Multi-unmanned aerial vehicle cooperative control method based on data link networking

Abstract

The invention relates to the technical field of unmanned aerial vehicle cooperative control, and discloses a multi-unmanned aerial vehicle cooperative control method based on data link networking. The method comprises the following steps of: constructing a cluster collaborative task data set comprising a flight path feature, a load state feature and a communication link feature; extracting dynamic cooperative nodes, including a cluster formation form switching node and a task target switching node, from the cluster formation form switching node and the task target switching node; generating a track cooperative trigger signal based on the dynamic cooperative node, and identifying a local topology adjustment node triggered by the formation form switching and a global topology reconstruction node triggered by the task target switching; generating a cooperative offset parameter containing a track offset and a communication delay offset according to the node classification; extracting cluster collaboration protocol parameters based on the collaboration offset parameters; and controlling the unmanned aerial vehicle cluster to execute a cooperative flight task according to the protocol parameters. The method can improve the cooperative efficiency and task adaptability of the unmanned aerial vehicle group in a dynamic environment, and is suitable for various multi-vehicle cooperative operation scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of cooperative control of unmanned aerial vehicles (UAVs), and in particular to a cooperative control method for multiple UAVs based on data link networking. Background Art

[0002] With the rapid development of drone technology, multi-drone collaborative mission execution has become a key approach to improving operational efficiency and adaptability. In scenarios such as military reconnaissance, disaster monitoring, and logistics distribution, single drones are often limited by payload capacity, flight time, and operational range, making it difficult to meet complex mission requirements. Therefore, multi-drone collaboration is needed to form a swarm system with complementary functions. Current multi-drone collaborative control methods generally lack adaptability to dynamic mission environments. For example, when the fleet's formation needs to change to avoid obstacles or adjust the mission area, traditional control methods often rely on preset formation transition rules, making it difficult to respond to emergencies in complex environments in real time. This can lead to track conflicts and reduced collaborative efficiency during formation reconfiguration. When switching mission objectives, such as from area search to fixed-point monitoring, the communication topology within the fleet needs to be adjusted to meet the information exchange requirements of the new mission. However, existing methods often use a global reconfiguration model to adjust the communication topology. This triggers a reconfiguration of the entire network regardless of the scope of the mission change. This not only increases communication resource consumption but can also affect the real-time performance of collaborative control due to data transmission delays during the reconfiguration process. Existing collaborative control protocols are typically designed for specific mission scenarios and lack versatility and reusability. When a fleet needs to perform multiple, complex tasks, the switching of control protocols becomes complex, making collaborative failures due to protocol incompatibility more likely to occur. Regarding collaborative parameter calculation, existing methods fail to adequately consider the correlation between track deviation and communication delay, often setting parameters separately. This makes it difficult to achieve collaborative optimization of track planning and communication quality in a dynamic environment, impacting the overall operational efficiency of the fleet. Summary of the Invention

[0003] The purpose of the present invention is to provide a multi-machine collaborative control method for unmanned aerial vehicles based on data link networking to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides a method for cooperative control of multiple UAVs based on data link networking, the method comprising: Constructing a swarm collaborative mission data set, wherein the swarm collaborative mission data set includes flight path characteristics, load state characteristics, and communication link characteristics; Extracting dynamic coordination nodes from the fleet coordination task data set, wherein the dynamic coordination nodes include a fleet formation switching node and a task target switching node; Generate a track coordination trigger signal based on the dynamic coordination node, and identify the cluster communication topology change node based on the track coordination trigger signal. The cluster communication topology change node includes a local topology adjustment node triggered by formation form switching and a global topology reconstruction node triggered by mission target switching; Generate collaborative offset parameters based on the classification of cluster communication topology change nodes, wherein the collaborative offset parameters include the track offset during formation form switching and the communication delay offset during mission target switching; Extracting swarm collaboration protocol parameters based on the collaboration offset parameters, wherein the swarm collaboration protocol parameters include a general flight protocol that is reusable across missions and a payload control protocol dedicated to a specific mission, where the specific mission is an operation scenario requiring a dedicated payload configuration or a differentiated collaboration strategy within a preset range; The UAV cluster is controlled to perform collaborative flight missions based on the swarm collaboration protocol parameters.

[0005] Preferably, the constructing of the cluster collaborative task data set includes: Extract the initial formation parameter set and corresponding flight time period based on flight path characteristics; Extracting a formation reconstruction parameter set and a corresponding switching period based on a formation morphology switching node; The initial formation parameter set and flight period are processed by the formation verification algorithm to generate a revised formation reconstruction parameter set; The initial formation parameter set and the revised formation reconstruction parameter set are fused to output the fleet collaborative task data set, in which the flight environment disturbance factor is introduced when the formation verification algorithm calculates the flight time deviation.

[0006] Preferably, generating a track coordination trigger signal includes: Extracting a formation reconstruction parameter set to generate a formation trigger signal; Extracting the corrected formation reconstruction parameter set to generate a task switching signal; Identify track coordination trends based on formation trigger signals and task switching signals; The communication link feature sampling frequency is dynamically adjusted according to the track collaboration trend to generate a track collaboration trigger signal.

[0007] Preferably, the identifying cluster communication topology change node includes: Calculate the similarity between the track coordination trigger signal and the track of the historical topology adjustment node; When the track similarity exceeds the preset similarity threshold, it is marked as a local topology adjustment node; When the track similarity is lower than the preset similarity threshold, it is marked as a global topology reconstruction node; The track similarity is calculated by combining the track spatial distribution overlap rate with the time synchronization deviation.

[0008] Preferably, generating the collaborative offset parameter includes: Extract the local topology adjustment nodes triggered by formation shape switching and mark them as formation offset nodes; Extract the global topology reconstruction node triggered by task target switching and mark it as the task offset node; Calculate the track position offset based on the formation offset node, and mark it as the track coordination offset parameter; Calculate the communication delay increment based on the task offset node, marked as the communication coordination offset parameter; The track cooperative offset parameter and the communication cooperative offset parameter are fused to generate the cooperative offset parameter.

[0009] Preferably, the extraction of cluster collaboration protocol parameters includes: The collaborative offset parameters are analyzed based on the flight path characteristics, and a cross-mission reusable trajectory control protocol is extracted as a universal flight protocol. Analyze the correlation between payload state characteristics and cooperative offset parameters, and extract a sensor scheduling protocol dedicated to a specific task as a payload control protocol; The general flight protocol is optimized through the track history data reuse verification algorithm.

[0010] Preferably, the control of the drone cluster to perform a collaborative flight mission includes: using the general flight protocol as a global track control benchmark, dynamically adjusting the working mode of the cluster sensors based on the payload control protocol; verifying the consistency of the protocol execution through the communication link characteristics, and triggering the protocol reload mechanism when it is detected that the communication delay offset exceeds the limit.

[0011] Preferably, the method also includes: establishing a swarm collaboration quality assessment model, inputting collaboration offset parameters and swarm collaboration protocol parameters; outputting a collaboration quality assessment value to a track collaboration trigger signal generation module; and dynamically correcting the update frequency of the formation reconstruction parameter set based on the collaboration quality assessment value.

[0012] Preferably, the swarm collaboration quality assessment model includes: constructing a weighted assessment function of track position offset and communication delay offset, introducing a flight environment disturbance factor as a correction term for the assessment function, and training the assessment function weight coefficient through historical mission execution data.

[0013] Preferably, the node for identifying cluster communication topology changes further includes: When a communication link interruption is detected or the data link quality is lower than the preset threshold, the topology degradation mechanism is triggered; Dynamically selecting the execution priority of a local topology adjustment node or a global topology reconstruction node based on the track similarity and the real-time communication quality evaluation result; The topology degradation mechanism maintains the basic formation communication capability through a preset minimum connected subgraph algorithm.

[0014] Compared with the prior art, the present invention has the following beneficial effects: By building a comprehensive data set for collaborative swarm missions, encompassing flight paths, payload status, and communication link characteristics, this provides comprehensive fundamental data support for collaborative control. This multi-dimensional data integration enables the swarm to comprehensively consider various influencing factors during mission execution, avoiding decision-making biases caused by missing a single data dimension. During the dynamic coordination node extraction phase, node identification is performed for two key scenarios: fleet formation switching and mission objective switching. This allows coordinated control to precisely focus on the core change points during the mission. This precise identification mechanism helps reduce the system resource consumption of irrelevant data processing and improves the pertinence and efficiency of control responses. Based on dynamic coordination nodes, track coordination trigger signals are generated and used to identify nodes with communication topology changes. This further differentiates the applicable scenarios for local topology adjustment and global topology reconstruction, enabling differentiated communication topology adjustments. For local mission changes, such as formation switching, local topology adjustments can reduce fluctuations in the communication network and maintain stable connections for most nodes. For global mission changes, such as switching mission objectives, global topology reconstruction ensures overall optimization of the communication link for the new mission. This flexible adjustment method balances communication efficiency and stability. By classifying nodes that change in the communication topology, collaborative offset parameters are generated. Track offsets and communication delay offsets are calculated for different mission scenarios, achieving collaborative optimization of track planning and communication quality. During formation switching, accurate calculation of track offsets avoids collision risks between drones while minimizing the impact of track adjustments on communication link stability. During mission target switching, consideration of communication delay offsets reduces the impact of data transmission delays on the collaborative execution of new missions, enabling dynamic adaptation of track planning and communication status. By extracting the swarm's collaborative protocol parameters through collaborative offset parameters, the system distinguishes between a general flight protocol that can be reused across missions and a payload control protocol dedicated to a specific mission, enhancing the flexibility and compatibility of the control protocol. The general flight protocol ensures the basic collaborative capabilities of the swarm when switching between different mission types, reducing the complexity of protocol switching. The dedicated payload control protocol can meet the personalized requirements of specific missions for payload operations, enabling the swarm to adapt to diverse mission scenarios while ensuring the execution accuracy of specific missions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a working principle diagram of the multi-machine collaborative control method of UAVs based on data link networking according to the present invention; Figure 2 Flowchart for track coordination trigger signal generation; Figure 3 Flowchart for node identification for cluster communication topology change; Figure 4 Flowchart generated for the collaborative offset parameters; Figure 5 Flowchart for extracting parameters of cluster collaboration protocol; Figure 6 This is the formation parameter and trigger signal analysis diagram; Figure 7 Adjust the analysis diagram for communication quality and topology; Figure 8 For formation offset and coordination parameter analysis. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0017] See also Figure 1 The present invention provides a method for cooperative control of multiple unmanned aerial vehicles based on data link networking, the method comprising: Flight path features record each UAV's latitude and longitude coordinates, altitude, velocity vector, and heading angle; payload state features include sensor operating mode, energy consumption rate, and remaining mission payload capacity; and communication link features include data link signal strength, transmission delay, and packet loss rate. Dynamic coordination nodes are extracted using a sliding time window analysis method. Heading angle mutation points are identified in the flight path features as formation configuration switching nodes, and operating mode transition points are detected in the payload state features as mission target switching nodes. Track coordination trigger signals are generated by comparing the rate of change of the Euclidean distance between dynamic coordination nodes within adjacent time windows. A trigger signal is generated when the rate of change exceeds a threshold. Graph theory analysis is used to identify nodes that change the communication topology of the swarm. A UAV communication connection graph model is established. The rate of change of node degree distribution in the graph is calculated based on the track coordination trigger signal. Nodes with a rate of change below 20% are marked as local topology adjustment nodes, and nodes with a rate of change above 50% are marked as global topology reconstruction nodes. The collaborative offset parameters are calculated using differential positioning. The track offset during formation switching is calculated using the position difference between the current track point and the preset track point. The communication delay offset during mission target switching is determined using the sliding average of the data link transmission delay. The swarm collaborative protocol parameters are extracted using a protocol template matching algorithm. The general flight protocol matches track control parameters with a similarity exceeding 85% from the historical mission database. The payload control protocol calls the preset sensor scheduling scheme from the protocol library based on the current mission type. The collaborative control of the drone swarm is achieved through a distributed decision-making system. Each drone node autonomously adjusts its flight status based on the received collaborative protocol parameters, while simultaneously exchanging status information in real time through the data link network to maintain formation consistency.

[0018] Example 1: See Figure 2 , the specific operational procedures of flight path feature processing, formation parameter generation and trigger signal generation mechanism. Airspace processing adopts three-dimensional grid division technology, and the mission airspace is discretized into a cubic unit grid with a side length of 50 meters. Each grid unit has a unique spatial coordinate identifier. The crossing events of the drone in the grid are recorded by the airborne positioning system, and the continuous coordinate points with millisecond-level timestamps constitute the core content of the initial formation parameter set. The automatic flight time division mechanism analyzes the time interval characteristics of continuous coordinate points. When the time interval between adjacent coordinate points exceeds two seconds, the system automatically generates a new flight time boundary identifier, which contains the starting coordinates of the time period and the heading vector data.

[0019] Formation parameter generation utilizes a dynamic reference system construction method. Within the initial formation parameter set, the system analyzes the communication signal strength parameters of each drone in real time and selects the node with the highest signal strength for more than five seconds as the virtual pilot. The remaining drone nodes establish a local coordinate system based on the virtual pilot, forming a position offset matrix containing relative azimuth, pitch angle, and Euclidean distance. This matrix data is updated at a 10 Hz frequency over the data link. The matrix dimensions correspond to the size of the swarm, and each element records the six degrees of freedom parameters of the relative position of the two drones.

[0020] Formation parameter verification utilizes a disturbance compensation model. The verification process first calculates the real-time standard deviation of all elements in the position offset matrix to establish a formation morphological stability index. Flight environment disturbance factors are acquired via onboard meteorological sensors and include three-dimensional wind speed vectors and atmospheric density parameters. The system establishes a three-layer altitude disturbance model: a weighting coefficient of 0.3 is assigned to the 0-100-meter altitude layer, a coefficient of 0.5 is assigned to the 100-300-meter layer, and a coefficient of 0.7 is assigned to the layer above 300 meters. The Monte Carlo simulation module imports the current formation parameter set into a kinematic model with perturbation factors and performs 500 random simulations. The simulation output is then differentially compared with the original position offset matrix to generate a set of correction parameters containing wind speed compensation values. These correction parameters encompass heading angle compensation, speed adjustment, and altitude correction, and are stored separately for each drone node.

[0021] The trigger signal is generated using matrix feature analysis technology. The formation trigger signal is derived from the real-time singular value decomposition of the position offset matrix. The system continuously monitors fluctuations in the matrix's maximum singular value. If it detects an increase in the singular value exceeding 1.5 times the preset baseline value over three consecutive sampling periods, it automatically generates a trigger signal containing the current formation disintegration risk level. This signal is broadcast to the entire cluster via the data link. The signal payload includes the virtual leader identification, the matrix instability coefficient, and a code for the recommended adjustment strategy.

[0022] The task switching signal is generated by the load mode state machine. Sensor operating modes are modeled using a twelve-state finite automaton, with each state containing three attributes: operating frequency, power consumption level, and data output rate. A hidden Markov model with sixteen hidden state nodes calculates the mode transition probability using the Viterbi algorithm. The model input is a three-dimensional vector of load state characteristics: energy consumption rate, number of sensors activated, and remaining data cache capacity. When the state transition probability is less than one in a thousand for ten consecutive samplings, the system generates a task switching event code. This code contains the original task identifier, the recommended switching mode code, and the sensor configuration plan index.

[0023] Communication sampling adjustment adopts a closed-loop feedback mechanism. The basic sampling frequency is set to a ten-hertz cycle. When the system detects the track coordination trend factor and the load state change factor at the same time, the sampling frequency adjustment program is started. The adjustment algorithm uses the heading angle change rate as the input variable and establishes a linear mapping relationship from zero to thirty hertz. The specific execution process is: the absolute value of the angle difference between the current heading angle and the previous sampling is calculated in real time, and converted into a frequency increment through the proportional coefficient. This increment is limited by a three-second sliding window to prevent frequency oscillation caused by instantaneous mutations. The frequency switching process adopts a smooth transition strategy, gradually adjusting to the target frequency with a gradient of five hertz per second.

[0024] Collaborative mission data integration utilizes a multi-source fusion strategy. The initial and revised formation parameter sets are aligned along the timeline in the data storage layer, with each flight period corresponding to a separate parameter file. The file structure contains sixteen data sections: a header stores a three-dimensional grid coordinate mapping table; a main body stores original and revised parameters grouped by drone node; and a tail records a summary of the environmental perturbation model configuration and a verification algorithm version identifier. An integrity check is performed before data is output. The verification algorithm calculates the file checksum and content hash value to ensure parameter consistency during data transmission. The entire data processing process is executed with real-time task priority within the embedded system, with the computation cycle strictly controlled within a 50-millisecond time window.

[0025] See Figure 6 , showing the heading angle changes, working mode status and collaborative trigger signal generation of the UAV formation. The upper chart shows the heading angle change curves of multiple UAVs, and the vertical dotted lines mark the key nodes for switching the formation form. These nodes usually appear at locations where the heading angle changes significantly, reflecting the adjustment of the overall movement direction of the formation. The middle chart shows the conversion process of the UAV working mode, including four states: initialization, standby, scanning and locking. The dotted lines mark the mission target switching nodes, which correspond to the sudden changes in the working mode. The lower chart shows the track collaborative trigger signal generated based on the dynamic collaborative node. The square mark indicates the local topology adjustment node, and the triangle mark indicates the global topology reconstruction node. These nodes are triggered when the communication quality deteriorates or the mission requirements change.

[0026] Example 2: See Figure 3, the specific implementation process of track similarity calculation, communication topology adjustment and exception handling mechanism. The application of dynamic time warping algorithm in track matching adopts sliding window comparison technology. The window length is set to twenty sampling points. Each sampling point contains three-dimensional coordinates and corresponding timestamp information. The historical topology adjustment event database maintains track pattern records in the last fifty missions. Each record stores a complete sequence of track points and topology change type identification. The similarity calculation process first resamples the current track signal to align it with the time resolution of the historical record, and then calculates the cross-correlation coefficient of the waveform profile. The preset similarity threshold is determined by dynamic statistical methods. The system regularly updates the distribution characteristics of historical similarity data and sets the threshold at the third quartile position. This value is adaptively adjusted as the mission execution experience accumulates.

[0027] The quantitative analysis of spatial distribution overlap is based on a three-dimensional voxel space occupancy model. The mission airspace is divided into cubic units with a side length of one meter, and a binary occupancy flag is set for each unit. The voxel occupancy of the current track and the historical track is stored in the form of a bitmap, and the system performs a bitwise AND operation to count the number of overlapping units. The time synchronization deviation measurement system deploys a precision clock synchronization network. The master node is equipped with an atomic clock source, and the slave node uses software phase-locked loop technology to track the master clock. The clock offset record contains a transmission delay compensation value, which is calculated by measuring the round-trip time of the data packet. The synchronization accuracy is maintained at the microsecond level, and abnormal clock deviation events trigger an immediate correction protocol.

[0028] Communication link interruption detection uses a multi-level judgment mechanism. The physical layer continuously monitors the carrier signal strength indicator. When the received power is lower than the sensitivity threshold for one hundred milliseconds, the link status diagnostic program is started. The data link layer counts the response timeout rate of the last ten transmission windows, and the application layer verifies the continuity of the heartbeat message. The activation of the topology degradation mechanism requires two conditions to be met: the physical layer interruption lasts for more than five hundred milliseconds and the application layer heartbeat loss count reaches three times. The minimum connected subgraph algorithm uses an improved Prim implementation, introducing a communication quality weight factor in the adjacency matrix construction phase, and giving priority to links with high signal strength to join the spanning tree. During the execution of the algorithm, unstable nodes are dynamically eliminated to maintain the minimum support structure of the basic communication network.

[0029] The real-time communication quality assessment system utilizes a composite indicator system. The signal strength indicator is normalized using a logarithmic scale, mapping the received power to a linear range between 0 and 100. The transmission delay indicator calculates percentile values ​​within a sliding window to eliminate the impact of instantaneous jitter. Packet loss rate statistics are based on sequence number continuity analysis to identify duplicate and out-of-order packets. These three indicators are weighted and combined to generate an overall assessment value. The weighting is dynamically adjusted based on the mission phase: the cruise phase prioritizes signal strength, the mission execution phase prioritizes latency, and the data return phase prioritizes packet loss rate. The assessment results are output in a five-level quantitative format, corresponding to different topology adjustment strategies.

[0030] The execution priority decision-making system is implemented using a fuzzy logic controller. Input variables include track similarity and communication quality assessment values, each divided into five fuzzy sets. The fuzzy set boundaries for the similarity variable use a trapezoidal membership function, while the communication quality variable uses a triangular membership function. The rule base contains 25 fuzzy inference rules, each defining the mapping between input and output variables. The defuzzification process uses the centroid method to calculate the precise output value, quantifying the degree of urgency on a continuous scale from zero to one hundred. The output interface maps this value into a five-level response code to drive the corresponding topology adjustment program.

[0031] The processing flow for a local topology adjustment node consists of two phases: neighbor discovery and link reestablishment. When a node is marked for local adjustment, the system first broadcasts a neighbor discovery request to collect node status information within a two-hop radius. The link selection algorithm evaluates the stability of candidate paths, calculated by combining historical connection duration and signal fluctuation variance. New links must meet minimum bandwidth requirements and verify bidirectional connectivity through a three-way handshake. During the adjustment process, the routing table entries for the original data flow are maintained to ensure that ongoing task data is not affected.

[0032] The global topology reconstruction process adopts a phased approach. The first phase collects state information from all network nodes and constructs a complete snapshot of the network topology. The second phase analyzes task requirements and resource distribution to calculate the optimal communication connection plan. The third phase implements a gradual reconstruction, adjusting connections based on node importance. Rollback checkpoints are set during the reconstruction process to automatically restore the system to its previous stable state when deterioration of key performance indicators is detected. A detailed execution log is recorded for each reconstruction operation, which is used to optimize subsequent decision parameters.

[0033] The exception handling mechanism includes predefined recovery plans and adaptive adjustment strategies. For momentary interruptions in the communication link, the system initiates a fast reconnection protocol, reestablishing the physical layer connection within 50 milliseconds. Sustained interruptions trigger a routing detour mechanism, maintaining network connectivity through backup relay nodes. If data link quality is detected to be consistently below the operating threshold, the system automatically switches to a streamlined communication mode, disabling non-critical data services. Resource allocation in degraded topology states utilizes strict priority scheduling to ensure bandwidth for the transmission of flight control commands and critical status information.

[0034] The collaborative analysis of track similarity and communication quality utilizes time alignment technology. The system establishes a unified time reference axis, precisely aligning the sampling points of the two metrics by timestamp. A joint analysis module detects temporal correlations in metric changes and identifies the causal relationship between communication quality degradation and track deviations. The analysis results are used to refine topology adjustment strategies. When track anomalies are confirmed to be causing communication issues, track correction is prioritized over network reconstruction. The system maintains a dynamically adjusted experience database, storing records of various anomaly scenarios to optimize subsequent decision-making.

[0035] The response mechanism's execution monitoring utilizes a distributed observation architecture. Each network node independently evaluates local execution performance and exchanges monitoring information via a data link. The central coordination node synthesizes these local observations to determine whether the global adjustment progress is meeting expectations. If execution deviations exceeding tolerances are detected, supplementary adjustment instructions are issued. Monitoring data is updated in real time to the topology management database, forming a closed-loop feedback control system. The entire process is designed for fault-tolerant execution. The failure of a single node will not affect the overall adjustment process, and the system automatically reassigns unfinished operations.

[0036] See Figure 7 , showing the communication quality evaluation and topology adjustment process of the drone cluster. The upper chart shows the change curve of the communication signal strength of each drone, reflecting the stability of the communication link. Signal strength fluctuations are affected by the flight environment, distance changes and interference factors. The middle chart shows the changes in communication delay. The increase in delay usually occurs when the formation form switches or the environmental interference increases. Latency optimization is the key to maintaining real-time collaborative control. The lower chart comprehensively evaluates the communication quality index and identifies it in combination with the topology adjustment node. The local topology adjustment node triggers small-scale communication optimization, and the global topology reconstruction node (triangle) triggers the communication structure adjustment of the entire formation.

[0037] Example 3: See Figure 4, the specific operational procedures for formation offset node identification, collaborative parameter calculation and fusion processing. The detection of formation offset nodes adopts the kinematic feature analysis method, and the key event nodes are identified by real-time monitoring of the changes in the motion state of the center of mass of the drone group. The system establishes a center of mass position calculation model in the global coordinate system, collects the three-dimensional coordinate data of all drones in the cluster every 200 milliseconds, and calculates its arithmetic mean as the instantaneous center of mass position. The generation of the velocity vector adopts the five-point differential method, and the instantaneous velocity is calculated using the position data of the current moment and the two sampling points before and after. When it is detected that the velocity vector direction angle changes by more than 15 degrees within the adjacent calculation cycle, the system marks the moment as a potential formation offset node candidate event. The candidate event needs to be verified for three consecutive sampling cycles to confirm that the speed change trend is continuous rather than an instantaneous disturbance before it is finally confirmed as a valid formation offset node and the complete motion state parameters of the node are recorded.

[0038] The relative coordinate system conversion process establishes a local reference system based on the formation's center of mass. The calculation of the transformation matrix takes into account the spatial distribution characteristics of the current formation's shape, and the local coordinate axis directions are determined through principal component analysis. The X-axis points in the direction of the formation's motion, the Y-axis is perpendicular to the X-axis and parallel to the ground plane, and the Z-axis completes the right-handed coordinate system. The conversion from global coordinates to local coordinates uses the following transformation relationship:

[0039] in, represents the global coordinates of the i-th UAV, is the coordinate of the formation center of mass, To consider the yaw angle , pitch angle and roll angle The rotation matrix of . The displacement difference is calculated in the local coordinate system to eliminate the influence of the overall movement of the formation and highlight the characteristics of the relative position change of each individual.

[0040] Communication latency measurement utilizes an adaptive heartbeat detection mechanism. The length of the probe packet is dynamically adjusted based on the current network load, starting at 64 bytes and compressed to 32 bytes when the channel is busy. Round-trip latency is measured by recording the time difference between sending a request and receiving a response, deducting a local processing delay compensation value. The sliding window average is calculated using an exponential weighting method, with new measurements weighted at 0.3 and historical values ​​weighted at 0.7, balancing real-time performance with stability. The latency delta is calculated by comparing the relative rate of change of the current average with a baseline value, determined through 30 seconds of continuous measurement during the initial communication link establishment.

[0041] The normalization of the coordination offset parameters uses the piecewise linear mapping method. The normalization benchmark of the track position offset is the maximum allowable spacing of the current formation, which is dynamically adjusted according to the mission phase: 50 meters in the assembly phase, 30 meters in the cruise phase, and 20 meters in the mission execution phase. The normalization benchmark of the communication delay offset uses the maximum allowable delay, taking into account the transmission distance and the number of hops. Calculate, where n is the number of hops and d is the transmission distance (km). The normalized result is clipped to ensure that the output value is within the range [0,1].

[0042] The dynamic weighting coefficient adjustment mechanism is linked to the mission type identification system. The mission type classifier analyzes the current payload state characteristics and operating mode sequence and outputs a probability distribution for the mission type. In reconnaissance mission mode, the track parameter weight is initially set to 0.6. When continuous area coverage is detected, it increases by 0.02 every 5 minutes, with an upper limit of 0.75. In strike mission mode, the communication parameter weight is initially set to 0.7 and increases to 0.8 when the weapon system is activated. A 5-second transition period is set during the weight adjustment process to prevent control oscillation caused by sudden parameter changes.

[0043] The identification of task offset nodes is based on the sensor state transition characteristics. The system defines six basic working states: initialization, standby, scanning, locking, data transmission, and maintenance. The state transition condition matrix records the allowed transition paths between each state. Unconventional transition sequences refer to transition combinations not defined in the matrix. When a state jump or reverse transition is detected, the transition sequence analysis process is initiated. Confirmation of a valid task offset node requires two conditions to be met: the state transition sequence contains at least two unconventional transitions, and the transition interval time is less than 50% of the normal value. The node record contains the transition path diagram and time characteristic parameters.

[0044] The parameter fusion process utilizes a layered weighted architecture. The first layer processes track position offsets, synthesizing the displacement differences between each UAV into a vector and calculating the overall deformation index of the formation. The second layer processes communication delay offsets, distinguishing between inter-UAV communication and ground station communication channels and calculating the delay impact coefficients for each. The third layer performs weighted fusion, selecting a weight template based on the mission type to generate comprehensive collaborative offset parameters. The fusion results are then smoothed using a Kalman filter to eliminate fluctuations caused by measurement noise.

[0045] The correlation analysis between formation and mission offsets utilizes event-synchronized detection technology. The system maintains a time-aligned event log, recording the occurrence and duration of each of the two types of offset nodes. When two types of events are detected occurring within a 5-second window, a correlation analysis process is initiated. This analysis compares the similarity of event characteristic parameters and establishes a possible causal relationship model. The offset parameters of strongly correlated events are jointly optimized, and the weight distribution strategy is adjusted to improve the overall consistency of coordinated control.

[0046] The abnormal excursion handling mechanism includes a three-level response strategy. Level 1 responds to minor excursions by compensating through local track fine-tuning; Level 2 addresses moderate excursions by adjusting the formation configuration and communication scheduling strategy; and Level 3 addresses severe excursions by initiating a mission reorganization process. The response level is determined based on the duration and gradient of the excursion parameters, using fuzzy logic decision-making. During the process, excursion trends are monitored in real time, and the response level is dynamically adjusted, forming a closed-loop control structure.

[0047] The historical migration pattern learning system continuously updates its library of typical migration scenarios. After each mission, the system extracts characteristic parameters of the migration event and, through cluster analysis, classifies it into existing scenarios or creates new ones. Scenario records contain triggering conditions, processing strategies, and effect evaluation data, which are used to optimize real-time decision-making for subsequent missions. The learning process uses an incremental update approach to preserve the continuity and adaptability of historical experience.

[0048] See Figure 8 , demonstrating the motion characteristics of a UAV formation and the calculation process of collaborative offset parameters. The upper graph shows the motion trajectory of the formation's center of mass, with Xs marking detected formation offset nodes. These nodes typically appear at locations where the trajectory direction changes significantly, reflecting adjustments to the formation's overall shape. The middle graph shows the curve of the formation's average speed, with dotted lines marking key nodes where the speed direction changes suddenly. These nodes require special attention, as they may trigger recalculation of collaborative control parameters. The lower graph compares the changing trends of two collaborative parameters: position offset and communication delay offset. Position offset reflects the formation's ability to maintain its shape, while communication delay offset reflects the efficiency of information synchronization. Together, they determine the quality of collaborative control.

[0049] Example 4: See Figure 5 , the specific operational procedures for trajectory control protocol extraction, sensor scheduling implementation and protocol execution monitoring. In a typical multi-aircraft collaborative reconnaissance mission scenario, the system optimizes the extraction and dynamic adjustment of the control protocol by analyzing the trajectory pattern characteristics in the historical mission library. The historical mission library adopts a hierarchical storage structure. The top-level index table records metadata such as mission type, environmental conditions and execution time, and the bottom-level data block stores detailed trajectory control parameter sequences. When a new collaborative flight mission is started, the feature matching engine first extracts the flight path feature descriptor of the current mission, including feature vectors such as track point distribution density, turning radius statistics and altitude change patterns. The matching process uses the nearest neighbor search algorithm to retrieve the ten most similar trajectory pattern records from the historical library. The selection of these records is based on the Euclidean distance calculation of the multidimensional feature space.

[0050] The generation process of the sensor scheduling protocol is tightly coupled to specific mission requirements. Taking the area search mission as an example, the system's predefined payload control rules include five basic scanning modes: spiral expansion scanning, grid line scanning, concentric circle scanning, random Brownian motion, and key area rescanning. Each scanning mode corresponds to a combination of sensor parameter configurations, including optical zoom, infrared sensitivity, scanning speed, and data sampling rate. During mission execution, the rule engine monitors changes in the target area's characteristics in real time. When signs of a suspicious target are detected, it automatically switches from spiral expansion mode to key area rescanning mode. The mode switching instruction includes a transition parameter sequence to ensure data continuity during sensor state transitions.

[0051] The protocol execution consistency verification mechanism operates in a distributed network environment. The verification node selection strategy is based on the communication topology, typically selecting three drones distributed in a triangle as the verification group. During each verification cycle, the master node broadcasts control commands with timestamps, and the verification nodes immediately respond with confirmation signals upon receipt. A voting mechanism counts the number and content of valid responses. When two or more verification nodes report the same execution status, the protocol execution is confirmed to be valid. Exception handling utilizes a progressive response strategy, triggering command retransmission upon the initial detection of an inconsistency. After three consecutive exceptions, the protocol reload process is initiated.

[0052] The following table shows the key parameter configurations of the sensor scheduling protocol in a typical reconnaissance mission.

[0053] Table 1: Parameter configuration of sensor scheduling protocol for area search mission Scan Mode Optical zoom ratio Infrared sensitivity level Scanning speed (m / s) Sampling rate (Hz) Applicable area type Helical expansion scan 4x 3 15 10 Unknown open area Raster Line Scan 6x 4 10 15 Structured Terrain Concentric circle scanning 5x 5 12 12 Point target area Random Brownian motion 3x 2 8 8 Complex interference environment Rescan key areas 8x 6 5 20 Suspicious target identification area The optimization process for the general flight protocol is achieved through iterative learning. Each individual in the genetic algorithm population represents a set of trajectory control parameters, including variables such as velocity profile, altitude profile, and turn rate. The fitness evaluation function comprehensively considers trajectory tracking accuracy, energy consumption, and mission completion time. A tournament strategy is used for selection, randomly selecting five individuals from the population and retaining the best one. A crossover operation uniformly mixes variables in the variable space, and a mutation operation introduces a Gaussian perturbation term. The optimization process runs continuously, with incremental learning performed between missions to gradually improve the adaptability of the protocol parameters.

[0054] The payload control protocol's dynamic adjustment mechanism responds to environmental change signals. When the optical sensor detects a change in lighting conditions exceeding a threshold, it automatically adjusts exposure parameters and filter settings. The infrared sensor's operating mode dynamically adjusts based on the ambient temperature gradient, increasing sensitivity in areas with high temperature differences. The scanning speed is adjusted in sync with the drone's flight speed to maintain a stable ground sampling interval. The data sampling rate is flexibly set based on the remaining storage system capacity and communication bandwidth, and a lossy compression algorithm is activated when resources are limited.

[0055] The following table shows the type determination of general flight protocols and payload control protocols.

[0056] Table 2: Standards and judgment basis for distinguishing between general flight protocols and payload control protocols Distinguishing dimensions General Flight Protocol Payload Control Protocol Task relevance Regardless of the mission type (e.g., formations must be maintained for all missions) Strongly related to the mission type (e.g., reconnaissance requires high sampling rate, strike requires weapon lock parameters) Multiplexing frequency Cross-task reuse rate ≥ 85% (only fine-tuning of environment parameters is required) Cross-task reuse rate ≤ 30% (core parameters need to be completely restructured) Adjust trigger conditions Only affected by flight environment disturbances (such as wind and temperature) Directly triggered by a change in mission objective (e.g., from "area search" to "fixed-point monitoring") Cross-task reuse mechanism and dedicated adaptation process: The reuse process of the general flight protocol: extract the "general protocol template library" (including 10 basic formation modes and 5 obstacle avoidance algorithms) from the historical mission database; input the "environmental disturbance factor" (such as wind speed and air pressure) of the current mission, and use the formula Fine-tuning (weight range 0.1-0.3); Verify the feasibility of reuse: If the adjusted parameters satisfy the tracking error ≤ 0.5 meters and the energy consumption rate ≤ 120% of the baseline value, reuse is carried out directly; otherwise, local parameter optimization is initiated (such as adjusting the turning radius threshold through genetic algorithm).

[0057] Dedicated adaptation process for the load control protocol: Call the "dedicated protocol framework" from the protocol library based on the task type (such as "power inspection" or "agricultural plant protection"); input the "communication delay offset" in the collaborative offset parameter to dynamically adjust the load parameters (for example, when the delay is greater than 200ms, turn off non-critical sensors to reduce the amount of data); through the "task-load matching algorithm" (formula: ) Verify the adaptability. When the matching degree is ≥80%, the protocol is locked. Otherwise, the manual intervention process is triggered.

[0058] The collaborative working logic of the two protocols: the general flight protocol is the "basic layer", which provides the global track control benchmark (such as the formation center of mass trajectory and the minimum safe distance), and all tasks share the same core logic; the payload control protocol is the "application layer", which dynamically adjusts the payload working mode according to the mission requirements based on the general protocol (for example, the general protocol maintains formation flight, and the payload protocol controls a certain drone to turn on the thermal imaging sensor); the collaborative trigger mechanism is that when the payload control protocol needs to adjust the drone position to optimize the sensor's field of view, it sends a "track fine-tuning request" (offset ≤ 2 meters) to the general protocol, and the general protocol uses the PID algorithm to achieve local position correction to avoid overall formation reconstruction.

[0059] The specific process of protocol parameter extraction is as follows: Generic flight protocol extraction: Input the coordination offset parameter ΔP (current track offset) and the environmental disturbance factor W (e.g., a wind speed weighting factor of 0.3-0.7). Track control protocol samples satisfying |ΔP_i - ΔP| × W ≤ 0.5 meters are retrieved from the historical mission database. A dynamic programming algorithm is used to calculate the spatial overlap between the sample track and the current track (threshold ≥ 85%). After selecting the optimal sample, the corrected heading angle is calculated using the formula: sample heading angle + ΔP × 0.02 × W corrected heading angle compensation. This generates the generic flight protocol.

[0060] Load control protocol extraction: Input the communication delay offset ΔT (sliding window average). When ΔT is less than 100ms, call the "standard sensor scheduling template" in the protocol library (for example, a sampling rate of 20Hz and no compression). When ΔT is greater than or equal to 100ms, initiate the delay adaptation rule: sampling frequency = 20Hz - (ΔT-100ms) / 20ms (minimum of 5Hz). Enable the LZW compression algorithm, with a compression ratio of 1 + ΔT / 200ms (maximum of 3:1). Simultaneously, the optimal sensor combination is selected using the balance coefficient α × ΔT + β × sensor accuracy (α = 0.6, β = 0.4) to generate the load control protocol.

[0061] The triggering conditions for the protocol reload mechanism are based on multi-level criteria. Communication delay offset is monitored using a sliding window statistical method with a window size of ten sampling periods. When more than half of the sample values ​​within the window exceed the threshold, a primary anomaly is flagged. The data link packet loss rate is calculated based on sequence number continuity analysis, and an upward trend over three consecutive statistical periods confirms a secondary anomaly. When dual anomaly conditions are met, the system initiates the protocol reload preparation procedure, including state snapshot storage and resource pre-release operations. The reload process utilizes an atomic operation design to ensure that the overall switching of control parameters does not occur in an intermediate state.

[0062] The implementation of the hierarchical control strategy relies on a condition monitoring network. The conventional level corresponds to normal mission execution, with sensors operating in balanced mode. The enhanced level is activated upon detection of critical targets, improving sensor accuracy and sampling density. The emergency level is activated when system resources are limited or interference is encountered, disabling non-essential functions to ensure core missions. Level transition decisions are based on multi-dimensional feature analysis, including indicators such as energy reserves, computational load, and communication quality. Each level has clearly defined functional boundaries and fallback conditions, forming a finite state automaton.

[0063] The distributed decision-making system's coordination mechanism is implemented through an information exchange protocol. Each drone node periodically broadcasts a status summary, including key data such as current location, payload status, and protocol version. Neighboring nodes receive this summary and update their local topology map to check cluster consistency. When local deviations in protocol execution are detected, a negotiation mechanism determines an adjustment strategy, prioritizing the correction solution with the smallest impact. Global inconsistencies trigger a resynchronization process, with the pilot node distributing baseline protocol parameters.

[0064] A historical data-reuse verification algorithm constructs parameter evolution trajectories. After each mission, the system extracts the actual flight path control parameters and compares them with the initial recommended parameters in the historical database. Parameter combinations that show significant improvements are marked as preferred cases and receive higher weight in subsequent matching. The verification process considers the specificity of mission scenarios to avoid cross-scenario parameter generalization errors. Preferred cases are disseminated using a restricted diffusion strategy, sharing only between drone nodes with similar mission types.

[0065] Example 5: The specific implementation process of the construction and dynamic correction mechanism of the collaborative quality assessment model. The input feature processing of the collaborative quality assessment model adopts multi-source data fusion technology, and the eight core feature parameters enter the neural network model through independent preprocessing channels. The mean and variance calculation of the track position offset adopts the sliding window statistical method, and the window length is dynamically adjusted according to the mission stage. The cruise stage is set to 30 seconds and the mission execution stage is shortened to 15 seconds. The statistical processing of the communication delay offset distinguishes between the uplink and the downlink, and calculates the delay distribution characteristics of each link separately. The current value of the flight environment disturbance factor is collected in real time by the airborne meteorological sensor array, and the predicted value is generated based on the time series analysis model, taking into account the pressure gradient and temperature change trend. The success rate statistics of the protocol execution consistency adopt the event counting method to record the voting results in each verification cycle. The overload frequency monitoring records the number of protocol resets per unit time to eliminate the statistical deviation caused by instantaneous fluctuations.

[0066] The neural network model's structural design utilizes a deep cross-feature fusion architecture. A feature cross layer is placed after the input layer to perform a nonlinear combination of track position offset and communication delay offset. The hidden layers utilize a stepped width design, with 128 neurons in the first layer, decreasing layer by layer to 16 neurons in the output layer. The activation function is chosen based on the characteristics of feature distribution. The hidden layer uses the ReLU function to process non-negative features, and the output layer uses the Sigmoid function to constrain the evaluation value within a standard range. The model is trained using mini-batch gradient descent, with each batch containing 32 samples. The learning rate is initially set to 0.001 and adjusted after each round of training using a cosine annealing strategy. The loss function is designed as a weighted binary cross entropy, which imposes asymmetric penalties for overestimation and underestimation of the evaluation value.

[0067] The weighted evaluation function is constructed using a hierarchical linear regression approach. The first layer processes the track position offset and establishes a linear mapping relationship between the position error and the evaluation value. The second layer superimposes the impact of the communication delay offset and captures the synergistic effect of the two types of offsets through interaction terms. The flight environment disturbance factor acts as a multiplicative correction term, applying an environmental attenuation coefficient to the basic evaluation value. The screening process of historical mission execution data uses time decay weighting, with recent mission data receiving a higher weight. Elastic network regression is used for weight coefficient training to balance the influence of L1 and L2 regularization terms to prevent overfitting. Moving window statistics are used for feature normalization to continuously update the mean and variance estimates.

[0068] A closed-loop control strategy is used to adjust the update frequency of the formation reconstruction parameter set. The input variable of the PID controller is the short-term rate of change of the collaborative quality assessment value, and the output variable is the adjustment amount of the parameter update interval. The proportional component quickly responds to fluctuations in the assessment value, the integral component eliminates steady-state errors, and the differential component suppresses overshoot. Control parameters are preset with baseline values ​​based on the mission type. Reconnaissance missions use a slower update rhythm, while strike missions adopt an agile update mode. A safety margin is set during the frequency adjustment process, with a minimum of no less than 1Hz and a maximum of no more than 10Hz to prevent system oscillation in extreme situations.

[0069] The online learning mechanism for the evaluation function weight coefficients uses an incremental update approach. Training data is streamed in real time, and the model performs parameter fine-tuning every five minutes. Feature importance analysis uses a permutation test, randomly shuffling single-dimensional feature values ​​to observe the magnitude of change in the evaluation value. The weight update step size for important features is increased accordingly, while a conservative adjustment strategy is adopted for less important features. Soft constraints are implemented on the weight coefficients, and the projected gradient method is used to ensure that each coefficient remains within a reasonable range. A rolling window of historical weight versions is maintained, enabling rapid rollback to a stable state.

[0070] The quantitative processing of flight environmental disturbance factors utilizes a multi-dimensional comprehensive evaluation. Three-dimensional wind speed data is vector-decomposed to calculate the strength of each axial component, and atmospheric density parameters are converted into aerodynamic influence coefficients. The disturbance level is divided into ten discrete levels, each with a corresponding control compensation strategy. The correlation model between environmental factors and assessment values ​​utilizes a piecewise linear approximation, establishing local linear relationships within different disturbance intervals. Real-time disturbance monitoring utilizes an outlier filter to eliminate misjudgments caused by transient sensor noise.

[0071] The output interface for collaborative quality assessment values ​​is designed with downstream system requirements in mind. The raw assessment value is filtered through a sliding average and converted to a percentage score. Score trends are calculated using first-order differences, with additional status markers for increasing, stable, and decreasing. Assessment results are distributed using a publish-subscribe model, with the trajectory planning module and communication scheduling module registering their respective assessment dimensions of interest. Key assessment events trigger an alert notification mechanism, which is broadcast to relevant nodes via the data link.

[0072] Historical mission data is stored and retrieved using a spatiotemporal index structure. Mission records are indexed by execution date and geographic region, supporting multi-dimensional queries. A data compression algorithm preserves key features while discarding redundant details. Retrieval utilizes an approximate nearest neighbor search to accelerate matching of similar mission patterns. Data updates are maintained under version control, retaining a record of all revisions for retrospective analysis.

[0073] The model's adaptive adjustment mechanism incorporates multiple protection strategies. When output values ​​are detected to be outside the acceptable range for five consecutive evaluation cycles, a model diagnostic process is initiated. This process checks for deviations in the input feature distribution and verifies model parameter anomalies. In the event of a serious anomaly, a simplified backup model is used to ensure uninterrupted basic evaluation functionality. The model recovery process utilizes a gradual hot restart, gradually reintroducing complex features. The entire evaluation system is designed as a fault-tolerant architecture, ensuring that a single module failure will not result in a loss of overall functionality.

[0074] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0075] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for cooperative control of multiple UAVs based on data link networking, characterized in that: include: Constructing a swarm collaborative mission data set, wherein the swarm collaborative mission data set includes flight path characteristics, load state characteristics, and communication link characteristics; Extracting dynamic coordination nodes from the fleet coordination task data set, wherein the dynamic coordination nodes include a fleet formation switching node and a task target switching node; Generate a track coordination trigger signal based on the dynamic coordination node, and identify the cluster communication topology change node based on the track coordination trigger signal. The cluster communication topology change node includes a local topology adjustment node triggered by formation form switching and a global topology reconstruction node triggered by mission target switching; Generate collaborative offset parameters based on the classification of cluster communication topology change nodes, wherein the collaborative offset parameters include the track offset during formation form switching and the communication delay offset during mission target switching; Extracting swarm collaboration protocol parameters based on the collaboration offset parameters, wherein the swarm collaboration protocol parameters include a general flight protocol that is reusable across missions and a payload control protocol dedicated to a specific mission, where the specific mission is an operation scenario requiring a dedicated payload configuration or a differentiated collaboration strategy within a preset range; The UAV cluster is controlled to perform collaborative flight missions based on the swarm collaboration protocol parameters.

2. The method for cooperative control of multiple drones based on data link networking according to claim 1 is characterized in that: The constructing of the cluster collaborative task data set includes: Extract the initial formation parameter set and corresponding flight time period based on flight path characteristics; Extracting a formation reconstruction parameter set and a corresponding switching period based on a formation morphology switching node; The initial formation parameter set and flight period are processed by the formation verification algorithm to generate a revised formation reconstruction parameter set; The initial formation parameter set and the revised formation reconstruction parameter set are fused to output the fleet collaborative task data set, in which the flight environment disturbance factor is introduced when the formation verification algorithm calculates the flight time deviation.

3. The method for cooperative control of multiple drones based on data link networking according to claim 2 is characterized in that: Generating a track coordination trigger signal includes: Extracting a formation reconstruction parameter set to generate a formation trigger signal; Extracting the corrected formation reconstruction parameter set to generate a task switching signal; Identify track coordination trends based on formation trigger signals and task switching signals; The communication link feature sampling frequency is dynamically adjusted according to the track collaboration trend to generate a track collaboration trigger signal.

4. The method for cooperative control of multiple UAVs based on data link networking according to claim 3 is characterized in that: The identification of cluster communication topology change nodes includes: Calculate the similarity between the track coordination trigger signal and the track of the historical topology adjustment node; When the track similarity exceeds the preset similarity threshold, it is marked as a local topology adjustment node; When the track similarity is lower than the preset similarity threshold, it is marked as a global topology reconstruction node; The track similarity is calculated by combining the track spatial distribution overlap rate with the time synchronization deviation.

5. The method for cooperative control of multiple UAVs based on data link networking according to claim 1, characterized in that: Generating the coordinated offset parameter includes: Extract the local topology adjustment nodes triggered by formation shape switching and mark them as formation offset nodes; Extract the global topology reconstruction node triggered by task target switching and mark it as the task offset node; Calculate the track position offset based on the formation offset node, and mark it as the track coordination offset parameter; Calculate the communication delay increment based on the task offset node, marked as the communication coordination offset parameter; The track cooperative offset parameter and the communication cooperative offset parameter are fused to generate the cooperative offset parameter.

6. The method for cooperative control of multiple UAVs based on data link networking according to claim 5 is characterized in that: The extraction of cluster collaboration protocol parameters includes: The collaborative offset parameters are analyzed based on the flight path characteristics, and a cross-mission reusable trajectory control protocol is extracted as a universal flight protocol. Analyze the correlation between payload state characteristics and cooperative offset parameters, and extract a sensor scheduling protocol dedicated to a specific task as a payload control protocol; The general flight protocol is optimized through the track history data reuse verification algorithm.

7. The method for cooperative control of multiple UAVs based on data link networking according to claim 6, characterized in that: The method of controlling a swarm of drones to perform a collaborative flight mission includes: using a general flight protocol as a global trajectory control benchmark, dynamically adjusting the working mode of the swarm's sensors based on a payload control protocol; verifying protocol execution consistency through communication link characteristics, and triggering a protocol reload mechanism when it is detected that the communication delay offset exceeds a limit.

8. The method for cooperative control of multiple UAVs based on data link networking according to claim 7 is characterized in that: It also includes: establishing a swarm collaboration quality assessment model, inputting collaboration offset parameters and swarm collaboration protocol parameters; outputting collaboration quality assessment values ​​to a track collaboration trigger signal generation module; and dynamically correcting the update frequency of the formation reconstruction parameter set based on the collaboration quality assessment values.

9. The method for cooperative control of multiple UAVs based on data link networking according to claim 8, characterized in that: The fleet collaboration quality assessment model includes: constructing a weighted assessment function of track position offset and communication delay offset, introducing a flight environment disturbance factor as a correction term for the assessment function, and training the assessment function weight coefficient through historical mission execution data.

10. The method for cooperative control of multiple UAVs based on data link networking according to claim 4, characterized in that: The node for identifying cluster communication topology changes also includes: When a communication link interruption is detected or the data link quality is lower than the preset threshold, the topology degradation mechanism is triggered; Dynamically selecting the execution priority of a local topology adjustment node or a global topology reconstruction node based on the track similarity and the real-time communication quality evaluation result; The topology degradation mechanism maintains the basic formation communication capability through a preset minimum connected subgraph algorithm.

Images