Unmanned aerial vehicle group dynamic path planning method and system based on cooperative intelligence

Abstract

The invention discloses an unmanned aerial vehicle group dynamic path planning method and system based on cooperative intelligence, and belongs to the field of unmanned aerial vehicle intelligent control, and the method comprises the steps: obtaining a navigation task of an unmanned aerial vehicle group, and marking N task nodes for the task full period of the navigation task; for a navigation task, positioning a key navigation path point; the task path planning module is connected with the cluster path planning module, traverses N task nodes and the key flight path points, carries out cluster division on an unmanned aerial vehicle cluster according to the course, carries out task path planning by taking a node single-machine target as guidance, and determines a path planning strategy; wherein in a manner of sequential planning based on N task nodes, offset calibration is carried out based on an unmanned aerial vehicle distribution network of an upper task node. The technical problems of low cooperation efficiency, high collision risk and poor task continuity caused by delay of centralized planning response of the unmanned aerial vehicle cluster, stiffness of static grouping resources and lack of a calibration mechanism in path error accumulation in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control of unmanned aerial vehicles (UAVs), and particularly to a method and system for dynamic path planning of UAV swarms based on collaborative intelligence. Background Art

[0002] Large-scale UAV collaborative operations have become an industry trend, but complex task scenarios pose higher requirements for path planning.

[0003] Existing single-UAV path planning methods are difficult to cope with the challenges of efficiency and safety in group collaboration. There is an urgent need for an intelligent planning scheme that can dynamically adjust and autonomously collaborate to meet the real-time response requirements in high-density and high-dynamic task scenarios. Summary of the Invention

[0004] This application provides a method and system for dynamic path planning of UAV swarms based on collaborative intelligence, aiming to solve the technical problems in the prior art, such as the late response of centralized planning for UAV swarms, the rigidity of static grouping resources, and the lack of calibration mechanism for path error accumulation, which lead to low collaborative efficiency, high collision risk, and poor task continuity.

[0005] In view of the above problems, this application provides a method and system for dynamic path planning of UAV swarms based on collaborative intelligence.

[0006] In the first aspect disclosed in this application, a method for dynamic path planning of UAV swarms based on collaborative intelligence is provided. The method includes obtaining the navigation tasks of the UAV swarm, identifying N task nodes for the entire task cycle of the navigation tasks; positioning key navigation points for the navigation tasks, where the key navigation points are marked with UAV numbers; connecting to a cluster path planning module, traversing the N task nodes and the key navigation points, dividing the UAV swarm into clusters according to the headings, and performing task path planning with the single-UAV target of the nodes as the guide to determine the path planning strategy; where the sequential planning based on the N task nodes is used as the method, and the UAV distribution network based on the upper-level task nodes is used for offset calibration.

[0007] Another aspect disclosed in this application provides a dynamic path planning system for a drone swarm based on collaborative intelligence. The system includes a navigation task acquisition module for acquiring the navigation tasks of the drone swarm and identifying N task nodes for the entire life cycle of the navigation tasks; a key waypoint positioning module for positioning key waypoints for the navigation tasks, where the key waypoints are marked with drone numbers; a path planning strategy determination module for connecting to a cluster path planning module, traversing the N task nodes and the key waypoints, clustering the drone swarm according to the heading, performing task path planning with the node single-machine target as the guide, and determining the path planning strategy; and an offset calibration module for performing offset calibration in the order of planning based on the N task nodes and using the drone distribution network based on the upper-level task nodes.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] Due to the technical solutions of discretizing nodes in the entire life cycle of the task, binding drone numbers to key waypoints, dynamically clustering according to the heading, and cross-node offset calibration, the technical problems of centralized planning response latency, static grouping resource rigidity, and lack of calibration mechanism for error accumulation in the prior art are solved, achieving the following technical effects: Through dynamic task decomposition and heading-consistent clustering, the multi-aircraft parallel planning efficiency is improved; based on the responsibility attribution of key waypoints and cross-node distribution network calibration, the cross-stage transmission of path deviations is suppressed; combined with cluster collision detection and local path smoothing optimization, the risk of formation conflicts is eliminated. Ultimately, the collaborative response speed of the drone swarm in a complex dynamic environment is improved, the task continuity is enhanced, and the resource utilization rate and safety reach the optimal balance of collaborative operations.

[0010] The above description is only an overview of the technical solutions of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features, and advantages of this application more obvious and understandable, the following specifically illustrates the specific embodiments of this application. Description of the Drawings

[0011] Figure 1 It is a schematic flowchart of a dynamic path planning method for a drone swarm based on collaborative intelligence provided by an embodiment of this application;

[0012] Figure 2 It is a schematic structural diagram of a dynamic path planning system for a drone swarm based on collaborative intelligence provided by an embodiment of this application.

[0013] Description of the reference numerals: Navigation task acquisition module 11, key waypoint positioning module 12, path planning strategy determination module 13, offset calibration module 14. Detailed Embodiments

[0014] The general idea of the technical solution provided by this application is as follows:

[0015] The embodiments of this application provide a method and system for dynamic path planning of an unmanned aerial vehicle (UAV) swarm based on collaborative intelligence. The navigation task is decomposed into N task nodes in the full cycle, and multi-UAV collaborative path planning is achieved by combining key waypoint positioning and heading-based dynamic cluster division. Specifically, first, the UAV numbers are bound to the UAVs through task decomposition and key waypoints to clarify the responsibility attribution; second, the clusters are divided according to the headings, and the local paths guided by the nodes are generated in parallel; finally, the path deviation is dynamically calibrated based on the upper node distribution network to suppress error accumulation.

[0016] After introducing the basic principle of this application, the various non-limiting implementation manners of this application will be specifically introduced below with reference to the accompanying drawings of the specification.

[0017] Embodiment 1

[0018] As Figure 1 shown, the embodiments of this application provide a method for dynamic path planning of an unmanned aerial vehicle (UAV) swarm based on collaborative intelligence, and the method includes:

[0019] Step S100: Obtain the navigation task of the UAV swarm, and identify N task nodes for the full cycle of the navigation task.

[0020] Specifically, the UAV swarm refers to a set of UAVs that can work collaboratively. These UAVs can be interconnected through a communication network to jointly execute complex tasks. The navigation task refers to a specific flight task that the UAV swarm needs to complete, which may include various types such as patrol, monitoring, and transportation. The task usually has clear goals and constraints. The full cycle of the task refers to the entire time process from the start to the end of the navigation task, including the preparation stage, execution stage, and end stage of the task. The task node refers to a key time point or event point set for task management and control during the full cycle of the task. These nodes can be the starting point, end point, turning point, or specific operation point of the task.

[0021] First, clarify the specific content and requirements of the navigation task, including the task's goals, constraints, expected time, etc. Then, conduct a full-cycle analysis and division of the task. During the full cycle of the task, N task nodes are identified according to the logical sequence and key events of the task. These nodes can be the starting point, key operation points in the middle, and end point of the task. To accurately identify these nodes, a geographic information system (GIS) is needed to determine the geographical location of the nodes, a trajectory planning algorithm is used to calculate the flight path of the UAVs, and a task simulation tool is used to verify the feasibility and effectiveness of the nodes.

[0022] By identifying N task nodes in the entire cycle of a mission, the execution process and key events of the mission can be understood more clearly, providing a basis for the flight path planning, task allocation, and collaborative work of unmanned aerial vehicles (UAVs). At the same time, this also helps to improve the execution efficiency and safety of the mission, ensuring that the UAV swarm can complete the mission smoothly according to the predetermined plan.

[0023] Step S200: For the navigation mission, locate the key waypoints, where each key waypoint is marked with a UAV number.

[0024] Specifically, a key waypoint refers to a geographical location or logical node that the UAV swarm must pass through or perform specific operations during a navigation mission. Its selection directly affects the efficiency and safety of mission completion. The UAV number marking assigns a unique UAV identifier to each key waypoint, clearly designating the individual UAV or UAV cluster responsible for performing the mission at that waypoint, ensuring responsibility attribution and path coordination.

[0025] First, extract the key target points and environmental constraint points in the navigation mission through mission parsing and a Geographic Information System (GIS) to generate an initial set of waypoints. Then, use a collaborative task allocation algorithm (such as distributed allocation based on the auction mechanism, Hungarian algorithm) to bind a UAV number to each waypoint: for example, in a logistics mission, according to the UAV's load capacity, current location, and priority, delivery point A is assigned to UAV-03, and obstacle B is responsible for bypassing by UAV-07. During this process, combine a Real-Time Kinematic Global Positioning System (RTK-GPS) with a UAV formation control platform (such as PX4 / ROS) to dynamically calibrate the waypoint coordinates, and synchronize the waypoint-number mapping relationship to the swarm through a mission management interface (such as the MAVLink protocol). If the mission changes dynamically (such as adding a temporary waypoint), adopt a dynamic reallocation strategy (such as the contract net protocol) to adjust the number attribution in real time to ensure mission continuity.

[0026] Through the precise binding of waypoints and UAV numbers, strong collaboration and high reliability in mission execution are achieved. This avoids multiple UAVs competing for the same waypoint.

[0027] Step S300: Connect to the swarm path planning module, traverse the N task nodes and the key waypoints, divide the UAV swarm according to the heading, conduct mission path planning with the single-node target as the guide, and determine the path planning strategy.

[0028] Specifically, the cluster path planning module is a core system integrating algorithms and communication protocols, used to coordinate the path planning of multiple UAVs, support distributed or centralized computing, and ensure that the global path is conflict-free and meets the mission constraints. Traversing task nodes and key waypoints means systematically scanning and prioritizing all nodes and key points during the entire mission cycle to determine the input conditions for path planning. Heading cluster division means grouping the UAV fleet according to the current or expected flight directions of the UAVs, so that the UAVs in the same group have the same heading, reducing the complexity of coordination. Node single-vehicle target means specifying a sub-goal for each task node for a dedicated UAV, ensuring clear responsibilities and focused resources. The path planning strategy refers to the flight path plan generated by integrating cluster division and node targets, including elements such as trajectories, speeds, and obstacle avoidance rules.

[0029] First, access the data of task nodes and key waypoints through the cluster path planning module, and use graph traversal algorithms (such as depth-first search or Dijkstra algorithm) to prioritize the nodes and waypoints to determine the starting point and the necessary path of the path planning. Subsequently, according to the real-time heading data (obtained through the UAV's on-board magnetometer or GPS direction sensor), use the clustering algorithm to divide the UAVs with similar headings into the same cluster. Then, guided by the node single-vehicle target, call the distributed path planning tool (such as Google OR-Tools or Fast-Planner) to generate a locally optimal path for each cluster. Finally, ensure that the paths across clusters are collision-free through the conflict detection and resolution module (such as the velocity obstacle method or the spatio-temporal corridor technique), and integrate them into the global path strategy and synchronize it to the UAV fleet for execution.

[0030] Through the heading cluster division and node-oriented path planning, the coordination efficiency and safety in complex scenarios are significantly improved.

[0031] Step S400: Among them, with the sequential planning based on the N task nodes as the method, perform bias calibration on the UAV distribution network based on the upper-level task node.

[0032] Specifically, sequential planning means generating path planning step by step according to the order of task nodes, and the execution result of the previous node provides input constraints for the subsequent node. The upper-level task node refers to the previous-stage node of the current planning node, and the UAV state after its execution constitutes the calibration benchmark. The UAV distribution network refers to the dynamic topological network formed by the actual spatial distribution and state of the UAV fleet when the upper-level node is completed, and is used to calibrate the initial conditions of the subsequent node planning.

[0033] First, the distributed state synchronization protocol is used to collect the data of the UAV distribution network completed by the upper node in real time (such as using the Redis stream database to store the GPS coordinates and remaining battery of each UAV), and compare it with the expected distribution of this node. Then, the calibration planning engine (such as Model Predictive Control MPC or Dynamic Window Approach DWA) is called to incorporate the offset into the path generation of the current node: for example, superimpose a compensation vector on the path planning of node "Delivery Point 2", or adjust the time window to make the UAV swarm arrive synchronously.

[0034] Through the dynamic calibration mechanism of the upper node distribution network, error suppression and resource coordination across task nodes are achieved.

[0035] Furthermore, for the entire life cycle of the navigation task, N task nodes are identified, including: setting preset navigation variables, where the heading variable is determined by the heading and navigation state, and the preset navigation variables are set through the preset variable scale; according to the entire life cycle of the task, using the preset navigation variables, the task is divided into stages, and N segmented navigation tasks are determined; based on the N segmented navigation tasks, the entire life cycle of the task is marked to determine the N task nodes.

[0036] Specifically, the preset navigation variables refer to a series of parameters or conditions set in advance to describe the navigation characteristics when planning the navigation task of the UAV swarm. These variables may include heading, speed, altitude, navigation state, etc., and are used to refine and control the flight behavior of the UAV. The heading variable specifically refers to the variable describing the flight direction of the UAV, usually represented by an angle or a direction vector. The heading variable is an important part of the preset navigation variables and determines the basic direction of the UAV during flight. The preset variable scale refers to the specific standard or range used to set and adjust the preset navigation variables. Task staging refers to dividing the entire navigation task into multiple relatively independent stages or parts according to a certain logic or condition, and each stage has clear goals and constraints, which is convenient for management and control. The segmented navigation task refers to the navigation task of each independent stage or part obtained after task staging, and these tasks together constitute the entire life cycle of the navigation task.

[0037] First, the preset navigation variables are set through the variable manager in the UAV flight control system, such as the Parameter module of PX4Autopilot. Then, the whole task cycle is analyzed using the threshold detection based on the sliding window and the state recognition of the hidden Markov model HMM: when the navigation variable reaches the preset scale, the stage segmentation point is marked, and N segmented navigation tasks are generated. For example, in a logistics distribution task, if the UAV detects that the heading deviates from the preset angle by more than 25° due to wind interference during flight, the stage switch is triggered, a "dynamic correction" node is added, and the path of the stage is replanned. Finally, the segmented tasks are bound to the geographic / logical nodes through the task topology module in ROS or the geo-fence function of the DJI Mobile SDK to form N task nodes.

[0038] By setting preset navigation variables, dividing the mission into stages, and marking mission nodes, the navigation missions of the drone swarm can be more finely controlled and managed. This helps to improve the efficiency and accuracy of mission execution and ensure that the drone swarm can successfully complete the mission according to the predetermined plan. At the same time, this also provides the basis and support for the collaborative work and dynamic path planning of the drone swarm.

[0039] Furthermore, a path planning strategy is determined, including: identifying a first task node, and determining a first key navigation point based on the task node; if the first key navigation point is an empty set, performing path planning according to a proximity principle to determine a first planned path.

[0040] Specifically, the first mission node refers to the first stage or target point that needs to be executed in the entire mission cycle, and is the starting point of path planning. The first key waypoint is a required position or logical operation point bound to the first mission node. If it is an empty set, it means that the node does not need to pass through a specific waypoint. The empty set means that the current mission node is not associated with any mandatory key waypoints, and the drone can freely choose a path. The proximity principle refers to a planning strategy that selects the shortest reachable path based on the Euclidean distance or cost function between the current drone position and the target node. The first planned path refers to the initial flight trajectory generated for the first mission node, which must meet constraints such as no collision and minimum energy consumption.

[0041] First, the first task node is determined by the priority sorting algorithm of the GPS / RTK-based positioning system or the task management platform. Then, the key waypoint matching engine (such as GIS geographic database query tool or graph-based path dependency analysis) is called to retrieve the waypoints associated with the node: if the matching result is an empty set, the nearest path planner (such as A* algorithm, RRT-Connect) is triggered to generate the first planned path based on the current position of the drone and the node target coordinates with the goal of minimizing the flight distance or time.

[0042] Significantly improve the flexibility and efficiency of path generation through empty set determination and nearest-neighbor planning strategy.

[0043] Further, after identifying the first task node, it includes: clustering the UAV swarm of the first task node according to the heading, determining M UAV clusters, where M is a positive integer less than or equal to the total number of the UAV swarm; traversing the M UAV clusters to perform cluster parallel planning and determining the first planned path.

[0044] Specifically, the M UAV clusters refer to the number of subgroups after division, and M ≤ the total number of UAVs. Cluster parallel planning means independently performing path planning for different clusters, making full use of distributed computing resources to improve the planning efficiency. The first planned path refers to the locally optimal flight path generated for each subgroup after cluster division, which needs to satisfy no conflicts within the cluster and collaborative constraints between clusters.

[0045] After identifying the first task node, first divide the UAV swarm through a heading clustering algorithm (such as heading angle grouping based on K-means or DBSCAN density clustering). Subsequently, call a distributed path planning framework (such as distributed nodes of ROS2 or a Kubernetes cluster scheduler) to allocate independent computing units for each cluster and perform path planning in parallel. During the planning process, coordinate the paths across clusters through a spatio-temporal synchronization protocol (such as an Apache Kafka message queue), and use a conflict detection tool (such as the Dynamic Window Approach DWA) to ensure no collisions between the paths of different clusters. Finally, integrate the paths of each cluster to form a global first planned path and send it to the UAV swarm for execution through the MAVLink communication protocol.

[0046] Achieve efficient collaboration of a large-scale UAV swarm through heading cluster division and parallel planning.

[0047] Further, if the first key waypoint is not an empty set, perform waypoint sequence identification on the first key waypoint; with the first key waypoint and the waypoint sequence as constraints, perform multi-point path planning based on the nearest-neighbor principle between points of the first key waypoint to determine the first initial planned path; perform smoothing processing on the first initial planned path to determine the first planned path.

[0048] Specifically, the flight path sequence identifier refers to assigning priority or order tags to key flight path points, specifying that the UAV must pass through these points in the designated order, which is usually used to avoid path intersections or meet mission timing requirements. The principle of proximity between points means that, on the premise of meeting the sequence constraints, a locally optimal path is selected based on the distance or cost function between flight path points, similar to the nearest neighbor optimization of the Traveling Salesman Problem (TSP). The first initial planned path refers to the preliminary flight trajectory generated according to the flight path point sequence and the principle of proximity, which may contain sharp turning angles or redundant fluctuations and needs to be further optimized. Smoothing processing refers to eliminating the mutation points in the path through curve fitting or trajectory optimization algorithms to improve flight stability and energy efficiency.

[0049] When the first set of key flight path points is non-empty, first, the flight path points are sequentially identified through a sequence planning engine (such as a path dependence analysis tool based on topological sorting or a dynamic priority algorithm), for example, using a genetic algorithm or an ant colony algorithm to solve the optimal access order. Then, with the sequence as the constraint, a multi-point path planner (such as a TSP variant algorithm, an LKH solver) is called to generate a locally optimal path between points, and the minimum total distance trajectory is calculated through the path optimization module of OR-Tools to form the first initial planned path. Subsequently, a trajectory smoothing algorithm is used to smooth the initial path.

[0050] Through the dual optimization of sequence constraints and path smoothing, the efficient and reliable execution of complex flight path point tasks is achieved.

[0051] Furthermore, determining the first planned path includes: traversing the M UAV clusters, performing path cluster collision analysis between clusters for the first task node to determine the first path constraint; traversing the M UAV clusters, performing single-vehicle collision analysis within the cluster for the first task node to determine the second path constraint; and performing path planning according to the first path constraint and the second path constraint.

[0052] Specifically, the first path constraint refers to the global restriction conditions that the paths between clusters must meet, usually set based on airspace control rules or mission priorities. Single-vehicle collision analysis refers to detecting the trajectory conflicts between individual UAVs within the same cluster to ensure flight safety within the cluster. The second path constraint refers to the local restrictions that the single-vehicle paths within the cluster need to meet, usually defined by the formation control protocol.

[0053] First, traverse the planned paths of all M clusters through an inter-cluster collision detection algorithm (such as conflict prediction based on spatio-temporal corridors or the velocity obstacle method) to analyze cross-cluster conflicts. For example, in a logistics task, if the paths of the "eastward cluster" and the "southward cluster" have a time overlap above an intersection, a first path constraint is generated. Then, use an intra-cluster conflict detection tool (such as KD-Tree-based nearest neighbor search or the dynamic window method) to analyze the trajectories of the drones within each cluster frame by frame. If the detected distance between individual drones is less than a safety threshold (such as 5 meters), a second path constraint (such as vertical stratification or speed adjustment) is imposed. For example, in a disaster rescue scenario, the drones in the same search and rescue cluster repel each other through the APF algorithm to maintain a minimum distance of 10 meters. Finally, integrate the two types of constraints and call a multi-objective optimization solver (such as the NSGA-II genetic algorithm or the CBS conflict constraint search) to generate a first planned path that meets global and local safety requirements, and send the constraint conditions and the path to the drone swarm through the MAVLink protocol.

[0054] Through the collaborative planning of dual collision constraints, the efficient and safe operation of the drone swarm is achieved.

[0055] Furthermore, after determining the first planned path, it includes: guiding the drone swarm to perform navigation management of the first task node according to the first planned path and transmitting back the drone distribution network; planning and determining the second planned path of the second task node; determining the starting distribution network of the drone swarm based on the second planned path, and performing coding mapping verification with the drone distribution network according to the drone coding to determine the node offset information; and performing offset compensation on the drone swarm according to the node offset information.

[0056] Specifically, navigation management refers to the process of controlling the drone swarm to fly according to the planned path, covering real-time operations such as navigation, formation maintenance, and obstacle avoidance. The drone distribution network refers to the dynamic topological network formed by the actual positions and states of the drone swarm after executing the first task node. The second task node refers to the next-stage target in the entire task cycle that is immediately after the first node. The second planned path refers to the flight trajectory generated based on the requirements of the second node, which needs to be seamlessly connected to the actual state of the drones. The starting distribution network refers to the expected starting positions and states of the second task node, which are calculated backward from the second planned path. The node offset information refers to the difference between the actual and expected positions / states.

[0057] During the execution of the first planned path, the state of the UAV is transmitted in real time through on-board sensors and cluster communication protocols (MAVLink, ZigBee) to construct a UAV distribution network. Subsequently, the task sequence planner (such as the TaskPlanner module of ROS) is called to generate the second planned path of the second task node, and the expected starting distribution network is calculated through an inverse kinematics deduction tool (such as the MATLAB Robotics Toolbox). Then, the encoding mapping engine (such as a hash table or the graph database Neo4j) is used to compare the actual and expected distributions aircraft by aircraft. Based on this, a bias compensation algorithm (such as model predictive control MPC or PID controller) is used to dynamically adjust the path: if it deviates due to wind disturbance, a detour trajectory is inserted into its second path through the dynamic window method, or the time delay is compensated by accelerating through the velocity feedforward control of the PX4 flight controller. Finally, the compensation strategy is synchronized to the entire cluster through a distributed consensus algorithm (such as the Raft protocol).

[0058] Through the dynamic bias compensation mechanism, the task robustness and continuity are significantly improved.

[0059] Furthermore, the bias compensation for the UAV swarm includes: determining the bias compensation method, which includes a first method based on the adjustment of the second planned path and a second method for correcting the deviation based on the UAV distribution network; and performing bias compensation on the UAV swarm according to the first method or the second method in the bias compensation method.

[0060] Specifically, the bias compensation method refers to the specific strategy for eliminating the deviation between the actual state of the UAV and the expected target, including two modes: the first method: indirectly compensate the current bias by modifying the trajectory, speed or time window of the subsequent planned path. The second method: directly adjust the actual state of the UAV swarm, and make it align with the expected distribution network through local path replanning or control instruction correction.

[0061] First, the type and magnitude of the deviation are quantified through state estimation based on Kalman filtering or covariance matrix analysis, and the first or second compensation method is decided: if the bias is mainly caused by environmental interference (such as wind) and the subsequent path can accommodate it, the first method is selected, and the path replanning tool (such as RRT*, A* variant algorithm) is called to adjust the second planned path. For example, in a logistics task, a shortcut path is inserted for the delayed UAV, and the update is sent through the global planner module of ROS; if the bias is caused by a hardware failure or a sudden obstacle and needs to be corrected immediately (such as the position deviating from the safety threshold), the second method is adopted, and direct correction is performed through a distributed control protocol (such as MAVLink instruction or PID controller). For example, in a disaster rescue, if the UAV deviates from the expected coordinates by 50 meters due to obstacle avoidance, a real-time waypoint instruction is sent through the position control mode of the PX4 flight controller to make it return to the target flight path.

[0062] Through a dual-mode compensation mechanism, efficient suppression of bias and improvement of task robustness are achieved.

[0063] In summary, the method for dynamic path planning of an unmanned aerial vehicle (UAV) swarm based on collaborative intelligence provided by the embodiments of the present application has the following technical effects:

[0064] 1. By decomposing the task into discrete nodes, binding key waypoints to specific UAVs, and dividing the swarm according to headings, the system realizes dynamic collaboration and efficient path planning. The modular design supports real-time adaptation to environmental changes, such as sudden obstacles or task adjustments. At the same time, the heading consistency significantly reduces path conflicts between swarms. The global bias calibration mechanism effectively suppresses error accumulation, ensures task continuity and reasonable resource allocation, and is applicable to collaborative operations in complex scenarios of large-scale UAV swarms.

[0065] 2. For task nodes without mandatory waypoints, the nearby planning strategy greatly simplifies the path generation process. The shortest path algorithm is directly used to quickly respond to task requirements, reducing computational complexity, especially applicable to emergency scenarios or rapid deployment of large-scale swarms. The free path mode enhances the flexibility of the system, seamlessly switches paths when sudden obstacles appear, and ensures the success rate and timeliness of tasks.

[0066] 3. The calibration mechanism based on the upper node distribution network realizes seamless connection between task stages. By dynamically correcting node deviations, it suppresses the cross-stage transmission of errors and ensures the overall coherence of the task. The elastic resource allocation strategy automatically adjusts task division of labor when some UAVs fail, ensures that key targets are completed first, and significantly improves the fault tolerance of the system.

[0067] Embodiment 2

[0068] Based on the same inventive concept as the method for dynamic path planning of an unmanned aerial vehicle (UAV) swarm based on collaborative intelligence in the foregoing embodiment, as Figure 2 shown, the embodiments of the present application provide a system for dynamic path planning of an unmanned aerial vehicle (UAV) swarm based on collaborative intelligence, and the system includes:

[0069] A navigation task acquisition module 11, configured to acquire the navigation task of the UAV swarm, and identify N task nodes for the entire task cycle of the navigation task; a key waypoint positioning module 12, configured to position key waypoints for the navigation task, where the key waypoints are marked with UAV numbers; a path planning strategy determination module 13, configured to connect to a swarm path planning module, traverse the N task nodes and the key waypoints, divide the UAV swarm into swarms according to headings, perform task path planning with the node single-machine target as the guide, and determine a path planning strategy; and a bias calibration module 14, configured to perform bias calibration in a manner of sequential planning based on the N task nodes and with the UAV distribution network based on the upper task nodes.

[0070] Further, the navigation task acquisition module 11 is further configured to perform the following steps: set preset navigation variables, where the heading variable is determined based on the heading and the navigation state, and the preset navigation variables are set through a preset variable scale; according to the entire task cycle, use the preset navigation variables to divide the task into stages, and determine N segmented navigation tasks; based on the N segmented navigation tasks, mark the entire task cycle to determine the N task nodes.

[0071] Further, the path planning strategy determination module 13 is further configured to perform the following steps: identify the first task node and determine the first key navigation point based on the task node; if the first key navigation point is an empty set, perform path planning according to the principle of proximity to determine the first planned path.

[0072] Further, the path planning strategy determination module 13 is further configured to perform the following steps: divide the UAV group of the first task node according to the heading to determine M UAV clusters, where M is a positive integer less than or equal to the total number of the UAV group; traverse the M UAV clusters to perform cluster parallel planning to determine the first planned path.

[0073] Further, the path planning strategy determination module 13 is further configured to perform the following steps: if the first key navigation point is not an empty set, perform navigation order identification on the first key navigation point; use the first key navigation point and the navigation order as constraints, and perform multi-point path planning based on the principle of proximity between points of the first key navigation point to determine the first initial planned path; perform smoothing processing on the first initial planned path to determine the first planned path.

[0074] Further, the path planning strategy determination module 13 is further configured to perform the following steps: traverse the M UAV clusters, perform path cluster collision analysis between clusters for the first task node to determine the first path constraint; traverse the M UAV clusters, perform single UAV collision analysis within the cluster for the first task node to determine the second path constraint; perform path planning according to the first path constraint and the second path constraint.

[0075] Further, the path planning strategy determination module 13 is further configured to perform the following steps: according to the first planned path, guide the UAV group to perform navigation management of the first task node and send back the UAV distribution network; plan and determine the second planned path of the second task node; determine the starting distribution network of the UAV group based on the second planned path, and perform encoding mapping verification with the UAV distribution network according to the UAV encoding to determine the node offset information; perform offset compensation on the UAV group according to the node offset information.

[0076] Further, the bias calibration module 14 is further configured to perform the following steps: determining a bias compensation method, where the bias compensation method includes a first method based on the adjustment of the second planned path and a second method based on the correction of the UAV distribution network; and performing bias compensation on the UAV swarm according to the first method or the second method in the bias compensation method.

[0077] Any step of the above-described method can be stored as computer instructions or programs in an unrestricted computer memory and can be called and recognized by an unrestricted computer processor to implement any one of the methods in the embodiments of the present application, and no redundant limitations are made here.

[0078] Further, the above-mentioned first or second may not only represent an order relationship, but may also represent a specific concept, and / or refer to the fact that multiple elements can be selected individually or in whole. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and variations.

Claims

1. A dynamic path planning method for drone swarms based on collaborative intelligence, characterized in that: include: Obtain the navigation mission of the drone group, and identify N mission nodes for the full mission cycle of the navigation mission; For the navigation mission, locate key waypoints, wherein the key waypoints are marked with drone numbers; Connecting the cluster path planning module, traversing the N task nodes and the key navigation points, clustering the drone group according to the heading, performing task path planning with the node single machine target as the guide, and determining the path planning strategy; Among them, bias calibration is performed on the drone distribution network based on the upper task nodes in a manner based on the sequential planning of the N task nodes.

2. The method for dynamic path planning of a drone swarm based on collaborative intelligence according to claim 1, characterized in that: For the entire mission cycle of the navigation mission, N mission nodes are identified, including: Setting preset navigation variables, wherein the heading variable is determined by the heading and the navigation state, and the preset navigation variable is set by the preset variable scale; According to the full mission cycle, the mission is divided into stages using the preset navigation variables to determine N segmented navigation tasks; According to the N segmented navigation tasks, the entire task cycle is marked to determine the N task nodes.

3. The method for dynamic path planning of a drone swarm based on collaborative intelligence according to claim 1, characterized in that: Determine the path planning strategy, including: Identify a first mission node and determine a first key waypoint based on the mission node; If the first key navigation point is an empty set, path planning is performed according to the proximity principle to determine the first planned path.

4. The method for dynamic path planning of a drone swarm based on collaborative intelligence as claimed in claim 3, characterized in that: After identifying the first task node, it includes: According to the heading, the drone group of the first task node is clustered to determine M drone clusters, where M is a positive integer less than or equal to the total number of drone groups; The M drone clusters are traversed, cluster parallel planning is performed, and a first planning path is determined.

5. The method for dynamic path planning of a drone swarm based on collaborative intelligence as claimed in claim 3, characterized in that: If the first key waypoints are not an empty set, marking the first key waypoints in a navigation order; Taking the first key navigation point and the navigation sequence as constraints, and based on the principle of proximity between points of the first key navigation point, multi-point path planning is performed to determine a first initial planned path; The first initial planned path is smoothed to determine a first planned path.

6. The method for dynamic path planning of a drone swarm based on collaborative intelligence according to claim 4, characterized in that: Determine the first planning path, including: Traversing the M drone clusters, performing path cluster collision analysis between clusters for the first task node, and determining a first path constraint; Traversing the M drone clusters, performing a single-machine collision analysis within the cluster for the first task node, and determining a second path constraint; Path planning is performed according to the first path constraint and the second path constraint.

7. The method for dynamic path planning of a drone swarm based on collaborative intelligence according to claim 1, characterized in that: After the first planning path is determined, it includes: According to the first planned path, guide the drone group to perform navigation management of the first task node and transmit back to the drone distribution network; Planning and determining a second planned path for a second task node; Determine a starting distribution network of the drone group based on the second planned path, perform coding mapping verification with the drone distribution network according to the drone coding, and determine node offset information; The drone swarm is bias compensated according to the node bias information.

8. The method for dynamic path planning of a drone swarm based on collaborative intelligence according to claim 7, characterized in that: Performing bias compensation on the drone group includes: Determine an offset compensation method, including a first method based on the second planned path adjustment and a second method based on the UAV distribution network correction; According to the first or second bias compensation method, bias compensation is performed on the drone group.

9. A dynamic path planning system for drone swarms based on collaborative intelligence, characterized in that: A system for executing the method for dynamic path planning of a drone swarm based on collaborative intelligence according to any one of claims 1 to 8, the system comprising: A navigation task acquisition module is used to acquire the navigation task of the drone group and identify N task nodes for the full task cycle of the navigation task; A key waypoint positioning module, used for positioning key waypoints for the navigation mission, wherein the key waypoints are marked with drone numbers; A path planning strategy determination module is used to connect to the cluster path planning module, traverse the N task nodes and the key navigation points, cluster the drone group according to the heading, perform task path planning with the node single machine target as the guide, and determine the path planning strategy; The bias calibration module is used to perform bias calibration on a UAV distribution network based on upper-level task nodes in a manner based on sequential planning of the N task nodes.

Images