A control algorithm for cooperative inspection of a drone cluster

Abstract

The present application relates to the technical field of unmanned aerial vehicle control, in particular to a control algorithm for cooperative inspection of unmanned aerial vehicle cluster, the present application is designed by innovative non-central architecture and double-neighbor system, when the node communication loss or failure of unmanned aerial vehicle cluster is up to 30%, it can still maintain more than 92% of task completion rate, fault tolerance is significantly better than prior art; its core only relies on three simple behavior rules, the requirement of single machine computing power and communication bandwidth is very low, it can adapt to low-cost civilian unmanned aerial vehicle platform, so that the cost of large-scale deployment is reduced by about 40%. Combined with dynamic weight adjustment mechanism and virtual leader guidance, the system can not only autonomously avoid dynamic obstacles in complex urban areas, but also accurately follow the preset inspection route, realizing the unity of flexibility and accuracy; and supports multi-hop Ad-hoc communication, maintains stable connection in the area without ground network; fault automatic switching, single machine exit does not affect the overall task; combined with fixed airport, it can realize 7*24 hours normal unmanned inspection.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control technology, specifically a control algorithm for collaborative inspection of UAV swarms. Background Technology

[0002] Drones have been widely used in many fields. As an efficient and flexible inspection method, drone swarm inspection has attracted more and more attention from front-line workers in recent years. By using a swarm of multiple drones for inspection, not only can a larger inspection area be covered, but the accuracy and efficiency of the inspection can also be improved, labor costs can be reduced, and inspection risks can be lowered.

[0003] Existing drone swarm inspection technologies mainly employ centralized control or pre-defined global path planning methods to inspect the inspection area.

[0004] However, when inspection personnel use the pre-set global path planning method for collaborative inspection of UAV swarms, the reliance on the central node for unified path planning means that the entire swarm is prone to paralysis if the central node fails or communication is interrupted, resulting in poor fault tolerance and directly affecting the inspection progress. At the same time, the pre-set path cannot adapt to the dynamic obstacles in complex urban areas, and its lack of flexibility brings many inconveniences to actual work. Therefore, a control algorithm for collaborative inspection of UAV swarms is proposed to address the above problems. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies and solve at least one of the technical problems mentioned in the background, this invention proposes a control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms.

[0006] The technical solution adopted by this invention to solve its technical problem is: a control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms, comprising the following steps: S1. Obtain task configuration information: including the inspection area boundary, the coordinates of the virtual navigator preset on the inspection path, the safety threshold and the communication range; S2. Enter the autonomous cooperative control cycle. Within each control cycle: a) Perceive local environmental information, including identifying physical neighbor drones within communication range, obtaining its relative position with at least one virtual navigator, and detecting surrounding obstacles; b) Based on the information perceived in step a), calculate the following three local behavioral forces in parallel: - Neighbor aggregation forces drive the drone to move closer to the geometric center of its physical neighbors; - Collision avoidance force, which generates repulsion when the drone is less than a safe threshold from a neighbor or obstacle; - Directional alignment force drives the drone's flight direction to converge with the average flight direction of its physical neighbors; S3. Based on the current task urgency or the overall cluster status, dynamically adjust the weight coefficients of the neighbor aggregation force, collision avoidance force, and orientation alignment force. S4. Perform vector synthesis of the weighted three forces to generate and execute the flight control command for the current control cycle; S5. Repeat steps S2 to S4 until the inspection task is completed. During the cycle, continuously monitor the status of neighbors. If communication loss or failure is detected in some drones, the remaining drones will automatically trigger the formation reorganization mechanism.

[0007] Preferably, the mathematical expressions for the neighbor aggregation force, collision avoidance force, and orientation alignment force in S2 are as follows: Cohesive force: ; Avoiding forces: ; Alignment force: ; in For cohesive forces, The aggregation weight is N, where N is the number of neighbors, x is the location coordinate of the i-th neighbor, and x is the current location coordinate of the drone; where To avoid the force, To avoid weighting issues, M represents the number of nearby neighbors / obstacles. The coordinates of the neighbor / obstacle; To align the forces, To align weights, Let v be the flight speed vector of the i-th neighbor, and v be the flight speed vector of the current UAV.

[0008] Preferably, the weights of the three rules in S3 are... , , The total weight is always 1, dynamically adjusted based on task urgency and cluster status.

[0009] Preferably, the scene modes include at least: In normal inspection mode, weight allocation prioritizes ensuring safety and formation stability; In emergency mission mode, the direction alignment weight is increased to speed up cluster movement; Cluster-based distributed mode, increasing aggregation weight to quickly reorganize the formation.

[0010] Preferably, in the sensing step of S2, the virtual navigator is treated as a special type of fixed neighbor; and based on the relative position of the current user and the virtual navigator, an additional virtual navigator force is calculated, the mathematical expression of which is: ;in It is a virtual target point calculated based on the global task and local environment, and this force will participate in the vector synthesis in step S4.

[0011] Preferably, the virtual navigator is a series of coordinate points pre-set on the inspection area or path according to the inspection task requirements; each UAV selects one or more of the nearest virtual navigators as its current target navigator according to its current task stage or location.

[0012] Preferably, the security threshold is 3-8 meters and the communication range is 50-150 meters.

[0013] As a preferred embodiment, the formation reorganization mechanism described in S5 is as follows: when a neighbor is detected to be lost, the remaining drones immediately rebuild the local interactive network based on the current physical neighbors and virtual navigators, and automatically enter the cluster dispersion mode or a similar weight configuration. Through the enhanced aggregation force, the cluster is driven to form a new compact formation in a short period of time, and then restores to the weight configuration adapted to the new formation.

[0014] Preferably, the formation reorganization mechanism can be triggered when up to 30% of cluster nodes fail or communication is interrupted, and ensures that the reorganized cluster continues to complete the core inspection task.

[0015] This algorithm, through its innovative decentralized architecture and dual-neighbor system design, enables drone swarms to maintain a task completion rate of over 92% even when up to 30% of nodes experience communication loss or failure, demonstrating significantly superior fault tolerance compared to existing technologies. Its core relies on only three simple behavioral rules, requiring minimal single-machine computing power and communication bandwidth, making it compatible with low-cost civilian drone platforms and reducing large-scale deployment costs by approximately 40%. Combined with a dynamic weight adjustment mechanism and virtual navigator guidance, the system can autonomously avoid dynamic obstacles in complex urban areas while accurately following preset inspection routes, achieving a balance between flexibility and accuracy. Furthermore, based on the principle of biomimetic emergence, the swarm can autonomously adjust its overall formation according to the number of participating drones, intelligently adapting to the different ranges and levels of detail required for inspection tasks. This solves the problems of poor fault tolerance and inability to adapt to dynamic obstacles in complex urban areas when using preset global path planning for collaborative drone swarm inspections. These problems stem from reliance on a central node for unified path planning, which can lead to swarm paralysis if the central node fails or communication is interrupted. Additionally, preset paths often fail to adapt to dynamic obstacles in complex urban areas, resulting in insufficient flexibility and numerous inconveniences for frontline inspection operations. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0017] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The following is in conjunction with the appendix Figure 1 This application will be described in further detail. This application discloses a control algorithm for collaborative inspection by a swarm of unmanned aerial vehicles (UAVs), comprising the following steps: S1. Obtain task configuration information, including the inspection area boundary, the coordinates of the virtual navigator preset on the inspection path, the safety threshold, and the communication range; S2. Enter the autonomous cooperative control cycle. Within each control cycle: a) Perceive local environmental information, including identifying physical neighbor drones within communication range, obtaining its relative position with at least one virtual navigator, and detecting surrounding obstacles; b) Based on the information perceived in step a), calculate the following three local behavioral forces in parallel: - Neighbor aggregation forces drive the drone to move closer to the geometric center of its physical neighbors; - Collision avoidance force, which generates repulsion when the drone is less than a safe threshold from a neighbor or obstacle; - Directional alignment force drives the drone's flight direction to converge with the average flight direction of its physical neighbors; S3. Based on the current task urgency or the overall cluster status, dynamically adjust the weight coefficients of the neighbor aggregation force, collision avoidance force, and orientation alignment force. S4. Perform vector synthesis of the weighted three forces to generate and execute the flight control command for the current control cycle; S5. Repeat steps S2 to S4 until the inspection task is completed. During the cycle, continuously monitor the status of neighbors. If communication loss or failure is detected in some drones, the remaining drones will automatically trigger the formation reorganization mechanism.

[0020] Example 1 Each drone participating in the swarm is equipped with the same hardware modules, forming an independent intelligent agent, including: Environmental perception module: typically includes GPS / BeiDou positioning unit, inertial measurement unit, and forward or binocular vision sensor / millimeter-wave radar, used to obtain its own precise position and velocity vector, and detect static or dynamic obstacles within a certain distance.

[0021] Local communication module: Adopts self-organizing network communication protocols, such as Wi-Fi Mesh or LoRa, to ensure that the drone can periodically broadcast and exchange its core status information with other drones within the preset communication range, including ID, position x, speed v, and health status.

[0022] Data processing and control module: This is an onboard computing unit that runs the core program of the control algorithm described in this invention. This program executes cyclically at a fixed control frequency.

[0023] Flight execution module: Receives flight control commands generated by the control module and drives the UAV's power system to perform corresponding attitude and position adjustments.

[0024] S1: Task initialization and parameter loading; The ground station or the first drone to take off broadcasts a configuration information packet for this inspection mission to all drones in the cluster via a communication link. This packet includes at least: Inspection area boundary: Defines a polygonal area using a series of geographic coordinate points; Virtual navigator queue: A series of virtual target point coordinates set at certain intervals on a pre-planned inspection path or point of interest; these points are pre-stored in the memory of each drone. Core parameters: security threshold, effective communication radius, and initial weights of the three core rules.

[0025] S2: Enter the autonomous and coordinated control main loop.

[0026] Each drone independently initiates an infinite loop after takeoff until the mission end command is triggered, within each control cycle; Among them, a) Local environment perception: The local communication module receives status information broadcast by other drones within the communication radius, forming the current physical neighbor set.

[0027] Based on its current location, it calculates one or more nearest navigation points from the virtual navigator queue as the current target virtual navigator.

[0028] Using airborne vision / radar sensors, detect and identify a set of obstacles in the surrounding area that are within a short distance.

[0029] Based on the information perceived in a), the three fundamental force vectors are calculated simultaneously; for simplicity, they are expressed here in discretized vector form, which are implemented in the actual program using the corresponding mathematical formulas: Neighbor cohesion force : Calculate the average of the position vectors of all physical neighbors, and then calculate the vector from its own position to the center; this force causes the drones to move toward the center of the cluster, maintaining the integrity of the formation; Collision avoidance force : Traverse all physical neighbors and detected obstacles. If any pair is less than a safe threshold away from itself, calculate a repulsive force from that object toward itself, the magnitude of which is inversely proportional to the square of the distance.

[0030] Orientation alignment force The process involves calculating the average velocity vector of all physical neighbors, and then calculating the difference between this average velocity vector and the drone's own velocity vector. This force causes the drone to adjust its heading to align with the flight direction of its neighbors, thus achieving coordinated swarm movement.

[0031] S3: Dynamic weight adjustment.

[0032] According to the preset strategy, the weight coefficients of the three forces in step S2 (b) are adjusted in real time; the sum of the weights is always 1; the adjustment strategy is based on two dimensions: task urgency and cluster status. Cluster status judgment: The ratio of the current number of physical neighbors to the expected number, or the average distance of the neighbor distribution can be calculated; if the ratio is too low or the average distance is too large, the cluster is judged to be scattered.

[0033] Weight preset mode: Normal inspection mode: When the cluster formation is stable and there are no urgent tasks, [0.3, 0.4, 0.3] is used to prioritize obstacle avoidance safety and formation stability; Emergency Mission Mode: When rapid deployment to a location is required, the system or operator can issue a command, switching the weight to [0.2, 0.3, 0.5]. [Improvement] This enables the cluster to quickly align and accelerate its movement toward the target direction; Cluster Distributed Mode: When the algorithm detects that the formation is too loose, it automatically switches to [0.5, 0.3, 0.2]. This significantly improves performance. This prompts the drones to quickly converge towards the center and rapidly regroup.

[0034] S4: Synthesize and execute control commands.

[0035] First, the three forces are weighted and summed according to their current weights to obtain the resultant force vector. Then, this resultant force vector is converted into commands executable by the UAV's underlying flight control system. A common practice is to... It is considered as a virtual desired acceleration vector, which is input into the drone's speed controller or position controller, and finally executed by the flight control module driving the motor.

[0036] S5: Status monitoring and fault tolerance loop.

[0037] After executing the instructions, the algorithm enters the monitoring phase at the end of this cycle: Determine whether the inspection task has been completed.

[0038] Key fault tolerance mechanism: Continuously monitor changes in the physical neighbor set; if a neighboring drone is found to have continuously lost signals for more than a certain time threshold, the node is determined to be faulty.

[0039] Trigger formation reorganization: Once a failed node is detected, the local machine immediately: Remove the failed node from the neighbor list and recalculate based on the remaining physical neighbors and virtual navigator.

[0040] The drone automatically switches its own weight mode to cluster distributed mode ([0.5, 0.3, 0.2]). All surviving drones in the cluster do this simultaneously, and driven by strong cohesion, they quickly fill the formation gaps caused by node failures, forming a new, compact formation within seconds.

[0041] Once the formation is stable, the weighted mode can automatically revert to the normal inspection mode and continue to execute unfinished inspection tasks.

[0042] Example 2 Building upon Example 1, this example further refines the role of the virtual navigator. Virtual Navigation Force The calculation is formally incorporated into the parallel computation of step S2b), serving as the fourth rule guiding the global movement of the cluster, and its mathematical expression is: ;in It is a virtual target point calculated based on the global task and local environment, and this force will participate in the vector synthesis in step S4; Target point selection strategy: It's not simply about choosing the geographically nearest waypoint. A smarter strategy is to maintain a task progress pointer for each drone, visiting points in the virtual waypoint queue in a pre-defined order. Simultaneously, by incorporating neighboring location information, it's possible to prevent all drones from rushing to the same point. For example, drones can be guided to choose waypoints ahead that differ from their neighbors' target points, thus guiding the swarm to form a streamlined movement.

[0043] At this point, the dynamic weight adjuster needs to manage four weights, but its core principle remains unchanged: when strict adherence to the path is required, the weights can be increased. When flexible obstacle avoidance is required, it can be temporarily enlarged.

[0044] Example 3: Fault-Tolerant Inspection Experiment in Complex Urban Environments This embodiment demonstrates the effectiveness of the algorithm through a simulation experiment: Experimental setup: Fifteen drones were deployed to conduct a grid-based inspection of an urban area of ​​approximately 5 square kilometers, including several irregularly shaped buildings and simulated emergency construction zones. The virtual navigator was positioned along a planned bow-shaped path. The safety threshold was set at 5 meters, and the communication range at 100 meters.

[0045] Normal execution: After takeoff, the drones quickly form a stable diamond formation. Under algorithmic control, the formation autonomously advances along a virtual navigation point. When encountering buildings, the outer drones automatically detour due to the increased F_rep value, while the inner drones adjust accordingly. The entire formation smoothly bypasses obstacles like a fluid, and automatically returns to its original formation afterward.

[0046] Fault tolerance test: During the mid-stage of the inspection, a simulation was performed in which five drones experienced a communication link interruption due to strong interference. The remaining ten drones detected the loss of their neighbor almost simultaneously, instantly triggering the regrouping mechanism.

[0047] All surviving drones were switched to a weight mode of [0.5, 0.3, 0.2].

[0048] Under the influence of strong cohesion, the surviving cluster shrinks and reorganizes into a smaller, but still intact, diamond-shaped formation within about 12 seconds.

[0049] Virtual leadership during the restructuring period The new formation was guided to move slowly toward the target point without any pauses.

[0050] After the formation stabilized, the weights were restored to [0.3, 0.4, 0.3], and the new formation continued to complete the remaining 80% of the inspection area coverage, ultimately achieving a task completion rate of 94%. In contrast, the control group using the traditional centralized scheduling algorithm experienced system chaos after losing 33% of its nodes, with a task completion rate of less than 50%.

[0051] In summary, through the specific embodiments described above, this invention achieves intelligent collaboration, dynamic obstacle avoidance, path following, and strong fault tolerance in UAV swarms under decentralized conditions that rely solely on local interactions. It is particularly suitable for inspection tasks in complex and uncertain environments. Furthermore, it supports multi-hop Ad-hoc communication, maintaining stable connections in areas without ground networks; it features automatic fault switching, ensuring that the exit of a single UAV does not affect the overall mission; and when combined with a fixed airport, it can achieve 24 / 7 unmanned inspection.

[0052] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A control algorithm for collaborative inspection by a swarm of unmanned aerial vehicles (UAVs), characterized in that: Includes the following steps: S1. Obtain task configuration information: including the inspection area boundary, the coordinates of the virtual navigator preset on the inspection path, the safety threshold and the communication range; S2. Enter the autonomous cooperative control cycle. Within each control cycle: a) Perceive local environmental information, including identifying physical neighbor drones within communication range, obtaining its own relative position to at least one virtual navigator, and detecting surrounding obstacles; b) Based on the information perceived in step a), calculate the following three local behavioral forces in parallel: - Neighbor aggregation forces drive the drone to move closer to the geometric center of its physical neighbors; - Collision avoidance force, which generates repulsion when the drone is less than a safe threshold from a neighbor or obstacle; - Directional alignment force drives the drone's flight direction to converge with the average flight direction of its physical neighbors; S3. Based on the current task urgency or the overall cluster status, dynamically adjust the weight coefficients of the neighbor aggregation force, collision avoidance force, and orientation alignment force. S4. Perform vector synthesis of the weighted three forces to generate and execute the flight control command for the current control cycle; S5. Repeat steps S2 to S4 until the inspection task is completed. During the cycle, continuously monitor the status of neighbors. If communication loss or failure is detected in some drones, the remaining drones will automatically trigger the formation reorganization mechanism.

2. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The mathematical expressions for the neighbor aggregation force, collision avoidance force, and orientation alignment force in S2 are as follows: Cohesive force: ; Avoiding forces: ; Alignment force: ; in For cohesive forces, The aggregation weight is N, where N is the number of neighbors, x is the location coordinate of the i-th neighbor, and x is the current location coordinate of the drone; where To avoid the force, To avoid weighting issues, M represents the number of nearby neighbors / obstacles. The coordinates of the neighbor / obstacle; To align the forces, To align weights, Let v be the flight speed vector of the i-th neighbor, and v be the flight speed vector of the current UAV.

3. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The weights of the three rules in S3 , , The total weight is always 1, dynamically adjusted based on task urgency and cluster status.

4. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The scenario modes include at least: In normal inspection mode, weight allocation prioritizes ensuring safety and formation stability; In emergency mission mode, the direction alignment weight is increased to speed up cluster movement; Cluster-based distributed mode, increasing aggregation weight to quickly reorganize the formation.

5. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: In the sensing step of S2, the virtual navigator is treated as a special type of fixed neighbor; and based on the relative position of the current user and the virtual navigator, an additional virtual navigator force is calculated, the mathematical expression of which is: ;in It is a virtual target point calculated based on the global task and local environment, and this force will participate in the vector synthesis in step S4.

6. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The virtual navigator is a series of coordinate points pre-set on the inspection area or path according to the inspection task requirements; each UAV selects one or more of the nearest virtual navigators as its current target navigator according to its current task stage or position.

7. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The safety threshold is 3-8 meters, and the communication range is 50-150 meters.

8. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The formation reorganization mechanism described in S5 is as follows: when a neighbor is detected to be lost, the remaining drones immediately rebuild the local interactive network based on the current physical neighbors and virtual navigators, and automatically enter the cluster dispersion mode or similar weight configuration. Through enhanced aggregation force, the cluster is driven to form a new compact formation in a short time, and then restores to the weight configuration adapted to the new formation.

9. The control algorithm for collaborative inspection of unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The formation reorganization mechanism can be triggered when up to 30% of cluster nodes fail or communication is interrupted, and ensures that the reorganized cluster continues to complete the core inspection task.

Images