Decentralized cluster formation control method and system based on virtual particle control

Through virtual particle control and decentralization methods, the problem of UAV formations relying on leaders and communication restriction in the existing technology is solved, flexible formations and efficient collaboration are achieved, and the communication and computing capabilities of UAV formations are improved.

CN118778684BActive Publication Date: 2024-11-15BEIJING ANGFEI TECH CO LTD
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202411281521.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-13
Publication Date
2024-11-15
Estimated Expiration
2044-09-13

AI Technical Summary

Technical Problem

The existing cluster formation control strategies rely on leader drones, with limited communication quality and flight stability, making it difficult to achieve flexible formation configuration and efficient drone collaboration.

Method used

Using a decentralized method based on virtual particle control, through virtual reference particle and inter-aircraft communication, each drone calculates flight control instructions to form a dynamic formation, and optimizes the communication topology and Computational resource management.

Benefits of technology

It realizes flexible configuration changes and efficient coordination of the drone formation, improves communication stability and computing capabilities, and adapts to complex tasks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118778684B_ABST
    Figure CN118778684B_ABST
Patent Text Reader

Abstract

The present invention provides a decentralized cluster formation control method and system based on virtual particle control, which relates to the field of control technology, including planning a reference route for a cluster formation task by a cluster control terminal; virtual reference particles perform particle motion according to the reference route, align once every preset time according to virtual particle information between machines, and the virtual particle information between machines is transmitted through machine-to-machine communication; a roll control instruction obtained by an L1 control strategy based on an expected route, an altitude control instruction superimposed on a waypoint altitude expectation as an altitude expectation input according to a formation altitude offset, and a speed control instruction giving an aircraft speed expectation according to a current position relationship between a UAV and a reference point; a secondary reference point position is obtained by calculating the relative position offset of each UAV from a central reference point in a corresponding formation, and the UAV finally forms a desired formation by catching up with the secondary reference point position.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to control technology, and in particular to a decentralized cluster formation control method and system based on virtual particle control. Background Art

[0002] Cluster formation control technology, which forms different formation configurations by controlling multiple control nodes, has broad application prospects in the field of UAV cluster reconnaissance.

[0003] Most of the existing cluster formation control strategies have a leader aircraft, and the formation configuration is completed based on the reference position information of the leader aircraft. In this way, the UAV cluster formation is limited by the communication quality between the UAVs and the flight stability of the leader. Summary of the invention

[0004] The embodiments of the present invention provide a decentralized cluster formation control method and system based on virtual particle control, which can solve the problems in the prior art.

[0005] According to a first aspect of the embodiments of the present invention,

[0006] Provides a decentralized cluster formation control method based on virtual particle control, including:

[0007] The cluster control terminal plans a reference route for the cluster formation mission; the virtual reference particle moves according to the reference route, and is aligned once every preset time according to the virtual particle information between machines, and the virtual particle information between machines is transmitted through inter-machine communication;

[0008] The multiple UAVs refer to the virtual particle information to calculate their own flight control instructions respectively, and the flight control instructions include: a roll control instruction obtained by the L1 control strategy based on the expected route, an altitude control instruction superimposed on the waypoint altitude expectation as an altitude expectation input according to the formation altitude offset, and a speed control instruction that gives the aircraft speed expectation according to the current position relationship between the UAV and the reference point;

[0009] By calculating the relative position offset of each drone from the central reference point in the corresponding formation, the secondary reference point position is obtained, and the drone finally forms the required formation by catching up with the secondary reference point position; wherein, different formation configurations are achieved by changing the relative position offset; when the multiple drones form a formation, the following conditions are met: the central reference point position of all drones is consistent with the direction of movement, and the position of a single drone can follow the position of its secondary reference point.

[0010] In an optional embodiment,

[0011] The virtual reference particle moves according to the reference route, and according to the virtual particle information between machines, an alignment is performed every preset time, including:

[0012] The virtual reference particle moves according to the reference route planned by the cluster control terminal. The north and east positions of the single-machine reference point are determined by the target position and the offset in the corresponding direction respectively, and the plane position of the secondary reference point is determined by the plane position of the single-machine reference point and the plane offset in the corresponding direction. The virtual reference particle of each UAV transmits the local virtual particle information through inter-machine communication, receives the virtual particle information transmitted by other UAVs according to a preset round-robin topology structure, and performs position alignment once every preset time using the received inter-machine virtual particle information. After alignment, the virtual reference particle positions and movement directions of the multiple UAVs are consistent. In the round-robin topology structure, each UAV has the same number of topological nodes, and the topological nodes are arranged according to the UAV number. UAVs give priority to communicating with other UAVs with adjacent numbers.

[0013] In an optional embodiment,

[0014] The roll control command derived from the L1 control strategy based on the expected route, the altitude control command superimposed on the waypoint altitude expectation as the altitude expectation input according to the formation altitude offset, and the speed control command given the UAV speed expectation according to the current position relationship between the aircraft and the reference point include:

[0015] The UAV calculates a roll control instruction, an altitude control instruction and a speed control instruction according to the expected route and the virtual reference particle information; wherein, the L1 control strategy is used to calculate the heading error based on the expected route and the current aircraft position, and the heading error is input into the roll angle instruction generation module to obtain the roll control instruction; the formation formation altitude offset is superimposed on the route waypoint altitude planned by the cluster control terminal, and the superimposed altitude value is input into the altitude control module as the altitude expectation to obtain the altitude control instruction; the sight line vector is calculated according to the relative relationship between the current aircraft position and the corresponding virtual reference particle position, and the sight line vector is multiplied by the virtual particle expected velocity to obtain the aircraft velocity vector, and the aircraft velocity vector is input into the velocity control module to obtain the velocity control instruction.

[0016] In an optional embodiment,

[0017] By calculating the relative position offset of each UAV from the central reference point in the corresponding formation, the position of the secondary reference point is obtained, and the UAV catches up with the position of the secondary reference point to finally form the required formation, including:

[0018] The cluster control terminal calculates the relative position offset of each UAV from the formation center reference point according to the expected formation formation, superimposes the center reference point position and the relative position offset to obtain the secondary reference point position of each UAV; each UAV preferentially flies to the corresponding secondary reference point position, and when the distance between the UAV and the secondary reference point position is less than a preset threshold, it is considered that the UAV has reached the formation sequence, and the state is adjusted to formation holding. In the formation holding state, the UAV flies at a preset speed; after all UAVs enter the formation holding state, the formation formation is formed; wherein, by adjusting the relative position offset of each UAV, the formation formation is dynamically changed during the flight.

[0019] In an optional embodiment,

[0020] The method further comprises:

[0021] The drone collects its own flight status information and surrounding environment information through airborne sensors, and uses the collected status information and environmental information to establish a link quality estimation model. The link quality estimation model comprehensively considers the distance between drones, relative speed, and signal strength, and quantitatively evaluates the quality and stability of the communication link between drones to obtain a link quality evaluation value;

[0022] According to the link quality evaluation value, all communicable drones are sorted, and the first N drones with the highest link quality evaluation value are selected as candidate communication objects, where N is dynamically adjusted according to the communication demand and the number of drones;

[0023] The drone adjusts the connectivity of the topological structure according to the mission requirements and formation. For the formation whose mission requires information sharing exceeding the preset sharing threshold, the drone builds a fully connected mesh topology. For the formation whose mission requires group collaboration, the drone builds a clustered star topology. For the formation whose mission requires data aggregation, the drone builds a tree topology.

[0024] The UAV estimates the future position and speed of itself and other UAVs through the motion prediction algorithm, and based on the link quality estimation model, predicts the trend of topological structure changes and makes preparations for communication adjustments in advance;

[0025] After adjusting the communication topology structure according to the link quality estimation model, the UAV broadcasts the topology change information to all other UAVs in the formation. The topology change information includes the UAV number, change time, and the list of neighbor nodes after the change. After receiving the topology change information, other UAVs update the locally stored formation communication topology structure.

[0026] In an optional embodiment,

[0027] The UAV estimates the future position and speed of itself and other UAVs through the motion prediction algorithm, and based on the link quality estimation model, predicts the change trend of the topology structure including:

[0028] Based on its historical trajectory data, the UAV uses the Kalman filter algorithm to recursively estimate the position and speed of the UAV. Through the two steps of prediction and update, the current position and speed of the UAV are estimated in real time. Assuming that the UAV maintains uniform linear motion in the prediction time domain, the position coordinates of the UAV at different times in the future are predicted.

[0029] A link quality estimation model based on Q-learning is constructed, in which the distance, relative speed and signal strength between the two ends of the link are used as state quantities, the predicted value of link quality is used as action quantity, and the link quality prediction error is used as reward value. The optimal link quality prediction action in each state is obtained through iterative update of Q value.

[0030] After the UAV obtains its predicted coordinates and those of other UAVs at different moments in the future, it calculates the predicted distances between different UAVs and inputs the predicted relative positions and speeds into the trained link quality estimation model to obtain the predicted link quality values ​​at different moments in the future. By calculating the normalized Hamming distance between the predicted topology and the current topology, the degree of change in the future topology can be quantitatively evaluated.

[0031] In an optional embodiment,

[0032] The method further comprises:

[0033] Select some drones with sufficient computing and storage resources in the drone formation and deploy mobile edge computing servers. The mobile edge computing servers use virtualization and containerization technologies to achieve flexible scheduling and isolation of computing resources, and design mobile edge computing applications based on microservice architecture.

[0034] When selecting the location of the mobile edge computing server, the power, computing resources, and communication capabilities of the drone itself are comprehensively considered, and the mobile edge computing servers are required to be evenly distributed within the formation to avoid communication congestion and single point failure caused by excessive concentration of servers;

[0035] After the mobile edge computing server is started, it broadcasts its own server identification, service type, and service capabilities within the formation. After other drones receive the broadcast information, they register the broadcast information of the mobile edge computing server in the local service list;

[0036] When the UAV processes the task locally, if it finds that the computing or storage resources are insufficient, it will initiate a task offloading request to the mobile edge computing server. The task offloading request includes the task type, input data, and expected completion time. The UAV sends the task offloading request to the mobile edge computing server through a round-robin topology structure.

[0037] After receiving the task offloading request, the mobile edge computing server decides whether to accept the request based on the task type and available resources. If accepted, the task is assigned to local computing resources for execution, and the calculation results are fed back to the requesting node through the round-robin topology structure. If rejected, the requesting node is informed of the rejection reason.

[0038] According to its own communication capability and link quality, the mobile edge computing server relays a drone's data packet to a target drone that is farther away than a preset distance threshold through multi-hop forwarding or store-and-forwarding, thereby expanding the communication coverage of the drone formation and optimizing the relay communication performance through priority management and power control.

[0039] According to a second aspect of the embodiments of the present invention,

[0040] Provides a decentralized cluster formation control system based on virtual particle control, including:

[0041] The first unit is used for planning a reference route for a cluster formation task by a cluster control terminal; a virtual reference particle moves according to the reference route, and an alignment is performed once every preset time according to the virtual particle information between machines, and the virtual particle information between machines is transmitted through inter-machine communication;

[0042] The second unit is used for multiple drones to refer to the virtual particle information and calculate their own flight control instructions respectively, wherein the flight control instructions include: a roll control instruction obtained by the L1 control strategy based on the expected route, an altitude control instruction superimposed on the waypoint altitude expectation as an altitude expectation input according to the formation altitude offset, and a speed control instruction that gives the aircraft speed expectation according to the current position relationship between the drone and the reference point;

[0043] The third unit is used to obtain the secondary reference point position by calculating the relative position offset of each drone from the central reference point in the corresponding formation, and the drone finally forms the required formation by catching up with the secondary reference point position; wherein, different formation configurations are achieved by changing the relative position offset; when the multiple drones form a formation, it satisfies: the central reference point position of all drones is consistent with the movement direction, and the position of a single drone can follow the position of its secondary reference point.

[0044] According to a third aspect of the embodiments of the present invention,

[0045] An electronic device is provided, comprising:

[0046] processor;

[0047] a memory for storing processor-executable instructions;

[0048] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0049] A fourth aspect of the embodiments of the present invention is:

[0050] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0051] The mobile edge computing solution proposed in this application deploys mobile edge computing servers in the drone formation, adopts virtualization and containerization technologies to achieve computing resource management, designs microservice architecture to develop computing applications, and realizes service discovery, task offloading, task execution, relay communication and other functions. It can significantly improve the computing, storage and communication capabilities of the drone formation and provide efficient and reliable support services for complex tasks.

[0052] The topology change trend prediction method proposed in this application predicts the future position of the drone through the Kalman filter algorithm, builds a link quality estimation model using the Q-learning algorithm, and introduces the normalized Hamming distance to measure the degree of topology change, which can effectively predict the topology structure change trend of the drone cluster. This method comprehensively considers the two factors of drone motion state and link quality, and improves the accuracy and reliability of topology change prediction. Predicting the topology change trend can help drones prepare for communication adjustments in advance and ensure the stability and robustness of communication when the formation structure and communication environment change. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of a decentralized cluster formation control method based on virtual particle control according to an embodiment of the present invention;

[0054] Figure 2 A schematic diagram of the topological structure of the inter-machine reference status information according to an embodiment of the present invention;

[0055] Figure 3 This is a schematic diagram of the location of inter-machine reference information according to an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram of the first cluster formation control according to an embodiment of the present invention;

[0057] Figure 5 This is a schematic diagram of the second cluster formation control according to an embodiment of the present invention;

[0058] Figure 6 It is a structural schematic diagram of a decentralized cluster formation control system based on virtual particle control according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0061] refer to Figures 1 to 5 , an embodiment of the present invention provides a decentralized cluster formation control method based on virtual particle control, the method comprising:

[0062] S101. The cluster control terminal plans a reference route for the cluster formation mission; the virtual reference particle moves according to the reference route, and the virtual particle information between machines is aligned once every preset time, and the virtual particle information between machines is transmitted through inter-machine communication;

[0063] S102. Multiple UAVs refer to the virtual particle information and calculate their own flight control instructions respectively, wherein the flight control instructions include: a roll control instruction obtained by the L1 control strategy based on the expected route, an altitude control instruction superimposed on the waypoint altitude expectation as an altitude expectation input according to the formation altitude offset, and a speed control instruction that gives the aircraft speed expectation according to the current position relationship between the UAV and the reference point;

[0064] S103. The secondary reference point position is obtained by calculating the relative position offset of each UAV from the central reference point in the corresponding formation, and the UAV finally forms the desired formation by catching up with the secondary reference point position; wherein, different formation configurations are achieved by changing the relative position offset; when the multiple UAVs form a formation, the following conditions are met: the central reference point position of all UAVs is consistent with the direction of movement, and the position of a single UAV can follow the position of its secondary reference point.

[0065] In an optional embodiment,

[0066] The virtual reference particle moves according to the reference route, and according to the virtual particle information between machines, an alignment is performed every preset time, including:

[0067] The virtual reference particle moves according to the reference route planned by the cluster control terminal. The north and east positions of the single-machine reference point are determined by the target position and the offset in the corresponding direction respectively, and the plane position of the secondary reference point is determined by the plane position of the single-machine reference point and the plane offset in the corresponding direction. The virtual reference particle of each UAV transmits the local virtual particle information through inter-machine communication, receives the virtual particle information transmitted by other UAVs according to a preset round-robin topology structure, and performs position alignment once every preset time using the received inter-machine virtual particle information. After alignment, the virtual reference particle positions and movement directions of the multiple UAVs are consistent. In the round-robin topology structure, each UAV has the same number of topological nodes, and the topological nodes are arranged according to the UAV number. UAVs give priority to communicating with other UAVs with adjacent numbers.

[0068] In an optional embodiment,

[0069] The roll control command derived from the L1 control strategy based on the expected route, the altitude control command superimposed on the waypoint altitude expectation as the altitude expectation input according to the formation altitude offset, and the speed control command given the aircraft speed expectation according to the current position relationship between the UAV and the reference point include:

[0070] The UAV calculates a roll control instruction, an altitude control instruction and a speed control instruction according to the expected route and the virtual reference particle information; wherein, the L1 control strategy is used to calculate the heading error based on the expected route and the current aircraft position, and the heading error is input into the roll angle instruction generation module to obtain the roll control instruction; the formation formation altitude offset is superimposed on the route waypoint altitude planned by the cluster control terminal, and the superimposed altitude value is input into the altitude control module as the altitude expectation to obtain the altitude control instruction; the sight line vector is calculated according to the relative relationship between the current aircraft position and the corresponding virtual reference particle position, and the sight line vector is multiplied by the virtual particle expected velocity to obtain the aircraft velocity vector, and the aircraft velocity vector is input into the velocity control module to obtain the velocity control instruction.

[0071] In an optional embodiment,

[0072] By calculating the relative position offset of each UAV from the central reference point in the corresponding formation, the position of the secondary reference point is obtained, and the UAV catches up with the position of the secondary reference point to finally form the required formation, including:

[0073] The cluster control terminal calculates the relative position offset of each UAV from the formation center reference point according to the expected formation formation, superimposes the center reference point position and the relative position offset to obtain the secondary reference point position of each UAV; each UAV preferentially flies to the corresponding secondary reference point position, and when the distance between the UAV and the secondary reference point position is less than a preset threshold, it is considered that the UAV has reached the formation sequence, and the state is adjusted to formation holding. In the formation holding state, the UAV flies at a preset speed; after all UAVs enter the formation holding state, the formation formation is formed; wherein, by adjusting the relative position offset of each UAV, the formation formation is dynamically changed during the flight.

[0074] In an optional embodiment,

[0075] The method further comprises:

[0076] The drone collects its own flight status information and surrounding environment information through airborne sensors, and uses the collected status information and environmental information to establish a link quality estimation model. The link quality estimation model comprehensively considers the distance between drones, relative speed, and signal strength, and quantitatively evaluates the quality and stability of the communication link between drones to obtain a link quality evaluation value;

[0077] According to the link quality evaluation value, all communicable drones are sorted, and the first N drones with the highest link quality evaluation value are selected as candidate communication objects, where N is dynamically adjusted according to the communication demand and the number of drones;

[0078] The drone adjusts the connectivity of the topological structure according to the mission requirements and formation. For the formation whose mission requires information sharing exceeding the preset sharing threshold, the drone builds a fully connected mesh topology. For the formation whose mission requires group collaboration, the drone builds a clustered star topology. For the formation whose mission requires data aggregation, the drone builds a tree topology.

[0079] The UAV estimates the future position and speed of itself and other UAVs through the motion prediction algorithm, and based on the link quality estimation model, predicts the trend of topological structure changes and makes preparations for communication adjustments in advance;

[0080] After adjusting the communication topology structure according to the link quality estimation model, the UAV broadcasts the topology change information to all other UAVs in the formation. The topology change information includes the UAV number, change time, and the list of neighbor nodes after the change. After receiving the topology change information, other UAVs update the locally stored formation communication topology structure.

[0081] For example, in order to obtain the flight status and surrounding environment information of the drone, the drone needs to be equipped with a variety of airborne sensors, including inertial measurement unit, global positioning system, laser radar, ultrasonic ranging, etc. Through these sensors, the drone can collect its own position, speed, heading, altitude and other status information in real time, as well as surrounding obstacles, terrain, weather and other environmental information.

[0082] The inertial measurement unit is composed of accelerometers, gyroscopes, magnetometers, etc., which can measure the angular velocity, acceleration, attitude angle and other motion information of the drone. Through coordinate system conversion and attitude solution, the angular velocity and acceleration output by the inertial measurement unit can be converted from the body coordinate system to the navigation coordinate system, and then through speed and position solution, the attitude angle, position, speed and other state quantities of the drone in the navigation coordinate system can be obtained.

[0083] The GPS receiver captures satellite signals and calculates the position, speed, time and other navigation information of the drone. However, the GPS signal is easily affected by occlusion and multipath effects, resulting in reduced positioning accuracy. Therefore, the inertial measurement unit and the GPS are usually combined for navigation, and the Kalman filter and other algorithms are used to achieve sensor fusion to obtain a more stable and reliable state estimation.

[0084] LiDAR measures the distance and direction between the drone and the obstacle by emitting a laser beam and receiving the reflected echo. By filtering, segmenting, clustering and other processing of the obstacle point cloud data obtained by the LiDAR scan, the location, size, contour and other attributes of the obstacle can be identified.

[0085] Ultrasonic sensors measure the distance to obstacles by emitting ultrasonic waves and detecting echoes. By properly arranging multiple ultrasonic sensors in different directions, obstacle distance information in different directions can be obtained. Ultrasonic sensors are suitable for obstacle avoidance and hovering control in close-range and low-altitude environments.

[0086] By integrating the drone status information and environmental information collected by the above sensors, complete drone flight perception data can be obtained. These data include the drone's own position, speed, attitude angle and other state quantities, as well as environmental information such as surrounding drones and obstacles.

[0087] Based on the drone status and environment information, a link quality estimation model can be constructed to evaluate the quality and stability of the communication links between different drones. In this paper, a neural network method is used to construct a link quality estimation model, and the mapping relationship between link characteristics and link quality is learned through offline training.

[0088] First, in the training phase, a large amount of simulation and experimental data is used to obtain link quality samples of drones in different scenarios. Each link quality sample contains a link feature vector and a corresponding link quality value. Link features include the following information: Euclidean distance between drones, relative speed between drones, azimuth angle difference (reflecting the relative attitude of drones), communication signal strength, link historical packet loss rate, etc.

[0089] Based on the above features, the neural network input is formed, and the link quality value is used as a label to train a feedforward neural network. The neural network takes the link features as input and outputs the corresponding link quality estimation value, which can quantitatively evaluate the communication quality and stability of the link. The network structure includes an input layer, multiple fully connected hidden layers and an output layer. The hidden layer nodes use the ReLU activation function, and the output layer uses the Sigmoid function to map the output to between 0 and 1, indicating the quality of the link.

[0090] The mean square error is used as the loss function during training, and the gradient descent method is used for parameter optimization. The network weights are continuously adjusted to make the output link quality estimation value as close as possible to the sample label value. Through multiple rounds of iterative training, the link quality estimation model can well fit the intrinsic relationship between link characteristics and link quality.

[0091] In the application stage, after the drone obtains its own status information with other drones, it extracts the link distance, relative speed, signal strength and other features, and inputs them into the trained link quality estimation model to obtain the quality evaluation value of the link with other drones. This method can dynamically predict the changing trend of link quality based on the real-time status and environment of the drone.

[0092] After obtaining the link quality evaluation values ​​of all communicable drones, it is necessary to select the drone with the most stable link as the priority communication object. The specific method is to sort all drones from high to low according to the link quality evaluation value, and select the top N drones with the highest evaluation value as candidate communication objects.

[0093] The number of candidate communication objects N can be dynamically adjusted according to the communication demand and the total number of drones. When the communication demand is high, such as low-latency, high-bandwidth data transmission tasks, the N value can be appropriately increased to increase the number of redundant links and ensure communication quality. When the number of drones is small, or the available frequency and time slot resources are limited, the N value can be reduced to reduce the number of candidate objects and limit communication overhead. Dynamic adjustment of the N value allows the system to strike a balance between performance and resource consumption.

[0094] After selecting the candidate communication object, the UAV adjusts the connectivity of the communication topology according to the mission requirements and formation. Different mission requirements and formation structures are suitable for different topological forms:

[0095] (1) For formations whose mission requires information sharing exceeding a preset sharing threshold, the drones build a fully connected mesh topology in which each node is directly connected to all other nodes. This enables rapid diffusion and synchronization of information among formation members.

[0096] (2) For formations that require teamwork, the drones construct a clustered star topology. Each cluster has a central leader node that connects to the cluster members and coordinates their actions. Clusters can maintain communication links through the leader node.

[0097] (3) For formations whose mission focuses on data aggregation and transmission to the core node, the drones build a tree topology. The core node serves as the root node and manages the data collection of the remaining branches.

[0098] By adjusting the topology connectivity for specific application scenarios, network efficiency and scalability can be optimized. Topology selection requires a balance between information sharing, collaboration scale, and communication efficiency.

[0099] In addition to adjusting the current topology, it is also very important to predict the trend of topological structure changes. The drone uses a motion prediction algorithm to estimate the future position and speed of itself and other drones. Based on the predicted position and the link quality estimation model established earlier, the trend of topological structure changes can be predicted.

[0100] Specifically, the drone calculates the predicted distance between drones at different moments in the future, as well as the predicted relative position and speed. These predicted parameters are then input into the link quality estimation model to obtain the predicted value of the link quality in the future time period. By comparing the predicted link quality value with the preset quality threshold, the connectivity of the links between different drones in the future can be determined, and the predicted network topology in the future can be obtained.

[0101] Predicting the trend of topological changes enables drones to prepare for communication adjustments in advance. If it is predicted that significant topological changes will occur in the future, such as the interruption of important communication links, drones can reserve communication resources in advance, adjust routing strategies, or adjust their roles and positions in the formation. Through this forward-looking preparation, the stability and robustness of communication can be guaranteed when the formation structure and communication environment change.

[0102] After adjusting the communication topology according to the link quality estimation model, the UAV broadcasts the topology change information to all other UAVs in the formation. The topology change information includes the UAV number, the change time, and the updated neighbor node list after the change.

[0103] Topology change information is transmitted through the broadcast channel using broadcast message packets. All drones within the communication range of the sending drone can receive this topology change information. After receiving the message packet, each drone updates the locally stored communication topology structure to reflect the latest changes.

[0104] The advantage of broadcast transmission is efficient and fast notification of topology changes. Since all drones can receive broadcast messages at the same time, there is no need for repeated forwarding or confirmation. This reduces communication overhead and ensures that all members of the formation have a consistent view of the topology.

[0105] However, considering the limited transmission range of drones, individual drones that are far away from the sending drone may not receive the topology change information immediately. To solve this problem, secondary forwarding of relay nodes can be introduced to enable the topology change message to be propagated to all drones in the formation. In addition, the validity period of topology information can be set, and topology synchronization can be performed regularly between drones to maintain a unified and up-to-date view of the entire network.

[0106] In an optional embodiment,

[0107] The UAV estimates the future position and speed of itself and other UAVs through the motion prediction algorithm, and based on the link quality estimation model, predicts the change trend of the topology structure including:

[0108] Based on its historical trajectory data, the UAV uses the Kalman filter algorithm to recursively estimate the position and speed of the UAV. Through the two steps of prediction and update, the current position and speed of the UAV are estimated in real time. Assuming that the UAV maintains uniform linear motion in the prediction time domain, the position coordinates of the UAV at different times in the future are predicted.

[0109] A link quality estimation model based on Q-learning is constructed, in which the distance, relative speed and signal strength between the two ends of the link are used as state quantities, the predicted value of link quality is used as action quantity, and the link quality prediction error is used as reward value. The optimal link quality prediction action in each state is obtained through iterative update of Q value.

[0110] After the UAV obtains its predicted coordinates and those of other UAVs at different moments in the future, it calculates the predicted distances between different UAVs and inputs the predicted relative positions and speeds into the trained link quality estimation model to obtain the predicted link quality values ​​at different moments in the future. By calculating the normalized Hamming distance between the predicted topology and the current topology, the degree of change in the future topology can be quantitatively evaluated.

[0111] For example, in order to predict the changing trend of the future UAV cluster communication topology structure, this paper proposes a method based on motion prediction and link quality estimation. The method mainly includes three steps: UAV motion state prediction, link quality estimation model construction and topology change trend evaluation.

[0112] The first step is to predict the motion state of the drone. Each drone uses the Kalman filter algorithm to recursively estimate the future position and speed based on its own historical trajectory data. The Kalman filter algorithm estimates the current position and speed of the drone in real time through two alternating steps of prediction and update. In the prediction step, the algorithm assumes that the drone maintains uniform linear motion in the prediction time domain, and predicts the prior estimate of the state at the current moment based on the state estimate value and state transfer matrix at the previous moment. In the update step, the algorithm combines the observation value and observation matrix at the current moment to correct the prior estimate of the state and obtain the posterior estimate as the initial value for the next moment prediction. By recursively executing the prediction and update steps, the predicted sequence of the drone's position coordinates at different moments in the future can be obtained.

[0113] The second step is to build a link quality estimation model. This paper uses the Q-learning algorithm based on reinforcement learning to build a link quality estimation model. The distance, relative speed and signal strength between the drones at both ends of the link are taken as state quantities, the predicted value of the link quality is taken as the action quantity, and the link quality prediction error is taken as the reward value. The optimal link quality prediction action in each state is obtained through iterative updates of the Q value table. Specifically, the state space and action space are first discretized and the Q value table is initialized. Then the Q value table is continuously updated through interaction with the environment. In each interaction step, an action is selected according to the current state to obtain the immediate reward and the next state, and the Q value is updated using the value function update formula of Q-learning. After multiple rounds of iterations, the Q value table converges to the optimal value. At this time, the action with the largest Q value is selected in each state, which is the optimal link quality prediction value in that state.

[0114] The third step is to evaluate the trend of topological changes. After the drone obtains the predicted coordinates of itself and other drones at different moments in the future, it calculates the predicted distances between different drones. At the same time, the predicted relative position and speed are input into the trained link quality estimation model to obtain the predicted link quality values ​​at different moments in the future. Based on the predicted link quality values, the connectivity between drones at different moments in the future can be judged and the corresponding predicted topological structures can be generated. In order to quantitatively evaluate the degree of change of the predicted topological structure relative to the current topological structure, the normalized Hamming distance is introduced as a metric. The normalized Hamming distance represents the ratio of the number of different edges between two topological structures to the total number of edges, and the value is between 0 and 1. The larger the normalized Hamming distance, the greater the difference between the two topological structures and the more significant the topological change. By calculating the normalized Hamming distance between a series of predicted topological structures and the current topological structure, the trend of topological structure change in the future can be quantitatively evaluated.

[0115] In summary, the topology change trend prediction method proposed in this paper predicts the future position of the drone through the Kalman filter algorithm, builds a link quality estimation model using the Q-learning algorithm, and introduces the normalized Hamming distance to measure the degree of topology change, which can effectively predict the topology structure change trend of the drone cluster. This method comprehensively considers the two factors of drone motion state and link quality, and improves the accuracy and reliability of topology change prediction. Predicting the topology change trend can help drones prepare for communication adjustments in advance and ensure the stability and robustness of communication when the formation structure and communication environment change.

[0116] In an optional embodiment,

[0117] The method further comprises:

[0118] Select some drones with sufficient computing and storage resources in the drone formation and deploy mobile edge computing servers. The mobile edge computing servers use virtualization and containerization technologies to achieve flexible scheduling and isolation of computing resources, and design mobile edge computing applications based on microservice architecture.

[0119] When selecting the location of the mobile edge computing server, the power, computing resources, and communication capabilities of the drone itself are comprehensively considered, and the mobile edge computing servers are required to be evenly distributed within the formation to avoid communication congestion and single point failure caused by excessive concentration of servers;

[0120] After the mobile edge computing server is started, it broadcasts its own server identification, service type, and service capabilities within the formation. After other drones receive the broadcast information, they register the broadcast information of the mobile edge computing server in the local service list;

[0121] When the UAV processes the task locally, if it finds that the computing or storage resources are insufficient, it will initiate a task offloading request to the mobile edge computing server. The task offloading request includes the task type, input data, and expected completion time. The UAV sends the task offloading request to the mobile edge computing server through a round-robin topology structure.

[0122] After receiving the task offloading request, the mobile edge computing server decides whether to accept the request based on the task type and available resources. If accepted, the task is assigned to local computing resources for execution, and the calculation results are fed back to the requesting node through the round-robin topology structure. If rejected, the requesting node is informed of the rejection reason.

[0123] According to its own communication capability and link quality, the mobile edge computing server relays a drone's data packet to a target drone that is farther away than a preset distance threshold through multi-hop forwarding or store-and-forwarding, thereby expanding the communication coverage of the drone formation and optimizing the relay communication performance through priority management and power control.

[0124] For example, in order to improve the computing and storage capabilities of drone formations and reduce data transmission delays, this paper proposes a solution to deploy mobile edge computing servers in drone formations. This solution uses some drones in the drone formation that have sufficient computing and storage resources, realizes flexible scheduling and isolation of computing resources through virtualization and containerization technologies, and uses microservice architecture to design mobile edge computing applications, providing low-latency, highly reliable computing and storage services.

[0125] Some drones with sufficient computing and storage resources are selected from the drone formation as the carrier platform of the mobile edge computing server. During the selection process, the drone's own power, computing resources, communication capabilities and other factors are comprehensively considered, and drones with sufficient power, rich computing resources and strong communication capabilities are given priority as server nodes.

[0126] In order to ensure the reliability and availability of services, mobile edge computing servers are required to be evenly distributed within the fleet to avoid communication congestion and single point failure caused by excessive concentration of servers. Specifically, the fleet is divided into multiple areas, and a certain number of server nodes are deployed in each area to make the servers evenly distributed in space. At the same time, the number and location of server nodes are dynamically adjusted to adapt to changes in fleet size and task requirements.

[0127] On the selected drones, virtualization and containerization technologies are used to deploy mobile edge computing servers. Virtualization technology creates multiple virtual machine instances to achieve logical division and isolation of computing resources and improve resource utilization efficiency. Containerization technology creates lightweight container instances to achieve rapid deployment and elastic scaling of applications and reduce operation and maintenance complexity. Mobile edge computing servers can flexibly schedule virtual machine and container resources, dynamically adjust resource configuration according to task load, and improve computing performance and service quality.

[0128] The mobile edge computing server adopts the microservice architecture to design the mobile edge computing application, splitting the complex application system into multiple loosely coupled microservice components. Each microservice component is responsible for an independent business function and interacts with other microservices through well-defined interfaces to achieve modularity and scalability of the application.

[0129] The advantages of microservice architecture are: first, the service granularity is small, a single service is simple and easy to develop, and can be developed in parallel by different teams to improve development efficiency; second, services communicate through lightweight communication protocols, which reduces the dependency between services and makes the system more flexible and maintainable; third, services can be independently deployed, upgraded and scaled according to demand, improving the availability and scalability of the system.

[0130] In the mobile edge computing scenario, the microservice architecture can divide computing, storage, communication and other functions into independent microservices and deploy them on different edge servers to achieve the decomposition and offloading of computing tasks. For example, computing-intensive tasks such as data preprocessing and feature extraction are designed as independent microservices and offloaded to edge servers with sufficient computing resources, while communication-intensive tasks such as data aggregation and state synchronization are designed as independent microservices and deployed on edge servers with strong communication capabilities. By reasonably dividing and deploying microservices, the advantages of mobile edge computing can be fully utilized to improve task execution efficiency and system performance.

[0131] After the mobile edge computing server is started, it needs to broadcast its own service information within the drone formation, including server identification, service type, service capability, etc. After receiving the broadcast information, other drones register the server information into the locally maintained service list and establish a mapping relationship between the server and the service type. The service discovery mechanism enables drones to dynamically perceive the available edge computing services within the formation, providing a decision-making basis for task offloading.

[0132] When the drone processes a task locally, if it finds that the computing or storage resources are insufficient to meet the task requirements, the task offloading mechanism is triggered. The drone selects a qualified edge server from the local service list according to the task type and sends a task offloading request to it. The task offloading request contains information such as task type, input data, and expected completion time, which is used to describe the computing requirements and service quality requirements of the task.

[0133] The drone sends the task offloading request to the selected mobile edge computing server through the round-robin topology. The round-robin topology can provide multiple redundant paths for task offloading and improve transmission reliability. Specifically, the drone constructs the offloading request as a data packet and sets the target address of the data packet to the address of the edge server. Then, the drone sends the data packet to the next hop neighbor node in the current round-robin topology. After the neighbor node receives the data packet, if it is not the target server, it will continue to forward the data packet to the next hop until the data packet reaches the target server.

[0134] After receiving the task offloading request, the mobile edge computing server first parses the request content and extracts key information such as task type, input data, and expected completion time. Then, the server evaluates its own computing resources and load status to determine whether it has sufficient computing power to complete the task. If the computing resources are sufficient, the server accepts the task offloading request and assigns the task to local computing resources for execution. The server selects a suitable virtual machine or container instance, passes the task and its input data to the selected instance, and starts the task execution process.

[0135] During the task execution process, the server monitors the execution progress and resource consumption of the task in real time, and dynamically adjusts resource allocation to ensure that the task is completed within the expected completion time. If the task execution is abnormal or timed out, the server can take fault-tolerant measures such as retry and migration to ensure the reliable execution of the task.

[0136] After the task is completed, the server encapsulates the task output and execution status into a feedback message and sends it to the requesting node through a round-robin topology. The feedback message is also transmitted in the form of a data packet along the reverse path of the task offloading request, and finally reaches the requesting node after a multi-hop relay. After receiving the feedback message, the requesting node extracts the task execution result, updates the local task status, and completes the entire task offloading process.

[0137] If the server finds that its own computing resources are insufficient to meet the task offloading request after evaluation, it will reject the request. The server sends a rejection message to the requesting node, stating the reason for the rejection, such as insufficient computing resources, unsupported task type, etc. After receiving the rejection message, the requesting node can choose to re-initiate the offloading request to other servers, or downgrade the task execution locally.

[0138] In addition to providing computing and storage services, mobile edge computing servers can also act as relay nodes to expand the communication coverage of the drone fleet. When a drone needs to send data to a target drone that is beyond the direct communication range, it can request the mobile edge computing server to relay and forward.

[0139] The mobile edge computing server selects multi-hop forwarding or store-and-forwarding mode for relay communication according to its own communication capability and link quality. In the multi-hop forwarding mode, after receiving the data packet from the source drone, the server immediately forwards it to the next hop node. After multiple forwardings, the data packet finally reaches the target drone. Multi-hop forwarding is suitable for scenarios with good link quality and high end-to-end delay requirements.

[0140] In the store-and-forward mode, after receiving the data packet from the source drone, the server temporarily stores it locally and waits for the right time to forward it to the next hop node. Store-and-forward is suitable for scenarios with large link quality fluctuations and high packet arrival rate requirements. The server can use node movement and topology changes to select a time window with good link quality and high transmission success rate for forwarding, thereby improving the reliable transmission performance of data packets.

[0141] In order to optimize the performance of relay communication, the mobile edge computing server adopts a priority-based packet scheduling mechanism and power control strategy. The server dynamically adjusts the scheduling order and forwarding time of the data packet according to the data packet's priority, lifetime, target node and other attributes, giving priority to the transmission of data packets with high importance and strong timeliness. At the same time, the server dynamically adjusts the transmission power according to the link status and interference environment, minimizing energy consumption and interference while ensuring the quality of communication.

[0142] To sum up, the mobile edge computing solution proposed in this paper deploys mobile edge computing servers in the UAV formation, adopts virtualization and containerization technologies to realize computing resource management, designs microservice architecture to develop computing applications, and realizes service discovery, task offloading, task execution, relay communication and other functions. It can significantly improve the computing, storage and communication capabilities of the UAV formation and provide efficient and reliable support services for complex tasks.

[0143] Another embodiment of the present invention is as follows:

[0144] The position information of the single-machine reference point is obtained by the L1 control strategy. The north and east reference point positions of the drone are P Tx, P Ty , V c is the reference speed of the drone, and the position control of the secondary reference point refers to the following expression:

[0145] SP Tx= P Tx+ Px, SP Ty= P Ty+ P y ,

[0146] in P x, P y is the plane offset of the formation in the north and east directions, SP Tx, SP Ty It is the plane position of the secondary reference point in the north and east directions.

[0147] The velocity of the single aircraft reference point is related to the position of the aircraft and the mass point. The reference point velocity control refers to the following expression:

[0148] ,

[0149] in is the plane position vector of the secondary reference point, is the UAV flight position vector, It is represented as the sight vector from the drone to the secondary reference point. To express the expected velocity of the virtual particle, the steady-state condition reached by the position of the virtual particle during the final single-machine stable tracking is:

[0150]

[0151] The topological structure of the inter-aircraft reference state information can be that the number of aircraft in each topological structure is 7, and the topological structure of each aircraft is the current numbered aircraft and the 6 aircraft after the number, that is, aircraft No. 1 refers to the aircraft information of aircraft numbers 1, 2, 3, 4, 5, 6, 7, and aircraft No. 2 refers to the aircraft information of aircraft numbers 2, 3, 4, 5, 6, 7, 8. When the number of aircraft numbers is less than 7, the flight information of numbers 1, 2, ... is sequentially referenced, that is, during the flight, aircraft No. 50 refers to the aircraft information of aircraft 50, 1, 2, 3, 4, 5, 6. This forms a round-robin topological structure.

[0152] According to the center reference point information between machines, the position information of the center reference point needs to be unified after a period of time. The position control expression of the reference information between machines is as follows:

[0153] ;

[0154] When the machine-to-machine reference point finally reaches stable control, the following conditions are met:

[0155]

[0156] The expected route of the responding aircraft node has been given in the above route planning, and the positions of the central reference point and the secondary reference point have been given in the reference point control strategy. The cluster single node controls the aircraft's roll attitude, altitude and speed according to the expected route and the position of the reference point.

[0157] First, single-node roll control is obtained by the L1 control strategy according to the desired route.

[0158] The single-node altitude control channel is superimposed on the waypoint altitude expectation according to the altitude offset required by the current formation and is input to the flight control as the altitude expectation. The altitude expectation expression is:

[0159] ,

[0160] The speed control channel of a single node is given according to the position relationship between the current aircraft and the reference point. The expected aircraft speed is V c , when the final single-node speed control is stable, it satisfies:

[0161]

[0162] By calculating the relative position offset of each aircraft from the central reference point in the corresponding formation, the position of the secondary reference point is obtained, and the position of the secondary reference point is caught up by the aircraft to finally form a formation.

[0163] When the final formation is formed, the following conditions should be met:

[0164] The center reference point position of all aircraft is consistent with the direction of movement, that is:

[0165]

[0166] The position of a single aircraft can follow the position of the secondary reference point, that is:

[0167]

[0168] Figure 6 FIG. 1 is a schematic diagram of a decentralized cluster formation control system based on virtual particle control according to an embodiment of the present invention. Figure 6 As shown, the system comprises:

[0169] The first unit is used for planning a reference route for a cluster formation task by a cluster control terminal; a virtual reference particle moves according to the reference route, and an alignment is performed once every preset time according to the virtual particle information between machines, and the virtual particle information between machines is transmitted through inter-machine communication;

[0170] The second unit is used for multiple drones to refer to the virtual particle information and calculate their own flight control instructions respectively, wherein the flight control instructions include: a roll control instruction obtained by the L1 control strategy based on the expected route, an altitude control instruction superimposed on the waypoint altitude expectation as an altitude expectation input according to the formation altitude offset, and a speed control instruction that gives the aircraft speed expectation according to the current position relationship between the drone and the reference point;

[0171] The third unit is used to obtain the secondary reference point position by calculating the relative position offset of each drone from the central reference point in the corresponding formation, and the drone finally forms the required formation by catching up with the secondary reference point position; wherein, different formation configurations are achieved by changing the relative position offset; when the multiple drones form a formation, it satisfies: the central reference point position of all drones is consistent with the movement direction, and the position of a single drone can follow the position of its secondary reference point.

[0172] According to a third aspect of the embodiments of the present invention,

[0173] An electronic device is provided, comprising:

[0174] processor;

[0175] a memory for storing processor-executable instructions;

[0176] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0177] A fourth aspect of the embodiments of the present invention is:

[0178] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0179] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0180] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A decentralized cluster formation control method based on virtual particle control, characterized in that: include: The cluster control terminal plans a reference route for the cluster formation mission; the virtual particles move according to the reference route, and align once every preset time according to the virtual particle information between machines, and the virtual particle information between machines is transmitted through inter-machine communication; The multiple UAVs refer to the virtual particle information to calculate their own flight control instructions respectively, and the flight control instructions include: a roll control instruction obtained by the L1 control strategy based on the expected route, an altitude control instruction superimposed on the waypoint altitude expectation as an altitude expectation input according to the formation altitude offset, and a speed control instruction that gives the aircraft speed expectation according to the current position relationship between the UAV and the reference point; By calculating the relative position offset of each drone from the central reference point in the corresponding formation, the secondary reference point position is obtained, and the drone catches up with the secondary reference point position to finally form the required formation; wherein, by changing the relative position offset, different formation configurations are achieved; when the multiple drones form a formation, the following conditions are met: the central reference point position of all drones is consistent with the direction of movement, and the position of a single drone can follow the position of its secondary reference point; The method further comprises: The drone collects its own flight status information and surrounding environment information through airborne sensors, and uses the collected status information and environmental information to establish a link quality estimation model. The link quality estimation model comprehensively considers the distance between drones, relative speed, and signal strength, and quantitatively evaluates the quality and stability of the communication link between drones to obtain a link quality evaluation value; According to the link quality evaluation value, all communicable drones are sorted, and the first N drones with the highest link quality evaluation value are selected as candidate communication objects, where N is dynamically adjusted according to the communication demand and the number of drones; The drone adjusts the connectivity of the topological structure according to the mission requirements and formation. For the formation whose mission requires information sharing exceeding the preset sharing threshold, the drone builds a fully connected mesh topology. For the formation whose mission requires group collaboration, the drone builds a clustered star topology. For the formation whose mission requires data aggregation, the drone builds a tree topology. The UAV estimates the future position and speed of itself and other UAVs through the motion prediction algorithm, and based on the link quality estimation model, predicts the trend of topological structure changes and makes preparations for communication adjustments in advance; After the UAV adjusts the communication topology structure according to the link quality estimation model, it broadcasts the topology change information to all other UAVs in the formation. The topology change information includes the UAV number, change time, and the changed neighbor node list. After receiving the topology change information, other UAVs update the locally stored formation communication topology structure. The UAV estimates the future position and speed of itself and other UAVs through the motion prediction algorithm, and based on the link quality estimation model, predicts the change trend of the topology structure including: Based on its historical trajectory data, the UAV uses the Kalman filter algorithm to recursively estimate the position and speed of the UAV. Through the two steps of prediction and update, the current position and speed of the UAV are estimated in real time. Assuming that the UAV maintains uniform linear motion in the prediction time domain, the position coordinates of the UAV at different times in the future are predicted. A link quality estimation model based on Q-learning is constructed, in which the distance, relative speed and signal strength between the two ends of the link are used as state quantities, the predicted value of link quality is used as action quantity, and the link quality prediction error is used as reward value. The optimal link quality prediction action in each state is obtained through iterative update of Q value. After the UAV obtains its predicted coordinates and those of other UAVs at different moments in the future, it calculates the predicted distances between different UAVs and inputs the predicted relative positions and speeds into the trained link quality estimation model to obtain the predicted link quality values ​​at different moments in the future. By calculating the normalized Hamming distance between the predicted topology and the current topology, the degree of change in the future topology can be quantitatively evaluated.

2. The method according to claim 1, characterized in that: The virtual particle moves according to the reference route, and according to the virtual particle information between machines, an alignment is performed every preset time, including: The virtual particle moves according to the reference route planned by the cluster control terminal. The north and east positions of the single-machine reference point are determined by the target position and the offset in the corresponding direction respectively, and the plane position of the secondary reference point is determined by the plane position of the single-machine reference point and the plane offset in the corresponding direction; the virtual particle of each UAV transmits the local virtual particle information through inter-machine communication, receives the virtual particle information transmitted by other UAVs according to a preset round-robin topology structure, and performs position alignment once every preset time using the received inter-machine virtual particle information. After alignment, the virtual particle positions and movement directions of the multiple UAVs are consistent; wherein, in the round-robin topology structure, each UAV has the same number of topological nodes, the topological nodes are arranged according to the UAV number, and the UAVs give priority to communicating with other UAVs with adjacent numbers.

3. The method according to claim 1, characterized in that The roll control command derived from the L1 control strategy based on the expected route, the altitude control command superimposed on the waypoint altitude expectation as the altitude expectation input according to the formation altitude offset, and the speed control command given the aircraft speed expectation according to the current position relationship between the UAV and the reference point include: The UAV calculates a roll control instruction, an altitude control instruction and a speed control instruction according to the expected route and the virtual particle information; wherein, the L1 control strategy is used to calculate the heading error based on the expected route and the current aircraft position, and the heading error is input into the roll angle instruction generation module to obtain the roll control instruction; the formation formation altitude offset is superimposed on the route waypoint altitude planned by the cluster control terminal, and the superimposed altitude value is input into the altitude control module as the altitude expectation to obtain the altitude control instruction; the sight line vector is calculated according to the relative relationship between the current aircraft position and the corresponding virtual particle position, and the sight line vector is multiplied by the virtual particle expected velocity to obtain the aircraft velocity vector, and the aircraft velocity vector is input into the velocity control module to obtain the velocity control instruction.

4. The method according to claim 1, characterized in that: By calculating the relative position offset of each UAV from the central reference point in the corresponding formation, the position of the secondary reference point is obtained, and the UAV catches up with the position of the secondary reference point to finally form the required formation, including: The cluster control terminal calculates the relative position offset of each UAV from the formation center reference point according to the expected formation formation, superimposes the center reference point position and the relative position offset to obtain the secondary reference point position of each UAV; each UAV preferentially flies to the corresponding secondary reference point position, and when the distance between the UAV and the secondary reference point position is less than a preset threshold, it is considered that the UAV has reached the formation sequence, and the state is adjusted to formation holding. In the formation holding state, the UAV flies at a preset speed; after all UAVs enter the formation holding state, the formation formation is formed; wherein, by adjusting the relative position offset of each UAV, the formation formation is dynamically changed during the flight.

5. The method according to claim 1, characterized in that The method further comprises: Select some drones with sufficient computing and storage resources in the drone formation and deploy mobile edge computing servers. The mobile edge computing servers use virtualization and containerization technologies to achieve flexible scheduling and isolation of computing resources, and design mobile edge computing applications based on microservice architecture. When selecting the location of the mobile edge computing server, the power, computing resources, and communication capabilities of the drone itself are comprehensively considered, and the mobile edge computing servers are required to be evenly distributed within the formation to avoid communication congestion and single point failure caused by excessive concentration of servers; After the mobile edge computing server is started, it broadcasts its own server identification, service type, and service capabilities within the formation. After other drones receive the broadcast information, they register the broadcast information of the mobile edge computing server in the local service list; When the UAV processes the task locally, if it finds that the computing or storage resources are insufficient, it will initiate a task offloading request to the mobile edge computing server. The task offloading request includes the task type, input data, and expected completion time. The UAV sends the task offloading request to the mobile edge computing server through a round-robin topology structure. After receiving the task offloading request, the mobile edge computing server decides whether to accept the request based on the task type and available resources. If accepted, the task is assigned to local computing resources for execution, and the calculation results are fed back to the requesting node through the round-robin topology structure. If rejected, the requesting node is informed of the rejection reason. According to its own communication capability and link quality, the mobile edge computing server relays a drone's data packet to a target drone that is farther away than a preset distance threshold through multi-hop forwarding or store-and-forwarding, thereby expanding the communication coverage of the drone formation and optimizing the relay communication performance through priority management and power control.

6. A decentralized cluster formation control system based on virtual particle control, used to implement the method described in any one of claims 1 to 5, characterized in that: include: The first unit is used for planning a reference route for the cluster formation task by the cluster control terminal; the virtual particle moves according to the reference route, and aligns once every preset time according to the virtual particle information between machines, and the virtual particle information between machines is transmitted through inter-machine communication; The second unit is used for multiple drones to refer to the virtual particle information and calculate their own flight control instructions respectively, wherein the flight control instructions include: a roll control instruction obtained by the L1 control strategy based on the expected route, an altitude control instruction superimposed on the waypoint altitude expectation as an altitude expectation input according to the formation altitude offset, and a speed control instruction that gives the aircraft speed expectation according to the current position relationship between the drone and the reference point; The third unit is used to obtain the secondary reference point position by calculating the relative position offset of each drone from the central reference point in the corresponding formation, and the drone finally forms the required formation by catching up with the secondary reference point position; wherein, different formation configurations are achieved by changing the relative position offset; when the multiple drones form a formation, it satisfies: the central reference point position of all drones is consistent with the movement direction, and the position of a single drone can follow the position of its secondary reference point.

7. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 5.

8. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Citation Information

Patent Citations

  • Method for navigation following type multi-agent formation path planning and storage medium

    CN113534819A

  • Finite time autonomous formation control method and system for cluster

    CN113655808A

  • Self-organizing formation method for unmanned aerial vehicle cluster

    CN114489127A

  • Single-route virtual long aircraft formation method based on geometric constraint rule

    CN114859959A

  • Multi-platform avionics system intelligent data routing method based on comprehensive communication efficiency

    CN115038137A