Path planning method for cooperative operation of multiple unmanned aerial vehicles

Abstract

The invention discloses a path planning method for cooperative operation of multiple unmanned aerial vehicles, and belongs to the technical field of unmanned aerial vehicle control. The method comprises the following steps: acquiring real-time state information of each unmanned aerial vehicle; dynamic tasks are allocated based on a lightweight auction protocol, and distance, energy consumption, task urgency and load are comprehensively considered in a bidding function; generating a global reference path for each unmanned aerial vehicle, and applying for a space-time reserved window of the three-dimensional space logic grid unit in a specific time interval from the coordination agent; the coordination agent performs conflict detection and feeds back a reservation result; each unmanned aerial vehicle carries out local re-planning in combination with an improved artificial potential field-fast extended random tree algorithm so as to avoid dynamic obstacles; and when the task is changed or the unmanned aerial vehicle fails, a task release and redistribution mechanism is triggered. According to the method, safe, efficient and extensible cooperative path planning of multiple unmanned aerial vehicles in a high-dynamic environment is realized, and the method is suitable for scenes such as logistics distribution, electric power inspection, emergency search and rescue and the like.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control, and more specifically to a path planning method for collaborative operations of multiple UAVs. Background Technology

[0002] With the rapid development of drone technology, single-drone operations are no longer sufficient to meet the demands of large-scale, high-efficiency missions. Multi-drone collaborative operations have become a research hotspot, widely applied in urban logistics, infrastructure inspection, agricultural plant protection, and disaster emergency response. In these applications, planning conflict-free, low-energy-consumption, and highly efficient flight paths for multiple drones is a core issue in ensuring system safety and efficiency.

[0003] In existing technologies, multi-UAV path planning is mainly divided into two categories: centralized and distributed. While centralized methods (such as mixed-integer linear programming and spatiotemporal A*) can obtain globally optimal solutions, their computational complexity increases exponentially with the number of UAVs, making them unsuitable for real-time demands in dynamic environments. Distributed methods (such as velocity obstacle-based methods, distributed RRT*, and consensus negotiation) offer some scalability, but are prone to deadlocks, oscillations, or implicit conflicts in high-density scenarios. Furthermore, they typically decouple task allocation from path planning, resulting in suboptimal overall performance. In addition, most current solutions do not adequately consider the following practical constraints: In scenarios with limited communication, where there is no central base station or 5G coverage is at the edge, the communication bandwidth between UAVs is limited, making it difficult to support high-frequency global synchronization; In scenarios with dynamic disturbances, random events such as sudden obstacles, task insertion / cancellation, and individual UAV malfunctions require the system to have rapid reconfiguration capabilities; In scenarios with secure isolation areas, the lack of explicit resource management mechanisms for the three-dimensional spatial-temporal dimensions, relying solely on post-event obstacle avoidance, presents potential collision risks.

[0004] While current research has attempted to introduce concepts such as spatio-temporal corridors or collision cones, these methods suffer from high computational overhead, difficulty in supporting large-scale clusters, and lack of closed-loop optimization with task allocation. Therefore, a novel multi-UAV cooperative path planning method is urgently needed, balancing security, real-time performance, robustness, and scalability. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing multi-UAV path planning methods in terms of dynamism, security, scalability, and task-path coordination, and to provide an efficient, robust path planning method suitable for large-scale clusters.

[0006] To achieve the above objectives, this invention proposes a path planning method for multi-UAV cooperative operations, which includes the following steps:

[0007] S1. Receive the tasks to be executed and obtain the real-time status information of each UAV, including location, remaining battery power, payload capacity and communication status.

[0008] S2. Based on state information, a lightweight auction protocol is used to dynamically allocate tasks to be executed, generating a task subset corresponding to each drone. Each drone bids for executable tasks according to a preset bidding function, which comprehensively considers task distance cost, expected energy consumption, task urgency and its own resource load.

[0009] S3. For each UAV's assigned task subset, construct a global reference path and request a spatiotemporal reservation window from the coordination agent. The spatiotemporal reservation window is the exclusive right to use a three-dimensional spatial logical grid cell within a preset time interval.

[0010] S4. The coordinating agent determines whether the requested spatiotemporal reservation window conflicts with other drones based on the conflict detection table. If there is no conflict, the reservation is approved. If there is a conflict, the agent returns to the conflict area and suggests a detour strategy.

[0011] S5. Each UAV generates a globally feasible path based on the approved spatiotemporal reserved window, and performs local replanning during execution by combining the improved artificial potential field-fast expanding random tree algorithm to avoid dynamic obstacles and other UAVs.

[0012] S6. When a drone failure or mission change is detected, the mission release and reallocation process is triggered, and the process returns to step S2.

[0013] Furthermore, the size of the three-dimensional spatial logical grid cell is dynamically adjusted based on the UAV's minimum safe distance and positioning accuracy, while the length of the time interval is determined based on the UAV's maximum cruise speed and control cycle. By dynamically adjusting the spatial grid size according to the UAV's minimum safe distance and positioning accuracy, and determining the time interval length based on the maximum cruise speed and control cycle, this scheme achieves refined and adaptive partitioning of spatiotemporal resources while ensuring flight safety. This avoids resource waste caused by excessive conservatism and improves the spatial utilization rate and time synchronization accuracy of path planning.

[0014] Furthermore, the coordination agent can be an edge computing node or a designated lead drone within the cluster. When communication is interrupted, each drone executes a conservative path continuation strategy based on locally cached spatiotemporal reservation window information, allowing entry only into reserved and unexpired grid cells. The coordination agent can be flexibly deployed on edge computing nodes or by the lead drone, enhancing the scalability and deployment adaptability of the system architecture. During communication interruptions, each drone executes a conservative continuation strategy based on locally cached reservation information, entering only confirmed safe spatiotemporal cells, effectively improving the system's fault tolerance and operational continuity in weak network or disconnected environments.

[0015] Furthermore, the improved artificial potential field-fast expanding random tree algorithm includes:

[0016] This algorithm models static obstacles as repulsive potential fields and other moving drones as dynamic repulsive sources with predicted trajectories, incorporating a social force model to simulate their avoidance intentions. During the sampling process of the artificial potential field-fast expanding random tree algorithm, low-potential regions are prioritized for exploration, and incremental pruning optimization is applied to the generated paths to meet real-time requirements. The algorithm models static obstacles and dynamic drones as repulsive potential fields and dynamic repulsive sources with predicted trajectories, respectively, and integrates a social force model to more realistically reflect the avoidance intentions among multiple agents. Simultaneously, prioritizing the exploration of low-potential regions and introducing incremental pruning optimization during sampling significantly improves the obstacle avoidance rationality, path smoothness, and real-time response capability of local replanning.

[0017] Furthermore, in the lightweight auction agreement, each drone represents a certain value for mission t. Calculate using the following formula:

[0018]

[0019] in, Let be the distance from the drone to mission t. Let t be the deadline for task t to execute. For the current time, The estimated energy consumption for performing task t. For the estimated execution time, For the current task load, , and This is a normalized reference value. , , and These are the weighting coefficients, and The bidding function comprehensively quantifies distance cost, energy efficiency, task urgency, and current load status, and achieves multi-objective balance through normalization and weight adjustment, making task allocation results more fair, efficient, and in line with actual operational constraints, effectively supporting collaborative decision-making among multiple UAVs under resource-constrained conditions.

[0020] Furthermore, the conflict detection table is stored in the form of a hash map, which can be identified as "grid cell ID + time slot number", and the value is a list of UAV IDs that have reserved the spatiotemporal cell. After receiving the reservation request, the coordination agent completes the conflict determination in O(1) time. Using the hash map structure with "grid cell ID + time slot number" as the key to store conflict information enables the coordination agent to complete the conflict determination in constant time O(1), which greatly reduces the coordination overhead and improves the processing efficiency and system scalability of spatiotemporal reservation approval under large-scale clusters.

[0021] Furthermore, the method described above is deployed in an edge computing architecture that supports 5G low-latency communication, ensuring that the transmission latency of control commands between the coordination agent and each drone does not exceed 50 milliseconds. Deploying the entire method in an edge computing architecture that supports 5G low-latency communication, and ensuring that the transmission latency of control commands between the coordination agent and drones does not exceed 50 milliseconds, provides reliable communication assurance for highly dynamic collaborative operations, significantly enhancing the system's real-time performance, stability, and applicability to high-speed scenarios.

[0022] To achieve the above objectives, the present invention also provides a storage medium storing a plurality of computer instructions, which, when executed, are used to implement the above-described path planning method for multi-UAV collaborative operations.

[0023] This invention discloses a path planning method for multi-UAV collaborative operations. By introducing a Spatio-Temporal Reservation Window (STRW) mechanism and a lightweight dynamic task auction protocol, it achieves joint optimization of tasks and paths, and combines an improved local replanning algorithm to ensure flight safety. Specifically, this invention first uses a lightweight auction protocol to distribute bidding and allocation of dynamically arriving tasks based on the real-time status of each UAV (including location, power, payload, communication capabilities, etc.). The bidding function comprehensively considers distance cost, energy consumption, task deadline, and current load to ensure balanced resource utilization. Subsequently, each UAV generates a preliminary global path for its assigned task and requests exclusive access to the three-dimensional spatial logical grid cells to be traversed within a certain number of time steps from the coordination agent. The coordination agent maintains a conflict detection table through a hash mapping structure, determines whether there is a spatiotemporal conflict in O(1) time, and provides feedback on approval or suggested detour strategies. After approval, the UAV generates the final globally feasible path and integrates the Artificial Potential Field (APF) method and the Rapid Expanding Random Tree (RRT*) algorithm for local replanning during execution to cope with dynamic obstacles and other unexpected behaviors of UAVs. When the system experiences task changes, additions, or UAV failures, the task release and reallocation process is automatically triggered to achieve system self-healing and reconstruction.

[0024] In summary, this invention achieves efficient and safe collaborative path planning for multi-UAV systems in dynamic and complex environments by integrating a lightweight auction mechanism, spatiotemporal reservation window coordination, and improved Artificial Potential Field-Rapidly Expanding Random Tree (APF-RRT) local replanning. Its beneficial effects include: task allocation considers distance, energy consumption, urgency, and load balancing, improving overall operational efficiency; the reservation and conflict detection mechanism of a three-dimensional spatiotemporal logical grid effectively avoids collisions between UAVs and ensures flight safety; local path planning possesses rapid response capabilities to dynamic obstacles and unexpected situations (such as UAV failure or mission changes), enhancing the system's robustness and adaptability; and the closed-loop design throughout the entire process supports dynamic release and reallocation of tasks, significantly improving the autonomy, flexibility, and reliability of multi-UAV collaborative operations.

[0025] It should be noted that the terms "first," "second," and similar terms used in this article are merely for describing the constituent elements of the technical solution and do not constitute a limitation on the technical solution, nor should they be interpreted as an indication or implication of the importance of the corresponding elements; elements with terms such as "first," "second," or similar terms indicate that at least one of the elements is included in the corresponding technical solution. Attached Figure Description

[0026] To more clearly illustrate the technical solution of the present invention and facilitate a further understanding of its technical effects, features, and objectives, the present invention will be described in detail below with reference to the accompanying drawings. The drawings constitute an essential part of the specification and are used together with Embodiment 1 of the present invention to illustrate the technical solution of the present invention, but do not constitute a limitation on the present invention.

[0027] Figure 1 This is a schematic flowchart of the method of the present invention;

[0028] Figure 2 This is a flowchart illustrating the process of local replanning using the artificial potential field-fast expanding random tree algorithm in the method of this invention. Detailed Implementation

[0029] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Of course, the specific embodiments described below are merely illustrative of the technical solutions of the present invention, and not intended to limit the present invention. Furthermore, the parts described in the embodiments or drawings are merely illustrative examples of relevant parts of the present invention, and not the entirety of the present invention. At the same time, all other embodiments obtained by those skilled in the art based on the embodiments of the present invention should naturally fall within the protection scope of the present invention.

[0030] like Figure 1 As shown in the flowchart, the path planning method for multi-UAV cooperative operations proposed in this invention includes the following steps:

[0031] S1. Receive the tasks to be executed and obtain the real-time status information of each drone, including location, remaining battery power, payload capacity, and communication status. Specifically, after the system starts, each drone periodically reports its status to the coordination agent (in this embodiment, a coordination service deployed on a 5G edge server) through its onboard sensors and communication module. Status information includes: GPS location, remaining battery power, maximum payload, and communication signal strength. Simultaneously, the task scheduling center issues multiple delivery tasks, each containing attributes such as target coordinates, cargo weight, and latest delivery time.

[0032] S2. Based on state information, a lightweight auction protocol is used to dynamically allocate tasks to be executed, generating a task subset for each drone. Each drone bids for executable tasks according to a preset bidding function, which comprehensively considers task distance cost, estimated energy consumption, task urgency, and its own resource load. Specifically, the coordination agent broadcasts all tasks to be allocated. Each drone runs the lightweight auction protocol locally: for each task t, the value is calculated. Each drone terminal submits its bid list to the coordinating agent; the coordinating agent allocates tasks according to the highest bid principle and notifies the winning drone.

[0033] S3. For each drone's assigned task subset, construct a global reference path and request a spatiotemporal reservation window from the coordination agent. The spatiotemporal reservation window represents the exclusive right to use a 3D spatial logical grid cell within a preset time interval. Specifically, the winning drone generates a preliminary path on a 3D raster map using the A* algorithm based on the target point of its currently accepted task, obtaining the sequence of grid cells to be traversed. Assuming a control cycle of 100ms and a maximum speed of 5m / s, each grid (10m side length) is expected to occupy 2 time slots (200ms). The winning drone requests the reservation rights of these grids from the coordination agent within the next 0-2s.

[0034] S4. The coordination agent uses a conflict detection table to determine whether the requested spatiotemporal reservation window conflicts with other drones. If there is no conflict, the reservation is approved; if there is a conflict, the agent returns the conflict area and suggests a detour strategy. Specifically, the coordination agent maintains a hash table, which can be identified as follows: The value is a set of drone IDs. Upon receiving the request for the winning drone, the system queries each item: if a certain... If the drone is already in use by another drone, mark the conflict; return to the conflict location and issue a suggestion, such as "delay 200ms to enter". "or "change of course" ".

[0035] S5. Each UAV generates a globally feasible path based on the approved spatiotemporal reserved window, and during execution, performs local replanning using an improved artificial potential field-fast expanding random tree algorithm to avoid dynamic obstacles and other UAVs. The winning UAVs fly along the global path. If an unconnected third-party UAV is detected entering the grid ahead, the improved APF-RRT* algorithm is immediately activated:

[0036] The dynamic obstacle is modeled as a moving repulsion source, and its trajectory is predicted for the next 2 seconds.

[0037] It biases towards low potential regions during RRT* sampling;

[0038] Generate a new local path that detours but still remains within the reserved spacetime window;

[0039] Continue execution, ensuring that the constraints of the spacetime reserved window are not violated.

[0040] In this embodiment, taking the winning drone as U and the current task as T as an example, the method of using the artificial potential field-fast expanding random tree algorithm for local path replanning in step S5 is explained. The winning drone U has obtained the globally feasible path of task T and holds the corresponding spatiotemporal reserved window set. Under the premise that the winning drone U flies along the reference trajectory at a preset control frequency, it is assumed that at the 3.2-second mark of flight, its onboard perception system (including millimeter-wave radar, visual SLAM, or ADS-B receiver) detects a third-party drone that is not connected to the network (denoted as U). Suddenly, it burst into the grid cell it was about to enter. And it is expected to be in the time slot It occupies this region within the corresponding time interval [3.2s, 3.4s).

[0041] At this moment, the winning drone U immediately triggers the local replanning module, such as Figure 2 As shown, perform the following sub-steps:

[0042] S5.1 Perform dynamic obstacle modeling and trajectory prediction:

[0043] Enter the current position of the obstacle. ,speed (Estimated from multiple frame point clouds or images), the obstacle type is determined (in this embodiment, it is determined to be a small quadcopter), and its maximum acceleration is estimated as... ;

[0044] Using a uniform acceleration motion model to predict the future of third-party drones The reachable set within:

[0045]

[0046]

[0047]

[0048] in, This refers to the set of approved reserved windows currently held by the winning drone U. Set to 3 meters as the safety radius;

[0049] Finally, a set of discretized predictions occupying spatiotemporal units is output:

[0050]

[0051] in, It is a collection of three-dimensional spatial logical grid units. It is a discrete time slot sequence. The time slot length (e.g., 0.2s).

[0052] S5.2 Constructing an improved artificial potential field:

[0053] Construct a gravitational field: pointing to the next waypoint on the original global path. Its potential energy is:

[0054]

[0055] in, This represents the gravitational coefficient in an artificial potential field;

[0056] Construct a repulsive field for each predicted occupied unit. If the current time , A time-sensitive window, corresponding to a spatial region. Superimposed Gaussian repulsive potential:

[0057]

[0058] in, For grid center, The distance at which the repulsive field acts is [distance]. The side length of the grid cell;

[0059] The total potential energy can be expressed as:

[0060]

[0061] Among them, weight The value increases as time approaches, reflecting the urgency of the situation and the increasing danger as the time draws closer.

[0062] S5.3, Perform improved PRT local replanning:

[0063] The planning scope is centered on the current location, with a radius of... The spherical region is the range;

[0064] Set a time window that only considers the future. Within a given time period, there are 5 time slots;

[0065] The sampling strategy is determined such that 80% of the random samples are biased towards low-potential regions (by rejecting samples, if...). (Then resampling) This represents a potential energy threshold for the artificial potential field, with 20% uniform sampling to ensure exploratory nature.

[0066] Perform node expansion to generate new nodes. Check its grid and their corresponding time slots Then, the spatiotemporal reserved window constraint is checked. If... If so, then discard the node;

[0067] Path optimization is performed, using RRT*'s reconnection mechanism to optimize path length, resulting in a final path. Must meet:

[0068]

[0069]

[0070] in, Indicates the physical location Functions that map to their respective grid cells.

[0071] S5.4 Perform path execution and constraint verification:

[0072] The generated local path Input to the underlying controller;

[0073] The controller tracks the trajectory at a frequency of 100Hz;

[0074] Simultaneously perform continuous verification: (1) Current location Does it belong to grid g? (2) Current time slot (3) Inspection Is it valid?

[0075] If the above verification rules are violated, an emergency brake will be immediately triggered and a request will be made to coordinate the agent to extend the reservation or hover and wait.

[0076] S6. When a drone failure or mission change is detected, the mission release and reallocation process is triggered, returning to step S2. Specifically, if the winning drone experiences an unexpected situation and cannot continue or complete the mission in time, the mission executed by that drone is automatically released. The coordinating agent rebroadcasts the mission, and other drones rebid based on the current status and take over the mission.

[0077] To achieve the above objectives, the present invention also provides a storage medium storing a plurality of computer instructions, which, when executed, are used to implement the above-described path planning method for multi-UAV collaborative operations.

[0078] It should be noted that the above embodiments are only for more clearly illustrating the technical solution of the present invention. Those skilled in the art will understand that the implementation of the present invention is not limited to the above content. Any obvious changes, substitutions or replacements made based on the above content do not exceed the scope of the technical solution of the present invention. Other implementations will also fall within the scope of the present invention without departing from the concept of the present invention.

Claims

1. A path planning method for multi-UAV cooperative operations, characterized in that, include: S1. Receive the task to be executed and obtain the real-time status information of each UAV, including location, remaining battery power, payload capacity and communication status. S2. Based on the status information, a lightweight auction protocol is used to dynamically allocate the tasks to be executed, generating a task subset corresponding to each drone. Each drone bids for the executable tasks according to a preset bidding function, which comprehensively considers the task distance cost, expected energy consumption, task urgency and its own resource load. S3. For each UAV's assigned task subset, construct a global reference path and request a spatiotemporal reservation window from the coordination agent. The spatiotemporal reservation window is the exclusive right to use a three-dimensional spatial logical grid unit within a preset time interval. S4. The coordination agent determines whether the requested spatiotemporal reservation window conflicts with other UAVs based on the conflict detection table. If there is no conflict, the reservation is approved. If there is a conflict, the agent returns to the conflict area and suggests a detour strategy. S5. Each UAV generates a globally feasible path based on the approved spatiotemporal reserved window, and performs local replanning during execution by combining the improved artificial potential field-fast expanding random tree algorithm to avoid dynamic obstacles and other UAVs. S6. When a drone failure or mission change is detected, the mission release and reallocation process is triggered, and the process returns to step S2.

2. The path planning method for multi-UAV cooperative operations as described in claim 1, characterized in that: The size of the three-dimensional spatial logical grid unit is dynamically adjusted according to the minimum safe distance and positioning accuracy of the UAV, and the length of the time interval is determined according to the maximum cruise speed and control cycle of the UAV.

3. The path planning method for multi-UAV cooperative operations as described in claim 1, characterized in that: The coordination agent is an edge computing node or a designated lead drone in the cluster; when communication is interrupted, each drone executes a conservative path continuation strategy based on the spatiotemporal reservation window information cached locally, allowing entry only into reserved and unexpired grid cells.

4. The path planning method for multi-UAV cooperative operations as described in claim 1, characterized in that, The improved artificial potential field-fast expanding random tree algorithm includes: Static obstacles are modeled as repulsive potential fields; Other moving drones are modeled as dynamic repulsion sources with predicted trajectories, and a social force model is introduced to simulate their avoidance intentions. In the sampling process of the artificial potential field-fast expanding random tree algorithm, low potential energy regions are explored first, and incremental pruning optimization is performed on the generated path to meet real-time requirements.

5. The path planning method for multi-UAV cooperative operations as described in claim 1, characterized in that, In the lightweight auction protocol, each drone's bid value for mission t is... Calculate using the following formula: in, Let be the distance from the drone to mission t. Let t be the deadline for task t to execute. For the current time, The estimated energy consumption for performing task t. For the estimated execution time, For the current task load, , and This is a normalized reference value. , , and These are the weighting coefficients, and .

6. The path planning method for multi-UAV cooperative operations as described in claim 1, characterized in that: The conflict detection table is stored in the form of a hash map and can be identified as "grid cell ID + time slot number". The value is a list of UAV IDs that have reserved the spatiotemporal cell. After receiving the reservation request, the coordination agent completes the conflict determination in O(1) time.

7. The path planning method for multi-UAV cooperative operation as described in any one of claims 1-6, characterized in that: The method is deployed in an edge computing architecture that supports 5G low-latency communication, and the control command transmission delay between the coordination agent and each of the drones does not exceed 50 milliseconds.

8. A storage medium, characterized in that: The system stores a plurality of computer instructions, which, when executed, are used to implement the path planning method for multi-UAV collaborative operation as described in any one of claims 1-7.

Images