A distributed intelligent control method for cooperative path planning of a UAV cluster

Abstract

The application discloses a kind of unmanned aerial vehicle cluster cooperative path planning distributed intelligent control method, it is related to unmanned aerial vehicle group cooperative control technical field, including, real-time acquisition unmanned aerial vehicle dynamic characteristic set, and carry out time synchronization, noise filtering and format integration, form local state data set;Local state data set is obtained through neighborhood cooperative analysis and temporary link planning, and relay chain layout data;Environment risk assessment and path feasibility analysis are carried out to relay chain layout data, and candidate path set is generated;Candidate path set is adjusted through local obstacle avoidance and dynamic constraint, and executable path sequence is generated, and through neighborhood communication and cooperative adjustment, cooperative execution path is generated;Cooperative execution path is integrated through task execution result and state update, and group state data is generated;Group state data is obtained through path feedback and cyclic optimization, and group cooperative achievement is obtained.Improvement task completion rate, cooperative efficiency and environmental adaptability.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm collaborative control technology, and in particular to a distributed intelligent control method for UAV swarm collaborative path planning. Background Technology

[0002] With the rapid development of UAV technology, especially in multi-UAV collaborative operation scenarios, path planning and intelligent control technologies have become research hotspots. Existing technologies mainly rely on centralized or semi-centralized control strategies, achieving path planning and task allocation for multiple UAVs through unified scheduling of UAV flight status, position, and task information. In centralized methods, the status information of all UAVs needs to be converged to a central control unit, which performs path calculation, task allocation, and obstacle avoidance planning. Semi-centralized methods typically involve information aggregation and collaborative decision-making within a local group or subset of UAVs. Current multi-UAV path planning methods have implemented various path generation strategies based on graph search, heuristic optimization, genetic algorithms, and artificial potential fields, achieving certain results in task coverage, path shortestification, and energy consumption optimization. These methods can effectively support UAV swarms in performing complex tasks under specific environments, such as terrain reconnaissance, logistics transportation, or disaster monitoring.

[0003] Multi-UAV cooperative path planning primarily relies on static or near-real-time information for path calculation, but its ability to support real-time acquisition of local dynamic characteristics of UAVs and multi-level cooperative adjustments is limited. Specifically, traditional methods often employ discretization or empirical strategies in neighborhood cooperative analysis, temporary link planning, and environmental risk assessment and dynamic adjustment of path feasibility, making it difficult to achieve global cooperative optimality while ensuring safety. Furthermore, existing solutions typically lack continuity and closed-loop optimization mechanisms for group state feedback and iterative optimization during task execution, thus affecting the stability and efficiency of path execution. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a distributed intelligent control method for collaborative path planning of UAV swarms to solve the problems of real-time executability and collaborative optimization of paths for UAV swarms in complex environments.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms, which includes real-time acquisition of UAV dynamic feature sets, and time synchronization, noise filtering and formatted integration to form a local state dataset.

[0008] Relay chain layout data is obtained by using local state datasets through neighborhood collaborative analysis and temporary link planning;

[0009] Environmental risk assessment and path feasibility analysis are performed on the relay chain layout data to generate a set of candidate paths;

[0010] The candidate path set is adjusted through local obstacle avoidance and dynamic constraints to generate an executable path sequence, and then a collaborative execution path is generated through neighborhood communication and collaborative adjustment.

[0011] The collaborative execution path is integrated with task execution results and status updates to generate group status data;

[0012] By using path feedback and iterative optimization, the group's state data can be used to obtain the results of group collaboration.

[0013] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms described in this invention, the specific steps for forming a local state dataset are as follows:

[0014] The dynamic feature set of the UAV is recorded and classified through continuous time series to form the original state vector, and then a synchronized corrected state vector is generated through time synchronization and multi-source correction.

[0015] The synchronous correction state vector is subjected to nonlinear filtering and dynamic normalization to generate a purified state vector;

[0016] The cleanup state vectors are integrated with neighborhood identifiers through a time series matrix to form a local state dataset.

[0017] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms described in this invention, the specific steps for obtaining relay chain layout data are as follows:

[0018] The local state dataset is used to form a preliminary topology structure through neighborhood drone identification and communication link analysis, and a temporary link layout vector is obtained through collaborative weight allocation and neighborhood link adjustment.

[0019] The temporary link layout vector is coupled with environmental risk and attitude adaptive adjustment to generate relay link layout data.

[0020] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms described in this invention, the specific steps for generating relay link layout data by coupling temporary link layout vectors with environmental risks and adjusting attitude adaptability are as follows:

[0021] The temporary link layout vector is used to generate an environmental risk dataset by collecting environmental monitoring data and calculating risk factors, and attitude adaptability index is generated by performing attitude feasibility analysis and dynamic constraint evaluation of relay nodes.

[0022] The attitude adaptability index is collaboratively scored and ranked to generate a priority sequence for relay chain layout. The relay chain node set is generated through local position fine-tuning and attitude feasibility verification.

[0023] The relay chain node set is integrated, node information is synchronized with global broadcast, and relay chain layout data is generated.

[0024] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms according to the present invention, the specific steps for generating the candidate path set are as follows:

[0025] Multi-source environmental element fusion and gridded risk mapping analysis are performed on relay chain layout data to form an environmental risk matrix;

[0026] The environmental risk matrix is ​​subjected to nonlinear fractional feasibility assessment and dynamic collaborative calculation to obtain the path adaptability vector, and a candidate path set is constructed through topological sequence sorting and feasibility filtering.

[0027] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms according to the present invention, the specific steps for generating an executable path sequence are as follows:

[0028] The candidate path set is sampled at equal intervals and multi-dimensional dynamic features are extracted to form a path feature vector. Then, a candidate feasible path set is generated by calculating and sorting the comprehensive executability score.

[0029] The candidate feasible path set is reconstructed through time parameterization mapping and continuous dynamic constraints to generate a dynamic feasible trajectory set, and a safe corrected trajectory set is generated through local obstacle avoidance detection and trajectory correction reconstruction.

[0030] The safety correction trajectory set is subjected to neighborhood consistency verification and concurrent conflict resolution to generate an executable path sequence.

[0031] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms according to the present invention, the specific steps for generating the collaborative execution path are as follows:

[0032] The executable path sequence is divided into time windows and voxelized in three-dimensional space to obtain trajectory summary records. Then, a trajectory conflict evaluation set is generated through neighborhood exchange and overlap analysis.

[0033] The trajectory conflict evaluation set is subjected to collaborative scoring and priority sorting to generate a trajectory collaborative score sequence. Then, the conflict-resolved trajectory group is obtained through neighborhood concurrent conflict resolution and local backoff correction.

[0034] The conflict resolution trajectory group is broadcast synchronously with neighborhood confirmation and consistency to generate a collaborative execution path.

[0035] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms according to the present invention, the specific steps for generating swarm state data are as follows:

[0036] The collaborative execution path is generated by collecting flight data and associating it with timestamps to produce trajectory execution records. Then, a clean trajectory execution dataset is obtained by filtering out noise, imputing missing values, and removing outliers.

[0037] Multidimensional index calculation and fractional comprehensive scoring and sorting are performed on the clean trajectory execution dataset to generate the overall group status index. Then, through summarization, integration and normalization, a group status data matrix is ​​generated.

[0038] The group state data is obtained by associating task stage identifiers with the group state data matrix and recording timestamps.

[0039] As a preferred embodiment of the distributed intelligent control method for UAV swarm collaborative path planning described in this invention, the specific steps for obtaining the swarm collaboration results are as follows:

[0040] The group state data is analyzed by comparing path deviation with execution difference to obtain difference feedback records. These records are then normalized and mapped to a unified performance coordinate system to generate a difference mapping matrix.

[0041] The difference mapping matrix is ​​optimized by collaborative optimization scoring and sorting to generate a priority sequence for trajectory optimization. Then, through cyclic adjustment and local reconstruction, a set of optimized trajectories is generated.

[0042] The optimized trajectory set is integrated with trajectory information and status information is synchronized to generate group collaborative results.

[0043] As a preferred embodiment of the distributed intelligent control method for collaborative path planning of UAV swarms described in this invention, the specific steps for integrating trajectory information and synchronizing status information of the optimized trajectory set to generate swarm collaborative results are as follows.

[0044] The optimized trajectory set is parsed and synchronized with a globally unified timestamp to generate time-aligned trajectory data. A clean status indicator dataset is generated by extracting status indicators and removing anomalies.

[0045] The collaborative contribution score of the clean state index dataset is calculated and sorted to generate a group collaborative priority sequence. Through trajectory integration and state synchronization, a unified group trajectory information set is generated.

[0046] By integrating and broadcasting a unified set of group trajectory information, collaborative results can be generated.

[0047] The beneficial effects of this invention are as follows: By combining neighborhood collaborative analysis and temporary link planning with environmental risk coupling and attitude adaptive adjustment, relay link layout data is generated, realizing the stable construction and environmental adaptive optimization of communication links between UAVs, and improving path feasibility and collaborative robustness. On this basis, through environmental risk assessment, candidate path generation, local obstacle avoidance and dynamic constraint adjustment, neighborhood conflict resolution and cyclic optimization, the invention achieves the beneficial effects of high-precision path planning, collaborative execution and continuous optimization of group state of UAV swarms in complex dynamic environments, thereby improving task completion rate, collaborative efficiency and environmental adaptability. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A flowchart of a distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms.

[0050] Figure 2 Flowchart for preprocessing local state datasets.

[0051] Figure 3 Flowchart for relay chain layout calculation.

[0052] Figure 4 Build a flowchart for the candidate path set. Detailed Implementation

[0053] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0054] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0056] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms, comprising the following steps:

[0057] S1. Real-time acquisition of UAV dynamic feature sets, followed by time synchronization, noise filtering, and formatting integration to form a local state dataset.

[0058] S1.1 The dynamic feature set of the UAV is recorded and classified through continuous time series to form the original state vector, and a synchronized correction state vector is generated through time synchronization and multi-source correction.

[0059] Furthermore, the dynamic feature set of the UAV is recorded and categorized through continuous time series. The output streams of each sensor are recorded frame by frame according to timestamps and classified and serialized according to feature type, and aggregated to form an original state vector arranged in chronological order with consistent fields. Time synchronization and multi-source correction are performed on the original state vector. Time synchronization eliminates clock offsets from various sources and aligns them to a unified time reference. Multi-source correction calibrates the dimensional differences, calibration deviations and baseline drift of different sources. Correction coefficients are obtained by using overlapping observation comparison or reference signal calibration and applied to the corresponding components to generate a synchronized correction state vector.

[0060] S1.2. Perform nonlinear filtering and dynamic normalization on the synchronous correction state vector to generate the purified state vector.

[0061] Furthermore, the synchronization correction state vector is scanned frame by frame according to time windows. Outlier detection based on median absolute deviation and median filtering are applied to suppress instantaneous impulse noise. Local resampling or interpolation is performed on time periods with significant jumps to maintain temporal continuity (significant jumps refer to situations where the change in the synchronization correction state vector between adjacent frames exceeds the normal fluctuation range, exhibiting sudden increases or decreases; for example, the current component of the synchronization correction state vector experiences a sudden and sharp increase followed by a rapid decrease in a continuous time window, or the voltage component exhibits continuous nonlinear shifts within a short period, causing a break in the time series trend; such time periods are considered significant jumps). A nonlinear filtering algorithm is used to estimate the state of each component of the synchronization correction state vector. The linear filtering algorithm can be either adaptive median filtering or particle filtering. By updating the filtering results within a time window using observation consistency assessment, it can take into account both sudden dynamics and long-term trends, thereby suppressing non-Gaussian noise and sporadic sensor errors. Dynamic normalization is performed on the synchronous correction state vector components after nonlinear filtering. Dynamic normalization uses a sliding window to statistically analyze local minimum, local maximum, local mean, and standard deviation, and uses interval scaling or Z-score transformation to map components with different dimensions to a unified dynamic range. During normalization, the relative change information of the time series is preserved, and the normalization parameters are smoothly updated during window sliding to avoid abrupt changes. The components that have undergone outlier removal, nonlinear filtering, and dynamic normalization are merged in the original field order to form a purified state vector.

[0062] S1.3 Integrate the cleanup state vectors with the neighborhood identifiers through a time series matrix to form a local state dataset.

[0063] Furthermore, the purification state vector is extracted into a sequence of adjacent purification state vectors according to continuous time changes. The purification state vector at each time moment is expanded into a row or column of a matrix according to the field order to ensure that the time order corresponds to the feature dimension. Indexing and metadata recording are performed on the time series matrix representation, including recording the start and end timestamps, source drone identifier and neighborhood identifier for each time series matrix representation, for subsequent retrieval by source and neighborhood. Multiple time series matrix representations from the same neighborhood are aligned and field consistency is checked according to timestamps. When there are gaps or missing timestamps, time is filled in by linear interpolation or nearest neighbor interpolation to maintain matrix continuity. The aligned time series matrix representations are bound to the corresponding neighborhood identifiers to generate a local state dataset.

[0064] S2. Obtain relay chain layout data by performing neighborhood collaborative analysis and temporary link planning on the local state dataset.

[0065] Existing methods typically perform static or near real-time analysis using the location and communication information of neighboring UAVs, and then rely on fixed rules or simple heuristic algorithms to generate link planning and obtain the relay chain layout. However, they have limited processing capabilities in local dynamic adjustment, collaborative weight allocation, and temporary link optimization, and easily overlook the impact of UAV dynamic status and environmental risks on the relay chain layout.

[0066] This invention performs neighborhood collaborative analysis on local state datasets, combines the dynamic characteristics of UAVs, communication link status, and environmental factors to plan temporary links, and simultaneously performs collaborative weight allocation and link fine-tuning on relay chain nodes. This enables the generation of more dynamic, reliable, and feasible relay chain layout data, achieving real-time optimization of UAV collaborative networks.

[0067] S2.1. The local state dataset is used to form a preliminary topology structure through neighboring UAV identification and communication link analysis, and a temporary link layout vector is obtained through collaborative weight allocation and neighboring link adjustment.

[0068] Furthermore, time-aligned UAV identifiers, location fields, velocity fields, and communication index fields are extracted from the local state dataset. Neighborhood UAVs are identified based on the location and velocity fields, and neighborhood relationships are determined through spatial distance calculation and motion coherence assessment. Communication link analysis is performed on the communication index fields from the local state dataset. Link feasibility is evaluated and line-of-sight information is labeled using received signal strength indication, round-trip delay estimation, and packet loss rate estimation. Based on the communication link analysis results, a preliminary topology is constructed using UAV identifiers as nodes and feasible links as edges. Cooperative weight allocation is performed on the preliminary topology. Link feasibility and task priority indices are normalized, and a cooperative score is calculated to rank link priorities. Neighborhood links are adjusted based on the cooperative score. Link reconstruction is completed by deleting infeasible links, selecting low-latency or high-reliability alternative links, and assigning relay responsibilities. Finally, the adjusted node pairs, corresponding link parameters, and link priority sequences are encoded into a temporary link layout vector in field order.

[0069] It should also be noted that motion coherence determination refers to the consistency analysis of the motion characteristics of different UAVs within the same time window in a local state dataset. This analysis is used to identify which UAVs are spatially coordinated or interconnected in their motion states. Motion coherence determination calculates the motion correlation coefficient or relative velocity difference between adjacent UAVs by comparing their velocity direction, velocity magnitude, and position change trends to determine whether they are in the same motion trajectory trend. For example, when two UAVs maintain similar velocity directions, similar displacement amplitudes, and stable relative distance changes over multiple consecutive time frames, their motions can be considered highly coherent, belonging to the same neighborhood motion group. Conversely, if their velocity directions are opposite or their position change trends are inconsistent, their motion coherence is considered low, and they do not constitute a neighborhood association.

[0070] S2.2. The temporary link layout vector is used to generate an environmental risk dataset through environmental monitoring data collection and risk factor calculation, and attitude adaptability index is generated through relay node attitude feasibility analysis and dynamic constraint evaluation.

[0071] Furthermore, relay node identifiers, location fields, link directions, and link parameters are extracted from the temporary link layout vector. Environmental monitoring data is collected based on ground meteorological observations, UAV airborne sensor observations, and external environmental perception data sources. This data includes time-series records of wind speed and direction, airflow turbulence indices, obstacle location and size, terrain obstruction information, and electromagnetic interference indices. The relay node locations in the temporary link layout vector are spatially registered with the environmental monitoring data collection results. Risk factors are calculated based on the spatial coverage area of ​​each relay node and link. These risk factors are assessed by analyzing the time-series characteristics of wind speed fluctuations, airflow disturbance frequency, and obstacle density around the relay node and its links, combined with terrain undulation and signal obstruction ratio, to evaluate flight path passability. The instantaneous electromagnetic interference intensity is also considered. The rate of change of time quantifies communication reliability. Risk factors are calculated by comprehensively considering the influence of the above-mentioned multi-source environmental parameters, reflecting the safety feasibility and link stability of relay nodes in the current environment. The risk factors are summarized to form an environmental risk dataset organized by relay nodes and link entries. At the same time, based on the attitude capability parameters of relay nodes in the temporary link layout vector, UAV dynamic constraint parameters and current aircraft status information, attitude feasibility analysis of relay nodes is carried out. Attitude feasibility analysis includes the reachability test of pitch, roll and yaw angle range and turning rate, the verification of thrust and torque margin, and the consistency test of flight envelope under the action of power supply and payload. The results of the attitude feasibility analysis of relay nodes and the unsatisfied items in the dynamic constraint assessment are recorded in parallel and attitude adaptability indexes are generated through executability discrimination rules.

[0072] It should also be noted that the executability judgment rule is determined by comparative analysis of the attitude changes, dynamic response, and flight stability of relay nodes in historical flight mission data and experimental flight records. The flight stability performance under different attitude parameter combinations (including pitch angle, roll angle, yaw angle, and their rate of change) is statistically analyzed, and different attitude parameter combinations are matched with dynamic constraint parameters and thrust margin. Through threshold division and conditional logic expressions, a set of parameter combinations that meet the conditions of attitude stability, dynamic reachability, and energy continuity is extracted. The conditional logic structure of each parameter combination set is formalized into a judgment criterion, forming an executability judgment rule, which is used to determine whether the actions of the relay node under given attitude and dynamic conditions are feasible.

[0073] S2.3. Perform collaborative scoring and ranking of attitude adaptability indicators to generate a priority sequence for relay chain layout. Then, generate a relay chain node set through local position fine-tuning and attitude feasibility verification.

[0074] Furthermore, executability levels, non-compliance lists, risk factors, link feasibility indicators, and task priority indicators are extracted item by item from the attitude adaptability indicators and environmental risk dataset. Following the executability level mapping rules, risk factor impact rules, and link feasibility judgment rules, each indicator is converted into a collaborative score using a scoring function. This collaborative score is used to sort relay nodes and links in descending order to form a relay chain layout priority sequence. Local position fine-tuning is then performed on each priority item in the relay chain layout priority sequence. This local position fine-tuning involves sampling with small displacements within the spatial coverage area of ​​the relay nodes and verifying attitude feasibility at each sampling position. Completed, the attitude feasibility verification includes pitch, roll, and yaw angle reachability verification, thrust and torque margin verification, and power supply and payload constraint verification. The attitude feasibility verification clearly marks the unsatisfactory items and removes unexecutable sampling positions. During the loop of local position fine-tuning and attitude feasibility verification, the verified positions, the corresponding attitude adaptability index update results, and the adjusted priority relationship in the relay chain layout priority sequence are recorded. The relay node identifiers, fine-tuned position fields, attitude adaptability index update results, and priority fields in the relay chain layout priority sequence that pass the verification are summarized in a unified format to form a relay chain node set.

[0075] It should also be noted that the executability level mapping rule is derived from the statistical modeling results of the correspondence between attitude adaptability indicators and dynamic constraint parameters. The executability level mapping rule establishes a mapping relationship between attitude stability and execution success rate by analyzing the feasibility distribution of different attitude combinations (including pitch, roll, and yaw angles) and actual execution results in multiple flight samples. This is used to convert the continuous range of attitude adaptability indicators into discrete executability levels, such as "fully executable," "conditionally executable," and "unexecutable."

[0076] The risk factor impact rules are derived from the historical statistical relationship analysis between risk factors in the environmental risk dataset and mission execution stability. Through multidimensional regression analysis of the correlation between environmental parameters such as wind speed fluctuations, obstacle density, terrain undulation, and electromagnetic interference change rate and link interruption rate or flight deviation rate, the influence of risk factors on link reliability and attitude control accuracy within different ranges is determined, thus forming risk factor impact rules. These rules are used to map the numerical values ​​of each risk factor in the environmental risk dataset into quantitative impact weights.

[0077] The link feasibility judgment rule is derived from the statistical matching of communication link analysis results and task execution success rate. By performing correlation analysis on the records of link maintenance time and data transmission integrity under different signal strength, round-trip delay, and packet loss rate conditions, the judgment threshold and feasibility level range of feasible links are determined, which are used to determine whether the link has the conditions for cooperative transmission in the relay link layout optimization.

[0078] The scoring function converts multidimensional indicators extracted from attitude adaptability metrics, environmental risk datasets, and link feasibility analysis into a unified numerical collaborative score. Taking executability level, risk factors, link feasibility metrics, and task priority metrics as inputs, the scoring function mathematically combines the normalized results of each metric to calculate the overall executability and collaborative effectiveness of each relay node and link. The scoring function is constructed based on historical task execution data and multi-factor correlation analysis results, ensuring comparability of different dimensional indicators under the same scale. This generates a collaborative score reflecting execution stability, communication reliability, and task urgency, which is used as the basis for prioritizing relay link deployment.

[0079] S2.4 Integrate the node information of the relay chain node set and synchronize it with the global broadcast to generate relay chain layout data.

[0080] Furthermore, the relay node identifier, fine-tuned position field, attitude adaptability index update result, and priority field from the relay chain layout priority sequence are extracted from the relay chain node set. Field integrity verification and timestamp consistency verification are performed on the extracted fields. When there is a duplicate relay node identifier, the nearest neighbor relationship is determined by comparing the position distance and priority field between nodes. The node record with higher priority is retained and its position and attitude index are updated, thus completing the merging of position nearest neighbors and resolution of priority conflicts. The merged and verified relay node entries are serialized into relay chain layout entries according to the field order and a synchronization timestamp and source UAV identifier are attached. Then, the consistency verification of the relay chain layout entries is performed and a message payload for broadcast is generated. The broadcast message payload is propagated hop by hop through the neighborhood communication protocol and the version comparison and overlay rules are performed at the receiving end to achieve global broadcast synchronization. The set of relay chain layout entries after receiving confirmation is summarized into relay chain layout data in a unified format.

[0081] It should also be noted that the overwrite rule is used to determine whether a new entry should replace an old one when the same relay chain layout entry already exists at the receiving end. The overwrite rule is a judgment criterion set during the communication protocol design phase based on the relationship between the old and new timestamps, the priority of the version number, and the trustworthiness of the source node. By comparing the synchronization timestamp and version identifier in the entry, if the new entry has a newer time and a higher version number, then overwriting is performed; if the versions are the same, the original record is maintained according to the trustworthiness of the source node, thereby ensuring the consistency and timeliness of the relay chain layout data across the entire network.

[0082] S3. Conduct environmental risk assessment and path feasibility analysis on the relay chain layout data to generate a set of candidate paths.

[0083] S3.1 Perform multi-source environmental element fusion and gridded risk mapping analysis on the relay chain layout data to form an environmental risk matrix.

[0084] Furthermore, relay node identifiers, location field sets, and attitude adaptability index update sets are extracted from the relay chain layout data. Simultaneously, environmental monitoring data such as wind speed, air pressure, humidity, obstacle density, signal interference intensity, and terrain undulation are collected. Spatial registration and temporal alignment of the environmental monitoring data are performed according to the spatial coordinates in the location field set to establish the correspondence between relay node locations and environmental elements. After spatial registration, a multi-source environmental element fusion operation is performed. Through feature dimension unification and numerical normalization, each environmental monitoring data is mapped to a unified risk expression range, resulting in a fused environmental element vector set. Then, the fused environmental... The feature vector set is used as input. The coverage area is divided into grids according to spatial resolution. Within each grid point, spatial distance interpolation and environmental feature gradient change analysis are used to interpolate and perform local risk analysis on the environmental features of neighboring relay nodes. The spatial distance interpolation determines the interpolation ratio based on the geometric proximity between the node and the grid point. The environmental feature gradient change analysis is used to identify areas of sudden wind speed changes, areas of concentrated obstacles, and areas of high risk of signal interference to generate grid point risk values. All grid point risk values ​​are reorganized into a two-dimensional matrix structure according to spatial coordinate order. Each element in the matrix corresponds to a grid point risk value, forming an environmental risk matrix.

[0085] S3.2 Perform nonlinear fractional feasibility assessment and dynamic collaborative calculation on the environmental risk matrix to obtain the path adaptability vector, and construct a candidate path set through topological sequence sorting and feasibility filtering operations.

[0086] Furthermore, based on the node order in the environmental risk matrix and relay chain layout data, possible paths are represented as grid sequence or node sequence. For each path, the grid risk value and path geometric cost are accumulated spatially to form a path local element set. A feasibility score is calculated for this path local element set. By combining the path's accumulated risk value with constraints such as path length, node reachability indicators, and communication link availability, a nonlinear fractional form is used to calculate the feasibility score for each path. The risk value is used as the numerator, and the sum of the length and reachability indicators is used as the denominator. A nonlinear transformation method is used to adjust the ratio curve to ensure numerical stability and dynamic response, forming the initial feasibility score for the path. Dynamic collaborative calculation is then performed, through communication between neighboring nodes... Information exchange and iterative updates integrate neighborhood link availability, time-varying wind field impact, and relay node attitude adaptability indicators into path feasibility scores, thereby generating modified feasibility values ​​that reflect the collaborative impact of the group. The modified feasibility value, maximum grid point risk, cumulative risk, and reachability indicators of each path are arranged in a fixed order to form a path adaptability vector. After obtaining the path adaptability vector set, the topology sequence is sorted, arranged in descending order according to the primary feasibility metrics and then serialized and sorted using secondary metrics to form a topology sequence. Feasibility filtering is performed, eliminating paths in the topology sequence that do not meet dynamic constraints, communication link constraints, or attitude feasibility verification. The remaining path items are then set into a candidate path set based on the sorting results.

[0087] S4. The candidate path set is adjusted through local obstacle avoidance and dynamic constraint adjustment to generate an executable path sequence, and a collaborative execution path is generated through neighborhood communication and collaborative adjustment.

[0088] S4.1. Perform equidistant sampling and multi-dimensional dynamic feature extraction on the candidate path set to form a path feature vector. Then, generate a candidate feasible path set by calculating and sorting the comprehensive executability score.

[0089] Furthermore, for each path in the candidate path set, interpolation is performed on the path geometry according to the principle of equidistant sampling to generate an equidistantly distributed waypoint sequence. The interpolation process can use linear interpolation or spline interpolation methods to ensure that the waypoint spacing remains consistent. At each waypoint location, multidimensional dynamic features are extracted, including the rate of change of heading, curvature, tangential velocity, normal acceleration, rate of change of acceleration, energy consumption indicators calculated based on velocity and acceleration, risk values ​​of grid points corresponding to the environmental risk matrix, path clearance information, and communication availability features obtained based on relay chain layout data and relay chain node sets. The multidimensional dynamic features are combined in a unified order to form a path feature vector, which includes local statistical features and cumulative statistical features to characterize the path in space. The system comprehensively considers the dynamic characteristics of the path, including time and energy consumption dimensions. It calculates a comprehensive executability score for the path feature vector by normalizing each multi-dimensional dynamic feature in the vector and combining it with the analysis of inter-feature correlations. A preset scoring formula is used to merge the normalized features into a comprehensive executability score, reflecting the feasibility of the path in terms of overall stability, drivability, energy efficiency, and communication continuity. This score serves as the basis for subsequent path ranking. All paths are ranked according to the comprehensive score, prioritizing paths with higher scores. The ranking results are then screened one by one using dynamic constraint verification, obstacle avoidance gap verification, and communication link availability verification. Paths that meet all verification conditions are organized and output as a candidate feasible path set based on the score results.

[0090] S4.2. The candidate feasible path set is reconstructed through time parameterization mapping and continuous dynamic constraints to generate a dynamic feasible trajectory set, and a safe corrected trajectory set is generated through local obstacle avoidance detection and trajectory correction reconstruction.

[0091] Furthermore, time-parameterized mapping is performed on each path in the candidate feasible path set. By establishing a time mapping relationship between equidistant sampling points along the path, the spatial path sequence is converted into a time-series trajectory. The time mapping process dynamically adjusts the sampling interval based on the UAV's velocity characteristics, heading rate of change, and attitude smoothness to ensure the trajectory's continuity in the time dimension and its physical executability. Continuous dynamic constraint reconstruction is then performed on the time-parameterized trajectory. The trajectory is progressively corrected using the UAV's maximum acceleration constraint, angular velocity continuity constraint, and attitude smoothness constraint, ensuring that the trajectory's velocity, acceleration, and angular velocity changes remain continuous in the time domain and satisfy the dynamic characteristics of the flight platform. Finally, after dynamic constraints... Local obstacle avoidance detection is performed on the corrected trajectory. By combining the environmental risk matrix and relay chain layout data, the minimum spatial distance, relative velocity, and safe clearance between trajectory waypoints and obstacle grid points are analyzed to mark trajectory segments with potential collision risks. For trajectory segments with risks, trajectory correction and reconstruction are performed. The waypoint positions are adjusted by local curve smoothing offset and attitude direction fine-tuning to ensure that the trajectory avoids obstacle areas while maintaining continuous dynamic constraints. After all trajectory corrections are completed, the integrity of the corrected trajectory set is verified to check the matching consistency of trajectory time series, velocity series, and attitude series. The trajectory set that meets the requirements of continuous dynamic constraints, obstacle avoidance constraints, and attitude feasibility is output as the safe corrected trajectory set.

[0092] It should also be noted that dynamic constraints refer to the physical motion limitations that a UAV must meet during flight, including the maximum allowable values ​​of speed, acceleration, and angular velocity, as well as the smoothness requirements for continuous changes in the trajectory. During time-parameterized mapping and trajectory correction, dynamic constraints ensure that the UAV does not exceed its thrust capability, inertial limits, or produce uncontrollable flight maneuvers in actual flight by checking and correcting the speed and acceleration changes at trajectory points segment by segment, thereby maintaining the executability and continuity of the trajectory in the time domain.

[0093] Obstacle avoidance constraints refer to the requirement that a trajectory must maintain a safe distance from obstacles, terrain, or environmental risk areas in space to avoid collisions or entering high-risk areas. During trajectory correction, obstacle avoidance constraints analyze the minimum spatial distance, relative velocity, and safe clearance between trajectory waypoints and obstacle grid points in the environmental risk matrix. This allows for local curve smoothing and attitude adjustments to potential collision trajectory segments, ensuring that the corrected trajectory maintains a safe distance from obstacles in space.

[0094] Attitude feasibility requirements mean that the pitch, roll, yaw angles, and turning rates of the UAV during trajectory flight must be within the dynamically achievable range and matched with the capabilities of the UAV airframe and propulsion system. During trajectory correction, attitude feasibility requirements ensure that the corrected trajectory is achievable in attitude control and will not lead to attitude loss by checking whether the attitude changes corresponding to the trajectory waypoints exceed the UAV's allowable angle range, turning rate, and power margin.

[0095] S4.3 Perform neighborhood consistency verification and concurrent conflict resolution on the safety correction trajectory set to generate an executable path sequence.

[0096] Furthermore, a neighborhood consistency check is performed on each trajectory in the safety correction trajectory set. By dividing the trajectory into neighborhood voxels in three-dimensional space and mapping the waypoints to the neighborhood voxels, the overlap and proximity of the trajectory with adjacent trajectories in terms of spatial position, velocity direction, and timestamp are extracted to form a neighborhood conflict index. Based on the neighborhood conflict index, concurrent conflict resolution is performed. For potential collisions or interference between trajectories, the local position, velocity, and time parameters of the waypoints are adjusted sequentially. Through local backtracking correction or fine-tuning of the heading, the trajectory avoids neighborhood conflicts while maintaining continuous dynamic constraints. After the conflict resolution of all trajectories is completed, the consistency of the adjusted trajectory set is verified to ensure that the neighborhood relationship between trajectories, communication link requirements, and path connection sequence meet the group coordination requirements. The trajectory set after neighborhood consistency verification and conflict resolution is organized according to time sequence and spatial sequence, and an executable path sequence is output.

[0097] S4.4. The executable path sequence is divided into time windows and voxelized in three-dimensional space to obtain trajectory summary records. Then, a trajectory conflict evaluation set is generated through neighborhood exchange and overlap analysis.

[0098] Furthermore, each trajectory in the executable path sequence is divided according to a unified timestamp to form a continuous time window sequence. Each time window contains the waypoint position, velocity, attitude, and communication status information of the trajectory within the time period. Within each time window, the waypoints of the trajectory are voxelized in three-dimensional space, and the waypoints are mapped to a spatial voxel grid. The set of trajectory waypoints and their temporal distribution contained in each voxel are recorded to form a trajectory summary record. Based on the voxelization results, a neighborhood exchange analysis is performed. For adjacent trajectory waypoint sets, the relative positional relationship and overlap of waypoints in voxel space are calculated, and overlap indices are extracted, including spatial overlap rate, duration of temporal overlap, and similarity of headings between neighborhoods. Based on the overlap indices, a conflict probability analysis is performed on all trajectory pairs. Trajectory waypoint pairs with spatial overlap rate and duration of temporal overlap significantly greater than the average of neighboring waypoints are marked as potential conflicts, and the conflict intensity and priority are calculated. The neighborhood exchange results and overlap indices of all trajectories within each time window are integrated to generate a trajectory conflict evaluation set.

[0099] S4.5. Perform collaborative scoring and priority sorting on the trajectory conflict evaluation set to generate a trajectory collaborative score sequence. Then, obtain the conflict-resolved trajectory group through neighborhood concurrent conflict resolution and local backoff correction.

[0100] Furthermore, a collaborative scoring calculation is performed on each waypoint pair in the trajectory conflict assessment set. Factors such as spatial overlap, duration of temporal overlap, heading similarity, and trajectory priority are comprehensively calculated to obtain a collaborative score for each trajectory. All trajectories are prioritized based on their collaborative scores to determine the adjustment order during conflict resolution, ensuring that high-priority trajectories maintain their original path continuity. Trajectories with potential conflicts are then subjected to neighborhood concurrent conflict resolution according to priority order. Spatial and temporal conflicts between trajectories are eliminated through local backtracking corrections, fine-tuning waypoint positions, and adjusting velocity and time parameters, while maintaining dynamic constraints and attitude continuity. After all conflict resolution is completed, the integrity of the corrected trajectories is verified to ensure consistency in neighborhood relationships, path continuity, and conflict resolution effects. The trajectory set after collaborative scoring, priority ranking, and local conflict resolution is then compiled and output as a conflict-resolved trajectory group.

[0101] S4.6. Generate a collaborative execution path by broadcasting the conflict resolution trajectory group through neighborhood confirmation and consistency synchronization.

[0102] Furthermore, for each trajectory in the conflict resolution trajectory group, neighborhood confirmation is performed. By analyzing the position of trajectory waypoints in three-dimensional space voxels, the coverage of time windows, and the communication link status of neighboring trajectories, other trajectories with direct neighborhood relationships with each trajectory are identified. The identified neighboring trajectories are broadcast synchronously and uniformly, sharing the timestamps, positions, velocities, and attitude information of trajectory waypoints within the neighborhood. This ensures that all relevant UAVs receive the same trajectory information. During the broadcasting process, potential minor differences between trajectories are dynamically corrected and synchronously adjusted to maintain consistency in spatial, temporal, and dynamic constraints, while ensuring the integrity and continuity of neighborhood communication links. After completing the neighborhood confirmation and synchronous adjustment of all trajectories, the set of corrected and consistent trajectories is organized, and a collaborative execution path is output.

[0103] S5. Integrate the collaborative execution path with task execution results and status updates to generate group status data.

[0104] S5.1. Collect flight data and associate it with timestamps to generate trajectory execution records for the collaborative execution path. Then, through noise filtering, missing value imputation, and outlier removal, obtain a clean trajectory execution dataset.

[0105] Furthermore, during the collaborative execution path flight, flight data is collected for each trajectory waypoint, including the UAV's position, speed, attitude, acceleration, and communication status. The collected data is then correlated with the unified timestamps of the trajectory waypoints to form a trajectory execution record. Noise filtering is performed on the trajectory execution record to remove abnormal fluctuations caused by sensor errors, environmental interference, or communication delays, ensuring smooth and continuous waypoint data. Missing values ​​in the trajectory execution record are imputed by using time-series information or spatial interpolation methods from neighboring waypoints to complete the missing waypoint data, ensuring trajectory integrity. At the same time, outliers in the trajectory execution record are removed by identifying and removing abnormal data whose position, speed, or attitude deviates significantly from neighboring waypoints to prevent outliers from affecting subsequent analysis. The waypoint data after noise filtering, missing value imputation, and outlier removal are then processed and output to generate a clean trajectory execution dataset.

[0106] S5.2 Perform multi-dimensional index calculation and fractional comprehensive scoring and sorting on the clean trajectory dataset to generate the overall state index of the group, and generate the group state data matrix through summarization, integration and normalization.

[0107] Furthermore, multi-dimensional indicators are extracted from waypoints of each trajectory in the clean trajectory execution dataset, including position accuracy, velocity stability, attitude consistency, trajectory deviation, acceleration variation, and neighborhood cooperative consistency. By statistically and dynamically analyzing each indicator along the time series, the performance of each trajectory under different dimensions is obtained. Fractional comprehensive scoring is performed on the multi-dimensional indicators. The indicators of each dimension are calculated according to a predetermined scoring formula to generate a comprehensive performance score for each trajectory. The comprehensive performance scores are then sorted to obtain the overall state index of the swarm, reflecting the coordination and stability of the UAV swarm during trajectory execution. The overall state index of all trajectories is summarized and integrated. The trajectory dimension and indicator dimension information are combined into a two-dimensional matrix. The normalization process is used to unify the dimensional range to ensure the comparability and overall consistency of each indicator. The results after multi-dimensional indicator calculation, fractional comprehensive scoring and sorting, summarization and integration, and normalization are sorted and output to generate a swarm state data matrix.

[0108] It should also be noted that the predetermined scoring formula is used to quantitatively evaluate the performance of each UAV or trajectory execution unit in the collaborative results of the group. It forms a single score value by combining or standardizing indicators such as trajectory execution accuracy, task completion, collaborative consistency, and time synchronization. For example, trajectory deviation, execution error, and collaborative contribution can be mapped to a unified score proportionally, with a higher score indicating better trajectory execution and collaborative effect. The indicators in the formula can be assigned example coefficients according to the task characteristics to reflect the relative importance of different factors, thereby generating a predetermined score that can be used for ranking and prioritization decisions.

[0109] It should also be noted that mission characteristics refer to the specific types of operations, mission objectives, and environmental or operational conditions involved in the execution of missions by the UAV swarm. For example, different missions such as inspection, transportation, monitoring, or formation flight have different focuses on trajectory accuracy, speed response, coordination consistency, and time synchronization. By analyzing the mission's operational objectives, execution constraints, and mission priorities, the relative weights and example coefficients of each indicator in the predetermined scoring formula are determined, so that the score can reasonably reflect the comprehensive performance of the UAV under the current mission.

[0110] S5.3. Perform task stage identification association and timestamp recording on the group state data matrix to obtain group state data.

[0111] Furthermore, the multi-dimensional performance indicators corresponding to each trajectory in the group state data matrix are associated with mission phase identifiers according to the flight mission process. The trajectory waypoints and corresponding performance indicators are mapped to different mission phases, such as takeoff, cruise, target tracking, and landing, forming mission phase association information. The waypoints and mission phase information of each trajectory are timestamped. Each waypoint, performance indicator, and mission phase information in the group state data matrix is ​​bound to a globally unified timestamp to ensure the integrity and traceability of the group state data in the time dimension. The waypoint data after mission phase identification association and timestamp recording are integrated and organized to output the group state data.

[0112] S6. By using path feedback and iterative optimization, the group state data is used to obtain the group collaboration results.

[0113] Existing methods typically provide local feedback through single-path execution results or static path planning, relying on simple accumulation or averaging to evaluate the group state. They lack multi-round cyclic optimization mechanisms, making it difficult to fully reflect the group's collaborative effect and unable to effectively adjust individual trajectories to improve overall collaborative performance.

[0114] This invention analyzes path deviation and execution difference in group state data, and combines a cyclic optimization strategy to adjust the group trajectory in multiple rounds and calculate collaborative scores, thereby generating unified and dynamically updated group collaborative results, achieving continuous optimization of the overall group performance and real-time feedback correction.

[0115] S6.1. The group state data is compared and analyzed by path deviation and execution difference to obtain difference feedback records. The difference is then normalized and mapped to a unified performance coordinate system to generate a difference mapping matrix.

[0116] Furthermore, the actual execution path and expected reference path of each trajectory are extracted from the group state data. The trajectory waypoint position, velocity, attitude and other multi-dimensional indicators are compared point by point. The deviation of each trajectory in terms of spatial position, velocity distribution and attitude change is calculated to form a difference feedback record. The difference feedback record is normalized to map the deviation values ​​of different indicators to a unified dimension range. The normalized difference feedback record is then mapped according to the group collaboration results, so that the deviation information of all trajectories can be compared and analyzed in the same coordinate system. The mapped difference feedback record is integrated according to the trajectory sequence to output the difference mapping matrix.

[0117] It should also be noted that the expected reference path refers to the trajectory sequence generated in advance based on mission planning, group collaboration results, or control strategies. It serves as a standard reference for the actual trajectory of each UAV, including the expected waypoint coordinates and motion states in dimensions such as spatial position, speed, and attitude, so as to facilitate point-by-point deviation analysis and performance evaluation of the actual trajectory.

[0118] S6.2 Perform collaborative optimization scoring and sorting on the difference mapping matrix to generate a trajectory optimization priority sequence, and generate an optimized trajectory set through cyclic adjustment and local reconstruction.

[0119] Furthermore, multidimensional deviation information of each trajectory is extracted from the difference mapping matrix. The collaborative optimization scoring method is used to comprehensively score the spatial position deviation, velocity change deviation, and attitude difference of the trajectory. The comprehensive scoring results are sorted to generate a trajectory optimization priority sequence to ensure that high-deviation trajectories receive higher optimization priority. Each trajectory is cyclically adjusted according to the trajectory optimization priority sequence. By locally correcting the waypoint position, velocity distribution, and attitude parameters of the trajectory segment by segment, local conflicts and discontinuities are eliminated. After each adjustment, continuity verification and feasibility checks are performed to ensure that the trajectory meets flight requirements under three-dimensional space and dynamic constraints. All cyclically adjusted and locally corrected trajectories are integrated to output the optimized trajectory set.

[0120] It should also be noted that the collaborative optimization scoring method is a calculation method used to quantify the degree of deviation and optimization priority of each trajectory in the collaborative execution of the group. The collaborative optimization scoring method extracts multi-dimensional deviation information of the trajectory in terms of spatial position, velocity distribution, and attitude parameters. It normalizes the deviations of each dimension according to a unified dimension and calculates a single comprehensive score through fractional combination or correlation analysis. This score reflects the urgency and improvement value of the trajectory in the overall collaborative execution. For example, indicators such as spatial position deviation, velocity change deviation, and attitude difference are mapped to scoring sub-items. Combined with the trajectory's collaborative contribution to the group, a comprehensive score is generated through a unified calculation rule. The comprehensive score is used to rank and form a trajectory optimization priority sequence, ensuring that trajectories with larger deviations are given priority for local correction and cyclical adjustment during the optimization process, thereby improving the overall consistency and safety of the group's collaborative execution.

[0121] S6.3. The optimized trajectory set is parsed and synchronized with a globally unified timestamp to generate time-aligned trajectory data. Then, a clean status indicator dataset is generated by extracting status indicators and removing anomalies.

[0122] Furthermore, the optimized trajectory set is parsed, and the waypoint positions, velocities, and attitude parameters of each trajectory are synchronized using a globally unified timestamp to form time-aligned trajectory data. This ensures that all trajectories can be directly compared and analyzed under a unified time reference. Multidimensional group state indicators, including position deviation, velocity changes, and attitude stability, are extracted from the time-aligned trajectory data. Anomaly removal methods are used to identify and remove significantly deviated state points, eliminating the influence of noise and outlier data, and ensuring the consistency and reliability of state indicators. All trajectory state data that have undergone time alignment, state extraction, and anomaly removal are integrated to generate a clean state indicator dataset.

[0123] S6.4 Calculate and sort the collaborative contribution scores of the clean state index dataset to generate a group collaborative priority sequence, and generate a unified group trajectory information set through trajectory integration and state synchronization.

[0124] Furthermore, a collaborative contribution score is calculated on the clean state index dataset to quantify the contribution of each UAV's state index to the overall goal during group collaborative execution. The scores are then ranked to form a group collaboration priority sequence, which clarifies the order of each UAV in group collaboration. Based on the group collaboration priority sequence, the trajectories of different UAVs are integrated, and the trajectories of different UAVs are combined in order of priority. During the integration process, the trajectory status is synchronized over time to ensure that all trajectories maintain consistent state indicators and spatial position relationships under a unified time reference. The integrated and synchronized trajectory data forms a unified group trajectory information set.

[0125] S6.5. By integrating and broadcasting a unified set of group trajectory information, group collaborative results are generated.

[0126] Furthermore, the unified group trajectory information set is integrated, and the trajectory status, position and coordination indicators of each UAV are summarized to form a complete information set containing the trajectory and status of all UAVs. Through broadcast synchronization, the integrated trajectory information is globally synchronized among the group of UAVs, so that each UAV obtains the same coordination information and ensures the consistency and real-time nature of information in the group. The group coordination results are generated through the processing completed by information integration and broadcast synchronization.

[0127] In summary, this invention achieves dynamic consistency of the cluster's internal state by: real-time acquisition of UAV dynamic characteristics and subsequent synchronous correction, filtering, and matrix integration to form a high-precision local state dataset, providing a reliable foundation for subsequent collaborative analysis and path planning; generating relay chain layout data through neighborhood collaborative analysis and temporary link planning, combined with environmental risk coupling and attitude adaptive adjustment, realizing stable construction and environmental adaptive optimization of communication links between UAVs, and improving path feasibility and collaborative robustness; and achieving the beneficial effects of high-precision path planning, collaborative execution, and continuous optimization of the group state of UAV clusters in complex dynamic environments through environmental risk assessment, candidate path generation, local obstacle avoidance and dynamic constraint adjustment, neighborhood conflict resolution, and iterative optimization, thereby improving task completion rate, collaborative efficiency, and environmental adaptability.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A distributed intelligent control method for collaborative path planning in unmanned aerial vehicle (UAV) swarms, characterized in that: include, Real-time acquisition of UAV dynamic feature sets, followed by time synchronization, noise filtering, and formatted integration to form a local state dataset; The specific steps to obtain relay chain layout data from the local state dataset through neighborhood collaborative analysis and temporary link planning are as follows: The local state dataset is used to form a preliminary topology structure through neighborhood drone identification and communication link analysis, and a temporary link layout vector is obtained through collaborative weight allocation and neighborhood link adjustment. Temporary link layout vectors are coupled with environmental risk and attitude adaptive adjustment to generate relay link layout data. Static or near real-time analysis is performed using the location and communication information of neighboring UAVs. Then, link planning is generated based on fixed rules or simple heuristic algorithms to obtain the relay link layout. Environmental risk assessment and path feasibility analysis are performed on the relay chain layout data to generate a set of candidate paths; The candidate path set is adjusted through local obstacle avoidance and dynamic constraints to generate an executable path sequence, and then a collaborative execution path is generated through neighborhood communication and collaborative adjustment. The collaborative execution path is integrated with task execution results and status updates to generate group status data; By using path feedback and iterative optimization, the group's state data can be used to obtain the results of group collaboration.

2. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that: The specific steps for forming the local state dataset are as follows: The dynamic feature set of the UAV is recorded and classified through continuous time series to form the original state vector, and then a synchronized corrected state vector is generated through time synchronization and multi-source correction. The synchronous correction state vector is subjected to nonlinear filtering and dynamic normalization to generate a purified state vector; The cleanup state vectors are integrated with neighborhood identifiers through a time series matrix to form a local state dataset.

3. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that: The process of generating relay link layout data by coupling temporary link layout vectors with environmental risks and adjusting them for attitude adaptation involves the following steps: The temporary link layout vector is used to generate an environmental risk dataset by collecting environmental monitoring data and calculating risk factors, and attitude adaptability index is generated by performing attitude feasibility analysis and dynamic constraint evaluation of relay nodes. The attitude adaptability index is collaboratively scored and ranked to generate a priority sequence for relay chain layout. The relay chain node set is generated through local position fine-tuning and attitude feasibility verification. The relay chain node set is integrated, node information is synchronized with global broadcast, and relay chain layout data is generated.

4. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that: The specific steps for generating the candidate path set are as follows: The location field of relay nodes is extracted from the relay chain layout data. Wind speed, air pressure, humidity, obstacle density, signal interference intensity and terrain undulation are registered and time-aligned according to spatial coordinates to establish the correspondence between relay nodes and environmental elements. The environmental elements are then processed in a unified dimension and normalized to form a fused environmental element vector set. The coverage area is divided into grids according to a preset spatial resolution, and spatial distance interpolation is performed at each grid point based on the geometric proximity relationship between the grid point and the neighboring relay node to generate grid point risk value. Multi-source environmental element fusion and gridded risk mapping analysis are performed on relay chain layout data to form an environmental risk matrix; The environmental risk matrix is ​​subjected to nonlinear fractional feasibility assessment and dynamic collaborative calculation to obtain the path adaptability vector, and a candidate path set is constructed through topological sequence sorting and feasibility filtering. The nonlinear fraction refers to a ratio expression that uses the cumulative risk value of the path as the numerator and the combined value of constraints such as path length and node reachability index as the denominator.

5. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that: The specific steps for generating the executable path sequence are as follows: The candidate path set is sampled at equal intervals and multi-dimensional dynamic features are extracted to form a path feature vector. Then, a candidate feasible path set is generated by calculating and sorting the comprehensive executability score. The candidate feasible path set is reconstructed through time parameterization mapping and continuous dynamic constraints to generate a dynamic feasible trajectory set, and a safe corrected trajectory set is generated through local obstacle avoidance detection and trajectory correction reconstruction. The safety correction trajectory set is subjected to neighborhood consistency verification and concurrent conflict resolution to generate an executable path sequence.

6. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that: The specific steps for generating the collaborative execution path are as follows: The executable path sequence is divided into time windows and 3D spatial voxels. Within each time window, the trajectory waypoints are 3D spatial voxels are performed and the waypoints are mapped to the spatial voxel grid. The trajectory waypoint set and time distribution contained in each voxel are recorded to obtain the trajectory summary record. Through neighborhood exchange and overlap analysis, a trajectory conflict evaluation set is generated. The trajectory conflict evaluation set is subjected to collaborative scoring and priority sorting to generate a trajectory collaborative score sequence. Then, the conflict-resolved trajectory group is obtained through neighborhood concurrent conflict resolution and local backoff correction. The conflict resolution trajectory group is broadcast synchronously with neighborhood confirmation and consistency to generate a collaborative execution path.

7. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that: The specific steps for generating the group state data are as follows: The collaborative execution path is generated by collecting flight data and associating it with timestamps to produce trajectory execution records. Then, a clean trajectory execution dataset is obtained by filtering out noise, imputing missing values, and removing outliers. Multidimensional index calculation and fractional comprehensive scoring and sorting are performed on the clean trajectory execution dataset to generate the overall group status index. Then, through summarization, integration and normalization, a group status data matrix is ​​generated. The group state data is obtained by associating task stage identifiers with the group state data matrix and recording timestamps.

8. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that: The specific steps to obtain the results of group collaboration are as follows. The group state data is analyzed by comparing path deviation with execution difference to obtain difference feedback records. These records are then normalized and mapped to a unified performance coordinate system to generate a difference mapping matrix. The difference mapping matrix is ​​optimized by collaborative optimization scoring and sorting to generate a priority sequence for trajectory optimization. Then, through cyclic adjustment and local reconstruction, a set of optimized trajectories is generated. The optimized trajectory set is integrated with trajectory information and status information is synchronized to generate group collaborative results.

9. The distributed intelligent control method for collaborative path planning of unmanned aerial vehicle (UAV) swarms as described in claim 8, characterized in that: The specific steps for integrating trajectory information and synchronizing status information of the optimized trajectory set to generate group collaborative results are as follows. The optimized trajectory set is parsed and synchronized with a globally unified timestamp to generate time-aligned trajectory data. A clean status indicator dataset is generated by extracting status indicators and removing anomalies. The collaborative contribution score of the clean state index dataset is calculated and sorted to generate a group collaborative priority sequence. Through trajectory integration and state synchronization, a unified group trajectory information set is generated. By integrating and broadcasting a unified set of group trajectory information, collaborative results can be generated.

Images