Control method for automatic battery changing and chemical adding of unmanned aerial vehicle

Abstract

The invention discloses an unmanned aerial vehicle automatic battery changing and chemical adding control method comprising the following steps: S1, a central control unit receives a service request signal sent by a to-be-supplied unmanned aerial vehicle, activates a global vision module to obtain an initial pose of the unmanned aerial vehicle in a parking apron coordinate system and land, and records a parking pose after the unmanned aerial vehicle is stably parked; s2, the central control unit analyzes a supply identifier in the service request signal, loads a corresponding exclusive control time sequence and a parameter set according to the supply identifier, and determines a target operation point based on a final parking pose to plan a motion path; and S3, the central control unit drives the multi-axis motion system and controls the end effector to execute the operation process corresponding to the supply identifier, so that the unmanned aerial vehicle is guided, the initial pose of the unmanned aerial vehicle under the parking apron coordinate system can be accurately obtained, the unmanned aerial vehicle is guided to land stably through a differential correction signal, and the unmanned aerial vehicle is guided to land stably. And finally, the parking posture is recorded as an operation reference, so that the operation error caused by the deviation of the landing position is effectively eliminated.

Description

Technical Field

[0001] This invention relates to the field of battery and pesticide replacement technology for agricultural drones, and in particular to a control method for automatic battery replacement and pesticide application for drones. Background Technology

[0002] With the rapid development of drone technology, it has been widely used in various fields such as agricultural plant protection, logistics and transportation, and emergency rescue. Especially in agricultural plant protection, drones, with their advantages of high operating efficiency, wide coverage, and adaptability to complex terrain, are gradually replacing traditional manual or mechanical spraying methods and have become an important tool in modern agriculture. Through pesticide application and intelligent operation, drones not only significantly improve pesticide utilization and reduce agricultural production costs, but also effectively reduce potential harm to the environment and human health, promoting the transformation and upgrading of agriculture towards high efficiency, greenness, and intelligence.

[0003] Currently, ground-based battery swapping and drug refueling operations for drones largely rely on manual labor or semi-automated equipment, resulting in low efficiency, poor consistency, and susceptibility to human error. While existing automated ground station technologies can achieve some auxiliary functions, they still have significant shortcomings in precise positioning, adaptive operation, and safety control. The existing technologies struggle to achieve high-precision perception and stable guidance of the drone's attitude, leading to significant positioning deviations at work sites. Furthermore, they lack intelligent identification and adaptive control mechanisms for different refueling tasks (such as battery replacement and drug refueling), hindering flexible adjustments to operating units and operational parameters. Additionally, the lack of sufficient integration of force control and torque closed-loop strategies in robotic arm motion planning and end-effector operations makes them susceptible to operational failures or equipment damage due to positioning errors or external interference, affecting system reliability and continuity.

[0004] Therefore, it is necessary to provide a control method for automatic battery swapping and drug dispensing of unmanned aerial vehicles (UAVs) to solve the above-mentioned technical problems. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides a control method for automatic battery swapping and drug dispensing of unmanned aerial vehicles (UAVs).

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a control method for automatic battery swapping and refueling of unmanned aerial vehicles (UAVs), the method being based on an automated ground station, the automated ground station including a central control unit, a multi-axis motion system, and an end effector installed at the end of the multi-axis motion system, the method comprising the following steps: S1: The central control unit receives the service request signal sent by the UAV to be resupplyed, activates the global vision module to obtain the initial pose of the UAV in the helipad coordinate system, guides the UAV to land, and records its final parking pose after the UAV comes to a stop. S2: The central control unit parses the replenishment identifier in the service request signal, loads the corresponding exclusive control timing and parameter set according to the result of "battery replacement" or "medicine refill", and determines the target operation point based on the final parking posture of the UAV to plan the motion path of the multi-axis motion system. S3: The central control unit drives the multi-axis motion system and controls the end effector to perform the work process corresponding to the replenishment mark.

[0007] In a preferred embodiment of the present invention, in step S1, the global vision module captures real-time images of the helipad area containing the UAV, and compares the UAV features in the image with a pre-stored standard UAV vision model using a template matching-based positioning algorithm to calculate the initial position of the UAV in the helipad coordinate system.

[0008] In a preferred embodiment of the present invention, in step S1, the central control unit calculates the guidance trajectory based on the initial pose and sends a differential correction signal to the UAV through a positioning beacon deployed on the helipad, guiding it to fly along the guidance trajectory to the predetermined landing center point.

[0009] In a preferred embodiment of the present invention, in step S1, during the landing process and after the UAV comes to a complete stop, the global vision module continuously updates its pose and records the final pose of the UAV after it has come to a complete stop as the parking pose.

[0010] In a preferred embodiment of the present invention, in step S2, the motion path planning adopts the RRT algorithm, with the target work point and the current position of the end effector as the starting point and the ending point, and the path search is performed in a preset three-dimensional environment model.

[0011] In a preferred embodiment of the present invention, the cost function of the RRT algorithm is defined as a weighted sum of path length, joint movement speed, and safe distance from the obstacle.

[0012] In a preferred embodiment of the present invention, in step S2, the trajectory planning module smooths the path point sequence using a fifth-order B-spline interpolation algorithm to generate a motion trajectory that is continuous in both velocity and acceleration.

[0013] In a preferred embodiment of the present invention, in S3, when the replenishment identifier is "battery replacement", the activated rigid operation unit performs battery disassembly and installation operations, including pressing the battery compartment unlock button with a linear servo push rod and hooking the battery handle with an electronically controlled self-locking hook.

[0014] In a preferred embodiment of the present invention, in S3, when the replenishment identifier is "medicine liquid filling", the activated flexible operation unit performs the opening and tightening operation of the medicine tank cover, including using a three-finger pneumatic flexible gripper to grasp the medicine tank cover and using a coaxial torque control module to control the tightening torque.

[0015] In a preferred embodiment of the present invention, in step S3, before executing the work process, a secondary fine positioning step based on end vision is further included. The end vision module captures a local three-dimensional point cloud image of the target work point and registers it with a pre-stored standard three-dimensional model to calculate the coordinates of the center point of the target work point.

[0016] This invention addresses the shortcomings of the prior art and has the following beneficial effects: This invention provides a control method for automatic battery swapping and refueling of unmanned aerial vehicles (UAVs). By setting up a dual positioning system with a global vision module and an end vision module, it can achieve high-precision perception and guidance of the UAV's position and attitude. It can accurately obtain the initial position and attitude of the UAV in the landing pad coordinate system, guide the UAV to land smoothly through differential correction signals, and finally record the parking position and attitude as the operation reference. This effectively eliminates the operational errors caused by landing position deviations and improves the positioning accuracy and reliability of the overall system.

[0017] This invention provides a control method for automatic battery swapping and drug dispensing of unmanned aerial vehicles (UAVs). By parsing the replenishment identifier in the service request, it can dynamically load the corresponding exclusive control timing sequence and parameter set, thereby realizing the intelligent and adaptive operation process. It can automatically identify different task requirements such as battery replacement or drug dispensing and match the corresponding rigid or flexible operating units. By using the RRT algorithm for collision-free path planning and combining it with the fifth-order B-spline interpolation algorithm to smooth and optimize the motion trajectory, it ensures that the robotic arm can complete the action efficiently and smoothly in different replenishment tasks, significantly improving the system's adaptability and operational efficiency.

[0018] This invention provides a control method for automatic battery swapping and refueling of unmanned aerial vehicles (UAVs). By employing force control and torque control models combined with real-time status monitoring, the method ensures the safety and stability of the operation process. During end-effector operation, force control is achieved through the electrically controlled self-locking hook of the linear servo push rod, or smooth twisting operation is achieved through the three-finger pneumatic flexible gripper and coaxial torque control module. At the same time, joint torque and operating parameters are monitored in real time, and safety protection mechanisms are triggered in a timely manner, significantly reducing the risk of system interruption and improving the continuity and reliability of unmanned refueling missions. Attached Figure Description

[0019] 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 recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1This is a flowchart illustrating a preferred embodiment of the present invention. Detailed Implementation

[0020] 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.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0022] like Figure 1 As shown, this invention provides a control method for automatic battery swapping and dosing of unmanned aerial vehicles (UAVs). The method is based on an automated ground station, which includes a central control unit, a multi-axis motion system, and an end effector installed at the end of the multi-axis motion system. The method includes the following steps: S1: The central control unit receives the service request signal sent by the UAV to be resupplyed, activates the global vision module to obtain the initial pose of the UAV in the helipad coordinate system, guides the UAV to land, and records its final parking pose after the UAV comes to a stop. S2: The central control unit parses the replenishment identifier in the service request signal, loads the corresponding exclusive control timing and parameter set according to the result of "battery replacement" or "medicine refill", and determines the target operation point based on the final parking posture of the UAV to plan the motion path of the multi-axis motion system. S3: The central control unit drives the multi-axis motion system and controls the end effector to perform the work process corresponding to the replenishment mark.

[0023] The following will describe each step in detail.

[0024] The execution carrier of this invention is an automated ground station, and the automated ground station system is a central control unit. Specifically, the central control unit is an industrial-grade computer equipped with an Intel Core i7-8665UE processor and 32GB DDR4 memory, running a real-time Linux operating system based on Ubuntu 18.04, and deploying a robot operating system as its core software framework.

[0025] The central control unit interacts with the multi-axis motion system, the end effector with integrated task-specific execution unit, and the multimodal sensing system via an industrial Ethernet bus for high-speed, deterministic data exchange.

[0026] The multi-axis motion system specifically refers to a six-axis collaborative robot, such as the Universal Robots UR10e model, which has a working radius of 1300 mm and a repeatability of ±0.05 mm, capable of covering the entire drone landing pad's workspace. Each of the six joints of the six-axis collaborative robot has a built-in torque sensor, providing the hardware foundation for subsequent force control operations. The structure and function of the end effector will be described in detail later; The multimodal perception system is a hierarchical perception network, which includes a global vision module deployed at a high ground station, an end vision module integrated on the end effector, a lidar system deployed around the work area, and an environmental monitoring module.

[0027] In step S1, after receiving the service request signal sent by the UAV to be resupplyed, the central control unit immediately activates the global vision module to start the UAV pose acquisition process. The service request signal includes the UAV's unique identification code, the current position coordinates provided by the onboard GPS / RTK system, and the resupply task type identifier, such as battery replacement or liquid refill.

[0028] The global vision module uses a high-resolution industrial camera, such as the Baslerac A2500-14gc model, equipped with a 1 / 1.8-inch CMOS sensor and a 6mm focal length lens, to capture real-time images of the apron area at a rate of 30 frames per second.

[0029] The image processing unit within the central control unit executes a template-matching-based localization algorithm: first, the image is converted to grayscale, then subjected to Gaussian filtering, and... Edge detection is used to extract the drone's contour features, and these features are then compared with a pre-stored standard drone visual model, which is formed by fusing multiple-angle images. Pose calculation is achieved by solving a least-squares problem, namely minimizing the reprojection error.

[0030] The formula for the reprojection error in pose calculation is: ; in, These are the feature points detected in the image. These are the corresponding 3D points in the standard visual model. It is the camera intrinsic parameter matrix. and These are the rotation matrix and translation vector to be solved. This is the camera projection function. The algorithm calculates the initial pose of the UAV in the helipad coordinate system, including three-dimensional coordinates. Euler angle posture The initial positioning accuracy can reach ±10 cm.

[0031] Based on the initial pose, the central control unit calculates the guidance trajectory and guides the UAV to the predetermined landing center point. A continuous path is generated based on the initial pose and landing point coordinates to ensure that the UAV descends at a constant speed. Specifically, the trajectory calculation employs a geometric method that combines piecewise linear interpolation with circular arc transition to ensure that the trajectory is continuous, smooth, and easy for UAVs to track. The process of calculating the guiding trajectory is as follows: The central control unit defines critical path points and generates segmented linear trajectories. The system uses the initial position of the UAV calculated by the global vision module as the starting point. The landing center point will be the endpoint. ; Subsequently, the system calculates a horizontal projection point located directly above the landing point and at the same initial altitude as the drone. The trajectory planning is divided into two linear parts: the first part is the horizontal correction trajectory from... arrive By linear interpolation formula Description, parameters The drone's horizontal movement is guided by a constant velocity change from 0 to 1, moving it horizontally above the landing point; the second vertical descent trajectory... arrive From the formula Description, parameters The speed changes uniformly from 0 to 1, guiding the drone to a vertical descent. Next, to solve the problem of two linear trajectories at the connection point To address the abrupt change in velocity direction, the system introduces a circular transition segment for smoothing. This method involves truncating a small segment from the end of the horizontal trajectory. and the initial segment of the vertical trajectory And calculate a relationship with line segment and A circular arc that is tangent to each other. The radius of this circular arc. This is a key parameter, and its value is strictly set according to the maximum centripetal acceleration limit of the drone. It is designed to ensure that the drone's turning action is extremely smooth when transitioning from horizontal flight to vertical descent, avoiding sharp turns or flight instability. Finally, the system synthesizes and outputs the final smooth trajectory. By seamlessly connecting the three parts—the horizontal straight segment, the transition arc segment, and the vertical straight segment—the system ultimately synthesizes a complete guide trajectory that remains continuous in both spatial position and velocity. The central control unit discretizes this smooth trajectory into a series of target point sequences corresponding to timestamps, and sends this sequence, along with control commands calculated in real time by the PID controller, to the UAV flight control system. This enables stable tracking of the predetermined trajectory, ensuring that the UAV autonomously and safely completes the landing process.

[0032] Meanwhile, the central control unit sends high-frequency differential correction signals to the UAV via ultra-wideband (UWB) positioning beacons deployed around the helipad to correct the UAV's positioning errors. The differential correction is based on real-time position deviation.

[0033] The formula for the PID controller in UAV flight control is: ; in, It is the positional error between the drone's current position and the target point on the guidance trajectory. , , It refers to proportional, integral, and differential gain. The calculated control variables are used to adjust the drone's flight attitude. The drone flies autonomously along the guided trajectory, with the global vision module continuously providing attitude feedback, forming a closed-loop control system to ensure stable flight.

[0034] During the drone's landing process, the global vision module continuously updates the pose at a frequency of 30Hz, determining the stationary state by comparing pose changes between consecutive frames. The landing gear contacts the pressure sensor on the tarmac, and the pose change is less than a preset threshold; for example, a change in position... mm, posture changes When the position reaches 0° and remains stationary for 1 second, the central control unit determines that the UAV has come to a complete stop. Finally, the system records the UAV's parking posture, which serves as the reference coordinate system for subsequent multi-axis motion system operations, eliminating errors caused by landing inconsistencies. The posture recording format is a six-degree-of-freedom vector [x,y,z,α,β,γ], where α, β, and γ represent rotation angles around the X, Y, and Z axes, respectively.

[0035] In step S2, the central control unit first parses the resupply identifier from the service request signal from the stationary UAV. This identifier is a predefined enumeration type variable with the value of "battery replacement" or "liquid replenishment". The parsing process is completed by an efficient string matching algorithm to ensure the accuracy and real-time performance of the instruction recognition.

[0036] Based on the parsing results of the supply identifier, the central control unit loads the corresponding dedicated control timing and parameter set dataset from the local solid-state drive and stores it in XML format with a strict structured schema.

[0037] For example, if the identifier is "battery replacement", a file named Battery_Replacement_Sequence.xml is loaded. This file defines in detail the motion sequence of rigid operating units (such as linear servo actuators and electronically controlled hooks), the expected motion trajectory of each joint, the target torque and speed curve of the servo motor, and parameters such as safety interlock conditions.

[0038] Conversely, if the identifier is "medicine filling", the Liquid_Refill_Sequence.xml file is loaded, which contains the control parameters of the flexible operating unit (such as the three-finger pneumatic gripper and coaxial torque module), the number of turns, the preset torque threshold, and the calibration data of the flow sensor.

[0039] Subsequently, the system uses the final parking pose of the UAV recorded in step S1 as a basis. The three-dimensional coordinates of the target work point are determined through coordinate transformation; For example, the offset of the battery compartment relative to the origin of the drone's fuselage coordinate system. It was foreseen; The target coordinates of the battery compartment unlock button or the center point of the medicine box cap in the world coordinate system of the helipad are obtained by calculating using the homogeneous coordinate transformation matrix. .

[0040] The formula for homogeneous coordinate transformation is: ; in, It is a 4x4 transformation matrix composed of the parking pose. These are the homogeneous coordinates of the work point in the fuselage coordinate system. This is used when determining the target work point. Then, the motion planning module is activated, responsible for planning the motion from the current position of the end effector for the multi-axis motion system (six-axis collaborative robot). arrive The collision-free motion path is determined. The path planning uses the Fast Exploratory Random Tree (RRT) algorithm.

[0041] The algorithm searches within a pre-constructed 3D voxel mesh environment, which integrates static obstacle information from the helipad, the UAV model, and the ground station itself. The cost function of the RRT algorithm... Set as total path length Maximum speed of movement of each joint The reciprocal (to encourage smooth movement), and the distance to the nearest obstacle. The weighted sum, mathematically expressed as: ; in, , , These are weighting coefficients adjusted based on safety and efficiency requirements. This cost function guides the algorithm to generate an optimal and safe path in joint space. The planned path point sequence... (in The vector representing the robot's joint angles is then fed into the trajectory planning module for smoothing. This module uses a quintic B-spline interpolation algorithm to fit the discrete path points. The quintic B-spline curve ensures that the generated trajectory is continuous in terms of position, velocity, and even acceleration, thus effectively avoiding impacts and vibrations during motion. Its mathematical expression is: ; in, It is a parameter The corresponding trajectory points, It is a control point. It is a quintic B-spline basis function; Finally, a time-parameterized, smooth joint space trajectory is generated, preparing for motion control in step S3.

[0042] In step S3, the central control unit first drives the multi-axis motion system to execute the motion trajectory planned in step S2, and controls the end effector to move to the vicinity of the target work point; then, the system starts the secondary fine positioning process based on end vision to improve the operation accuracy; the end vision module captures the local three-dimensional point cloud image of the target work point, and registers it with the pre-stored standard three-dimensional model through the iterative nearest point algorithm to calculate the precise coordinates of the center point of the target work point.

[0043] The mathematical expression of the ICP algorithm is to minimize the distance between point clouds: ; in These are points in the model's point cloud. These are points in a scene point cloud. It is a rotation matrix. It is a translation vector; by solving this optimization problem, the system compensates for the positioning error to the millimeter level, ensuring that the end effector is accurately aligned with the work point.

[0044] Specifically, based on the parsing result of the replenishment identifier, the central control unit activates the corresponding end effector unit to execute the dedicated operation process; if the identifier is for battery replacement, the rigid operation unit is activated, and the linear servo push rod presses the battery compartment unlock button, whose force control model is based on a spring-damped system. ; in For output force, This is the stiffness coefficient. The damping coefficient is... To prevent positional deviations, the pressing pressure is kept stable within a preset threshold; simultaneously, the electronically controlled self-locking hook engages the battery handle, and the motor current is monitored. Estimate gripping force ,in With a constant torque, safe gripping is achieved. The multi-axis motion system completes battery disassembly and installation in force control mode, avoiding impact.

[0045] If the replenishment identifier is liquid medicine filling, the flexible operation unit is activated, the three-finger pneumatic flexible gripper grabs the medicine tank lid, and the coaxial torque control module performs the screwing operation.

[0046] Torque control uses a proportional-derivative algorithm: ; in, For command torque, From the perspective of expectation, From a practical perspective, To account for angular errors, ensure that torque is limited within a safe range.

[0047] During drug dispensing, the flow sensor monitors the instantaneous flow rate. The cumulative flow is calculated by integrating over time: ,when Once the set value is reached, the dispensing will automatically stop; after completion, the grippers will tighten the medicine tank cap again, and the torque module will ensure a tight seal.

[0048] During operation, the system monitors the status of each joint in real time and reads data from the torque sensor. It is then compared with the expected value τd; if the deviation exceeds the threshold, an emergency stop is triggered. .

[0049] Upon completion of the mission, the central control unit sends a confirmation signal to the drone and controls the multi-axis motion system to reset to the standby position. The entire process forms a closed-loop control, ensuring the safety and reliability of the operation, and ultimately completing the drone's automatic battery swap or resupply. Example 1

[0050] In S1, the central control unit receives the service request signal sent by the DJI T30 drone, activates the global vision module to capture real-time images of the helipad area, calculates the initial pose of the drone in the helipad coordinate system through a template matching-based positioning algorithm, calculates the guidance trajectory, sends differential correction signals to the drone through the ultra-wideband positioning beacon deployed on the helipad, guides it to fly along the guidance trajectory to the predetermined landing center point, and records its final parking pose after the drone has come to a complete stop. In S2, the central control unit parses the replenishment identifier in the service request signal, loads the corresponding dedicated control timing and parameter set according to the result of "battery replacement" or "medicine refill", and determines the target operation point coordinates of the battery compartment unlock button or the center point of the medicine box cover through homogeneous coordinate transformation based on the final parking posture of the UAV. The RRT algorithm is used to plan the collision-free motion path of the multi-axis motion system in the preset three-dimensional environment model, and the fifth B-spline interpolation algorithm is used to smooth the path point sequence to generate a continuous motion trajectory. In S3, the central control unit drives the multi-axis motion system to move along the planned trajectory to the vicinity of the target work point. Then, the end vision module is activated for secondary fine positioning. The captured 3D point cloud is registered with the pre-stored standard model through an iterative nearest-point algorithm to calculate the precise coordinates of the work point. If the replenishment label is "battery replacement", the rigid operation unit is activated to control the linear servo push rod to press the battery compartment unlock button with a preset force of 5N, and to control the electronically controlled self-locking hook to grab the 2.2kg battery. If the replenishment label is "medicine filling", the flexible operation unit is activated to control the three-finger pneumatic flexible gripper to grab the medicine tank cover with a gripping force of 8N, and to control the coaxial torque control module to perform a screwing operation with a target torque of 2.5N·m. During the operation, the joint torque and operating force / torque are monitored in real time and compared with the set threshold.

[0051] Comparative Example 1: In S1, the ground station receives the UAV service request, activates a single global vision module for coarse positioning, guides the UAV to land in the predetermined area of ​​the helipad, and records its theoretical parking posture. In S2, the system parses the supply identifier, loads a fixed control program, and plans the point-to-point motion path of the multi-axis motion system based on the idealized operation point coordinates preset by the UAV model and a simple linear interpolation algorithm. In S3, the central control unit drives the multi-axis motion system directly to the preset theoretical target point, omitting the end-effector precision positioning step. If a battery is being swapped, the universal two-finger parallel pneumatic gripper (replaced with a push rod fingertip) is controlled to attempt to press the battery unlock button, and then the fingertip is replaced to attempt to grip the battery handle. If medication is being added, the same gripper (replaced with a fingertip adapted to the medicine tank lid) is controlled to grip the medicine tank lid and perform a screwing operation with a fixed number of rotations and motor speed. The system does not have real-time force / torque closed-loop control function and relies only on basic physical limits and emergency stop buttons as safety measures.

[0052] Comparative Example 2: In S1, the system uses the UAV's GPS signal and ground beacon for joint positioning to guide the UAV to land and record its approximate pose; In S2, the system parses the service request, but regardless of the supply identifier, it loads the same set of general control timing and motion parameters, estimates the target work point based on the approximate pose, and uses the pre-recorded teaching path as the motion trajectory of the multi-axis motion system. In S3, the central control unit drives the multi-axis motion system to replay the teaching path; when performing a battery swapping task, a multi-purpose mechanical finger combination mechanism is used to sequentially press the button and grasp the battery; when performing a drug dispensing task, the same mechanism is used to attempt to grasp and rotate the medicine tank lid; all operations are performed based on position control mode, without binocular vision precision positioning and force / torque sensor feedback, and the system compensates for accumulated errors through periodic position calibration.

[0053] Experimental example: Experimental objective: To verify the performance differences between Example 1 and Comparative Examples 1 and 2 in the automatic battery swapping and dosing control of UAVs, with a focus on evaluating key indicators such as alignment accuracy, operation success rate, operation time, and system stability.

[0054] Experimental methods: Using a combination of simulation and physical testing, 200 cycles of battery swapping and chemical dosing operations were performed on each of Example 1, Comparative Example 1, and Comparative Example 2. The testing environment included a standardized helipad, a DJI T30 drone model, and an automated ground station system. Data was collected and recorded in real time using high-precision sensors such as vision systems, torque sensors, and timers.

[0055] Experimental steps: Preparation phase: Set up the test environment, initialize the automated ground station system, and calibrate the sensors; Test execution: For each embodiment and comparative example, the following operations were performed in sequence: the drone sent a service request to simulate the landing process; the battery swapping operation was performed (including battery unlock button alignment, battery grabbing and replacement); the medication filling operation was performed (including medication tank cover alignment, unscrewing and tightening, and medication filling); and the key parameters of each operation were recorded.

[0056] Data collection: Continuously record alignment accuracy, success rate, torque, operation time, and number of system interruptions over 200 cycles; Data analysis: The average values ​​of each indicator were calculated and compared. The results are shown in Table 1.

[0057] Table 1:

[0058] As shown in Table 1: The control method of Example 1 is significantly better than Comparative Example 1 and Comparative Example 2 in terms of positioning accuracy, operation success rate, work efficiency and system stability.

[0059] Example 1 uses the template matching algorithm of the global vision module for initial coarse positioning, and then combines the iterative nearest point algorithm of the end vision module for secondary fine positioning. This dual guarantee mechanism will improve the initial positioning accuracy at the decimeter level to the final alignment accuracy at the millimeter level. In contrast, Comparative Example 1 relies on a single global vision system, while Comparative Example 2 even uses GPS joint positioning with larger errors. Both lack a fine end-point correction process, which amplifies the alignment error by several times or even ten times. In terms of core control strategy, Example 1 adopts a closed-loop control algorithm based on a physical model, which is the key to its high success rate and operation quality. For battery replacement, Example 1 uses a force control model to precisely control the pressing and gripping force; for medication dispensing, a torque control algorithm is used to precisely regulate the tightening torque. This real-time feedback and adjustment of force and torque makes the operation both smooth and precise. In contrast, Comparative Example 1 and Comparative Example 2 both employ simple open-loop position control, mechanically executing preset actions. They cannot adapt to the minute changes and uncertainties in actual contact, resulting in problems such as grasping failure and insufficient torque.

[0060] In terms of motion planning, Example 1 uses the RRT algorithm to search for collision-free paths and combines it with fifth-order B-spline interpolation to generate smooth trajectories. This makes the movement of the robotic arm both efficient and stable, significantly shortening the operation time. In contrast, Comparative Example 1 uses simple linear interpolation, which may result in suboptimal paths and abrupt changes. Comparative Example 2 simply replays pre-recorded paths, failing to adapt to the subtle differences in each operation, leading to low efficiency and frequent interruptions. Example 1 constructs a complete real-time status monitoring and safety protection system. By continuously monitoring parameters such as joint torque and operating force and comparing them with safety thresholds, it can promptly trigger abnormal protection, thereby improving the robustness and reliability of the system.

[0061] In contrast, the safety measures in Comparative Examples 1 and 2 are relatively rudimentary, relying mainly on physical limits, resulting in a much higher number of system interruptions in 200 cycle tests.

[0062] In summary, the superiority of Example 1 does not stem from a breakthrough in a single technology, but rather from an advanced algorithm and control model system that works in synergy, consisting of "high-precision perception and positioning + intelligent closed-loop force control + optimized motion planning + comprehensive status monitoring". Comparative Examples 1 and 2, on the other hand, simplified or adopted outdated technologies at various stages, resulting in performance gaps in the overall system.

[0063] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A control method for automatic battery and drug replacement of a UAV, the method being based on an automated ground station, the automated ground station comprising a central control unit, a multi-axis motion system, and an end effector mounted at the end of the multi-axis motion system, characterized in that, The method comprises the following steps: S1: the central control unit receives the service request signal sent by the unmanned aerial vehicle to be supplied, activates the global vision module to obtain the initial pose of the unmanned aerial vehicle in the apron coordinate system, guides the unmanned aerial vehicle to land, and records the final parking pose of the unmanned aerial vehicle after the unmanned aerial vehicle is parked stably; S2: the central control unit analyzes the supply identifier in the service request signal, loads the corresponding exclusive control time sequence and parameter set according to the result of "battery replacement" or "liquid filling", and determines the target work point based on the final parking pose of the unmanned aerial vehicle to plan the motion path of the multi-axis motion system; S3: the central control unit drives the multi-axis motion system to control the end effector to execute the work process corresponding to the supply identifier. 2.The control method for automatic battery and drug replacement of the UAV according to claim 1, characterized in that: In the S1, the global vision module captures real-time images of the apron area containing the unmanned aerial vehicle, and compares the unmanned aerial vehicle features in the images with the pre-stored standard vision model of the unmanned aerial vehicle through a positioning algorithm based on template matching, to calculate the initial position of the unmanned aerial vehicle in the apron coordinate system. 3.The control method of claim 1, wherein: In the S1, the central control unit calculates a guide track based on the initial pose, and sends a differential correction signal to the unmanned aerial vehicle through the positioning beacon arranged in the apron, to guide the unmanned aerial vehicle to fly to a predetermined landing center point along the guide track. 4.The control method of claim 1, wherein: In the S1, the global vision module continuously updates the pose during the landing process of the unmanned aerial vehicle and after the unmanned aerial vehicle is parked stably, and records the final pose of the unmanned aerial vehicle after the unmanned aerial vehicle is completely parked as the parking pose. 5.The control method for automatic battery and drug replacement of the UAV according to claim 1, characterized in that: In the S2, the RRT algorithm is used for planning the motion path, and the target work point and the current position of the end effector are used as the starting point and the ending point for path searching in a preset three-dimensional environment model. 6.The control method of claim 5, wherein: The cost function of the RRT algorithm is defined as the weighted sum of the path length, the joint motion speed and the safe distance from the obstacle.

7. The control method for automatic battery and drug replacement of the unmanned aerial vehicle according to claim 1, characterized in that: In the S2, the path point sequence is smoothed through the five times B-spline interpolation algorithm by the trajectory planning module, to generate a motion trajectory which is continuous in speed and acceleration. 8.The control method of claim 1, wherein: In the S3, when the supply identifier is "battery replacement", the activated rigid operation unit performs the dismounting and mounting operation of the battery, including pressing the battery compartment unlocking button by using a linear servo push rod and hooking the battery handle by using an electrically controlled self-locking hook. 9.The control method of claim 1, wherein: In the S3, when the supply identifier is "liquid filling", the activated flexible operation unit performs the unscrewing and screwing operation of the medicine box cover, including grabbing the medicine box cover by using a three-finger pneumatic flexible gripper and controlling the screwing torque by using a coaxial torque control module. 10.The control method of automatic battery and drug replacement of the UAV according to claim 1, characterized in that: In the S3, before the work process is executed, a secondary precise positioning step based on end vision is further included, a three-dimensional point cloud image of the local target work point is captured by the end vision module, and the standard three-dimensional model is matched to calculate the coordinates of the center point of the target work point.

Images