the cost of formation adjustment. It combines multiple boundary conditions such as the current formation health status, remaining mission requirements, UAV energy consumption, and flight constraints, and uses a mixed integer programming algorithm to find the global optimum. It quickly calculates the optimal reconfiguration scheme with the minimum adjustment cost and accurately outputs reconfiguration instructions such as formation adjustment, node replacement, link reconstruction, and task reallocation. Under the premise of ensuring formation flight safety and mission continuity, it minimizes the energy consumption cost and motion loss of formation reconfiguration, and achieves fault self-healing and dynamic adaptation.

[0065] In one embodiment of this application, the fault-tolerant protection module includes a heartbeat monitoring unit, a data packet verification unit, and a fault isolation unit. The heartbeat monitoring unit sends a liveness signal to neighboring nodes at a dynamic adaptive cycle. The data packet verification unit uses cyclic redundancy check combined with forward error correction coding. The fault isolation unit dynamically blocks the control process of the faulty node through virtualization container technology.

[0066] It should be noted that the heartbeat monitoring unit adopts a dynamic adaptive heartbeat cycle mechanism. In the case of stable formation flight and low mission load, the heartbeat transmission cycle is extended to reduce communication energy consumption. In the case of formation maneuvering, high-density operation and complex environment, the monitoring cycle is shortened. The unit sends equipment survival signals to neighboring nodes at high frequency and monitors the online status and link connectivity of each node in real time. It can quickly identify abnormalities such as node offline, crash, and link interruption.

[0067] The data packet verification unit adopts a dual protection mechanism combining cyclic redundancy check and forward error correction coding. Cyclic redundancy check can accurately detect the loss, disorder, and tampering of transmitted data packets, while forward error correction coding can autonomously complete data error correction in scenarios with a small number of data transmission errors without retransmitting data, effectively ensuring the accuracy and integrity of cross-node data transmission in the formation.

[0068] The fault isolation unit relies on lightweight virtualization container technology to dynamically block and isolate the abnormal control processes of faulty drone nodes, distinguishing between normal business processes and faulty processes. It can effectively prevent the abnormal commands and erroneous data of faulty nodes from spreading in the formation network without shutting down the entire machine, thus greatly improving the fault tolerance and environmental adaptability of the drone formation.

[0069] In one embodiment of this application, the flight control execution module has a built-in model predictive controller. Based on the deviation between the expected trajectory output by the cooperative scheduling decision module and the real-time state of the machine, the motor speed control quantity is generated through rolling optimization, and formation cooperative constraints are introduced as hard constraints for model predictive control.

[0070] It should be noted that during the operation of the module, it receives target instructions such as the expected flight trajectory, formation position, and mission attitude from the collaborative scheduling decision module in real time, and simultaneously collects the real-time flight status data such as position, speed, attitude, and energy consumption of the machine, and continuously compares the deviation between the expected trajectory and the real-time status of the machine.

[0071] The model predictive controller is based on the UAV dynamics model to predict the flight status in the near future. It uses a rolling optimization algorithm to iteratively solve for the optimal control quantity in real time, accurately generate the UAV motor speed control quantity, and realize dynamic correction of the flight trajectory.

[0072] Meanwhile, this module innovatively uses formation coordination constraints as hard constraints for model predictive control, embedding global constraints such as safe distance between UAVs, formation topology restrictions, and coordinated flight speed matching into the control algorithm. While ensuring that individual UAVs accurately track target trajectories, it strictly avoids problems such as formation collisions, formation disorder, and coordination mismatch caused by individual UAV autonomous adjustments, achieving dual protection of precise individual UAV control and stable formation coordination.

[0073] In one embodiment of this application, the global state perception module further includes a dynamic obstacle behavior prediction submodule, which performs time-series prediction of the motion trajectory of nearby moving obstacles based on a long short-term memory network, and inputs the prediction results into the collaborative scheduling decision module for collision avoidance trajectory planning.

[0074] It should be noted that this submodule is supported by a long short-term memory network algorithm, which fully leverages the advantages of long short-term memory network in extracting features from time-series data and predicting dynamic trends. It can collect historical motion data of nearby moving obstacles in real time from millimeter-wave radar and binocular vision sensors, including the real-time position, speed, direction of motion, and attitude changes of the obstacles.

[0075] By using a pre-trained time-series prediction model, the system accurately predicts the trajectory, movement trend, and activity range of obstacles over a future period, outputting the predicted dynamic movement of the obstacles. The predicted obstacle behavior results are then input into the collaborative scheduling decision module in real time, providing advance prediction basis for global collision avoidance trajectory planning, formation adaptive adjustment, and node task avoidance. This effectively mitigates the collision risk between dynamic obstacles and the UAV formation, significantly improving the safety and operational stability of UAV formation flight in complex dynamic environments.

[0076] It should be noted that, in this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0077] The present application and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present application. The actual structure is not limited to this. In conclusion, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present application, such design should fall within the protection scope of the present application.

Claims

1. A distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system, characterized in that, It includes a distributed self-organizing network communication module, a global state awareness module, a cooperative scheduling decision-making module, a flight control execution module, a formation dynamic reconfiguration module, and a fault tolerance protection module, among which, The distributed self-organizing network communication module enables point-to-point real-time data interaction between each UAV terminal, autonomously constructs a decentralized, multi-hop redundant formation communication network, and completes bidirectional synchronous transmission of flight status data, mission instructions, and environmental perception data. The global state perception module collects the flight attitude, position coordinates, power status, and equipment condition data of the drone in real time. At the same time, it receives shared state data from other drone terminals in the formation, as well as external environmental obstacles and airspace situation data, to complete the formation's global state fusion perception and state anomaly identification. The collaborative scheduling decision module includes a multi-agent reinforcement learning scheduling model, which autonomously completes distributed task decomposition, dynamic task allocation, flight trajectory collaborative planning and formation timing scheduling based on the formation's global perception data, preset task requirements and the performance parameters of each UAV terminal, thereby realizing autonomous collaborative decision-making of the formation. The flight control execution module receives scheduling instructions and trajectory planning instructions output by the collaborative scheduling decision module, precisely controls the flight attitude, flight speed and flight path of the aircraft, synchronously matches the formation collaborative flight constraints, and completes the flight control execution operation; The formation dynamic reconstruction module monitors abnormal conditions such as offline, fault, and airspace obstacle interference of the formation UAVs in real time. Based on the global status data, it quickly calculates the optimal formation reconstruction strategy and dynamically adjusts the formation topology, UAV flight position and task division to achieve adaptive formation reconstruction. The fault-tolerant protection module performs real-time monitoring and fault-tolerant processing of issues such as communication anomalies, data packet loss, single-machine failures, and scheduling command conflicts. Through data redundancy verification, command backup and resending, and fault node isolation mechanisms, it maintains the continuous execution of formation flight control scheduling operations.

2. The distributed UAV formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The distributed self-organizing network communication module adopts a wireless self-organizing network based on a dynamic source routing protocol, which supports UAV terminals to discover and maintain multi-hop redundant routes on demand, realizes fast route repair when the network topology changes, and adopts a hybrid access method of frequency division multiplexing and time division multiplexing to support parallel and conflict-free transmission of multiple nodes.

3. The distributed UAV formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The global state perception module includes an airborne differential GPS positioning unit, a nine-axis inertial measurement unit, a binocular vision sensor, and a millimeter-wave radar. It achieves spatiotemporal synchronous fusion of multi-source heterogeneous data through a lossless Kalman filter algorithm.

4. The distributed UAV formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The multi-agent reinforcement learning scheduling model adopts a deep reinforcement learning framework based on the multi-agent near-end policy optimization algorithm. Its state space includes the relative geometric configuration of the formation, task priority and energy constraints, and its action space includes speed increment, heading angle offset and task switching instructions.

5. A distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The collaborative scheduling decision module also has a built-in distributed consensus mechanism. Each UAV terminal reaches a consensus on the task allocation result through a limited number of iterations of communication, and uses a blockchain smart contract to record the scheduling decision log.

6. A distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The formation dynamic reconstruction module includes a topology evaluation submodule and a reconstruction strategy generation submodule. The topology evaluation submodule quantifies the formation health in real time based on graph connectivity and elastic redundancy indicators. The reconstruction strategy generation submodule uses a mixed integer programming algorithm to calculate the optimal reconstruction scheme with the minimum adjustment cost.

7. A distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The fault-tolerant protection module includes a heartbeat monitoring unit, a data packet verification unit, and a fault isolation unit. The heartbeat monitoring unit sends a liveness signal to neighboring nodes at a dynamic adaptive cycle. The data packet verification unit uses cyclic redundancy check combined with forward error correction coding. The fault isolation unit dynamically blocks the control process of faulty nodes through virtualization container technology.

8. A distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The flight control execution module has a built-in model predictive controller. Based on the deviation between the expected trajectory output by the cooperative scheduling decision module and the real-time state of the machine, it continuously optimizes and generates motor speed control quantities, and introduces formation cooperative constraints as hard constraints for model predictive control.

9. A distributed unmanned aerial vehicle (UAV) formation intelligent flight control and collaborative scheduling system according to claim 1, characterized in that, The global state perception module also includes a dynamic obstacle behavior prediction submodule, which performs time-series prediction of the motion trajectory of nearby moving obstacles based on a long short-term memory network, and inputs the prediction results into the collaborative scheduling decision module for collision avoidance trajectory planning.

Images